Download - tree desertification planning for african agroecosystem

FACULTEIT LANDBOUWKUNDIGE EN TOEGEPASTE BIOLOGISCHE WETENSCHAPPEN

ACADEMIC YEAR 2001 - 2002

METHODOLOGY FOR TREE SPECIES DIVERSIFICATION PLANNING FOR AFRICAN AGROECOSYSTEMS

METHODOLOGIE VOOR DIVERSIFICATIEPLANNING VOOR BOOMSOORTEN IN AFRIKAANSE AGROECOSYSTEMEN

DOOR

IR. ROELAND KINDT

THESIS SUBMITTED IN FULFILMENT OF THE REQUIREMENT FOR THE DEGREE OF DOCTOR (PHD) IN APPLIED BIOLOGICAL SCIENCES

PROEFSCHRIFT VOORGEDRAGEN VOOR HET BEHALEN VAN DE GRAAD VAN DOCTOR IN DE TOEGEPASTE BIOLOGISCHE WETENSCHAPPEN

OP GEZAG VAN

RECTOR: PROF. DR. A. DE LEENHEER

PROMOTOR

PROF. DR. IR. P. VAN DAMME

CO-PROMOTOR

DR. A. J. SIMONS

DEAN

PROF. DR. IR. O. VAN CLEEMPUT

iii

The author, the promotor and co-promotor give authorisation to consult and to copy parts of this work for personal use only. Any other use is limited by Laws of Copyright. Permission to reproduce any material contained in this work should be obtained from the author.

De auteur, de promotor en de co-promotor geven de toelating dit doctoraatswerk voor consultatie beschikbaar te stellen, en delen ervan te copiëren voor persoonlijk gebruik. Elk ander gebruik valt onder de beperkingen van het auteursrecht, in het bijzonder met betrekking tot de verplichting uitdrukkelijk de bron vermelden bij het aanhalen van de resultaten uit dit werk.

Prof. Dr. ir. P. Van Damme

Promotor

Dr. A. J. Simons

Co-promotor

ir. R. Kindt

Author

(PVD & RK)

Ghent University

Faculty of Agricultural and Applied Biological Sciences

Department Plant Production

Laboratory of Tropical and Subtropical Agronomy and Ethnobotany

Coupure links 653, B-9000 Ghent, Belgium

[email protected]

&

(AJS & RK)

International Centre for Research in Agroforestry (ICRAF)

Domestication of Agroforestry Trees Programme

PO Box 30677, Nairobi, Kenya

[email protected]

[email protected]

v

CONTENTS ACKNOWLEDGMENTS

viii

SAMENVATTING VAN HET PROEFSCHRIFT (summary)

x

CHAPTER 1

Introduction

R Kindt

1

CHAPTER 2

The Relationship between Species Diversity within an Ecosystem and its Stability and Productivity

R Kindt

21

CHAPTER 3

Description of the Survey Areas and Data Collection Methods

R Kindt

33

CHAPTER 4

Methodology

R Kindt

51

CHAPTER 5

The Study of On-Farm Tree Species Diversity in Western Kenya to Plan for Agroecosystem Diversification

R Kindt, AJ Simons & P Van Damme

73

CHAPTER 6

The Study of Random and Proximity-based Tree Species Diversity on Farms in Western Kenya Using Exact Species Accumulation Curves

R Kindt, P Van Damme & AJ Simons

91

vi

CHAPTER 7

Studying On-farm Tree Species Richness and Evenness in Western Kenya with Diversity Profiles

R Kindt, P Van Damme & AJ Simons

101

CHAPTER 8

Do Farm Characteristics Explain Differences in Tree Species Diversity among Western Kenyan Farms?

R Kindt, AJ Simons & P Van Damme

117

CHAPTER 9

Comparison of Tree Species Composition of Farms in Western Kenya Using Constrained Ordination. I. Analysis for Separate Use-groups

R Kindt, P Van Damme, AJ Simons & H Beeckman

135

CHAPTER 10

Comparison of Tree Species Composition of Farms in Western Kenya Using Constrained Ordination. II. Analysis for Separate Niches

R Kindt, P Van Damme, AJ Simons & H Beeckman

155

CHAPTER 11

Farmers’ Decision-making in Managing On-farm Tree Diversity

R Kindt, P Van Mele & P Van Damme

171

CHAPTER 12

Intraspecific Diversity of Trees on Farms in Western Kenya – Results from Simulations of Individual-based Metapopulation Dynamics

R Kindt, AJ Simons, P Van Damme & D Reheul

191

CHAPTER 13

Comparing Species Richness and Evenness Contributions to On-Farm Tree Diversity for Datasets with Varying Sample Sizes from Kenya, Uganda, Cameroon, and Nigeria with Randomised Diversity Profiles

R Kindt, A Degrande, L Turyomurugyendo, C Mbosso, P Van Damme & AJ Simons

207

vii

CHAPTER 14

Species Diversity on Farms in Cameroonian, Kenyan and Ugandan Landscapes

R Kindt, J Weber, A Lengkeek, JM Boffa, A Degrande, JG Jourget, D Van Oijen , C Mbosso, P Van Damme & AJ Simons

223

CHAPTER 15

General Conclusions and Recommendations for Future Research and Development Activities

R Kindt

251

LITERATURE LIST

269

APPENDIX I

Matrices with Information on Tree Species Identities, Farm Characteristics and Species Distribution over Farms in Western Kenya

285

APPENDIX II

Questionnaires Used in the Western Kenyan Surveys

313

viii

ACKNOWLEDGMENTS

Many people have helped me to complete this thesis. I sincerely hope that I managed to mention all of them in this section. If you do not see yourself mentioned, this only reflects a lapse in my short-term memory.

First, I want to thank my wife Gladys Mumbi Wacira for her patience with me when I was bringing some work home and to tolerate when I was traveling outside of Nairobi or to faraway places in my mind. I have enjoyed all the time that we have shared and I am looking forward to many years in the future, especially now that we are trying to expand our family.

I am very grateful to my supervisor in the International Centre for Research in Agroforestry, Dr. Tony Simons, for providing me with the possibility of conducting the research that lead to results listed in this thesis, and as co-promotor for intellectually stimulating me and helping me to complete this thesis.

Prof. Dr. ir. Patrick Van Damme has been of equal importance to the success of this thesis. I am very grateful for his supervision and the time and detail that he has devoted to providing comments on this work.

I appreciate the farmers of Cameroon, Kenya, Nigeria, and Uganda, who provided time and information during data collection for this thesis, very much. I hope that I have been able to demonstrate the diversity of their knowledge in this thesis, and that its results may contribute to improvement of their livelihood. I am very grateful for Joseph Njeri who assisted greatly during data collection in western Kenya.

I thank Prof. Dr. ir. Hans Beeckman, Prof. Dr. ir. Dirk Reheul and Dr. ir. Paul Van Mele for contributions to some chapters of this thesis. I also feel very grateful for the comments given by Paul on the general information requirements for this thesis.

I am very grateful for the review of the doctoral manuscript by the members of the examination committee: Prof. Dr. ir. M. Verloo (Gent University, Belgium, chair), Prof. Dr. ir. Patrick Van Damme (Gent University, promotor), Dr. Tony Simons (ICRAF, Nairobi, co-promotor), Prof. Dr. ir. Hans Beeckman (Gent University), Prof. Dr. ir. Bernard De Baets (Gent University), Prof. Dr. Paul Goetghebeur (Gent University), Prof. Dr. ir. Robert De Wulf (Gent University), Prof. Dr. ir. François Malaisse (Université de Gembloux, Belgium), Prof. Dr. Jean-Pierre Ottoy (Gent University), Dr. ir. Olivier Thas (Gent University) and Dr. ir. Paul Van Mele (CABI, UK).

I wish to thank VVOB (the Flemish Association for Development Co-operation and technical Assistance) for seconding me to ICRAF. Of the VVOB headquarters office, I especially would like to thank Dr. ir. Jaak Lenvain and Dr. Rudy Poulussen. Lut Laenen-Fox has been extremely helpful as representative for VVOB in Kenya. I would like to thank Prof. Dr. ir. Jef Coosemans for the comments that he provided in his capacity as scientific advisor for VVOB associates in ICRAF.

I also feel very grateful towards DFID, USAID, and IFAD who provided funding that enabled to conduct fieldwork in Cameroon, Kenya, Nigeria, and Uganda.

Joris, Veerle, Said, and Aron De Wolf always managed to provide me with a very warm welcome in Kisumu, and making my fieldwork in western Kenya very enjoyable. I appreciate Joris’ assistance with methodological issues of planning and analysis very much.

ix

Very important to the conceptualization and completion of this thesis have been colleagues within ICRAF. I would like to acknowledge them in alphabetical order: Kwesi Atta-Krah, Liz Betser, Jean-Marc Boffa, Wim Buysse, Richard Coe, Ian Dawson, Ann Degrande, Joris De Wolf, Steve Franzel, Anne-Marie Izac, Hannah Jaenicke, Bashir Jama, Ard Lengkeek, Jens-Peter Barnekow Lillesø, Hans Lindqvist, Charlie Mbosso, Amadou Niang, Frank Place, Jane Poole, Thomas Raussen, Ralph Rommelse, Ralph Roothaert, Stephen Ruigu, Zac Tchoundjeu, Levand Turyomurugyendo, Tom Vandenbosch, Meine Van Noordwijk, Bruno Verbist, François Verdeaux, Grégoire Vincent, John Weber, James Were, and Robert Zomer. I also thank all former and present members of the ICRAF Management Committee and Board of Trustees, especially Dennis Garrity, Anne-Marie Izac, Roger Leakey and Pedro Sanchez, for their support. I am very grateful for the comments and suggestions provided by the members of the Advisory Committee to ICRAF on Genetic Resource Activities and the members of the Board Commissioned External Review of the ICRAF Tree Domestication Programme.

In ICRAF, I also appreciated the assistance of many people who are not researchers but who have assisted me throughout the years. These people are too many to list here, but I would like to give special mention to Stella Muasya and Josina Kimotho.

Of the Laboratory of Tropical and Subtropical Agronomy and Ethnobotany of the Faculty of Agricultural and Applied Biological Sciences of the University of Ghent, I thank Annita Goethals, ir. Tinneke Dirckx and ir. Xavier Scheldeman for assistance that they provided with administrative issues.

I would like to thank my VVOB-colleagues at ICRAF for their companionship a nd assistance: ir. Wim Buysse, ir. Bruno Cammaert, ir. Ann Degrande, ir. Elke Delvoye, Dr. ir. Johan Desaeger, ir. Joris De Wolf, ir. Tom Vandenbosch, ir. Bruno Verbist, ir. Inge Vissers and ir. Katty Wauters. The assistance of ir. Jan Beniest at ICRAF has also been greatly appreciated.

Last but not least, I would like to thank my parents, Daniel Kindt and Rita Vermander, my brothers Wouter and Brecht, my Belgian and Kenyan extended families, and friends from Belgium, Kenya and other countries, for their friendship and support.

x

SAMENVATTING VAN HET PROEFSCHRIFT

In het eerste hoofdstuk worden de 27 hypotheses die onderzocht werden belicht. Ook wordt het onderzoek gesitueerd binnen de onderzoeksagenda van ICRAF rond het beheer van natuurlijke rijkdommen in verschillende tropische regio’s. Het onderzoek was vooral exploratief – met verschillende numerische-ecologie technieken werden secties binnen agro-ecosystemen van significant lagere diversiteit gecategorizeerd die later het mikpunt kunnen zijn voor diversificatie-initiatieven.

In het tweede hoofdstuk worden de resultaten van een literatuurstudie naar het verband tussen de diversiteit van een ecosysteem en zijn stabiliteit en productiviteit voorgesteld. Verschillende ecologische experimenten en modellen hebben aangetoond dat er een positieve relatie bestaat. Deze relatie is echter afhankelijk van spatio-temporele diversiteit en diversiteit in karakteristieken van de soorten in het systeem.

Het derde hoofdstuk beschrijft de zes studiegebieden in Kenya, Uganda, Kameroen en Nigeria. De gerandomizeerd-gestratificeerde monstername van boerderijen op basis van afstand tot bosreservaten en socio-economische parameters wordt toegelicht.

Het vierde hoofdstuk belicht de gebruikte analyze-technieken, waaronder enkele nieuwe methodes.

Het vijfde hoofdstuk onderzoekt hoe de belangrijkste gebruiken van bomen over boerderijen verdeeld zijn. Een positieve relatie tussen het aantal soorten op een boerderij, en het aantal gebruiken van soorten op deze boerderij wordt bewezen. Niet iedere boer gebruikt dezelfde soort op dezelfde manier – zowel soorten als informatie over gebruiken kunnen wijdser verspreid worden in het gebied.

In hoofdstuk zes wordt de accumulatie van soorten met gebiedsgrootte geanalyseerd met een snellere en meer accurate manier dan de gangbare methode. Gebruiken met lagere alpha- en/of beta-diversiteit werden in kaart gebracht.

Het zevende hoofdstuk introduceert een nieuwe techniek om de diversiteit van een systeem te ontbinden in de invloeden van soortenrijkdom, proportionele gelijkheid en grootte-orde van het proefgebied. Met de methode worden gebruiken binnen het West-Kenyaanse landschap van lagere diversiteit – rijkdom en/of proportionele gelijkheid - geselecteerd.

In hoofdstuk acht wordt de relatie tussen karakteristieken van boerderijen en hun diversiteit onderzocht door middel van multiple regressiemodellen. Significante relaties worden aangetoond, maar niet alle variatie kon gemodelleerd worden.

Hoofdstukken negen en tien bekijken verschillen en soortencompositie van boerderijen via een combinatie van twee nieuwe ordinatietechnieken. Een significant gedeelte van de variatie wordt gemodelleerd. Uitzonderlijke boerderijen worden aangetroffen met een soortensamenstelling die niet typisch is voor andere boerderijen van hetzelfde type.

In hoofdstuk elf worden de resultaten voorgesteld van participatorisch onderzoek naar de manier waarop boeren de diversiteit op hun boerderijen willen beheren. De algemene resultaten toonden aan dat boeren meer bomendiversiteit willen, vooral omdat ze de elkaar aanvullende karakteristieken van verschillende boomsoorten appreciëren.

Hoofdstuk twaalf onderzoekt de dynamiek van genetische diversiteit in metapopulaties van de meeste onderzochte boomsoorten. De meeste boomsoorten kwamen geaggregeerd voor in het landschap. Hoewel de simulatiemodellen geplaagd werden door de momenteel beperkte kennis inzake reproductieve ecologie van de meeste tropische boomsoorten, toonden deze modellen aan

xi

dat een gestratifieerde collectiemethode van zaden, zaailingen of stekken door boeren de genetische erosie van deze soorten kan beperken.

In hoofdstukken dertien en veertien worden verschillende diversiteitsfacetten die in eerdere hoofdstukken onderzocht werden voor westelijk Kenya ook onderzocht voor de andere studiegebieden.

Hoofdstuk vijftien bekijkt de hypotheses uit hoofdstuk een en bespreekt hoe de aanzienlijke diversiteit die aangetroffen werd in de studiegebieden gebruikt kan worden binnen initiatieven die deze systemen verder willen diversifiëren. De thesis wordt afgesloten met een literatuurlijst, informatie gecollecteerd in westelijk Kenya en de formulieren die gebruikt werden tijdens de inventarissen en interviews.

CHAPTER 1 INTRODUCTION

R KINDT

INTRODUCTION

2

The research findings that are presented in this thesis are the results of activities that were conducted by R. Kindt and colleagues in the International Centre for Research in Agroforestry. R. Kindt was seconded to ICRAF by the Flemish Association for Development Co-operation and Technical Assistance (VVOB).

A short introduction is given first about the mission of ICRAF, after which a description of the domestication concept developed by ICRAF is presented. Finally, an introduction is given to some of the data analysis methods employed in this thesis.

1.1. History and Mission of ICRAF

ICRAF was established as the International Council for Research in Agroforestry in 1978 to promote agroforestry research in developing countries, with a focus on research, training, and information activities in sub-Saharan Africa. It joined the Consultative Group on International Agricultural Research in 1991 to conduct strategic research on agroforestry on a global scale, changing its name from Council to Centre and expanding into Southeast Asia and Latin America. Currently, ICRAF has active regional programmes in Eastern and Central Africa, Humid West Africa, Southern Africa, the Sahel, Southeast Asia and Latin America, whereas its headquarters are located in Nairobi. In 2002, ICRAF will change its name to the World Agroforestry Centre and adopt the tagline ‘transforming lives and landscapes‘ as an indication of its evolution into a 21st century institute.

The mission of ICRAF is to conduct innovative research and development on agroforestry, strengthen the capacity of its partners, enhance worldwide recognition of the human and environmental benefits of agroforestry, and provide scientific leadership in the field of integrated natural resource management. The mission will be accomplished through the combination of the best of science with farmer knowledge in a wide range of strategic alliances along the research-development continuum (ICRAF 2000).

Agroforestry refers to the growing of trees on farms to improve livelihoods and protect the environment. The technical definition of agroforestry adopted by ICRAF is (Leakey 1996):

Agroforestry is a dynamic, ecologically-based, natural resource management practice that, through the integration of trees on farms and in the agricultural landscape, diversifies and sustains production for increased social, economic and environmental benefits.

Simons (1996) mentions that this definition conveys the idea that various agroforestry practices play different roles in the ecological succession towards ‘climax agroforests’ at a landscape scale. The term ‘Trees’ in the agroforestry definition refers to woody or ligneous plants, which include ‘trees’ (woody plants with a single main stem, =7m), ‘small trees’ (< 7 m), ‘shrubs’ (multi-stemmed woody plants), ‘lianas’ or ‘woody vines’ (woody plants that generally require a support), and ‘bamboos’.

Agroforestry is a form of natural resource management (NRM) – the management of resources (natural capital) generated by natural biogeochemical processes and solar energy that produce flows of desirable products and services. Agroforestry provides economic and ecosystem functions at multiple geographical scales ranging from the farm, watershed/village/landscape, national/regional to the global scale. At the global scale, for instance, incorporating trees into farming systems leads to carbon sequestration, greenhouse gas regulation, biodiversity conservation, poverty reduction, and decreased human migration. At the farm level, agroforestry leads to greater prosperity by providing marketable products such as lumber, building poles, firewood, fodder, fruits and medicines. At the farm level, trees also improve soil fertility by fixing atmospheric nitrogen and recycling nutrients, holding moisture where it is needed, reducing soil erosion, and helping to ensure the stability of future production (ICRAF 2000).

CHAPTER 1

3

ICRAF’s integrated natural resources management (INRM) agenda builds on the results of the Green Revolution, but differs from it in four ways (ICRAF 2000):

• the farmers whose needs are addressed are the poor in marginal or depleted environments;

• the research includes beneficiaries other than farmers, such as community-level landusers, national and global policy makers, and the concerned public;

• the approach focuses on heterogeneous environments, which require a flexible range of management options; and

• the method builds upon the production and ecosystem service functions that natural capital fulfils in agriculture. Such functions increase productivity while ensuring the stability of these increases.

The INRM agenda of ICRAF corresponds to the advocacy of a “greener revolution”, an environmentally sustainable revolution, that considers gains and losses in values of ecosystem goods and services (see Chapter 2) by converting natural ecosystems to agriculture, now that human population size and per capita consumption are assumed to be the biggest drivers of global environmental change (Tilman et al. 2001a).

1.2. Tree Domestication in ICRAF

The Tree Domestication Programme of ICRAF’s Research Division conducts research on the domestication of trees and landscapes, operating under the following definition of domestication (Simons in prep. – this is an updated definition from Simons 1997):

Domesticating agroforestry trees involves accelerated and human induced evolution to bring species into wider cultivation through a farmer-driven or market-led process. This is a science-based and iterative procedure involving the identification, production, management, and adoption of high quality germplasm. High quality germplasm in agroforestry incorporates dimensions of productivity, fitness of purpose, viability, and diversity. Strategies for individual species vary according to their functional use, biology, management alternatives, and target environments. Domestication can occur at any point along the continuum from the wild to the genetically transformed state. The intensity of domestication activities warranted for a single species will be dictated by a combination of biological, scientific, policy, economic, and social factors. In tandem with species strategies are approaches to domesticate landscapes by investigating and modifying the uses, values, intraspecific diversity, ecological functions, numbers, and niches of both planted and naturally regenerated trees

The research documented in this thesis has contributed to the ‘domestication of the landscape’ concept and methodology that was recently developed in ICRAF. In contrast to domestication of agroforestry species aimed at accelerated evolution of individual species, domestication of the landscape proposes modifications to agroecosystems based on diversity already present in these systems. Simons (in prep.) distinguished four potential interventions of landscape domestication: (i) “replacement” of a tree by a tree of the same species, (ii) “substitution” of a tree by a tree of a different species, (iii) “addition” of new trees, or (iv) modifications in tree “management”. Several analysis methods of diversity in the investigated agroecosystems, and how this information can be utilized to design interventions are documented in the various chapters.

As the research presented in this thesis was conducted within the objectives of ICRAF’s Tree Domestication Programme of diversification of the species composition of agricultural landscapes, the general objectives of the investigations of this thesis can be summarized as:

INTRODUCTION

4

• to document the distribution of species diversity in several landscapes

• to investigate possibilities for diversification of these landscapes

Palmer et al. (2001) differentiated between species that are globally important agroforestry species for domestication, regionally important species, and ‘shield species’ that are collectively important for farm incomes and the maintenance of ecosystem services. They further stated that on-farm management of genetic resources of shield species was a vitally important component of a domestication strategy. Genetic diversity contributes to landscape biodiversity as biodiversity can be structured into genes, populations, species, functional ecological groups, and ecosystems. Molecular methods that sample genetic material and allow estimations of genetic diversity in tree populations have been used for priority species for domestication in ICRAF (examples of such research conducted in ICRAF are Dawson & Powell 1999; Russell et al. 1999 and Lowe et al. 2000). For practical reasons, genetic diversity per se has not been sampled during the surveys documented in this thesis. However, information on origins and types of germplasm (propagules for sexual multiplication of plants such as seeds or seedlings, or vegetative multiplication such as cuttings or grafts), on the distribution of trees, and general reproductive biology characteristics of tropical tree species enabled simulations of genetic diversity dynamics for various scenarios of on-farm germplasm management. Actual sampling of genetic diversity and measurement of reproductive ecology characteristics of individual species would obviously have allowed for better investigations, but time and resources did not permit this and thus a decision was made to focus on the interspecific component of landscape diversity.

1.3. Format and Sequence of the Chapters

Most of the chapters of the thesis were developed as articles in the required formats of the targeted journals for submission. Here, the articles have been modified into a common format. To avoid duplication, the sections that described the study areas and the methods were removed and added into separate chapters (Chapter 3 & 4). For the same reason of presenting information in a more condensed format, all literature is combined into a common bibliographic list. Chapter 15 presents some general conclusions from the investigations, and makes suggestions for future research. The objective of the following chapter (Chapter 1) is to outline the objectives of the investigations within the context of the objectives of ICRAF, and to demonstrate how the various chapters address various aspects pertaining to diversity and diversification of agroecosystems.

Since scientific knowledge on the influence of species’ diversity on ecosystem productivity and stability evolved substantially during the research period, and to better document the results of a literature survey on the subject, one chapter specifically documented the role of diversity in ecosystems (Chapter 2). We realized that our investigations were performed on anthropogenic systems (agroecosystems), and that the typical measure of productivity of a natural ecosystem as biomass differs from the economical measurement of agricultural productivity. However, we included Chapter 2 since we referred to the role of diversity in ecosystem functioning (which included agroecosystems) explicitly when discussing options for diversification in various chapters. Although Chapter 2 appears separately, information on the function of diversity was not removed from the other chapters.

Figure 1.1 depicts a general sequence by which some of the analyses were conducted. The on-farm surveys aimed at collecting information about all tree species that occurred on selected farms. All information was stored in Microsoft Access databases that were designed for data storage and analysis. Data analysis often started by constructing a sites × species matrix that contained the abundance (number of trees) of each species on each farm (ecological data matrices

CHAPTER 1

5

such as sites × species matrices are the typical basis for many ecological analyses: Legendre & Legendre 1998 pp. 52-53). Some matrices only listed trees that provided a specific product or service (summarized here as use), only listing an abundance > 0 for a particular cell in case the farmer had communicated to use the species for that particular purpose on his or her farm. Other matrices only listed trees that occurred in a specific on-farm niche, defined by the location on the farm and the establishment pattern of trees in that location (Chapter 10). A substantial part of this thesis deals with comparisons among and between these matrices through statistics that describe differences in diversity. For example, in Chapter 7, diversity profiles are calculated from matrices that each only included trees that provided a specific use. Another example is Chapter 8, where the diversity of fruit trees are compared among farms, based on calculations from the fruit matrix. Thus, comparisons among and within uses or niches actually refer to comparisons among or within matrices. Figure 1.2 provides an illustration of these two types of comparison modes.

With the sites × species matrix as the basis for many analyses, identification of all tree species that occurred on each farm was crucial. Species identification in the field posed some serious problems in some cases, however. Not all species could be confirmed to botanical species when analyses were made, although herbarium samples were taken (deposited in ICRAF), vernacular names were collected, and repeated visits were made in the field with species identification guides. In Kenya, Beentje (1994) was used as the essential reference for indigenous species, and Mabberley (1997) for exotic (and naturalized – which were still classified ‘exotic’ in this thesis) species. We used the sequence of botanical families by Cronquist (1981). The major identification problem was often the lack of reproductive organs, while farmers had removed some trees before these could be identified in the field or by herbarium specimens. However, information of table I.1 and table I.3 (Appendix I), and the information stored in the databases, will allow later investigation of the potential influence of species identification on the results. As many methods used logarithmic scales to compare the diversity of various systems, the influence of unidentified species on the results is expected to be minimal, however. In addition, this thesis has focused more on analysis methods rather than the results of these analyses – although results of the analyses have lead to a current field-testing phase of landscape domestication in the target areas.

For some types of data analysis presented in this thesis, appropriate software was not available. To perform these analyses, several programs were written in FORTRAN by the author. The source code for these programs is provided in Appendix III.

1.4. Hypotheses Tested in Chapters 5 - 14

Chapters 5-12 describe various analyses that were conducted on data collected from western Kenya to test a range of hypotheses. Chapters 13 and 14 test hypotheses on data collected from a larger set of surveys in four countries. Here, is listed the 27 hypotheses that were tested in these chapters, and what the consequences of accepting or rejecting a hypothesis would be in terms of conducting biodiversity surveys or planning the management of on-farm diversity. Some hypotheses were tested in several chapters, and were therefore provided with the same number.

Most often, hypotheses dealt with the possibility of obtaining some ranking from low to high diversity within or among the matrices used as inputs in the investigations. As diversity can be measured in many different ways – which is demonstrated throughout this thesis – many different diversity rankings are theoretically possible. The consequence of the existence of rankings is that certain development or research projects that are designed to enhance the on-farm diversity of tree species will have the possibility to concentrate their efforts on subsections of lower biodiversity of a landscape. One justification to concentrate on subsections of lower diversity is that, on average, larger effects are expected from diversification of a system of low

INTRODUCTION

6

diversity. For example, the effect of adding one new species where only one species is present is expected to be greater than adding one new species to ecosystems where already 50 species are present – in situations where diversity has an effect on the functioning of the ecosystem. In case a project aims at improving the functioning of the ecosystem through diversification, for instance by improving the stability of its production, then efforts of adding one species would be more efficient if such species was added to a system of lower diversity. Thus, to maximize the effect of diversification, it is better to introduce a species in a subsection where diversity is currently low. Diversification where diversity is low could also be justified from the perspective of limiting risk. For example, an agroecosystem with two species that provide timber and fifteen species that produce firewood could be categorized as more risky of failing to produce timber than failing to produce firewood due to potential pests, diseases, or other factors that affect production. One of conditions of this relationship, however, is that the fifteen species that provide firewood do not have a substantially higher risk (in the order of 15/2) of not producing that product than that associated with the two species that provide timber.

Obviously, species affect the functioning of an ecosystem through their interactions with the environment and with the other species, and not merely by their presence or absence. Therefore, fewer species may offer a larger range in these interaction traits than offered by more species. Thus, the points presented above imply that more species offer a larger set of species’ characteristics, as provided for instance when each species has a random set of traits. In addition, the same number of species may offer a large or a small variation in traits, depending on the distribution of trees among those species. In cases where 99% of trees belong to the same species, then the variation in traits will be low, whereas when each species has the same abundance, a larger variation in traits will be offered. Thus, evenness in the abundance of species will also effect the variation in traits and by consequence the functioning of the ecosystem. Therefore, making species more evenly distributed may also effect ecosystem functioning. As diversity can conceptually be dissociated into richness (the number of species) and evenness (the equality among abundances of the component species), diversification can also be dissociated into enhancing richness (adding more species) and enhancing evenness (making the abundance distribution more even). Diversification, in the context of this thesis, should therefore not be equated to species addition. The effect of diversity (and thus diversification) on ecosystem processes, and conditions for positive effects of diversity on these processes, are detailed in Chapter 2.

In summary, a large section of this thesis was devoted to categorizing the diversity of subsections of agricultural ecosystems. In case subsections of smaller diversity can be identified, then diversification efforts that are targeted towards these subsections are expected to have larger effects than random introductions of species or trees would have.

The following hypotheses were posed:

Chapter 5 documents how sites × species matrices can be formed based on information on species’ uses collected from each farm. As the formation of such matrices was an important step before the analyses of Chapters 6 - 9 could be conducted, the various options available to form these matrices were scrutinized. Subsequently, some simple statistics that rank categories of uses in terms of diversity were calculated from these matrices.

CHAPTER 1

7

H1 There is a positive relationship between the number of species on a farm, and the number of uses of these species on the farm

? Accepting this hypothesis would indicate that farmers establish more species so that more uses can be obtained from trees on their farms. In case the objective of a project would be to maximize the number of species that are grown on farms (e.g. for conservation-through-use), then farmers would prefer to grow species for a large range of purposes, rather than growing many species for a limited number of purposes. A positive relationship could indicate that redundancy (the average number of species per use) would not increase much with species richness, and, therefore, that diversification within a use may still be important even if many species and uses are encountered on a farm.

H2 The number of uses of a species in the entire survey equals the number of uses of a species on each farm

? Accepting this hypothesis implies that the collection of ethnobotanical information can be simplified, as information of the distribution of species over farms and information of the uses of a species will suffice to obtain the information on the uses of a species on each particular farm. In other words, the use matrix described above can be constructed by means of a sites × species matrix, and a general list with all the uses of each species. In case the hypothesis is rejected, then information will need to be collected for the uses of each species on each farm. Accepting this hypothesis could mean that farmers do not share information on species’ uses well.

H3 Recording information only on the main use of each species on a farm provides an adequate representation of the distribution of diversity of species of similar on-farm use

? Accepting this hypothesis implies that the collection of ethnobotanical information can be simplified, as only information on the main use of a species on a farm will suffice to obtain a representative use matrix. Rejecting the hypothesis means that information will need to be collected for all uses of each species on each farm

H4 Alpha diversity, i.e. the average number of species per farm, is the same for each use

? Rejecting this hypothesis implies that uses can be ranked from low to high average diversity, using the diversity statistic of alpha diversity. Under such conditions, development projects could target those uses of low diversity, for instance when the aim would be to have at least two species for each use.

H5 Gamma diversity, i.e. the total number of species observed in the survey, is the same for each use

? Rejecting this hypothesis implies that uses can be ranked from low to high total (landscape-scale level) diversity. Development projects could then target uses of low diversity in the entire landscape (these are not necessarily the same uses with the smallest diversity on the average farm as investigated by hypothesis 4). A project could for example set the objective of having a minimum of five species for every use that occurs in a specific area (the relationship between diversity and function applies to each spatial scale, not only at the smallest scale, as described in Chapter 2).

INTRODUCTION

8

H6 Evenness, as expressed by the proportion of the most abundant species in the entire survey, is the same for each use

? Rejecting this hypothesis implies that uses can be ranked from low to high total landscape-scale diversity. Here, however, diversity was not expressed in terms of richness (as for hypothesis 5), but in terms of evenness in the abundances of the component species. The objective of a project could for instance be that within each use not more than 50% trees belong to the same species. As for hypotheses 5 and 6, such objective could be linked to limit risks when one species would not provide the product or service at a given moment.

H7 Species density, the average number of trees per species per ha, is the same for each grouping of species of similar on-farm use

? Rejecting this hypothesis means that uses can be ranked from low to high density. In case the risk of genetic erosion (the loss of genetic diversity) would be positively correlated with density (through effects from limited geneflow), and assuming no effects from reproductive biology then rankings in terms of density would also be rankings in terms of risk for genetic erosion. Certain projects could use this information to focus their efforts on uses with high risks for genetic erosion.

Chapter 6 documents species accumulation patterns, which are associated with the average species richness of 1, 2, 3, ... accumulated sites, and spatial distribution patterns of species richness. In contrast with earlier research, a novel approach was employed where a formula was specified that allows the calculation of the average richness from information on the frequency of each species, whereas in earlier research accumulation curves were obtained through randomization. Species accumulation curves provide information on alpha diversity, the average number of species per farm, and beta diversity, providing information on differentiation in species composition among sites. The two parameter model,

zcAS =

where S stands for species richness, c for alpha diversity, A for the area sampled and z for beta diversity is widely used in the ecological and biogeographical literature (see Chapter 6), and describes how species richness at specific spatial scale (e.g. A = 10 ha) is determined by alpha and beta diversity. This chapter complements chapter 5 through a sample-scale specific measurement of beta diversity.


? See Chapter 5


? See Chapter 5

CHAPTER 1

9

H8 Beta diversity, i.e. the average difference in species composition among farms, is independent of sample size

? The consequence of accepting this hypothesis is that species accumulation, the average increase of species richness when species richness is measured on a larger number of randomly selected farms, can be categorized by two parameters only, through the model specified above. When species accumulation is more complex, beta diversity will need to be categorized for each intermediate scale (2, 3, 4 ... farms) in between 1 farm and all farms. The consequence for the design of interventions is that in systems with high beta diversity, alpha diversity can be enhanced relatively easily by distributing more widely some of the species that are only present in a subsection of the landscape now. In systems with low beta diversity, substantial enhanced alpha diversity could only be possible through introduction of new species in the landscape.

H9 Sample size has no influence on the rank-order in terms of species richness between uses

? The consequence of rejecting this hypothesis means that one use may contain more species on 10 farms than another use, while the other use may contain more species on 50 farms than the first use. Therefore, neither the first nor the second use can be identified as the use of lowest diversity, without specifying at which spatial scale diversity is measured. Projects that aim at enhancing diversity will need to specify at which scale diversity needs to be enhanced, in order to be able to select the uses of lowest diversity.

H10 Villages contain fewer species than expected by random chance

? Accepting this hypothesis means that farms within villages are more similar in species composition than expected when species would be distributed at random over farms. This means that that species are aggregated within villages. Under these conditions, distributing species and information on their use more randomly in the landscape – for instance through farmer-farmer exchange visits – will enhance the species richness of each village.

? Aggregation of species within villages probably reflects the history of cultivation of particular species in an area.

Chapter 7 describes a new mathematical procedure by which diversity can be dissociated into species richness (the number of species) and evenness (in the abundance distribution). Changes in diversity when scaling from one farm to the complete survey are investigated. Chapter 7 complements Chapter 5 through a more accurate measurement of evenness. All analyses in this chapter are done for the entire survey. Diversity is thus calculated for all species and all trees encountered.


? See Chapter 5

H11 Diversity, as measured for the species composition of the entire survey, is the same for each use

? Rejecting this hypothesis implies that uses can be ranked from low to high landscape-scale diversity. In contrast with Chapter 5 (hypotheses 5 & 6), the rank-order that will be obtained will not be dependent on the type of diversity statistic that is used, i.e. statements that one use is more diverse do not need to be accompanied by “... when diversity was measured as...”. Rejecting the hypothesis allows projects to identify uses of lowest diversity.

INTRODUCTION

10

H12 Evenness, as measured for the species composition of the entire survey, is the same for each use

? Rejecting this hypothesis implies that uses can be ranked from low to high total landscape-scale evenness. Rejecting the hypothesis allows projects to identify uses of lowest evenness.

? Hypotheses 11 and 12 will not be accepted or rejected simultaneously for all theoretical cases. When one use has substantially higher richness, it will also have substantially higher diversity than a second use that could have higher evenness. In other words, higher evenness and higher richness imply higher diversity, but higher diversity does not imply higher evenness and higher richness.

H13 A use with higher gamma diversity will also have higher evenness

? Accepting this hypothesis implies that uses can be ranked from low to high total landscape-scale diversity, only by measuring gamma diversity. Under these conditions, measurement of diversity and diversification can be approached easily mathematically, as projects should only be concerned with the number of species, not with the differences in the abundance distribution.

Whereas Chapters 5-7 deal with methods of comparing the diversity among sites × species matrices (see above), Chapters 8-10 deal with methods of comparing the diversity within matrices. Chapters 8-10 investigate whether important differences in diversity can be detected among the rows of these matrices, thus among the farms. The analyses of Chapters 8-10 were repeated for each matrix. As the relationships between characteristics of farms with diversity are investigated in a statistical manner, those results should apply to a wider area, allowing for extrapolation of the results and design of area-wide interventions.

Chapter 8 shows how much of the diversity of an individual farm can be explained by multiple linear regression analysis with socio-economic characteristics of the farm and the location of farms in relation to forests as explanatory variables. The regression analyses allowed to investigate whether factors such as the gender of the head of the household, wealth, family size, education levels, or proximity to forests, could influence the diversity of tree species on a farm.

H14 Socio-economic characteristics and the spatial location of the farm both explain variation in diversity and abundance among farms

? Accepting this hypothesis means that socio-economic factors influence on-farm diversity. Under these conditions, development projects have various options to target their efforts at subsections of the landscape, depending on the objectives of these projects. In case a project would prefer to design interventions that would benefit a predefined group of clients (e.g. poor farmers, female-headed households), then those uses could be targeted for which the selected types of clients have substantially lower diversity than other types of clients. Another pathway for these projects could be to select the uses of lowest diversity first (potentially by using one of the methods documented in the previous chapters), and then target the interventions at the types of farms that have lowest diversity within these uses. Both pathways would lead to the formulation of objectives of such project as for example “enhancing fruit-tree diversity on female-headed farms”, which could be justifiable as farms of that type would be substantially lower in diversity.

CHAPTER 1

11

H15 The same area of land, consisting of many or few randomly-accumulated farms, has the same diversity and abundance

? Rejecting this hypothesis means that fragmentation of farmland in an area has an influence on the diversity encountered in that area. Such pattern has implications for the conservation of species on farms. It may also influence ecosystem functioning, as a different number of species would be present on the same unit of land after x years. The hypothesis tests whether diversification or simplification may already happen in a landscape without the direct influence of development projects. It could be that further subdivision of farms would lead to fewer species that are maintained in this landscape, or the reverse phenomenon of more species in a fragmented landscape could be observed. Projects should consider such temporal dynamics in species diversity in their planning, especially for trees that take several years to reach maturity.

? This hypothesis is linked to hypotheses 5, 8 & 9, adding a temporal dimension to the investigations – despite the fact that our surveys were conducted within a limited time frame which did not allow for direct investigations of the dynamics of diversity over time. Given the fact that it takes several years to alter the tree composition of a landscape, projects should ideally measure their impact by comparing the expected diversity when no interventions would have been done with the changes they provoked.

Chapters 9 and 10 are examples of constrained ordination techniques, where the relationship between farm characteristics and the species composition of these farms is analyzed. Such methods offer an alternative way of measuring beta diversity (Chapter 6), and provide information on the most important species in which farms of various types differ. Analyses in Chapter 9 were repeated for each group of species of similar on-farm use, while analyses were done in Chapter 10 for trees that were established in the same on-farm niche.

H16 Socio-economic characteristics and the spatial location of the farm both explain variation in species composition among farms

? Accepting this hypothesis means that socio-economic factors influence on-farm composition. For example, poorer farmers may have fewer trees of species x on their farms than richer farmers. Similarly as discussed for hypothesis 14, development projects then can target their efforts at subsections of the landscape. In case a project would for example define its objectives of “enhancing fruit-tree diversity on female-headed farms” (either based on results of methods as described in Chapter 8, or not), then the conditions under which hypothesis 16 is accepted could show that the species composition of female-headed farms differs substantially from that of male-headed farms. In such case, the diversity on female-headed farms would be enhanced by promoting species on those farms that feature more on male-headed farms for the moment. A project could, however, also choose to promote species that do not dominate any subsections of the landscape for the moment.

INTRODUCTION

12

H17 Relationships between farm characteristics and the abundance of a particular species corresponds to the relationships between farm characteristics and the relative species composition of the farm

? Accepting this hypothesis means that the interpretation of ordination results is straightforward. Rejecting the hypothesis means that a farm with a large abundance of a particular species is not necessarily a farm where this species dominates its composition. It is possible that farm x has more trees of species a, but that the species composition of farm y is more dominated by species a. For instance, the 100 trees of species a on farm x may constitute 10% of all trees on this farm, while the 50 trees of the species on farm y may constitute 90% of all trees on that farm. When the hypothesis is accepted, the design of interventions is simpler. The hypothesis is directly linked to hypothesis 16 as it implies that increasing the abundance of species x on subsections of the landscape where this species has lower abundance would directly result in a more homogeneous species composition for the entire landscape, without altering the abundance of other species. Otherwise, in the example that we provided, increasing the abundance of farm y from 50 to 60 trees would result in farm y differing more in species composition with farm x. Making the species composition more homogeneous could be a strategy of enhancing the evenness of species in a landscape.

Chapter 11 provides some evidence of farmer management of interspecific diversity based on open-ended interviews with farmers. This chapter investigates whether farmers already appreciate diversity within certain uses, and what the reasons for their appreciation are. These reasons are not necessarily those of enhancing the functioning of the agroecosystem as mentioned above. Therefore, observing local reasons for diversity could provide another pathway by which diversity could be promoted in the area – either for the local reasons that farmers provide, or for the ecological reasons stemming from ecological relationships between diversity, stability and productivity.

H18 The diversity of species that is present on farms is desired by farmers

? Rejecting this hypothesis means that the diversity that is present on farms is not necessarily desired by farmers. This hypothesis was tested because many tree species can regenerate spontaneously, and could therefore be encountered because they were only tolerated by farmers but not actively desired. Rejecting this hypothesis has implications for the design of projects. For instance, when only one out of thirty species used for timber is desired by farmers for this purpose, then a project aimed at diversifying may still opt to concentrate efforts on timber trees, as diversity of timber trees could decline rapidly.

H19 For a particular use, farmers have reasons to maintain more than one species

? Accepting this hypothesis means that farmers are managing diversity within uses. Under these situations, development projects could find it easier to promote diversity, as diversification would correspond to farmers’ knowledge, perceptions, and practices. Possibly, win-win situations could occur when a diversification project would promote alternative species that would be appreciated by farmers, and that would enhance agroecosystem function.

CHAPTER 1

13

H20 For a particular use, farmers have reasons not to maintain species in perfectly even distributions

? Accepting this hypothesis means that farmers are not willing to establish species in the most diverse distribution that is theoretically possible where each species has the same number of trees. Projects that aim to diversify a landscape should incorporate information on the abundance distribution that is desired by farmers.

? Previous projects have often recorded information on preference rankings of among species. Differences in preference will probably result in differences among the abundance of each species. Obtaining a preference ranking, however, does not mean that farmers would only want to establish the most preferred species (hypothesis 19). Hypothesis 20 implies that more even distributions may be promoted, but that perfectly even distributions may not be realistic.

Chapter 12 investigates consequences of present or potential management methods of intraspecific diversity based on an individual-based metapopulation model. A metapopulation model describes the distribution of individuals within subpopulations (a metapopulation is a group of subpopulations, in a similar way as a population is a group of individuals), with higher chances of geneflow within subpopulations than among subpopulations. As stated above (section 1.2), most of this thesis concentrated on interspecific rather than intraspecific diversity. However, since information was collected on origins and types of germplasm for each tree encountered in the landscape, some simulations of how genetic diversity would change were possible. With the focus on species in this chapter, some analyses of this chapter focused on differences among columns (=species) of the overall sites × species matrix, whereas Chapters 8-10 looked at differences among rows (=farms) of such matrices.

H21 Species are aggregated within farms and villages

? Accepting this hypothesis means that more trees of the same species will be encountered on the same farm and the same village, than expected from a random distribution. The implication of such situation is that geneflow is expected to be spatially structured. In situations where a tree is expected to have a higher chance of obtaining pollen from trees that occur on the same farm than from trees that occur on different farms, and where trees are spatially aggregated, a metapopulation model will describe the actual situation better.

? This hypothesis is related to hypothesis 7, which dealt with species density. Spatial aggregation of species, as investigated through hypothesis 21, means that although species density may be very low (e.g. < 1 tree/ha), each tree may exchange genes with many conspecific trees in case most of these trees occur in the direct surroundings. Thus, species density may not be a very accurate criterion to investigate the chances of genetic erosion, because it is still possible that subpopulation sizes will be relatively large under situations where species density is low.

INTRODUCTION

14

H22 Changes in the way in which germplasm is collected will affect genetic diversity

? Whereas farmers have little control over geneflow by pollen, a major difference with natural ecosystems is that farmers have direct control over geneflow by seed or cuttings. For situations where species are aggregated (hypothesis 21), geneflow will differ drastically where farmers collect most seeds or cuttings from trees that occur on their own farm to collection schemes where most germplasm is collected from other farms. Accepting hypothesis 22 implies that when germplasm collection patterns are changed, genetic diversity dynamics would be different, by consequence. Projects that would aim to limit genetic erosion could then advocate for the best germplasm collection systems in an area. For example, bulking all seed that is collected in a village and letting farmers select at random from this bulk, rather than collecting only seed from the own farm, could limit genetic erosion drastically.

H23 Farmers prefer to collect most trees from the own farm

? Accepting this hypothesis has implications for the expected geneflow when projects would not change the current preferences and practices of farmers in terms of germplasm acquisition. The existence of preferences of the own farm would indicate that projects that aim to change anthropogenic geneflow in an area should address these preferences, for example by explaining the negative consequences of current practices. The balance between the benefits that farmers obtain by collecting from the own farm and the benefits of collecting from other sources should be investigated, if projects aim at altering current practices.

? Information on the currently preferred origins to investigate influences on genetic diversity is only relevant when hypothesis 22 can be accepted.

Chapters 13 and 14 provide a synthesis of results from several surveys. These chapters used some of the analysis methods of the previous chapters.

In Chapter 13 , information from Cameroon, Nigeria, south-western Uganda, and western Kenya is used. In this chapter, we compared results from different surveys that collected information from a different number of farms, a different number of individuals, and a different area. We investigated whether sample size had an influence on diversity. In case such influence is detected, comparisons will only be meaningful if based on the same sample size, for example 50 farms for each survey, or 10,000 trees for each survey.

CHAPTER 1

15

H24 The diversity rank-order of surveys depends on whether comparisons are based on an equal sample size of the same number of sites, the same number of trees, or the same area

? In case this hypothesis holds true, then ranking one use as less diverse than another use will depend on the sample size that is used. A development project should then select the most appropriate sample size that is used to compare diversity. For example, if the project would be more concerned with the results for farmers, then comparing results based on the same number of farms may be more appropriate. If a project would attach higher value to diversity per unit area, then the reference basis of the same area would be more relevant.

? This hypothesis will only be relevant when surveys differ in the number of sites, numbers of individuals/site and average area/site. Otherwise, a reference basis of the same number of sites will automatically be a reference of the same number of individuals and area. Typically, biodiversity surveys differ in individuals/site, and therefore testing hypothesis 24 is relevant in most situations.

? This hypothesis is related to hypothesis 11, which compared diversity among matrices that were all obtained from the same sample size of 201 farms and 116 ha. Thus, implicitly hypothesis 11 is based on the sample size of the same number of farms and the same area that is sampled. Therefore, when hypothesis 24 would be accepted for the comparisons of Chapter 7, a different rank-order would be obtained of the uses if based on the same number of trees.

H25 Distance from forests influences diversity of villages

? If this hypothesis is accepted, then development projects could design interventions based on distance to forests. For example, if a project would aim at diversifying the fruit tree diversity of villages, whereas villages close to a forest would have higher diversity, then the project could decide to target villages that are further away from forests.

? This hypothesis is related to hypothesis 14. The analyses differ in that hypothesis 14 compares the diversity of each individual farm based on the location of the village to which they belong, whereas hypothesis 25 compares the total diversity observed in each village.

In Chapter 14 , information is analyzed from Cameroon, central Kenya, central Uganda, and western Kenya. Some analyses that were presented in previous chapters for western Kenya were also reported in this chapter. Analyses were repeated for each use and for each niche in each survey.


? See Chapter 5


? See Chapter 5


? See Chapter 5

INTRODUCTION

16


? See Chapter 5


? See Chapter 6


? See Chapter 6


? See Chapter 6


? See Chapter 7


? See Chapter 7


? See Chapter 7


? See Chapter 8


? See Chapter 9

? We only tested the influence of spatial location, not socio-economic characteristics, on species composition, however.

H26 The diversity rank-order of gamma diversity of surveys depends on whether comparisons are based on an equal sample size of the same number of sites or the same number of trees

? In case this hypothesis holds true, then ranking one use as less diverse than another use will depend on the sample size that is used. A development project should then select the most appropriate sample size that is used to compare diversity.

? Although the implication of this hypothesis is similar to that of hypothesis 24, the method by which we investigated this hypothesis differed from the method used in Chapter 13, and not only because hypothesis 24 tested diversity and not only richness. In Chapter 13, we always accumulated farms until we obtained a fixed reference value of x farms, y individual trees, or z ha sampled. In Chapter 14, we either accumulated farms or individual trees. For the accumulation of individual trees, trees were selected at random from the overall pool of individuals. The sampling analogy would be that for accumulation of farms, farms were picked at random from a large bag containing all farms, while for accumulation of individuals, individuals were picked at random from a large bag containing individual trees.

CHAPTER 1

17

H27 A larger percentage of on-farm trees are exotic than the percentage of exotic species on-farm

? Accepting this hypothesis means that the percentage of exotic species poorly reflects the percentage of exotic trees. A development project that is focused on conservation-through-use might focus more on the percentage of trees than the percentage of species. This hypothesis shows that the correct statistic should be used.

? This analysis is actually related to the analysis of richness and evenness, but now not comparing subsections based on use or niche, but considering two groups of trees: exotics and indigenous (or naturalized) trees.

DATABASE

Farm

Species

Trees

•Spatial position•Gender head•Size•...

•Uses•Exotic / Indigenous•...

•Number•Planted / Retained•Niche•Age•...

Characteristics

Farms

Species

Farms

Farms

Diversity statisticAverage diversity

statistic Randomize

Species

Speciesaccumulation

curves

Diversity profileaccumulation

patterns

Regression

Constrained ordination

Diversity profilesRank-abundance

and density-abundance plots

Sum

Accumulation

Use or nicheSubset

Figure 1.1. Linkages between the farms × species matrix and various analysis methods used in this thesis.

Characteristics

Farms

Species

Farms

FruitSpecies

Farms

Timber

-0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8 carpap

manind

psiguasyzcum

EbuchiebeShimutuMutambiMadidi

Speciesaccumulation

Speciesaccumulation

0 20 40 60 80 100 120 140 160 180 200

Number of farms

0

20

40

60

Nu

mb

er o

f sp

ecie

s

fruittimber

Ordination

Only trees of thatspecies that provide

fruit on that farm

Only trees of thatspecies that providetimber on that farm

Figure 1.2. Some analyses in this thesis were comparisons among matrices (example of comparison between species accumulation for fruit and for timber, Chapter 6), while other analyses were

comparisons within matrices (example of constrained ordination of farms for fruit, Chapter 9)

CHAPTER 2 THE RELATIONSHIP BETWEEN SPECIES DIVERSITY WITHIN AN ECOSYSTEM AND ITS STABILITY AND

PRODUCTIVITY R KINDT

ECOSYSTEM DIVERSITY, PRODUCTIVITY AND STABILITY

22

One of the objectives of tree domestication is to diversify agroecosystems to render them more stable and more productive (ICRAF 2000). The literature information that is presented here indicates that, in general, there is a positive relationship between ecosystem diversity (D), and ecosystem stability (S) and productivity (P). The literature, however, lists specific conditions under which the D~S and D~P relationships exist.

Most often in the listed studies, diversity is measured by species richness, the number of species present in the ecosystem, although some other measures of diversity have also been employed (formulas for diversity indices, such as the Shannon diversity index, are listed in Chapter 6).

In D~P and D~S studies, productivity is most often measured in terms of biomass (either above-ground or total biomass). Most often in studies of agricultural systems, however, productivity is measured as (harvested) yield, which is expressed in monetary units (typically US$/ha). Recently, natural ecosystems have also been evaluated economically (e.g. Bockstael et al. 1995; Ayensu et al. 1999; Costanza et al. 1999; Nunes & van den Berg 2001). Such studies include information on the value of ecosystem goods (e.g. food, timber, genetic resources, medicines) and services (e.g. water purification, flood control, coastline stabilization, carbon sequestration, pollination, waste treatment, biodiversity conservation, soil generation, disease regulation, maintenance of air quality, aesthetic and cultural benefits), since trade-offs among those goods and services have become the global rule rather than the exception. For example, a country may increase food supply by converting a forest to agriculture, but will, in consequence, decrease the supply of clean water, timber, biodiversity, and flood control (Ayensu et al. 1999). Nunes & van den Berg (2001) state that the available economic valuation estimates provide a very incomplete perspective on (at best the lower bounds of) the value of biodiversity changes, since not all benefits are validated. In this chapter and elsewhere in this thesis, however, ecosystem productivity was measured as biomass. Making certain assumptions on the link between species’ biomass and economic value, however, shows that D~P would apply both to the biomass and the economic value of an ecosystem (see below).

Greater stability in D~P and D~S studies is frequently measured as reduced (temporal) variation in biomass. However, stability of an ecosystem under perturbations can also be measured in terms of persistence (time before change), resistance (the degree of change) or resilience (time to return to equilibrium) (Pimm 1984). Organisms may therefore contribute differently to different aspects of stability – trees, for example, are less resilient but more persistent than short-lived organisms on a time scale measured in years (Pimm 1984). Agroforestry systems may therefore be more persistent but less resilient than agricultural systems without trees. For this thesis, the stability measure of decrease in biomass variation was adopted, especially since this was the measure used in most investigations by mechanistic models constructed to study D~P and D~S in parallel.

2.1. The Relationship between Species Diversity and Ecosystem Stability in Surveys and Ecological Experiments

Elton (1958) and Hutchinson (1959) provide theoretical grounds to prove positive D~S. Arguments that these authors used included the observation that simple population models had higher oscillations, the observation that small oceanic islands were more easily invaded, and the theoretical insight that more diverse systems provided more alternative pathways for energy to reach consumers. However, despite these longstanding theoretical grounds, positive D~S was still described as controversial at the end of the 20th century, because experiments and models had led to contradictory conclusions (Tilman et al. 1996). Johnson et al. (1996) indicated that four hypotheses had been proposed for D~S: (i) the diversity-stability hypothesis referring to a (linear)

CHAPTER 2

23

increment of stability with diversity; (ii) the rivet hypothesis pointing to a catastrophic collapse of stability below a certain diversity threshold level; (iii) the redundancy hypothesis referring to compensating species behaviours (increments in their abundance) in systems of moderate diversity, with decrease in stability below a specific diversity threshold; and (iv) the idiosyncratic hypothesis referring to no relationship between diversity and productivity. Chapin et al. (1998) also formulated positive D~S in the form of a hypothesis for further testing in more complex experiments than the ones that had been conducted already.

Among the earlier field studies that investigated the form of the relationship, Frank & McNaughton (1991) documented a positive relationship (R2=0.58) between the Shannon diversity index and 1 minus the cumulated changes in species proportions before and after drought. Rodríguez & Gómez-Sal (1994), using the same measures, demonstrated a negative relationship (R2=0.38). Tilman & Downing (1994) and Tilman (1996) showed a positive relationship (R2=0.21-0.22) between species richness and the relative rate of change in plant community biomass before and after drought.

Besides yielding contradictory results, the regression models used in the above studies only explained small amounts of variation (expressed by the R2 parameter, the coefficient of multiple determination = the square of the multiple correlation coefficient R; Legendre & Legendre 1998, p. 524) of stability by using diversity as an explanatory factor. This was especially the case in the two last studies where less than 50% in stability was explained. The low amount of explained variation for positive D~S as in Tilman (1996) indicated that a number of species-poor systems had relatively high stability, and vice versa that some species-rich systems had lower stability. Such pattern indicates a higher chance for systems with higher diversity to be more stable, but no certainty that the more diverse system will be more stable. The observed pattern further shows that pre-conditions exist for positive D~S, as otherwise all diverse systems would have had higher stability. Johnson et al. (1996), who reviewed a number of D~S studies of grasslands, hypothesised that seasonality was the underlying process, for example. Cottingham et al. (2001) mentioned that it is often difficult to isolate the effects of species richness from those of the identity of particular species used in experiments. Models (see below) have demonstrated that diversity in characteristics of species is more important than species diversity per se.

2.2. The Relationship between Species Diversity and Ecosystem Productivity in Surveys and Experiments

Darwin already hypothesised that more diverse plant communities would have more effective resource exploitation and thus greater community productivity (McNaughton 1994). However, as for D~S, D~P studies have led to contradictory results. The Waide et al. (1999) survey of ecological literature of over 200 studies (excluding intensively managed systems) found that 30 % documented unimodal, 26 % positive linear, 12 % negative linear, and 32 % non significant D~P. Classification of studies on ecological, geographical, or taxonomic scales also resulted in a mixture of relationships. For example, the respective percentages for studies (n=50) within plant communities were 24, 22, 12, and 42 %. Methodological problems may have influenced some results, since Oksanen (1996) warned that the relationship between abundance and biomass would be unimodal when fixed quadrat sample sizes are used for measurements, excluding any relationships between species. Loreau et al. (2001) expect that unimodal patterns of P~D in multilocational studies are caused by co-varying factors that influence productivity, such as soil fertility, climate, or herbivory – whereas experiments should investigate whether diversity alone has any effect. In any case, D~P was also listed at the end of the 1990s as a research area that needed further investigation (Chapin et al. 1998).


24

In analogy to D~S studies, analysis of D~P experiments, besides yielding contradictory results, managed only to explain relatively low percentages of variation. The Tilman et al. (1996) documentation of a positive relationship between exp(Shannon index) and total plant cover had R2=0.18. In a multiple linear regression analysis of the same experiment, Tilman et al. (1997a) found that only functional group diversity (for a definition of functional ecological groups, see below), and not species diversity, significantly contributed to productivity, but with R2=0.09. The positive relationships found by Hooper & Vitousek (1998) between functional group richness and the reduction of inorganic N pools had R2=0.26-0.35. Naeem et al. (1994) reported a positive link between species richness of different growing chamber communities and cover, and a negative link with the percentage of transmittance of photosynthetic radiation. At the end of the study, the standard error bars for productivity of communities with high (31 species) and medium (15 species) richness overlapped, however. The negative relationship between species richness and harvested biomass documented by Smith & Rushton (1994) had R2=0.26. Conclusions about low percentages of variation can be similar as for D~S: even in situations where there is a general positive D~P, there is no guarantee that a particular system of higher diversity will be more productive. Therefore, specific conditions for a positive D~P of a particular system must exist.

The study that has ignited much recent controversy about D~P is the BIODEPTH pan-European grassland experiment, from which a positive relationship between diversity (species richness and functional group richness) and above ground biomass was reported that explained 17.7 % of variation (Hector et al. 1999). The controversy focused on the possibilities that D~P resulted from ‘overyielding’ or from species’ sampling effects (Huston et al. 2000; Kaiser 2000a, 2000b; Purvis & Hector 2000). Overyielding is defined as a synergistic relationship among species so that the productivity of the mixture is higher than that of any species growing in isolation. The ‘sampling effect’ is based on a model whereby the biomass of a species mixture equals the monoculture biomass of the most productive species (Loreau & Hector 2001; the dominance scenario described by Yachi & Loreau 1999 – section 2.3). Such model results in D~P as random species mixtures that contain more species will have greater chances of sampling the more productive species. However, under the sampling effect, each combination of species that contains the most productive species will have the highest productivity – including monocultures of the most productive species. The controversy originated from experimental limitations since not every possible combination of species could be tested because of the enormous number of species combinations that are possible (the number of combinations can be calculated as St!/(Sm! (St-Sm)!), where St stands for total species richness and Sm for species richness in the mixture). Loreau et al. (2001) mention that testing the hypothesis that there is a minimal set of complementary species that can explain diversity effects ideally requires replicated testing of the performance of all species combinations at all diversity levels, which is often difficult. They suggest that future biodiversity experiments should focus on what individual species do, in order to investigate the hypothesis that a few dominant species suffice to explain D~P.

Loreau & Hector (2001), analyzing a subset of BIODEPTH data, distinguished between ‘selection effects’ and ‘complementarity effects’ of biodiversity, and were able to demonstrate positive D~P through complementarity. Complementarity among species can result from niche differentiation (resource partitioning) or facilitation among species (positive interactions). Complementarity can be negative, however, as it is measured as the change in the average relative yield of the mixture (formulas and a theoretical example of the calculation of selection and complementarity effects have been provided in section 2.7). Loreau & Hector (2001) described the sampling effect as too restrictive as a general alternative to complementarity. One reason they provided is that a community may not be dominated by a single species. Another reason they listed stated that sampling may also result in selection of complementary species and not necessarily only of those species that are more productive in monocultures. They therefore used the selection effect described by Price (1970) for evolutionary genetics, which can be calculated as

CHAPTER 2

25

the difference between the net biodiversity effect (calculated as observed yield – yield expected from monoculture yields and species’ proportions, see section 2.7) and the complementarity effect.

In response to the recent controversy, Tilman et al. (2001b) investigated D~P in a 7-year grassland experiment. At the end of the experiment, total and above-ground biomass of 16-species plots was 22% and 27% greater than for 8-species plots, respectively. In statistical analyses where species with small monoculture biomass were removed, total biomass remained significantly dependent on species number. The hypotheses that D~P resulted solely from (i) short-lived transients caused by species with high growth rates; (ii) sampling effects of low-viability species; or (iii) sampling effects of the most productive species, could be rejected, thus indicating that niche complementarity contributed significantly to D~P.

Distinguishing between selection effects and complementarity effects would also benefit analysis of crop mixture studies. Trenbath (1974) found that species mixtures produced more than monocultures in 20 percent of 572 experiments. Other examples of crop mixtures performing better have been provided in Dover & Talbot (1987) and Trenbath (1999). Swift et al. (1996) stated that intercrops might perform better, but that better performance could not be generalised as it depended on how much the component crops use resources differently. Zhu et al. (2000) demonstrated that growing two different rice varieties resulted in higher productivity as damages by diseases were reduced.

2.3. Investigation of Biodiversity Effects in Model Ecosystems

Because in biodiversity experiments often only a fraction of all possible species combinations can be tested, theoretical studies provide a more extensive method to test biodiversity effects. Models are also ideally suited to investigate processes that potentially contribute to D~S or D~P, since it are these processes that are simulated in the model ecosystems. Recent models have shown that incorporation of differentiation in species’ characteristics (corresponding to environmental heterogeneity) is the underlying mechanism that explains D~S and D~P. Tilman (1994) stated that continuous interspecific trade-offs between species abilities (related to species morphology, physiology and behaviour) to deal with different environmental constraints would result in a potentially unlimited number of species that can coexist in a habitat with spatial heterogeneity in resource supply. The models and theory can be expanded to incorporate intraspecific differentiation, since the models are simply based on heterogeneity in characteristics of component individuals, not on the mechanisms of geneflow and evolution within the ecosystems.

Tilman et al. (1997c) concluded, based on three models of randomly chosen species competing for a single or two resources, that average productivity would increase and variability would decrease if mixtures contained more species. For competition for a single resource, a monoculture of the most productive species would have resulted in the same productivity – the ‘sampling effect’. For heterogeneous habitats, some of the more diverse communities always had higher productivity than less diverse communities. The underlying effect of D~P and D~S in the three models was variation among species in parameters that describe the way in which they obtain resources.

Loreau (1998) built a model based on an ecosystem limited by a single nutrient, with species differing in resource use efficiency and resource depletion zones. Productivity increased with diversity based on complementarity and higher mean resource-use intensity at higher diversity levels. Richness had diminishing effects on productivity.


26

Yachi & Loreau (1999) demonstrated the ‘insurance effect’ of species richness on ecosystem productivity in fluctuating environments through a reduction in temporal variance of productivity and an increase in the temporal mean of productivity in case of ‘dominance’. Dominance indicates situations where ecosystem productivity can be approximated by the most productive species, while equivalence implies that ecosystem productivity will be the average productivity of individual species. In real ecosystems, productivity is expected to lie in between both dominance and equivalence extremes. Asynchronicity of species responses to environmental conditions was the basis for the diversity effects, so that perfect correlations of species responses resulted in no biodiversity effect. For the dominant productivity scenario, increasing the variance of species responses decreased ecosystem redundancy (= the smallest species richness level where the maximum temporal mean of productivity is reached).

Nijs & Roy (2000) built another model involving different species that differed in nutrient uptake competing for a single resource,. These authors demonstrated that more species-rich and more inter-specifically different communities had higher productivity. Their model showed the sampling effect where species vary in their exponential growth characteristics.

Ives et al. (2000) expanded their investigations of D~S and D~P (which also demonstrated the insurance effect, Ives et al. 1999) to model ecosystems with several trophic levels (predator-prey relationships). They demonstrated that variability in prey or predator densities primarily depended on the number of species-environment interactions rather than the number of species-species interactions. The reduction in variance was eliminated when all species experienced environmental variation in the same way.

All the models cited above have demonstrated that differences in species-environment interaction and environmental heterogeneity result in D~S and D~P for random species mixtures in environments with random characteristics. Tilman (2001) explained that random species mixtures were needed to test pure effects of diversity. Berendse (1994) pointed out that natural communities are not assembled at random, but species by species, and that continuous selection affects interactions among species. McCann (2000) mentioned that the condition of generally weak bioenergetic consumer-resource interactions needed to be added so that diversity could enhance stability. Such statements indicate that, although D~S and D~P have been demonstrated for theoretical ecosystems, understanding of the functioning of an actual ecosystem will need to be based on characterization of species, environment and species-environment interactions. The concept that other species are part of the environment of a particular species needs to be modeled as well to simulate effects such as facilitation among species. Effects of the spatio-temporal distribution of environmental characteristics on D~S and D~P will need to be explored more extensively, rather than assuming random distribution.

The multi-species model of Norberg et al. (2001) (reviewed in Tilman 2001) tackled some of the problems mentioned in the previous paragraph. They simulated the link between evolutionary dynamics and ecosystem dynamics by modeling the influence of temporal dynamics of the environment on distribution of species’ abundances. Biomass of each species was a function of its phenotype, the value of the environmental variable, total biomass of the community, and external sources (typically dispersal). Their model demonstrated that an adaptive ecosystem would have about 6 % higher productivity than an ecosystem consisting of the single species with the highest possible long-term productivity. Short-term productivity was reduced by increased phenotypic variance of the species in the ecosystem. A net positive D~P was caused as phenotypic variance results in a higher rate at which traits can track environmental change. External inputs were necessary to maintain phenotypic variance, which indicated that sites should not be isolated from the surrounding area.

CHAPTER 2

27

2.4. Functional Ecological Groups

Studies of species-rich ecosystems often attempt to reduce complexity by investigating functional groups rather than individual species. The findings from recent models on the importance of differences between species characteristics in understanding D~P and D~S also point to the possibility of clustering species with similar characteristics in functional groups. Functional groups can be defined as clusters of species that play the same role in maintaining and regulating ecosystem processes (Gitay et al. 1996). When one species is removed from its functional group, other species may increase in abundance taking over its role. Schulze & Mooney (1994) point out that functional groups will be differently defined depending on the processes that are being studied. Norberg et al. (2001) define functional groups as clusters of species that share similar resources and predators. Colasanti et al. (2001) differentiate between functional groups and functional types, where one species may be simultaneously a member of several functional groups, but not of several functional types. Keystone species can be seen as monospecific functional groups. Where macro-level functional integrity (such as productivity) of ecosystems may be preserved by maintaining functional groups, their structural integrity (such as resilience) may be weakened when diversity within groups is diminished (De Leo & Levin 1997). Chapin et al. (1997) make a similar analysis as De Leo & Levin (1997), and state that as time scales increase, ecosystems will experience a wider range of conditions, and that the importance of diversity among functionally similar species will increase. They conclude, therefore, that no two species are ecologically redundant, even if they appear similar in their ecosystem effects under one particular set of environmental conditions. A debate however exists in the conservation literature on whether the functional group concept could be used to plan conservation. Gitay et al. (1996) for example pointed out that not all functions of an individual species within an ecosystem are known, so that the effects of species removals cannot be predicted, while Walker (1995) stated that the functional group concept was necessary for practical conservation planning.

The discussion on functional groups is actually focused on the question whether all but one of the species within the same functional group are redundant (i.e. ‘absolute redundancy’ within functional groups), or whether functional groups are merely collections of species with similar ecosystem functions. The models with two limited resources presented by Tilman et al. (1997c), and small variation in species characteristics in the model developed by Yachi & Loreau (1999), are theoretical cases where absolute redundancy of a species may not occur. Increasing the diversity for systems with moderate or even high diversity may still result in small enhancements of stability or productivity.

Notwithstanding the question whether absolute redundancy occurs in a specific ecosystem, models have demonstrated saturating effects of increased biodiversity, resulting from a process whereby each extra species added to a mixture where each species has random characteristics, will on average differ less from the species that are already present. Holmes (1998) stated that, because the effects of biodiversity on ecosystem functioning saturated, the biggest benefits of biodiversity enhancement could be expected for ecosystems that are at present lowest in diversity, that is to say particularly agroecosystems and forest plantations. Tilman (1999b) stated that total primary productivity could be expected to increase 35-70% when increasing plant species diversity from one to about 20 species through a combined sampling effect and niche differentiation, implicitly acknowledging limited effects of increasing diversity beyond 20 species. Cottingham et al. (2001 eq. 1) described that variability of an ecosystem (var(E)) can be dissociated into the sum of the variance of the individual species and their summed covariances:

+= ∑∑∑

=

−

==

N

i

i

jji

N

ii SSSE

1

1

11

),cov(2)var()var( .


28

The consequence of this dissociation is that when species vary independently, their covariance is zero, and the variance of the ecosystem equals the sum of the variances of species. However, when species do not vary independently, overall variability may increase or decrease. Therefore, factors that change the summed variances or covariances should be the major focus of D~S studies. Cottingham et al. (2001), therefore, advocated calculating variances and covariances for each experimental unit to investigate how species covariances changed across diversity gradients. Van Noordwijk & Ong (1999), based on theoretical consideration of variance dissociation, concluded that where covariance was zero, system variance would decrease with N-1. Increasing N will result in smaller effects on system variance under these conditions. Considering the role of species characteristics in D~S and D~P, enhancing diversity by introducing species with characteristics that are less correlated with the characteristics of species that are already present will result in larger effects. Information on functional groups (clusters based on species-environment relationships) can therefore guide diversification if such information is used to select species with characteristics that are currently rare in the ecosystem and for which there is an ecosystem niche.

2.5. The Role of Disturbance

The vulnerability of complex ecosystems such as tropical rainforests and coral reefs has been cited as evidence for a negative D~S (e.g. Dover & Talbot 1987: 24). As models have demonstrated that diversity effects are based on variation in species characteristics’ in respect to environmental characteristics, disturbance effects should also be investigated within the context of species-environment interaction.

Disturbance has been listed as an explanatory factor for diversity. Reice (1994) stated that natural systems are usually not in equilibrium and that habitat disturbance and heterogeneity result in the enhancement of biodiversity. Phillips et al. (1994) found that dynamism (the mean of recruitment and mortality, indicating small-scale disturbance) explained most of the variation in species richness among 25 mature forests from all major tropical regions, and that more dynamic forests were more species rich and more productive. Wills et al. (1997) found that intraspecific negative density-dependent effects (for example a negative correlation between the basal area of conspecifics, and the ratio of (births-deaths)/total of trees of the species) were more common than interspecific effects in maintaining diversity in rainforests. They listed biotic effects (especially pests) as more likely causes for diversity rather than abiotic effects (such as nutrient removal and toxicity), and therefore advocated for the retention of all pathogens to preserve rainforest biodiversity. Givnish (1999) proposes that “random drift over evolutionary time in the relative effectiveness of density-dependent control by specialist natural enemies may account both for the observed distribution of tropical tree abundance and for the repeated dominance of particular taxa in separate but ecologically similar sites”. Nevo (2001) lists evidence that genetic diversity is maintained by environmental diversity, and that this pattern surprisingly also occurs for non-coding levels of DNA.

The intermediate disturbance theory is based on the competitive exclusion principle whereby superior competitors ultimately replace all other organisms that are more superior in colonization. The highest richness of species that vary in colonization and competition traits then occurs at intermediate levels of disturbance (Van der Maarel 1993; Weithoff et al. 2001). Tilman (1994) points out that successional dynamics also explain the maintenance of biotic diversity because trade-offs exist between colonization and competition traits of species. Recurring disturbances in an infinitely large habitat then allow for the persistence of an unlimited number of species. McNaughton (1994) indicated that diversity would act as a stabiliser of ecosystem function if higher diversity resulted from greater niche packing, with disturbances creating niches

CHAPTER 2

29

that exist or existed elsewhere in the landscape followed by establishment of suitable species in new niches. Loreau et al. (2001) mention that diversity loss at a regional scale or dispersal limitations due to landscape fragmentation may reduce the pool of potential colonists and limit the potential to change species composition to locally adjust to environmental changes.

Novel disturbances may not initiate succession, and such disturbances would not lead to higher biodiversity. Tilman & Lehman (2001) mention that novel environmental changes may lead to substantial losses of species richness, especially where species abundances are limited by multiple environmental factors. Woodruff (2001) identified the current anthropogenic homogenisation of biotas as the major driver for the current biodiversity crisis, when global rates of species extinction are at an all time high when calculated for the last 65 million years. Chapin et al. (1997) mention that the current global extinction rate is 100 – 1000 times greater than prehuman levels.

Tilman & Lehman (2001) mention that novel environmental constraints would create conditions for new diversification. Kirchner & Weil (2000) and Novacek & Cleland (2001) state, however, that recoveries from mass extinctions in the fossil record are measured in millions of years. Kirchner (2002) expects that our planet will be biologically depleted beyond the lifespan of the entire human species, since investigations of the geological record revealed that evolution did not accelerate quickly in response to rapid extinction. A possible explanation for this process is that extinction does not only eliminate species, but also ecological niches. Ecosystems must, therefore, first increase in complexity so that there are new niches – new roles for new organisms in the ecosystem – to fill, before new species can evolve. In analogy, Thompson (1999) stated that the history of evolution and biodiversity was fundamentally a history of coevolution, since most living organisms required a combination of their own genetic machinery and that of one or more other species to survive and reproduce.

Although disturbance that maintains spatio-temporal environmental heterogeneity in the ecosystem drives D~P and D~S, Ives et al. (1999) stated that higher species richness was also an insurance against unknown future changes. Pachepsky et al (2001) demonstrated that a relationship between time of reproduction and (progeny per time unit/lifetime) and weak stochastic disturbance reproduced natural biodiversity patterns (a log-normal abundance distribution of species or plant physiological types), while initial environmental heterogeneity in nutrient availability was not necessary. Tilman (2000) listed four reasons why species could coexist, based on interspecific trade-offs between (i) competitive abilities and dispersal abilities; (ii) competitive abilities and susceptibility to disease; (iii) abilities to live off average conditions and abilities to exploit resource pulses; and (iv) abilities to compete for alternative resources in a heterogeneous landscape. Based on the modeling work of Pachepsky et al. (2001), trade-offs in time of reproduction and fecundity could be added as a fifth reason. Pachepsky et al. (2001), by modeling plant physiological types rather than species, further demonstrated that trade-offs in individual traits enabled coexistence and that intraspecific variation should therefore not be ignored. It is possible that other trade-offs enable coexistence of various species or plant types – the role of disturbance is then to remove some individuals and enable the establishment of new species or plant types.

Disturbance leading to greater diversity, and diversity leading to greater stability (D~S), may seem paradoxical. However, Van der Maarel (1993) listed disturbance ‘for’ stability as the best description for the relationship between both factors. Patchiness, time-scale, and intensity of disturbance were listed by this author as major factors in the relationship between disturbance and stability, with disturbances of low intensity and small extent leading to stable communities, while highly intense disturbances at the landscape level may not result in stability. Van Noordwijk & Ong (1999) and Loreau et al. (2001) also pointed at the various scales in ecosystems, where stability at lower level is not necessary for stability at higher levels and vice versa. The role of disturbance in D~S and D~P primarily lays in maintaining diversity both of species (individual traits) and environment. Any successful effort to constrain the natural variability of a system


30

could eventually lead to its fragility (De Leo & Levin 1997). Scheffer et al. (2001) also state that natural disturbance promotes diversity and renewal processes in ecosystems. These authors further advocate to concentrate conservation efforts more on eliminating gradual changes that affect ecosystem resilience since alternative stable states exist for many major ecosystems. Whereas natural disturbances lead to ecosystem stability, slow changes in variables of land use, nutrient stocks, soil properties and biomass of long-lived organisms may lead to catastrophic shifts between ecosystem states, from which recovery to the alternative ecosystem state will not be easy since hysteresis is involved.

2.6. Conclusions

Ecological theoretical considerations, surveys, experiments and models have demonstrated that there is a positive, but saturating and conditional, relationship between D~S and D~P. Conditions for positive D~S and D~P include diversity and trade-offs in traits of species (or individuals), diversity in environmental characteristics, and disturbance that maintains turnover of species and spatio-temporal variation. Diversity of species that are more similar in traits will have smaller effects on S and P, than similar diversity that offers a larger variation in traits. Differences in the variation of species traits at a fixed diversity level were probably responsible for the low percentages of explained variation in experiments, while larger variation in traits at lower species diversity would result in negative D~S and D~P in some surveys. To understand D~S and D~P in ecological surveys and experiments (i.e. in the field), not only species diversity should be measured, but also heterogeneity in species’ traits, characteristics of the environment, and the relationships between species’ and environmental characteristics – factors that are not as easily measured as species richness and yield. Cottingham et al. (2001) concluded that future efforts should focus not only on demonstrating that diversity has an effect on stability, but also on identifying when this phenomenon is likely to occur. Loreau et al. (2001) state that experiments are needed in which both diversity and environmental fluctuations are controlled. Whether heterogeneity is better measured by the discrete distribution of categories of species and environmental conditions, or by documenting the variance in characteristics of species and the environment, is an issue that needs to be resolved. Norberg et al. (2001) proposed to use phenotypic variance as ‘D’, whereas Tilman (2001) proposed a hybrid measurement by combining species richness and phenotypic variance. Similarly to the options of classification or ordination of vegetation, the nature of spatio-temporal arrangement in the focus ecosystem will probably determine the best abstraction method.

The studies have indicated that in a heterogeneous environment, diversity of species will lead to larger stability and productivity. The models have, however, also demonstrated the reverse phenomenon, that in a homogeneous environment, choice of appropriate species at low diversity levels will enable stable and productive ecosystems. Agroecosystem managers have two options when attempting to increase stability and productivity in a heterogeneous environment with low species diversity: (i) to increase species diversity (mimic the diversity of the natural system); or (ii) to reduce environmental heterogeneity. However, the practical questions for the first option are where and when which species should be introduced – Van Noordwijk et al. (1994), discussing the “not to put all eggs in one basket” option for agroecosystem management stated that the practical questions were ‘which baskets?’ and ‘how to allocate eggs?’ rather than ‘how many baskets?’. For the second option, the practical question is how all heterogeneity, including pest and disease dynamics, can be reduced in a sustainable matter and what the consequences of global climatic changes would be.

In the introduction of this chapter, we stated that D~P would also have been recorded if productivity had been measured economically, rather than in terms of biomass, in case certain

CHAPTER 2

31

assumptions were made on the relationship between the biomass of a species and its economic value. The ecological models assumed random variation among growth rates of component species. (Such variation among species lead to the analytical problems of distinguishing between sampling and complementarity effects in biodiversity experiments, as documented in section 2.2.) Conceptually, growth rates could be interpreted as those for the economical value of each species rather than for their biomass, if we assume that a (random) fraction of total biomass is harvested for every species and that a (random) species-specific value is obtained for each unit of harvested biomass. Possibly, the species with highest biomass would not have the highest economic value. However, as long as trade-offs exist among species in their environmental responses, then the models would document positive D~P under conditions of a fluctuating environment. These conditions exclude the case where the same species always has the highest productivity, including the situation where only one species has economic value. Theoretically, dynamics of supply and demand could counterbalance changes in economic value in relation to changes in biomass ( i.e. a higher economic value when less biomass is produced, whereas we assumed a fixed value per unit of harvested biomass). Some sets of market dynamics could then result in the case where one species always has the highest productivity. Therefore, the absence of such market dynamics are an additional condition for positive D~P, when P is measured in economical units.

For agroecosystem management, other factors than the stability or the productivity of the complete agroecosystem may be important. Farmers may for instance opt to produce various products and services that are needed by an individual household, which do not necessarily constitute largest potential productivity. Concentration on those species with the largest productivity or profitability could be very risky in a fluctuating market environment. The findings on saturating D~S and D~P and consideration of diversification to limit risks in agroforestry production systems, strongly advocate to increase diversity especially where diversity is low. In analogy to the biodiversity hotspot approach to conservation (Myers et al. 2000), spots of low diversity could be identified in agroecosystems for which diversification would result in largest effects on stability and productivity.


32

2.7. Addendum: Additive Partitioning of Biodiversity Effects

Loreau & Hector (2001) proposed partitioning the net effect of biodiversity in ecosystem productivity into complementarity and selection effects to study the function of biodiversity. Complementarity and selection effects are based on information on the yield observed of each species (i) in monocultures (Mi) and in the mixture (YOi), and the proportion of each species in the mixture (RYEi). Table 2.1 provides information on calculation of complementarity and selection from these statistics through a theoretical example. Based on the figures, it can be demonstrated that the net biodiversity effect for the mixture, YO-YE = 40, can be partitioned into the complementarity effect, 3 (? RY)average (M)average = 3 (0.1)(200) = 60, and the selection effect, 3 cov(? RY, M) = -20.

Table 2.1. Calculation of statistics used to calculate complementarity and selection effects. M RYE YO YE RYO ?RY ? Y (1) (2) (3) (4)=(2)(1) (5)=(3)/(1) (6)=(5)-(2) (7)=(3)-(4)=(5)(1)-(2)(1)=[(5)-(2)](1)=(6)(1)

species1 100 0.5 70 50 0.7 0.2 20 species2 200 0.4 100 80 0.5 0.1 20 species3 300 0.1 30 30 0.1 0 0 total 200 160 40 average 200 0.1 M: monoculture yield; (R)Y(o/e): (relative) observed/expected yield

CHAPTER 3 DESCRIPTION OF THE SURVEY AREAS AND DATA

COLLECTION METHODS R KINDT

SURVEY AREAS

34

Information that was analyzed in this thesis was collected during six surveys of on-farm tree diversity conducted in tropical Africa during 1998 - 2001. Figures 3.1 - 3.4 show the locations of the study sites. Since the selection of farms within each agroecosystem was partially based on their relative position to nearby forests, the location of major forest zones has been included in figure 3.2. Information was collected from complete farms defined as all land that was managed by a household.

Although information was ultimately available from six locations in four countries (Cameroon, central Kenya, central Uganda, Nigeria, south-western Uganda, and western Kenya), most analyses that are reported in this thesis (Chapters 5 - 12) were based on data from western Kenya only. This was the location of the pilot survey for on-farm diversity studies. For this reason, information is more extensively provided for the western Kenyan site. Additional description of the study sites, especially for the other surveys, can be found in the cited references. Questionnaires and databases used for data collection, storage, and analysis were similar in the various studies.

3.1. Western Kenya

3.1.1. Study Area

Complete tree species inventories were made in 201 farms in the Vihiga and Kakamega districts of western Kenya in the period of January 1998 – April 1999. The farms consisted of 50 (in one case 51 – we discovered during the survey that one farmer had already handed over part of his farm to his son) randomly selected households in four villages.

In a follow-up survey on 78 of the 201 inventoried farms (May 1999 – April 2000), farmers were interviewed more in depth about the management of interspecific diversity of trees on their farms (Chapter 11). For the second survey, it was intended to collect information on all 50 farms of one village, and a subset of 68 farms of the other villages. The subset of farms of each village was selected in a randomly-stratified manner using information on socio-economic characteristics and species richness collected in the first survey. Due to time constraints caused by delay in data collection and entry (the last interviews were only completed in December 2001), information could only be analyzed for 78 of the 118 selected farms (table 3.1).

Figure 3.3 shows that the study area belongs to the Victoria Basin forest-savanna mosaic (AT0721). This ecoregion is noted for its high species diversity and endemism which results from the mixture of habitat types. These include more than 310 tree species, 280 species of birds, 220 species of butterflies, and 100 species of moths. The forest habitats in the ecoregion have been mostly replaced by savanna, farmland, and pasture. The remaining forest patches are small and fragmented. (source: http://www.nationalgeographic.com/wildworld/terrestrial.html).

Warner (1995) describes common features of farming systems in eastern Africa. Most systems provide security of land tenure, rural households experience labour shortages and a growing dependency on off-farm income, and trees are widely planted throughout the region. Households place more reliance on New World crops such as maize (Zea mays), cassava (Manihot esculenta), potatoes (Solanum tuberosum), sweet potatoes (Ipomoea batatas), and beans (Phaseolus vulgaris), while beer brewing is everywhere important.

The study area lays within the East and Central African Bimodal Highlands, characterised by altitudes above 1000 m above sea level (all altitudes recorded in this chapter are above sea level) and bimodal rainfall of more than 1000 mm per annum (Hoekstra 1988). Warner (1995) provided some general features of this zone, which includes Burundi, Kenya, Rwanda, and Uganda. Trees are retained for fuel in the homestead area, exotic fruit trees are preferably planted in the

CHAPTER 3

35

homestead area, while indigenous and exotic trees (especially Markhamia spp., Ficus spp., Euphorbia spp. and Grevillea spp.) are planted on boundaries, in fields and in homesteads. Farmers prefer to plant trees that address several of their needs for poles, construction materials, fuel, and soil fertility improvement. Selected indigenous trees are retained. In some systems, indigenous trees are expected to grow naturally and should not be planted.

The highlands of Kenya cover an area of 85,000 km2 and accommodate 8 - 10 million inhabitants, corresponding to 15% of Kenya’s total land area and 40-50% of its total population (Hoekstra 1988). The annual rainfall in the area is positively correlated to altitude, except where features such as lakes interfere (Warner 1993). Of the total area, 59% consists of undulating landscapes with slopes between 2-8%. The soils are erosion prone because of the high land use intensity (Shepherd et al. 1997). Holmgren et al. (1994) described the Bimodal Highlands in Kenya as an area where tree cover on farms is positively correlated with population density and has thus been increasing over the last decades.

Carter & Crowley (unpublished results) provide a description of changes in landuse in western Kenya from 1900-1990. Landuse changed drastically in the area from 1915 onwards through measures from the colonial administration. These measures included the formation of the North Kavirondo Native Reserve to which the African population was restricted, the imposition of taxes, and the restriction of Africans to cotton (Gossypium spp.), maize (Zea mays), and sesame (Sesamum indicum) as cash crops. Before that period, agriculture had been characterized by mixed farming, with sorghum (Sorghum bicolor), millet (Pennisetum glaucum), indigenous vegetables, maize, and bananas (Musa spp.) as the staple foods, and sweet potatoes (Ipomoea batatas) as hunger foods. Because of the colonial measures and the introduction of the iron hoe, the area under maize expanded and intensified, leading to land degradation, clearance of forest, and food shortages. The imposed shift of focus in the farming system to cash generation, in combination with population growth, also lead to male migration (as not enough cash could be generated through maize), the adoption of exotic vegetables for cash, and the increase of area under bananas as food supplement and cash. Farmers also started growing trees for the fuelwood market of Kisumu City. The colonial government tried to counter the negative effects of their measures by making soil conservation compulsory, by initiating afforestation, by stimulating farmers to grow root crops to increase food security, and by gazetting much of the southern part of Kakamega Forest, but negative effects prevail until today. Consolidation and limited expansion of land holdings, the conversion of bush, scrub and riparian woodland to arable land, the decline of numbers of trees in land, and the increase of trees in hedges characterized the period after Independence. In the same period, crossbred cattle were introduced for milk with Napier grass (Pennisetum purpureum) as feed (which also became a cash crop).

The study area is inhabited predominantly by the Luhya (Luyia) ethnic group and belongs to agroecological zone Upper Midlands 1 (Tea-Coffee Zone) described by Jaetzhold and Schmidt (1983). UM1 is a zone with permanent cropping possibilities consisting of two or three seasons. In the zone, altitude ranges 1500-1800 m, annual mean temperature ranges 18.1-20.4 °C, and annual rainfall ranges 1600-2000 mm. The limitation to one ethnic group and one agro-ecological zone was a deliberate choice to limit potential influences on diversity characteristics observed.

Four villages were selected within the area, each located in a different stratum as identified by Bradley et al. (1985) and Bradley (1991), through the interpretation of low level aerial photographs and confirmed during an on-farm survey. Table 3.1 summarises information on each stratum and lists the villages where the interviews were undertaken.

The selection of villages coincided with a gradient towards the Kakamega Forest National Reserve (0º10’-0º21’ N (UTM 18.43-38.70), 34º47’-34º58’ E (UTM 698.48-718.88)). Kakamega Forest is the easternmost remnant of the Guineo-Congolian rain forest that in the past millennia stretched across the entire expanse of central Africa. The reserve is situated at an altitude of 1400

SURVEY AREAS

36

- 2300 m and encompasses an area of 24,000 ha. Up to 20% of all Kenyan plant and animal species occur only in this forest, including 75% of all butterflies. The reserve contains a unique mixture of highland and central African lowland species. Some 380 species of plants occur in swamps, riverine, and hardwood forest areas, glades and the shallow forest around the edge of the reserve. There are over 150 documented species of woody trees, shrubs and vines, 170 species of herbs (of which 60 are orchids) and 62 species of ferns. There are over 350 species of birds that reside in the forest, including rare snake-eating birds, while several primate species are also present (Cox et al. 1999; Ipalei 1999).

3.1.2. Socio-economic Characteristics

Various surveys had previously collected information on household characteristics. David & Swinkels (1994) provide one of the more recent sources of household characteristics of Vihiga and Kisumu districts. Results of this survey (n=75) showed that 87% of households had at least one cow. Households with crossbred cattle were wealthier. Over half of the household heads were 51 or older. The average number of resident family members was 7.3 persons for Vihiga, of which 3.7 were adults (> 15 years old). For both districts, 13% of heads of household had not received any education while 32% had received some form of secondary education. Heads of household were 71% male, 12% de jure female (single, divorced or widowed woman) and 17% de facto female (husband being away for long periods). Using permanent houses, corrugated iron roofed houses and thatch-roofed houses as wealth indicators categorised 5% of households as wealthy, 73% as average, and 22% below average, respectively. Data collected by Kindt (1997) through a survey in Vihiga and Kakamega districts (n=53) gave similar results for the percentages of heads of household older than 51, male-headed households and thatch-roofed houses, and for education levels of heads. Data from this survey differed for wealth indicators with 12% permanent houses as opposed to the 5% reported by David & Swinkels (1994).

Table 3.2 shows quantitative household and farm characteristics used as explanatory variables in regression (Chapters 8 & 14) and ordination (Chapter 9, 10 & 14) analyses, while table 3.3 gives an overview of the respective qualitative characteristics. The analyses could not be done for the complete dataset because of missing values in household characteristics – the complete set of explanatory characteristics was available on only 183 of the 201 inventoried farms (tables 3.2 and 3.3). For the regression and ordination analyses, qualitative factors were analysed by s-1 dummy explanatory variables coding for the s states of the factor. Quantitative variables were ranged within the [0,1] interval by subtracting the minimum value and subsequently dividing by the maximum value.

Comparing values presented in tables 3.2 and 3.3 with those collected in other surveys revealed that a reasonable socio-economic cross-section of the population was interviewed. Farm sizes as communicated by farmers were not verified in the field, but averages for Ebuchiebe (0.52 ha), Shimutu (0.93 ha) and Mutambi (0.42 ha) corresponded to the averages of the strata presented in table 3.1. An exception was the average farm size in Madidi of 0.40 ha. In comparison with characteristics from previous surveys, the present survey contained fewer households with cattle (71%), fewer household heads were older than 51 (38%), the number of resident household members was smaller (5.8), fewer household heads had received some secondary education (18%) and fewer male-headed households were encountered. The observed differences were however not very large. For each explanatory variable, the coefficient of multiple determination (R2) was calculated using the other variables as explanatory variables in a regression model. This statistic expresses the amount of variance in the response variable that can be explained by explanatory variables (Legendre & Legendre 1998, p. 524). When the variance inflation factor of a variable (defined as (1-R2)-1) is large (> 20 i.e. R2 > 0.95), then its regression coefficient is unstable (ter

CHAPTER 3

37

Braak 1986). As can be seen from tables 3.2 and 3.3, no variable had large R2 and, therefore, variance inflation factors were low.

Data averages for villages (not presented) showed that household characteristics were not evenly distributed over villages. Shimutu village was most different from the other three villages with more cattle (but fewer crossbreeds), fewer residents, lower education levels, and more male-headed households than the other villages.

Comparing the household characteristics of the second survey (tables 3.2 and 3.3), revealed that the subset of villages still represented a reasonable socio-economic cross-section of the population, despite the fact that information was not analyzed for all 118 farms for which the second interview was planned (see section 3.1.1). Only variables that coded for villages had moderate inflation factors due to the strong correlation between the dummy variable for Ebuchiebe and Shinyalu (R2 = 0.77). The strong correlation was the result from the fact that only four farms belonged to other villages.

3.1.3. Data Collection on Tree Species

For every tree species encountered on a farm, information was collected on the abundance in particular on-farm niches of the various cohorts (all trees established in the same period) by participatory interviews with household members involving farm walks, tree counting by the interviewer, and data recording using pre-tested questionnaires (Appendix II). On-farm niches for trees refer to the location on the farm and the establishment pattern of trees at the location. The niches that were distinguished for western Kenya were for example trees in the homestead area, trees mixed in cropland, trees on contours in cropland, trees on external boundaries of the farm, trees on internal boundaries on the farm, trees in woodlots, and trees in fallows.

Free responses on tree uses were obtained on a species-by-species basis. These answers were postcoded during data entry in the databases that were created for data analysis and storage. Respondents were also requested to name the main use of the species on the farm. Tree circumferences were measured 30 cm above ground (recommended for agroforestry species, MacDicken et al. 1991; Stewart & Salazar 1992) for every cohort. From tree circumferences and cohort abundance, cross-sectional area of species was calculated.

Information was provided by the farming household on the origins and the forms of germplasm of each tree cohort, and the expected replacement age of each cohort. Origins of germplasm were postcoded in categories including the own farm, other farms of the same village and outside the village. Forms of germplasm were classified as sexual with known origin of the mother tree (seeds and seedlings), wildlings (natural regeneration), and cuttings. Farmers were asked to provide the preferred form and origin of germplasm for each species. Where farmers did not have an opinion on the preferred form and origin, calculations for metapopulation dynamics (Chapter 12) used the most frequently preferred forms and origins of the species within each village, if these could be determined.

In a follow-up survey, a subset of farmers (table 3.1) was interviewed about desired modifications in tree composition using a mapping tool that documented present tree composition in various on-farm niches. The respondents were also asked for reasons for these modifications, why they had not been done, and specifically about the reasons related to desired tree diversity within each use category (Chapter 11). The questionnaire for this second survey is provided in Appendix II.

SURVEY AREAS

38

3.2. Cameroon and Nigeria

This part of the study was targeted at the humid forest zone of West and Central Africa, with four case-study villages in Southern Cameroon and two in Southeast Nigeria. The zone is located between 10°N and 6°S, and 30°W and 35°E and is characterized by altitudes of less than 1000 m above sea level. Annual rainfall ranges between 1400 mm and 4000 mm with bimodal distribution. The main daily temperature varies between 24 and 27°C. The soils are mainly ferric acrisols. The vegetation goes from matured secondary rain forest to degraded forest depending on demographic pressure and cultivation intensity. Cropping systems of the region are mainly based on slash-and-burn practice and produce food crops for home consumption and local markets. Tree crops, such as cocoa and coffee are usually cultivated as small-scale plantations mixed with other fruit trees, medicinal plants and timber species (Degrande et al. in prep.).

The case-study communities (located in four villages in Cameroon and two villages in Nigeria) were selected against a range of socio-economic (ethnic group, demographic pressure, access to markets, land availability) and biophysical criteria (agro-ecology) and households were sampled based on a participatory wealth-ranking exercise (Degrande et al. in prep.). During the wealth-ranking exercise, two men and two women from each village ranked all households in five well-being categories. From each category, four households were chosen at random, except for one village due to its small size.

Table 3.4 provides some information on the villages where the surveys were conducted. Whole-farm inventories of all exotic and indigenous trees, whether planted or not, were conducted in two Cameroonian villages in the period of 1999 - 2000. Previous (1998) fruit tree inventories were done in all villages for a study specifically targeted at the management of fruit trees. The definition of “farm” used in the study includes homegardens, food crop fields, fallow land, cocoa and coffee plantations, oil palm fields and/or small orchards, managed by a household.

Tables 3.5 and 3.6 show information on the farm characteristics that were used in multiple regression analyses (Chapter 14). The only variables with moderate variance inflation factors (R2

> 0.60) were those that coded for male heads and for the origin of the household head with respect to the village (indigenous/allogenous) due to the strong correlation between these factors (R2 = 0.75).

3.3. South-western Uganda (Kabale)

The on-farm tree diversity survey was conducted in the Kigezi Highlands in parts of Kabale district, located in western Uganda (bordering Rwanda). The Kigezi Highlands are part of the East and Central African Bimodal Highlands (see section 3.1.1).

The Kigezi Highlands are primarily an agricultural zone. The altitude ranges from 1200 to about 2800 m, annual mean minimum and maximum temperatures are 10 and 23°C, respectively, and annual bimodal rainfall ranges 1000-1500 mm. The district, especially where farms were surveyed, is characterized by ragged terrain with flat-topped hills dissected by steep-sided valleys. Gardens in valleys are usually reclaimed swamps. The area is heavily terraced as a result of colonial by-law enforcement, although terraces are increasingly crushed to provide productive soils to lower parts. Soils are fine textured and easily leached and eroded, especially on lower and mid-hill slopes. The landholding per household in Kabale ranges from 0.2 to 9.1 ha and is scattered over several hills in 2-20 plots (averaging 7) (Turyomurugyendo in prep.).

Figure 3.5 provides a map with the locations of the 928 plots where the survey was conducted in 2000 and 2001. Five parishes (clusters of villages) were selected on the gradient between the Bwindi Impenetrable National Park (BINP) and Kabale town, with two parishes adjacent to

CHAPTER 3

39

BINP, two parishes adjacent to Kabale town, and one parish at intermediate position in the gradient (names and positions are provided in table 3.7).

BINP is located from 0°53' - 1°08' S and 29°35' - 29°50'E and has an area of 32,092 ha. The park borders with Zaire to the west and its distance by road to the nearest main town of Kabale to the south-east is 29 km. Bwindi is characterised by steep hills and narrow valleys, with a general incline from the northern and western areas (below 1750 m), to the south-western corner (above 2250 m). The forest gets the name 'impenetrable' from the dense cover of herbs, vines, and shrubs inhabiting the valley bottoms. Bwindi is one of the few large expanses of forest in East Africa where lowland and montane vegetation communities meet. Combined with its probable role as a Pleistocene refuge, this situation has led to an extremely high biodiversity. Current evidence indicates that Bwindi is the most diverse forest in East Africa for tree species (more than 200 species) and ferns (more than 104 species), as well as other taxa (fauna). In recognition, Bwindi was selected by IUCN's Plant Programme as one of the 29 forests in Africa most important for conserving plant diversity. Bwindi is believed to hold the richest faunal community in East Africa, including over 214 species of forest birds (336 species in total), 120 species of mammals (including 7 species of diurnal primates), and 202 species of butterflies (84% of the country's total). Highly significant is the presence of almost half of the world's population of mountain gorillas Gorilla gorilla berengei (about 300 out of 650) (Turyomurugyendo in prep.).

Parish leaders were requested to stratify villages. Criteria that leaders used included distance to BINP, distance to a lake, and differences in soil characteristics or temperature. Four villages were selected at random within the various strata. Based on a list provided by the village leader (or based on a list of vaccinations in one instance), two male-headed and two female-headed households were selected at random within each village. Tables 3.8 and 3.9 provide an overview of socio-economic characteristics that were collected during the survey. The wealth ranking (1? 5 = poorest? richest) was calculated as the average ranking on three criteria (type of house, livestock numbers and furniture) (Guinand 1996). Farm sizes were calculated from the estimated length and width of each plot by counting steps.

Household and farm characteristics were not used in multiple regressions, since ethnobotanical information was not available when the article that synthesised results from the various survey areas (Chapter 14) was written. The only variables with moderate variance inflation factors (R2 > 0.60) were those that reflected the gender of the household head. As most farms were male-headed, de jure female household heads formed a strongly correlated category with the male-headed category (R2 = 0.82).

3.4. Central Kenya (Meru)

A survey was conducted that collected information on tree species diversity in central Meru district (one of the 13 districts in the Kenyan Eastern Province), adjacent to Mount Kenya. This study area belongs to the East and Central African Bimodal Highlands (see section 3.1.1). Meru people predominantly practise mixed farming, i.e. crop cultivation and animal husbandry. Maize (Zea mays), beans (Phaseolus vulgaris), potatoes (Solanum tuberosum), sorghum (Sorghum bicolor), pigeon peas (Vicia faba), green grams (Vigna radiata), cassava (Manihot esculenta), yams (Dioscorea spp.), arrowroots (Maranta arundinacea) and millet (Pennisetum glaucum) are used as staple crops. Cash crops include coffee (Coffea arabica), tea (Camellia sinensis), tobacco (Nicotiana tabacum), cotton (Gossypium spp.), miraa (Catha edulis), and macadamia nuts (Macadamia integrifolia). There are six gazetted forests in Central Meru district covering a total area of 86,955 ha, including Mount Kenya (Lengkeek et al. in prep.).

SURVEY AREAS

40

Mount Kenya (5199 m) is the second highest mountain in Africa after Kilimanjaro and is a vital water catchment on which seven million people depend. The forest zone hosts important populations of several threatened animal species. Mount Kenya became an International Biosphere Reserve in 1978, while Mount Kenya National Park / National Forest was inscribed on the IUCN World Heritage List in 1997. The Natural Forest (70,520 ha) is located between 1600 and 3100 m. There are 81 endemic higher plant species to Mt. Kenya, which are possibly relic forms of plants that were more widely distributed during the Pleistocene period. Vegetation varies with altitude and rainfall, with a rich alpine and sub-alpine flora. The lower limit of the indigenous forest on Mt. Kenya now mainly stands at 2000 - 2500 m. Upland Gallery Forest (1700 - 2300 m) contains a more humid rainforest on the east side in which truly giant trees, many with massive buttressed trunks, are common. The Bamboo zone (2300 - 2600 m) includes the most extensive stands of dense bamboo in East Africa, forming a distinct crescent around the mountain. The Hagenia-Hypericum Parkland (2600 - 3500 m), Giant Heather (2900 - 3200 m), Afro-alpine (3200 – 4000 m), and Nival (> 4000 m) zones form the other altitude-dependent vegetation zones (Lengkeek et al. in prep.).

The survey followed the framework of participatory on-farm species screening trials that were implemented earlier. For these trials, farmers were selected from ‘catchment groups’ that implemented a common land management plan under co-ordination of the Ministry of Agriculture (MoA). For the trials, three groups were selected within relatively similar agro-ecological zones (Upper Midlands 2 and 3), and based on different location towards the forest (0, 12 and 25 km). Some information on conditions of the catchment groups is presented in table 3.10 (Lengkeek & Carsan in prep.).

Farmers were selected that were willing to participate in tree planting trials, and that had at least one goat, sheep, or pig because the experiments included fodder species. The chair of the catchment group and MoA staff were requested to select four poor, four intermediate and four rich farmers in a randomly-stratified manner. They used wealth criteria of farm size, type of house, and number of animals. The request specified to select within each wealth group two male-headed and two female-headed farmers, but the exercise resulted in some wealth groups that included only one female-headed farmer (Lengkeek & Carsan in prep.).

The species diversity studies were conducted in 2001 on 35 farms that had joined the species trials. Of only 32 farms, a complete set of socio-economic criteria that was used in regression analysis was available (tables 3.11 and 3.12). Education level was measured as 0 (no education), 1 (primary 1-4), 2 (primary 5-8), 3 (secondary) and 4 (adult education). Tables 3.9 and 3.10 show the quantitative and qualitative socio-economic characteristics that were used. The only variables with moderate variance inflation factors (R2 > 0.60) were farm size and wealth. The higher variance inflation was not caused by strong correlation of both factors, as ANOVA indicated important influences of the other variables as well in explaining the variance of farm size or wealth.

3.5. Central Uganda (Mabira)

The survey in Central Uganda collected information in 2001 from 105 farms that were arranged in five ‘axes’ that started from the Mabira Forest Reserve at angles of about 72°. The Mabira Forest Reserve is located in between the town of Kampala (45 km) and Jinja (20 km), at 1070 - 1340 m, and in the Ugandan Mukono District. The study area is located within the East and Central African Bimodal Highlands (see section 3.1.1). The reserve consists of a main block of the Mabira Main Forest Reserve (29,950 ha), the Namawanyi/Namananga Forest Reserve (256

CHAPTER 3

41

ha), and the Kalangala Falls. It is estimated that Mabira has 312 species of trees and shrubs, 287 of birds, 23 of small mammals, 218 of butterflies and 97 of moths (Boffa et al. in prep.).

On each axis, one village was selected within a distance of less than 1 km to the Mabira Forest Reserve, one village between 5-7 km to this forest, and one village at a distance between 12-19 km. Table 3.13 provides the names of the villages and their locations. Exceptionally, farms belonged to two villages for the position furthest from the forest on axis 4, therefore the survey was conducted in 16 villages. Within each village, a randomly-stratified sample was taken of a male-headed and female-headed household nested within three wealth categories. A seventh farm was selected within each village from which the head was known to be a forest user, somebody who regularly collected forest products (often for medicinal use).

Tables 3.14 and 3.15 show the quantitative and qualitative socio-economic characteristics used as explanatory variables in regression analysis. Because information was missing, only 97 households had complete information for the variables that were used in the statistical analyses. The only variables with R2 larger than 50% were the age of the household head or partner, and the years that the present head was head of the household. These larger R2 mainly resulted from positive correlation between the two variables (R2=0.44).

SURVEY AREAS

42

Table 3.1. Information on the strata of Kakamega and Vihiga districts, and information on the villages (arranged West-East or far-close to Kakamega Forest) where the interviews were held

Villages (average (UTM) co-ordinates)

Strata and information from Bradley et al. (1991, pp. 121-126) Households interviewed in first and second survey

‡

Ebuchiebe (0º 5.52’ (10.15) N, 34º 35.81’ (677.71) E)

Stratum A1 has 185 homesteads km 2, extremely small farms (0.53 ha), with most families having at least one adult male working away, the residual family members concentrating on food crops. Woody biomass amounts to 21.2% of groundcover, of which 59.9% present on-farm.

50 (48) 49 (47)

Madidi (0º 3.66’ (6.74) N, 34º 40.32’ (686.08) E)

Stratum E is characterised by average values (112 homesteads km 2, farm size of 0.87 ha, 20.6% of groundcover by woody biomass of which 45.6% on-farm).

50 (46) 0 (0)

Mutambi (0º 4.34’ (8.00) N, 34º 45.81’ (696.27) E)

Stratum A2 is similar to A1 in terms of farm sizes (0.49 ha) and population density (195 homesteads km 2). Woody biomass accounts to 23.8% of groundcover, of which 67.2% is present on-farm.

50 (44) 6 (4)

Shimutu (0º 11.28’ (20.79) N, 34º 48.89’ (701.98) E)

Stratum F2 has 74 homesteads km 2, average farm sizes of 1.12 ha, and residual forest in valley bottoms. On-farm (17.6%) woody biomass (23.8% groundcover) is significantly smaller than in the other areas.

51 (45) 23 (21)

‡ Figures between brackets indicate households where complete socio-economic data were available; summary statistics for these households are provided in tables 3.2 and 3.3

Table 3.2. Quantitative household characteristics (before ranging in the 0-1 interval, except for R2) in the western Kenyan surveys Survey Statistic Farm size

(acres) (1 acre =

0.4005 ha)

Number of crossbred

cattle

Number of local cattle

Number of cattle

Number of years

household head

Maximum age of head and partner

Number of resident children

Level of schooling

1 Average 1.4 0.16 1.6 - 21.0 53.5 4.2 5.0 (n=183) Minimum 0.25 0 0 - 0 19 0 0 Maximum 5.5 5 10 - 78 86 12 12 St. dev. 1.02 0.67 1.74 - 17.11 13.90 2.52 3.59 R2 0.44 0.36 0.27 - 0.27 0.45 0.12 0.39 2 Average 1.6 - - 1.6 19.6 53.5 4.4 4.5 (n=72) Minimum 0.25 - - 0 0 19 0 0 Maximum 5 - - 10 78 86 10 12 St. dev. 1.02 - - 2.04 16.80 14.38 2.46 3.09 R2 0.45 - - 0.45 0.29 0.62 0.28 0.44 (St. dev.: standard deviation, R2: variance explained by other variables, Level of schooling was coded as 0 for no education, 1 through 8 for Standard 1 to 8 and 9 through 12 for Form 1 to 4)

Table 3.3. Qualitative household characteristics of the western Kenyan surveys Village Type of household head Type of house

Values Number % R2 Values Number % R2 Values Number % R2 Survey 1 (n=183)

Ebuchiebe 48 26 0.42 Male 114 62 0.47 Permanent 20 11 0.35 Shimutu 45 25 0.54 Female de jure 28 15 0.43 Iron roofed 117 64 - Mutambi 44 24 0.37 Female de facto 41 23 - Thatch roofed 46 25 0.16 Madidi 46 25 -

Survey 2 (n=72) Ebuchiebe 47 65 0.83 Male 46 64 0.54 Permanent 10 14 0.49 Shimutu 21 29 0.84 Female de jure 10 14 0.44 Iron roofed 41 57 - Other 4 6 - Female de facto 16 22 - Thatch roofed 21 29 0.28 (R2: variance explained by other variables, only provided for the factors with dummy coding; female (de jure): widowed or divorced female household head; female (de facto): female head because husband lives away)

CHAPTER 3

43

Table 3.4. Information on the study villages in Cameroon and Nigeria (source: Degrande et al. in prep.) (n=112)

Village (Country)

Geographical Position

Inventory Farms Agro-ecology Population density

Average farm size

Chopfarm (Cameroon)

3°57' N 9°15' E

Only fruit trees

14 Low montane Forest Fairly low 2.1 ha

Elig-Nkouma (Cameroon)

4°06' N 11°24' E

All trees 20 Degraded humid forest Medium 2.6 ha

Nko’ovos (Cameroon)

2°55' N 11°21' E

All trees 19 Humid forest Low 6.0 ha

Makenene (Cameroon)

4°52' N 10°48' E

Only fruit trees

19 Transition between humid forest and savanna

High 6.2 ha

Ilile (Nigeria)

5°19' N 6°55' E

Only fruit trees

20 Degraded humid forest High 0.7 ha

Uguwaji (Nigeria)

6°25' N 7°32' E

Only fruit trees

20 Transition between forest and savanna

Medium 0.8 ha

Table 3.5. Quantitative household characteristics (before ranging in the 0-1 interval, except for R2) for the Cameroonian villages with complete inventories (n=39) Statistic Farm size (ha) Wealth ranking Age of head Size of household Education level Average 4.3 3.1 46.9 7.2 1.5 Minimum 0.11 1 22 1 0 Maximum 15.02 5 80 25 3 St. dev. 3.74 1.39 14.12 5.19 0.64 R2 0.39 0.22 0.29 0.34 0.41

Table 3.6. Qualitative household characteristics for the Cameroonian villages with complete inventories (n=39)

Characteristic Category coded ‘1’ Farms % R2 Category coded ‘0’ Farms % Further from forest Elig-Nkouma 20 51 0.42 Nko'ovos 19 49 Type of household head Male 31 79 0.63 Female 8 21 Indigenous in village Yes 32 82 0.65 No 7 18 Use of hired labour Yes 28 72 0.18 No 11 28

SURVEY AREAS

44

Table 3.7. Information on the study villages in south-western Uganda (source: Turyomurugyendo in prep.) (n=80)

Parish Villages Position Average Latitude

Average Longitude

Number of farms

Kashasha Ihunga, Katojo, Kitahurira, Ndeego Close to forest 1° 8‘ S 29° 48’ E 16 Kitojo Bitanwa, Kyogo, Nkukuru, Nyakaranga Close to forest 1° 1’ S 29° 47’ E 16 Kagarama Hamurambi, Ishanga, Kaashekye, Murambo Far from forest 1° 15’ S 29° 55’ E 16 Nangara Butoboore, Ikamiro, Kihorongwa, Rugoma Far from forest 1° 11’ S 29° 58’ E 16 Butoboore Kyantobi, Nyakabungo, Nyakihanga, Rukinda Intermediate from forest 1° 12’ S 29° 55’ E 16

Table 3.8. Quantitative household characteristics (before ranging in the 0-1 interval, except for R2) for the south-western Ugandan survey (n=80) Statistic Farm size (ha) Wealth ranking Maximum age

of head and partner

Number of children

Number of plots

Years head Education level

Average 5.2 2.6 46.1 3.7 11.6 14.4 3.2 Minimum 0.27 1.17 22 0 2 1 0 Maximum 45.87 4.50 79 11 35 60 12 St. dev. 7.75 0.67 16.14 2.41 6.42 12.27 3.73 R2 0.40 0.40 0.64 0.28 0.56 0.55 0.43

Table 3.9. Qualitative household characteristics for the south-western Ugandan survey (n=80) Parishes Type of household head

Values Number % R2 Values Number % R2 Kashasha 16 20 0.60 Male 37 46 0.85 Kitojo 16 20 0.53 Female de jure 39 49 0.86 Kagarama 16 20 0.50 Female de facto 4 5 - Nangara 16 20 0.43 Butoboore 16 20 -

CHAPTER 3

45

Table 3.10. Description of farmer groups for land management for the central Kenyan survey (n=40) (source: Lengkeek & Carson in prep.)

Descriptor Catchment group Name Igoji Kigane Nchoroiboro Location Igoji Nkubu Ruiri Agro-ecological zone Upper Midlands 2

(Coffee Zone) Upper Midlands 2

(Coffee Zone) Upper Midlands 3

(Marginal Coffee Zone) Mean annual rainfall (mm) 500 – 2200 500 - 2200 500 - 1800 Total Population in 1999 45,066 54,924 40,604 Average farm size of trials (acres) 5.4 3.1 6.0 Soils Well drained, very

deep loam to clay Well drained, extremely

deep loam clay Well drained, moderately deep dark reddish brown firm cracking clay with humic topsoil

Distance to the forest 25 km 12 km 0 km Altitude (m) 1353 - 1586 1497 – 1674 1524 – 1761 Longitude 37° 40’ E 37° 39’ E 37° 38’ E Latitude 0° 11’ S 0° 04’ S 0° 09’ N Number of farms * 12 (10, 10) 12 (11, 11) 16 (14, 11)

* Figures between brackets indicate farms where biodiversity survey was conducted, and the number of farms of the biodiversity survey where complete socio-economic data were available; summary statistics for the latter category of households are provided in tables 3.11 and 3.12

Table 3.11. Quantitative household characteristics (before ranging in the 0-1 interval, except for R2) for the central Kenyan survey (n=32) Statistic Farm size

(ha) Distance from

forest (km) Age of trial participant

Number of children

Wealth Education level

Average 1.6 11.9 48.9 4.6 1.7 2.3 Minimum 0.22 0 30 0 1 0 Maximum 6.92 25 65 10 3 4 St. dev. 1.35 10.28 10.10 2.15 0.74 1.03 R2 0.65 0.30 0.43 0.13 0.63 0.39

Table 3.12. Qualitative household characteristics for the central Kenyan survey (n=32) Characteristic Category coded ‘1’ Farms % R2 Category coded ‘0’ Farms % Type of household head Male 20 63 0.21 Female 12 38 Income generation from the farm Yes 17 53 0.36 No 15 47 Off-farm employment Yes 8 25 0.22 No 24 75

SURVEY AREAS

46

Table 3.13. Information on the study villages in central Uganda (source: Boffa et al. in prep.) (n=105) Village name Axis Distance from

Mabira Forest Average Latitude Average Longitude Number of farms ‡

Kiwaala 1 < 1 km 0° 28‘ N 32° 54’ E 7 (6) Kayanja 1 5-7 km 0° 23’ N 32° 52’ E 7 (7) Mpoma 1 12-19 km 0° 26’ N 32° 44’ E 7 (6) Kitayunja 2 < 1 km 0° 34’ N 32° 56’ E 7 (7) Ntonto 2 5-7 km 0° 38’ N 32° 21’ E 7 (7) Mindi 2 12-19 km 0° 40’ N 32° 51’ E 7 (7) Kkungu 3 < 1 km 0° 31’ N 33° 02’ E 7 (7) Bukasa 3 5-7 km 0° 34’ N 33° 04’ E 7 (7) Kiteredde 3 12-19 km 0° 44’ N 33° 02’ E 7 (5) Kyambogo 4 < 1 km 0° 30’ N 33° 01’ E 7 (7) Kiira 4 5-7 km 0° 28’ N 33° 09’ E 7 (7) Kalega 4 12-19 km 0° 21’ N 33° 13’ E 4 (4) Namwendwa 4 12-19 km 0° 30’ N 33° 04’ E 3 (3) Kitoola 5 < 1 km 0° 22’ N 32° 59’ E 7 (7) Buwuma 5 5-7 km 0° 21’ N 33° 05’ E 7 (4) Mutwe 5 12-19 km 0° 16’ N 32° 59’ E 7 (6) ‡ Figures between brackets indicate households where complete socio-economic data were available; summary statistics for these households are provided in tables 3.14 and 3.15

Table 3.14. Quantitative household characteristics (before ranging in the 0-1 interval, except for R2) for the central Ugandan survey (n=97) Statistic Wealth ranking Maximum age of head

and partner Number of children Years head

Average 2.0 45.6 4.7 20.2 Minimum 1 19 0 1 Maximum 3 76 20 55 St. dev. 0.80 14.23 3.57 13.44 R2 0.06 0.52 0.15 0.53

Table 3.15. Qualitative household characteristics for the central Ugandan survey (n=97) Distance to forest Type of household head Forest user

Values Number % R2 Values Number % R2 Values Number % R2 Close 34 35 0.28 Male 54 56 0.32 Yes 15 15 0.11 Far 31 32 0.29 Female de jure 13 13 0.22 No 82 85 - Intermediate 32 33 - Female de facto 30 31 -

a

c

b

Figure 3.1. Location of the study sites. Average locations of villages are shown, which were based on the coordinates of individual farms and/or plots. a: locations within Africa; b: locations in

Cameroon (4 villages) and Nigeria (2 villages); c: locations in Uganda (west) and Kenya (east).

a

e

b dc

Figure 3.2. Location of the study sites. Average locations of villages are shown, which were based on the coordinates of individual farms and/or plots. a: locations in Uganda (west) and Kenya (east) (=Figure 3.1c); b: locations in south-western Uganda; c: locations in central Uganda; d: locations in western Kenya; e: locations in central Kenya; b-e: relative position of the edge (centre for c) of the

nearest forest to the villages where the surveys were conducted

Lake TurkanaMount ElgonRift ValleyAberdaresMount KenyaMount Kilimanjaro

Lake Victoria

KENYA

TANZANIA

SOMALIA

ETHIOPIA

UGANDA

SUDAN

DRC

RWANDA

BURUNDI

AT0101Albertine Riftmontane forests

AT0108East Africanmontane forests

AT0721Victoria Basinforest-savannamosaic

Figure 3.3. Location of the eastern African study sites (circles) on a map of the terrestrial ecosystems of the world (URL http://www.nationalgeographic.com/wildworld/terrestrial.html). The ecosystems to which the study locations belonged are indicated in the legend, descriptions of the ecosystems can be found at the website.

SURVEY AREAS

50

Figure 3.4. Satellite image of the East African area portrayed in figure 3.3. Mountains, Rift Valley and lakes indicated in figure 3.3 can clearly be observed. (URL: http://www.nationalgeographic.com/mapmachine).

ButobooreKagaramaKashashaKitojoNangara

8.0*105 8.1*105 8.2*105 8.3*105 8.4*1059.85*106

9.86*106

9.87*106

9.88*106

9.89*106

Figure 3.5. Distribution of plots of the surveyed farms in south-western Uganda. Coordinates are UTM (WGS 1984). Symbols

refer to the parish to which the plot with the homestead area belongs.

Page not printed

CHAPTER 4 METHODOLOGY

R KINDT

METHODOLOGY

52

This chapter summarises the data analysis methods that were used to arrive at the results reported in Chapters 5-14. The specific methods used in each chapter were indicated there for fast cross-referencing.

4.1. Formulation of Use-group and Niche Matrices

Analogous to the concept of ecological functional groups (species that play the same role in maintaining and regulating ecosystem processes, Gitay et al. 1996), use-groups were defined as groups of species providing similar products or services to the farm household. Studying use-groups is similar to studying functional groups – redundancy within use-groups can be studied to estimate how well groups are buffered against species losses. Groups with low redundancy can be identified.

Use-group delineation and composition followed the information provided by each household, since uses were listed for all species occurring on each farm. Free responses were obtained on tree uses that were postcoded during data entry and checking. Species could contribute to several use-groups. Farms × species matrices were formed for the main use-groups, expressed as the percentage of farms on which they occurred – other use-groups occurred on fewer farms. Table 14.1 (Chapter 14) lists the main use-groups and farm percentages.

Several farms × species matrices were formed following several rules for inserting abundance > 0 in a specific matrix cell. These matrices formed the basis for several subsequent analyses (Chapter 1, figure 1.1 & 1.2). The consequences of using various rules to formulate these matrices were one of the main areas researched in Chapter 5.

Use-groups (i.e. matrices) defined by species occurrence and use as recorded at individual farms are referred to as Urec. Abundance > 0 was recorded for a cell in case the specific farmer (listed in rows) had communicated to use the particular species (listed in columns) for the particular use (product or service). Urec was used in Chapters 13 & 14, except for the western Kenyan survey that used Uadj (see below).

Because it is possible that not all actual uses of a particular species were recorded for each farm (e.g. missing data – see Chapter 5), information was adjusted by always including species in a group if more than half of and more than five households where the species occurred mentioned the particular use. Use-group definition based on this adjustment is referred to as Uadj. Uadj was used in Chapters 6-9 & 11.

The intensity of data collection makes duplication of the survey in other locations difficult. Data collection in other locations could be shortened by only recording the main use of the species at each farm, or by attributing all potential uses of a species to each farm. These use-group definitions are referred to Umain and Uall, respectively. The effects of these simplifications in data collection were investigated.

Niche matrices were formed in an analogous way as Urec. These matrices only included abundance > 0 for a cell in case the particular species (listed in columns) occurred in a specific niche on a particular farm (listed in rows). Table 14.1 (Chapter 14) lists the main niches that were encountered in the surveys – other niches occurred on a lower percentage of farms than specified in this table.

CHAPTER 4

53

4.2. Measurement of Diversity

4.2.1. Alpha and Gamma Diversity

Species richness (S) refers to the number of species that were encountered on a specific farm, in a specific village, or in a specific survey.

Alpha diversity was analysed by taking the average number of species per farm (Savg = S1 of section 4.2.2) across all farms ( i.e. n=201 for western Kenya). 95% confidence interval limits were calculated by (Hayek & Buzas 1997 eq. 4.4) as

201

ˆ96.1ˆ

σµ ± .

Use-group gamma diversity was analysed by the total number of species in the survey (Stot). For Stot, confidence interval limits could not be calculated because only one value was obtained.

4.2.2. Species Accumulation Curves

Species accumulation curves show the trend in which additional species are encountered when a larger area is sampled. The

zcAS =

model proposed by Arrhenius (1921) is often used to describe the relationship between area (A) and species richness (S). These curves can be calculated by two main approaches (Condit et al. 1996; Chazdon et al. 1999). One approach involves calculating the average number of species from a nested series of expanding quadrats. The other approach calculates the average number of species by randomly adding new sites. (Yet another approach involves random accumulation of individuals rather than sites – this approach is briefly outlined in the discussion of Chapter 14.)

We compared species accumulation patterns by randomly sampling farms (‘random richness’), with patterns resulting from randomly sampling farms within the same village, before sampling randomly from other villages (‘proximity-based richness’). The second village sampled belonged to the same half of the survey area.

Chances of sampling individual species were calculated by a new approach based on the hypergeometric distribution. Where species i occurs on fi of Ftot sites, the expected average species richness after N random site additions for Stot species equals

SN = ∑=

totS

i 1

(1 - ∏= +

+N

a 1 tot

itot

1a-F1 a -f-F

) = ∑=

totS

i 1

(1 -

−NF

NfF totitot ).

This equation can be deducted by estimating the chance of not sampling species i in fi randomly selected sites. The last part of the equation corresponds to the rarefaction method that was developed by Sanders (1968) and corrected by Hurlbert (1971), applied to a dataset documenting presence (1) or absence (0) of each species on each site. A program (ExactS) can be obtained from the authors that calculates SN from a sites × species matrix over the entire range of accumulated sites (Kindt 2001a).

The accuracy of this new method, and especially the gains in calculation time have been detailed in table 4.1 and figure 4.2, calculating the average richness for 100 randomly accumulated farms for all species (i.e. table I.3). Simulations were conducted on a Micron Millennium computer with a x86 Family 6 Model 5 Stepping 2 Genuine Intel processor with 130,468 kb RAM. Speciesrichref Kindt (2001i, see below) was used to calculate the species richness for s randomizations. Table 4.1 and figure 4.2 show that the average species richness is quickly approximated, as are the 95%

METHODOLOGY

54

confidence interval and standard deviation. For s = 100,000, a smooth and symmetrical histogramme is obtained. For s = 10,000, the species richness that was most frequently encountered (i.e. the peak value of the histogramme) was closest to the average species richness.

It has to be remarked here that even 1,000,000 randomizations only reflect a small fraction of all possible combinations of 100 farms out of a total of 201. In total, approximately 1.8 E59 combinations can be obtained. It would take 9.8 E56 seconds or 3.1 E49 years to calculate all values at a speed of 0.005457 seconds per randomization (calculated as the average for 1,000,000 randomizations, table 4.1), which would enable to calculate the exact value. It is therefore more practical to calculate the exact average species richness through the new approach introduced here based on the hypergeometric distribution which only takes 0.17 seconds.

Comparisons with other software packages that calculate species accumulation curves based on randomizations showed that not as accurate results could be obtained for Biodiversity Professional (maximum 50 randomizations, McAleece et al. 1997) and PC-ORD (fixed at 500 randomizations, McCune & Mefford 1999) since the number of randomizations is restricted in these packages. Substantially longer calculation time is required for EstimateS (Colwell 1997), which calculates estimators for regional species richness as well, however. The standard deviation that is provided in PC-ORD and EstimateS allows for reliable calculation of the 95% confidence interval. These programs do not provide the histogramme of values obtained during randomizations. In addition, they do not provide maximum or minimum values. The maximum value can be an important value when planning to minimise areas needed for conservation (see Chapter 15). We can therefore recommend using ExactS to calculate the exact average species richness of subsamples, and Speciesrichref if confidence intervals and ranges of values are also of interest.

Proximity-based richness was calculated by multiplying SN within the village with the probability of the village sequence. Proximity-based SN after N (=50) random site additions for example equals

∑=

4

1i

(vj Ftot SNj)

with vj the number of farms within village j and SNj the expected average species richness in village j.

Parameter values zN of the non-linear model SN= NzcN were calculated for 2 = N = 201. By equalling c to the average richness of a farm (the expected value for S1), the exponent zN can be calculated for each number of accumulated farms as:

zN = ln(SN c-1) (ln(N))-1.

Assuming that AN = N A1, above model is related to the S=cAz model proposed by Arrhenius (1921) which describes the relationship between area and S. Although we encountered substantial variation in farm sizes (coefficient of variation=0.72), our approach calculated SA for accumulations of the average farm size (Chapter 6). Parameter c can be defined as alpha diversity as it describes the average site diversity, and parameter z as beta diversity as it expresses the differences in species composition among sites (Chapter 6).

4.2.3. Randomization Tests on the Influence of Sample Size on Species Richness

Using a Monte-Carlo approach, random and ‘randomly-clustered’ richness was calculated for specific sample sizes (table 6.2 in Chapter 6; table 14.2 in Chapter 14) based on 100,000 random farm sequences. One of the sample sizes corresponded to the size of the smallest village of the respective survey.

CHAPTER 4

55

The average random richness was calculated by randomization to check the values obtained through the hypergeometric distribution. The Speciesrichref program developed to calculate random richness (Kindt 2001i) provided besides the average richness of s (e.g. 100,000) random site sequences also the range of values (minimum - maximum), and 95 % confidence interval limits. Confidence intervals were determined by selecting the thresholds with counts equal (or just below) 2.5% permutations and equal (are just above) 97.5% of permutations. Unlike average random richness, which only depends on the frequency of each species, these latter statistics also depend on the aggregation of species within sites, and can therefore only be calculated through a randomization procedure (see Chapter 6). Figure 4.2 and table 4.1 (discussed in the section 4.2.2) shows the influence of the number of randomizations on the results.

Randomly-clustered richness was calculated by grouping farms first in four random subsamples corresponding to the size of the villages sampled, before calculating the average, ranges and 95 % confidence interval limits of the average richness of the four subsamples (Kindt 2001h, Speciesrichvillage). This approach could be conceptualised as creating ‘artificial villages’ by randomly assigning farms to these villages. Such approach is a null model analysis as specified in Gotelli (2001), an investigation that compares a statistic (here: species richness) calculated for the real data set with the same statistic of null communities generated by a set of rules for randomisation from the real data set. For our null communities we kept the sizes of the four subsamples equal to the sizes of the actual villages, and thus eliminated sample-scale effects on diversity. The ranges of randomly-clustered richness were compared to proximity-based SN, as the latter value corresponds to the actual distribution of farms and associated richness in villages.

As for random richness, the influence of the number of randomizations on the results was investigated. The results of this analysis presented in figure 4.3 and table 4.2 show that, as for random richness, the average species richness is quickly approximated, as is the 95% confidence interval. For s = 100,000, a smooth and symmetrical histogramme is obtained. For s = 10,000, the species richness that was most frequently encountered (i.e. the peak value of the histogramme) was closest to the average species richness. The results show that the average richness of four clusters of 50 farms closely approximates the exact average richness of 50 randomly accumulated farms (see Chapter 6). A smaller range of values is obtained for randomly-clustered richness compared to random richness (figures 4.2 & 4.3, Chapter 6).

4.2.4. Diversity Profiles

Although many equate diversity to species richness (the number of species encountered), diversity is a function of the number of species, and the evenness in distribution of species’ abundances (Magurran 1988 pp. 7-8; Purvis & Hector 2000). We developed a new method that discriminates between both facets of diversity.

The Rényi series provides diversity profile values (Ha) based on a scale parameter value (“scale”) a (= 0 and a ? 1) (Tóthmérész 1995 eq. 1; Legendre & Legendre 1998 eq. 6.31; Rennols & Laumonier 2000 eq. 1):

( )α

α

α −= ∑

1log ip

H ,

where pi = proportion of item i (abundance/total). In section 4.7, we demonstrated how Rényi profiles were calculated for four theoretical ecosystems.

We used base e to calculate all logarithms. It can be demonstrated that values of the Rényi profile at the respective scales of 0, ˜ 1, 2 and 8 are related to species richness S

H0 = ln(S),

METHODOLOGY

56

Shannon H (Magurran 1988 eq. 2.17; Condit et al. 1996 eq. 3; Legendre & Legendre 1998 eq. 6.1)

∑−== ii ppHH log1

Simpson D-1 (Magurran 1988 eq. 2.27; Legendre & Legendre 1998 eq. 6.41)

∑ −− == ))ln()ln( 1212 ipDH ,

and Berger-Parker d-1 (Magurran 1988 eq. 2.31)

)ln()ln( 1max

1 −−∞ == pdH

diversity indices.

System a is more diverse than system b if all values of the diversity profile corresponding to system a are larger. Systems that have intersecting profiles consist of one system that is richer but not more evenly distributed. We separated the contributions of evenness on profile values by investigating the difference

Ha-H0 = ln(Ea) = ln(eHa/S).

We defined system a as more evenly distributed than system b in case all values of ln(Ea) are larger, and referred to a plot of ln(Ea) against a as an evenness profile (see discussion of Chapter 6). Ea = 1 corresponds to systems that are perfectly evenly distributed – these systems will have perfectly horizontal diversity profiles. Since H8 is only determined by the proportion of the dominant (=most abundant) species, E8 provides an insight in the contribution of the dominant species to evenness. Therefore, systems with larger E8 have a more evenly distributed dominant species. Systems with intersecting evenness profiles consist of one system where the dominant species is more evenly distributed but the other species less evenly.

The values of the series for the various species groups were calculated for scales a ∈{0, 0.1, 0.25, 0.5, 0.75, 1.25, 1.5, 1.75, 2, 2.5, 3, 10, 100, 200}. Species proportions were calculated for the complete sample (201 farms). Values were calculated for H1 and H8 (since we expanded the Rényi series to include these values) by Shannon and Berger-Parker diversity indices directly.

After expanding the Rényi series to include the Shannon index, the ratio

ln(E1) / ln(E8 )

provides an indication of the evenness in distribution of the other species (excluding the dominant species).

H8 was the only value for which confidence intervals for the expected values for the survey area could be calculated as it is obtained solely from the proportion of the dominant species (i.e. the Berger-Parker diversity index), while other values in the diversity series include an effect of species richness. The calculation of the confidence interval for this statistic is outlined in section 4.2.6.

4.2.5. Accumulation Patterns of Diversity Profiles

Similarly to the randomization approach to calculating random richness (section 4.2.3), we used a Monte-Carlo approach of random site additions to obtain accumulation surfaces for diversity profiles by calculating the average Ha for each accumulated number of sites for the scales we differentiated for the complete sample (section 4.2.4). This was a new approach to analyzing the effects of sample size on diversity.

CHAPTER 4

57

We calculated average values and ranges (minimum and maximum) for the entire range of accumulated farms (1? n) based on 10000 random site sequences. The FORTRAN programme developed to carry out the computations (Kindt 2001d) can be obtained from the author.

Average values, ranges (minimum and maximum) and 95% confidence intervals were calculated for diversity and evenness profiles for 100,000 random farm sequences (Kindt 2001e, 2001f, 2001g). Since different patterns of profile accumulation will result from differences in the distribution of species abundances over farms, comparisons of diversity and evenness of several locations are only meaningful if they are based on the same sample size (see Chapter 6). As indicated in table 13.1, we had three options how equal sample sizes could be determined: (a) based on the number of farms sampled (Fref), (b) based on the area sampled (Aref), and (c) based on the number of trees sampled (Nref). In Chapter 13, we calculated profiles for each country and for each village based on the country or village with the smallest sample size on Fref, Aref and Nref (table 13.1). For example, for comparisons among countries, Fref = 39 as fewest farms were sampled in Cameroon, and Aref = 115.72 ha as the total area surveyed was smallest in Kenya. In Chapters 6 & 7, we only considered sample sizes determined by the number of farms (Fref). Chapter 13 was the only chapter where also Aref and Nref were considered.

Because the average farm size and average farm abundance varied among countries and among villages, calculations based on Fref, Aref and Nref were expected to yield different results. For example, villages with greater tree density will reach the Nref at fewer farms. It is easy to calculate average values for Fref as the exact reference can be reached within each random sequence. It is unlikely, however, to reach the exact Aref and Nref by randomly adding farms from surveys that did not standardise these sample sizes. Therefore, we calculated profiles for Aref and Nref by selecting the accumulated number of farms within each random sequence that was closest to the reference value. In some cases, the first randomly selected farm may contain more than twice the reference value, in which case zero is the closest number of farms to the reference value. In such cases, the diversity profile can not be calculated. It is also impossible to calculate diversity profiles where the accumulated number of farms contains no trees, although such cases did not occur in our datasets. We calculated average diversity profiles based on the number of valid sequences (determined by S>0).

We counted the number of farm sequences that yielded profile values equal or above thresholds ranging from 0 to ln(S) (the maximum value possible at each scale) in steps of 0.01. We similarly counted the number of farm sequences that yielded evenness values above thresholds ranging from -ln(S) to 0. These data allowed determination of 95% confidence intervals by selecting the thresholds with counts equal (or just below) 2500 and equal (are just above) 97,500. The maximum and minimum profile values obtained were also recorded.

Figure 4.4 and table 4.3 show that accurate values for average profile values are quickly obtained for subsamples of 100 farms (Fref). For randomizations = 10,000, smooth histogrammes are obtained. The 0.01 interval that was most frequently observed ( i.e. the peak of the histogrammes) contained the average profile value. An increasing scale of the diversity profile corresponds to a wider histogramme (see Chapter 7 & 13). The width of the histogramme increased with the number of randomizations – in case the maximum values would be of particular interested, then a large number of randomizations will be needed (see section 4.2.2 and Chapter 15). The analysis presented in figure 4.4 indicates that the 100,000 randomizations used throughout this thesis produced reliable results for average profile values and 95% confidence intervals.

4.2.6. Diversity and Evenness Statistics

Although a complete characterisation of the diversity and evenness of a system involves calculation of diversity and evenness profiles (section 4.2.4, Chapter 7), we used specific diversity

METHODOLOGY

58

and evenness statistics in some chapters to simplify calculations or comparisons. The Shannon diversity index H, Simpson diversity index D-1 and inverse Berger-Parker index d-1, which are all values at specific scales of the Rényi series Ha (section 4.2.4) were calculated directly from information on species’ frequencies.

In Chapter 8, logarithmic species richness, Shannon, and logarithmic Simpson and inverse Berger-Parker indices were calculated for each farm and use-group. As ln(0) can not be determined, the range of values obtained only span farms where the use-group occurred. The statistic ln(S+1) included all farms.

An alternative inverse Berger-Parker index (d-1csa) was calculated based on cross-sectional areas of

the dominant species and use-group, not on abundance as the other diversity indices and diversity profiles.

95% confidence interval limits for the frequency of the dominant species (cluster sampling) were calculated by (Hayek & Buzas 1997 eq. 8.28)

121

22

)1(

)(96.1 −

=

−

−±

∑nmn

ppmp

n

iii

,

mi = abundance of farm i; pi = species frequency at farm i; p = species frequency in the survey; n = number of farms; m = total abundance.

These were used to calculate confidence limits for d-1 and d-1csa. The reciprocal values of the

confidence interval limits were reported as confidence interval limits for d-1, natural logarithms of these reciprocal values as confidence interval limits for H8 .

The inverse Berger-Parker index was also calculated for the dominant species from the broken stick distribution with the same species richness as each use-group (d-1

bs). Broken stick distribution corresponds to random distribution of individuals over species (Hayek & Buzas 1997; Legendre & Legendre 1998). The formula used to calculate the expected frequency of the dominant species for a group with Stot species was (Legendre & Legendre 1998 eq. 6.49)

∑=

−−Stot

xtot xS

1

11 .

The broken stick value is less strict than the inverse Berger-Parker index based on completely even distribution. This value equals Stot because completely evenly distributed species each have a frequency of Stot

-1.

As the Berger-Parker is only determined by the frequency of the dominant species, it expresses evenness of a system. Evenness of species distribution was also analyzed by J (Magurran 1988 eq. 2.22; Legendre & Legendre 1998 eq. 6.44)

J = )ln( S

H

and E (Legendre & Legendre 1998 eq. 6.50 and 6.49)

E = ∑− )(ln)( ii yEyE

H

with ∑=

−−=S

ixi xSyE 11)(

CHAPTER 4

59

indices based on the Shannon diversity for equal distribution and for broken-stick distributions for the same species richness, respectively. These indices were only calculated for households where species richness was higher than one to avoid division by zero.

4.3. Use-group Abundance and Density

Use-group species abundance (Navg) was calculated by the average of individual species abundance on farms. Use-group density (Davg) was calculated by dividing species abundance on farm by farm size. Species density for use-groups at the survey scale (SDtot) was calculated by Davg/Stot. Species density was calculated in a similar matter after removing the dominant species from each use-group (SDtot-d).

In Chapter 8, in analogy to species richness (section 4.2.6), ln (N) and ln (N+1) indicate calculations of species richness and tree abundance for households where the use-group was present, and all households including those with zero abundance, respectively.

4.4. Regression, Correlation and Non-Parametric Analysis

4.4.1. Correlations

Correlation parameters were calculated by the S-Plus 2000 (Professional Release 1) package. Following Kolmogorov-Smirnov goodness of fit tests (S-Plus) for normality, non-parametric Spearman ? and Kendall t rank correlation and significance tests were calculated in case the distributional assumptions for Pearson’s correlation r were not met.

In Chapter 8, correlations were calculated between several diversity statistics (section 4.2.6) for the same farm and same use-group, and between the Shannon index of the same farm and different use-groups. High correlation values for the first aspect would indicate that the choice of diversity index did not influence diversity rankings very much ( i.e. that the ranking at one specific scale of the Rényi series was the same as at another scale of the Rényi series). High correlation values for the second aspect would indicate the possibility of classifying farms as more diverse for a number of groups based on diversity information for only one group.

4.4.2. Robust MM Regression

Regression analysis for most figures used the Robust MM Regression technique provided by S-Plus. The robust fit is minimally influenced by outliers in the dependent and independent variables, and can be used when random variation in the data is not Gaussian. In one case (figure 5.3), results of robust regression based on the natural logarithm (y~a+bln(x)) were presented because substantially more variation was explained by that approach.

4.4.3. Multiple, Stepwise and Partial Regression Analysis

In Chapter 8, the influence of farm characteristics on farm diversity statistics (H, ln (S), J and ln (N+1), section 4.2.6) as response variables were studied using multiple linear regressions (S-Plus). Qualitative factors were coded as dummy explanatory variables. Quantitative variables were ranged within the [0,1] interval by subtracting the maximum value and subsequently dividing by the maximum value (see Chapter 3). Regression coefficients can then be compared more easily, whereas the significance of regressions is not affected. S-Plus was used to select explanatory

METHODOLOGY

60

variables using stepwise linear regression (S-Plus uses the Akaike Information Criterion to guide backward and forward selection).

Where residuals of calculated models were not normally distributed (Kolmogorov-Smirnov goodness-of-fit test in S-Plus), the significance of coefficients was calculated by 9999 permutations of the residuals of the full regression model by the randomisation programme of Legendre (1999a) based on Anderson & Legendre (1999). Before using this programme, attempts were made to normalise response variables using the Box-Cox normalisation procedure.

By coding for separate villages, and because multiple regression coefficients are partial regression coefficients, the influence of farm and household characteristics independent from their broader geographic location can be evaluated. The proportion of variation that is exclusively explained by spatial and non-spatial explanatory variables was calculated by partial linear regression as outlined by Legendre & Legendre (1998 pp. 528-532).

We used a similar multiple regression approach in Chapter 14 to test whether certain farm characteristics could explain the species richness or abundance in use-groups. Since all farms were included in the calculations, including those where particular uses contained no trees, response variables were transformed to ln(X+1). We constructed a general model for all use-groups by including dummy variables for the use category (table 14.3).

4.4.4. Tests for Differences among Farms and Use-Groups

In Chapter 8, multiple regression (section 4.4.3.) was also used to investigate whether significant differences existed among the average values for ln (S), H, J, E, ln (d-1), and ln (N) for different use-groups and farms (section 4.2.6) by coding the different use-groups as dummy variables. Where residuals of calculated models were not normally distributed, we calculated 99,999 permutations by Legendre (1999a).

The significance of differences between groups was also calculated in Chapter 8 by the non-parametric Kruskal-Wallis rank tests (S-Plus) and Multi-Response Permutation Procedures (PC-ORD 4 package, McCune & Mefford 1999). MRPP provide a chance-corrected within-group agreement statistic (A), which is a descriptor of within-group homogeneity. When A equals 1, then all items are identical within groups, while A equaling zero means that group heterogeneity equals expectations by chance.

4.4.5. More Detailed Analysis of the Influences of Farm Size on Diversity and Abundance

In Chapter 8, the influence of farm size on farm diversity (species richness S and Shannon diversity H) was studied here by fitting the models

baNS =

and baNH =

(N: farm size/smallest farm size)

in S-Plus. Since parameters a can be calculated directly as average diversity of farms of the smallest size, models with prefixed parameter a and models where this parameter was estimated were compared.

The influence of farm size was further evaluated by calculating H, S and N for random combinations of farms of the same size and subsequently comparing values for the same

CHAPTER 4

61

combined area size. Because of the large number of combinations possible to calculate diversity statistics of accumulated farms of the same size (there are for instance 1820 = 16! (12! 4!)-1 possibilities of combining 4 of the 16 farms of 0.25 acres), the programme EstimateS 5 (Colwell 1997) was used. This program calculates average H, S, and N for random additions of sites; results presented here were based on 5000 randomisations. For S, the exact average species richness for randomly accumulated sites was calculated as well (section 4.4).

4.5. Analysis of Species Composition

4.5.1. Ordination Method

Differences in species composition and influences of farm characteristics were investigated by canonical ordination. Ordinations were conducted for all species occurring on a farm and for each use-group separately. All 183 farms (out of the 201 surveyed) with complete information on environmental characteristics were included in the analysis, including those farms where a particular use-group was not represented.

Differences in species composition can be analyzed by plotting each farm by its species abundances on axes representing each species and subsequently investigating how close farms are to one another in the multidimensional space. Ordination methods seek a representation in a lower number of dimensions of this space.

Ordinations of farms were obtained through the transformed data approach described by Legendre & Gallagher (2001). The transformed data approach provides an alternative to the distance-based Redundancy Analysis (RDA) approach of Legendre & Anderson (1999). RDA is a canonical ordination method in which the ordination axes are combinations of explanatory variables included in a second matrix. The ordination method further follows the principal components analysis (PCA) methodology so that consecutively calculated axes represent lower amounts of variance.

RDA was based on transformed farm – species abundance matrices in which the Euclidean distance between sites equalled their Hellinger distance (Rao 1995), one of several ecological distance measures that have been developed to express difference in species composition (e.g. Legendre & Legendre pp. 274-288). The Hellinger distance between site (farm) 1 and 2 is calculated by

∑= ++

−

S

j

jj

nn

nn

1

2

2

2

1

1

(nxj abundance of species j at site x; S: species richness).

When transforming each abundance value by

+x

xi

nn ,

the Euclidean distance between site vectors of the transformed matrix will equal the Hellinger distance of the original data. Only species occurring on more than five farms were included in the calculations, and their abundance was transformed by ln (nxj+1) before the Hellinger distance transformation. Legendre & Legendre (1998) and Legendre & Gallagher (2001) described the Hellinger distance as a more meaningful distance measure than those used under classical RDA (Euclidean distance) or canonical correspondence analysis (Chi-square distance), especially when

METHODOLOGY

62

less weight is to be given to rare species. The transformed data method is also to be preferred over CCA if richer sites are not to be given higher weight. The program used for the Hellinger transformations was written by Legendre (1999b).

For RDA, both polynomial (PRDA) and linear (LRDA) forms were calculated (Makarenkov & Legendre 1999, 2002). The polynomial (non-linear) form allows for second-order relationships between ordination axes and environmental characteristics, whereas the linear (classical) form only includes first-order relationships. The program used for RDA (PRDA and LRDA) was Polynomial RdaCca (Makarenkov & Legendre 1998).

Only ordination figures obtained through PRDA have been included. Eigenvectors were scaled to length 1, so that in ordination diagrammes, distances among sites (farms) approximated the Hellinger distance. This is RDA scaling type 1 as outlined in Legendre & Legendre (1998 pp. 585-586).

Projection of a site at right angle on a species vector approximates its value (Legendre & Legendre 1998 pp. 585-586). Sites (farms) were represented by symbols with size corresponding to the number of farms with the same species composition. Sites where the group was not represented correspond to the origin before centring.

Explanatory variables were represented differently when they referred to categorical or continuous variables (tables 3.2 and 3.3). Categorical variables were represented by the centroids of the fitted scores of sites belonging to the respective category. Classes of nominal variables are better displayed by centroid points (ter Braak & Smilauer 1998 p. 167). An advantage of centroid representation is that centroids for all states (including the one state that is a linear combination of the other states) can be calculated easily. Centroids can be interpreted in a similar way as individual site scores in terms of species composition. The vector connecting two states indicates the expected change in species composition if a farm would change from one state to the other (ter Braak & Smilauer 1998 p. 170). Continuous variables were represented by the multiple correlation coefficients of environmental variables and their squared value with fitted site scores of each axis (Makarenkov & Legendre 1999; Makarenkov & Legendre submitted). Linear (Pearson) correlations between environmental characteristics and fitted site scores were also calculated (represented with suffix ‘1’ in the diagrammes). Subsequently, vector lengths were multiplied by five; this does not change their interpretation, but improves clarity of diagrammes (Legendre & Legendre 1998 p. 586). Angles between species and continuous variable vectors reflect their (multiple) correlations. Correlations were calculated between site scores and fitted site scores for the axes represented (species-environment correlation).

In ordination diagrammes, the circle of equilibrium contributions with radius n/2 (n: number of species) was represented. According to Legendre & Legendre (1998 p. 402), only descriptors (species) whose vector lengths exceed the circle (represented by an ellipsoid when the axes do not have the same scale) significantly contribute to the formation of two PCA axes. Although RDA calculates the PCA solution on fitted values of species variables after multiple regression and not on original species values, we only included vectors in the diagrammes whose length exceeded the equilibrium circle. For the same reason of enhancing clarity of diagrammes, only centroids at distance of 0.2 n/2 from the origin were included.

Variance explained by the first (two) axes of RDA was compared to the variance explained by the first (two) axes in PCA; this is an expression of the level of constraint introduced by canonical ordination. The correlation was calculated between site scores on the first RDA and PCA axis. For PCA, also the number of axes that should be considered following the broken stick distribution criterion (the number of axes with eigenvalues larger than the broken stick values) was provided (Legendre & Legendre 1998 p. 410).

CHAPTER 4

63

4.5.2. Partial Ordination

Independent contribution of sets of explanatory variables on the ordination was investigated by partial canonical analysis (Borcard et al. 1992; Legendre & Legendre 1998 pp. 605-612; ter Braak & Smilauer 1998 p. 124). This approach is similar to partial linear regression (sectin 4.4.3). The variance explained by only using village information (Vv), and the respective variance by only using household characteristics (Vh) were obtained. Some of the total variance (Vt, calculated when using both sets of variables) may have been explained jointly by both sets. The variance that was explained exclusively by village information can be obtained as Vve=Vt-Vh, and the respective variance for household characteristics as Vhe=Vt-Vv. Joint variance explained equals Vj=Vt-Vve-Vbe=Vt-(Vt-Vh)-(Vt-Vv)=Vv-Vve=Vh-Vhe. Vj will have a negative sign when Vt<Vve+Vbe (equivalent to Vv<Vve or Vh<Vhe).

4.5.3. Combining Ordination and Multiple Regression Analysis

Another method by which the influence and significance of environmental characteristics on differences in species abundance were investigated was multiple regression analysis (section 4.4.3). Only species abundances (ln(n+1) transformed) for those species whose vector lengths exceeded the equilibrium contribution circle in ordination diagrammes were investigated. Coefficients obtained through multiple regression including all variables were coded as ra, while regression coefficients obtained after stepwise regression were indicated as rs. Both regression coefficients are partial regression coefficients; where a variable is retained by stepwise regression, it also has a significant contribution to the variance that is explained. Regressions only describe patterns in variance of sites for a single species, whereas ordination highlights patterns in variance of sites with different species composition. Regressions, however, describe all variance, while two-dimensional ordination diagrammes only represent a fraction of all variance. Ordination and regression analyses results are thus complementary.

4.5.4. Comparisons with Results of Previous Surveys

Bradley et al. (1985) and Bradley (1991) provided information on differences in species composition between strata A1 and A2 (see Chapter 3). Interpretation of stacked columns (Bradley et al. 1985, Bradley 1991 p. 153) revealed the percentage of farms with particular tree species in three on-farm niches in the Ebusikhale (stratum A1, n=90) and the Kegoye (stratum A2, n=60) sublocations. The 95% confidence interval for the species frequencies was provided by (Hayek & Buzas 1997 eq. 8.21):

1)1(

01.2−−

±n

ppp

(p: species frequency; n: number of farms).

Lauriks et al. (1999) differentiated five hedge types based on clustering of relative species frequencies of 63 samples each taken near the centre of a 25 km2 cell of a grid covering Vihiga and Siaya (bordering on the west) districts. For the Shimutu village, which is situated in Kakamega district, the percentages for the adjacent cells South and Southeast of the Shimutu cell were mentioned. The 95% confidence intervals (50 samples) were calculated using the clustered sampling method (Hayek & Buzas 1997 eq. 8.28, see section 4.2.6).

METHODOLOGY

64

4.6. Metapopulations

4.6.1. Investigation of Clustering of Species and Density-Abundance Relationships

In the first set of analyses, we investigated the spatial structuring of populations of each species that was inventoried. Such analyses preceded the simulations of dynamics of genetic diversity, since discovery of non-random spatial patterns implies that models that ignore spatial structure would not simulate geneflow accurately. We differentiated between random distribution, aggregation (defined as a higher than average chance of finding a tree of the same species in close proximity of another tree) and hyperdistribution (a less than average chance of finding a conspecific individual). The situation of aggregation especially represents a case where metapopulation models (a metapopulation consists of many local populations connected by migration, Hanski 1999) simulate actual geneflow better, since individuals will have higher chances of receiving genes from the same subpopulation in neutral models of genetic diversity. We present selectively-neutral models where presence or absence of specific alleles does not influence the chances of survival of an individual or the chances of mating between two individuals. Alternative models could build in mechanisms of avoidance of assortative mating, and hence potentially smaller chances of geneflow within subpopulations.

In analogy to Condit et al. (2000) (see also Ostling et al. 2000), the distribution of species density was investigated by calculating the average density of conspecifics at increasing distance from each individual tree, standardised by subsequent division by average density (which allows comparison among species with various population densities). Average densities of conspecifics at short distance of 1, >1 and <1 indicate perfectly random distribution, aggregation and hyperdistribution of species, respectively.

Although co-ordinates of each farm were recorded by a Global Positioning System receiver, small farm sizes and the fact that only one recording was made within each farm prevented accurate estimation of between-farm distances (the largest recorded distance between farms within each village were 0.73, 0.98, 1.01 and 1.33 km). Co-ordinates of individual trees were not recorded. Therefore, calculations were based on the three categories of origins distinguished above, while limiting the “outside the village” origin to the other villages. Defining Nxy and Axy as species abundance and size of farm y in village x, then average density of conspecifics within the same farm, within the same village and outside the village can be calculated as respectively:

∑∑∑∑

v ivi

v ivi

A

N ,

∑∑∑∑∑∑

≠

≠

v i ijvj

v i ijvj

A

N, and

∑∑∑∑∑∑∑∑

≠

≠

v i vw jwj

v i vw jwj

A

N

(for all v and i where Nvi > 0).

These values are standardised by dividing them by:

∑∑∑∑

v ivi

v ivi

A

N

(for all v and i).

The Density-Abundance relationship describes that species with wider distributions (larger P, the fraction of occupied habitat islands) are locally more abundant (larger w, density) than species with more narrow distributions. Hanski & Gyllenberg (1997) used the nonlinear model:

waP

Plog)

1log( +=

−

CHAPTER 4

65

to describe this relationship. In analogy, we investigated the relationship between the fraction of occupied farms and species density.

4.6.2. Metapopulation Simulations

We investigated changes in neutral genetic variation (variation that is not influenced by selective pressures) for a diploid species by software (Kindt 2001c) based on random selections of alleles from the seasonal genepools of germplasm (seed, wildlings or cuttings, see Chapter 3) and pollen. The algorithm replaced each tree older than a specified age by a new individual. Trees were only replaced after they had contributed to the genepools (e.g. a tree was only cut and replaced after it had produced seed – we often encountered the farmer practice of harvesting seed from trees that just had been cut). Allelic identities of the replacing tree were based on information on origins and forms of germplasm (Chapter 3), and parameter values describing geneflow, determined for each tree separately (table 12.1). As potential origins for pollen and germplasm for every tree that was being replaced, we distinguished between the same tree, the same subpopulation, the other subpopulations, and outside sources. The actual tree from which alleles came was selected at random once the type of origin was determined (e.g. if it was established that seed originated from other subpopulations than the one where the new tree would be established, while 25 trees in the other subpopulations had produced seed, then each of these trees had a chance of 0.04 of being selected as actual mother tree). Alleles originating from outside the metapopulation always had new identities.

For germplasm types for which it was certain that the mother tree belonged to the same subpopulation as the subpopulation from which germplasm was obtained (i.e. seeds or cuttings), the old tree was not replaced in a particular season when the origin did not contain germplasm. For wildlings, the chance was calculated that the subpopulation of the mother tree differed from the subpopulation from which the germplasm originated, based on the number of trees with germplasm in each subpopulation and geneflow parameters (table 12.1). The value of a random number then determined whether the same subpopulation or other subpopulations were chosen as origin. In a similar fashion, the chances were calculated that the tree that provided pollen was different from the tree that provided germplasm (table 12.1) (such scenario was not relevant for cuttings). We simulated avoidance of selfing by setting the chance that pollen originated from the same tree to zero.

At the completion of each randomisation (after establishments of all new trees in season t), the algorithm calculated the proportion of homozygous loci (Hom, the percentage of trees where both alleles had the same identity), and Rényi diversity series values using base 2 (section 4.6). 2H0 corresponds to allele richness, H1 to the Shannon diversity index, 2-H2 to the Simpson index and 2-H8 to the frequency of the most frequent allele. The Simpson index represents the probability that two alleles selected at random, with replacement (an allele can be selected twice), have the same identity.

The expected homozygosity of individuals in generation t equals the inbreeding coefficient Ft, i.e. the probability that two alleles sampled at random from generation t-1 are identical by descent. For Wright-Fisher idealised populations of size N (Caballero 1994; Yeh 2000), the inbreeding coefficient equals

tt N

F )21

1(1 −−= .

A Wright-Fisher idealised population (described in the remaining text as randomly-mating population)

(...) consists of an infinite, randomly mated base population subdivided into infinitely many subpopulations, each with a constant number, N, of breeding individuals per generation. In each

METHODOLOGY

66

subpopulation, parents produce an infinite number of male and female gametes into a large pool from which only 2N are sampled and united to produce the N zygotes of the following generation. All individuals survive from birth to adulthood. Both the sampling of gametes and their union (including self-fertilization) are random, so that all parents have an equal chance of producing offspring (...) generations do not overlap, and only autosomal loci are considered (Caballero 1994)

As initially (t=0) each allele was given a unique identity and alleles originating from outside had a new identity, identity in state of alleles after t seasons meant identity by descent. The Simpson index calculated from the allele frequencies obtained in season t thus equals Ft+1 – drift (reduction in genetic diversity as a result from finite sampling) begins one generation before inbreeding (reduction in genetic diversity or homozygosity of the average individual) (Caballero 1994). The average and range of the Simpson index was also calculated directly (i.e. not from the respective Rényi profile value). This statistic formed the basis for calculation of the effective population size (i.e. the theoretical population with random mating that would show the same reduction in genetic diversity as the actual population that is investigated) (Caballero 1994; Yeh 2000):

1)1/(11 )))1(1(2( −+

+−−= tte FN .

We calculated genetic diversity for a randomly mating hermaphrodite or monoecious metapopulation of 50 individuals (consisting of five subpopulations of size 10 propagated by wildlings, Agemax = Agemat = 0; Flow = Pt = Ps = Pm = Po = Ws = Wo = 1, see table 12.1) after 50 seasons. We compared obtained values with results after 50 seasons for various scenarios of non-random mating for metapopulations of the same size (table 12.3), based on 100,000 randomizations.

We also investigated changes in neutral genetic variation of the 175 species in each village (metapopulations) based on the abundance of tree cohorts (trees of the same age) on farms (subpopulations). We calculated average results obtained after 50 seasons. Most tropical trees are self-incompatible (Boshier 2000). Differences in the time at which pollen is released and stigmas are receptive in a flower and spatial separation between anthers and stigmas promote outcrossing in many selfcompatible plants (Holsinger 2000). We therefore set Pt=0 except when we had information that the species was self-compatible. Other parameter values were set as Flow = 0.8, Ps = Ws = 20, Pm = 1, Po = 0.75, Wo = 0.995 and FM = 1. Cohort ages (years) communicated by farmers equaled 2Age0, the seasonal age of the individual trees. We assumed two flowering seasons per year for each species and, therefore, expressed age in seasons – the bimodal rainfall pattern results in two periods per year that are ideal for tree establishments. Agemax of each species was determined as the maximum of (i) the weighted average age of the oldest trees on each farm and (ii) the weighted average replacement ages of trees. Agemat was calculated as the minimum of 10 and (INT(Agemax/5) + 1). Trees of unknown age were provided with a random age ∈ [0, Agemax] (phenotypic differences, tree management, and small population sizes prevented a ccurate estimations of age from height, cross-sectional area, or volume, Kindt unpublished data).

The no-change scenario used the farmer-preferred origins and forms of germplasm as communicated by farmers for each species and each farm, thus assuming that farmers would not alter their preferences. Other scenarios that were tested were: (b) germplasm of the replacing tree (seedlings, unless the preferred form was cuttings) came from the old tree; (c) all germplasm (seedlings, unless the preferred form was cuttings) was selected randomly from the same farm; and (d) all germplasm (wildlings) were selected randomly from the same village. The latter scenario, the only one that used Ws = Wo = 1, simulated bulking seed within the village and making random selections from this bulk. Such simulations tested the consequences of different interventions aimed at altering the intraspecific management of trees on farm. We calculated 1000 randomizations for species with abundance < 1000, and 100 randomizations for the other species.

CHAPTER 4

67

Table 4.1. Results for species richness (S) for all farms and species (western Kenya) for subsamples of 100 farms obtained from the exact calculation and from randomizations. The 95% confidence interval limits are indicated by 2.5% and 97.5%. These limits were calculated based on standard deviation (St. dev.) in the two columns at the right-hand side, otherwise calculations were based on the histogramme of obtained values. References for the software are provided in the text. Software Rando-

mi-zations

Time Time / rando-

mization

S Mini-mum

Max-imum

2.5% 97.5% St. dev. 2.5% 97.5%

ExactS NA 0.17 NA 137.6732 NA NA NA NA NA NA NA Speciesrichref 10 0.17 0.0170 139.2000 130 148 129 148 5.81 127.81 150.59 Speciesrichref 100 0.68 0.0068 136.9000 121 154 124 150 6.38 124.40 149.40 Speciesrichref 500 2.76 0.0055 137.2140 121 154 125 149 6.22 125.02 149.40 Speciesrichref 1000 5.49 0.0055 137.4500 120 154 124 150 6.28 125.14 149.76 Speciesrichref 10,000 54.06 0.0054 137.6740 117 158 124 150 6.33 125.26 150.09 Speciesrichref 100,000 570.27 0.0057 137.6770 107 163 124 150 6.34 125.25 150.10 Speciesrichref 1,000,000 5456.90 0.0055 137.6743 107 166 124 150 6.34 125.25 150.10 EstimateS 500 1167.00 2.3340 137.4500 NA NA NA NA 6.37 124.96 149.94 PC-ORD 500 193.00 0.3860 137.6200 NA NA NA NA 6.34 125.20 150.04 Biodiversity Professional 50 35.00 0.7000 137.9600 NA NA NA NA NA NA NA

Table 4.2. Results for randomly-clustered species richness (S) for all farms and species (western Kenya) obtained from randomly attributing farms to artificial villages with the same size of the actual villages in western Kenya (50, 50, 50 and 51). The 95% confidence interval limits are indicated by 2.5% and 97.5%. The exact species richness for 50 randomly accumulated farms was provided as well. References for the software are provided in the text. Software Randomi-

zations Time Time /

rando-mization

S Minimum Maximum 2.5% 97.5%

ExactS NA 0.17 NA 109.2527 NA NA NA NA Speciesrichvillage 10 0.30 0.0300 108.5500 108.00 109.25 108.00 109.25 Speciesrichvillage 100 1.42 0.0142 109.3000 107.25 112.25 107.25 111.50 Speciesrichvillage 1,000 12.28 0.0123 109.2448 105.75 112.50 106.75 111.50 Speciesrichvillage 10,000 128.48 0.0128 109.2593 104.75 112.75 106.75 111.50 Speciesrichvillage 100,000 1326.02 0.0133 109.2528 103.25 113.75 106.75 111.50 Speciesrichvillage 1,000,000 12682.87 0.0127 109.2529 103.25 114.75 106.75 111.50

Table 4.3. Results at specific scale parameter values (a) for Rényi diversity profile values (Ha) for all farms and species (western Kenya) obtained from randomizations for subsamples of 100 farms. The 95% confidence interval limits are indicated by 2.5% and 97.5%, Avg indicates the average profile value. Randomi-zations

a Avg Mini-mum

Maxi-mum

2.5% 97.5% a Avg Mini-mum

Maxi-mum

2.5% 97.5%

10 0 4.9351 4.87 5.00 4.87 5.00 1 2.8049 2.74 2.90 2.74 2.90 2 2.4056 2.31 2.53 2.31 2.53 8 1.7226 1.63 1.80 1.63 1.80

100 0 4.9182 4.80 5.04 4.82 5.02 1 2.7622 2.65 2.90 2.65 2.86 2 2.3602 2.14 2.53 2.21 2.48 8 1.7300 1.43 1.97 1.55 1.91

1,000 0 4.9222 4.79 5.04 4.82 5.02 1 2.7609 2.58 2.92 2.65 2.86 2 2.3551 2.13 2.58 2.20 2.48 8 1.7138 1.40 2.02 1.50 1.89

10,000 0 4.9238 4.76 5.06 4.82 5.02 1 2.7581 2.57 2.92 2.64 2.86 2 2.3523 2.09 2.58 2.20 2.48 8 1.7160 1.35 2.03 1.50 1.89

100,000 0 4.9239 4.67 5.09 4.82 5.02 1 2.7581 2.52 2.94 2.64 2.86 2 2.3525 2.01 2.60 2.20 2.48 8 1.7171 1.27 2.05 1.50 1.89

1,000,000 0 4.9238 4.67 5.11 4.82 5.02 1 2.7584 2.50 2.97 2.64 2.86 2 2.3528 2.01 2.60 2.20 2.48 8 1.7170 1.27 2.11 1.50 1.89

METHODOLOGY

68

4.7. Addendum

An illustration is provided how Rényi diversity profile values can be calculated from the abundance distribution of species, in addition to the formulas provided in section 4.6. Calculations are provided in tabular format for four ecosystems. The equivalence between their profile values at scales 1 and 8, and the Shannon and logarithm of the inverse Berger-Parker indices is provided in the columns on the right-hand side. The diversity profiles for the four systems are provided in figure 4.1.

System Species Abun- p Diversity profile values Shannon Berger- dance a 0 0.99999 2 100 Parker A 1 10 0.47619 pa 1 0.476194 0.226757 6E-33 p lnp -0.3533 = pmax 2 6 0.285714 pa 1 0.285718 0.081633 3.92E-55 p lnp -0.35793 3 4 0.190476 pa 1 0.190479 0.036281 9.64E-73 p lnp -0.31585 4 1 0.047619 pa 1 0.04762 0.002268 6E-133 p lnp -0.14498 Sum 21 1 4 1.000012 0.346939 6E-33 -1.172066 0.47619 Formula

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

∑− pp ln

)1

ln(maxp

Profile H0 ˜ H1 H2 ˜ H8 H1 H8 Value 1.386294 1.172068 1.058607 0.749432 1.172066 0.741937 B 1 10 0.238095 pa 1 0.238099 0.056689 4.73E-63 p lnp -0.34169 = pmax 2 10 0.238095 pa 1 0.238099 0.056689 4.73E-63 p lnp -0.34169 3 6 0.142857 pa 1 0.14286 0.020408 3.09E-85 p lnp -0.27799 4 6 0.142857 pa 1 0.14286 0.020408 3.09E-85 p lnp -0.27799 5 4 0.095238 pa 1 0.09524 0.00907 7.6E-103 p lnp -0.22394 6 4 0.095238 pa 1 0.09524 0.00907 7.6E-103 p lnp -0.22394 7 1 0.02381 pa 1 0.02381 0.000567 4.7E-163 p lnp -0.08899 8 1 0.02381 pa 1 0.02381 0.000567 4.7E-163 p lnp -0.08899 Sum 42 1 8 1.000019 0.173469 9.46E-63 -1.865213 0.238095 Formula

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

∑− pp ln

)1

ln(maxp

Profile H0 ˜ H1 H2 ˜ H8 H1 H8 Value 2.079442 1.865215 1.751754 1.442579 1.865213 1.435085 C 1 17 0.708333 pa 1 0.708336 0.501736 1.06E-15 p lnp -0.24426 = pmax 2 1 0.041667 pa 1 0.041668 0.001736 9.5E-139 p lnp -0.13242 3 1 0.041667 pa 1 0.041668 0.001736 9.5E-139 p lnp -0.13242 4 1 0.041667 pa 1 0.041668 0.001736 9.5E-139 p lnp -0.13242 5 1 0.041667 pa 1 0.041668 0.001736 9.5E-139 p lnp -0.13242 6 1 0.041667 pa 1 0.041668 0.001736 9.5E-139 p lnp -0.13242 7 1 0.041667 pa 1 0.041668 0.001736 9.5E-139 p lnp -0.13242 8 1 0.041667 pa 1 0.041668 0.001736 9.5E-139 p lnp -0.13242 Sum 24 1 8 1.000012 0.513889 1.06E-15 -1.171194 0.708333 Formula

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

∑− pp ln

)1

ln(maxp

Profile H0 ˜ H1 H2 ˜ H8 H1 H8 Value 2.079442 1.171203 0.665748 0.348324 1.171194 0.34484 D 1 10 0.25 pa 1 0.250003 0.0625 6.22E-61 p lnp -0.34657 = pmax 2 10 0.25 pa 1 0.250003 0.0625 6.22E-61 p lnp -0.34657 = pmax 3 10 0.25 pa 1 0.250003 0.0625 6.22E-61 p lnp -0.34657 = pmax 4 10 0.25 pa 1 0.250003 0.0625 6.22E-61 p lnp -0.34657 = pmax Sum 40 1 4 1.000014 0.25 2.49E-60 -1.386294 0.416667 Formula

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

α

α

−∑

1)ln( p

∑− pp ln

)1

ln(maxp

Profile H0 ˜ H1 H2 ˜ H8 H1 H8 Value 1.386294 1.386294 1.386294 1.386294 1.386294 1.386294

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

H

A A AA A A A A A A A A

AA A

CC

C

C

C

C

C

CC

CC

C

C C C

D D D D D D D D D D D D D D D

B B BB B B B B B B B B

BB B

1.0

1.6

2.7

4.5

7.4

12.2

20.1

EX

P(H

)

Same richness Same evenness

Almost sameShannon index

Largest frequency ofthe dominant species

A

B

C

D Figure 4.1. Diversity profiles for four hypothetical ecosystems. Symbols indicate trees of the same species. System A has the same richness as system D, which is lower than the richness of system B = C. System A has the same evenness as system B. System D has a perfectly even distribution, and system C the least even distribution. System B is the most diverse, and system D is more diverse than

system A. System C can not be ranked as more or less diverse than systems A and D, as it is richer but less evenly distributed than these other two systems.

0 1 2 3

100

105

110

115

120

125

130

135

140

145

150

155

160

165

170

175

randomizations = 10time / randomization = 0.0170

0 1 2 3 4 5 6 7 8 9 10 11

100

105

110

115

120

125

130

135

140

145

150

155

160

165

170

175


0 10 20 30 40 50 60 70

100

105

110

115

120

125

130

135

140

145

150

155

160

165

170

175

randomizations = 1,000time / randomization = 0.0055

0 100 200 300 400 500 600 700

100

105

110

115

120

125

130

135

140

145

150

155

160

165

170

175


0 1,000 2,000 3,000 4,000 5,000 6,000 7,000

100

105

110

115

120

125

130

135

140

145

150

155

160

165

170

175


0 10,000 20,000 30,000 40,000 50,000 60,000 70,000

100

105

110

115

120

125

130

135

140

145

150

155

160

165

170

175

randomizations = 1,000,000time / randomization = 0.0055

Figure 4.2. Number of permutations, excluding zero counts (horizontal axis) with particular species richness (vertical axis) for 100 randomly accumulated farms. The boxplot shows minimum, maximum and 95% confidence intervals. The horizontal reference line shows the calculated average species richness obtained through randomization, the diamond shape shows the average species

richness obtained through the new approach based on the hypergeometric distribution.

0 1 2 3

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115


0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115


0 10 20 30 40 50 60 70 80 90 100

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115


0 100 200 300 400 500 600 700 800 900

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115


0 1,000 2,000 3,000 4,000 5,000 6,000 7,000 8,000 9,000

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115


0 10,000 20,000 30,000 40,000 50,000 60,000 70,000 80,000 90,000

100

101

102

103

104

105

106

107

108

109

110

111

112

113

114

115


Figure 4.3. Number of permutations, excluding zero counts (horizontal axis) with particular randomly-clustered species richness (vertical axis) for farms that are randomly allocated to clusters of sizes 50, 50, 50 and 51, which corresponds to the sizes of the villages encountered in western Kenya. The boxplot shows minimum, maximum and 95% confidence intervals. The horizontal reference line

shows the calculated average species richness obtained through randomization, the diamond shape shows the average species richness obtained through the hypergeometric distribution.

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 200

1

2

3

4

5

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf


0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 220

1

2

3

4

5

0 0.1 0.25 0.5 0.75 1 1.25 1.51.75 2 2.5 3 10 100 inf


0 20 40 60 80 100 120 140 160 180 2000

1

2

3

4

5

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf


0 200 400 600 800 1000 1200 1400 1600 1800 20000

1

2

3

4

5

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf


0 2000 4000 6000 8000 10000 12000 14000 16000 18000 200000

1

2

3

4

5

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf


0 40000 80000 120000 160000 2000000

1

2

3

4

5

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf


Figure 4.4. Number of permutations, excluding zero counts (scale on lower horizontal axis) with particular diversity profile values (vertical axis) for 100 randomly accumulated farms. The upper horizontal axis shows the scales of the profile values (same as figure 4.1), the connected diamond shapes show the average diversity profile calculated through the randomizations. The boxplot shows minimum, maximum and 95% confidence intervals. Histogrammes and boxplots are presented at the right and left of the specific diversity profile scale, respectively. Not all histogrammes and boxplots

were included for clarity of the figure.

CHAPTER 5 THE STUDY OF ON-FARM TREE SPECIES DIVERSITY IN WESTERN KENYA TO PLAN FOR AGROECOSYSTEM

DIVERSIFICATION R KINDT, AJ SIMONS & P VAN DAMME SUBMITTED TO FORESTS, TREES AND LIVELIHOODS

A survey was conducted in western Kenya involving a complete tree census and collection of ethnobotanical information on 201 farms with the final objective of diversifying on-farm tree species composition through domestication. In total, 102,138 trees were surveyed belonging to 175 species, with 60 different uses, averaging 3.7 uses per species. On-farm species richness and use-group richness were positively correlated (Spearman ?=0.65). Average number of species and uses per farm were 16.6 (15.7-17.5) and 9.7 (9.4-10.1), respectively. Alpha and gamma diversity, and richness and evenness were investigated for the most frequent use-groups in the landscape. Major differences were detected, which allow domestication to target groups of low species diversity. When removing the dominant species, five groups had average species densities lower than one tree ha -1 while only two groups had densities higher than 10. This points to potential problems associated with small population sizes.

AGROECOSYSTEM DIVERSIFICATION

74

5.1. Introduction

One of the purposes of agroforestry tree domestication is the enhancement of the stability and productivity of agroecosystems by diversifying on-farm tree species composition (species richness and evenness of species abundances). Diversification and intensification of land use through the domestication of agroforestry trees (woody perennials) is one of the three pillars of ICRAF’s research agenda (ICRAF 1997; ICRAF 2000). In this chapter, methods are described for grouping tree species according to their uses on farms and analysing the diversity of the use-groups to which these trees belong. Information on group diversity resulting from these methods can be used to measure the impact of interventions on on-farm diversity and steer interventions to maximise impact in the direction of greater agroecosystem diversity. The inclusion of information on species uses in the analysis – rather than just analysing tree species diversity for complete farms – is related to the ultimate aim of enhancing on-farm productivity through increased diversity of species used.

Field experiments and ecological models have shown that there is a positive, but conditional relationship, between species diversity and ecosystem stability and productivity (for example Frank & McNaughton 1991; Tilman & Downing 1994; Naeem et al. 1994; Tilman 1996; Tilman et al. 1996; Tilman et al. 1997a, 1997b; Hooper & Vitousek 1998; Hector et al. 1999; Ives et al. 1999; Yachi & Loreau 1999; Nijs & Roy 2000; Ives et al. 2000). Findings of models and experiments involving crop mixtures in agroecosystems showed the same positive relationship (Dover & Talbot 1987; Trenbath 1999; Van Noordwijk et al. 1994; Swift et al. 1996; Van Noordwijk & Ong 1999). Whether the higher productivity associated with higher diversity was the result of overyielding (a synergistic relationship among species so that the productivity of the mixture is higher than any species growing in isolation), or the result of sampling more productive species in more species-rich mixtures (a phenomenon described as the sampling effect), however, remains controversial (Huston et al. 2000; Kaiser 2000a, 2000b; Purvis & Hector 2000). In the latter case, each combination of species that includes the most productive species would have the highest productivity, including monocultures. McCann (2000) mentioned that the condition of generally weak bioenergetic consumer-resource interactions needed to be added so that diversity could enhance stability.

Research has demonstrated that diversity only explains a small percentage of variation in the data so some species-poor ecosystems can still have relatively high stability or productivity and vice versa. Waide et al. (1999) reviewed over 200 relationships between diversity and productivity in the ecological literature (excluding intensively managed systems) and reported that 30% were unimodal, 26% positive linear, 12% negative linear, and 32% not significant. For natural ecosystems, different and probably interacting factors could lead to higher species richness, compared to managed ecosystems. Tilman (1999) describes four explanations for high species richness, including local spatial heterogeneity, nonequilibrium conditions, interactions among at least three trophic levels, and neighbourhood recruitment limitation, all given appropriate trade-offs in species traits. In general, species coexist through evolutionary persistent trade-offs in their abilities to deal with factors that constrain their fitness and abundance (Tilman 2000).

Theoretical and field studies have shown that diversity in species traits and environmental heterogeneity are explanatory factors for the diversity-productivity and diversity-stability relationships. Heterogeneity in environmental and species characteristics results in saturating effects of increased diversity because randomly added species will each differ less from species already present. Chapin et al. (2000) described how diversity had diminishing effects on ecosystem functioning because of niche overlap and increment of redundancy. The studies confirmed the redundancy hypothesis of the options for the relationship between diversity and stability listed by Johnson et al. (1996) who described that species can be arranged in functional groups, so that ecosystem stability will only suffer when all species are removed from a particular functional group.

CHAPTER 5

75

Gitay et al. (1996) and Chapin et al. (2000) pointed out that all the functions of an individual species within an ecosystem are rarely understood, so that the effects of species removal cannot be predicted. Furthermore, De Leo & Levin (1997) warn that long-term stability of the ecosystem may suffer from reducing diversity within functional groups, whereas Ives et al. (1999) stated that diversity was still the best guarantee against unknown future changes. The sampling effect is only valid for sites where environmental conditions do not change (on spatial or temporal scales) to the effect that a different species becomes the most productive.

Because of saturating effects of diversity on productivity and stability, the approach taken in the study presented here was to identify use-groups with low diversity (in analogy with identification of functional groups with low redundancy) because more important effects on ecosystem functioning can be expected from increasing the diversity in groups with lower diversity levels. Diversity enhancement is considered at both the farm (alpha diversity) and landscape level (gamma diversity, within the same agroecosystem), as stability and productivity enhancement by diversity can be expected to occur at these scales because heterogeneity of species traits and environmental characteristics occurs at both levels.

The first set of analyses reported deals with the delineation of use-groups. The remainder of the paper compares the diversity of various groups. By calculating statistics describing the diversity for various tree products such as fruits and medicines, and services such as soil fertility improvement or boundary demarcation, use-groups with lower diversity can be identified and subsequently targeted for domestication. This approach to domestication is concerned with addressing interspecific diversity. Data on group abundance are used to assess intraspecific diversity of individual species.

5.2. Materials and Methods

Formulation of Use-group matrices (4.1), alpha and gamma diversity (4.2.1), diversity and evenness statistics (4.2.6), use-group abundance and density (4.3), correlations (4.4.1), robust MM regression (4.4.2)

5.3. Results

5.3.1. Distribution of Uses over Species

On the 201 inventoried farms, 102,138 trees were counted for 175 species. Use information was obtained for 3,331 species-farm combinations. This results in an average number of species per farm of 16.6 (15.7-17.5, P=0.95) (P: probability, in this case indicating a 95% probability interval for the mean), which is a relatively high number considering the average farm size of 0.58 ha (0.52-0.64 ha, P=0.95). Farmers listed 60 different uses for the species on their farms. The average number of uses per species was 3.7 (Uadj, 3.3-4.1, P=0.95). The frequency distribution of uses is presented in figure 5.1. Four species (each occurring on only one farm) had no uses. Considering the average number of uses and figure 5.1, diversity of use-groups could be underestimated severely by only studying the main uses of species (Umain) since only 23 species (13%) had a single use.

Not all uses of individual species were adopted on every farm. The average number of uses per species per farm was 2.0 (1.9-2.1, P=0.95). When analyzing the uses per species for each individual farm, 48.2% of farm-species combinations had two uses, compared to 29.7 and 15.1%


76

of combinations having three and one use(s) respectively. The other cases of zero, four and five uses were much rarer with respective percentages of 1.0, 6.0, and 0.5.

The fact that not all uses were adopted on farms can also be concluded from figure 5.2, which shows that most species occur below the line (not drawn) of total number of uses = average number of uses. Of the 69 species on the reference line, 63 only occurred on one farm.

There was a significant but low positive correlation between the total number of uses and the average number of uses of species (Spearman ?=0.56, P=0; Kendall t=0.39, P=0). The regression analysis provided in figure 5.2 confirms this conclusion, but the equation explained relatively little variation.

Because firewood was the most important secondary use and not very prominent as main use (firewood was the main use in only 20% of cases where it was mentioned, table 5.1), the average number of uses was analysed excluding firewood. On individual farms, the average number of uses per species thus dropped to 1.2 (1.1-1.3, P=0.95). The regression of the average number of uses excluding firewood on the total number of uses (including firewood) is included in figure 5.2. The parallel regression lines of figure 5.2 (where one includes and one excludes firewood) shows that firewood is on average one of the two uses that each species has on the farm. Species with few uses in total have about one other use on average, while species with many uses have about 1.5 other uses on average.

Because not all potential uses are realised, calculations should be based on actual uses on individual farms as provided by Urec and Uadj as otherwise sizes of use-groups (especially rare groups) could be overestimated.

5.3.2. Effects of Sample Size on Numbers of Uses Recorded per Species

The effects of species frequency (number of households) and the number of uses that was recorded per species was investigated. Figure 5.3 shows a positive relationship that was confirmed by Spearman rank correlation (?=0.82, P=0), Kendall tau (t=0.59, P=0) and linear regression (36% of variation explained). A positive relationship was also revealed for the number of uses per species and the total number of trees counted (Spearman ?=0.76, P=0; Kendall t=0.54, P=0). The positive relationship could be the result of sampling so that rarer uses had a higher chance to be detected in more frequent species. The sampling effect could also have played a role in that farmers had more chances of discovering additional uses for species that are more frequent in the area. The possibility that more species with more uses would be established or retained (this was the actual hypothesis for the relationship between species frequency and total number of uses) seems less likely. The reason is that the average number of uses does not increase very much with species frequency as shown in figure 5.4 (only an increase of 0.6 uses from rare to frequent species). Moreover, only low correlations were detected using Spearman (?=0.30, P=0.0001) and Kendall (t=0.19, P=0.0001) correlation tests, and the regression explained relatively little variation.

For the analysis of use-group diversity, the findings again suggest to use Urec and Uadj definitions. Doing otherwise (using Uall) would over-emphasize rare uses. The Uall definition allows analysis of expected changes in use-group diversity if species would be used for all their potential purposes on each farm where they occur. A possible shortcoming for this type of analysis could be that the potential uses of rare species would be underestimated because these had smaller chances of being sampled.

Related to figures 5.3 and 5.4 is the analysis of the total number of uses per household versus the number of species per household. The data suggest that higher richness of species on farms is established to provide new uses (figure 5.5; Spearman ?=0.65, P=0; Kendall t=0.47, P=0;

CHAPTER 5

77

regression explaining 39% of variation). The average number of species per use also increases for farms with more species (figure 5.6; Spearman ?=0.70 P=0; Kendall t=0.50 P=0; regression explaining 48% of variation). Farms with higher species richness therefore also have higher use redundancy.

5.3.3. Frequency of Use-groups over Farms

Figure 5.7 shows the distribution of use-groups over households where they were mentioned (Uadj). It shows that some uses are very rare (19 occur only on one farm, seven only on two farms and five only on three farms), while five groups are nearly ubiquitous (occurring on more than 195 farms = 97%). The average number of uses per household was 9.7 (9.4-10.1, P=0.95), while the frequency distribution of farms with a certain number of uses is shown in figure 5.8.

Listed in table 5.1 is the number of times that a use was mentioned, counted both as number of farms × number of species with a certain use, and as number of farms (i.e. the use-group frequency). The effects of various use-group definitions can thus be compared. Only studying the main use of species would result in underestimation of use-group frequency, especially for those groups that include more secondary uses. Firewood is the prime example of such a use-group, but also for many other groups focussing on Umain would result in incorrect evaluations. Assigning all uses to all households (Uall) resulted in large overestimation of frequencies of most groups. The recommendation, therefore, is to analyze group diversity based on uses actually recorded for species at particular farms.

Fruit and beverage were the only use-groups in which Umain and Uall definitions would not lead to major discrepancies with Urec and Uadj, which means that these were the only groups where actual use was close to (for beverage exactly the same as) main and potential use. The greater frequencies of the Uall grouping (only beverage occurred on fewer than 199 farms = 99%) mean that alpha diversity can be increased substantially for most groups by obtaining the potential use from each species-farm combination. For example, if farmers used each species for timber that had already been used by a least one farmer for this purpose, then the average number of species used for timber per farm would double on average. It is possible, however, that not all the potential uses were obtained because particular species do not have optimal traits to provide those uses. Using the same example of timber, it is possible that many farmers do not use particular species on their farms for timber since they are not satisfied with the timber quality of the species, or with its growth rate. For the scenario where species that are present now do not have optimal traits, alpha diversity would be better enhanced by more widely distributing species with better performance for the particular use (and not only uses in species already present).

No major differences were detected between the figures resulting from the Urec and Uadj definitions. Because some information was evaluated as missing (i.e. a specific use was expected of a species when more than 50% and more than five farmers reported the particular use for the species), Uadj was selected as the optimal definition to be used in analyses of group differences. Another reason to use Uadj was that use information was actually missing for 97 farm × species cases. This resulted from overlooking species when interviewing the household about uses.

The average number of species per group was derived from the other two statistics provided in table 5.1 and thus reflects the average number of species per farm where the use occurs. Using the threshold level of a minimal ‘redundancy’ of one species (corresponding to a minimum of two species), soil fertility, charcoal, beverage, and fodder can be identified as groups with lowest average buffering against non-performance of the most frequent species (since for these groups, less than one species would remain in the group). These were the groups with lowest farm frequencies as well.


78

5.3.4. Diversity Characteristics of Use-groups

Table 5.2. provides characteristics relating to alpha and gamma diversity of use-groups, and relating to richness and evenness of species distribution in these groups.

When analyzing alpha and gamma diversity of groups (Uadj) based on Stot and Savg, four pools of use-groups seem to materialise: (i) high gamma diversity, comprising firewood and shade (pool A); (ii) medium gamma diversity, containing timber, medicine, and ornamental (pool B); (iii) moderate gamma diversity, consisting of fruit, boundary, construction, soil fertility, and charcoal (pool C); and (iv) low gamma diversity, containing beverage and fodder (pool D). Differences in Savg and associated relatively narrow confidence intervals make comparison of alpha diversity possible. Within pool A, firewood has higher alpha diversity, but both firewood and shade have quite high alpha diversity (high alpha diversity has been defined here as Savg > 2 or a redundancy of one). Within pool B, timber is the only use-group with high alpha diversity. Pool C contains two use-groups with high alpha diversity (fruit and boundary demarcation) (with a possible inclusion of construction based on the upper boundary of the confidence interval). Both use-groups of pool D have very low alpha diversity. Analysis based on Urec would have resulted in similar conclusions. However, for Umain only firewood and fruit would have been identified as use-groups with high alpha diversity, while for Uall beverage would have been the only use-group with low alpha diversity.

When analysing diversity of species within use-groups at the survey or landscape scale groups (Uadj), species richness corresponds to the gamma diversity that was discussed above. Use-group evenness can be compared because average values differed substantially, although confidence intervals were quite wide. In some cases, the lower limit of the confidence interval for the frequency of the dominant species included 0. In these cases, the upper boundary of the confidence interval for the inverse Berger-Parker index could not be calculated. Both use-groups in pool A have high evenness of distribution. Values higher than 3 are considered to be high and less than 1.5 low; these limits correspond to the dominant species contributing 1/3 or 2/3 of total group abundance respectively. All use-groups in pool B had medium evenness with the dominant species however taking more than half of total group abundance on average. Within pool C, boundary demarcation had high evenness (allowing a slightly smaller lower confidence limit), whereas fruit and construction had low evenness. Beverage had low evenness in pool D. As for Savg, using Urec grouping would have resulted in similar conclusions. Differences occurred in Umain grouping for shade (very wide confidence interval) and charcoal (low evenness). For Uall definition, only fruit could be identified as a group with low evenness, while several use-groups progressed to very even distribution.

Studying d-1csa for Uadj groups revealed that some use-groups (fruit, medicine, and ornamental)

progressed to more evenly distributed species. This means that dominant species in these groups based on cross sectional area dominated less than dominant species based on species abundance. Firewood showed the opposite trend, while for other groups d-1 and d-1

csa patterns were similar.

Comparing d-1 with evenness values obtained through the broken stick distribution (d-1bs) showed

that most use-groups had lower evenness values than those of the broken stick distribution. Exceptions were the shade (with a wide confidence interval) and fodder groups. Comparing the broken stick values with d-1

csa revealed that ornamental and fodder had higher upper confidence limits. In the case of the ornamental group, the confidence limit was even higher than the maximum possible value (Stot) which resulted from the large width of the confidence interval (despite the width of the interval, ornamental can be defined as a group with high evenness).

CHAPTER 5

79

5.3.5. Abundances and Density of Use-groups

Table 5.3 shows farm abundance, use-group density and average species densities in a use-group. The same pattern of differences between Urec, Uadj, Umain, and Uall that occurred in previous sections was observed once more. Values for Urec and Uadj showed minor differences, while values for Umain and Uall led to under- and overestimation, respectively.

Values for average use-group abundance per farm (Uadj) were quite high, except for fodder. Small farm sizes (average 0.52-0.64 ha, P=0.95) resulted in relatively high use-group densities. Firewood, timber, boundary demarcation, construction, and beverage were groups with average densities higher than 200 trees ha -1. Gamma diversity, however, caused substantially lower average species densities. Two groups (shade and medicine) had SDtot values < 1 tree ha -1, while five groups had values < 2 trees ha -1. When analyzing densities by removing the dominant species (SDtot-d), values were much lower illustrating the uneven distribution in many groups. Not surprisingly, groups with the lowest d-1 values (table 5.2) were also groups with the largest differences between SDtot and SDtot-d. Five groups had SDtot-d values = 1 tree ha -1. Only two groups (boundary demarcation and beverage) had values > 10 trees ha -1.

5.4. Discussion

The results indicate a complex pattern of distribution of species and uses over farms. Because the basic management unit is the farm, the most relevant option to analyse use-group diversity seems to be to base analysis on species distribution over households and information collected on all uses of the species at the particular farm (i.e. Urec or Uadj).

There was a large difference and low correlation between the total number of uses of a species, and the average number of uses per species per farm. This phenomenon could be interpreted in terms of gaps in knowledge on how particular species could be used, or as a lack in perceived need for a certain use. The lack in perceived need for the use could result from the fact that the species is only used when no other (better) species can provide the use ( i.e. species of last resort). Other possibilities could be that the use responds to rare needs only occurring in a few households, or because the use is only needed infrequently on farms.

Another area where the difference between the actual and potential uses emerged was in household occurrence of certain use-groups. For instance, medicinal use was observed in 112 farms (56%), but given farm species composition potentially occurred in 199 farms (99%). The information provided in Figure 7 and Table 1 therefore points to potential interventions. One strategy could be to increase the frequency of some use-groups in the landscape where these uses do not occur. Groups of medium occurrence could be selected for wider distribution such as medicine, soil fertility, or fodder occurring now on respectively 112 (56%), 136 (68%) and 47 farms (23%) (Uadj). A combined strategy could involve also targeting those farms with a low total number of groups (for instance lower than eight uses per household, Figure 8) and supplement additional uses to these farms. Such interventions would increase the alpha diversity of the more widely distributed use-groups and could also increase gamma diversity provided that new species contribute to the group.

One question that needs to be solved, however, is in how far households that use none of the trees on their farms for a particular use would wish to obtain the use from their own farms. For instance, 23% of households obtained tree fodder from their farms, while 71% of households had cattle (the number of small ruminants was not recorded). The question to be solved is how many of these 71% see the need to obtain tree fodder from their own farms to feed their animals, and whether this number equals the 23% that currently obtain tree fodder from their farms.


80

Some farmers could indeed choose to use other fodder sources, for instance grass from the own farm or tree fodder from other farms. A related question is how many farmers without livestock could produce and sell fodder – maybe even more than 71% of farmers could see this as a valid option. A second-generation survey is currently underway focusing on these and similar questions, collecting information on so-called farmer-desired diversity per farm. Because the desire of households to realize a particular use on their own farm is currently unknown, it seems more appropriate to focus on use-groups with medium occurrence than use-groups with low occurrence, as the latter use-groups could be limited to rare needs.

The question whether differences Uadj and Uall are the result of gaps in knowledge on uses of species, rarity of need for the use, the choice to obtain the product from off-farm sources (for example through purchase from other farmers) or from alternative on-farm sources thus remains unanswered. Both average number of uses per species and use-group frequencies highlight the potential value of extension messages on alternative uses of species (in case the hypothesis of gaps in knowledge on species uses holds true). Widespread extension of information on potential uses of species that do not occur on all farms at present could result in increased on-farm diversity if this information would encourage farmers to incorporate new species in their farm.

An increment of alpha diversity (Savg) would buffer use-groups of individual farms better against random low performance of individual species resulting from heterogeneity of environmental conditions, which could include weather conditions, diseases, pests, or market failures. Arnold (1995) stated that farmers plant trees in pursuit of their livelihood goals of income generation, risk management, household food security and optimum use of available land, labour and capital. Warner (1995) stated that farmers plant trees based on their needs for tree products, decline in access to off-farm resources, lack of alternative products, increasing market opportunities and declining on-farm labour. The analysis presented here does not distinguish whether a particular use-group has more on-farm consumption or more sale value. Although we collected such information, we did not make the distinction as it could be equally important to diversify use-groups mainly used for sale and use-groups mainly for direct use.

Groups with low alpha diversity (Savg) were identified and could be targeted for diversification efforts. Gamma diversity (Stot) provides suggestions on how alpha diversity can be improved: for those groups with higher gamma diversity (belonging to pools A, B, and possibly C), a wider distribution of existing species within the area would offer one method of enhancing alpha diversity. For groups with low gamma diversity, the solution would rather be to introduce new species or to promote alternative uses for species that are already present. Increasing gamma diversity could also result in increased stability and productivity at the landscape level (under conditions as discussed in the introduction). Therefore, identification of use-groups with low gamma diversity and subsequent increment of gamma diversity could have a direct impact at this level. High alpha diversity of Uall groups suggests that promotion of a higher number of alternative uses would increase alpha diversity to reasonable levels for most use-groups.

The difference between the Uadj and Uall groups potentially points to farmer-perceived redundancy within groups – some species could provide a certain use, but are not used in particular households because other species serve the same purpose. The questions to be resolved, however, when promoting additional species are: (i) the quality of production of these species; (ii) the complementarity in production by these species with species that are already used (for instance seasonal differences); and (iii) differences in species’ traits which are the basis for enhancement of the stability of production.

Information on richness and evenness of species abundance distribution within use-groups revealed that the evenness of many groups was very low. The evenness of the majority of use-groups was lower than the evenness associated with broken stick distribution, a reference-value that could be used as a target for evenness increment. Even for the scenario where all possible

CHAPTER 5

81

products are obtained from the species that are present on-farm (Uall), low evenness values prevailed. This means that even if all species would be used, evenness would still be low. Fewer use-groups had high evenness than the number of groups with high richness. From a diversification point of view, increment of evenness thus seems to be a higher priority. Low evenness implies that groups have weak buffering of performance against the loss of the dominant species, while disturbances might target species with large abundance (see for instance Wills et al. 1997).

Species densities in many groups were very low, especially if the dominant species was removed from the group. The large effect of removing the dominant species on calculated average species densities is another indication of the need for more even distribution of species. As described in Kindt & Lengkeek (1999), most species in the western Kenya survey area occur in very low densities. The inverse J-shape of species rank-abundance plots is also the typical situation for tropical rain forests (Kohyama 1991; Magnussen & Boyle 1995). Because of this distribution pattern, most species occur in considerably lower densities than calculated even when the dominant species is removed from each group. Therefore, an analysis of the specific density of each species is needed.

The average density per species (SDtot) showed that, in case species were completely evenly distributed (i.e. when each species would have the same density Davg), 10 use-groups would have more than 1 tree ha -1. This figure is higher than the typical density of canopy trees of tropical rain forests of one or fewer adults per hectare (Chase et al. 1996). It may not be realistic, however, to promote perfectly even distribution of species over groups, for instance in those cases when farmers prefer particular species. Therefore, some species may be maintained in very small densities in the landscape. Small plant population sizes result in rapid genetic erosion with an expected loss of genetic diversity per generation of minimum (2 × population size)-1 (the cases where expected losses are smaller, listed for example by Caballero 1994, do not apply). Simons et al. (1994) stated that genetic variation must be maintained in domesticated agroforestry species to allow for a variety of user needs and environmental conditions. To evaluate the influence of small densities on the genetic variation that can be maintained in on-farm tree populations, information is also needed on the spatial distribution of these species in the area. Species are aggregated when the average density of conspecifics in close proximity to each tree is higher than overall density. Condit et al. (2000) reported that nearly every tree species in six tropical forests had an aggregated distribution, and that the rarest species were the most aggregated. For species with small landscape densities, genetic erosion would be limited by encouraging farmers to grow these species in an aggregated pattern, which would also correspond with the natural distribution of species.

Information on population sizes of species is also needed to evaluate the potential of their conservation through use. Conservation through use may be the best option for many species considering forest decline in the area (Kindt & Lengkeek 1999). The genetic diversity of these species will need to be maintained to allow for long-term conservation. Small population sizes, however, should not be a reason not to use particular species in agroecosystems. In such cases, however, continuous interventions (germplasm inputs) will be needed to ensure genetic diversity is maintained, and genetic diversity will need to be conserved elsewhere either through in situ or ex situ approaches. This type of intervention for low-density species could be considered as strict domestication of the landscape (analogous to the strict definition of domestication sensu Harlan 1975) as these species will depend on human interventions to survive in the agroecosystem.

The analyses focused on species richness and diversity, while productivity is also related to abundance. A diverse use-group may still not cover all demands for products or services that are needed on an individual farm, while a less diverse use-group could provide surplus. Linked to this observation is the fact that use-groups were not validated in terms of importance to the farm household (e.g. economic importance, importance for food security). Diversification could be


82

targeted towards more important use-groups, rather than targeted towards those groups which have low diversity. Because of small farm sizes and since the current average number of uses per household is ten, the decision could be made not to promote all uses. Decreasing the number of uses per farm could result in higher profitability per farm. An analogy is the criterion introduced by Van Noordwijk et al. (1997) of the shape of the relationship between biodiversity and profitability. If the shape of the trade-off curve indicates that initial diversity loss would result in large gains in profitability, then these authors suggest that a segregation (specialisation) approach may be more appropriate – if increment of profitability is the major goal for the landscape. Similarly, Van Noordwijk & Ong (1999) indicated that the value of diversity in agroecosystems strongly depended on the ability of farmers to derive value from a large number of components, and not from one dominating component.

With fewer use-groups on a farm, their within-group diversity could be increased to higher levels than possible with many use-groups. However, reduction in the number of use-groups per farm could result in substantially greater risks to individual farmers. The evenness of inter-group diversity was not investigated, but such investigation could provide insights in potential reduction of use-group number that would allow for greater evenness of uses. Reduction of use-groups on farms does not necessarily lead to reduction of diversity at the landscape level (i.e. beta diversity increment can buffer for alpha diversity decrease), so that stability at the landscape scale may not suffer. Interestingly, species population size arguments as presented above also advocate for approaches that focus on both alpha and beta diversity.

Some other aspects relating to diversity, and productivity and stability that require further investigation are listed below.

• One aspect that was not included in the analysis is that not all trees of a species necessarily contribute to the uses listed for the species. For instance, only half of the trees of a species used for firewood (156 or 89% of the 175 species encountered, and 15.2 or 92% of the average 16.6 species occurring per farm were used for firewood) could be actually used for that purpose. Linking species abundance to productivity should also incorporate information on intra-specific variation in productivity, while different parts of each tree may be used for a particular purpose. The analysis presented, therefore, still includes an aspect of potential use rather than actual use. With the purpose of detecting groups of low diversity, this level of detail may not be necessary.

• The local value of diversity was not included in the analysis, whereas local values may include other components than productivity. Thrupp (1998) mentioned values of agrobiodiversity in addition to increased productivity and stability, including diversification of income opportunities to farmers, reducing risks, increasing efficiency of resource use, increasing nutritional values, and provision of sources of medicines and vitamins. Swift et al. (1996) mention the diverse set of goals that farming communities may have such as minimising risks, attaining minimum production or preserving cultural traditions. A second-generation survey attempts to document farmers’ expectations related to the diversity within the different use-groups. A related issue is the implementation of agroecosystem diversification. Harrington (1996) distinguished between agroecosystem diversification through design (involving participatory research on indigenous knowledge about system diversity) and demand-led diversification (where farmers follow market signals to diversify their farming systems). He pointed out that the former approach had not led to widespread adoption by farmers, while the second approach does not necessarily foster sustainable management. He advocated for a combination of both approaches. Swift et al. (1996) contrasted agroecosystem design methods of introducing small changes into traditional agricultural practices with the total conversion of a natural ecosystem into systems that only contained desired elements irrespective of background ecological conditions. They distinguished a third approach where the

CHAPTER 5

83

agroecosystem is reconstructed using biological diversity rather than rejecting it. Information on areas in the agroecosystem where tree diversity is lower is not the only information necessary for successful implementation of participatory agroecosystem diversification. Reasons may exist why low diversity of some groups is locally preferred, so that provision of greater options to farmers (new information on how species that are present can be used, or wider distribution of species) would not lead to greater diversification. We believe, however, that the approach presented offers important information on interventions points on which diversification could focus, even if interventions would only indicate that local management is not constrained by the number of options that are currently available.

• The analyses were conducted at the landscape level (comparing alpha and gamma diversity), whereas differences in diversity also occur among individual farms. Obviously, greater differences among farmers result in larger gamma diversity, but only average alpha diversity was considered. Kindt et al. (Chapter 8) analysed diversity statistics of individual farms and their differences in species composition since individual farms (or villages) are the true targets for increasing diversity.

• The aspect of intra-specific diversity was discussed above. Another aspect that could be taken into consideration for domestication of species at the landscape scale could be supra-specific diversity. For instance, when buffering against diseases is sought, diversifying botanical families may be more effective than only diversifying species within genera.


84

Table 5.1. Characteristics of the 12 most frequent use-groups based on various group definitions (see methods). Use-group Urec Uadj Umain Uall occur. farm

oc. sp. avg. occur. farm

oc. Sp.

Avg. occur. farm

oc. sp. avg. occur. farm

oc. sp. avg.

Firewood 2834 201 14.1 3053 201 15.2 594 175 3.4 3165 201 15.7 Fruit 920 201 4.6 944 201 4.7 906 201 4.5 948 201 4.7 Timber 623 184 3.4 775 196 4.0 288 130 2.2 1623 201 8.1 Shade 554 157 3.5 597 166 3.6 172 105 1.6 2795 201 13.9 Boundary demarcation 473 191 2.5 571 197 2.9 411 186 2.2 1852 201 9.2 Construction 355 191 1.9 389 197 2.0 283 182 1.6 1415 201 7.0 Medicine 239 108 2.2 251 112 2.2 158 86 1.8 1374 199 6.9 Ornamental 194 87 2.2 208 96 2.2 74 51 1.5 1616 200 8.1 Soil fertility improvement 156 106 1.5 196 136 1.4 89 76 1.2 1170 200 5.9 Charcoal 124 80 1.6 124 80 1.6 18 16 1.1 1498 201 7.5 Beverage 74 62 1.2 74 62 1.2 74 62 1.2 74 62 1.2 Fodder 46 40 1.2 53 47 1.1 25 23 1.1 454 199 2.3 occur.: number of times the use was mentioned; farm oc.: number of households where the use was mentioned; sp. avg.: average number of species per farm and per use for those farms where the use was mentioned

CHAPTER 5

85

Table 5.2. Diversity characteristics of the 12 most frequent use-groups based on various group definitions (see methods) Use Statistic Use-groups

Group defi-nition

Fire-wood

Fruit Timber Shade Boun-dary

demar-cation

Con-struc-tion

Medi-cinal

Orna-mental

Soil fertility impro-vement

Char-coal

Beve-rage

Fodder

Urec Stot 156 25 49 84 34 20 58 53 27 27 4 7 Savg 14.1 4.6 3.1 2.8 2.4 1.8 1.2 1.0 0.8 0.6 0.4 0.2 (lcl) 13.3 4.4 2.8 2.3 2.1 1.6 1.0 0.7 0.6 0.5 0.3 0.2 (ucl) 14.9 4.8 3.4 3.2 2.6 1.9 1.4 1.2 0.9 0.8 0.5 0.3 d-1 5.3 1.4 1.7 7.3 3.0 1.2 1.5 2.0 1.8 2.0 1.2 2.2 (lcl) 4.4 1.3 1.5 4.0 2.5 1.2 1.1 1.4 1.1 1.3 1.0 1.5 (ucl) 6.5 1.5 2.0 42.6 3.9 1.4 2.1 3.0 4.3 4.0 1.5 4.5 d-1csa 2.9 3.2 1.9 6.8 4.1 1.3 7.6 6.0 3.9 2.9 1.3 1.6 (lcl) 2.6 2.9 1.7 5.1 2.9 1.2 3.6 3.2 2.6 1.9 1.2 1.1 (ucl) 3.3 3.7 2.2 10.3 7.3 1.4 n.a. 48.5 8.2 5.3 1.5 2.9 d-1bs 27.7 6.6 10.9 16.8 8.3 5.6 12.5 11.6 6.9 6.9 1.9 2.7 Uadj Stot 156 25 49 84 34 20 58 53 27 27 4 7 Savg 15.2 4.7 3.9 3.0 2.8 1.9 1.2 1.0 1.0 0.6 0.4 0.3 (lcl) 14.4 4.5 3.5 2.6 2.6 1.8 1.0 0.8 0.8 0.5 0.3 0.2 (ucl) 16.0 4.9 4.2 3.4 3.1 2.0 1.5 1.3 1.1 0.8 0.5 0.3 d-1 5.6 1.4 1.6 7.7 3.2 1.3 1.5 2.0 1.9 2.0 1.2 2.4 (lcl) 4.7 1.3 1.4 4.2 2.6 1.2 1.2 1.5 1.2 1.3 1.0 1.6 (ucl) 6.8 1.5 1.8 44.8 4.2 1.4 2.1 3.0 4.7 4.0 1.5 4.9 d-1csa 3.1 3.3 1.6 7.3 2.8 1.3 7.7 6.4 3.0 2.9 1.3 1.7 (lcl) 2.8 2.9 1.5 5.4 2.2 1.3 3.7 3.4 2.1 1.9 1.2 1.1 (ucl) 3.5 3.8 1.7 11.0 3.8 1.4 n.a. 56.5 5.4 5.3 1.5 3.5 d-1bs 27.7 6.6 10.9 16.8 8.3 5.6 12.5 11.6 6.9 6.9 1.9 2.7 Umain Stot 70 25 40 47 22 6 49 30 12 10 4 3 Savg 3.0 4.5 1.4 0.9 2.0 1.4 0.8 0.4 0.4 0.1 0.4 0.1 (lcl) 2.6 4.3 1.2 0.7 1.9 1.3 0.6 0.3 0.4 0.0 0.3 0.1 (ucl) 3.3 4.7 1.7 1.0 2.2 1.5 1.0 0.5 0.5 0.1 0.5 0.2 d-1 3.9 1.4 2.4 3.2 2.9 1.2 1.4 2.0 1.7 1.1 1.3 2.8 (lcl) 3.0 1.3 1.7 1.4 2.4 1.1 1.0 1.2 1.2 1.0 1.0 1.4 (ucl) 5.4 1.5 3.8 n.a. 3.7 1.3 2.7 8.9 3.2 1.3 1.9 173.7 d-1csa 2.9 3.1 5.2 4.8 3.8 1.2 4.5 3.2 1.9 1.3 1.3 1.7 (lcl) 2.2 2.7 2.9 2.6 2.5 1.2 1.8 1.4 1.3 1.0 1.2 1.2 (ucl) 4.1 3.6 21.3 29.7 7.9 1.3 n.a. n.a. 3.5 2.0 1.5 3.0 d-1bs 14.5 6.6 9.3 10.6 6.0 2.4 10.9 7.5 3.9 3.4 1.9 1.6 Uall Stot 156 25 49 84 34 20 58 53 27 27 4 7 Savg 15.7 4.7 8.1 13.9 9.2 7.0 6.8 8.0 5.8 7.5 0.4 2.3 (lcl) 14.9 4.5 7.6 13.2 8.8 6.7 6.4 7.6 5.5 7.1 0.3 2.1 (ucl) 16.6 5.0 8.6 14.6 9.6 7.4 7.3 8.5 6.1 7.8 0.5 2.4 d-1 5.9 1.4 2.0 3.8 4.4 2.6 2.0 4.7 2.4 2.3 1.3 1.5 (lcl) 5.0 1.3 1.8 3.3 3.8 2.2 1.8 3.7 1.7 2.0 1.0 1.3 (ucl) 7.3 1.5 2.3 4.5 5.3 3.0 2.3 6.7 3.8 2.7 1.9 1.7 d-1csa 2.9 3.2 1.9 6.8 4.1 1.3 7.6 6.0 3.9 2.9 1.3 1.6 (lcl) 2.6 2.9 1.7 5.1 2.9 1.2 3.6 3.2 2.6 1.9 1.2 1.1 (ucl) 3.3 3.7 2.2 10.3 7.3 1.4 n.a. 48.5 8.2 5.3 1.5 2.9 d-1bs 27.7 6.6 10.9 16.8 8.3 5.6 12.5 11.6 6.9 6.9 1.9 2.7 Gamma pool

A C B A C C B B C C D D

Stot: total number of species in the survey; Savg: average number of species per farm (including farms where use was not mentioned); d-1: inverse Berger-Parker index based on species abundance; d-1csa: inverse Berger-Parker index based on species cross sectional area; d-1bs: inverse Berger Parker index based on the broken stick distribution; lcl: lower confidence limit (p=0.95); ucl: upper confidence limit (p=0.95); gamma pool: pooling of use-groups based on Stot


86

Table 5.3. Density and abundance of the 12 most frequent use-groups based on various group definitions (see methods) Use Statistic Use-groups

Group Defi-nition

Fire-wood

Fruit Timber Shade Boun-dary

demar-cation

Con-struc-tion

Medi-cinal

Orna-mental

Soil fertility impro-vement

Char-coal

Beve-rage

Fodder

Urec Navg 448 46 101 23 201 106 13 26 57 17 95 2.1 (lcl) 389 37 81 14 176 87 5 16 17 8 48 0.9 (ucl) 506 55 122 31 225 125 21 36 96 26 143 3.4 Davg 1090 96 245 59 583 225 25 100 91 33 211 8.5 (lcl) 932 74 177 34 467 179 11 50 39 17 118 0.3 (ucl) 1248 118 314 83 699 270 39 149 143 49 303 16.7 SDtot 7.0 3.9 5.0 0.7 17.2 11.2 0.4 1.9 3.4 1.2 52.6 1.2 SDtot-d 5.8 1.4 2.3 0.6 14.6 2.4 0.2 0.9 2.3 0.6 20.0 0.9 Uadj Navg 477 46 139 24 213 109 13 26 59 17 95 2.3 (lcl) 416 37 117 15 188 90 5 16 19 8 48 1.0 (ucl) 537 55 161 33 238 128 21 36 99 26 143 3.5

Davg 1153 97 356 61 603 231 25 100 97 33 211 8.7 (lcl) 990 75 268 36 488 185 11 50 45 17 118 0.5 (ucl) 1316 119 443 85 718 276 40 150 150 49 303 16.8 SDtot 7.4 3.9 7.3 0.7 17.7 11.5 0.4 1.9 3.6 1.2 52.6 1.2 SDtot-d 6.2 1.4 3.7 0.6 13.7 2.4 0.2 0.9 2.5 0.6 20.0 1.0 Umain Navg 32 45 12 3.9 191 99 6.9 1.9 11 2.1 95 0.7 (lcl) 25 35 8 1.2 166 80 -0.1 0.5 5 0.0 48 0.2 (ucl) 40 54 16 6.5 215 117 13.9 3.3 17 4.1 143 1.2 Davg 58 94 22 8.9 561 214 14 6.6 22.6 3.1 211 2.2 (lcl) 44 73 14 2.7 446 168 2 0.5 9.9 0.6 118 0.7 (ucl) 71 116 30 15.1 676 259 26 12.6 35.3 5.6 303 3.8 SDtot 0.8 3.8 0.6 0.2 25.5 35.6 0.3 0.2 1.9 0.3 52.6 0.7 SDtot-d 0.7 1.4 0.4 0.1 22.4 8.2 0.1 0.1 1.0 0.03 20.0 0.7 Uall Navg 503 46 171 319 379 220 174 184 174 199 95 47 (lcl) 441 37 148 281 345 190 147 161 125 173 48 38 (ucl) 564 55 195 358 412 249 201 207 223 225 143 56 Davg 1217 97 418 752 948 504 360 485 383 465 211 115 (lcl) 1051 75 329 631 816 406 299 388 286 371 118 86 (ucl) 1383 119 508 874 1080 601 422 582 479 559 303 144 SDtot 7.8 3.9 8.5 9.0 27.9 25.2 6.2 9.1 14.2 17.2 52.6 16.4 SDtot-d 6.6 1.4 4.7 6.7 22.9 16.4 2.9 6.8 8.8 10.5 20.0 6.6 Navg: average number of trees per farm; Davg: average density of trees (ha-1); SDtot: average species density (ha-1); SDtot-d: average species density when the dominant species is removed (ha-1); lcl: lower confidence limit (p=0.95); ucl: upper confidence limit (p=0.95)

CHAPTER 5

87

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Number of uses per species

0

5

10

15

20

25

30

35

40

45

Nu

mb

er o

f sp

ecie

s

0

5

10

15

20

25

Per

cen

tag

e o

f sp

ecie

s

Figure 5.1. Frequency distribution of species with a specific total number of uses

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14

Total number of uses

0

1

2

3

4

0

1

2

3

4

Ave

rage

num

ber

of u

ses

regression including firewood

regression excluding firewood

Figure 5.2. The average number of uses per species versus the total number of uses per species. Sizes of circles correspond to the

number of species with the same characteristics (the biggest circle corresponds to 35 species). The lines correspond to robust linear regression fits and 95% confidence intervals (8% of variation explained including firewood; 9% excluding firewood)


88

0 20 40 60 80 100 120 140 160 180 200

Species frequency (number of households)

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

0

1

2

3

4

5

6

7

8

9

10

11

12

13

14

To

tal n

um

ber

of u

ses

per

sp

ecie

s

Figure 5.3. The total number of uses per species over species frequency. The size of the circle corresponds to the number of

species with the same use and frequency characteristics (largest circle corresponds to 30 individual species). The lines correspond to robust linear regression fits and 95% confidence intervals (36% of variation explained)

0 20 40 60 80 100 120 140 160 180 200

Species frequency (number of households)

0

1

2

3

4

0

1

2

3

4

Ave

rag

e n

um

ber

of

use

s

Figure 5.4. Average number of uses per species per household over species frequency. The size of the circle corresponds to the number of species with the same use and frequency characteristics (largest circle corresponds to 30 individual species). The lines

correspond to robust linear regression fits and 95% confidence intervals (3% of variation explained)

CHAPTER 5

89

0 5 10 15 20 25 30 35 40 45 50

Number of species per household

0

5

10

15

20

25

30

0

5

10

15

20

25

30N

um

ber

of u

ses

per

ho

use

ho

ld

Figure 5.5. The number of uses per household versus the number of species per household. Circle size corresponds to the

number of households with the same characteristics (largest circle corresponds to 7 households). The lines correspond to robust linear regression fits and 95% confidence intervals (39% of variation explained)

0 5 10 15 20 25 30 35 40 45 50


0

1

2

3

0

1

2

3

Ave

rag

e n

um

ber

of s

pec

ies

per

use

Figure 5.6. The average number of species per use per household compared to household species richness. Circle size

corresponds to the number of households with the same characteristics (largest circle corresponds to 7 households). The lines correspond to robust linear regression fits and 95% confidence intervals (48% of variation explained)


90

1 2 3 4 5 6 7 8 15 16 18 23 26 28 47 62 80 96 112

136

166

196

197

201

Number of households where use group was recorded

0

5

10

15

20

Nu

mb

er o

f use

gro

up

s

Not frequent use groups

0

5

10

15

20

25

30

Per

cen

tag

e o

f use

gro

up

s

Frequent use groups

Figure 5.7. Frequency of use-groups expressed as number of use-groups recorded at a specific number of farms

0 5 10 15 20 25 30

Number of uses per household

0

5

10

15

20

25

30

35

Nu

mb

er o

f h

ou

seh

old

s

0

5

10

15P

erce

nta

ge

of

ho

use

ho

lds

Figure 5.8. Frequency of farms with a certain number of uses

CHAPTER 6 THE STUDY OF RANDOM AND PROXIMITY-BASED TREE SPECIES DIVERSITY ON FARMS IN WESTERN KENYA

USING EXACT SPECIES ACCUMULATION CURVES R KINDT, P VAN DAMME & AJ SIMONS SUBMITTED TO JOURNAL OF APPLIED ECOLOGY

With the objective of promoting diversification of on-farm tree species composition through domestication, a survey was conducted in western Kenya involving a complete tree census, tree measurement and collection of ethnobotanical information in 201 small-scale farms (0.1-2.2 ha). Exact values of average expected species richness at each scale from farm to complete survey were calculated for random and proximity-based (randomly adding farms within the same village first) species accumulation based on the hypergeometric distribution. Alpha and beta diversity statistics were calculated for the 12 most frequent use-groups of trees using the SN=cNz(N) model. Monte-Carlo analysis involving 100,000 random site sequences determined ranges in species richness obtained for subsamples, and proved that random distribution of species over farms would result in higher proximity-based species richness at intermediate scales. Important differences detected between the use-groups in alpha and beta diversity can help in targeting agroecosystem diversification efforts in an often tree-rich but species-poor landscape.

SPECIES ACCUMULATION CURVES

92

6.1. Introduction

One of the purposes of agroforestry tree domestication is enhancement of stability and productivity of agro-ecosystems by diversifying on-farm tree species composition (presence and abundance). Diversification and intensification of land use through domestication of agroforestry trees is one of the three pillars of ICRAF’s research (ICRAF 1997; ICRAF 2000). In this chapter, a methodology of studying tree species diversity at various scales (ranging from the individual farm to the complete survey) is presented and discussed.

Ecological experiments and models have been able to show that there is a positive but conditional relationship between species diversity, and ecosystem stability and productivity (for example Naeem et al. 1994; Tilman 1996; Tilman et al. 1997a, 1997c; Hector et al. 1999; Yachi & Loreau 1999; Ives et al. 2000; Nijs & Roy 2000). Findings of models and experiments involving crop mixtures in agroecosystems showed the same positive relationship (Dover & Talbot 1987; Van Noordwijk et al. 1994; Swift et al. 1996; Trenbath 1999; Van Noordwijk & Ong 1999). However, whether higher productivity associated with higher diversity resulted from overyielding (a synergistic relationship among species so that productivity of the mixture is higher than any species growing in isolation), or from greater chances of sampling more productive species in more species-rich mixtures (a phenomenon described as the sampling effect), remains controversial (Huston et al. 2000; Kaiser 2000a, 2000b; Purvis & Hector 2000). In the latter case, each combination of species that includes the most productive species would have the highest productivity, including monocultures. McCann (2000) mentioned that the condition of generally weak bioenergetic consumer-resource interactions needed to be added so that diversity could enhance stability.

The models mentioned above (Tilman et al. 1997b; Yachi & Loreau 1999; Ives et al. 2000; Nijs & Roy 2000) have shown that diversity in traits of species and environmental heterogeneity are explanatory factors for diversity-productivity and diversity-stability relationships. Heterogeneity in species characteristics results in saturating effects of increased diversity because randomly added species will each differ less from species already present. The studies (e.g. Tilman 1996 for stability, Hector et al. 1999 for productivity) demonstrated that diversity only explains a small amount of variation in the data so that some species-poor ecosystems can still have relatively high stability or productivity and vice versa. This variation probably also results from differences in heterogeneity in species traits at fixed species richness. Because of saturating effects of diversity on productivity and stability, the approach taken in this chapter was identification of use-groups with low diversity, since, on average, more important effects may be expected from increasing diversity within groups with lower diversity levels.

Most species richness data refer to natural or semi-natural ecosystems. We focus on farm-level tree diversity in landscapes where farmer management dominates the presence of trees, but spontaneous regeneration of trees still occurs. Diversity enhancement is considered at all levels from farm (alpha diversity) to complete survey level as, on average, stability and productivity enhancement by diversification can be expected at each level where heterogeneity of species traits and environmental characteristics occurs.

To accurately describe changes in species richness (S) with increasing farm number (N), the expected average species (SN) accumulation curves were calculated from the hypergeometric distribution. S obtained from random additions of farms was compared to proximity-based (randomly adding farms within the same village first) S to explore ways of enhancing richness at intermediate scales.

CHAPTER 6

93

6.2. Material and methods

Formulation of use-groups (4.1), species accumulation curves (4.2.2), randomization tests on the influence of sample size on species richness (4.2.3)

6.3. Results

6.3.1. Form of Species Accumulation Curves and Values of Associated Parameters

Species accumulation curves for all species and the 12 most common use-groups are presented in figure 6.1. In figure 6.1, five different pools of species groups can be characterised in terms of Stot = S201 (i) highest richness, including firewood and all species (pool A); (ii) high richness, only including shade (pool B); (iii) medium richness, including medicine, ornamental and timber (pool C); (iv) moderate richness, including fruit, construction, soil fertility, boundary demarcation and charcoal (pool D); and (v) low richness, including fodder and beverage (pool E).

Figure 6.2 illustrates that random beta diversity (zN) has decreasing values at increasing scale (N 2? 201), except for construction where values were increasing. Within the five pools, individual groups show distinctive differences in random alpha (c, table 6.1) and beta diversity (table 6.1, figure 6.2). Random beta diversity of groups of the same pool had the same rank order (smaller or larger values) for all scales, except for construction and fruit in pool D. Within pool A, firewood had smallest alpha and beta diversity. In pools C, D, and E (again with the exception of construction), groups with smaller alpha diversity had larger beta diversity. Figure 6.2 shows that the decrease of zN is non-linear compared to ln(N), whereas comparing zN to N (figure not provided) showed stronger non-linear trends.

Alpha diversity of many use-groups is quite low. Only five (six when also including construction with c=1.935) use-groups have buffering (redundancy) against loss of one species on the average farm by having at least one alternative species.

Proximity-based SN (figure 6.1) and beta diversity (table 6.1, figure 6.2) followed a more complex pattern than random SN and beta diversity, since few new species were added when only a few sites could be sampled from the same village and vice versa. Obviously, alpha diversity and Stot were the same for random and proximity-based sampling. At intermediate scale, proximity-based beta diversity was lower than random beta diversity, even for beverage at N=50 (not clearly shown in the table and figures) with proximity-based beta diversity = 2.9950229 < 2.9950248 = random beta diversity. Generally, values of proximity-based beta diversity decreased with increasing scale. However, shade, charcoal, fruit, fodder and beverage were groups where z50 < z100.

6.3.2. Effects of Sample Size on Species Richness

Table 6.2 shows that 100,000 randomizations yielded random S100 and S50 that were very close to the exact calculations based on the hypergeometric distribution. Table 6.2 further shows the large ranges obtained for random S – even when investigating the 95% confidence intervals for subsamples of 100 sites. Confidence intervals for 100 site subsamples, however, allow discriminating the same pools determined above from Stot, with only the intervals of boundary demarcation of pool D overlapping with those of groups belonging to pool C.

S50 was similar for random and randomly-clustered sampling – which was expected for the average of 50 randomly selected farms and the average for four random subsamples of 50 farms. Intervals for proximity-based sampling were however substantially smaller, as the ranges of randomly-clustered richness were within the confidence interval of random richness. The only


94

overlaps in ranges of randomly-clustered richness between pools occurred between boundary demarcation and two groups of pool C.

The ranges obtained for randomly-clustered richness were in general above the proximity-based richness. The lowest ratio of proximity-based S50 to randomly-clustered S50, 0.77, corresponded to boundary demarcation, the highest ratio, 0.96, to construction. Proximity-based S50 was higher than the minimum values of randomly-clustered richness for medicine, charcoal, fruit, construction, and beverage. The former value only occurred within the confidence interval of charcoal and construction, although for charcoal the value may not belong to the actual interval since we calculated the interval in a conservative manner. Not observing the actual proximity-based richness in 100,000 random sequences is not surprising as we calculated that approximately 7.7 E+117 arrangements of 4 × 50 sites could be created from 201 samples.

For total S of the use-groups (Stot= S201), no estimations of ranges are available as only one sample of 201 farms was taken. Therefore, the effect of sampling 201 farms on the accumulation curve could not be estimated.

6.4. Discussion

Crawley & Harral (2001) and Rosenzweig (2001) cite several reasons why species accumulate in number with increasing sample size, including sampling effects for rarer species, spatial clumping and segregation, habitat effects, higher speciation and lower extinction rates, and lower chances of area-wide anomalies. Ritchie & Olff (1999) describe that species-area patterns will emerge if larger species only detect resources in larger patches within their habitat, but tolerate lower concentrations of resources than smaller species. Hanski & Gyllenberg (1997) demonstrated that interspecific differences in abundance and metapopulation dynamics result in species-area and distribution-abundance curves. Using the hypergeometric distribution to calculate the results from random sampling, we calculated species-area curves that occur when not every species is present on each site. The species accumulation patterns (and distribution-abundance patterns, Kindt, unpublished results) were also the result from anthropogenic effects, especially through decisions of individual farming households to obtain germplasm or to protect species.

Accumulation curves presented in this article show how SN increases (or declines when investigating for N 201? 1) when farms are randomly added (removed), when addition (removal) is not based on their proximity, or when all farms belonging to the same village are sampled first. The random site accumulation approach used here, however, also included an element of area contiguity because each site was a contiguous area. Hanski (1998) indicates that habitat destruction is likely to be a combination of loosing a contiguous region of the landscape, loosing a complete habitat fragment or reduction in area of an individual fragment. For each fragmentation pattern, an accumulation curve could be constructed, therefore many accumulation curves could be constructed for the same area. However, every accumulation curve will converge towards the same S1 and Stot.

Using the hypergeometric distribution to calculate random SN at intermediate scales yielded exact results, what contrasts to calculating these values by random additions of sites (e.g. the Monte-Carlo approach in this chapter; Colwell 1997; Rennolls & Laumonier 1999). Since the calculation by means of the hypergeometic distribution only used information on the farm/site-frequency of each species, it means that clustering of species within sites has no influence on the average species accumulation. This contradicts the statement by Gotelli & Colwell (2001) that sample-based accumulation curves can not be obtained theoretically as they also depended on the spatial distribution of individuals. The spatial distribution will, however, influence the range of values observed for a particular sample size (e.g. 50 farms), as some random sequences will obtain 50

CHAPTER 6

95

farms that contain many different species, while some other random sequences will obtain 50 farms that share many species.

Using the hypergeometric distribution to calculate proximity-based SN also contrasts to the practice of calculating the average richness of consecutive bisections of an area (e.g. Plotkin et al. 2000). Calculating SN by the second approach does not consider other contiguous subsamples of the same size that could be formed from combinations of the smallest sample size that overlap bisections. However, since we did not consider farm proximity within villages, most additions within villages were not expected to result in contiguous areas – because of the small farm and village sizes, we did not expect distance among farms within the same village to have a strong effect on beta diversity. An approach that would first determine combinations of sites resulting in contiguous areas, followed by calculating SN, could generate figures that are more accurate. These methods could either determine all combinations resulting in contiguous areas of the same size, followed by calculating SN through the hypergeometric distribution, or approximate SN through randomisations. Calculation of randomly-clustered SN as presented in this paper could subsequently investigate how frequently the observed proximity-based richness was observed.

After equalling parameter c of the SN= NzcN model to S1, we demonstrated that a constant parameter zN could not accurately describe species accumulation in the groups we differentiated, which implies that beta diversity was not constant with scale. Moreover, we observed non-linear relationships between zN and N or ln(N). Plotkin et al. (2000) observed that log-log species-area curves for five 50 ha tropical forest plots did not have a constant slope, and therefore expanded the S=cAz model with a polynomial term. Other studies also showed overestimation of slopes for large areas, and underestimation for small areas (May & Stumpf 2000, Crawley & Harral 2001) with the classical power function model. Lomolino (2000) argued that the species-area relationship could be sigmoidal as the species richness of relatively small islands species seemed to be independent of island size, while an asymptote of species richness was observed for the largest islands. Losos & Schluter (2000) showed that the slope increased significantly beyond a threshold island size. Harte et al. (1999) demonstrated that self-similarity of species distributions within an area corresponded to a constant z value, but also mentioned that it is extremely unlikely that the species-area relationship would have constant z values across scales. Calculations (Richard Coe, ICRAF, unpublished results) demonstrated that the two-parameter species accumulation model actually represents rare cases of random or proximity-based species accumulation associated with very even species abundance distribution.

The SN= NzcN (N: accumulated number of sites) and SA= AzcA (A: accumulated area) models are conceptually similar. By calculating the average size of one site (A1), SA for any size of area AN = N A1 can be calculated by N-based models by determining N. A possible criticism on calculation of species accumulation models in this paper could be that farm sizes vary, and that therefore random runs of site additions involved random additions of various farm sizes. A similar criticism could be that farm shapes differ, while patch shape influences S per unit area (Condit et al. 1996; Harte et al. 1999; Maddux et al. 1999). Differences in diversity of various on-farm niches such as external boundaries, woodlots and homestead areas and varying proportions of these niches associated with shape differences (Kindt, unpublished data) could amplify shape effects on diversity. Interpretation of accumulation curves should however not pose any problems, as long as one realises that accumulation curves resulted from double averaging. The horizontal axis represents accumulation of the average farm (size and shape), whereas the vertical axis shows associated SN. Accumulation curves thus still present the expected average species richness for N farms.

In analogy to Fonseca & Ganade (2001) who calculated how functional groups are lost when species become randomly extinct, our methodology could be used to calculate how functional groups are lost when an area fragmentises at random. Calculating species accumulation patterns


96

from the hypergeometric distribution can also prove useful to separate effects of area reduction on alpha and beta diversity. Gonzalez (2000) for example differentiated between species loss through fragmentation by random sampling from the formerly continuous landscape, and subsequent species loss through fragmentation from community relaxation. The second process is a longer-term dynamical process where species that were included in fragmented patches by random sampling, disappear through time as they are sink species (species that do not reproduce enough to prevent their extinction after fragmentation) or because rare area-wide species extinctions occur (Rosenzweig 1999, 2001). The second process could be described by reductions in alpha diversity with time (c(t), t: time), since relaxation can be studied on areas of the same size (i.e. ideally not reflected in changes of the parameter that describes beta diversity). For the first process, we recommend calculating the expected loss in richness from random sampling by the hypergeometric distribution, rather than calculating the expected loss by the classical model with constant parameter z.

Equalling average diversity at one site to alpha diversity may seem arbitrary, while beta diversity is usually not expressed as the exponent z of power function models describing species accumulation. Stot values provide indications of gamma diversity differences of use-groups, although these values are almost certainly underestimates. However, as Loreau (2000) pointed out, there is no precise ecological prescription on the division between scales. This author differentiated between multiplicative (ß=?/a) and additive (ß=?-a) approaches of defining beta diversity. Describing beta diversity as exponent z of species accumulation models anticipates the problems described in Loreau (2000) associated with the scale at which alpha diversity is measured. This description of beta diversity also conforms to the concept of intra-biogeographical-province species accumulation curves (Rosenzweig 1999, 2001), and observations that often much of the structure of local assemblages can be modelled as random draws from regional species pools (Griffiths 1999; Gaston 2000).

Predicting S for a larger area than the sample area (“abundifaction”, in the context of this chapter means extrapolation of accumulation curves) is difficult (Hayek & Buzas 1997 pp. 344-345). Rennols & Laumonier (1999) advocated studying random subsamples of the total and the sample area to be able to extrapolate species-area curves. Plotkin et al. (2000) concluded that their models provided the most accurate framework to predict species richness from small samples. However, they first calculated parameters describing the complete form of species accumulation curves, and then calculated differences in calculations of Stot based on the variation in S for subsets of the area. The equivalent analysis for the method described here would be to first determine zN parameters, and then analyze the influence of parameter c estimation from subsets of data. Plotkin et al. (2000) observed accurate predictions because the five forests they studied mainly differed in parameter c (May & Stumpf 2000), thus mainly differed in alpha diversity. Large ranges in S observed for subsets of 100 and 50 sites as observed in this article caution against extrapolation of accumulation curves based on parameter zi estimations. However, in case Stot of a survey approximates the average that would be obtained by taking various samples of the same survey area size (Stot), then extrapolation of parameters zN (N>Ntot) could allow for more reliable estimations of SN>Ntot.

Species accumulation curves showed differences between use-groups, which allows targeting interventions towards use-groups of lower SN. If stability of production, or productivity, of individual farms would be linked to S at the individual farm level (conditions such as heterogeneity in environmental and species trait characteristics were discussed in the introduction), one target of domestication research and interventions could be to increase alpha diversity of use-groups. Interventions could focus on groups with low alpha diversity (for example, construction, medicine, soil fertility, charcoal, and fodder). Interventions may focus on other groups based on other criteria, but calculated alpha diversity values will still provide benchmark values of diversification.

CHAPTER 6

97

Gamma and beta diversity could provide some answers on how alpha diversity could be improved. For groups with high gamma diversity, diffusion between farmers of germplasm and information on species usage could be promoted. Reasons to promote species that are already present in the landscape could be related to species conservation through use (Kindt & Lengkeek 1999), biosafety precautions, or ecological suitability considerations. For groups with low gamma diversity, interventions can attempt to introduce additional species, or to discover or promote alternative uses of species present. Improving gamma diversity could contribute to improving the stability (or productivity) of agroecosystems at the survey level. Using the same line of thought, approaches that would result in higher beta, but similar alpha and gamma diversity, could also improve ecosystem functioning, as SN for subsets of sites would be larger. Our results show that the average richness observed in the four villages (proximity-based SN) was substantially lower than the richness of random subsamples of the same size (randomly-clustered SN). This finding implies that a more random distribution of existing species over farms within the area would result in higher proximity-based beta diversity and SN, and potentially better agroecosystem functioning.

Whereas this work identifies where to intervene in a tree-rich but often species-poor landscape, diversification should be farmer driven and not merely an exercise in curve shifting. Probably the most important factor to consider is that alpha diversity of individual farms may be quite high for use-groups with low average alpha diversity, and vice versa, so that relevance of diversification may be farm- and use-group dependent. Warner (1993) stated that farmers, confronted with deforestation, were most likely to establish trees for products for which non-tree alternatives were not available. Construction and timber could be such groups, whereas herbs could provide medicine, or residues from crops, firewood. Diversification of the former use-groups may therefore be more relevant. Botanical similarity (family, genus) or similarity in species’ characteristics (e.g. Pachepsky et al. 2001) could be another factor to consider when planning for diversification interventions. Even when targeting use-groups of higher alpha diversity, however, analysis of gamma and beta diversity as presented above can aid in designing interventions seeking to diversify agroecosystems.


98

Table 6.1. Parameter values for N=2, 50, 100, 150 and 201 for the SN= NzcN model for species accumulation curves for all species and species belonging to use-groups. Species group c z201 Random accumulation Proximity-based accumulation z2 z50 z100 z150 z2 z50 z100 z150 All 16.572 0.444 0.624 0.483 0.460 0.450 0.581 0.442 0.448 0.445 Firewood 15.194 0.439 0.618 0.476 0.454 0.445 0.573 0.442 0.447 0.440 Shade 2.970 0.630 0.911 0.706 0.665 0.644 0.884 0.670 0.656 0.641 Medicine 1.249 0.724 0.934 0.773 0.746 0.734 0.916 0.738 0.741 0.730 Ornamental 1.035 0.742 0.946 0.804 0.774 0.756 0.915 0.753 0.754 0.748 Timber 3.856 0.479 0.671 0.516 0.494 0.485 0.625 0.485 0.485 0.482 Boundary demarcation 2.841 0.468 0.694 0.510 0.489 0.477 0.617 0.441 0.463 0.467 Charcoal 0.617 0.713 0.906 0.774 0.745 0.728 0.889 0.753 0.738 0.724 Soil fertility 0.975 0.626 0.735 0.665 0.646 0.634 0.700 0.607 0.620 0.628 Fruit 4.697 0.315 0.414 0.332 0.321 0.317 0.398 0.317 0.314 0.313 Construction 1.935 0.440 0.266 0.402 0.431 0.439 0.261 0.392 0.413 0.433 Fodder 0.264 0.618 0.957 0.731 0.659 0.631 0.908 0.678 0.639 0.623 Beverage 0.368 0.450 0.856 0.491 0.455 0.449 0.680 0.463 0.455 0.449

Table 6.2. Averages (SN), ranges (minimum: Min., maximum: Max.), and lower and upper 95% confidence interval limits (2.5%, 97.5%) for species richness obtained for random (r) and randomly-clustered (rc) subsamples (N: subsample size) obtained from random site sequences. Exact expected averages calculated for random (r) and proximity-based (p) accumulation (Accum.). Species group N Exact calculations Results from 100,000 randomizations Accum. SN Subsamples SN Min. Max. 2.5% 97.5% All 100 r 137.67 r 137.67 111.0 162.0 124.0 150.0 50 r 109.25 r 109.24 81.0 143.0 94.0 124.0 50 p 93.32 rc 109.25 104.4 113.8 106.9 111.5 Firewood 100 r 122.82 r 122.83 98.0 145.0 110.0 134.0 50 r 97.63 r 97.63 72.0 129.0 84.0 111.0 50 p 85.59 rc 97.63 92.7 101.6 95.5 99.6 Shade 100 r 63.38 r 63.38 47.0 78.0 55.0 71.0 50 r 46.73 r 46.72 27.0 66.0 36.0 56.0 50 p 40.86 rc 46.73 42.7 50.0 44.9 48.3 Medicine 100 r 38.83 r 38.84 18.0 57.0 28.0 48.0 50 r 25.58 r 25.57 9.0 47.0 15.0 36.0 50 p 22.41 rc 25.58 22.3 28.1 24.2 26.8 Ornamental 100 r 36.49 r 36.50 20.0 49.0 28.0 44.0 50 r 23.93 r 23.93 8.0 41.0 15.0 33.0 50 p 19.68 rc 23.93 20.6 26.5 22.5 25.3 Timber 100 r 37.57 r 37.57 23.0 48.0 31.0 43.0 50 r 28.89 r 28.89 17.0 43.0 22.0 35.0 50 p 25.71 rc 28.89 26.3 31.3 27.7 30.0 Boundary demarcation 100 r 27.03 r 27.03 18.0 34.0 22.0 31.0 50 r 20.86 r 20.86 10.0 32.0 15.0 26.0 50 p 15.97 rc 20.86 18.3 22.8 19.7 22.0 Charcoal 100 r 19.10 r 19.09 7.0 27.0 13.0 24.0 50 r 12.72 r 12.72 2.0 25.0 6.0 19.0 50 p 11.72 rc 12.72 10.0 14.3 11.7 13.8 Soil fertility 100 r 19.07 r 19.09 8.0 27.0 13.0 24.0 50 r 13.12 r 13.13 3.0 24.0 7.0 19.0 50 p 10.47 rc 13.12 10.8 14.8 12.2 14.0 Fruit 100 r 20.63 r 20.62 14.0 25.0 16.0 24.0 50 r 17.20 r 17.19 10.0 24.0 13.0 21.0 50 p 16.23 rc 17.20 15.5 18.3 16.4 18.0 Construction 100 r 14.10 r 14.10 4.0 20.0 9.0 18.0 50 r 9.36 r 9.36 2.0 18.0 4.0 14.0 50 p 8.98 rc 9.36 7.0 10.8 8.4 10.3 Fodder 100 r 5.49 r 5.49 3.0 7.0 3.0 7.0 50 r 4.57 r 4.57 1.0 7.0 2.0 6.0 50 p 3.74 rc 4.57 3.8 4.8 4.2 4.8 Beverage 100 r 3.00 r 2.99 1.0 4.0 2.0 4.0 50 r 2.50 r 2.50 1.0 4.0 1.0 4.0 50 p 2.25 rc 2.50 2.0 2.6 2.4 2.6

allfirewoodshade

0 25 50 75 100 125 150 175 200

0

25

50

75

100

125

150

175

medicineornamentaltimber

0 25 50 75 100 125 150 175 200

0

10

20

30

40

50

60

boundarycharcoalsoil fert.fruitconstruct

0 25 50 75 100 125 150 175 200

0

5

10

15

20

25

30

35

fodderbeverage

0 25 50 75 100 125 150 175 200

0

1

2

3

4

5

6

7

a b

c d

Figure 6.1. Random species accumulation curves for all species and for the 12 most frequent use-groups in the landscape (horizontal axis: number of farms, vertical axis: number of species). Symbols

(represented for each decade only) and dotted lines correspond to average accumulated species richness, and full lines to average accumulated species richness when farms are sampled in the same village first (proximity-based accumulation). a: species’ pools A and B, b: pool C, c: pool D, d: pool E (see text)

allfirewoodshade

1.0 2.7 7.4 20.1 54.6 148.4 403.4

0.4

0.5

0.6

0.7

0.8

0.9

1.0

medicineornamentaltimber

1.0 2.7 7.4 20.1 54.6 148.4 403.4

0.4

0.5

0.6

0.7

0.8

0.9

1.0

fruitconst.

boundarycharcoalsoil fert.

1.0 2.7 7.4 20.1 54.6 148.4 403.4

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0fodderbeverage

1.0 2.7 7.4 20.1 54.6 148.4 403.4

0.4

0.5

0.6

0.7

0.8

0.9

1.0

a b

c d

Figure 6.2. Changes in parameter zN of the Nz

N cNS =)

model for random species accumulation curves for all species and for the 12 most frequent use-groups in the landscape (horizontal axis: number of farms on ln-scale, vertical axis: parameter zN). Symbols (not all values included) and dotted lines correspond to random species richness, full line to proximity-based species richness, dashed

lines connect the extreme values of z2 and z201. a: species’ pools A and B, b: pool C, c: pool D (symbols: see Fig. 1), d: pool E (see text)

(Page not included)

CHAPTER 7 STUDYING ON-FARM TREE SPECIES RICHNESS AND EVENNESS IN WESTERN KENYA WITH DIVERSITY

PROFILES R KINDT, P VAN DAMME & AJ SIMONS SUBMITTED TO ENVIRONMENTAL AND ECOLOGICAL STATISTICS

Species diversity is a function of the number of species and the evenness in the abundance of the component species. Diversification, therefore, can involve adding new species, or making species more evenly distributed. We derived evenness profiles from the Rényi diversity series, which allowed ranking systems in richness and ranking them in evenness. We applied the methodology to investigate differences in diversity among the main functions of trees on western Kenyan farms. No group had perfectly even distributions. Evenness could especially be enhanced for construction materials, fruit, ornamental, firewood, timber and medicine, which included some of the most species-rich groups of the investigated landscape. When considering only the evenness in the distribution of the dominant species, timber, medicine, fruit and beverage ranked lowest (> 60% of trees belonged to the dominant species of these groups). Interestingly, these were also the groups that are mainly grown by farmers to provide cash through sales. For our data, evenness profiles allowed for easier discrimination of differences in evenness than provided by the alternative method of the Lorenz order. The influence of sample size on Rényi diversity profiles was investigated, which is an expansion of species accumulation curves as also the sample size – evenness relationship can be investigated. Accumulation patterns indicated that most groups had abundance distributions most similar to the log series model. Shade and fodder had an abundance distribution more typical of the log normal model. Since the calculation of (accumulation patterns of) diversity and evenness profiles pose no computational difficulties, we advocate to use these methods to fully study diversity and evenness of systems of interest.

DIVERSITY PROFILES

102

7.1. Introduction

Diversity means different things to different people. Most often in natural or agricultural systems, species counts are provided as the measure of diversity. Continuing this logic, diversification means adding more species. Species diversity, however, is a function of the number of species, and the evenness in distribution of species’ abundances (Magurran 1988 pp. 7-8; Purvis & Hector 2000). Options for diversification can therefore be dissociated into interventions that target species richness, and those that target species’ evenness.

In the realm of agroforestry, underpinning the need for diversification is the desire to enhance the stability and productivity of agroecosystems (ICRAF 1997 pp. 6-7; ICRAF 2000 pp. 24-25). Field experiments and ecological models have shown the positive but conditional relationships between species diversity, and (agro)ecosystem stability and productivity (Kindt et al. – Chapter 2, 4 & 5). The conditions for the relationships are heterogeneity in environmental and species’ characteristics. Adding few individuals of a new species to a system will, therefore, have a smaller effect than adding more individuals that result in a more even abundance distribution of all species composing the system. Similarly, making the abundances of species already present in a system more evenly distributed may have larger effect than adding new species. Diversification with the aim of enhancing ecosystem stability and productivity should therefore seek a balance in increment of species richness and evenness.

We investigated the diversity of various groups of species that contributed to similar service or production functions on farms in western Kenya, as our target was diversification within these functions, and not of overall species diversity. Since heterogeneity in characteristics of species results in saturating effects of increased diversity on ecosystem stability (Holmes 1998), we expect larger effects from increasing diversity within functions (“use-groups”) with lower diversity. Ranking of use-groups in diversity, therefore, allows prioritising their scope for diversification measured as the average expected effect of adding one species.

Diversity ordering techniques were designed to differentiate more diverse from less diverse systems. Tóthmérész (1995) was able to show that the Rényi profile, which we use in this chapter to rank use-groups, is one of the most useful ordering techniques. We expanded the use of the Rényi series to ranking of evenness by dissociating the contributions of species richness and species’ evenness to diversity profile form. Doing so, we could prioritise use-groups for species richness and for species evenness increment. In addition, such approaches provide the opportunity to model effects of replacement, substitution and expansion at fixed and varying tree densities.


Formulation of use-groups (4.1), diversity profiles (4.2.4), accumulation patterns of diversity profiles (4.2.5)

7.3. Results

7.3.1. Diversity Ordering and Examination of Richness and Evenness Contributions

Figure 7.1 shows the Rényi diversity ordering of all species and the twelve most frequent use-groups. By examining the values at scales 0, 1, 2, and 8, species richness and values of Shannon, Simpson, and Berger-Parker diversity indices can be inferred. The many intersections in the figure show the difficulties to order most groups in diversity. Table 7.1 provides parameters describing the pattern of the various accumulation curves. Figure 7.2 shows the Rényi evenness

CHAPTER 7

103

ordering. This figure shows that many groups have intersecting profiles, so that many groups can not be ranked in evenness.

We differentiated between five pools of species groups based on their total species richness (H0): (i) pool A - very high richness, including all species and firewood; (ii) pool B - high richness, including shade; (iii) pool C - medium richness, including medicine, ornamental and timber; (iv) pool D - moderate richness, including fruit, construction, soil fertility, boundary demarcation and charcoal; and (iv) pool E - low richness, including fodder and beverage. Some differences and similarities in diversity and evenness within and between pools can subsequently be detected (figures 7.1 and 7.2).

Groups of pool A and B were more diverse than the other groups. The evenness profile shows that shade (pool B) has a more even distribution than groups of pool A. Diversity profiles of groups of pool C intersected all groups of pool D and E, with the exception of fruit and construction (pool D) and beverage (pool E). The evenness profiles, however, show that ornamental and timber had lower evenness than all groups of pool D, with exception of construction. The evenness profile of medicine intersected the other groups of pool C. All groups of pool C had intersecting evenness profiles, therefore none could be classified as more evenly distributed within the group.

Most groups within pool D could be ranked in diversity, since their diversity profiles did not intersect. Boundary demarcation is the most diverse and evenly distributed group within the pool. Construction is the least diverse group of the pool and fruit the second least diverse. Their evenness profiles intersect however. Charcoal and soil fertility had intersecting diversity and evenness profiles. As their species richness is the same, charcoal can be categorised as the group where the dominant species is more evenly distributed, but where the other species are less evenly distributed. Within pool E, fodder is more diverse and more evenly distributed. This group has the most even distribution of all groups, resulting in intersections with diversity profiles of all groups of pool C and D, with the exception of boundary demarcation.

Values of H8 (figure 7.1 and table 7.1) lower than 0.5 show that the dominant species of timber, medicine, fruit, and beverage contains more than 60 percent of all trees of these use-groups. Values of H8 between 0.5 and 1.5 indicate that the dominant species contains between 22 and 60 percent of all trees for soil fertility, charcoal, ornamental, fodder and boundary demarcation. The dominant species contains a smaller percentage of trees for the other groups. However, values of E8 (figure 7.2) show that the dominant species for all species and firewood are not very evenly distributed. Medicine had the least evenly distributed dominant species.

Values of (H0 - H1)( H0 -H8 )-1 provide an insight in evenness distribution of species within groups when excluding the dominant species. Fodder has, besides the most evenly distributed dominant species, also the most evenly distributed other species. Shade ranks second in distribution of other species. Construction, timber, and ornamental rank last in the distribution of other species.

7.3.2. Effects of Sample Size

Figure 7.3 shows the accumulation patterns for diversity profiles of the various use-groups we distinguished. Several accumulation patterns can be observed. Diversity profile values increased with increasing numbers of sites for all scales for shade and fodder. Diversity profiles reached stable values from 50-100 farms onwards, except for small scales (a<1) for timber, fruit, construction, and beverage. The other groups had an intermediate pattern as Shannon index values (a=1) reached relatively stable values with increasing numbers of sites (see also figure 7.6), while other profile values kept changing. Of these groups, medicine was the only group where profile values at larger scales (a>1) were decreasing for subsamples larger than 100 farms.

DIVERSITY PROFILES

104

Figure 7.4 shows the average profile values and associated 95 percent confidence intervals and ranges for 100 farm subsamples. The figure indicates that, on average, similar profiles are obtained as for the full sample. However, profiles, especially medicine, fodder, and shade, had large confidence intervals. The confidence intervals still allow classifying use-groups in the richness pools that we differentiated above. The only overlapping confidence intervals were for the groups of pool C and boundary demarcation (pool D). Within pools, most confidence intervals of diversity profile values overlap. However, within pool D, boundary demarcation has the most even distribution of other species, while the opposite phenomenon occurs for construction. Fodder is more evenly distributed than beverage in pool E.

Figure 7.3 suggests that extrapolating profile values should model a non-linear change of Ha with increasing farm number as for species richness, except where stable values were obtained. Table 7.1 (H8 ,l and H8 ,u) indicates the ranges that are expected for the average value of H8 when all farms would be sampled in the survey area. Relatively large ranges were even obtained for groups where stable values were obtained at the respective scale (timber, fruit, construction, and beverage). However, these ranges were among the smallest. Large ranges of H8 for the complete sample (table 7.1) corresponded to large ranges of the statistic for 100 farm subsamples (figure 7.4).

7.4. Discussion

7.4.1. Interpretation of Diversity and Evenness Profiles

We calculated the evenness profile as Ha-H0 = ln(Ea). This is an expansion of the decomposition of the Shannon diversity index by Hayek & Buzas (1997 eq. 14.1) as:

H = ln(S) + ln(E) = ln(S) + ln(eH/S).

Rousseau et al. (1999) used the Lorenz order to determine differences in evenness between systems. The Lorenz curve is calculated by connecting points representing the cumulative proportion of species against the cumulative proportion of abundances, where species abundances are arranged in ascending order. If some values of the Lorenz curve of system a are greater and none are smaller than those of system b, then system a is more evenly distributed. If the Lorenz curves intersect however, then systems are non-comparable. We provided the Lorenz curves for all groups in figure 7.5. Comparing the curves of figure 7.2 and 7.5 shows that a similar ordering of groups in terms of evenness is obtained: fodder is the most diverse, while most other groups have intersecting profiles (for instance boundary demarcation and shade or medicine). It can be demonstrated numerically that systems with the same Lorenz curve will have the same evenness profile, and that this phenomenon can only occur for systems with nxS (nx=1,2,3... and S fixed) species each.

We interpreted

ln(E8 ) = H8 - H0

as the contribution the dominant species to evenness. By substituting

H8 - H0 = ln(p-1) - ln(S) = -ln(Sp),

it can be demonstrated that for two systems with the same E8 ,

Sa pa = Sb pb.

The same condition can be deducted for the last sections of Lorenz curves of two systems (connecting the cumulative proportion of Stot-1 species to (1,1)) to reflect the same evenness. Where Sapa > Sbpb, or E8 of system a is smaller, one of the conditions for system a to be less even

CHAPTER 7

105

according to the Lorenz or Rényi evenness order is fulfilled. For instance, figures 7.2 and 7.5 show that beverage has a less evenly distributed dominant species than fodder.

Rousseau et al. (1999) recommended using the k-dominance plot ordering method, a method of which a modified version also performed well in the analyses conducted by Tóthmérész (1995), but they did not consider the Rényi series. An advantage of the Rényi profiles over (modified) k-dominance curves is that they allow for discrimination between species richness (scale 0), evenness (decline of the profile from H0 with increasing scale, ln(Ea)) and the influence of the dominant species on evenness (value at scale 8) of various systems by comparing values on the same Ha axis. The logarithm in the formula makes visual comparisons of systems with many species easier. For instance, intersections can be distinguished easier in figure 7.2 than in figure 7.5. Another advantage of the Rényi profiles is that the values of the widely used Shannon and Simpson diversity indices are provided, what facilitates comparisons with other studies that used these indices. Because diversity is represented on a single axis versus fixed scales, it is also easier to show the effects of sample size on diversity by plotting sample size on a third axis as in figure 7.3. We therefore recommend using the Rényi series to compare richness and evenness of various systems, especially if the analysis includes investigation of sample size effects.

7.4.2. Using Only One Diversity Index

The many intersections in diversity profiles have the result that ranking of groups based on a single diversity index will often be influenced by the relative contributions of evenness and richness in the calculation of the index. For example, using the Shannon index will rank medicine as more diverse, and ornamental as less diverse, while usage of the Simpson index would result in the reverse order. Fodder and medicine have a similar Shannon index value, but fodder has fewer species but larger reciprocal Simpson and Berger-Parker indices. Where diversity profiles intersect, it is not possible to rank one system as more diverse, and using a single diversity index will therefore result in incorrect conclusions.

Because evenness profiles may intersect, calculating evenness based on the Shannon diversity index as J=H/ln(S) (Magurran 1988 eq. 2.22; Hayek & Buzas 1997 eq. 13.9) or as E = eH/S (Hayek & Buzas 1997 eq. 13.10) and ranking groups on J or E values will result in similar inaccuracies as ranking these groups in diversity based on a single diversity index. The interested reader may, for example, investigate evenness profiles for one system with 4 species of abundance 10, 6, 4 and 1 with another system with 8 species of abundance 50, 13, 12, 12, 12, 12, 11 and 9, which have similar Shannon evenness, but intersecting evenness profiles. For practical purposes (see below), when systems with lowest comparative evenness need to be selected, the decomposition of the Shannon index may prove sufficient. However, since the calculation of evenness profiles can be achieved easily (moreover, since software was developed that directly provides this information from farms × species matrices) we would still advocate to use the evenness profile for reasons of complete characterisation of evenness. Information on the evenness of the dominant species, as provided by an evenness profile, also offers an interesting statistic.

The decomposition of the diversity profile into Ha = H0 + ln(Ea) explains that Shannon and Simpson indices should be perfectly correlated for systems that have the same evenness profile. Magnussen and Boyle (1995) found strong linear correlations between both indices (0.98) based on mathematical modeling of species’ distributions. As the models they used actually described the evenness in species distribution, this high correlation was to be expected.

DIVERSITY PROFILES

106

7.4.3. Effects of Sample Size

Magurran (1988, p. 72) indicates that a Rényi series value with larger scale parameter value has reduced sensitivity to sample size. Gimaret-Carpentier et al. (1998) observed that the Simpson index (a=2) reached stable values at lower sample sizes compared to the Shannon index (a=1). We observed these patterns for those groups where stable values were obtained at all larger scales (timber, fruit, construction, and beverage). Shade and fodder, however, showed the other extreme with increasing profile values with increasing sample size, and not asymptotic values.

Hayek & Buzas (1997 pp. 378-390), based on May (1975), proposed to investigate accumulation patterns of H (figure 7.6), ln(E) and ln(E)/ln(S) (figure 7.7) to choose the best abundance distribution model (“SHE analysis”) as distributions often converge. Reaching asymptotic values for these statistics with increasing sample size would indicate that the species abundance distribution corresponds to respectively a log series, a broken stick (ln(E)˜ -0.42 for S > 30) or log normal distribution. We could, therefore, conclude that most groups (and especially timber, fruit, construction and beverage) have abundance distributions more typical of the log series distribution (constant H), while the distributions of shade and fodder corresponded more to a log normal distribution (constant ln(E)/ln(S)). Possibly, some log series patterns resulted from not sampling all species. May (1975) describes that less complete samples often conform to the log series distribution, while samples with a large number of species are expected to be log-normally distributed (as a result from the Central Limit Theorem). Therefore, we can not predict whether larger sample sizes would have resulted in a smaller number of groups with patterns typical of the log series distribution than we observed here.

As Rényi profile accumulation patterns can be calculated for various abundance distribution models, we suggest to expand the SHE analysis of Hayek & Buzas (1997 pp. 378-390) to the complete diversity profile, as in figure 7.3. For example, for a log series (and a geometric series), the Simpson and Berger-Parker indices are also expected to reach constant values (May 1975). This was a phenomenon that we could observe most clearly (specific figures not included, but see figure 7.3) for the groups with the clearest asymptotic H pattern. Analysis with simulated datasets showed an asymptotic pattern of Ha=1 for a log series distribution that was not observed for a lognormal distribution (Kindt unpublished data). Since asymptotic H1 and H2 are only expected for the log series distribution, the findings of Gimaret-Carpentier et al. (1998) most likely correspond to studies of systems more conform to this distribution.

4.4. Designing Interventions Based on Diversity Profiles

Analysis of diversity profiles provided insights in diversity structure of species groups. The value of using diversity profiles lays not only in determining which groups are more diverse. Diversity profiles also allow discrimination between richness and evenness contributions to diversity. If the intervention would attempt to increase diversity for two groups with intersecting profiles, then for one group improvement of richness could be attempted, while improvement of evenness would be the target for the other group.

The lack of evenness in the distribution of the dominant species and the many steep decreases in profiles with increasing scale parameter value showed that diversity could be increased substantially in many use-groups by targeting evenness. No group had perfectly evenly distributed species. The log series distribution represents a substantially less even distribution than associated with the log normal distribution (May 1975). Evenness increment could be achieved by encouraging farmers to establish trees in more even numbers or by more species-even germplasm distribution. In combination with efforts to improve alpha diversity (the average species richness on an individual farm) in the landscape, this suggests for promotion of less frequent species both in terms of farm frequency as in terms of total abundance. The analysis shows use-groups with

CHAPTER 7

107

steep diversity and evenness profiles where such diversity improvements would be most useful: i.e. the construction, fruit, ornamental (although this group will probably not constitute a priority to farmers), firewood, timber and medicine groups. When considering the frequency of the dominant species only (not how evenly this species is distributed), timber, medicine, fruit and beverage are the groups with frequencies larger than 60 percent. Interestingly, these are also the use-groups that could be categorised as providing more cash income to farmers (‘high value trees’ that could be selected to be of higher priority for domestication for this reason).

Returning to our objective of diversification to stabilise or increase productivity, it may not be necessary to seek perfectly even distribution in species’ abundances. Since species diversity effects result from spatio-temporal environmental heterogeneity and differences in species characteristics relating to their performance within the environmental matrix, the most stable or productive system will result from optimal matching of richness and evenness in species characteristics with richness and evenness of environmental characteristics. However, including features such as potential diseases or market collapses into the environment in which tree species are grown on farms shows the risks associated with growing the best performing species in case a single species would be most suited to the major part of the spatio-temporal physical environment.

One of the factors that we did not include in our analysis was difference in farm abundance among use-groups and among farms. Some species may be more farm size dependent in their abundance, for instance species used to improve soil fertility, while others may be less dependent, for instance species used to shade the homestead area. For use-groups with larger farm size dependence, recommendations for diversification could include farm size differences as small farms may not offer great scope for enhancement of evenness.

7.5. Conclusions

The analysis of diversity did not include all aspects that could influence decisions on alterations in tree species composition on farms. Kindt et al. (Chapter 4) listed some potential aspects. These included potential differences in importance attributed by farmers to different groups so that diversification of a more important but more diverse group could be given priority over that of a less diverse but also less important group. Another aspect was possible differentiation among farms in alpha diversity, so that farms with low diversity for a specific use-group could be targeted, rather than only targeting groups with lower diversity at the survey level. However, the methods presented here provide meaningful insights in relationships between richness, evenness, and sample size of use-groups. They thus provide accurate guidance for attempts in alterations of these characteristics. Obviously, they allow also for detailed monitoring of the impact of interventions of these characteristics by providing a baseline to compare diversity before and after interventions.

DIVERSITY PROFILES

108

Table 7.1. Parameters of the diversity profiles of the most frequent use-groups. H1 was calculated directly as the Shannon diversity index. Confidence interval for H8 (H8 ,l : lower limit; H8 ,u : upper limit) is based on the confidence interval for the dominant species. Use-group H0 H8 H0 - H8 (H0 – H1)/( H0 – H8 ) H8 ;l H8 ;u All 5.16 1.78 3.39 0.70 1.61 1.98 Firewood 5.05 1.71 3.33 0.71 1.55 1.92 Shade 4.43 2.04 2.39 0.60 1.43 3.80 Medicine 4.06 0.41 3.65 0.68 0.15 0.76 Ornamental 3.97 0.68 3.29 0.80 0.38 1.11 Timber 3.89 0.48 3.41 0.82 0.36 0.61 Boundary 3.53 1.17 2.35 0.67 0.97 1.43 Charcoal 3.30 0.70 2.60 0.77 0.30 1.39 Soil fertility 3.30 0.64 2.66 0.70 0.17 1.56 Fruit 3.22 0.32 2.90 0.74 0.24 0.41 Construction 3.00 0.24 2.75 0.85 0.17 0.32 Fodder 1.95 0.87 1.08 0.31 0.45 1.59 Beverage 1.39 0.22 1.17 0.76 0.04 0.43

CHAPTER 7

109

Table 7.2. Abundance distribution of all species and the most frequent use-groups. Cells indicate the number of species with a given abundance in the group. ‡ Abun-dance Class

All Fire-wood

Shade Medicine Orna-mental

Timber Boun-dary

Charcoal Soil fertility

Fruit Con-struction

Fodder Beverage

1 44 36 23 18 17 14 5 6 7 5 5 0 1 2 18 18 10 10 8 3 1 5 2 3 2 1 1 3 8 7 8 3 6 1 1 0 0 1 1 0 0 4 6 7 3 6 2 3 1 2 0 2 0 0 0 5 5 6 1 0 1 3 0 1 3 0 1 0 0 6 5 2 2 2 2 3 2 0 0 0 0 0 0 7 3 3 0 2 1 2 0 0 0 1 1 0 0 8 1 2 2 0 1 0 0 0 0 0 0 0 0 9 1 0 2 0 0 1 1 1 1 0 0 0 0

10 5 3 2 2 0 0 0 0 1 1 0 0 0 11 2 3 0 0 2 1 0 1 0 1 3 0 0 12 1 1 2 1 0 0 0 0 0 0 0 0 0 13 1 0 0 0 0 1 0 1 0 0 0 0 0 14 2 1 1 1 1 0 0 0 0 0 1 0 0 16 0 0 2 0 0 0 1 1 1 0 0 0 0 17 2 2 0 0 1 0 0 0 0 2 0 0 0 18 0 1 1 0 0 1 0 0 0 0 0 0 0 19 1 2 0 1 1 0 1 0 0 0 0 0 0 20 1 0 0 2 0 0 0 0 0 0 1 0 0 21 1 1 1 0 0 0 0 1 0 1 0 0 0 22 2 1 2 0 0 0 0 0 1 0 0 0 0 23 0 0 0 1 1 1 0 0 0 0 0 0 0 24 0 0 0 0 0 0 2 0 0 0 0 0 0 25 4 4 1 0 0 1 1 0 0 0 0 0 0 26 1 0 0 0 0 0 0 0 0 0 0 0 0 28 2 2 0 0 0 1 0 0 0 0 0 0 0 29 0 0 0 2 0 1 0 1 0 0 0 0 0 30 1 1 0 0 1 0 0 0 0 0 0 0 0 31 0 0 0 0 0 0 0 0 1 0 0 1 0 32 1 1 0 0 1 1 0 0 0 1 0 0 0 33 1 1 0 0 0 0 0 0 0 0 0 0 0 36 1 1 0 0 1 0 1 0 1 0 0 0 0 40 0 0 0 0 0 0 0 1 0 0 0 0 0 41 1 1 1 0 0 0 0 0 0 0 0 0 0 42 1 1 0 1 0 0 0 0 0 0 0 0 0 43 1 1 0 0 0 0 0 0 0 0 0 0 0 47 1 1 0 0 0 1 0 0 0 0 0 0 0 48 0 0 0 0 0 0 0 1 0 1 0 0 0 49 2 2 1 0 0 0 1 0 0 0 0 0 0 50 2 2 1 0 1 0 0 0 0 0 0 2 0 51 1 1 1 0 0 0 0 0 0 0 0 0 0 52 1 1 0 0 0 0 1 0 1 0 0 0 0 53 0 0 0 0 0 0 0 1 0 0 0 0 0 55 1 0 0 0 0 0 0 0 0 0 0 0 0 56 0 0 0 0 0 0 0 0 1 0 0 0 0 60 0 0 0 0 0 0 0 0 0 0 0 1 0 65 0 0 1 0 0 0 0 0 0 0 0 0 0 68 0 0 0 0 0 0 0 0 1 0 0 0 0 69 0 0 1 0 0 0 0 0 0 0 0 0 0 70 1 1 0 0 0 0 0 0 0 0 1 1 0 72 1 0 0 0 0 0 1 0 0 0 0 0 0 74 2 2 0 0 0 0 0 0 0 0 0 0 0 75 1 1 1 0 1 0 0 0 0 0 0 0 0 77 1 1 0 0 0 0 0 0 0 0 0 0 0 79 1 1 0 0 0 1 0 0 0 0 0 0 0 81 1 2 0 1 0 0 0 1 0 0 0 0 0 83 0 0 1 2 0 0 0 0 0 0 0 0 0 86 1 1 0 0 0 0 0 0 0 0 0 0 0 88 1 1 0 0 0 0 0 0 0 0 0 0 0 91 0 0 0 0 0 0 0 1 0 0 0 0 0 94 0 0 1 0 0 0 0 0 0 0 0 0 0 96 1 0 0 0 0 0 0 0 0 0 0 0 0

100 0 0 0 1 0 0 1 0 0 0 0 0 0 101 0 1 0 0 0 0 0 0 0 0 0 0 0 102 1 0 0 0 0 0 0 0 0 0 0 0 0 103 0 0 0 0 1 0 0 0 0 0 0 0 0 108 0 0 0 0 0 0 1 0 0 0 0 0 0 110 1 1 0 0 0 1 0 0 0 0 0 0 0 111 0 0 1 0 0 0 0 0 0 0 0 0 0 113 1 1 0 0 0 0 1 0 0 0 0 0 0 119 0 0 0 0 0 0 1 0 0 0 0 0 0

DIVERSITY PROFILES

110

Table 2 (cont.) Abun-dance Class

All Fire-wood

Shade Medicine Orna-mental

Timber Boun-dary

Charcoal Soil fertility

Fruit Con-struction

Fodder Beverage

122 0 0 1 0 0 0 0 0 0 0 0 0 0 145 1 1 1 0 0 1 0 0 0 0 0 0 0 150 0 0 0 0 0 0 0 0 1 0 0 0 0 161 0 0 1 0 0 0 0 0 0 0 0 0 0 162 1 1 0 0 0 0 0 0 0 0 0 0 0 165 1 1 0 0 0 0 0 0 0 0 0 0 0 188 0 0 0 0 1 0 0 0 0 0 0 0 0 189 1 0 0 0 0 0 0 0 0 0 0 0 0 191 0 0 0 0 0 0 0 0 0 0 0 1 0 201 1 1 1 0 0 1 0 0 0 0 0 0 0 203 0 0 0 1 0 0 0 0 0 0 0 0 0 225 0 0 1 0 0 0 0 0 0 0 0 0 0 246 0 0 0 0 0 0 0 0 1 0 0 0 0 248 1 1 0 0 0 1 0 0 0 1 0 0 0 250 0 0 0 0 0 0 0 0 0 0 1 0 0 269 0 0 1 0 0 0 0 0 0 0 0 0 0 277 1 1 0 0 0 0 0 0 0 1 0 0 0 296 1 1 0 0 0 1 0 0 0 0 0 0 0 318 0 0 0 0 0 1 0 0 0 0 0 0 0 320 0 0 0 0 0 0 0 0 0 0 1 0 0 335 1 1 0 0 0 0 0 0 0 0 0 0 0 364 1 1 0 0 0 0 0 0 0 1 0 0 0 370 1 0 0 0 0 0 1 0 0 0 0 0 0 404 0 0 0 0 0 0 1 0 0 0 0 0 0 413 0 0 1 0 0 0 0 0 0 0 0 0 0 454 1 1 0 0 0 0 0 0 0 0 0 0 0 465 0 0 1 0 0 0 0 0 0 0 0 0 0 476 1 1 0 0 0 0 1 0 0 0 0 0 0 495 1 1 0 0 0 0 0 0 0 0 0 0 0 505 0 0 1 0 0 0 0 0 0 0 0 0 0 608 0 0 0 0 0 0 1 0 0 0 0 0 0 617 1 0 0 0 0 0 0 0 0 1 0 0 0 618 0 0 1 0 0 0 0 0 0 0 0 0 0 621 0 0 1 0 0 0 0 0 0 0 0 0 0 849 1 1 0 0 0 0 0 0 0 1 0 0 0

1016 0 1 0 0 0 0 0 0 0 0 0 0 0 1030 0 0 0 0 0 0 0 0 1 0 0 0 0 1119 0 0 0 0 0 1 0 0 0 0 0 0 0 1276 0 0 0 0 0 0 0 1 0 0 0 0 0 1693 0 0 0 0 0 0 0 1 0 0 0 0 0 1733 0 0 0 1 0 0 0 0 0 0 0 0 0 1861 1 1 0 0 0 0 0 0 0 0 0 0 0 1869 0 0 0 0 0 0 0 0 1 0 0 0 0 1896 0 0 0 0 1 0 0 0 0 0 0 0 0 2019 1 1 0 0 0 0 0 0 1 0 0 0 0 2523 1 1 0 0 0 0 1 0 0 0 0 0 0 2686 0 0 0 0 1 0 0 0 0 0 0 0 0 2702 1 1 0 0 0 0 0 0 0 0 0 0 0 3783 1 1 0 0 0 0 0 0 0 0 0 0 1 4019 1 1 0 0 0 0 0 0 0 0 1 0 0 4458 1 1 0 0 0 0 1 0 0 0 0 0 0 5930 1 0 0 0 0 0 1 0 0 0 0 0 0 6261 1 1 0 0 0 0 1 0 0 0 0 0 0 6300 0 0 0 0 0 0 0 0 1 0 0 0 0 6729 1 1 0 0 0 0 0 0 0 1 0 0 0 7788 1 1 0 0 0 1 1 0 0 0 0 0 0

13258 1 1 0 0 0 0 1 0 0 0 0 0 0 15397 1 1 0 0 0 0 0 0 0 0 0 0 1 17244 1 1 0 0 0 1 0 0 0 0 1 0 0

‡ This table was included in the submitted article to Environemtal and Ecological Statistics , which requires that raw data are provided, and was not referred to specifically in this section (as description of study area and data collection methods were combined in Chapter 3). The table allows for rapid calculation of diversity and evenness profiles (figures 7.1 and 7.2).

CHAPTER 7

111

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 200 inf

alpha

0

1

2

3

4

5

6

Hallfirewoodshademedicineornamentaltimberboundarycharcoalsoil fertilityfruitconstructionfodderbeverage

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

244.7

403.4

EX

P(H

)

Figure 7.1. Diversity profiles based on the Rényi series Ha for all species and species belonging to particular use-groups

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 200 inf

alpha

allfirewoodshademedicineornamentaltimberboundarycharcoalsoil fertilityfruitconstructionfodderbeverage

-4

-3

-2

-1

0

ln(E

)

0.02

0.03

0.05

0.08

0.14

0.22

0.37

0.61

1.00E

Figure 7.2. Evenness profiles based on the Rényi series Ha for all species and species belonging to particular use-groups

a b c

d e f

Figure 7.3. Accumulation curves for diversity profiles for various use-groups, represented as three-dimensional contours at full-sample profile values. Alpha values of 4, 4.5 and 5 correspond to alpha values of 10, 100 and 8 respectively. Use-groups: a: firewood, b: shade, c: medicine, d: ornamental, e: timber, f: boundary demarcation

g h i

j k l

Figure 7.3 (cont.d). Accumulation curves for diversity profiles for various use-groups, represented as three-dimensional contours at full-sample profile values. Alpha values of 4, 4.5 and 5 correspond to alpha values of 10, 100 and 8 respectively. Use-groups: g: charcoal, h: soil fertility, i: fruit, j: construction, k: fodder, l: beverage

DIVERSITY PROFILES

114

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

6H

allfirewoodshademedicineornamentaltimberboundarycharcoalfruitconstructionsoil fertilityfodderbeverage

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

244.7

403.4

EX

P(H

)

Figure 7.4. Average profile values, 95 percent confidence intervals (rectangles) and ranges (lines) for 100 farm subsamples based on 100,000 random site sequences. All profile values were calculated for the same scales, but groups were presented at different

alpha/scale values for better discrimination of group intervals and ranges. (inf.: 8 )

0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0

Cumulative proportion of species

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Cu

mu

lativ

e p

rop

ort

ion

of

abu

nd

ance

allfirewoodshademedicinetimberornamentalboundarycharcoalfruitconstructionfertilitybeveragefodder

Figure 7.5. Lorenz curves for all species and species belonging to particular use-groups

CHAPTER 7

115

0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 160 170 180 190 200

Number of farms

0

1

2

3

Sh

ann

on


Figure 7.6. Accumulation curves for the Shannon diversity index for all species and species belonging to particular use-groups.

Horizontal reference lines indicate full-sample values for 11 accumulation curves.

10 20 30 40 50 60 70 80 90 100110 120 130 140 150 160 170 180 190 200

Number of farms

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

ln (

E)

/ ln

(S

)


Figure 7.7. Accumulation curves for ln(E)/ln(S) for all species and species belonging to particular use-groups. Horizontal

reference lines indicate full-sample values for two accumulation curves.

CHAPTER 8 DO FARM CHARACTERISTICS EXPLAIN DIFFERENCES IN TREE SPECIES DIVERSITY AMONG WESTERN KENYAN

FARMS? R KINDT, AJ SIMONS & P VAN DAMME SUBMITTED TO AGROFORESTRY SYSTEMS

With the objective of planning diversification of on-farm tree species composition, a survey was conducted in western Kenya involving a complete tree census and collection of ethnobotanical information on 201 farms. Differences between farms in diversity of the 12 most frequent use-groups were analysed by species richness, Shannon, Simpson and Berger-Parker diversity indices, and Shannon evenness and equitability. A large range of values was detected among farms and use-groups. Multiple linear regression of diversity statistics on household characteristics indicated significant relationships. However, these relationships generally explained low percentages of variation (ranging 2-44%). The connection between household characteristics and use-group diversity allows targeting specific household types with lower diversity. Farm size had a positive relationship with diversity of most use-groups. However, accumulation curves revealed that the same area carried a larger abundance and diversity when it was composed of a greater number of smaller farms. If the pattern of further subdivision of farmland in the survey area continues and the same differences between smaller and larger farms prevail, then larger diversity per unit area can be expected. Because smaller farms contain smaller diversity, however, diversification with the aim of enhancing or stabilising productivity of individual farms may become an important priority in the survey area. The results presented allow for the identification of individual farms, use-groups, and household types for which diversification is more relevant, and at the same time allow for impact monitoring.

EXPLANATORY FACTORS OF FARM DIVERSITY

118

8.1. Introduction

The objective of tree domestication research in western Kenya is the diversification of the tree species composition present in agroecosystems (Kindt & Lengkeek 1999). Making agroecosystems ecologically more stable and economically more rewarding through diversification and intensification of land use by the domestication of agroforestry trees is one of the three pillars of ICRAF’s research (ICRAF 1997; ICRAF 2000). This chapter provides an analysis of differences in tree species diversity among farms. Diversity will contribute to ecosystem productivity and stability under conditions of heterogeneity in species traits and environmental characteristics (Kindt et al. – Chapter 2, 4 & 5). Since the relationship between diversity and ecosystem functioning is saturating, diversification is expected to contribute more to ecosystem functioning if it increases the diversity of functional groups with the lowest diversity (Kindt et al. – Chapter 2, 4 & 5).

The analysis of differences in species diversity had two objectives: to identify whether and where there exist substantial differences in species diversity among farms, and to identify whether potential differences could be linked to farm characteristics. The first objective deals with the possibility of identifying and targeting farms of low diversity. The second objective relates to the possibility of targeting specific types of households. Because of population growth and subsequent farm divisions in the area (Bradley 1991; Holmgren et al. 1994), special attention was given in a final set of analyses to the relationship between farm size and tree species diversity.


Formulation of use-groups (4.1), alpha and gamma diversity (4.2.1), diversity and evenness statistics (4.2.6), regression, correlation and non-parametric analysis (4.4)

8.3. Results

8.3.1. Range of Diversity Parameters.

Figures 8.1 to 8.8 show box plots for the distribution of the diversity characteristics of the various use-groups for H, ln (S), J, E, ln (d-1), ln (N), ln (S+1) and ln (N+1) respectively. These figures show the large range in diversity among farms, where some farms have relatively low diversity and other farms relatively high diversity. At the same time, the figures show that although some use-groups are more diverse on average, some farms may still have low diversity for these groups. Most important, the figures show that there is already high diversity in the agroecosystems studied.

If we set the value of ln(2) as a benchmark value for H (this value corresponds to two equally distributed species, or a higher number of less equally distributed species), then we can see from Figure 8.1 that a large number of farms already had larger values, whereas the majority of farms are below this level for the majority of use-groups. Firewood and fruit were the only use-groups where more than half of the farms had values above the benchmark. Beverage and fodder were the only use-groups were all farms had lower diversity than the benchmark. No single group had values above the benchmark for all farms.

When studying figure 8.2 with information on ln (S) and taking the same benchmark value of ln(2) (corresponding to 2 species – equally distributed species have the same value of ln (S) at all scales of the Rényi series, Kindt et al. - Chapter 6), then similar conclusions can be drawn as for

CHAPTER 8

119

H. Because ln (S) can only take discrete values, box plots are probably not the ideal representation for the distribution of this statistic. They were maintained, however, for consistency with representation of other diversity statistics. Firewood (and fruit, ignoring outliers) is the only use-group where all farms had values above the benchmark, whereas other use-groups had values above and below the benchmark. Compared with H, values were higher which points to the fact that species were not equally distributed within use-groups.

Figure 8.3 (J) presents results that confirm that most use-groups were dominated by a limited number of species, although some farms had almost equally (J˜ 1) distributed species. Again, a large range in values of farms can be observed. Figure 8.4 (E) provides a similar picture as for J. In some groups (fodder and medicine), more than half of the farms showed higher diversity than the one associated with the broken stick distribution. However, Figures 8.3 and 8.4 did not include farms where ln (S) equalled zero for the use-group.

Figure 8.5 shows that most farms had values below the ln(2) value for ln (d-1). The benchmark value corresponds here to the most dominant species having half of the total abundance within groups. Firewood was the only group where more than half of the farms had higher values than the benchmark value.

Figure 8.6 shows the large range in abundance of use-groups over farms. Most use-groups showed a large range in abundance, while large differences among use-group means were also observed.

Figure 8.7 demonstrates that half of the farms had no trees contributing to the beverage, charcoal, fertility, fodder, and ornamental uses. These were the same use-groups with the lowest values in figure 8.2, with exception of the ornamental group. Boundary demarcation, firewood, fruit, and timber were groups where 75% of farms had two or more species. Only including farms where the use was present (figure 8.2) shows that shade joined the use-groups with this characteristic.

Figure 8.8 highlights those use-groups that could be observed in figure 8.7 where particular uses are not common. The figure also shows the range in abundance that could be observed in figure 8.6.

8.3.2. Differences among Use-groups

Table 8.1 provides the results of non-parametric significance tests on coefficients calculated by multiple linear regression. Table 8.1 shows that average values of diversity statistics of most use-groups differed significantly from the overall mean, as shown above in the box plot diagrammes. There were significant differences among the means of use-groups, but at the same time large differences within groups obscured the group mean differences (R2 values were not very high). Kruskal-Wallis Rank Sum tests and MRPP for each diversity statistic confirmed the differences between groups. Some use-groups were less diverse on average and could be targeted by domestication projects. For instance, fertility, beverage, charcoal, fodder and ornamental had lowest average H and ln(S) values.

8.3.3. Correlations of Diversity Statistics

Spearman rank correlation coefficients calculated among diversity indices belonging to the same farms and the same groups are provided in table 8.2. Table 8.2 illustrates that correlations among diversity statistics are generally strong. Strength of correlation followed similarity in Rényi series scale parameter value. Because of weaker correlations between parameters at different scales in the Rényi series, absolute diversity ordering proves complicated. Absolute ordering implies that


120

all values in the series of one system are higher or lower than all the values of another system. Figure 8.9 confirms the complications with diversity ordering. Perfect ordering would result in monotonically increasing ln (d-1) with increasing ln (S). Farms that are more diverse in absolute terms would thus appear to the top and to the right of less diverse farms in figure 8.9. The figure shows that many farms will have intersecting diversity profiles. This implies that some farms have larger species richness but smaller equality in species proportions. Figure 8.9 also shows that H offers a reasonable compromise to order farms in diversity as values at the upper and right sides of the graph have higher H values in general.

J and E parameters were strongly correlated (table 8.2), while correlations with other diversity statistics increased with the Rényi scale parameter (which could be expected as indices higher in the series are more strongly linked to evenness of species distribution). Because of the weaker correlations of evenness parameters with diversity statistics, ln (S) or H together with J or E present more information on variation in the data.

Correlations of ln (N) and the diversity statistics show that increasing abundance is more strongly related to species richness, than the evenness of their distribution. Correlations throughout the Rényi series are positive, so increasing tree abundance on an individual farm is associated with an increase in diversity. However, negative correlation with J and E show that the farms with greater abundance have less equally distributed species than less abundant farms. Correlations however are such that their predictive power is weak; more abundant farms are not necessarily more diverse and less evenly distributed than less abundant farms.

8.3.4. Correlations among Shannon Indices of the Same Farm

Table 8.3 provides information on the correlation of the Shannon index for different use-groups belonging to the same farms. The general picture is that correlations were very weak: most are not significantly different from zero. It is therefore difficult to classify farms as more or less diverse for all use-groups based on the information of one use-group. Diversity parameters should be calculated for all groups. Correlations mainly result from the same species used for several purposes, for example boundary markers that also produce fruit, construction, or firewood, or have ornamental value. These correlations also point to various purposes related to on-farm niches, e.g. trees in homestead areas used for shade, fruit, medicine and ornamental purposes.

8.3.5. Influences of Household Characteristics

Table 8.4 indicates the coefficients calculated by multiple linear regression analysis for spatial and household characteristics as explanatory variables for diversity statistics. The results show that although 33 regressions proved significant at the 0.1% probability level (this corresponds with a Bonferonni correction for 50 tests for a significance level of 5%), their explanatory power (R2) was low. In other words, there were significant differences between the mean values of the diversity statistics, but the explanatory variables were not very effective in predicting the values of diversity statistics at individual farms.

Partial linear regression results indicated that for many use-groups and diversity statistics, household characteristics explained more variation than spatial characteristics. S and N for all trees, construction, firewood, fruit, timber, and soil fertility are examples. For other use-groups however, spatial characteristics explained more variation; examples of this are beverage, boundary markers, and ornamental species. A third pattern occurred where both types of variables explained similar fractions of diversity (for example H for construction; H and S for medicine and shade). In the last four cases, the sum of exclusive variation explained by spatial and

CHAPTER 8

121

household characteristics exceeded 100%. This means that the combination of both types of characteristics explained the variation better than these characteristics separately (Legendre & Legendre 1998 p. 533).

Detection of trends in village diversity based on the gradient towards the Kakamega Forest National Reserve proved difficult. Shimutu, which is located closest to the forest, was lowest in overall N, and N for construction, fodder, timber, medicine, shade and soil fertility, and lowest H and S for fruit and shade. Largest values at the other end of the gradient (Ebuchiebe) were only observed for N of soil fertility. Cases where none of the intermediate villages (Madidi and Mutambi) had higher values included N for construction and timber, and H and S for fruit and shade. The results for other use-groups and diversity characteristics did not relate to forest distance, but showed that characteristics other than farm location relative to the forest would be better in explaining differences in diversity. However, the only case where Shimutu had the largest diversity statistic was observed for J of construction, which was a regression coefficient very close to the one for Ebuchiebe.

Analysis of household and farm characteristics showed, among others, the following trends. Male-headed households had highest N for all trees, firewood, medicine, and soil fertility, highest H for timber and highest S for fruit, shade and soil fertility. For many use-groups, de jure female-headed households had similar diversity characteristics as male-headed households, but they contrasted with de facto female-headed households. Examples are S for all trees, construction, firewood, timber and ornamental trees and H for shade trees and N for construction, fodder, and timber. De jure female-headed households had lowest H for boundary demarcation and soil fertility, lowest N for shade trees and highest N for boundary demarcation.

Farm size was positively related to diversity statistics of many use-groups, and relationships were very strong as shown by the coefficients and their significance. Exceptions were decreasing fruit H and J despite increasing S and N for this group, and decreasing ornamental S and shade J. The area-species relationship predicts increasing S with increasing farm size, therefore these results are not surprising. The importance of the area effect is more interesting, however. Whether larger farms are more or less diverse than predicted by the area-species relationship is the subject of the next section.

Coefficients for indicators of wealth (housing type and number of cattle) seem to indicate that wealthier households have lower overall H, S and J, greater N for all trees, beverage, boundary demarcation, firewood, fodder, timber, ornamental, and shade, but smaller N for construction, fruit and medicine. Contradictory results were obtained for charcoal.

Number of years the head of household was in charge of the farm was associated with higher H for construction and fruit, higher S for fruit and construction, higher N for medicine, and lower N for boundary demarcation. The maximum age of household head or partner was positively associated with H of medicine and shade, and overall and fruit, medicine and shade S. This household characteristic was negatively associated with fertility H and S, and construction S. The latter result contrasts with the finding of the number of years that a head was in charge of the farm.

Number of children was positively correlated with H and S of all trees, boundary, fruit and medicine, H of timber, and N of construction. The only negative correlation was with S of the charcoal. Where schooling was retained after stepwise regression, its coefficient was negative, e.g. H for all trees, boundary demarcation, firewood, and soil fertility.


122

8.3.6. Influences of Farm Size

The relationship between area and diversity is normally not expressed by a linear relationship as discussed above, but through a power function model. For species richness and Shannon diversity, the fitted power function models were 186.0423.12 NS = (rse = residual standard error = 6.04) and 115.0297.1 NH = (rse=0.468). Respective models where parameters a were calculated beforehand were 188.0375.12 NS = (rse=6.03) and 138.0239.1 NH = (rse=0.467). Parameters a were closely approximated in models where this parameter was estimated.

Figures 8.10 and 8.11 show the fitted models with prefixed parameters a (these models had slightly smaller residual standard error) and the range of observed values. As observed in the previous section, there remained considerable unexplained variation. Some farms with a small size had relatively high diversity, while some farms with larger sizes had relatively low diversity. The average diversity for a specific farm size can be predicted quite precisely, as 95% confidence intervals were small. The figures further demonstrated that a linear model would provide a similar fit to the data.

Table 8.5 demonstrates that random addition of several farms resulted in most cases in higher H, S, and N than for single or fewer farms of the same combined size. This table also points to the area-species relationship – larger farms have, on average, higher diversity. The results indicate that when several farms are combined, diversity increases more than expected from a pure farm size increment effect.

8.4. Discussion and Conclusions

The analyses show the large range in diversity characteristics among farms and use-groups. Some use-groups were more diverse on average throughout the survey but there were exceptions – some diverse groups included some farms with low diversity and some less diverse groups included some farms with high diversity. This means that use-groups with low average alpha diversity can be detected at the survey scale, but that diversification may not be appropriate for all farms since the diversity of some farms was already quite high.

Correlations were high among diversity indices calculated for the same farms and same use-groups, but low among different uses for the same farms. Therefore, diversity indices should be calculated separately for use-groups, while a reasonable ranking of diversity of farms and use-groups could be obtained from calculation of a single diversity index. Magnussen & Boyle (1995) also found strong linear correlations between H and D-1 (0.98) based on mathematical modeling of species distributions. Because correlations become weaker at larger scale differences in the Rényi series, it is impossible to rank use-groups absolutely in diversity. Absolute ranking only occurs when all values of the Rényi diversity series rank in the same order, otherwise one system is richer and the other system possibly more evenly distributed (Kindt et al. - Chapter 6). Our recommendation is to compare farms based on a diversity index related to species richness (S or H) in combination with an evenness indicator (J or E). Despite the fact that J and E decreased for more diverse groups, richer groups still had more evenly distributed species according to the inverse Berger-Parker index.

Regression analysis showed that various household and farm characteristics had significant relations with diversity characteristics of use-groups. Many of the relationships could point to causal effects of on-farm management of tree species diversity. One example is the relationship on a farm between number of children and diversity of fruit and medicine trees. Another example is the relationship between wealth and smaller abundance of fruit, construction, and medicine trees as this relationship could indicate that wealthier households prefer to purchase

CHAPTER 8

123

these products rather than to produce them on their farm. A third example is the higher number of trees for boundary demarcation for de jure female headed households. Maybe legally female-headed households have greater problems with recognition by male neighbours of farm boundaries. An example of potentially spatially related causes for diversity is the lower abundance of construction, and timber trees in the village closest to Kakamega Forest.

Some relationships were consistent with results reported by David & Swinkels (1994): wealthier households having more trees. However, the results contrasted with findings by David & Swinkels (1994) that male-headed households sold more poles and timber: our results showed a contrast between de facto female-headed households and the other two household types. Simple regressions confirmed the findings of David & Swinkels (1994): the abundance of construction and timber was positively associated with male-headed households and not significantly related to de jure female headed households (P(F) for de jure and de facto female heads was respectively 0.74 and 0.97). The difference between the multiple and single regression results show the value of the first type of analysis in isolating effects by removing possible effects from other variables.

Differences among male-headed, de jure female-headed and de facto female-headed households showed a complex pattern. Male-headed farmers differed from female-headed households in some cases, whereas for other groups and diversity statistics male-headed and de jure female heads were more similar. Den Biggelaar (1995) found that female household heads in southern Rwanda were permitted to take decisions to plant trees, which is normally forbidden for women in Rwanda. Female household heads, however, did not plant more trees than wives of male heads, but their reasons for planting trees were more similar to reasons provided by male heads. Analysis of our dataset led to the contrasting finding that female heads established more trees than wives of male heads (averaging 115 trees per farm versus 4), but fewer than male heads (430). This finding could indicate a further evolution in tree planting practices, where permissions to plant trees also resulted in more planted trees by female heads of households.

Jarvis et al. (2000) mention that gender, age, wealth or social status affect farmers’ knowledge, actions and access to resources regarding the maintenance of crop diversity. Long et al. (2000) provide some examples of the influence of wealth, age and gender on crop diversity. Citing examples of multiple regression analyses, Jarvis et al. (2000) also listed farm size, family size, and years of education as significant explanatory factors for farmer variety choices. The results from our study show that variables that influence crop diversity (wealth, age, gender, farm size, family size, years of education) also influence on-farm tree diversity.

Whereas the relationships revealed by multiple regression were significant, their explanatory power was low. Low predictive values are common in diversity studies. For example, functional group and species richness explained only 17.7% of variation in the productivity of grasslands (Hector et al. 1999). Similar research on the relationship between species richness and grassland plant cover provided R2 values of 0.18 (experiment) and 0.16 (native grassland) (Tilman et al. 1996). Other examples are a R2 value of 0.203 of relative resource use versus functional group richness reported by Hooper & Vitousek (1998), and R2 values of 0.12 – 0.18 between species richness and the size of the largest individual exotic invader described by Levine (2000).

Cromwell & van Oosterhout (2000) used a multiple linear regression approach to investigate crop diversity in two districts in Zimbabwe. They found that farm/household characteristics explained 86% of the variation in the number of crops grown, 80% of the variation in the number of crop varieties, and 50% of the variation in the proportion of the farm devoted to small grains. Considerably more variation was explained than in our study. However, some of the explanatory variables they used had an evident link to observed diversity, such as the fact whether small grains were valued by the household and whether respondents grew all varieties that they liked.

Low R2 values indicated differences between average values with large percentages of unexplained variation (i.e. many exceptional cases against the overall trend). Data available here


124

only allow for speculation about the reason for the unexplained variation. One possibility could be that various types of farmers do have different strategies, but that many farmers are currently not able to establish the preferred diversity because of germplasm availability problems. Following this hypothesis, correlations among household characteristics and diversity would increase when more germplasm would become available to farmers. Another possibility is that some traditional practices that limited certain types of households establishing trees (such as taboos against women planting trees as described by Bradley, 1991 and David & Swinkels, 1994) are currently disappearing. Correlations among household characteristics and diversity would decrease when new customs develop following the latter hypothesis. One method by which future trends in use-group diversity versus household characteristics can be tested is by participatory research on desired levels of diversity. Such a survey is currently underway in the area.

Combining several smaller farms resulted in larger average diversity and abundance for the same area than observed in undivided farms. Larger farms still had higher diversity than smaller farms, however. The higher diversity of combined farms indicates that beta diversity effects (differences in species composition among farms) caused higher diversity in farm combinations. The analysis of the relationship between (combined) farm size, and diversity and abundance, was repeated for separate villages to control for beta diversity between villages. Results of these analyses (not provided here) showed the same pattern as observed for random combinations of farms. Results that differences in farm area alone can not explain all differences in diversity, as observed in our survey, suggest that households manage tree species diversity on their farms. The differences in diversity between combinations of farms and undivided farms also allow speculation about potential effects on diversity from further subdivision of farms. If the same trends remain valid, then the expectation is that alpha diversity would further decrease for individual farms, but that beta diversity would more than compensate for this decrease so that diversity of the same area of farm size would increase. With decreasing diversity of individual farms, tree diversification may become more relevant for individual farms. Higher diversity per unit area could, however, imply that higher stability would already be present at the agro-ecosystem level. Higher species richness and abundance per unit area offer larger opportunities for tree species conservation through use in the landscape (Kindt & Lengkeek 1999). However, additional information is needed to check whether farmers would be interested in maintaining rare species on smaller farms, and for which species sustainable genetic management is possible and practical in such landscape (Kindt et al. – Chapter 11).

The findings of higher tree abundance per unit area on smaller farms confirmed the “more people, less forests, more trees” postulation as formulated by Simons et al. (pers. comm.) citing examples from China, Kenya, Sri Lanka, Thailand and Vietnam. The literature cited by these authors included findings from Bradley et al. (1985) (based on data from Kakamega and Vihiga districts, to which our survey area belonged) and from Holmgren et al. (1994) who confirmed the relationship for the total of highland areas of Kenya. Such model of a U-shaped curve of tree biomass over population density and time was also described for Sahelo-Sudanian shrub-grasslands of north-east Nigeria (Mortimore et al. 1999), indicating that increased tree biomass with population density may be a global phenomenon for areas with high population densities.

CHAPTER 8

125

Table 8.1. Mean, standard deviation, and range for six diversity statistics together with differences between use-groups calculated by randomisation tests on coefficients calculated by multiple linear regression on dummy variable coding all groups except fertility. The group mean for fertility therefore corresponds to the intercept, while means for the other groups correspond to the addition of the intercept and the coefficient. The probability of the Kruskal-Wallis rank sum test and Multi-Response Permutation Procedure probability with chance-corrected within-group agreement statistic A describe the significance of differences among groups. Parameter H ln (S) J E ln (d-1) ln (N)

Mean 0.608 1.022 0.600 0.740 0.332 3.522 St. dev. / Range 0.586 2.431 0.877 3.761 0.271 0.984 0.330 1.213 0.371 1.791 1.969 8.183 Regression R2 / P. 0.495 0 0.666 0 0.121 0 0.125 0 0.330 0 0.469 0 Variable Coeff. P. Coeff. P. Coeff. P. Coeff. P. coeff. P. coeff. P. Intercept / Fertility 0.146 0.260 0.567 0.696 0.086 2.352 Beverage -0.106 0.097 -0.130 0.093 -0.252 0.003 -0.308 0.003 -0.071 0.127 1.596 1e-5 Boundary 0.437 1e-5 0.661 1e-5 0.051 0.250 0.056 0.301 0.263 1e-5 2.511 1e-5 Charcoal 0.036 0.535 0.036 0.609 0.023 0.719 0.026 0.739 0.018 0.668 0.121 0.551 Construction 0.199 3e-5 0.355 1e-5 0.003 0.938 0.006 0.910 0.107 0.001 1.675 1e-5 Firewood 1.328 1e-5 2.387 1e-5 -0.01 0.797 -0.053 0.315 0.668 1e-5 3.486 1e-5 Fodder -0.077 0.271 -0.171 0.046 0.213 0.055 0.266 0.049 -0.04 0.361 -1.056 1e-5 Fruit 0.844 1e-5 1.217 1e-5 0.100 0.021 0.106 0.044 0.448 1e-5 0.849 1e-5 Timber 0.417 1e-5 0.956 1e-5 -0.101 0.022 -0.132 0.013 0.178 1e-5 1.898 1e-5 Medicine 0.299 1e-5 0.333 1e-5 0.20 8e-5 0.242 1e-4 0.201 1e-5 -0.811 1e-5 Ornamental 0.125 0.024 0.299 1e-5 -0.064 0.234 -0.082 0.208 0.061 0.133 -0.107 0.576 Shade 0.589 1e-5 0.745 1e-5 0.182 7e-5 0.211 1e-4 0.359 1e-5 -0.290 0.082 Kruskal-Wallis P. 0 0 0 0 0 0 MRPP A / P. 0.327 0 0.469 0 0.091 0 0.095 0 0.222 0 0.313 0

Table 8.2. Spearman rank correlations between various diversity statistics calculated for the same farm and use-group (upper half) and probability of these correlations equalling zero (lower half)

ln (S) H ln (D-1) ln (d-1) J E ln (N) a 0 1 2 8 - - - Ln (S) 1.000 0.896 0.847 0.812 -0.078 -0.120 0.582 H 0.000 1.000 0.990 0.973 0.576 0.541 0.387 Ln (D-1) 0.000 0.000 1.000 0.993 0.702 0.670 0.346 Ln (d-1) 0.000 0.000 0.000 1.000 0.766 0.737 0.324 J 0.006 0.000 0.000 0.000 1.000 0.998 -0.529 E 0.000 0.000 0.000 0.000 0.000 1.000 -0.547 Ln (N) 0.000 0.000 0.000 0.000 0.000 0.000 1.000

a: Rényi scale parameter value (see methods)

Table 8.3. Spearman rank correlations of Shannon diversity indices between use-groups belonging to the same farms (upper half) and significance of these correlations (lower half). Correlations with probability lower than 10% are presented in bold.

Beve-rage

Boun-dary

Char-coal

Con-struction

Fire-wood

Fodder Fruit Timber Medi-cine

Orna-mental

Shade Fertitlity

Beverage 1.000 -0.055 -0.204 -0.143 -0.134 n/a -0.003 -0.030 0.228 0.246 0.163 0.094 Boundary 0.669 1.000 -0.024 -0.125 0.387 0.121 -0.239 0.048 0.047 0.227 -0.003 0.038 Charcoal 0.279 0.826 1.000 -0.064 0.119 -0.270 -0.066 0.103 -0.182 0.039 -0.099 -0.205 Construction 0.267 0.081 0.564 1.000 0.264 0.006 -0.005 0.421 -0.001 -0.231 -0.020 0.089 Firewood 0.293 0.000 0.287 0.001 1.000 0.332 -0.067 0.371 0.100 0.056 -0.060 0.081 Fodder n/a 0.411 0.262 0.966 0.024 1.000 -0.052 0.342 0.247 0.296 0.234 0.066 Fruit 0.979 0.001 0.556 0.937 0.343 0.722 1.000 0.129 0.219 0.130 0.198 0.106 Timber 0.813 0.501 0.359 0.000 0.000 0.020 0.070 1.000 -0.034 0.113 0.070 -0.060 Medicine 0.121 0.619 0.198 0.992 0.291 0.150 0.020 0.722 1.000 0.146 0.460 -0.080 Ornamental 0.144 0.027 0.805 0.024 0.584 0.110 0.202 0.270 0.271 1.000 0.358 -0.146 Shade 0.221 0.966 0.404 0.789 0.436 0.142 0.011 0.368 0.000 0.001 1.000 0.122 Fertility 0.542 0.655 0.131 0.304 0.344 0.674 0.215 0.451 0.434 0.229 0.190 1.000

Table 8.4. Coefficients of household characteristics as explanatory variables on diversity characteristics (Shannon, ln (species richness), Shannon evenness and ln (group abundance+1) for various use-groups of species after stepwise linear regression (n: number of observations). The significance of coefficients for explanatory variables (shown below characteristics between brackets) as provided by multiple linear regression where residuals were normally distributed (Kolmogorov-Smirnov test), otherwise calculated by permutation tests (indicated by *). Total explained variation is expressed by the multiple correlation coefficient R2, with probability P(F). Results of partial linear regression are presented for spatial (village) and non-spatial (household and farm) characteristics (with percentage of total explained variation below).

Group

Statistic (sample size)

St.dev. Range

R2

(P(F))

Ex-clusive spatial

variation

Ex-clusive non-

spatial variation

Intercept Ebuchie-be village (farthest

from forest)

Shimutu village (closest

to forest)

Mutambi village

Male headed

Female headed

(de jure) (no

husband)

Farm size

(acres) (1 acre = 0.4005

ha)

Perma-nent house (most

wealthy)

Thatch roofed house (least

wealthy)

Number of cross-

bred cattle

(wealth)


(wealth)

Years being head

Maxi-mum age head or partner

Number of

resident children

Highest level of schoo-ling of head

All H 0.49 0.20 0.06 0.11 1.62 -0.35 -0.22 -0.28 0.79 -0.75 0.25 -0.28 (n=183) 2.35 (6e-7) (0.33) (0.55) (0) (3e-4) (0.045) (0.003) (3e-4) (0.008) (0.121) (0.015) ln (S) 0.40 0.24 0.05 0.23 2.27 -0.15 -0.27 0.19 0.16 0.86 -0.32 0.23 0.34 (n=183) 2.44 (4e-8) (0.23) (0.94) (0) (0.020)* (6e-4)* (0.007)* (0.079)* (1e-5)* (0.139)* (0.097)* (0.010)* J 0.16 0.15 0.06 0.05 0.65 -0.07 -0.09 -0.16 -0.10 (n=183) 0.87 (5e-6) (0.44) (0.38) (0) (0.004)* (0.001)* (0.062)* (0.007)* ln (N+1) 0.84 0.20 0.06 0.16 5.55 -0.26 -0.31 0.24 0.29 1.53 -0.27 (n=183) 6.23 (2e-7) (0.28) (0.78) (0) (0.092)* (0.080)* (0.139)* (0.015)* (3e-5)* (0.042)* Beverage ‡ H 0.09 0.05 n/a n/a 0.02 0.12 (n=59) 0.56 (0.079) (0.065) (0.062)* ln (S) 0.28 0.08 0.07 0.04 0.09 -0.21 0.30 (n=59) 1.09 (0.071) (0.84) (0.49) (0.071) (0.032)* (0.093)* ln (N+1) 2.23 0.48 0.44 0.03 0.32 -0.67 3.24 1.23 0.86 (n=183) 8.03 (0) (0.92) (0.06) (0.171) (0.023)* (1e-5)* (0.062)* (0.034)* Boundary H 0.45 0.28 0.24 0.06 0.73 -0.62 -0.28 -0.22 -0.14 0.36 0.34 -0.16 (n=179) 1.67 (5e-10) (0.88) (0.21) (0) (0) (0.003) (0.008) (0.090) (0.036) (0.016) (0.137) ln (S) 0.55 0.28 0.25 0.04 1.05 -0.79 -0.38 -0.35 0.62 0.28 (n=179) 2.07 (2e-11) (0.90) (0.16) (0) (1e-5)* (7e-4)* (5e-4)* (0.003)* (0.094)* J 0.27 0.08 0.06 0.02 0.66 -0.17 -0.12 (n=146) 0.97 (0.001) (0.72) (0.28) (0) (0.001)* (0.047)* ln (N+1) 1.41 0.07 0.03 0.04 5.33 -0.52 -0.56 0.47 -0.45 -0.97 (n=183) 6.94 (0.014) (0.48) (0.60) (0) (0.034)* (0.026)* (0.105)* (0.056)* (0.047)* Charcoal H 0.36 0.06 n/a n/a 0.24 -0.20 (n=74) 1.42 (0.035) (0) (0.032)* ln (S) 0.48 0.06 n/a n/a 0.49 -0.42 (n=74) 1.94 (0.034) (0) (0.033)* J 0.33 0.44 n/a n/a 0.71 -0.55 (n=25) 0.98 (2e-4) (0) (7e-4)* ln (N+1) 1.57 0.11 0.04 0.07 0.47 -0.81 -0.60 1.89 0.63 -1.85 0.85 (n=183) 6.42 (0.001) (0.39) (0.66) (0.154) (0.004)* (0.072)* (0.011)* (0.117)* (0.059)* (0.136)*

Table 8.4 (cont.d)

Group


St.dev. Range

R2

(P(F))

Spatial variation

Non-spatial

variation

Intercept Ebuchie-be village

Shimutu village

Mutambi village

Male headed

Female headed

(de jure)

Farm size

(acres)

Perma-nent house

Thatch roofed house

Cross-bred cattle


Years being head

Maxi-mum age

Number of

children

Schoo-ling of head

Construction H 0.28 0.25 0.13 0.12 0.15 0.19 -0.08 0.41 0.24 (n=179) 1.25 (2e-10) (0.53) (0.51) (3e-4) (4e-5)* (0.068)* (4e-4)* (0.006)* ln (S) 0.36 0.25 0.10 0.16 0.33 0.27 0.18 0.13 0.50 0.39 -0.26 (n=144) 1.79 (4e-9) (0.41) (0.66) (0) (1e-5)* (0.005)* (0.103)* (1e-4)* (0.001)* (0.053)* J 0.30 0.16 0.09 0.07 0.59 0.12 0.14 -0.10 -0.15 0.11 (n=179) 0.98 (1e-4) (0.60) (0.43) (0) (0.067)* (0.044)* (0.142)* (0.002)* (0.152)* ln (N+1) 1.36 0.21 0.02 0.20 2.6314 -0.55 0.67 0.61 3.14 -1.19 0.85 (n=183) 6.95 (1e-7) (0.10) (0.97) (0) (0.029) (0.004) (0.048) (0) (0.111) (0.057) Firewood H 0.48 0.20 0.04 0.11 1.61 -0.28 -0.18 -0.23 0.78 -0.68 -0.30 (n=183) 2.22 (4e-7) (0.24) (0.58) (0) (0.002) (0.091) (0.014) (3e-4) (0.012) (0.009) ln (S) 0.40 0.22 0.05 0.21 2.29 -0.16 -0.25 0.22 0.22 0.82 0.28 (n=183) 2.37 (2e-8) (0.25) (0.93) (0) (0.013)* (9e-4)* (0.001)* (0.012)* (1e-5)* (0.028)* J 0.16 0.15 0.05 0.07 0.64 -0.07 -0.08 -0.16 -0.11 (n=183) 0.87 (6e-6) (0.36) (0.46) (0) (0.011)* (0.003)* (0.054)* (0.002)* ln (N+1) 0.85 0.20 0.05 0.17 5.32 0.46 0.30 1.33 -0.31 (n=183) 6.23 (2e-8) (0.25) (0.82) (0) (7e-4)* (0.013)* (4e-5)* (0.017)* Fodder ‡ H 0.19 0.11 n/a n/a 0.12 -0.12 -0.12 (n=43) 0.67 (0.095) (0) (0.098)* (0.059)* ln (S) 0.24 0.11 n/a n/a 0.16 -0.16 -0.16 (n=43) 0.69 (0.083) (0) (0.091)* (0.053)* ln (N+1) 0.83 0.18 0.13 0.06 0.39 -0.70 -0.78 -0.52 0.30 0.42 0.87 0.58 (n=183) 4.20 (6e-6) (0.71) (0.34) (0.034) (2e-5)* (1e-5)* (0.001)* (0.037)* (0.030)* (0.048)* (0.038)* Fruit H 0.47 0.14 0.03 0.05 0.94 -0.22 -0.38 0.29 0.37 (n=183) 1.90 (1e-5) (0.23) (0.39) (0) (0.009) (0.051) (0.051) (0.018) ln (S) 0.39 0.18 0.07 0.14 1.14 -0.28 -0.15 0.10 0.37 -0.20 0.20 0.27 0.35 (n=183) 2.30 (2e-5) (0.41) (0.78) (0) (2e-4)* (0.021)* (0.071)* (0.026)* (0.026)* (0.143)* (0.074)* (0.007)* J 0.25 0.17 0.03 0.10 0.72 0.11 0.09 -0.44 (n=182) 0.95 (1e-7) (0.21) (0.60) (0) (0.008)* (0.040)* (2e-5)* ln (N+1) 1.02 0.23 0.03 0.15 3.07 -0.49 -0.36 -0.42 2.44 -0.58 (n=183) 5.26 (6e-9) (0.14) (0.69) (0) (0.009) (0.090) (0.028) (0) (0.009) Timber H 0.39 0.09 0.04 0.04 0.51 -0.19 0.17 (n=179) 1.78 (1e-4) (0.46) (0.46) (0) (0.004) (0.004) ln (S) 0.59 0.14 0.04 0.10 0.82 -0.28 0.47 0.35 0.32 (n=179) 2.56 (1e-5) (0.28) (0.69) (0) (0.004)* (2e-5)* (0.012)* (0.106)* J 0.23 0.02 n/a n/a 0.47 -0.27 (n=160) 0.95 (0.042) (0) (0.042) ln (N+1) 1.49 0.13 0.01 0.13 3.43 -0.51 0.67 0.56 2.25 -0.44 (n=183) 6.95 (7e-5) (0.11) (0.99) (0) (0.069)* (0.012)* (0.108)* (5e-4)* (0.069)*

Table 8.4 (cont.d)

Group


St.dev. Range

R2

(P(F))

Spatial variation

Non-spatial

variation

Intercept Ebuchie-be village

Shimutu village

Mutambi village

Male headed

Female headed

(de jure)

Farm size

(acres)

Perma-nent house

Thatch roofed house

Cross-bred cattle


Years being head

Maxi-mum age

Number of

children

Schoo-ling of head

Medicine H 0.47 0.29 0.17 0.17 0.04 -0.22 -0.56 0.62 0.42 0.55 (n=104) 1.60 (1e-6) (0.58) (0.57) (0.702) (0.034)* (1e-5)* (0.006)* (0.025)* (0.002)* ln (S) 0.62 0.37 0.25 0.21 0.09 -0.46 -0.86 1.11 0.52 0.71 (n=104) 2.39 (4e-9) (0.67) (0.57) (0.55) (4e-4) (0) (1e-4) (0.023) (0.001) J 0.24 0.15 0.09 0.08 0.71 0.19 -0.15 0.13 (n=60) 0.86 (0.023) (0.64) (0.56) (0) (0.012)* (0.115)* (0.117)* ln (N+1) 1.30 0.25 0.21 0.09 0.40 -0.46 -1.07 0.87 0.27 1.98 -1.65 0.64 (n=183) 6.57 (7e-9) (0.86) (0.35) (0.086) (0.053)* (1e-4)* (5e-4) (0.133)* (4e-4)* (0.021)* (0.108)* Ornamental H 0.37 0.08 n/a n/a 0.41 -0.28 -0.21 -0.16 (n=90) 1.38 (0.049) (0) (0.010)* (0.062)* (0.103)* ln (S) 0.60 0.32 0.23 0.07 0.78 -0.82 -0.56 -0.43 0.21 0.42 -0.52 0.40 (n=90) 2.07 (2e-5) (0.73) (0.22) (0) (1e-5)* (0.002)* (0.002)* (0.136)* (0.026)* (0.124)* (0.018)* J 0.33 0.21 0.16 0.06 0.40 0.43 0.25 0.24 (n=48) 0.97 (0.012) (0.76) (0.30) (0) (0.008)* (0.085)* (0.059)* ln (N+1) 1.72 0.29 0.25 0.02 2.50 -2.09 -2.11 -1.21 -0.44 1.31 (n=183) 5.99 (6e-12) (0.87) (0.08) (0) (1e-5)* (1e-5)* (1e-4)* (0.081)* (0.117)* Shade H 0.60 0.13 0.08 0.07 0.43 -0.44 0.27 0.24 0.39 (n=151) 2.10 (2e-4) (0.59) (0.54) (8e-4) (2e-4)* (0.017)* (0.133)* (0.076)* ln (S) 0.74 0.14 0.08 0.09 0.61 -0.55 0.30 0.26 0.57 (n=151) 3.13 (9e-5) (0.53) (0.62) (1e-4) (2e-4)* (0.013)* (0.116)* (0.027)* J 0.23 0.14 0.01 0.12 0.75 0.07 -0.19 -0.09 -0.26 0.15 (n=115) 0.95 (0.004) (0.12) (0.89) (0) (0.134)* (0.067)* (0.115)* (0.033)* (0.114)* ln (N+1) 1.41 0.18 0.11 0.09 1.67 -0.80 -1.47 -0.51 -0.53 1.17 0.58 1.01 (n=183) 5.83 (8e-6) (0.60) (0.53) (0) (0.004)* (1e-5)* (0.067)* (0.056)* (0.055)* (0.074)* (0.041)* Soil Fertility H 0.28 0.10 0.01 0.09 0.31 0.09 -0.11 0.11 -0.18 -0.26 (n=123) 1.40 (0.027) (0.17) (0.96) (9e-4) (0.131)* (0.116)* (0.060)* (0.146)* (0.008)* ln (S) 0.42 0.08 0.05 0.05 0.40 0.23 0.13 -0.38 -0.24 (n=123) 1.79 (0.025) (0.58) (0.58) (0.002) (0.011)* (0.082)* (0.038)* (0.103)* J 0.32 0.13 n/a n/a 0.75 0.08 -0.22 -0.41 (n=123) 0.98 (0.180) (1e-4) (0.475) (0.359) (0.032) ln (N+1) 1.61 0.07 0.03 0.05 0.94 0.42 -0.53 0.45 1.62 (n=183) 7.24 (0.009) (0.53) (0.72) (0) (0.13)* (0.11)* (0.069)* (0.020)* ‡ Regression analysis not performed for J because of small sample size (beverage: n=10; fodder: n=6).

CHAPTER 8

129

Table 8.5. Average values for species richness (S), tree abundance (N) and Shannon diversity (H) based on 5000 random additions of farms of the same original size and provided by the models with prefixed parameter a (see text). Figures between brackets are the exact average species richness (see text). Statistic Original farm Average diversity for (combined) farm size size 0.25 0.5 0.75 1 1.5 2 2.5 3 4 5 S 0.25 (n=16) 12.38 21.04 26.05 29.68 36.46 41.72 46.83 51.38 59.00 n/a (12.37) (19.94) (25.50) (29.98) (37.10) (42.82) (47.68) (51.92) (59.00) 0.5 (n=32) 14.75 n/a 22.83 28.61 33.00 36.49 39.94 45.45 51.19 (14.75) (22.68) (28.39) (32.98) (36.88) (40.30) (46.18) (51.22) 0.75 (n=20) 14.60 n/a 22.44 n/a n/a 31.89 n/a n/a (14.60) (22.22) (27.54) 1 (n=37) 15.46 n/a 23.73 n/a 29.48 34.28 38.51 (15.46) (23.61) (29.55) (34.36) (38.43) 1.5 (n=25) 18.60 n/a n/a 27.20 n/a n/a (18.16) (27.46) 2 (n=26) 15.93 n/a n/a 26.96 n/a (17.31) (26.72) 2.5 (n=12) 18.25 n/a n/a 26.06 (18.17) (26.76) 3 (n=10) 19.91 n/a n/a (19.91) 4 (n=4) 21.75 n/a (21.75) 5 (n=2) 20.67 (20.67) Model 12.38 14.10 15.21 16.06 17.33 18.29 19.07 19.74 20.83 21.72 N 0.25 (n=16) 280 590 840 1081 1632 2196 2786 3387 4549 n/a 0.5 (n=32) 330 n/a 660 1014 1384 1672 1991 2581 3298 0.75 (n=20) 440 n/a 879 n/a n/a 1720 n/a n/a 1 (n=37) 490 n/a 983 n/a 1459 1953 2472 1.5 (n=25) 600 n/a n/a 1186 n/a n/a 2 (n=26) 500 n/a n/a 1000 n/a 2.5 (n=12) 510 n/a n/a 956 3 (n=10) 706 n/a n/a 4 (n=4) 680 n/a 5 (n=2) 1658 H 0.25 (n=16) 1.24 1.69 1.90 2.02 2.18 2.28 2.34 2.37 2.43 n/a 0.5 (n=32) 1.60 n/a 1.93 2.05 2.13 2.20 2.26 2.32 2.36 0.75 (n=20) 1.28 n/a 1.57 n/a n/a 1.87 n/a n/a 1 (n=37) 1.42 n/a 1.72 n/a 1.88 1.99 2.08 1.5 (n=25) 1.68 n/a n/a 1.91 n/a n/a 2 (n=26) 1.67 n/a n/a 1.95 n/a 2.5 (n=12) 2.09 n/a n/a 2.13 3 (n=10) 1.85 n/a n/a 4 (n=4) 1.77 n/a 5 (n=2) 1.45 Model 1.24 1.36 1.44 1.50 1.59 1.65 1.70 1.75 1.82 1.87

beve

rage

boun

dary

char

coal

cons

truct

ion

fert

ility

firew

ood

fodd

er

frui

t

med

icin

e

orna

men

tal

shad

e

timbe

r

Use groups

0

1

2

0

1

2

H

beve

rage

boun

dary

char

coal

cons

truct

ion

fert

ility

firew

ood

fodd

er

frui

t

med

icin

e

orna

men

tal

shad

e

timbe

r

Use groups

0

1

2

3

4

ln(S

)

1.000

2.718

7.389

20.086

54.598

S

Figure 8.1. Range of the Shannon diversity index over farms for different use-groups. The

reference line is drawn at ln(2). Figure 8.2. Range of species richness over farms for different use-groups. The reference line is

drawn at ln(2).

beve

rage

boun

dary

char

coal

cons

truct

ion

fert

ility

firew

ood

fodd

er

fruit

med

icin

e

orna

men

tal

shad

e

timbe

r

Use groups

0

1

0

1

J

beve

rage

boun

dary

char

coal

cons

truc

tion

ferti

lity

firew

ood

fodd

er

fruit

med

icin

e

orna

men

tal

shad

e

timbe

r

use groups

0

1

0

1

E

Figure 8.3. Range of the Shannon evenness (based on equal distribution) over farms for different use-groups. The reference line is drawn at 0.5

Figure 8.4. Range of the Shannon equality (based on the broken stick distribution) over farms for different use-groups. The reference lines are drawn at 0.5 and 1

beve

rage

boun

dary

char

coal

cons

truct

ion

ferti

lity

firew

ood

fodd

er

fruit

med

icin

e

orna

men

tal

shad

e

timbe

r

use groups

0

1

2

ln(1

/d)

1.000

2.718

7.389

1/d

beve

rage

boun

dary

char

coal

cons

truc

tion

ferti

lity

firew

ood

fodd

er

frui

t

med

icin

e

orna

men

tal

shad

e

timbe

r

use groups

0

1

2

3

4

5

6

7

8

ln(N

)

1.0

2.7

7.4

20.1

54.6

148.4

403.4

1,096.6

2,981.0

N

Figure 8.5. Range of the inverse Berger-Parker index over farms for different use-groups. The

reference line is drawn at ln(2). Figure 8.6. Range of tree abundance over farms for different use-groups

beve

rage

boun

dary

char

coal

cons

truc

tion

ferti

lity

firew

ood

fodd

er

fruits

med

icin

e

orna

men

tal

shad

e

timbe

r

Use groups

0

1

2

3

4

ln(S

+1)

1.000

2.718

7.389

20.086

54.598

S+1

beve

rage

boun

dary

char

coal

cons

truc

tion

ferti

lity

firew

ood

fodd

er

fruits

med

icin

e

orna

men

tal

shad

e

timbe

r

use groups

0

1

2

3

4

5

6

7

8

ln(N

+1)

1.0

2.7

7.4

20.1

54.6

148.4

403.4

1,096.6

2,981.0

N+1

Figure 8.7. Range of species richness for different use-groups, including those farms where the

group does not occur. The reference line is drawn at ln(3) (S=2 as in the other figures). Figure 8.8. Range of tree abundance for different use-groups, including those farms where the

group does not occur).

0 1 2 3 4

ln(S)

0

1

2

0

1

2

ln(1

/d)

0.2

0.5

0.7

1.0 1.2

1.2

1.5

1.5

1.5

1.7

1.7 1.7

1.7

1.9

2.2

Figure 8.9. Species richness versus inverse Berger-Parker index of individual farms and individual use-groups. The contour grid corresponds to the Shannon diversity index.

CHAPTER 8

133

0 1 2 3 4 5 6

Farm size (acres)

0

10

20

30

40

50

0

10

20

30

40

50S

pec

ies

rich

nes

s

Figure 8.10. Relationship between farm size and species richness. Full line is the fitted species-area relationship model (see text),

dotted lines the 95% confidence intervals. Symbol size corresponds to farm size (maximum 6).

0 1 2 3 4 5 6

Farm size (acres)

0

1

2

3

0

1

2

3

Sh

ann

on

div

ersi

ty in

dex

(lo

g e

)

Figure 8. 11. Relationship between farm size and Shannon diversity index. Full line is the fitted Shannon-area relationship model

(see text), dotted lines the 95% confidence intervals.

CHAPTER 9 COMPARISON OF TREE SPECIES COMPOSITION OF FARMS IN WESTERN KENYA USING CONSTRAINED ORDINATION I. ANALYSIS FOR SEPARATE USE-GROUPS

R KINDT, P VAN DAMME, AJ SIMONS & H BEECKMAN SUBMITTED TO JOURNAL OF APPLIED ECOLOGY

One of the objectives of tree domestication research is the diversification of the tree species composition of agroecosystems. We investigated whether substantial differences existed in the species composition of farms, so that diversification efforts could be guided by these differences – for example by introducing species that are dominant in one subsection of the landscape into subsections where these species are currently rare. Since we aimed to target diversification towards species that contributed to the 12 most important functions of trees on farms, we investigated various farms × species matrices that only included trees that provided the particular function on a particular farm. Ordinations of farms were obtained through Hellinger-distance-based linear and polynomial Redundancy Analysis (LRDA and PRDA), which all produced significant results expect LRDA for charcoal. Variance partitioning indicated significant influences of farm location, while socio-economic characteristics had significant influences on 7 LRDA and 9 PRDA ordinations. PRDA explained substantially more variation than LRDA (44% versus 18%). PRDA only explained 50% of the variation explained by unconstrained Principal Components Analysis on the first axis, believed to correspond to a “real-world” gradient. The relationship between farm location, farm characteristics and species composition were documented in various ordination diagrammes. These results were confirmed by multiple regression analyses using the abundance of dominant species as response variable, which implies that farms with a higher composition of a particular species also had higher abundance of this species. Despite the fact that unexplained variation implies that some farms differ in species composition with the majority of farms of the same type (same location or same socio-economic group), diversification can be guided for most farms by information on the dominant species of each type of farm. Diversification can either be achieved by introducing dominant species from other sections of the landscape, or by introducing rare species. The analysis would particularly be useful for groups of low average diversity that are dominated by only two species: construction, medicine, charcoal, beverage and fodder.

SPECIES COMPOSITION OF USE-GROUPS

136

9.1. Introduction

One of the objectives of tree domestication research in general and more specific in western Kenya is the diversification of tree species composition in agroecosystems (Kindt & Lengkeek 1999). Making agroecosystems ecologically more stable and economically more rewarding through diversification and intensification of land use by domestication of agroforestry trees is one of the three pillars of ICRAF’s research (ICRAF 1997; ICRAF 2000). This chapter provides an analysis of on-farm tree species composition.

The analysis of on-farm species composition had two objectives: to identify whether and where substantial differences in species composition exist between farms, and to test whether these differences are linked to characteristics of the respective farms. Where important differences in species composition exist, diversification could attempt to introduce new species in different sites. Where species composition is mainly related to characteristics of sites, diversification projects could be guided by these characteristics, allowing targeting specific types of households or areas.

9.2. Material and Methods

Formulation of use-groups (4.1), analysis of species composition (4.5)

9.3. Results

Figures 9.1 and 9.2 show ordinations of farms based on their species composition, while figures 9.3 - 9.14 provide ordinations for species belonging to particular use-groups. Table 9.1 provides a summary of the ordinations.

All ordinations were significant, except the PRDA for the charcoal group. PRDA explained substantially more variance than LRDA (the respective averages being 44% and 18% of explained variance). For eight groups, this resulted in two axes of PRDA explaining more variance than all the canonical variance explained by LRDA. Both methods, however, only explained a fraction of the total variance, what resulted in the first axis explaining substantially less variance than PCA. The average variance explained on the first axis of PRDA was only 50% of that of the first PCA axis. Two axes of PRDA (represented in the figures) explain on average 23% of variance – two axes of PCA explain 50% of variance on average.

For some PCA ordinations, the broken stick criterion revealed that many axes were significant. This points to data sets with a lot of variance that can only be represented on a large number of orthogonal axes. Whether this variance points to many ‘real world’ gradients, or whether these are distortion axes induced by the ordination method (see discussion) could not be resolved. The number of axes to be studied was related to the number of species differentiated in the diagrammes. Except for medicine, shade and soil fertility, the species number differentiated in the diagrammes equalled the number of broken stick axes plus 0-3.

Because the site scores on the first RDA and PCA axes were strongly correlated, those sites were arranged on a similar gradient. PCA calculates the hypothetical gradient that expresses most variance on the first axis. RDA calculates a hypothetical gradient in the same way, with the restriction that this gradient is calculated as a combination of environmental characteristics (‘direct gradient analysis’).

CHAPTER 9

137

Table 9.2 provides a summary of the variance partitioning of the various use-groups. Differences between villages were significant for all use-groups. For the charcoal, fodder, medicine, and soil fertility use-groups, using environmental characteristics alone did not result in significant relationships with PRDA ordination. For the timber and ornamental groups, using PRDA instead of LRDA made the ordination significant. The reverse was the case for the medicine group.

Polynomial versus classical RDA resulted in differences in Vj. Vj was only larger for PRDA for the beverage and boundary demarcation groups. Ordination for all species was the only case where Vj was smaller and positive. Vj of negative sign means that the two groups of environmental characteristics have effects of opposite signs, and that both environmental characteristics together explain the variance much better (Legendre & Legendre 1998 p. 533). Because of the higher amount of variance explained, the significance of the ordination, and the interaction terms between village and household characteristics that can occur in PRDA, all ordination diagrammes presented were the result of PRDA using all environmental characteristics – even where use of environmental characteristics alone was not significant. Another reason to present this type of ordination is that site scores that were presented were not fitted site scores based on environmental characteristics, but site scores that depended on species scores.

The following discussion describes the interpretation of the ordination diagrammes based on vector lengths and positions. Strong influence of environmental characteristics on species composition was only expected in ordination diagrammes where vectors lengths were long and angles between vectors were small. Only vectors corresponding to linear correlation were analyzed. Regression coefficients from multiple regression confirming or rejecting trends observed in diagrammes were mentioned. The complete list of regression coefficients obtained from stepwise regression is provided in table 9.3. Because some species had the same abundance distribution for different uses, the regression coefficients were only mentioned once and reference was made to the respective use-group.

Figures 9.1 and 9.2 show the ordination for all species and households. Figure 9.1 mainly highlights differences in species composition between Ebuchiebe, Shimutu, and Mutambi villages. Figure 9.2 features differences between Madidi and Mutambi on the vertical axis. Ebuchiebe is characterized by Euphorbia tirucalli (regression coefficients mentioned for the boundary demarcation use), Markhamia lutea (construction), Sesbania sesban (soil fertility), Grevillea robusta (timber) and Syzygium cuminii (fruit). Shimutu has more Psidium guajava (fruit), Tithonia diversifolia (boundary demarcation), Croton macrostachyus (firewood), Lantana camara (boundary demarcation), and Dracaena fragrans (boundary demarcation). Mutambi households contain more Coffea arabica (beverage), Buddleja davidii (boundary demarcation), and Camellia sinensis (beverage). Madidi features more Dracaena fragrans (like in Shimutu, see boundary demarcation), Buddleja davidii (like in Mutambi, see boundary demarcation) and Cupressus lusitanica (boundary demarcation). Not all correlations expected between species and household characteristics from the ordination could be confirmed. The relationship between farm size and Croton macrostachyus and Psidium guajava were confirmed (see discussions for firewood and fruit), but the one between farm size and Tithonia diversifolia not (firewood discussion). The positive link between the number of local cattle and Lantana camara observed in figure 9.2 could not be confirmed by regression analysis (ra=-0.46, P=0.60).

Figure 9.3 reveals that the beverage group only consisted of two species, Camellia sinensis and Coffea arabica . Mutambi village differs from the other villages as it has higher presence of Camellia sinensis (rs=1.79, P=0.0001) and Coffea arabica (rs=1.97, P=0.0001). Ebuchiebe village had less Coffea arabica (rs=-0.37, P=0.0001). Farms with more Coffea arabica had more de jure female heads (rs=0.81, P=0.001), more permanent houses (rs=1.05, P=0.001), and heads with higher levels of schooling (rs=0.85, P=0.01) and higher age (rs=1.13, P=0.02). Regression analysis could not confirm that heads longer in charge of the farm was linked to abundance of this species (ra=-0.57, P=0.41). Camellia sinensis was mainly linked to the number of crossbred cattle (rs=1.98, P=0.05)


138

and farm size (rs=1.13, P=0.02). Farms with less Camellia sinensis had more de jure female heads (rs=-0.68, P=0.06).

Figure 9.4 shows that species composition for the boundary demarcation use-group mainly differed for Buddleja davidii, Lantana camara, and Euphorbia tirucalli. The first species is mainly prominent in Mutambi and Madidi villages (negative correlations for the two other villages of rs=-2.18, P=0.0001 and rs=-2.01, P=0.0001 respectively). The species is also negatively related to farms with permanent houses (rs=-0.73, P=0.04). The second species seems to dominate Shimutu, although its vector length is not very long. The regression analysis only highlighted higher absence of the species in Ebuchiebe (rs=-1.50, P=0.0001) and Mutambi (rs=-0.60, P=0.07), but the centroid position of Madidi indicated higher presence of the species in this village as well. The species is positively related to farm size (rs=2.82, P=0.0002). Euphorbia tirucalli dominates Ebuchiebe (rs=1.96, P=0.0001) and is more absent in Shimutu (ra=-1.22, P=0.0009). Regression did not confirm the positive relation to the number of children (rs=0.60, P=0.41). Dracaena fragrans, Cupressus lusitanica, and Tithonia diversifolia only contributed marginally to the ordination (shorter vector lengths) indicating their higher presence in Shimutu. Regression results were only explicit for Tithonia diversifolia (rs=1.62, P=0.0001). Directions of vectors and regression coefficients (table 9.3) for D. fragrans and Cupressus lusitanica indicated higher presence in Madidi as well. Regression coefficients for Cupressus lusitanica show that the species dominates Madidi and only to a lesser degree Shimutu – this pattern can be observed in figure 9.10 where the species is well represented.

Figure 9.5 shows that farms mainly differ in charcoal group composition for Eucalyptus saligna and Markhamia lutea. Eucalyptus saligna is more dominant in Mutambi (rs=0.57, P=0.002). Markhamia lutea is more prominent in Madidi and less in Ebuchiebe (Ebuchiebe: rs=-0.80, P=0.002; Shimutu: rs=-0.56, P=0.06; Mutambi: rs=-0.50, P=0.06). Significance tests for this group (table 9.2) revealed that farm characteristics did not have a significant influence on ordinations. However, information on the diagramme for farm size, number of crossbred cattle, and age of household head agreed with partial regression results.

Figure 9.6 indicates that the two species dominating charcoal also dominate the construction group. Markhamia lutea is more prominent in Ebuchiebe (rs=1.14, P=0.0001) and is positively associated with farm size (rs=2.91, P=0.0001), permanent houses (rs=0.79, P=0.02), and the number of years that the household head was in charge of the farm (rs=0.83, P=0.06). A positive correlation with age (ra=0.62, P=0.31) could not be confirmed, but the age vector was short and had a wider angle with the Markhamia lutea vector. No major village differences existed for Eucalyptus saligna (nor were significant coefficients detected), but de facto female households heads had fewer trees of the species (rs=0.86, P=0.001 and rs=0.76, P=0.032 for male and de jure female headed households respectively).

Figure 9.7 shows the ordination of farms and species used for firewood. It is almost an identical ordination as for all species – most species have firewood as one of the given uses (Kindt et al. - Chapter 5). Ordination mainly shows species typical for Ebuchiebe, Shimutu, and Mutambi. In Ebuchiebe, Euphorbia tirucalli (boundary demarcation), Sesbania sesban (soil fertility), Syzygium cumini (fruit), Markhamia lutea (construction) and Grevillea robusta (timber) dominate. Species characterising Shimutu are Croton macrostachyus (rs=1.76, P=0.0001), Tithonia diversifolia (boundary demarcation) and Lantana camara (boundary demarcation). In Mutambi, typical species are Buddleja davidii (boundary demarcation), Coffea arabica (beverage) and Camellia sinensis (beverage). Cupressus lusitanica and Harungana madagascariensis are not well represented in the diagramme. Their vectors indicate that they occur less in Ebuchiebe (respectively boundary demarcation and rs=-1.03, P=0.0003). Table 9.3, however, indicates a low coefficient for Shimutu for Harungana madagascariensis. Apart from village characteristics, the only other environmental characteristic that features prominently in the diagramme is farm size (linked to Tithonia diversifolia, but ra=-0.09, P=0.89 and Croton macrostachyus, r s=1.11, P=0.01). Number of children (linked to Euphorbia tirucalli

CHAPTER 9

139

but ra=0.60, P=0.41) is represented by a smaller vector. The link between Lantana camara and number of local cattle was discussed for Figs 1 & 2.

The fodder group represented in figure 9.8 is another group with differences in composition for two species in first two RDA dimensions. The major difference detected was for Madidi village with more Sesbania sesban and Leucaena leucocephala. Table 9.3 confirms these differences with significant negative regression coefficients for all other villages. The ordination was not significant for farm characteristics. Number of crossbred cattle however had a position that was confirmed by regression (rs=1.07, P=0.01).

Figure 9.9 mainly highlights differences in species composition for fruit between Ebuchiebe and Shimutu. Syzygium cumini and Mangifera indica are more prominent in Ebuchiebe, while Psidium guajava is more prominent in Shimutu. The respective coefficients for Ebuchiebe and Shimutu confirmed this trend (table 9.3). Carica papaya was another species contributing to the ordination diagramme. The negative association of this species with number of crossbred cattle could not be confirmed (ra=-0.27, P=0.12). Larger size farms have more Psidium guajava (rs=2.82, P=0).

Figure 9.10 shows that differences in species composition in the timber group are mainly characterized by differences in Markhamia lutea, Cupressus lusitanica, and Zanthoxylum gillettii. Species that contributed less are Grevillea robusta, Syzygium cumini, and Eucalyptus saligna. Ebuchiebe is characterized by Markhamia lutea (rs=1.02, P=0.0001), Grevillea robusta (rs=0.40, P=0.003) and Syzygium cumini (fruit). Mutambi is characterized by Zanthoxylum gillettii (rs=0.39, P=0.001). Madidi features more Cupressus lusitanica (see table 9.3 where the other villages have negative coefficients). When considering household characteristics, the positive influence of permanent houses (rs=2.09, P=0.0001) and negative influence of thatch roofed houses (rs=-0.49, P=0.12) on Cupressus lusitanica is prominent. At the level of associations of the species with age of household head, farm size and number of local cattle, only the first relationship could be confirmed (respectively rs=-0.49, P=0.12, ra=1.11, P=0.21 and ra=0.37, P=0.69). The latter environmental characteristic, however, had a very short vector, while the first variable had the longest vector length.

Figure 9.11 shows that species composition for medicine mainly differed for Harungana madagascariensis and Azadirachta indica . The ordination shows higher presence of the first species in Mutambi (rs=0.94, P=0.0001) and lower presence in Ebuchiebe (rs=-0.41, P=0.07). Household characteristics did not have a significant influence on ordination (table 9.2), while vectors were short. However, partial regression confirmed association of Azadirachta indica with age of household head (rs=0.30, P=0.06) and number of children (rs=0.31, P=0.04).

The ornamental use-group is presented in figure 9.12. Only three species had significant contributions to the diagramme: Terminalia mantaly, Buddleja davidii and Cupressus lusitanica . The main differences highlighted in the diagramme are differences between Ebuchiebe, Shimutu, and Madidi. The first two villages feature less Buddleja davidii and Cupressus lusitanica , while the opposite phenomenon can be observed for Madidi. Regression coefficients (table 9.3) confirmed this observation. Positive correlation of Terminalia mantaly with permanent houses could not be confirmed (ra=0.14, P=0.53). The species was positively related to the number of crossbred cattle (rs=0.51, P=0.03), which is another indicator of wealthier households, and the highest level of schooling of the household head (rs=0.22, P=0.04).

Figure 9.13 shows the ordination for the shade group. Three species feature prominently. Mangifera indica is positively correlated with Madidi (negative correlations for other villages, table 9.3). Croton macrostachyus is associated with Mutambi, although regressions could not confirm this observation (ra=0.14, P=0.34). Shimutu has more Bischofia javanica (rs=0.18, P=0.03). Of species that were not well represented, Madidi featured more Croton megalocarpus and Syzygium cumini (negative correlations for other villages, table 9.3), Mutambi and Madidi more Psidium guajava and Eriobotrya japonica (negative correlations for the other villages, table 9.3), while Mutambi had more Zanthoxylum gillettii (rs=0.21, P=0.0001). A positive relationship between Bischofia javanica and farm


140

size could not be confirmed (ra=0.003, P=0.99), while the expected negative relationship with Mangifera indica had an opposite regression coefficient (rs=0.54, P=0.03).

Figure 9.14 shows the ordination for the soil fertility enhancement group. The major gradient for villages can be detected for Sesbania sesban, with higher presence in Ebuchiebe (rs=0.55, P=0.04) and Madidi (other village coefficients), and lower presence in Mutambi (rs=-0.41, P=0.002) and Shinyalu (rs=-0.97, P=0.002). Species that were more dominant in Mutambi were Coffea arabica (rs=0.73, P=0.0001), Carica papaya (rs=0.22, P=0.003), Ricinus communis (not confirmed with ra=0.06, P=0.37) and Camellia sinensis (rs=0.55, P=0.001). Tephrosia vogelii featured more in Shimutu (rs=0.30, P=0.002). Farm characteristics were not significant (table 9.2). However, the negative relationship between years that the household head was in charge and Sesbania sesban was confirmed (rs=-0.80, P=0.08).

9.4. Discussion

9.4.1. Ordination Analysis Results

Where species respond to the same gradients, relationships in the data can be represented adequately in a few dimensions (Økland 1996). The many axes that would need to be analyzed with unconstrained ordination for some groups (e.g. all species, firewood, and shade) could express that many gradients are present in these data. Økland (1999) however warned that (polynomial) distortion axes may occur because species never meet all assumptions of the species response models, and also pointed out that noise axes might occur as well. Therefore, not all axes necessarily express true gradients, and an ordination diagramme reflecting a small amount of variance may be relevant.

Distortion and noise axes result in normal percentages of variance explained ranging 20-50% (ter Braak & Smilauer 1998 p. 121; Økland 1999). This makes it difficult to assess which fraction of remaining variance in the data could be explained by unmeasured environmental variables. However, if we hypothesise that the first axis calculated by PCA is a structure axis, then we could assess whether additional environmental variables could potentially explain more variance by the difference in variance explained on the first PCA (constrained) and RDA (unconstrained) axis. High correlations between RDA and PCA site scores indicate that the first axis describes the same gradient. Another method by which gradient differences were tested was by looking at species that contributed significantly to the first two PCA axes. As indicated in table 9.3, most species that contributed to the ordination by PRDA also contributed to the PCA ordination. Some species that did not contribute were highlighted in the tables. In general, more species contributed to the RDA than to the PCA ordination diagrammes. Only three species contributed to PCA ordinations on the first two axes and not to RDA ordinations: Bridelia micrantha for firewood, Eriobotrya japonica for fruit and Terminalia mantaly for shade. The first two species only contributed marginally to PCA ordinations. Because gradients expressed on first PCA and RDA axes were largely similar, additional environmental variables could allow RDA axes to reconstruct the PCA axes better. Despite the complexity caused by potential noise and distortion axes, it is thus still possible to assess potential additional amounts of variance explained by additional sets of explanatory variables by investigating how well the dominant gradient revealed by PCA is reconstructed by RDA.

As indicated by Økland (1999), distortion and noise axes pose no problems to the variance partitioning method. Neither do these axes cause problems to compare the variance explained by LRDA versus PRDA. PRDA explained higher amounts of variance and the method was able to reconstruct the first PCA axes better, thus better explaining the dominant gradient in the data. Village differences alone were significant for all ordinations, while farm and household

CHAPTER 9

141

characteristics alone were significant for the majority of ordinations. Because of negative Vj where only village differences were significant, ordination provided by using all environmental characteristics was still appropriate for all ordination analyses (more variance was explained).

Multiple regression explained low percentages of variance (table 9.3). The highest amount of variance explained was for Coffea arabica for beverage (49%). Low amounts imply that single species were not strongly related to explanatory variables and that a lot of unexplained variance remained. This resulted in difficulties in reconstructing PCA axes in the dataset, even where an axis is mainly correlated with the same single species of which nearly all variance is explained (e.g. ordination for beverage, charcoal and soil fertility).

The trends from ordination diagrammes were confirmed by multiple linear regression for most strong effects (where vector lengths were longer, or with centroids far from the origin). This result also means that the Hellinger transformation applied to the data did not effect the results very much, meaning that farms with higher abundance of the species also had higher frequency

+x

xi

nn

(see methods) of the species. Another consequence of this result is that the ordination diagrammes explained meaningful amounts of variance, although these were small if expressed in fractions of the total variance present.

Jarvis et al. (2000) state that gender, age, wealth or social status affect farmers’ knowledge, actions and access to resources regarding the maintenance of crop diversity. Long et al. (2000) provide some examples of the influence of wealth, age and gender on crop diversity. Jarvis et al. (2000) also listed farm size, family size, and years of education as explanatory factors for farmer variety choices. Our results show that variables that influence crop choices (wealth, age, gender, farm size, family size, and education) also influenced on-farm tree composition.

In many ordination diagrammes, the vector representing the multiple linear correlation had a different angle from the vector representing the linear correlation. One example is the difference between the schooling vectors in figure 9.1. These differences and the higher amount of variance explained by PRDA imply that significant second-order relationships between species composition and farm characteristics exist. However, considering these factors adds to the complexity of the analysis (see next section).

9.4.2. Ordination for Diversification

As indicated above, environmental characteristics had a significant effect on differences in species composition on farms. Interpretation of ordination diagrammes and multiple regression analysis revealed environmental characteristics with large and significant effects. However, the amount of variance explained was still small. This was also true when investigating the first axis expected to be a structural gradient. These results are analogous with those from Kindt et al. (Chapter 8) investigating differences in diversity and abundance among use-groups. Environmental characteristics can be used as a guide to plan domestication activities based on differences in species composition. However, some farms will differ in composition from the majority of farms with the same characteristics. For instance, although for boundary demarcation most farms of Ebuchiebe were dominated by Euphorbia tirucalli, some farms had a composition that was more typical of Shimutu or Madidi. For a single species (e.g. Euphorbia tirucalli as a boundary marker), large amounts of unexplained variance remained as well. Exceptional cases, however, do not prevent targeting of interventions aiming at diversifying the agricultural landscape. For the example of Ebuchiebe, interventions can focus on diversifying from Euphorbia tirucalli, which could be relevant for the majority of, but not all, farms in the village.


142

PRDA included second-order relationships between environmental characteristics and ordination axes. Considering second-order effects (for instance the relationship between Euphorbia tirucalli and the level of schooling) explains differences in species composition better, but adds to the complexity of the design of interventions.

Diversification can either involve distributing dominant species found in one group of households to another one, or making rarer species more dominant. Both approaches increase alpha diversity (the average diversity of one farm) in a similar way, but have alternative effects on beta diversity (the differences between farms as observed by the changing slope of the species-area curve, Kindt et al. – Chapter 6). The result of distributing dominant species more widely on species accumulation curves pattern will be larger species richness values for few randomly accumulated farms, than the values of the curve corresponding to the original distribution. Making less frequent species more dominant will result in the two species accumulation curves to coincide at a larger number of farms (Kindt, unpublished data). The latter approach results in a larger number of accumulated farms showing a greater species richness. It is possible that the greater richness will influence the stability of the agroecosystem, due to the positive relationship between diversity and ecosystem stability under conditions of heterogeneity in species’ traits and environmental characteristics (Kindt et al. – Chapter 2, 5 & 6). However, a decision will need to be made by farmers on appropriateness of increasing more dominant or more rare species. The first approach could have the advantage that the species has better tested traits making it suitable for the particular use. It is also possible that a rare suitable species was overexploited. The two approaches reflect the two major pathways in landscape diversification as described by Van Noordwijk et al. (1997). The integration approach is mainly an approach of increasing alpha diversity, while the segregation approach focuses more on beta diversity.

Analysis of species composition supplements analysis of alpha diversity of use-groups. The construction, medicine, charcoal, beverage, fodder, ornamental and soil fertility groups were those that were identified with alpha diversity lower than two (Kindt et al. – Chapter 5 & 6). Ordination diagrammes of the first five groups showed that only two species significantly contributed to the ordinations. The dominant species approach would involve wider distribution of both species. Farms with lower presence of either species can be identified in the diagrammes and could be targeted when distributing the other species. The rare species approach would involve selecting species that did not contribute to the ordination and distributing them more widely, or to introduce a new species. The same farms as identified for the dominant approach (farms dominated by a single species) could be targeted.

CHAPTER 9

143

Table 9.1. Summary information on the ordination of all species and various use-groups. Variance is expressed as percentage of total variance.

Ordination Sp Sp (2)

Var PRDA

Sign (99)

Var LRDA

Sign (9999)

Ax Var PCA (1)

Var LRDA

(1)

Var PRDA

(1)

Cor. PCA- RDA

Var PCA (2)

Var PRDA

(2) All species 70 15 42.23 0.01 23.52 0.0001 12 12.17 8.62 9.60 0.99 20.64 15.74 Beverage 2 2 50.33 0.01 34.26 0.0001 0 63.32 29.68 38.58 0.99 100.0 50.33 Boundary

demarcation 27 6 48.44 0.01 30.30 0.0001 5 30.71 16.78 19.79 0.99 47.74 28.13

Charcoal 21 2 44.27 0.21 11.56 0.0108 1 50.47 6.88 14.92 -0.99 61.23 21.05 Construction 18 2 42.49 0.02 13.67 0.0035 2 61.80 9.79 20.17 0.99 81.14 29.52 Firewood 68 13 43.21 0.01 24.01 0.0001 12 12.49 8.68 9.53 0.98 21.81 16.07 Fodder 7 2 48.41 0.01 14.14 0.0003 0 32.54 9.23 18.20 0.85 60.50 30.66 Fruit 13 4 36.01 0.01 14.97 0.0001 1 24.86 7.04 10.28 -0.91 46.62 16.92 Timber 31 6 41.27 0.01 19.63 0.0001 5 23.77 8.23 11.42 0.99 37.96 18.36 Medicine 26 2 43.51 0.01 10.59 0.0040 4 22.82 4.14 8.79 0.89 42.54 15.57 Ornamental 31 3 47.60 0.01 11.78 0.0006 3 25.36 5.66 10.50 0.83 43.29 19.04 Shade 50 8 42.21 0.01 12.27 0.0001 9 13.35 3.94 6.51 -0.93 23.55 10.78 Soil fertility

enhancement 22 6 45.84 0.01 17.17 0.0001 1 53.53 12.07 20.06 0.99 61.99 24.41

(Sp: number of species in the site-species matrix analyzed; Sp (2): number of species included in the ordination diagramme with the first two axes; Var PRDA/LRDA/PCA (y): variance explained by polynomial RDA/linear RDA/PCA (for first y axes); Sign(99/9999): significance of ordination of lefthand column for 99/9999 permutations; Ax: number of axes to consider based on the broken stick distribution; Cor. PCA – RDA: correlation between the site scores on the first PCA and PRDA axis)

Table 9.2. Variance partitioning between village and household characteristics for all species and various use-groups. Variance is expressed as percentage of total variance.

Ordination Vv LRDA/PRDA

Sign (9999)

Vh LRDA

Sign (9999)

Vh PRDA

Sign (99)

Vve LRDA

Vve PRDA

Vhe LRDA

Vhe PRDA

All species 16.22 0.0001 10.20 0.0001 27.56 0.01 13.32 14.67 7.30 26.01 Beverage 25.61 0.0001 13.19 0.0014 33.26 0.01 21.07 17.07 8.65 24.72 Boundary demarcation 22.11 0.0001 12.81 0.0001 32.04 0.01 17.49 16.40 8.19 26.33 Charcoal 5.97 0.0001 5.99 0.4723 33.67 0.27 5.57 10.60 5.59 38.30 Construction 6.14 0.0007 8.00 0.1180 33.83 0.02 5.67 8.66 7.53 36.35 Firewood 7.63 0.0001 10.46 0.0001 28.21 0.01 13.55 15.00 16.38 35.58 Fodder 8.34 0.0001 5.22 0.7247 28.43 0.48 8.92 19.98 5.80 40.07 Fruit 7.63 0.0001 7.53 0.0674 24.79 0.06 7.44 11.22 7.34 28.38 Timber 9.61 0.0001 5.69 0.6370 28.74 0.01 13.94 12.53 10.02 31.66 Medicine 4.80 0.0001 10.39 0.0001 30.72 0.21 0.2 12.79 5.79 38.71 Ornamental 6.01 0.0001 6.49 0.2854 35.31 0.06 5.29 12.29 5.77 41.59 Shade 5.88 0.0001 7.70 0.0059 31.35 0.01 4.57 10.86 6.39 36.33 Soil fertility enhancement 11.96 0.0001 5.36 0.6559 30.49 0.20 11.81 15.35 5.21 33.88

(V v/h (e) LRDA/PRDA: (exclusive) linear/polynomial variance explained by village/household characteristics Sign(99/9999): significance of ordination of lefthand column for 99/9999 permutations)

Table 9.3. Coefficients of village and household characteristics as explanatory variables on species abundance information for various use-groups of species after stepwise linear regression. The significance of coefficients for explanatory variables (shown below characteristics between brackets) as provided by multiple linear regression (indicated by *) where residuals were normally distributed, otherwise calculated by permutation tests. Total explained variance is expressed by the multiple correlation coefficient R2, with probability P(F).

Ordination

Species (other use-groups where the species occurred with the same abundance pattern over households)

R2

(P(F))

Inter-cept

Ebuchie-be village (farthest

from forest)


to forest)

Mutambi village

Male headed

Female headed (de jure)

(no husband)

Farm size (acres)

(1 acre = 0.4005

ha)


wealthy)


wealthy)

Number of cross-

bred cattle

(~wealth)


(~wealth)

Years being head


Number of

resident children

Level of schoo-ling of head

Beverage Camellia sinensis 0.217 -0.04 1.79 -0.68 1.42 1.98 (firewood, all) (6e-9) (0.870)* (0.0001) (0.0571) (0.0449) (0.0547) Coffea arabica 0.491 -0.72 -0.37 1.97 0.81 1.05 -1.57 1.13 0.85 (firewood, all) (0) (0.041)* (0.0748) (0.0001) (0.0010) (0.0014) (0.0355) (0.0162) (0.0135) Boundary Buddleja davidii ‡ 0.394 2.33 -2.18 -2.01 0.44 -0.73 -0.46 -0.87 1.03 demarcation (firewood, all) (2e-16) (0)* (0.0001) (0.0001) (0.1546) (0.0446) (0.0776) (0.1135) (0.0492) Dracaena fragrans 0.192 1.34 -1.88 -1.23 1.77 (all) (3e-8) (0)* (0.0001) (0.0003) (0.0049) Cupressus lusitanica 0.275 1.51 -1.78 -0.74 -1.30 2.09 -0.49 1.91 (firewood, timber, all) (1e-10) (0.001)* (0.0001) (0.0487) (0.0006) (0.0001) (0.1181) (0.0038) Lantana camara 0.242 2.78 -1.50 -0.60 2.82 -1.06 -0.52 -1.33 -1.11 (firewood, all) (2e-8) (0)* (0.0001) (0.0737) (0.0002) (0.0152) (0.0908) (0.0431) (0.1183) Tithonia diversifolia ‡ 0.289 0.40 1.62 -0.50 -0.47 (firewood, all) (3e-13) (0.029)* (0.0001) (0.0189) (0.1281) Euphorbia tirucalli 0.266 1.95 1.96 -1.22 0.46 (firewood, all) (5e-12) (0)* (0.0001) (0.0009) (0.1239) Charcoal Eucalyptus saligna 0.124 -0.41 0.57 0.80 0.69 -0.77 0.59 (3e-4) (0.192)* (0.0021) (0.0573) (0.0975) (0.0335) (0.0357) Markhamia lutea 0.128 0.44 -0.80 -0.56 -0.50 0.89 0.79 -1.72 0.89 (0.001) (0.125)* (0.0019) (0.0576) (0.0530) (0.1344) (0.0146) (0.0300) (0.0572) Construction Markhamia lutea 0.284 0.85 1.14 2.91 0.52 -1.88 0.83 (firewood, all) (2e-11) (1e-5)* (0.0001) (0.0001) (0.1217) (0.0182) (0.0580) Eucalyptus saligna 0.156 2.17 0.86 0.76 2.02 0.95 (timber) (4e-6) (0)* (0.001)* (0.032)* (4e-4)* (0.06)* Firewood Croton macrostachyus 0.475 -0.10 1.76 0.42 1.11 1.26 -0.57 0.58 (all) (0) (0.742) (0.0001) (0.0216) (0.0112) (0.0018) (0.0942) (0.0332) Harungana madagascariensis 0.301 0.23 -1.03 -0.82 0.72 0.28 3.57 0.45 -1.64 (all) (2e-11) (0.328)* (0.0003) (0.0071) (0.0077) (0.1565) (0.0001) (0.0422) (0.0335) Fodder Sesbania sesban 0.100 0.39 -0.29 -0.39 -0.33 1.07 (8e-4) (4e-4)* (0.0223) (0.0023) (0.0128) (0.0143) Leucaena leucocephala 0.150 0.32 -0.33 -0.40 -0.35 0.28 (7e-6) (0)* (0.0001) (0.0001) (0.0001) (0.0756) ‡: Species that did not contribute significantly to the first two PCA axes

Table 9.3 (cont.d) Ordination

Species R2

(P(F))

Inter-cept

Ebuchie-be

Shimutu Mutam-bi

Male headed


Farm size Perma-nent house

Thatch roofed house

Number of cross-

bred cattle


Years being head


Number of

resident children


Fruit Carica papaya 0.084 0.53 0.28 -0.64 0.67 (0.001) (0.003)* (0.0449) (0.0016) (0.0256) Syzygium cumini ‡ 0.287 0.42 -0.93 -0.51 0.18 1.06 -0.68 (firewood ‡, timber ‡, all ‡) (1e-11) (0)* (0.0001) (0.0001) (0.0434) (0.0005) (0.0567) Mangifera indica 0.191 0.96 -0.63 -0.36 1.09 -0.24 -0.24 (3e-7) (0)* (0.0001) (0.0011) (0.0001) (0.0961) (0.1112) Psidium guajava 0.216 2.06 -0.71 -0.64 2.82 (all ‡) (2e-9) (0)* (0.004)* (0.011)* (0)* Timber Grevillea robusta 0.165 -0.32 0.40 -0.27 0.52 0.37 0.68 0.40 (firewood, all) (1e-5) (0.177)* (0.0033) (0.0485) (0.0791) (0.0042) (0.0263) (0.0469) Markhamia lutea 0.195 -0.07 1.02 0.55 -0.57 -1.05 0.84 0.69 (8e-7) (0.765)* (0.0001) (0.0071) (0.0174) (0.0350) (0.0378) (0.0921) Zanthoxylum gillettii ‡ 0.191 -0.08 -0.12 0.39 0.44 0.29 (1e-7) (0.438)* (0.1461) (0.0001) (0.0196) (0.1033) Medicine Harungana madagascariensis 0.210 0.16 -0.41 -0.61 0.94 1.74 -1.13 (6e-8) (0.349)* (0.0682) (0.0201) (0.0001) (0.0025) (0.0834) Azadirachta indica 0.041 -0.05 0.30 0.31 (0.023) (0.596)* (0.0571) (0.0407) Ornamental Terminalia mantaly 0.053 0.10 0.51 0.22 (0.007) (0.066)* (0.0308) (0.0413) Buddleja davidii 0.151 0.92 -1.20 -1.13 -0.45 0.70 (1e-5) (1e-4)* (0.0001) (0.0001) (0.0852) (0.1083) Cupressus lusitanica ‡ 0.183 1.35 -1.27 -1.20 -0.96 -0.39 1.37 (9e-7) (0)* (0.0001) (0.0001) (0.0003) (0.0736) (0.0494) ‡: Species that did not contribute significantly to the first two PCA axes


Species R2

(P(F))

Inter-cept

Ebuchie-be

Shimutu Mutam-bi

Male headed



Thatch roofed house

Number of cross-

bred cattle


Years being head


Number of

resident children


Shade Mangifera indica 0.231 0.91 -0.46 -0.87 -0.53 -0.30 0.54 -0.32 (2e-8) (0)* (0.0003) (0.0001) (0.0001) (0.0114) (0.0278) (0.0309) Croton megalocarpus 0.116 0.46 -0.44 -0.57 -0.29 0.23 0.26 -0.64 (0.001) (6e-5)* (0.0009) (0.0001) (0.0404) (0.0217) (0.1319) (0.1121) Syzygium cumini ‡ 0.180 0.34 -0.31 -0.51 -0.35 0.14 0.32 0.20 -0.51 -0.24 (2e-5) (4e-4)* (0.0005) (0.0001) (0.0002) (0.0386) (0.1042) (0.0706) (0.0671) (0.1152) Psidium guajava ‡ 0.090 0.06 -0.11 -0.16 0.23 0.11 -0.31 (0.005) (0.028)* (0.0038) (0.0021) (0.0375) (0.0642) (0.0413) Eriobotrya japonica ‡ 0.148 -0.17 -0.23 -0.34 0.42 0.47 0.30 (3e-5) (0.259)* (0.0008) (0.0014) (0.0501) (0.0178) (0.0238) Zanthoxylum gillettii 0.173 -0.03 0.21 0.10 -0.31 0.27 (7e-7) (0.610)* (0.0001) (0.0181) (0.0114) (0.0077) Croton macrostachyus 0.097 -0.34 0.83 0.96 -0.45 0.65 (0.001) (0.194)* (0.0387) (0.0080) (0.1577) (0.0069) Bischofia javanica 0.115 0.166 0.18 -0.16 0.15 0.52 -0.25 (6e-4) (0.087)* (0.0265) (0.0609) (0.0244) (0.0413) Soil fertility Ricinus communis 0.045 0.21 0.24 -0.20 -0.18 (0.04) (0.011)* (0.0676) (0.0951) (0.0240) Carica papaya ‡ 0.082 0.23 0.22 0.11 -0.37 -0.21 (0.004) (0.032)* (0.0028) (0.0654) (0.0190) (0.0626) Coffea arabica ‡ 0.143 -0.03 0.73 0.37 (1e-6) (0.637)* (0.0001) (0.0573) Camellia sinensis ‡ 0.276 0.01 0.55 0.85 0.35 2.61 -0.57 (4e-11) (0.977)* (0.0015) (0.0329) (0.1407) (0.0023) (0.0843) Tephrosia vogelii ‡ 0.082 0.02 0.30 -0.29 0.20 (0.002) (0.8450)* (0.0016) (0.1022) (0.1297) Sesbania sesban 0.195 1.32 0.55 -0.97 -0.41 0.29 1.46 -0.80 -0.60 (firewood, all) (2e-6) (2e-5)* (0.0364) (0.0022) (0.1218) (0.1455) (0.0091) (0.0762) (0.0780) ‡: Species that did not contribute significantly to the first two PCA axes

CHAPTER 9

147

Ebu

Shi

Mut

Madfedjfedftha

-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6

Axis I (9.6 % tot. var., 22.8 % can. var., spec-env. corr. 0.89)

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6A

xis

II (6

.1 %

tot.

var

., 14

.7 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.88)

size

ncros

nloc

heady

age

nchil

school

size1

nloc1heady1

age1

nchil1

school1


buddav

camsin

cofrob

cromac

cuplusdrafra

euptir

grerob

harmad

lancam marlutpsigua

sessessyzcum

titdiv

Figure 9.1. Ordination plot for all species. % tot. var.: percentage of total variance explained on the axis; % can. var.: percentage of canonical variance explained on the axis; spec-env. cor.: species-environment correlation.

EbuShi

Mut

Mad

fedjfedftha

-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6


-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

Axi

s III

(4.3

% to

t. v

ar.,

10.1

% c

an. v

ar.,

spec

-en

v. c

orr

. 0.8

2)

brimic

buddav

camsin

cofrob

cromac

cuplusdrafra

euptir

grerob

harmad

lancam

marlut

sesses

syzcum

titdiv

size

ncros

nloc

heady

age

nchil

school

size1

nloc1

heady1

age1

nchil1

school1


Figure 9.2. Ordination plot for all species. Abbreviations as in figure 9.1.


148

-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1


-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Axi

s II

(11.

7 %

to

t. va

r., 2

3.3

% c

an. v

ar.,

spec

-en

v. c

orr

. 0.5

6) camsin

cofspp

size

ncros

nloc

heady

age

nchil

school

size1

ncros1

nloc1

heady1

age1

nchil1school1


Ebu Shi

Mut

Madmale

fedj

fedf permirontha

Figure 9.3. Ordination plot for the beverage use-group. Largest symbol size (Ebuchiebe village) corresponds to 47 farms and origin before species centring. Abbreviations as in figure 9.1.

EbuShi

MutMad

fedjfedfperm

-0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9


-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Axi

s II

(8.3

% t

ot.

var

., 17

.2 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.76)

buddav

cuplusdrafraeuptir

lancamtitdiv

size

nloc

age

nchilschool

size1

nloc1

age1

nchil1


Figure 9.4. Ordination plot for the boundary demarcation use-group. Largest symbol size (Ebuchiebe village) corresponds to 18 farms. Abbreviations as in figure 9.1.

CHAPTER 9

149

-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1


-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Axi

s II

(6.1

% t

ot.

var

., 13

.8 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.75) eucsal

marlut

sizencros

nloc

heady

age

nchilschool

size1

ncros1

nloc1

age1


Ebu

Mut

Madperm

Figure 9.5. Ordination plot for the charcoal use-group. Largest symbol size (Ebuchiebe village) corresponds to 33 farms and origin before species centring. Abbreviations as in figure 9.1.

-0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0


-1.0

-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Axi

s II

(9.4

% t

ot.

var

., 22

.0 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.64)

eucsal

marlut

size

ncros nlocheady

age

school

size1heady1

age1

school1


EbuMutMadfedj

fedfperm

Figure 9.6. Ordination plot for the construction use-group. Largest symbol size (Mutambi and Madidi village) corresponds to 11 farms. Abbreviations as in figure 9.1.


150

Ebu

Shi

MutMad

fedjfedftha

-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7


-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Axi

s II

(6.5

% t

ot.

var

., 15

.1 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.85)

buddav

camsin

cofrob

cromac

cuplus

euptir

grerob

harmad

lancammarlut

sessessyzcum

titdiv

size

ncros

nloc

headyage

nchil

school

size1

nloc1heady1

age1

nchil1

school1


Figure 9.7. Ordination plot for the firewood use-group. Abbreviations as in figure 9.1.

-0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9


-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Axi

s II

(12.

5 %

to

t. v

ar.,

25.7

% c

an. v

ar.,

spec

-en

v. c

orr

. 0.6

8)

leuleu

sesses

size

ncros

ncros1


Mut Mad

Figure 9.9. Ordination plot for the fodder use-group. Largest symbol size (Ebuchiebe village) corresponds to 43 farms and origin before species centring. Abbreviations as in figure 9.1.

CHAPTER 9

151

-0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7


-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

Axi

s II

(6.6

% t

ot.

var

., 18

.4 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.57)

carpap

manind

psiguasyzcum

sizencros

nloc

heady

nchil

size1ncros1


EbuShi

Figure 9.9. Ordination plot for the fruit use-group. Largest symbol size (Shimutu village) corresponds to 2 farms. Abbreviations as in figure 9.1.

-1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5


-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

Axi

s II

(6.9

% t

ot. v

ar.,

16.8

% c

an. v

ar.,

spec

-env

. cor

r. 0

.74)

cuplus

eucsal

grerob

marlut

syzcum

zangil

size

ncros

nloc

headyage

nchil school

size1

ncros1

nloc1age1

nchil1


Ebu

Shi

Mut

Madfedj

permtha

Figure 9.10. Ordination plot for the timber use-group. Largest symbol size (Mutambi village) corresponds to 5 farms. Abbreviations as in figure 9.1.


152

-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0


-1.0

-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

Axi

s II

(6.8

% to

t. v

ar.,

15.6

% c

an. v

ar.,

spec

-en

v. c

orr

. 0.5

8)

harmad

azaind

size

ncros

nloc

heady

age

nchil

school

nloc1

heady1

age1nchil1


Ebu Mut

Madfedfperm

tha

Figure 9.11. Ordination plot for the medicine use-group. Largest symbol size (Shimutu village) corresponds to 26 farms and origin before species centring. Abbreviations as in figure 9.1.

-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7


-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

Axi

s II

(8.5

% t

ot.

var

., 17

.9 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.65)

buddav

cuplus

terman

size

ncros

nlocheady

nchil

school

size1

ncros1school1


Ebu

Shi

Mad perm

Figure 9.12. Ordination plot for the ornamental use-group. Largest symbol size (Ebuchiebe and Shimutu village) corresponds to 32 farms and origin before species centring. Abbreviations as in figure 9.1.

CHAPTER 9

153

Shi

Mut

Mad

fedj

fedf

permtha

-0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8


-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

Axi

s II

(4.3

% t

ot.

var.

, 10.

1 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.66) bisjav

cromac

cromeg

erijap

manind

psigua

syzcum

zangil

size

nloc

age

size1

age1


Figure 9.13. Ordination plot for the shade use-group. Largest symbol size (Shimutu village) corresponds to 17 farms and origin

before species centring. Abbreviations as in figure 9.1.

-1.1 -1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6


-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Axi

s II

(4.4

% to

t. v

ar.,

9.5

% c

an. v

ar.,

spec

-en

v. c

orr

. 0.7

5)

camsincarpap

cofrob

riccom

sesses

tepvog

ncros nloc

heady

ncros1

heady1


EbuShi

Mut

Madfedf perm

Figure 9.14. Ordination plot for the soil fertility group. Largest symbol size (Ebuchiebe village) corresponds to 29 farms. Abbreviations as in figure 9.1.

CHAPTER 10 COMPARISON OF TREE SPECIES COMPOSITION OF FARMS IN WESTERN KENYA USING CONSTRAINED ORDINATION II. ANALYSIS FOR SEPARATE NICHES

R KINDT, P VAN DAMME, AJ SIMONS & H BEECKMAN SUBMITTED TO AMBIO

One of the objectives of tree domestication research is the diversification of the tree species composition of agroecosystems. We investigated whether substantial differences existed in the species composition of farms, so that diversification efforts could be guided by these differences – for example by introducing species that are dominant in one subsection of the landscape into subsections where these species are currently rare. Since we aimed to target diversification at specific on-farm niches, we investigated farm × species matrices that only listed trees that occurred either in homesteads, mixed in cropland, woodlots, fallows, external boundaries, internal boundaries, or on contours in cropland. Ordinations of farms were obtained through Hellinger-distance-based linear and polynomial Redundancy Analysis (LRDA and PRDA) were all significant except for crop contours. Variance partitioning indicated significant influences of farm location, while socio-economic characteristics had significant influences on 6 LRDA and 4 PRDA ordinations. PRDA explained substantially more variation than LRDA (47% versus 14%). PRDA only explained 49% of the variation explained by unconstrained Principal Components Analysis on the first axis, believed to correspond to a “real-world” gradient. The ordination results were confirmed by multiple regression analyses using the abundance of dominant species as response variable. The relationship between location and species composition differed with those of previous surveys. Differences in sampling intensity and the fact that locations and time of sampling differed in each survey could have caused the differences. Compositional differences among surveys reflected more differences in species proportional abundance, rather than presence or absence of dominant species, thus changes in abundance of component species would easily lead to changes in the species composition type.

SPECIES COMPOSITION OF NICHES

156

10.1. Introduction

One of the objectives of tree domestication research in general and more specific in western Kenya is the diversification of tree species composition in agroecosystems (Kindt & Lengkeek 1999). Making agroecosystems ecologically more stable and economically more rewarding through diversification and intensification of land use by domestication of agroforestry trees is one of the three pillars of research of the International Centre for Research in Agroforestry (ICRAF 1997; ICRAF 2000).

Addressing questions on species distribution in the agricultural landscape, species composition was investigated for various on-farm niches. On-farm niches for trees refer to the location on the farm and the establishment pattern of trees at the location. The niches that were distinguished for western Kenya were trees in the homestead area, trees mixed in cropland, trees on contours in cropland, trees on external boundaries of the farm, trees on internal boundaries on the farm, trees in woodlots, and trees in fallows. The distribution of trees in on-farm niches has been recorded in several surveys of agroforestry species (e.g. Ponce et al. 1991; Thijssen et al 1993), including previous surveys in the same area (Franzel pers. comm.; Kindt 1997). Categories of on-farm niches should be largely self-explanatory, easily identifiable in the field and from aerial photographs, and correspond to local experience with tree establishment (Bradley 1991 p. 166).

The analysis provided in this chapter supplements analysis of various types of functions of tree species in the studied western Kenyan agroecosystem.


Formulation of niche matrices (4.1), analysis of species composition (4.5)

10.3. Results

10.3.1. Comparison with Previous Surveys

Table 10.1 mentions farm frequency of species in three niches as recorded by Bradley (1991) and during the survey. The clearest difference between the trends observed by Bradley (1991) are the higher percentage of farms with Euphorbia tirucalli in Ebuchiebe than Mutambi. However, the confidence intervals for the percentages of the species just overlapped. The higher frequency of Persea americana and Psidium guajava in cropland in Mutambi and lower frequencies of Sesbania sesban and Markhamia lutea in this village and niche were confirmed (but confidence intervals just overlapped for Markhamia lutea). However, all frequencies were higher in the later survey, a pattern that was also observed for many other species and niches (e.g. Markhamia lutea in hedges, Psidium guajava in other niches). The higher percentages introduced differences between Ebuchiebe and Mutambi for Mangifera indica with a higher frequency in the first village. Results from both surveys indicate that most species belong to several niches.

Table 10.2 lists all species mentioned by Lauriks (1996) and Lauriks et al. (1999). The species that also contributed to the ordination diagramme for the external boundary (Buddleja davidii and Cupressus lusitanica) were added to the table. The major differences of the results of our survey with those obtained by Lauriks et al. (1999) and Lauriks (1996) were the lower frequency of Euphorbia tirucalli in Mutambi and Madidi villages, and the higher frequency of the species in Ebuchiebe. In the first two villages, the effect is caused by the higher proportion of Cupressus lusitanica , Buddleja davidii and Dracaena steudneri. Ebuchiebe had a lower proportion of Lantana

CHAPTER 10

157

camara. The difference for Ebuchiebe could have resulted from the sampling frame of 25 km2 cells used by Lauriks et al. (1999). The cells east and west to the Ebuchiebe cell belonged to hedge type 1, which is a hedge type that describes the Ebuchiebe species composition very well. The co-ordinates of the farm sampled by Lauriks (1996) (677.25 E, 12.71 N) indicates that the position was about 3 km north of Ebuchiebe. The sampling from this author therefore does not exclude the boundaries of the hedge type 1 subregion extending further west or east than the 25 km2 grid boundaries of the Ebuchiebe cell. The hedge types that corresponded most to the composition recorded in Mutambi, Madidi and Shimutu villages was hedge type 3 containing a mixture of Dracaena steudneri, Euphorbia tirucalli, Lantana camara and Tithonia diversifolia. Cells north-west (for the first village) and south of the cells to which these villages belonged had this species composition.

The results from table 10.2 also indicate that differences between hedges were mainly differences in frequencies of species, not between occurrence of these species. Hedge types 1 and 2 shared six of the seven species used to categorize differences between these types, while at least five of these species were encountered in each village of the recent survey.

10.3.2. Results of Constrained Ordination

Table 10.3 provides information on the ordinations by niches. Figures 10.1 to 10.7 present the ordination diagrammes. The complete list of regression coefficients obtained from stepwise regression is provided in table 10.5.

All ordinations were significant, except LRDA and PRDA for crop contours. PRDA explained substantially more variation than LRDA (respective averages are 47 and 14% of explained variation). Both methods however only explained a fraction of total variation, what resulted in the first axis explaining substantially less variation than PCA. The average variation explained by the first axis of PRDA was only 49% of that of the first PCA axis. Two axes of PRDA (corresponding to the diagrammes included) explain on average 17% of variation – two axes of PCA explain 35% of variation on average.

Except for crop contours, the broken stick criterion revealed that many axes (minimum five) were significant. This points to data sets with a lot of variation that can only be represented on several orthogonal axes. The number of significant axes corresponded to the number of species differentiated in the diagrammes.

The site scores on the first RDA and PCA axes were strongly correlated, except for homesteads and crop contours. These two groups were also the groups where variation explained on first and second PCA axes was almost similar. This indicates that PCA did not identify a dominant first gradient in the data, and that any two orthogonal axes belonging to the same plane would have expressed a similar amount of variation. RDA expressed another first axis belonging to a similar plane, as species that significantly contributed to the ordination were similar.

Table 10.4 provides a summary of the variance partitioning of the various niches. Differences between villages were significant for all niches. LRDA had significant ordination for all niches except for crop contours. PRDA was not significant for crop contours and not highly significant (P=0.09) for woodlots and tree fallows. LRDA and PRDA had different joint explained variation (Vj). Vj was always smaller for PRDA. Ordination for external boundaries was the only case where Vj was positive. A negative sign for Vj means that both environmental characteristics together explain the variance better than the sum of their individual effects (Legendre & Legendre 1998 p. 533).

Figure 10.1 shows the ordination for species occurring mixed in cropland. Mutambi is characterized by Coffea arabica (rs=2.06, P=0.0001) and Camellia sinensis (rs=1.71, P=0.0001).


158

Madidi and Ebuchiebe contain more Sesbania sesban, Madidi and Mutambi more Psidium guajava (negative coefficients for the other villages, table 10.5). Ebuchiebe contains more Markhamia lutea (rs=0.38, P=0.04). The higher abundance for Persea americana, Carica papaya , Mangifera indica and Cajanus cajan for Madidi could only be confirmed for the latter species (table 10.5). For Persea americana and Mangifera indica , only the lower abundance in Ebuchiebe was confirmed (rs=-0.60, P=0 and rs=-0.15, P=0.08, respectively). Permanent houses have more Coffea arabica (rs=1.16, P=0.0005) and Markhamia lutea. (rs=0.45, P=0.08) and less Carica papaya (rs=-0.60, P=0.0004). The relationship between Coffea arabica and the number of years that the head was in charge of the farm could not be confirmed (this pattern also occurred for the beverage group), but the relationship with de jure female household heads was (rs=0.79, P=0.001). Farm size was positively correlated with the presence of Markhamia lutea (rs=0.81, P=0.05). Camellia sinensis presence and age of the household head correlation could not be confirmed (ra=0.40, P=0.60), but the vector length for this species was short.

Figure 10.2 shows the species composition for the homestead area. Only three species are well represented. Croton macrostachyus is typical for Shimutu (rs=1.00, P=0.0001). Mangifera indica is more typical of Ebuchiebe and Madidi (table 10.5). Lower presence of Psidium guajava in farms with thatch-roofed houses was confirmed (rs=-0.42, P=0.009), but not higher presence in permanent houses (rs=-0.33, P=0.15). The species also occurred more in households with de jure female heads (rs=0.33, P=0.09). Croton macrostachyus occurred more on farms with heads being longer in charge (rs=0.53, P=0.06), with more cattle of local race (rs=1.08, P=0.003) and less on farms with thatch-roofed houses (rs=-0.33, P=0.02). The relationship with farm size could not be confirmed (ra=0.12, P=0.73), but the angle between both vectors was larger. Of species that were not well represented, Markhamia lutea was associated with permanent houses (rs=0.33, P=0.03) and the time the household head was in charge of the farm (rs=0.76, P=0.001). Syzygium cumini was more abundant in Madidi (table 10.5).

Figure 10.3 provides the ordination of species occurring in woodlots. Mutambi had higher proportion of Harungana madagascariensis (rs=0.60, P=0.02) and Bridelia micrantha (rs=0.23, P=0.02). Farm size was associated with Eucalyptus saligna (rs=2.69, P=0.002) and fewer de facto household heads had the species (positive coefficients for the other head categories, table 10.5). The link between the species and the number of cattle of local race could not be confirmed (ra=0.31, P=0.77). Croton macrostachyus was associated with the age of the household head (rs=0.80, P=0.02), Mangifera indica was not (ra=-0.13, P=0.34), but both species vectors were short. Harungana madagascariensis was associated with the level of schooling of the household head (rs=0.76, P=0.02) and negatively linked with permanent houses (rs=-0.52, P=0.09). The positive links with the number of crossbred cattle and the number of children could not be confirmed (respectively ra=-0.56, P=0.49 and ra=0.65, P=0.14).

Figure 10.4 shows the ordination of species and farms for tree fallows. Differences between villages were not evident. More Senna didyomobotrya (rs=0.06, P=0.09) and Bridelia micrantha (rs=0.23, P=0.01) occurred in Mutambi. The village was also characterized by less Psidium guajava (rs=-0.41, P=0.09). The clearest gradient is present for farms with larger sizes. These farms have more, Psidium guajava (rs=1.57, P=0.004), Croton macrostachyus (rs=1.40, P=0.0001), Harungana madagascariensis (rs=1.62, P=0.0001), Markhamia lutea (rs=0.87, P=0.004) and Bridelia micrantha (rs=1.15, P=0.0001). Senna didyomobotrya was positively linked with the number of years a head was in charge (rs=0.15, P=0.03). Farms with more P. guajava had fewer crossbred cattle (rs=-1.15, P=0.10).

Figure 10.5 shows the ordination for the external boundary. The main differences in species composition are those revealed among villages. Ebuchiebe contains more Euphorbia tirucalli (rs=1.94, P=0.0001) and Markhamia lutea (rs=0.71, P=0.003). Shimutu contains more Lantana camara (rs=0.73, P=0.05), Tithonia diversifolia (rs=1.6, P=0.0001) and, Psidium guajava (rs=0.49, P=0.04). Mutambi and Madidi contain more Buddleja davidii (negative coefficients for the other

CHAPTER 10

159

villages, table 10.5). A similar pattern occurring for Cupressus lusitanica was not confirmed (the coefficient for Mutambi was also negative, table 10.5). Dracaena fragrans is more dominant in Shimutu and Madidi (negative coefficients for the other villages). The relationship between farm size and Lantana camara could not be confirmed (ra=0.94, P=0.31). The age of the household head was positively related to Cupressus lusitanica (rs=2.01, P=0.002), as was a permanent house (rs=2.20, P=0.0001). The latter variable had the opposite relationship with Buddleja davidii than apparent from the ordination (rs=-0.68, P=0.05). The positive relationship between this species and number of crossbred cattle could not be confirmed (ra=0.54, P=0.60), but the relationship with number of children (a longer vector but with a wider angle) was (rs=0.88, P=0.09).

Figure 10.6 shows that the ordination for the internal boundary only distinguishes two species. Buddleja davidii is mainly associated with farms occurring in Madidi (negative coefficients for the other villages, see table 10.5) and with more children (rs=0.79, P=0.002). Cupressus lusitanica was associated with farms with permanent houses (rs=0.89, P=0.0009) and more cattle of local race (rs=0.91, P=0.02). The association between the species and the number of crossbred cattle could not be confirmed (ra=-0.26, P=0.69).

Figure 10.7 shows the ordination for species occurring on contours. The main ordination is based on differences for Sesbania sesban and Psidium guajava. Ebuchiebe contains more Sesbania sesban (rs=-0.14, P=0.03), while Madidi contains more Psidium guajava (negative coefficients for the other villages, table 10.5).

10.4. Discussion

10.4.1. Comparison with Previous Surveys

The results from the survey differed in various aspects with the stratification obtained from two previous studies, while other patterns were confirmed. The specific locations where samples were taken were different in all surveys, while the samples were also taken at different times and at different intensity (only one sample per cell in one case versus minimum 50 in the other surveys). Lauriks et al. (1999) mentioned that some of their findings contrasted with those of Bradley (1991). They found other hedge types in some parts of Vihiga than the type dominated by E. tirucalli. When comparing the results from Bradley (1991) and Lauriks et al. (1985), it is also apparent that the more intense sampling by the first survey revealed farms that did not contain the dominant hedge species. One example is the occurrence of Lantana camara on approximately a quarter of farms of the A1 stratum described by Bradley (1991), a species that dominated the hedge type for the same area in the results of Lauriks et al. (1999). These results agree with the finding that some farms appeared in the ordination diagrammes with composition more typical of other villages (see discussion below). With farms belonging to the same village differing much in species composition, the sampling method of Lauriks et al. (1999) could have sampled farms with hedge types that were less frequent in the village or cell. The regional distribution of hedge types in Vihiga could thus also have been an artefact created by particular selections of farms. For instance, hedge type 3 was only sampled three times and types 4 and 5 twice – their distribution could be more an indication of being less frequent in the whole area than their occurrence in the specific location. An indication for potential sampling effects are also farms encountered by Lauriks (1996) with two hedge types: the farm in the cell north of Madidi was sampled twice and contained a hedge more similar to type 1 and a hedge more typical of type 2.

The results thus indicated that the strata as obtained by various surveys did not agree completely with each other. The results however confirmed a spatial partitioning of species composition since typical species compositions for each village could be determined for most use-groups and niches. The spatial pattern of species composition over the survey area is therefore expected to


160

either be more patchy than described by the previous surveys (i.e. consisting of a smaller-scale mosaic of various types of composition), or still consisting of large areas with dominant hedge types but with more irregular boundaries. Lauriks et al. (1999) concluded that their method was not detailed enough to reveal intra-district differences. To investigate which pattern of species composition provides a better description, a more detailed spatially distributed sampling scheme should be used (e.g. 1 km2 cells or transect surveys). Because some farms were found with exceptional species composition in this and the Bradley surveys, the sampling scheme should include various samples per village.

The method used by Lauriks et al. (1999) was based on the squared difference in species profiles. As indicated by Legendre & Gallagher (2001), the distance between species profiles should not be used to analyse ecological gradients. However, if Lauriks et al. (1999) would have used a distance measure such as the Hellinger distance used in the ordinations presented here, differences in species composition would still have prevailed.

10.4.2. Ordination Results

As we observed for ordinations for use-groups (Kindt et al. - Chapter 9), ordinations and multiple regressions did not explain all variation in the data, while PRDA and PCA site scores on the first axis were strongly correlated (with the exception of homesteads and crop contours). These results indicate that single species were, but not strongly, related to the explanatory variables that we used. The ordination diagrammes and regression results showed that various exceptional cases occurred where sites with similar environmental characteristics had dissimilar abundances and species compositions.

Although not all variation of Hellinger-distances among sites could be explained by RDA, the consistency between RDA and multiple regression results and the significance of the relationships between species and environmental characteristics allow to consider environmental characteristics to guide diversification. As we discussed for results of use-groups (Kindt et al. - Chapter 9), exceptional cases do not prevent targeting of interventions. For example, the ordination diagrammes show typical species compositions of various villages. Species typical of one village could be introduced in another, but in some exceptional farms, the species would already occur.

The analyses in this article show differences in species composition of separate on-farm niches. These analyses therefore allow for the design of interventions for separate niches. In the previous paper, we mentioned that diversification could be targeted towards use-groups with lower diversity of the average farm (alpha diversity). Average species diversity on crop contours is 0.2 (1.9 when only including those farms where the niche is represented), internal boundaries 0.9 (2.4), fallows 1.3 (6.3), woodlots 2.9 (3.9), external boundaries 4.5 (4.8), homesteads 5.0 (5.6), and cropland 6.5 (6.6). Crop contours (figure 10.7) and internal boundaries (figure 10.6) could therefore be the focal niches for diversification. The ordination diagrammes show that both groups are dominated by two species only.

However, when we analyze niche diversity by the frequency of the dominant species by summing up species abundances over all farms (the Berger-Parker diversity index, Magurran 1988 eq. 2.31), internal boundaries and crop contours are the most diverse with relative frequencies of 0.16 and 0.19, respectively. Frequencies of dominant species in homesteads are 0.22, in external boundaries 0.30, tree fallows 0.41, cropland 0.56, and woodlots 0.65. It is obvious that the choice of the niche with the lowest diversity depends on the criterion used. Choice of the niche to diversify will in reality also depend on importance attributed to its diversification by farmers, while it has to be kept in mind that diversity for the same niche varies among farms so that diversification of the niche may not be relevant for all farms. However, the analysis presented

CHAPTER 10

161

here enables to select niches where most farms have low diversity and provides information on species composition of each farm that can be used to guide interventions that seek farm diversification.

Studies of species-rich ecosystems often attempt to reduce complexity by investigating functional groups rather than individual species. Functional groups can be defined as clusters of species that play the same role in maintaining and regulating ecosystem processes (Gitay et al. 1996). Schulze & Mooney (1994) point out that functional groups will be differently defined depending on the processes studied. Norberg et al. (2001) define functional groups as clusters of species that share similar resources and predators. Colasanti et al. (2001) differentiate between functional groups and functional types, where one species may be simultaneously a member of several functional groups, but not of several functional types. When one species is removed from its functional group, other species belonging to the group may increase in abundance (number of individuals) taking over its role. As De Leo & Levin (1997) pointed out, macro-level functional integrity (such as productivity) of ecosystems may be preserved by maintaining functional groups while reducing diversity within groups, but their structural integrity (such as resilience – the time of the ecosystem to recover after disturbance) may be weakened.

A similar analysis as performed above for on-farm niches could be applied to functional ecological groups: identification of groups with lowest diversity in combination with investigation how various sites differ in species composition, which leads to suggestions for interventions seeking to increase site diversity. However, the concept of using functional groups to plan conservation has been a topic of controversy. Walker (1995) is in favour arguing that conservation efforts should focus on important species. Gitay et al. (1996) and Chapin et al. (2000) pointed out the possibility that not all functions of an individual species within an ecosystem are known, so that effects of species removals cannot be predicted. The same reasons to target or not to target conservation apply to targeting of diversification.


162

Table 10.1. Farm frequency (%) of species mentioned in Bradley et al. (1985) and Bradley (1991) (B.) and in our survey (95% confidence interval). Species A1 / Ebusikhale / Ebuchiebe A2 / Kegoye / Mutambi Hedge Homestead Cropland Hedge Homestead Cropland B. Survey B. Survey B. Survey B. Survey B. Survey B. Survey Euphorbia tirucalli 45 92 (80-99) - 0 - 0 95 68 (50-81) - 0 - 0 Persea americana - 10 (1-19) 21 54 (40-68) 15 60 (50-74) - 4 (0-10) 38 46 (30-60) 65 88 (80-97) Cupressus lusitanica 24 20 (9-31) - 12 (3-21) 22 6 (1-13) 50 32 (20-45) - 6 (1-13) 22 0 Mangifera indica - 6 (1-13) 28 64 (50-78) - 32 (20-45) - 0 30 24 (10-36) - 44 (30-58) Markhamia lutea 15 58 (40-72) - 18 (7-29) 27 58 (40-72) 0 20 (9-31) - 14 (4-24) 3 28 (20-41) Sesbania sesban - 0 - 16 (5-27) 27 66 (50-80) - 2 (0-6) - 4 (0-10) 0 34 (20-48) Lantana camara 27 14 (4-24) - 0 - 0 0 30 (20-43) - 4 (0-10) - 2 (0-6) Eucalyptus saligna 21 26 (10-39) 24 34 (20-48) 23 44 (30-58) 22 10 (1-19) 0 22 (10-34) 25 36 (20-50) Psidium guajava 18 44 (30-58) 0 42 (30-56) 0 24 (10-36) 0 26 (10-39) 30 56 (40-70) 35 58 (40-72)

Table 10.2. The species composition in hedges provided by Lauriks et al. (1999) for hedge types and Lauriks (1996) for hedge composition of individual 25 km2 cells (L. village) or cells south or southeast (L. S or SE) from the villages we surveyed, and as obtained from our survey (with 95% confidence interval). Species Composition of hedge types in percent (village) Hedge type 1 (Mutambi, Madidi) Hedge type 2 (Ebuchiebe) Shimutu L. L.

Mu. L.

Ma. Mutambi Madidi L. L.

Eb. Ebuchiebe L.

S L. SE

Shimutu

Euphorbia tirucalli

64.3 62.7 91.5 18.0 (12.0-25.4)

17.0 (11.0-23.6)

5.2 2.1 74.0 (63.0-84.9)

32.7 0.0 8.9 (3.8-14.0)

Dracaena steudneri 2.3 19.6 2.5 13.0 (3.2-22.1)

18.0 (10.0-25.2)

0.8 0.0 0.3 (0.0-0.8)

40.4 0.0 24.0 (16.0-32.2)

Lantana camara 13.8 0.0 0.0 9.0 (2.6-15.4)

13.0 (7.2-17.8)

52.9 84.1 5.6 (0.6-10.5)

1.9 17.9 21.9 (13.9-29.9)

Markhamia lutea 6.0 0.0 0.0 0.7 (0.2-1.1)

3.9 (0.1-7.7)

5.1 0.0 7.4 (3.0-11.8)

0.0 0.7 2.4 (1.0-3.9)

Tithonia diversifolia

2.6 0.0 0.0 0.7 (0.0-1.8)

0.0 6.1 0.0 0.0 0.0 68.9 20.0 (12.0-28.9)

Aloe sp. 0.0 0.0 0.0 0.0 0.0 3.3 0.0 0.0 0.0 0.0 0.0 Psidium guajava 3.0 0.0 4.5 0.8

(0.3-1.2) 1.6

(0.6-2.5) 2.1 4.8 3.7

(1.4-6.0) 15.4 0.0 6.3

(3.5-9.1) Buddleja davidii - - - 27.5

(14.0-40.8) 14.6

(7.5-21.8)) - - 0.0 - - 0.0

Cupressus lusitanica

- 0.0 0.0 26.5 (12.0-41.4)

26.8 (17.0-36.7)

- 0.0 4.4 (0.0-11.4)

0.0 0.0 6.9 (2.3-11.5)

Other species 8.1 17.7 1.5 3.8 23.1 24.4 11.2 4.6 9.6 12.5 9.6

CHAPTER 10

163

Table 10.3. Summary information on the ordination of various on-farm niches. Variance is expressed as percentage of total variance.

Ordination Sp Sp (2)

Var PRDA

Sign (99)

Var LRDA

Sign (9999)

Ax Var PCA (1)

Var LRDA

(1)

Var PRDA

(1)

Cor. PCA- RDA

Var PCA (2)

Var PRDA

(2) Cropland 61 10 39.70 0.01 17.12 0.0001 11 12.68 6.48 7.45 0.94 22.98 12.49 Homestead 62 7 40.41 0.01 13.79 0.0001 9 11.63 4.01 5.63 -0.12 22.00 9.38 Woodlot 54 5 37.88 0.05 12.10 0.0006 5 33.78 4.12 10.21 -0.99 46.96 14.82 Tree fallow 51 7 53.87 0.02 10.36 0.0184 5 31.98 5.61 13.43 -0.99 42.32 19.23 External boundary

50 8 42.70 0.01 22.01 0.0001 11 20.63 10.28 13.31 0.99 33.44 20.54

Internal boundary

34 2 46.70 0.06 12.97 0.0002 5 24.59 5.48 10.36 0.99 38.35 17.55

Crop contours 19 2 66.18 0.44 8.76 0.1637 2 22.98 3.84 13.37 -0.10 42.34 22.53 (Sp: number of species in the site-species matrix analyzed; Sp (2): number of species included in the ordination diagramme with the first two axes; Var PRDA/LRDA/PCA (y): variance explained by polynomial RDA/linear RDA/PCA (for first y axes); Sign(99/9999): significance of ordination of lefthand column for 99/9999 permutations; Ax: number of axes to consider based on the broken stick distribution; Cor. PCA – PRDA: correlation between the site scores on the first PCA and PRDA axis)

Table 10.4. Variance partitioning between village and household characteristics for various on-farm niches. Variance is expressed as percentage of total variance.

Ordination Vv LRDA/PRDA

Sign (9999)

Vh LRDA

Sign (9999)

Vh PRDA

Sign (99)

Vve LRDA

Vve PRDA

Vhe LRDA

Vhe PRDA

Cropland 10.3 0.0001 8.62 0.0002 27.11 0.01 8.50 12.59 6.82 29.40 Homestead 6.12 0.0001 8.10 0.0006 28.66 0.04 5.69 11.75 7.67 34.29 Woodlot 3.79 0.0027 8.78 0.0078 28.66 0.09 3.32 9.22 8.31 34.09 Tree fallow 2.97 0.0137 8.48 0.0193 40.44 0.09 1.88 13.43 7.39 50.90 External boundary 14.98 0.0001 10.60 0.0001 29.88 0.01 11.41 12.82 7.03 27.72 Internal boundary 4.63 0.0001 9.41 0.0009 35.81 0.06 3.56 10.89 8.34 42.07 Crop contours 2.52 0.0250 6.44 0.3135 51.10 0.31 2.32 15.08 6.24 63.66

(V v/h (e) LRDA/PRDA: (exclusive) linear/polynomial variance explained by village/household characteristics; Sign(99/9999): significance of ordination of lefthand column for 99/9999 permutations)

Table 10.5. Coefficients of village and household characteristics as explanatory variables on species abundance information for various on-farm niches after stepwise linear regression. The significance of coefficients for explanatory variables (shown below characteristics between brackets) as provided by multiple linear regression (indicated by *) where residuals were normally distributed, otherwise calculated by permutation tests. Total explained variance is expressed by the multiple correlation coefficient R2, with probability P(F). Ordination

Species (other niches where the species occurred with the same abundance pattern over households)

R2

(P(F))

Inter-cept

Ebuchie-be village (farthest

from forest)


to forest)

Mutambi village

Male headed


(no husband)

Farm size (acres)

(1 acre = 0.4005

ha)


wealthy)


wealthy)

Number of cross-

bred cattle

(~wealth)


(~wealth)

Years being head


Number of

resident children


Mixed Coffea arabica 0.514 -0.45 -0.33 2.06 0.79 1.16 -1.51 0.76 0.48 In (0) (0.176)* (0.0979) (0.0001) (0.0012) (0.0005) (0.0312) (0.0925) (0.1435) Cropland Camellia sinensis 0.214 -0.08 1.71 -0.55 1.18 1.41 (1e-8) (0.708)* (0.0001) (0.0880) (0.0622) (0.1232) Psidium guajava ‡ 0.120 0.568 -0.22 -0.38 0.55 (4e-5) (0)* (0.0927) (0.0044) (0.0003) Mangifera indica 0.016 0.44 -0.15 (0.085) (0)* (0.0838) Persea americana 0.175 1.60 -0.60 -0.22 -0.67 -0.82 -0.53 (2e-6) (0) (0)* (0.151)* (0.093)* (0.009)* (0.006)* Carica papaya 0.090 0.38 -0.60 0.65 (2e-4) (0.008)* (0.0004) (0.0111) Cajanus cajan ‡ 0.171 0.43 -0.51 -0.49 -0.49 0.20 (9e-7) (3e-5)* (0.0001) (0.0001) (0.0001) (0.1276) Sesbania sesban 0.137 1.11 -0.93 -0.54 1.40 -0.71 (3e-5) (0)* (0.0001) (0.0064) (0.0021) (0.0559) Markhamia lutea 0.11 0.49 0.38 -0.34 0.38 0.81 0.45 (8e-4) (0.002)* (0.0448) (0.0909) (0.0847) (0.0536) (0.0768) Eucalyptus saligna 0.025 1.03 -0.82 (0.034) (0)* (0.0339) Homesteads Syzygium cumini ‡ 0.172 0.10 0.21 -0.13 -0.15 -0.15 0.37 0.36 (8e-6) (0.064)* (0.0032) (0.0608) (0.0367) (0.1039) (0.0813) (0.0183) Mangifera indica 0.202 0.43 0.18 -0.43 -0.26 0.86 -0.22 (1e-7) (4e-5) (0.0968) (0.0007) (0.0193) (0.0003) (0.0886) Azadirachta indica ‡ 0.026 0.06 0.32 (0.028) (0.067) (0.0309) Psidium guajava 0.150 0.26 0.33 1.43 -0.33 -0.42 0.63 (2e-5) (0.135)* (0.0881) (0.0002) (0.1471) (0.0086) (0.0574) Markhamia lutea ‡ 0.088 0.04 0.33 0.76 (3e-4) (0.641) (0.0281) (0.0011) Eucalyptus saligna 0.091 0.32 0.62 0.40 -0.45 (6e-4) (0.002) (0.0002) (0.0203) (0.046) Croton macrostachyus 0.318 -0.19 1.00 -0.33 1.08 0.53 0.55 (2e-13) (0.247) (0.0001) (0.0163) (0.0028) (0.0565) (0.0092)


Species R2

(P(F))

Inter-cept

Ebuchie-be

Shimutu Mutam-bi

Male headed



Thatch roofed house

Number of cross-

bred cattle


Years being head


Number of

resident children


Woodlots Eucalyptus saligna 0.134 1.32 -0.73 -0.62 0.98 0.88 2.69 1.67 (3e-4) (0.003)* (0.0407) (0.1243) (0.0094) (0.0758) (0.0020) (0.0209) Mangifera indica 0.036 0.18 -0.09 -0.15 -0.11 (0.085) (6e-4)* (0.07) (0.13) (0.12) Croton macrostachyus 0.109 -0.10 0.61 0.80 (3e-5) (0.606) (0.0005) (0.0196) Bridelia micrantha ‡ 0.077 -0.13 0.23 0.35 0.47 0.23 (0.006) (0.191) (0.0147) (0.1137) (0.0526) (0.0960) Harungana madagascariensis 0.206 0.13 -0.66 -0.78 0.60 1.67 -0.52 0.76 (3e-7) (0.586)* (0.0078) (0.0051) (0.0163) (0.0044) (0.0863) (0.0154) Fallows Senna didymobotrya ‡ 0.069 -0.09 0.06 0.15 0.14 (0.005) (0.016) (0.0935) (0.0335) (0.0056) Bridelia micrantha 0.201 0.09 0.23 -0.19 1.15 -0.48 -0.57 -0.41 (5e-7) (0.329)* (0.0104) (0.0565) (0.0001) (0.0928) (0.0121) (0.0183) Markhamia lutea 0.100 -0.23 -0.21 0.87 -0.70 0.54 (8e-4) (0.104) (0.1378) (0.0041) (0.0714) (0.0354) Harungana madagascariensis 0.226 0.27 -0.20 -0.23 1.62 -0.64 -0.61 -0.62 (4e-8) (0.027)* (0.0814) (0.0955) (0.0001) (0.0907) (0.0434) (0.0121) Croton macrostachyus 0.167 0.123 -0.23 -0.22 -0.22 1.40 -0.70 -0.31 (1e-5) (0.258)* (0.0356) (0.1049) (0.0743) (0.0001) (0.0566) (0.1544) Psidium guajava 0.099 0.32 -0.34 -0.41 1.57 -1.15 (9e-4) (0.084) (0.1211) (0.0861) (0.0036) (0.0981) Sesbania sesban ‡ 0.045 -0.01 0.16 0.18 (0.016) (0.783) (0.0558) (0.0339)


Species R2

(P(F))

Inter-cept

Ebuchie-be

Shimutu Mutam-bi

Male headed



Thatch roofed house

Number of cross-

bred cattle


Years being head


Number of

resident children


External Markhamia lutea 0.117 0.77 0.71 -0.36 0.57 -1.66 Boundaries (2e-4) (0)* (0.0035) (0.1405) (0.0981) (0.0378) Euphorbia tirucalli 0.257 2.15 1.94 -1.12 (3e-12) (0)* (0.0001) (0.0013) Buddleja davidii 0.309 1.65 -1.75 -1.70 -0.68 -0.53 0.88 (7e-13) (0)* (0.0001) (0.0001) (0.0538) (0.0396) (0.0945) Cupressus lusitanica 0.310 0.73 -2.01 -1.12 -1.19 2.20 2.01 1.29 (3e-12) (0.102)* (0.0001) (0.0023) (0.0014) (0.0001) (0.0021) (0.0313) Dracaena fragrans 0.198 1.29 -1.79 -1.05 -0.79 1.74 (5e-8) (0)* (0.0001) (0.0013) (0.0533) (0.0054) Lantana camara 0.162 1.94 -1.20 0.73 -0.76 -1.13 (2e-6) (0)* (0.0009) (0.0508) (0.0435) (0.0629) Tithonia diversifolia ‡ 0.274 0.22 1.56 -0.53 -0.50 (2e-12) (0.059) (0.0001) (0.0571) (0.0172) Psidium guajava 0.163 1.02 0.49 -0.34 1.51 -1.01 (2e-6) (3e-5)* (0.0375) (0.1050) (0.0050) (0.0217) Internal Buddleja davidii 0.191 0.95 -1.04 -0.90 -0.41 0.35 -0.61 0.79 Boundaries (1e-6) (3e-4)* (0.0001) (0.0001) (0.0439) (0.0819) (0.0880) (0.0224) Cupressus lusitanica 0.141 0.30 0.89 0.91 -0.48 -0.46 (2e-5) (0.066) (0.0009) (0.0205) (0.1174) (0.1500) Crop Sesbania sesban 0.022 0.031 0.14 Contours (0.043) (0.365)* (0.0344) Psidium guajava 0.161 0.35 -0.23 -0.20 -0.15 -0.19 0.28 0.65 -0.22 (6e-5) (2e-4)* (0.0033) (0.0282) (0.0527) (0.0006) (0.1094) (0.0232) (0.1267)

‡: Species that did not contribute significantly to the first two PCA axes

CHAPTER 10

167

-0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8


-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7A

xis

II (5

.0 %

to

t. v

ar.,

12.7

% c

an. v

ar.,

spec

-en

v. c

orr

. 0.7

0)

cajcaj

camsin

carpap

cofrob

eucsal

manind

marlut

perame

psiguasesses

sizencros

nloc

heady

age

schoolsize1

ncros1nloc1heady1

age1


Ebu Shi Mut

Mad

fedj

fedf

perm

irontha

Figure 10.1. Ordination plot for species occurring mixed in cropland. % tot. var.: percentage of total variance explained on the axis; % can. var.: percentage of canonical variance explained on the axis; spec-env. cor.: species-environment correlation. Largest

symbol size (Ebuchiebe village) corresponds to 2 farms and origin before species centring.

-0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5


-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

Axi

s II

(3.8

% to

t. v

ar.,

9.3

% c

an. v

ar.,

spec

-en

v. c

orr

. 0.6

3)

cromac

eucsal

manind

marlut

azaind

psigua

syzcum

size ncros

nloc

heady

agenchil

school

size1 ncros1

nloc1

heady1

age1nchil1

school1


EbuShi

MutMad

fedj

fedfpermiron

tha

Figure 10.2. Ordination plot for species occurring in homesteads. Largest symbol size (Mutambi village) corresponds to 6 farms and origin before species centring. Abbreviations as in figure 10.1.


168

-0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 1.1


-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Axi

s II

(4.6

% t

ot.

var.

, 12.

1 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.73)

brimic

cromac

eucsal

harmad

manind

size

ncros

nloc

heady

age

nchilschool

size1

ncros1

nloc1

age1

nchil1

school1


EbuShi

Mut

Madfedjfedfperm

Figure 10.3. Ordination plot for species occurring in woodlots. Largest symbol size (Mutambi village) corresponds to 15 farms and origin before species centring. Abbreviations as in figure 10.1.

-1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6


-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

Axi

s II

(5.8

% t

ot.

var.

, 10.

8 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.69)

brimic

cromac

harmad

marlut

psigua

sendid

sesses

size

ncros

nloc

heady

age

nchilschool

size1

ncros1

nloc1

heady1

age1 nchil1

school1


EbuShiMut

fedfpermtha

Figure 10.4. Ordination plot for species occurring in fallows. Largest symbol size (Ebuchiebe village) corresponds to 42 farms and origin before species centring. Abbreviations as in figure 10.1.

CHAPTER 10

169

-0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8


-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

0.4

0.5

0.6A

xis

II (7

.2 %

to

t. va

r., 1

6.9

% c

an. v

ar.,

spec

-en

v. c

orr

. 0.7

6)

buddavcuplus

drafra

euptir

lancam

marlut

psigua

titdiv

size

ncrosnloc

heady

age nchil

school

size1

ncros1

nloc1

age1 nchil1EbuchiebeShimutuMutambiMadidi

Ebu

Shi

MutMad

fedjfedfperm

tha

Figure 10.5. Ordination plot for species occurring on external boundaries. Largest symbol size (Mutambi village) corresponds to five farms and origin before species centring. Abbreviations as in figure 10.1.

-0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0


-1.0

-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

Axi

s II

(7.2

% t

ot.

var.

, 14.

4 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.70)

buddav

cuplus

sizencrosnloc

heady

age

nchil

schoolsize1

ncros1 nloc1

age1

nchil1

school1


EbuShi MutMadfedjfedf

perm

Figure 10.6. Ordination plot for species occurring on internal boundaries. Largest symbol size (Ebuchiebe village) corresponds to 37 farms and origin before species centring. Abbreviations as in figure 10.1.


170

EbuMadfedf

-1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3


-1.1

-1.0

-0.9

-0.8

-0.7

-0.6

-0.5

-0.4

-0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

Axi

s II

(9.2

% t

ot.

var

., 13

.8 %

can

. var

., sp

ec-e

nv.

co

rr. 0

.64)

psigua

sesses

ncros

school

school1


Figure 10.7. Ordination plot for species occurring on crop contours. Largest symbol size (Ebuchiebe village) corresponds to 43 farms and origin before species centring. Abbreviations as in figure 10.1.

CHAPTER 11 FARMERS’ DECISION-MAKING IN MANAGING ON-FARM

TREE DIVERSITY R KINDT, P VAN MELE & P VAN DAMME IN PREPARATION

Tree domestication research and agroforestry are aimed at diversifying and sustaining agricultural landscapes for increased social, economic and environmental benefits. A vast body of ecological literature has demonstrated the function of species diversity in enhancing the stability and the productivity of ecosystems. In this chapter, we focused on local reasons for management of species diversity as documented through on-farm surveys. Farmers wanted more species on their farms, and the species accumulation curve for the species wanted was above that for species currently present on their farms. Such pattern contradicts the hypothesis that much diversity would be spontaneously regenerating tree species that are not actually desired. However, since we based our investigations on participatory maps of diversity that was currently present, we could not investigate what the consequence would be if many species would not regenerate spontaneously any longer, since this represents a hypothetical situation. There was a strong relationship between the richness present on a farm and the richness desired, which potentially indicated limitations in knowledge on alternative species that limits levels of desired diversity. For only one product, medicine, all farmers expected that enough would be produced on their farm at the desired abundance. This finding indicates that farmers prefer to obtain several products from their farm, rather than maximizing production for one product. For the same product or service, most farmers preferred to have several species. The most common reasons for diversity that were mentioned were complementarity among species for a particular use, or continuous supply. Risk avoidance also featured frequently. For timber, many farmers explained to still be experimenting with several species as they did not know their production characteristics. The situation that farmers already appreciate diversity offers an ideal framework for diversification. The fact that many farmers tested new species by evaluation one tree that would then become the seed source for more trees on the farm if the species tested positively suggests that some species may suffer genetic erosion. For some species, the introduction of more diverse germplasm may therefore be necessary.

MANAGEMENT OF ON-FARM DIVERSITY

172

11.1. Introduction

The objective of tree domestication research in western Kenya is the diversification of the tree species composition present in agroecosystems (Kindt & Lengkeek 1999). Making agroecosystems ecologically more stable and economically more rewarding through the diversification and intensification of land use by the domestication of agroforestry trees is one of the three pillars of ICRAF’s research (ICRAF 1997; ICRAF 2000).

Ecological surveys, experiments and models have demonstrated the positive, but conditional and saturating, relationship between diversity and ecosystem productivity and stability that is based on heterogeneity in environment and species traits (Kindt et al. – Chapter 2, 5 & 6). In this article, we focus on the reasons that farmers provided for maintaining tree diversity on their farms as communicated during semi-formal interviews. The open-ended nature of the questions that we asked farmers did not limit responses to the role of diversity in ecosystem behaviour, and therefore enabled capturing some other reasons to maintain various species on the same farm. The results are presented according to some of the topics that were discussed with farmers.

11.2. Results

11.2.1. Changes in Diversity and Abundance

Based on the diversity that was encountered on their farms in the first survey, a subset of farmers were interviewed in a second survey about the changes that they wanted in species composition on their farms. They were asked whether they wanted to add some new species, whether some species had to be removed, and what the changes in abundance of each species would be. The discussions were conducted based on participatory mapping of the species diversity within each on-farm niche.

Figure 11.1 shows the average species accumulation by randomly adding farms for the species that are present and for the species that are desired. Figure 11.1 indicates that farmers want more species on their farms at all scales from 1 to 78 farms. The hypothesis can, therefore, be rejected that the diversity that was encountered during the first survey was merely a short-term transient that resulted from spontaneous regeneration of many species that were not wanted, but only tolerated, by farmers.

Table 11.1 reveals that differences among the curves could be explained almost entirely from differences in alpha diversity of farms – between the differences in the average richness of a single farm that was present and that was wanted. Beta diversity, expressed by the exponent in the Nz

N aNS =ˆ model (Kindt et al. – Chapter 6), was almost similar (the maximal difference was 0.00835 for z2, calculated from the data used to draw figure 11.1). This finding implies that farmers want on average one extra species on their farm, while the differences among farms in species composition remained similar.

Investigations were done for the most frequent use-groups, all trees that provided a certain product or service that was provided on more than 25 farms. Table 11.1 shows that the average increase of species richness was mainly reflected in average increases for fruit, timber, and medicine, while fewer species would be used mainly for firewood, on average.

Figure 11.2 shows the diversity profile for all species occurring on farm versus the species wanted (Ha against a; a more diverse system has larger Ha than a less diverse system for all a; Kindt et al. – Chapter 7). No major differences were observed (two-sample Kolmogorov-Smirnov test: P=0.9996). Figure 11.2 also shows the higher desired species composition for the total survey that can be observed in figure 11.1 (S; H0 = ln(S)). However, the desired trees were less evenly

CHAPTER 11

173

distributed since the respective evenness profile (Ha - H0, figure not provided) was lower. The difference between the proportion of the dominant species (pdom; H8 = ln(1/pdom); pdom is the Berger-Parker diversity index) was very small (29% versus 30%, calculated from the data used to create figure 11.1). This pattern indicates that the higher richness of figure 11.1 was caused by additional species of low abundance, which did not alter the other facet of diversity, evenness, very much.

Table 11.1 provides information on average changes in evenness of individual farms through the Shannon diversity index (H1) which combines information on richness and evenness, and the Berger-Parker index which only reflects the proportion of the dominant species. The similar Shannon diversity index for two of the three groups with largest increments in richness, timber and medicine, shows that the richness increment was accompanied by a less even distribution in species. Although pdom is lower for these groups, the dominant species is less evenly distributed for these groups, if H8 - H0 is calculated. Fruit was the only group with relatively (compared to the other use-groups, absolute increments were small) large increments for richness and evenness.

Figure 11.3 shows that there is a strong relationship between the richness present on a farm, and the richness desired for that farm. The two farms with highest richness at present would still be increased in diversity, whereas the farm with smallest richness at present also had the smallest desired richness. No saturation point for diversity could be observed (e.g. farmers desiring maximum 20 species). The strong relationship revealed in figure 11.3 between present and desired richness shows that the average increase of richness is not mainly caused by increases for farms of low richness. This figure confirms that high species richness on farm is desired by farmers, and rejects the hypothesis that high richness observed on a farm would be temporary as species that regenerated spontaneously had not yet been removed.

Figure 11.4 shows that the relationship between the number of uses present and the number of uses desired on a farm is positive, as for the relationship for species richness, but not as strong as the latter relationship. The figure also indicates that, in general, fewer uses were desired than present. This relationship is probably related to the fact that the interview was not conducted for each use that was maintained for each farm, however, rather than reductions in the number of uses derived from the species that were maintained.

When requested when new trees would be established, the majority of farmers communicated that they would be established in the same or next year. Farmers that could not convey a planting year (29% of cases) mentioned that establishments would be done in the near future 86% of the time. Only one farmer explained that some trees would be planted three year after the interviews, while no farmer mentioned planting 2 years after the interviews.

Farmers were asked whether species that were not wanted any longer would be protected if they regenerated naturally. The answer was only positive for eight cases, in which case they would be cut and used as firewood after a few years. Again, this finding rejects the hypothesis that much diversity could be explained by temporary occurrence of spontaneous diversity without prior desire to maintain this diversity.

The unexplained variation of the regression represented in figure 11.3 shows that many farmers wanted either fewer or more species than the actual number on their farms. We investigated whether differences at a fixed diversity level could be linked to characteristics of these households by multiple linear regression analyses that used diversity present and household characteristics as explanatory factors for desired diversity. Seventy-six percent of variation in wanted richness of all trees on a farm was explained, of which 70% could be explained by richness alone, and 18% explained by household characteristics exclusively. Considering richness present, farmers in the village close to Kakamega Forest, de jure female-headed households, and wealthier households wanted lower diversity levels (regression coefficients have not been


174

reported here). Farmers that were longer in charge of their farms wanted lower diversity, while older farmers still wanted more diversity. The major pattern that we observed was large variation for each type of household and farm, however. This finding indicates that individual farmer strategies regarding diversity exist that are not linked to household characteristics only (e.g. female-headed, number of children). Regression results (not represented) for species richness and abundance of the most frequent use-groups analyzed separately showed similar trends of small percentages of variation explained by household characteristics.

Figure 11.5 indicates that there is only a very weak relationship between the logarithmic Berger-Parker indices. Some farmers wanted a higher proportion of the dominant species, while others wanted a smaller proportion, almost regardless of the actual proportion on their farms. The poor relationship in Berger-Parker indices could be related to the weak relationship between abundance present and abundance wanted, presented in figure 11.6. This figure further shows that almost all farmers wanted more trees on their farm.

Information from table 11.1 for changes in average abundance of use-groups shows that abundance would be increased for each group. The larger average abundance was not reflected in substantial increments in diversity (S, H, and BP of table 11.1 are all diversity indices). Hence, establishment of greater number of trees on farms was not primarily focused on average increments of species richness or evenness of use-groups, despite the opportunities (adding new species, increasing the abundance of currently less-frequent species) offered for diversification at larger abundance.

We expected that the desired abundance for each use-group as communicated by farmers would reflect the abundance needed to satisfy all household needs. When we asked farmers whether less, equal, or more than the household needs would be provided, we obtained a range of results, however. The respective percentages of farmers that would obtain less, equal, or more trees for use-groups that provided products were for charcoal (21, 32, 47), for construction (25, 29, 46), for firewood (42, 17, 42), for fruit (14, 26, 60), for timber (11, 13, 76) and for medicinal trees (0, 5, 95). Firewood, which is a use that is needed daily, was the group with the largest perceived lack in production. Medicine, which is not needed every day and only in small volumes, was the only group where all households mentioned that enough would be produced from the own farm. The percentages of farmers that would not obtain enough mainly indicate limitations in farm size, which, in combination with the management option of producing various tree products and services rather than concentrating on fewer use-groups, results in less production than needed to satisfy all needs of some farmers.

Often when not enough was produced, the product would be bought. Some farmers with not enough firewood would collect it from land of neighbours that is left fallow. One farmer explained that he would select the standing tree at the neighbour’s farm to be bought for firewood. Some farmers indicated that not enough firewood could be produced because many trees were meant for another main purpose. Some farmers that would not obtain enough charcoal mentioned that they would use firewood instead. Farmers with not enough fruit would sometimes obtain it free from neighbours. One farmer mentioned to lack the knowledge of proper tree management to produce enough fruit. One farmer explained that he wanted to conserve many timber species for future generations, but that this resulted in not being able to produce enough, what would have happened if he had established only the best species.

Asked what would be done with surplus production, most farmers answered that the surplus would be sold to neighbours, markets, or dealers for the specific product. Some farmers with a surplus of construction wood would convert it into firewood, timber, or charcoal, or materials for fencing. Farmers with surplus firewood stated that it would either be kept for later use, or sold, or given to neighbours. Farmers with more fruit production than needed by the family would sometimes give it to visitors or relatives. Some farmers mentioned that fruit would

CHAPTER 11

175

otherwise rot or be eaten by birds. Farmers with more timber than needed, would sell it to furniture makers or timber dealers. Farmers with surplus medicine would often give if to neighbours, although many farmers sell medicine to neighbours. Medicine sold at markets would often be processed first.

As we did not measure actual production, or actual household needs, it was impossible to investigate relationships between diversity, abundance, and productivity. We could therefore not conduct parallel analysis to those of ecological experiments that investigate the relationship between diversity and productivity. Finding linkages between diversity and productivity at fixed abundance as in ecological experiments, could be one rationale to promote diversity on farms.

11.2.2. Reasons for Changes in Abundance of Individual Species

We could have asked farmers directly why there was a difference between the actual number of trees and species and their desired numbers. As we discovered that some species were desired in larger abundance than desired, and other species in lower abundance, we interviewed farmers about changes in individual species as we were interested in knowing why some species would be reduced in number and others increased. We felt that these questions better reflected the dynamics in diversity on farms – opposite to asking why there was a net increment or decrease –, and would focus answers more on actual reasons for these changes.

Differentiating between species, 28 species (23%) were desired in lower abundance (including 8 species that were not wanted anymore). For 13 (11%) species, the same abundance was wanted. For 82 species (66%), more trees were wanted, which included 16 new species. The overall increase in abundance (section 11.2.1) was therefore not a general trend for each species. The high percentage of species that was wanted in higher abundance, and the explanations that farmers could provide to increase their abundance, confirm that diversity is desired on farms.

Almost all farmers that wanted a species in higher abundance explained this by its good characteristics, either related to its productivity or suitability to the on-farm niche. The question was asked why these species were not established in the desired abundance immediately. The main reasons that farmers provided were: germplasm was not available yet (86% of cases), a change in opinion about abundance needed (22%), recent changes in land allocation or having recently moved (12%) and/or space that had to be created by removing other species on the farm first (15%). The germplasm restrictions were often related to waiting until more wildlings (naturally regenerated trees) had occurred on the farm. As species were often represented by only one individual on the farm, most wildlings will probably be descendants from this single mother tree. Some farmers mentioned that trees had to be mature before cuttings could be taken. The wish to obtain seedlings of good quality was a far less frequent explanation for delays in establishing species at their desired abundance. Changes in opinion on the desired abundance were often related to older trees that would not produce that much any longer, new insights in the potential of some species (we did not probe how these insights had emerged), or additional needs of the household such as family growth. In a few cases, new boundary conflicts with neighbours resulted in a higher abundance needed for trees that demarcate external boundaries.

The main reasons why some species were wanted in lower abundance were that the species was not very suitable for the particular on-farm niche (45%), that the species did not perform very well (23%), or the related reason that other species performed better (23%). Most frequently mentioned for niche compatibility was that the species cast too much shade on crops in cropland or on the crops of the neighbour (for species that demarcated external boundaries). For the homestead area, reasons were that the species took too much space, produced excessive leaf litter, competed with other species, or had dangerously weak branches. Space requirements for other species was also mentioned for crop contours, which were reserved by some farmers for


176

Napier grass (Pennisetum purpureum) grown for fodder. Although one farmer explained that Spathodea campanulata had low importance, the farmer still wanted to keep one tree so that the species would not disappear from the farm. Bad production of a species was sometimes inferred from the performance of one individual tree (e.g. one farmer explained that Mangifera indica , a frequent species in the area, did not perform very well on the farm based on the performance of one tree). In two cases, farmers explained that their father knew the purpose of particular species, but that they did not, and that the species was therefore not wanted anymore on the farm. In four cases, the farmer explained that disease problems and not knowing how to properly manage the species were reasons not to want a particular species any more. In seven cases, the explanation was that fewer trees would produce enough.

In analogy to asking why farmers did not establish trees in higher abundance immediately where this abundance was desired, farmers were also asked why they established some species in higher abundance than required later. The reasons that farmers provided were: the farmer had expected the species to perform better (23%), the decision was only later made which species to establish (21%), not enough germplasm was available yet of a better species (11%), or that cropland was expanded later (5%). Where the farmer had expected better performance, the farmers sometimes mentioned their surprise because the species performed better at the neigbhour’s farm. Farmers that mentioned that they had only made decisions later motivated their decision by stating that small trees did not harm crop yield very much or simply explained that they had not had the time to remove species of lower priority. Where farmers expanded cropland, sometimes also soil fertility improving or soil erosion controlling species were wanted in lower abundance. Respondents reasoned in these cases that the fertility had already been added or that erosion had been controlled, which indicates that not all farmers understand the underlying ecological principles for service functions of trees. In seven cases, species that were established by ancestors were not wanted any longer. In one case, the farmer had expected the species to produce less, but since the species was more productive, fewer trees were wanted now.

Examples as cited above indicated that many farmers were experimenting what the best abundance of a particular species would be on their farms. Some farmers were testing the validity of local believes on the function of some indigenous tree species. Some farmers believe that Dracaena steudneri trees would prevent that thunderstorms would harm the farm. One farmer explained that he would not plant the species any longer, as it was not very effective for that purpose. Another farmer, however, explained that he would increase the abundance of the species from one to ten trees as he experienced that one tree was not very effective in preventing thunderstorms and, therefore, wanted to test whether more trees would provide the service.

Information on changes in abundance was also collected during the inventory phase. Farmers were asked whether each tree encountered would be replaced later, and, if answered in a positive manner, whether replacement would be done by the same species. The answers provided did not correspond very well to the information collected on desired species and abundances in the later interview. Thirty-five species (33%) were not mentioned as a species that replaced a tree present on farm during the first interview, while only seven species were not wanted anymore according to the information of the second interview. The opposite situation, where a species was indicated of being replaced but was not wanted anymore on the farm according to the second interview, only occurred once. The fact that replacement of a tree during the first interview was specific to the niche in which the tree was growing was not responsible for the differences between both interviews. Species that were not recorded as replaced during the first interview would still be established in the same niche according to the second interview. It is still possible that the differences between the two interviews are related to different spaces within each niche where the species could be planted. We expect, however, that differences between the interviews were related to the fact that farmers often wanted to keep a species on their farm albeit by keeping

CHAPTER 11

177

only one tree of the species as some farmers indicated, and that the second interview was better in revealing this strategy of farmers.

11.2.3. Reasons Provided by Farmers for Establishing and Maintaining Tree Diversity

After the participatory mapping exercise, and questions that related to desired changes in abundance of particular species, discussions were held with farmers that focused on the composition of species of use-groups that were maintained on the farm.

The first topic of the discussion formed the reasons of maintaining several or only one species within a particular use-group. Table 11.2 summarizes information for the most frequent use-groups, with post-classified reasons for their diversity (information in the following paragraphs for use-groups separately gives a better indication of the actual reasons that farmers provided). Differences in characteristics and continuous supply were the most common reasons for diversity. The reason of “differences in characteristics of species” indicates that farmers perceived that mixtures of species provided various characteristics to a use-group that were difficult to combine in a single species (e.g. fast growth and quality of timber). The explanation that each species had good characteristics was included as it occurred frequently, although this category is not an actual explanation why several species were desired. This category probably points to risk avoidance strategies in case one species in the use-group would not provide the use due to pests, diseases or other problems that would prevent that a particular species would provide the product or the service. Experimenting whether particular species would perform well indicates farmer willingness of diversification. The ‘testing’ category further shows that desired farm composition (section 11.2.1) may change over time – fewer species could be desired if species test negatively, or more species could be desired in case farmers test new species on their farms.

Respondents that desired different species for firewood mentioned the different growth and maturity rates and the fact that most species had another main purpose. Therefore, many species were needed to produce enough daily-needed firewood. Some farmers mentioned that they wanted to avoid that species would become extinct on their farm, what would happen if requirements of firewood would result in all its trees to be cut. This type of response does not directly reflect a farmer strategy of diversification, since extinction of species by firewood usage could be prevented by establishing enough trees of the species on the farm. However, the fact that farmers opt to use several species in lower abundance for firewood (with higher perceived extinction risks) rather than few species in higher abundance, probably reflects risk avoidance strategies. Few farmers said that the diversity simply resulted from natural regeneration. As these species were still included in the maps that reflected the diversity desired on farms, spontaneous recruitment of species does not mean that these species are not wanted – maybe these farmers are just happy with the spontaneous diversity and do not perceive the need to actively plant additional trees. Potentially the reasons of spontaneous regeneration and avoidance of species extinction are related, reflecting a strategy of protecting the various species that regenerate spontaneously which results in the desired quantity and diversity of trees that provide firewood.

Many farmers with different fruit species mentioned that they wanted to provide choice to family members and to customers, that the fruits had different tastes, or that some fruits were better for home consumption and others were better for sale. Some farmers explained that fruits differed in vitamin contents while a balanced supply was needed. Seasonality and differences in maturity were another important reason (of ensuring continuous supply) for diversity within this group. Sometimes the possibility of pest or disease attacks on particular species was mentioned (Citrus spp. often experience such problems), thus indicating risk avoidance strategies.

When farmers wanted several timber species, they mentioned different qualities of timber provided by different species. Some species were for sale and others for direct use on the farm.


178

Many farmers wanted to provide choice to customers. Several farmers listed different growth rates, other uses besides timber, and the need for a continuous supply as reasons for diversity. Some farmers said that all the species were marketable, as they had hard wood and linear timber. Risk avoidance was an important reason in the timber group because stem deformations were often expected and could not be predicted, and farmers felt that growing a diversity of species would result in a greater percentage of trees of good form. Many farmers mentioned that they wanted to test the growth characteristics of various new species for timber. This finding indicated that farmers were not familiar with timber production characteristics (quantity and quality) of several species, which is probably related to the longer time needed to obtain the product and greater differences in price/quality for timber.

Diversity within the construction wood was explained most often by needing two species for house construction: one species that was elastic and did not crack when nailed (characteristics typically provided by the local Markhamia lutea), and another species with fast growth and straight poles (most often provided by Eucalyptus spp.). One farmer said that it was a local taboo to use two species for house construction. Farmers that mentioned that several species had good characteristics explained that all species were strong, straight, termite resistant, long-lived, and with good sprouting after coppicing.

An important reason for different shade species was differentiation in shade density, so that the family had the option to choose the most suitable species for the conditions of a specific day. One farmer explained that different species offered different features to look at when sitting below them, which indicates that also personal or sentimental values can play a role in decision-making on diversity. Differences in growth rates and the fact that some species had other functions were given as reasons for diversity as well. One farmer explained that although mango had the densest shade, it was not safe to sit beneath it when it had fruits so that an alternative species had to be available under these circumstances. Farmers that stated that the various species all had good characteristics mentioned that they occurred in the homestead area, that they were rarely removed for other products and had a long lifespan, had wide crowns and thick shades. One farmer mentioned the possibility that one species would shed its leaves. Another farmer explained that it is prestigious to have shade trees around the house.

Some farmers with different species for boundary demarcation explained that some species were recognized by the governmental land surveyors and neighbours for land demarcation, while other species made the boundary more compact, were more beautiful, or produced timber besides demarcating the boundary. Some farmers mentioned the different growth rates of the species so that the fence could be compact shortly after tree establishments. Farmers that explained that each species had good characteristics said that they all were allowed on the boundary as they did not hinder the crops of the neighbours much, had small diameters, could be coppiced, and had a long lifespan. Some farmers made the differentiation between species that were better on external boundaries (between farms) and species that were better on internal boundaries (mainly around the homestead). On external boundaries, recognition of boundary demarcation by a particular species was more important, while on internal boundaries ornamental value had higher importance. The reason that different species were grown in different niches thus reflected differences in species characteristics – these reasons could have been classified in the first category for diversity (“species differ in characteristics”).

Some farmers with different medicinal species mentioned that they treated different types of diseases. A frequent reason that was mentioned was that species mixtures would be more effective, especially if no precise diagnosis could be made. Difference in growth rates was provided as well as a reason to grow diversity. One farmer mentioned that medicinal species were often attacked by diseases, so that many species were needed in order to ensure production of medicine.

CHAPTER 11

179

In the soil fertility group, neither difference in species’ characteristics nor continuous supply were mentioned. It could be concluded from table 11.2 that soil fertility improvement was only group where farmers did not perceive compatibility in characteristics of several species. This finding is maybe related to the short-term nature of service provision within this group through 3-6 month fallows and/or similarity among the nitrogen-fixing species. The finding could also reflect on-farm soil fertility research conducted by ICRAF on some farms that were included in the survey. As one of the objectives of the soil fertility research is selection of the best species for soil fertility improvement, farmers could be indicating to wait for results of this research and the concept of a superior species excludes consideration of compatibility among species.

Some farmers with various species for charcoal mentioned the reason that maturity varies among species, whereas charcoal is needed continuously. One farmer described differences in maturity as a risk limitation strategy because it was not known which species would reach maturity first.

For ornamental species, differences in appearance were mentioned as a reason for diversity within the group. The fact that several species had ornamental value was also provided as a reason for diversity within this group.

When farmers were asked for reasons for having only one species in a use-group, the most commonly provided reason was that the one species was the only species with good characteristics. The single species wanted for timber had straight and fast growth, and hard wood. Farmers explained that Eucalyptus spp. provided all necessary woody house construction elements, had fast growth, resprouted rapidly after coppicing, had a long lifespan, tolerated competition in a woodlot well, had straight stems, and was not very susceptible to diseases. Mangifera indica was the only species desired for shade as some farmers explained that the species produced a dense shade, had a large crown, did not shed its leaves often, and had a long lifespan. For boundary demarcation, farmers often mentioned that the species should be recognized by the governmental land surveyors and by neighbours. Other reasons that some farmers gave were that the species should be compact and propagated easily by cuttings. One farmer mentioned that it should be possible to nail barbed wire on the stem of the species. Some farmers mentioned that Azadirachta indica was the only species with an effective medical treatment, often mentioning a particular ailment. Some farmers explained that Sesbania sesban had proven its soil improvement properties, had fast growth, produced firewood, and was compatible with crops. In some cases farmers mentioned that the species was recommended by researchers and stated that the species produced nodules. The desire for mono-specific ornamental use-groups was explained as Terminalia mantaly having systematic branching and leafing patterns and not shedding its leaves often.

Table 11.3 gives information that touches on the other aspect of diversity besides richness: the evenness (or lack of evenness) in the abundance of the various species. The discussions were held with the same farmers that were interviewed about reasons to maintain more than one species in a use-group (their number is provided in table 11.2). The reasons that farmers provided were post-classified in the categories provided in table 11.3, whereas the discussion of the various use-groups provides the reasons that farmers gave more literally.

Difference in species performance was the most common reason provided for differences in abundance. Difference in growth rate, the age of reaching maturity, and the quality of the product were most frequently mentioned. For firewood, differences in coppicing potential, niche compatibility (the suitability to grow the species in a woodlot, on boundaries or mixed in cropland), thorniness, and tree size were listed. For fruit, farmers listed differences in yield, fruit taste, fruit size, seasonality (fruiting during the dry season), marketability, and susceptibility to diseases. For timber, the likelihood of stem deformations, tree size, marketability, niche requirements, and susceptibility to diseases were mentioned. For construction, differences in straightness, lifespan, coppicing ability, competition in woodlot arrangements, and the number of


180

products used for construction were listed. For shade, differences in tree size, shade density, frequency of leaf shedding, and the strength of branches were mentioned. For boundary demarcation, differences in appearance, recognition as boundary markers, size, drought tolerance, and thorniness were mentioned. Farmers provided differences in marketability, effectiveness, tree size, and the number of useful parts were provided for the medicine group.

The other important reason for differences in species abundance within the same groups that was provided was the other functions that some species had. One farmer for example explained to select the best Cupressus lusitanica individuals to grow into timber, while other trees were pruned within the fence to form a uniform surface. Reasons related to germplasm included difficulties to obtain germplasm, natural regeneration, easiness of germination, or the fact that cuttings of smaller sizes could be used. Small overall group abundance was a reason given for similarity in species’ abundances. Small group abundance was related to farm or family size, or niche requirements. New species are tested in small abundance. This practice probably often results in poor appreciation of a particular species as frequently only one tree is evaluated. The practice could also lead to severe inbreeding where farmers only test one tree and later use this tree as sole germplasm source for the farm.

In some cases, the same reason was used by some farmers to explain establishment of fewer trees, and by other farmers to explain establishment of more trees for the same species. For example, some farmers explained that fewer Euphorbia tirucalli trees were needed because they were recognized for boundary demarcation and thus not removed, while other farmers provided the boundary demarcation property as the reason to establish more trees. The abundance needed probably relates to the relationship among neighbours. Another example is that some farmers explained a low abundance by natural regeneration, what was a reason for high abundance for other farmers. In the former case, farmers said that the species only regenerated naturally, while in the latter case, the species also regenerated naturally. A third example was one farmer not establishing many Eucalyptus spp. trees because of their high marketing potential (a reason for their high abundance according to other farmers) because this factor made them also more likely to be stolen from the farm. A final example is high medicinal efficiency, which was the reason for some farmers to establish many and for other farmers to establish few trees of Azadirachta indica .

11.2.4. The Relationship between Species Ranking and Diversity

For each use-group and farm, information was collected on the preference ranking of each species. We asked farmers to rank species as this exercise is often conducted during Participatory Rural Appraisal to plan on-farm tree planting activities (e.g. Franzel et al. 1996). The hypothesis that we wanted to test was whether diversity within a group could be caused by differences among farmers. Since our results indicated that farmers desired diversity for the use-groups on their farms (see above), investigating the relationship between species ranking and diversity focused more on influences on the evenness within a group (i.e. that species with high priority would be established in much higher abundance).

Table 11.4 shows that no use-group had a single species that ranked first, since two (construction) up to 16 species (timber, shade) got first ranking on some farms. Therefore, if each farmer only established the species with first priority, use-groups would contain at least two species. For most use-groups, trees with first ranking constituted less than 50% of all trees in the use-group, which shows that farmers wanted to maintain many species of lesser priority. The only exceptions were boundary demarcation, construction, and soil fertility improvement. For these groups, however, species with first priority were only 13% - 32% of species that would be maintained on farms, which shows, again, that many species of lesser priority would be maintained.

CHAPTER 11

181

For some use-groups, the overall species of first priority could be determined quite easily based on the percentage of farmers that gave first priority to the particular species. The prime examples with high percentages of farmers that choose the same species are Eucalypus saligna for construction, Warburgia ugandensis for medicine and Sesbania sesban for soil fertility improvement, and, to a lesser extend, Euphorbia tirucalli for boundary demarcation. For these use-groups, the overall species of first priority also was dominant (having the largest abundance within the group).

For the other use-groups, however, other species were more frequent, which was often linked to other purposes of the species. Examples are Markhamia lutea for charcoal (proportion 0.50, also used for construction), Euphorbia tirucalli for firewood (0.32, also used for boundary demarcation), or Cupressus lusitanica as ornamental species (0.91, also used for boundary demarcation). This finding shows that a use-group may be dominated by a certain species as it has several purposes, and not because the species had the best ranking. The various purposes of some species thus result in the fact that, for some use-groups, the best species does not have the largest abundance. For the other use-groups, it was also more difficult to determine an overall species of first priority, as the highest percentage of farmers that gave the same species first ranking was only just above 50% (for fruit and shade).

The general pattern that we observed was, therefore, that although there was a relationship between priority and abundance of species (especially for boundary demarcation, construction, and soil fertility improvement), while farmers still opted to establish a range of species of lower priority. Although farmers could categorise some species as more important than other species, they were still interested in growing diversity.

11.3. Discussion

We observed that most of the diversity that was encountered on farms during an inventory was desired by farmers. During participatory mapping, most species were maintained. Farmers could explain why some species would be increased in abundance, and why several species were wanted within most use-groups. Moreover, rather than finding that only a fraction of diversity was actually desired, while the other fraction was only transient diversity caused by natural regeneration, we found that farmers wanted more diversity than what was recorded in their fields.

We should be careful with the interpretation of the results that we obtained, however. The exercise that collected information on diversity desired by farmers was based on a map of the diversity that was encountered on a farm. A part of ‘desired’ diversity could have been caused by this method, that could have simulated ‘protection’ of some species. The question remains whether many species would be actively planted if they would not regenerate naturally. It is a difficult question to investigate, however, since the situation of ‘no natural regeneration’ is hypothetical in the landscapes that we investigated. Discussions with farmers of hypothetical situations are not very useful in most cases as these only yields hypothetical answers (Fergus Sinclair, University of Bangor, pers. comm.).

Changes in composition would be realized in a very short time span, so that the information on desired composition could have been influenced by the easiness by which this composition could be attained from the present composition, and only reflect short term planning objectives of farmers. Although we explained to farmers that germplasm problems should not influence the ‘ideal farm’, farmers may still have reflected on current germplasm problems, which may also have restricted farmers in ‘establishing’ more species on ‘ideal’ farms. Knowledge problems could be the reasons that some farmers only wanted few species (those farms that are plotted in the lower-left corner of figure 11.3), so that increased information on the uses of species could


182

influence farmers in establishing more species. Better knowledge on species’ uses could increase diversity on farms that have low diversity at present, especially since we could not observe any saturation point for diversity on farms, and since farmers were not only desiring to grow species of first priority on their farms.

We found several indications that knowledge was a limiting factor for on-farm diversity. Some farmers explained that they did not know what their ancestors were using a species for. One farmer explained that neighbours were using bark from a species that occurred on his farm, but that they were not willing to explain the exact treatment. Knowledge problems are also reflected in indications of farmers that experimented with new species. Many farmers tried to enhance their knowledge by testing the performance of particular species on their farms. Since farmers were not only growing species of first priority, we expect that farmers were testing new species for diversifying the species composition of particular uses. A problem could be that farmers often based their judgment of the performance of a species based on the performance of a single tree, thus ignoring intraspecific variation. Some farmers expressed their surprise that the species was doing very well on their neighbour’s farm, but not on theirs. On the other hand, some farmers were testing species outside or at the limits of their ecological range – for example, one farmer was testing cocoa (Theobroma cacao), another farmer cashew (Anacardium occidentale) – so that testing one tree could be sufficient. Testing species with a single individual could result in reductions of genetic variation, as this single tree would often be the mother tree for the other trees to be established on the farm in case it performed well.

As farmers were increasing the abundance three-fold on the ideal farms (table 11.1), one could question how realistic they had been as not enough space may be available on their farms to sustain the abundance. Maybe a more realistic exercise would have been to limit farmers to the actual abundance, and let them choose one species for each available ‘planting hole’. We wanted to let farmers free in changing abundance of each species, however, to investigate changes in diversity related to changes in abundance. In addition, our experience was that space was still available on many farms to increase tree abundance.

As we expect that information on changes in diversity may be for the short term, and not realistic in all cases, the real test for changes in abundance and diversity will be by monitoring actual modifications that farmers make in on-farm composition. It is very likely that as farmers’ experience in existing and new species on their farms increases, that on-farm composition will progressively differ more from the present composition. Our results indicate, however, that future tree composition on farms, if different, will continue to harbor diversity. As described by Holmgren (2001), the planning process can be considered to consist of five steps, including inventory (collection of data), evaluation (organization of data), strategy (general direction for changes), design (particular direction of changes) and management (implementation), while feedback is necessary at every stage. Holmgren (2001) further states that only the landholder can effectively conduct whole-farm planning, as he or she is the only person that can consider all strategic, practical, and technical elements of the entire farm. As we found relatively small percentages of variation in abundance and richness explained by household characteristics, we also indicated that individual perceptions and management strategies differ among farmers of the same socio-economic characteristics. Under these conditions, the role for research or development seems to lay in providing farmers with information on consequences of possible management strategies (e.g. genetic erosion, increases of soil erosion) and information on other options (e.g. alternative species within a use-group, tree management).

It is interesting to note that fruit, timber, and medicine were the use-groups with largest increments in abundance. These use-groups were identified by ICRAF and its partners as priority groups to increase farmers’ incomes through agroforestry, and trees belonging to these groups were described as ‘high-value trees’ (ICRAF 2000). It is likely that farmers made a similar analysis,

CHAPTER 11

183

and were concentrating more on the marketable tree products. The analogous analysis suggests that activities that focus on these use-groups may be implemented easily.

The fact that most tree species provided several uses may have created the artifact that diversity was richer in some use-groups than strictly desired by farmers. This may especially have happened for the diversity of firewood, as firewood is a secondary product of most species. Adding a new species for timber, for instance, also resulted in addition of a new firewood species. Some of our discussions on diversity within a use-group may therefore not have been entirely relevant, as diversity was a by-product by diversity in other groups. Some farmers realized this confusing effect by explaining that diversity in the firewood group was caused by diversity in other use-groups.

We recorded some of the multiple criteria that farmers are using in decision-making regarding the trees that are maintained on a farm for the same purpose. Most criteria were based on the quantity and quality of the provision of products or services by trees. Growth rate, or growth after coppicing, reflect criteria of quantity. Taste, straightness, or density of the canopy are indicators of quality. Other criteria included personal values (“enjoyable to look at”, “preservation for future generations”), while other criteria were community values (taboos). At the same time, the fact that several use-groups were maintained on a farm shows that farmers are interested in growing diversity for the various products and services that can be provided by a diversity of species. Overall, our findings indicate that farmers are mainly interested in growing tree diversity for the variation in products or services that can be provided among and within use-groups. Within a use-group, tradeoffs in use characteristics were often listed as important criteria for diversity, for example beauty versus legality, dry season versus humid season production, or preference in the market versus preference by the family. Species within a use-group are therefore preferably compatible in characteristics. As, theoretically, for a new species with random characteristics, compatibility within a use-group can be achieved more easily when a use-group has few (at the extreme, only one) species, projects that aim at diversifying on-farm species composition could target use-groups that are currently of lower diversity. These groups are also buffered the least against production loss through non-performance of one species. Interventions could focus on introducing new species for the particular purpose, more widely distributing species that provide the purpose on few farms currently, or promoting new uses of species. Obviously, evenness should be considered besides richness, as groups of lower evenness will offer less compatibility and less buffering against risks.

Some of the reasons that farmers provided for diversity in tree species composition were analogous to reasons provided for farmers growing a diversity of crop varieties. Jarvis & Hodgkin (2000) list five aspects of farmer decision making that seem likely to have most influence the extent and distribution of genetic diversity. These include decisions on which agro-morphological characteristics are important, farming practices, planting niche, population sizes, and seed sources, of which we recorded criteria that apply to the first three decisions (we did not investigate farmer decision-making regarding intraspecific diversity in this paper). Reasons cited for the interest of farmers in crop diversity include the exploitation of different microhabitats (e.g. Colfer 1991; Haugerud & Collison 1990), different uses (e.g. Bedigian & Harlan 1983; Shigeta 1990), different maturities (e.g. Clawson 1985; Richards 1993), risk avoidance and yield stability (e.g. Voss 1992; Merrick 1990), perceptual differences (e.g. Boster 1985) and cultural diversity (e.g. Cleveland 1993). Other examples that farmers manage different crop varieties are provided in van der Heide & Tripp (1996). Farmers thus seem to make similar decisions regarding tree diversity.

Our results also agree with the information presented in Arnold & Dewees (1995) on the reasons that farmers grow trees on farms. In summarising the information from Arnold & Dewees (1995), Arnold (1995) conclude that farmers plant trees in pursuit of their livelihood goals of income generation, risk management, household food security and optimum use of available land, labour and capital. Tree species and varieties are thus planted to produce various products


184

and various services. Scherr (1995) for example reports on the results of a tree planting project in that trees are planted for building poles, fruit, green manure, shade, fencing, fuelwood, timber, ornamental, fodder and other uses. This researcher further stated that although most of the tree species provide several products and services, on-farm products and services are typically provided by several tree species.

In conclusion, we documented that farmers desire on-farm tree diversity. Our results also indicated that some use-groups and some farmers of lower diversity could be targeted by projects that aim at diversifying agroecosystems. These projects could mainly focus on the provision of additional information on species’ uses and consequences of management options, as management strategies of farmers of the same socio-economic characteristics are likely to differ widely, even when tree information and germplasm are distributed more widely than at present.

CHAPTER 11

185

Table 11.1. Average diversity (with standard deviation) that was present versus diversity wanted on a single farm for all species and for species that belonged to specific use-groups. Averages were calculated based on the farms where the use-group was present or wanted. S: species richness, H1: Shannon diversity index, BP: inverse Berger-Parker index (1/pdom), N: abundance. Group Number

of farms where

present

Number of farms where

wanted

S present

S wanted

H1 present

H1 wanted

BP present

BP wanted

N present

N wanted

All 78 78 15.8 (5.1) 16.6 (5.0) 1.5 (0.5) 1.4 (0.5) 2.4 (1.1) 2.1 (0.8) 416 (281) 1360 (921) Firewood 78 78 14.4 (4.6) 13.2 (4.2) 1.5 (0.5) 1.3 (0.4) 2.3 (1.0) 2.0 (0.7) 398 (886) 1210 (886) Fruit 78 78 4.7 (1.6) 5.0 (1.7) 1.0 (0.5) 1.2 (0.4) 1.9 (0.8) 2.4 (0.9) 39 (83) 59 (83) Construction 78 77 2.2 (0.7) 2.1 (0.7) 0.5 (0.3) 0.4 (0.3) 1.4 (0.3) 1.3 (0.3) 94 (530) 305 (530) Timber 77 77 4.4 (1.9) 4.8 (2.1) 0.7 (0.4) 0.7 (0.5) 1.4 (0.4) 1.5 (0.6) 107 (659) 493 (659) Boundary demarcation 77 76 2.4 (1.2) 2.3 (1.1) 0.4 (0.4) 0.4 (0.4) 1.4 (0.5) 1.3 (0.4) 213 (615) 796 (615) Shade 65 65 4.0 (2.1) 3.8 (1.9) 0.9 (0.5) 0.8 (0.5) 1.9 (0.9) 1.8 (0.8) 35 (428) 135 (428) Medicine 50 49 1.9 (1.1) 2.5 (1.6) 0.5 (0.5) 0.6 (0.5) 1.5 (0.7) 1.7 (0.8) 3 (29) 13 (29) Soil fertility improvement 53 49 1.5 (0.9) 1.5 (0.8) 0.2 (0.3) 0.2 (0.3) 1.1 (0.4) 1.1 (0.3) 23 (274) 126 (274) Charcoal 29 27 2.1 (1.4) 2.0 (1.3) 0.3 (0.4) 0.3 (0.4) 1.2 (0.4) 1.2 (0.3) 12 (27) 12 (27) Ornamental 28 22 1.4 (0.6) 1.5 (0.7) 0.2 (0.3) 0.2 (0.3) 1.1 (0.3) 1.1 (0.2) 8 (118) 16 (118)

Table 11.2. Main categories of reasons provided by farmers to have several species in a use-group. Figures indicate the number of farmers that listed these reasons, with figures between brackets reflecting the percentage of the farmers that desired diversity (>1 species) of the use-group, except in the first column where the figure expresses the percentage of all farmers with the use-group on their farms. Use-group Households Main reasons for diversity that desire

minimum two species

Species differ in charac-teristics

Continuous supply

Each species has good charac-teristics

Species use different niches

Risk avoidance

Testing

Firewood 78 (100) 2 (3) 28 (36) 23 (29) 0 (0) 5 (6) 0 (0) Fruit 77 (99) 25 (32) 44 (57) 14 (18) 1 (1) 3 (4) 2 (3) Timber 71 (92) 27 (38) 18 (25) 15 (21) 1 (1) 10 (14) 15 (21) Construction 68 (88) 49 (72) 3 (4) 14 (21) 0 (0) 2 (3) 0 (0) Shade 58 (89) 16 (28) 14 (24) 27 (47) 0 (0) 1 (2) 0 (0) Boundary demarcation 57 (75) 26 (46) 3 (5) 16 (28) 8 (14) 1 (2) 0 (0) Medicine 35 (71) 23 (66) 9 (26) 9 (26) 0 (0) 0 (0) 2 (6) Soil fertility improvement 15 (31) 0 (0) 0 (0) 10 (67) 0 (0) 0 (0) 0 (0) Charcoal 14 (52) 2 (14) 5 (36) 3 (21) 0 (0) 1 (7) 0 (0) Ornamental 9 (41) 2 (22) 1 (11) 2 (22) 0 (0) 0 (0) 0 (0)


186

Table 11.3. Main categories of reasons provided by farmers for differences in desired tree number among various species that belong to the same use-group. Figures indicate the number of farmers that listed these reasons, with figures between brackets reflecting the percentage of the farmers that desired diversity (>1 species) of the use-group Use-group Main reasons for differences in abundance Species differ in

performance Species have other

purposes Differences in

germplasm availability

Small total group abundance

Some species are only tested

Firewood 30 (38) 49 (63) 3 (4) 0 (0) 0 (0) Fruit 56 (73) 14 (18) 5 (6) 1 (1) 2 (3) Timber 28 (39) 33 (46) 4 (6) 2 (3) 7 (10) Construction 53 (78) 5 (7) 4 (6) 0 (0) 0 (0) Shade 26 (45) 18 (31) 1 (2) 4 (7) 1 (2) Boundary demarcation 42 (74) 4 (7) 2 (4) 0 (0) 0 (0) Medicine 19 (54) 6 (17) 1 (3) 1 (3) 3 (9) Soil fertility improvement 7 (47) 4 (27) 0 (0) 0 (0) 0 (0) Charcoal 6 (43) 2 (14) 3 (21) 0 (0) 0 (0) Ornamental 3 (33) 2 (22) 0 (0) 0 (0) 0 (0)

Table 11.4. Distribution of species of first priority within use-groups. The first two columns list the distribution of those trees that got first priority on a minimum of one farm. The other columns document the distribution of the overall species of first priority (the species that got the highest number of first rankings). Group Number of Percentage of Dominant species with ranking 1 species with

priority 1 (% of all species)

trees with first priority within a

group

Identity Number (% of farms) of farms

with rank 1

Number (% of farms) with rank

2

Percentage of group

abundance Firewood 11 (11) 31 Eucalyptus saligna ° 24 (36) 22 (33) 20 Fruit 4 (21) 21 Persea americana ° 36 (52) 18 (26) 16 Timber 16 (42) 19 Grevillea robusta ° 22 (37) 11 (19) 3 Construction 2 (13) 81 Eucalyptus saligna ° 52 (98) 1 (2) 81* Shade 16 (28) 40 Mangifera indica ° 23 (52) 11 (25) 3 Boundary demarcation 6 (32) 68 Euphorbia tirucalli ° 23 (56) 9 (22) 49* Medicine 6 (16) 44 Warburgia ugandensis 20 (69) 2 (7) 28* Soil fertility improvement 3 (14) 58 Sesbania sesban 8 (80) 0 (0) 70* Charcoal 8 (33) 14 Bridelia micrantha 2 (22) 0 (0) 4 Ornamental 4 (17) 7 Terminalia mantaly ° 2 (40) 0 (0) 3 °: exotic species, *: the species has the largest abundance in the group

CHAPTER 11

187

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80

Number of farms

0

10

20

30

40

50

60

70

80

90

100

110

120N

um

ber

of

spec

ies

0.4

0.5

0.6

Par

amet

er z

of

S=a

N^z

richness presentrichness wantedparameter z presentparameter z wanted

Figure 11.1. Species accumulation curve for the umber of species wanted and the number of species present (left axis), and the

accumulation curve for parameter z of the model S=aNz(N) (right axis).

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

EX

P(H

)diversity wanteddiversity present

Figure 11.2. Diversity present versus diversity wanted represented by the Rényi diversity profile for fixed scales of the profile.


188

0 5 10 15 20 25 30 35

Species richness present

0

5

10

15

20

25

30

35

0

5

10

15

20

25

30

35

Sp

ecie

s ri

chn

ess

des

ired

Figure 11.3. Number of species wanted versus number of species present. Full line represents the robust linear regression and

dotted lines the 95% confidence intervals (54% variation explained)

5 10 15

Number of uses present

0

5

10

15

0

5

10

15

Num

ber

of u

ses

desi

red

Figure 11.4. Number of uses wanted versus number of uses present. Full line represents the robust linear regression and dotted lines the 95% confidence intervals (16% variation explained). The size of the symbol corresponds to the number of farms with

the same number of uses present and desired (maximum = 8).

CHAPTER 11

189

0.0 0.5 1.0 1.5 2.0

ln(inverse Berger-Parker index present)

0.0

0.5

1.0

1.5

2.0

0.0

0.5

1.0

1.5

2.0ln

(in

vers

e B

erg

er-P

arke

r in

dex

des

ired

)

Figure 11.5. Inverse Berger-Parker index (1/pdom) of trees present versus trees wanted. Full line represents the robust linear

regression and dotted lines the 95% confidence intervals (8% variation explained)

1.5 2.0 2.5 3.0 3.5

Abundance present (log 10)

1.5

2.0

2.5

3.0

3.5

4.0

1.5

2.0

2.5

3.0

3.5

4.0

Ab

un

dan

ce w

ante

d (

log

10)

Figure 11.6. Relationship between abundance present and abundance wanted. Full line represents the robust linear regression and

dotted lines the 95% confidence intervals (14% variation explained)

CHAPTER 12 INTRASPECIFIC DIVERSITY OF TREES ON FARMS IN WESTERN KENYA – RESULTS FROM SIMULATIONS OF

INDIVIDUAL-BASED METAPOPULATION DYNAMICS R KINDT, AJ SIMONS, P VAN DAMME & D REHEUL IN PREPARATION

We investigated the distribution of species on farms in four villages in western Kenya, and what the influence of this distribution was on intraspecific diversity under the present farmer management using information on the preferred origins and forms of germplasm of each individual tree. In total, 160 species were investigated out of the 175 species encountered in the landscape surveys. Results showed that species were aggregated within farms and within villages. We therefore expected that the preferred origin of the own farm would have strong effects on the dynamics of intraspecific diversity. However, as information on reproductive ecology and intraspecific diversity currently present in the metapopulations (350 situations where a species occurred in one of the villages, with farms harbouring subpopulations) was very scarce, we could only make approximate investigations based on stratified-randomized geneflow. Since we assumed that selfing was not possible for most species, and that pollen had 25% chance of originating from outside the village – using some general data on reproductive ecology of tropical tree species – obtaining germplasm from the tree that was replaced resulted in a smaller percentage of homozygous trees than by obtaining the germplasm at random from the same farm or at random from the same village. The highest proportion of homozygous trees were obtained for species with an abundance of 2-10 trees, but the highest number of homozygous trees is expected for species with abundance > 10. The differences among the different scenarios of germplasm acquisition were small, however, which was related to comparing genetic diversity after 50 flowering seasons, assuming no inbreeding in the first season, and a relatively high percentage of long-distance pollen migration. Small population sizes for most species indicate, however, that present levels of genetic diversity may be very low, and that introduction of more diverse germplasm might increase genetic diversity substantially. Our metapopulation model could be used when more information becomes available. Simulations with the model showed that good approximations were obtained for theoretical metapopulations.

GENETIC DIVERSITY IN METAPOPULATIONS

192

12.1. Introduction

One of the purposes of agroforestry tree domestication is the enhancement of the stability and productivity of agroecosystems by diversifying on-farm tree species composition (species richness and evenness of species abundances). Diversification and intensification of land use through the domestication of agroforestry trees (woody perennials) is one of the three pillars of ICRAF’s research agenda (ICRAF 1997; ICRAF 2000). In this chapter, analyses of genetic diversity of species that are grown on farms in western Kenya are conducted through individual-based metapopulation simulations based on the observed distribution of trees on farms, tree age distribution, and forms and origins of germplasm used by farmers. Although knowledge of reproductive ecology of most tropical tree species is very limited (Boshier 2000), we believe that using general assumptions of geneflow in tropical species still provides insights that can aid in formulating recommendations on intraspecific management of trees on farms.


Metapopulations (4.6)

12.3. Results

12.3.1. The Distribution of Species on Farms

Figure 12.1 shows the rank-abundance graph of the 175 species encountered in the survey. Only for 47 species (27%), more than 50 trees were counted. Only 30 species (17%) occurred in densities larger than one tree ha -1. Of 44 species (25%), only a single individual was encountered, while for 18 species (10%) only two trees were counted.

Investigating the density distribution of species showed that all species were aggregated within farms since the normalised average farm density was greater than one (above the reference line in figure 12.2). The general trend presented in figure 12.2 is that species with lower abundance have a more aggregated. This pattern obviously occurred for species with an abundance of 1 as they could only occur on one farm – the range of values observed for species of this abundance only indicates the range in farm sizes. Eucalyptus saligna occurring in the highest abundance had the lowest aggregation, with a value of 1.03 for the normalised within-farm density. However, the second most abundant species, Camellia sinensis, and some other species with abundances larger than 1000, clearly had an aggregated distribution. For many species, the trees were aggregated within villages as well, as can be seen from within-village values above the reference line.

Figure 12.3 shows the relationship between the logit-transformed farm frequency and the logarithm of species density. In accordance to the Density-Abundance relationship, species that occurred on a larger number of farms had, on average, a higher abundance as indicated by the regression line. However, many species with high densities had lower farm frequencies (represented below the regression line), while few species had the opposite distribution.

12.3.2. Origins and Types of Germplasm

Table 12.2 shows a cross-tabulation of the actual and farmer-preferred forms and origins of germplasm, averaged over all species and expressed in percentage of total abundance. Table 12.2 shows that cuttings were the most important germplasm source for trees. The percentages of seedlings, wildlings, and seed were each only about half of the percentage of cuttings. Fruit was

CHAPTER 12

193

not a major germplasm source. The majority of trees were obtained from within the same village, with slightly more than half of the trees obtained from the own farm. Most cuttings were obtained from neighbours, most seedlings from outside the village, and most wildlings and seed from the own farm.

Investigating the farmer-preferred origins showed that the own farm was preferred in a higher percentage than the actual number of trees originating from this origin. Farmers especially expressed to prefer more cuttings and seed from within the own farm. Seed had 10% larger preference than the actual number of trees established by this form of germplasm. The 10% difference corresponded to fewer seedlings that farmers preferred to obtain. The pattern that emerges is that farmers would produce their own seedlings from seed obtained from their own trees, rather than sourcing them from outside. Cuttings were preferred in a similar amount as for the actual preferences, but they would preferably not be obtained from the neighbours. The actual and preferred number of wildlings did not differ much, where the majority would be obtained from the own farm. Of the wildlings that were obtained on the farm, 95.5% were not transplanted.

12.3.3. Genetic Diversity in Hypothetical Metapopulations

Figure 12.4 shows the Rényi profiles calculated on the allelic frequencies obtained after 50 seasons. Table 12.3 provides the averages and ranges of profile values at scales 0, 2 and 8, and information on the evenness of the distribution of the dominant allele, H8 -H0.

Random mating resulted in four alleles remaining in the metapopulation after 50 generations, with the dominant allele occupying about half of the loci (pdom=0.49). The age-structured metapopulation (scenario 1) was the most diverse, while the two metapopulations where 1% of pollen was obtained from the outside were also more diverse than the ideal population. However, on average, these populations did not contain many more alleles (respectively 6.8, 6.7, and 5.4) than the ideal population, while the dominant allele was less evenly distributed. Obtaining 50% wildlings and 50% cuttings resulted in diversity that was very similar to the diversity of the ideal population, with an average of 4.01 alleles.

The metapopulation where trees only had 50% chance of flowering each season was the least diverse containing 2.1 alleles on average with pdom=0.70. The metapolation with a female/male ratio of 4 was the second least diverse (2.6 alleles, pdom=0.64). Although the dominant allele had a larger frequency than the dominant allele in the ideal population, these alleles were more evenly distributed.

The metapopulations where trees had higher chances of obtaining pollen from the same tree or the same subpopulation were less diverse than the ideal population. Within these metapopulations, the metapopulation with the lowest chance of obtaining pollen from the same tree (Pm=0, resulting in a chance of 0.10) was the most diverse (3.9 alleles, pdom=0.505). The subpopulations with the highest chance of pollination from the same tree (Pt=Ps=20, resulting in a chance of 0.6452) had the lowest diversity. A higher chance of obtaining germplasm from the same subpopulation in the latter two metapopulations resulted in higher diversity. Avoidance of selfing (Pt=0) or only obtaining seedlings from the same subpopulation resulted in higher diversity than the ideal population, with alleles=4.05 and pdom=0.492, and 4.09 and 0.487, respectively. The differences with the ideal population, however, of the metapopulations with higher or lower selfing, or higher or lower chances of obtaining germplasm from the same subpopulation, were not that large, spanning average allelic richness of 2.6 – 4.09.

Investigation of differences between the directly calculated Simpson index and 2-H2 revealed that direct calculation resulted in values that were on average 2.39 (1.12 - 3.47)% higher, with highest


194

differences for lowest values. Calculation of Ne from 2-H2 would thus result in higher estimations. However, both values indicated the same trends.

Observed homozygosity (table 12.3) differed most strongly from the expected heterozygosity (based on the allelic frequencies) for those metapopulations that had higher chances of obtaining pollen from the same tree or subpopulation. Calculating the ratio between Ne based on the observed heterozygosity, and Ne based on the Simpson index (table 12.3), resulted in ratios of 0.77 (Pt=20), 0.54 (Pt=Ps=20), 0.50 (Pt=Ps=Ws=20) and 0.92 (Pm=0). Increases in homozygosity thus followed the trend of lower diversity (see above), while obtaining more germplasm from the same metapopulation increased homozygosity. The mixed-scenario metapopulation, which was the third most diverse in allele frequencies, had the lowest heterozygosity (21.7) and ratio (0.27). Avoidance of selfing or obtaining germplasm only from the same subpopulation lowered homozygosity, however not to a large degree (respective ratios of 1.02 and 1.001). Ratios for the other metapopulations were also within the 0.99 (50% flowering) - 1.04 (age-structured) interval.

12.3.4. Changes in Genetic Diversity for Species Encountered on Farms in Western Kenya

Figure 12.5 shows the average number of homozygous trees encountered for the various scenarios of germplasm forms and origins. Obviously, avoidance of selfing resulted in no homozygous trees encountered for species with abundance 1. The highest proportion of homozygous trees was encountered for species with abundance between 2 and 10. Values of the proportion of homozygous trees above the reference line (dot-dashed) representing the reciprocal of species abundance indicate a chance of obtaining, on average, at least 1 homozygous tree. Comparison with the reference therefore indicates that in most cases no homozygous trees are expected for species with abundance < 10, while more homozygous trees are expected for species with larger abundance. The fact that many metapopulations had values larger than 0 in figure 12.5 however indicates that, even for metapopulations where on average fewer than one homozygous individual occurred, in some cases homozygous individuals were obtained

The reference line intersected the Supersmoother (S-Plus 2000) lines of species dominated by wildlings around abundance 22, and species dominated by seeds or seedlings around abundance 37 for scenario a, which used the farmer-preferred forms and origins. Only obtaining sexually generated germplasm from the replaced tree (scenario b) resulted in an intersection at abundance 200, while obtaining these forms of germplasm randomly from the farm (scenario c) resulted in an intersection at abundance 17. The scenario of selecting seeds at random from the village (scenario d) resulted in an intersection around 20. Scenario d resulted in the largest number of metapopulations with abundance < 10 above the reference line (5.5%), while scenario c resulted in the largest number of metapopulations of abundance = 10 above the reference line (53.7%).

Figure 12.6 shows the average expected number of homozygous trees encountered if alleles would be sampled at random from the metapopulations. For scenario a, intersections for sexually obtained germplasm were observed for abundance 4, and for cuttings at abundance 5. Intersections for sexually obtained forms of germplasm occurred at abundances 3 in scenarios b, c and d. Obviously, no intersection occurred for cuttings in scenario b. An intersection occurred for this form of germplasm in scenario c at abundance 14. Scenario c resulted in the largest number of metapopulations of sizes <10 and = 10 above the reference line (31.1% and 98.2% respectively). The smallest number of metapopulations < 10 above the reference line was observed for scenario a (17.1%), and the smallest number of metapopulations = 10 for scenario b (58.5%).

Figure 12.7 shows information on allelic richness and the frequency of the dominant allele. In general, the metapopulations with the largest numbers of alleles had the lowest frequency for the

CHAPTER 12

195

dominant allele, as the correlation between both factors was high for the various scenarios (0.97-0.98). Species with larger abundances had the largest diversity (richness and evenness) on average. Propagating trees through cuttings resulted in more evenly distributed alleles at a fixed allelic richness level, which is indicated by the line corresponding to this form of germplasm being above the line corresponding to sexually formed germplasm. Only for scenario b and for species completely dominated by cuttings was the proportion of the dominant allele equal to the allelic richness, i.e. the alleles were perfectly evenly distributed. This obviously resulted from loosing none of the alleles when obtaining cuttings from the same tree. In the other cases, the dominant allele was about four times more frequent than the perfect even distribution, with an increase of the frequency at larger richness.

12.4. Discussion

12.4.1. Species Distributions and Origins and Forms of Germplasm

In figure 12.1, we observed the inverse J-shaped rank-abundance distribution typical of tropical rain forests (Kohyama 1991; Magnussen & Boyle 1995; Alvarez-Buylla et al. 1996). The aggregation of species corresponded to aggregation patterns observed in several tropical forests (Condit et al. 2000). The density-abundance relationship could have resulted from metapopulation dynamics, since a simple metapopulation model resulted in that particular relationship (Hanski & Gyllenberg 1997).

The low percentage of wildlings (17%, table 12.2), and thus of natural regeneration, shows that the management of tree species diversity on farms (especially tree establishments and removals) also resulted in patterns of species distributions that are similar to those observed in natural ecosystems. Investigation of the influence of tree abundance on the results of table 12.2 (where percentages were calculated on the total number of trees) by linear regression did not indicate that trees of lower abundance were more frequently established by natural regeneration (R2=0.023 of proportion of wildlings against village-abundance). However, the 10 most frequent species-village combinations (53% of abundance) represented a very low percentage of all trees obtained from wildlings (2.5%). The two most frequent species within villages (21.6% of abundance) represented a higher percentage of trees obtained by cuttings (44.6%).

The origins of seedlings recorded in this survey contrasted with findings of a study conducted by Kindt (1997) in a larger area that included the survey area, where only 9% of farmers obtained germplasm from their own farm. However, information in Kindt (1997) was only collected from one species on each particular farm (selection was geared at collecting ethnobotanical information from a large range of species, while interviews were limited in time), while data of table 12.2 represented all species. Results from other surveys conducted respectively in Peru (Brodie et al. 1997), Kenya and Uganda (Lengkeek et al. in prep.) and peri-urban Nairobi (Muriuki & Jaenicke in prep.) also showed the importance of the own farm and the neighbours’ farms as sources of germplasm.

12.4.2. Calculations on Idealised Metapopulations

We calculated the expected Ne for generation t+1 based on the Simpson index observed in generation t. We thus avoided calculating Ne for generation t which should be based on sampling without replacement (e.g. Vitalis & Couvet 2001). Basing Ne on the average Simpson index value also avoided calculating harmonic rather than arithmetic means of Ne values.


196

Our simulations showed a strong increase of identity by descent in all the metapopulations that we studied, including those metapopulations where 1% mutations or migration from outside were allowed. A relatively low range of alleles remained in season 50 (averaging 2.1 – 6.8) from the original set of 100 or more (with 1% mutations, 50 new alleles are expected to have occurred from generation 1-50 as 100 gametes are needed in each generation). The parallel pattern of the diversity profiles indicates that differences in diversity are mainly the result of differences in allelic richness, rather than differences in the distribution in the evenness of the alleles. (Differences in evenness are reflected by differences in steepness of the diversity profiles, Kindt et al. – Chapter 6). Therefore, differences in Ht mainly occurred because more or fewer alleles were lost, rather than caused by changes in their distribution. Allowing new alleles to enter the metapopulation, in contrast, resulted in larger losses of evenness.

We did not obtain the average Ne (=50) expected for the randomly-mating population, although we obtained a close approximation (50.3). The range in values obtained in table 12.3 indicates the stochastic variation in genetic diversity in actual populations – it is possible that all alleles are lost in one generation, or that none disappear in 50 generations. Not obtaining the exact Ne did therefore not come as a surprise.

Our results also agreed with studies that calculated the expected average values for several exceptions to randomly-mating populations. For instance, Ne of a population with a female/male ratio of 4 equals 32

111 )4()4( −−− += mfe NNN (Caballero 1994; Yeh 2000),

whereas we obtained Ne=30.5. The average Ne of 50.8 for metapopulations where selfing is not allowed corresponded to the predicted values (˜ N+0.5) (Caballero 1994). The observation of Ne of 24.8 in the metapopulations where trees only had 50% chance of flowering also corresponded to the effective population size concept. For models with overlapping generations, Caballero (1994) only considered the asymptotic, smaller Ne. We observed higher values of Ne, but considering that trees in season 50 were in generation 10 shows that the obtained homozygosity was substantially higher than expected for non-overlapping generations (9.56). Wang (1997) calculated the inbreeding coefficient for metapopulations consisting of subpopulations of the same size, receiving the same proportions of migrant pollen and seed. Using our software on metapopulations of 20 subpopulations of 20 individuals each with selfing rate of 0.1, receiving 10% of migrant pollen and no migrant seed, or receiving no migrant pollen and 10% migrant seed, resulted in similar values of F20= 0.205 and 0.159 as obtained by Wang (1997).

Wright’s F-statistics (Wang 1997; Whitlock & Barton 1997; Nagylaki 1998; Kaj & Lascoux 1999; Yeh 2000) investigate genetic diversity at the subpopulation level. As our focus was on genetic diversity in the total metapopulation and observed heterozygosity, we did not report F-statistics above, although these were calculated. Pt=20 resulted in FIS (reduction of heterozygosity in an individual due to non-random mating in the subpopulation) of 0.16, and FST (reduction in heterozygosity due to metapopulation subdivision) of 0. Respective values of FIS and FST for Pt=Ps=20, Pt=Ps=Ws=20 and Pm=0 were: 0.43 and 0.06, 0.42 and 0.13, and 0 and 0.04. The F statistics thus reflected the level at which non-random mating occurred.

In our calculations, each other subpopulation had the same chance of being selected as a source for germplasm or pollen. Spatially-realistic metapopulation models, such as the incidence function model (Hanski 1998; Hanski 1999 pp. 81-87), calculate the colonization potential of a habitat patch from its distance from other patches. Stepping-stone (e.g. Epperson 1999; Hu & Ennos 1999; Austerlitz et al. 2000) and grid-based models (e.g. Tilman et al. 1997) are other examples of metapopulation models based on spatial arrangement of subpopulations. Differences in the chance of each subpopulation to exchange genes with other subpopulations could be incorporated in calculations by providing specific chances (e.g. Pij and Wij) of geneflow from

CHAPTER 12

197

subpopulation i to subpopulation j, based on subpopulation locations to expand the calculations documented in table 12.3. The weights correspond to the migration matrices considered in various metapopulation models (e.g. Whitlock & Barton 1997; Nagylaki 1998; Epperson 1999; Vitalis & Couvet 2001). The calculations could be expanded to individual trees, or take into account characteristics such as boundaries between subpopulations, or patterns in pollinator migration or seed dispersal.

Giving an allele that originates from outside the metapopulation a new identity is conceptually similar to infinite allele models. Gavrilets (1999) discussed the metaphor of holey adaptive landscapes in which metapopulations can be represented by networks of connected genotypes of higher fitness, whereas holes correspond to genotypes with zero fitness. Natural selection selects against mutations resulting in genotypes closer to holes, therefore not every mutation has the same chance of remaining in the metapopulation. K-allele models (e.g. Vitalis & Couvet 2001) limit the number of possible allelic states. It is obvious that our calculations are based on identity by descent, and that alleles that are not identical by descent may therefore be identical in state.

12.4.3. Calculations for the Species Encountered on Farms in Western Kenya

Obtaining the germplasm from the same tree resulted in smaller percentages of homozygous trees (also under random sampling from the metapopulation – figure 12.6) than obtaining germplasm at random from the same farm or from the village. For small metapopulations, sourcing germplasm randomly from the metapopulation resulted in more homozygous trees, while in larger metapopulations sampling from the same subpopulation resulted in more homozygous trees. The underlying process seemed to be that sampling from the same tree ensured that more trees contributed to the next generation, and that therefore fewer alleles were lost, especially since selfing was avoided. The larger homozygosity of larger metapopulations where germplasm was obtained from the same subpopulation is then probably caused by differences in subpopulation sizes, with larger drift in the small subpopulations. Obtaining germplasm from the same tree thus seemed to indicate a positive relationship between the degree of stratification in the sampling and the obtained homozygosity. However, the differences among the results of the various scenarios were very small, as patterns in the subfigures of figures 12.5 and 12.6 were similar. The similar patterns are probably caused by the short time horizon in our investigations of 50 seasons only, considering the perennial nature of trees (the weighted average maximum age of trees in the metapopulation model was 19.1 seasons, for example).

Advising farmers to obtain their germplasm from the tree that is replaced obviously does not enable farmers to select superior mother trees. The findings on the importance of stratification of the sample suggest that, where farmers select superior mother trees, the best method of maintaining genetic diversity will be by documenting germplasm origins, so that a reasonable population size can be maintained. Even in situations where farmers do not select superior material, mixing germplasm with documented origins may enable maintaining larger effective population sizes. Our findings show that sampling should be done in a randomly-stratified manner (i.e. ensuring that every mother tree contributes to the next generation) to prevent genetic erosion from germplasm mixtures. Stratified sampling could also avoid outbreeding depression in case populations had undergone evolutionary adaptations to differing local conditions.

Many plants are not diploids as in our simulations, but polyploids (Mahy et al. 2000; Miller & Venable 2000; Wu et al. 2001) containing more than two haploid sets of chromosomes. Polyploids are either allopolyploids derived from distinct genomes, autopolyploids derived from similar genomes, while many intermediate types exist (Wu et al. 2001). Polyploids generally harbour more genetic variation than their diploid relatives (Mahy et al. 2001). The simulations could easily be expanded to polyploid species by introducing additional rules how the larger


198

genepool is sampled to check whether higher average genetic diversity levels are obtained in these species (see below).

Although our calculations were based on a single locus, our predictions of the percentage of trees that are homozygous for a single locus are indicative of the average percentage of loci within the genome of a tree that are expected to be homozygous. Information would however be required on the levels of association between alleles (linkage disequilibrium or ‘hitchhiking’) and whether alleles are deleterious or not, to arrive at precise predictions for entire genomes (e.g. Deng & Fu 1998; Saccheri et al. 1999).

In our calculations, and due to a lack of documentation, we made various assumptions on geneflow, such as the chance of pollination from the same tree, the amount of germplasm that is likely to originate from outside the village, the chance that a tree had flowers in a particular season, or that species were diploid. Other studies have also suffered from lack of information. Cain et al. (2000) pointed out that the patchy nature of many landscapes makes long distance seed dispersal of many spatially isolated species a necessary, although unusual, event, but that models that incorporate information on the spatial arrangement of populations typically suffer from poor information on dispersal. Hanski (1999) mentions that a significant amount of habitat is often unoccupied, indicating limitations in movement of many species. Dynesius & Jansson (2000) described that species at higher latitudes are often more vagile than tropical species (which would be caused by Milankovitch climate cycles, periodical changes in the orbit of the Earth leading to variations in insolation and climatic change), while Dalling et al. (1998) and Hubbell et al. (1999) report that dispersal limitations in tropical forests result in their higher species richness. Alvarez-Buylla et al. (1996) mentioned that metapopulation models for tropical forest areas all assume some naturally caused longer distance dispersal. Cordeiro & Howe (2001) found that recruitment of animal-dispersed trees was three times smaller in small forest fragments (< 9 ha) than in larger fragments of three times their size, whereas recruitment of wind- and gravity dispersed trees was unaffected. As most tropical trees bear fruit adapted for animal dispersal, these authors, therefore, expect tree recruitment limitations for most species following forest fragmentation. To estimate the effects of seed dispersal on genetic diversity, information is also needed on other factors explaining tree recruitment. Curran et al. (1999) for example listed seedling establishment as a major bottleneck in dipterocarp recruitment. Clark et al. (1999) present a literature review on studies of recruitment limitation in forests and concluded that information was needed on the various contributing factors of fecundity, dispersal, establishment, and growth and mortality of various growth stages of trees, especially since these factors interact.

Momose et al. (1998) list various long-distance pollinators observed in Lambir (Malaysia). Hamrick & Nason (2000) mention that pollen flow can be quite extensive (>25%) over distances of one kilometre – this figure inspired the parameter value of Po=0.75 in our calculations. Although individual insect pollinators may only fly in between near neigbours, pollen carryover (i.e. pollen transfer from flower A? B by pollinator X and subsequently from B? C by pollinator Y) may result in pollination of several hundred meters (Hamrick & Murawski 1990; Dawson et al. 1997). Chase et al. (1996) found that isolated trees could act as stepping stones for geneflow among populations. Young & Boyle (2000) indicate that pollen flow can be high in fragmented populations, provided vectors can pass non-forest habitat. Young & Merriam (1994) and White et al. (2002) showed that fragmentation could actually lead to an increase in pollination distances. Aldrich & Hamrick (1998) indicated further complexities in metapopulation dynamics as they found that 68% of seedlings in forest remnants originated from remnant adults in surrounding pastures. Stacy et al. (1996) reported the combined effects of subpopulation size and species aggregation, so that plants in small clusters received more pollen from outside than plants occurring in larger clusters or in more even distributions. These studies point towards potentially complex patterns in pollen flows. To investigate the influence of pollen flow on genetic diversity, pre- and postzygotic processes should be considered (e.g. self-incompatibility, Boshier 2000).

CHAPTER 12

199

More precise predictions of changes in genetic diversity should not only rely on measurements of the factors mentioned above, but also consider factors such as differences in flowering phenology or the influence of agroforestry practices on flowering and pollen and seed vectors. Tree removal could be specified for each tree and each season, possibly linked with productivity of the individual tree and farmer decision-making, while better information could be obtained on actual subpopulation sizes. The greatest weakness in our approach in estimating genetic diversity, however, was not the uncertainty in tree-specific parameter values, but calculating the probability in identity by descent from the present generations of trees, ignoring effects of past geneflow and population sizes, and assuming that each allele that originated from outside had a new identity. In case initial genetic diversity would have been quite low – which is likely if patterns of germplasm acquisition have a long history in the area (table 12.2) –, then actual heterozygosity would decline much faster. However, as we wanted to investigate the influence of various management options on genetic diversity, we were still provided with changes in genetic diversity that could be used for comparative purposes.

In how far loss of genetic variation poses dangers to inbreeding depression remains to be investigated for each particular species and situation. In comparison with other plant species, trees contain high levels of molecular diversity (Hamrick & Murawski 1990; Hamrick et al. 1992). In contrast with other plants, genetic variation partitions within rather than among tree populations (Hamrick et al. 1992). Trees often carry a heavy genetic load of deleterious recessive alleles (Boshier 2000). Lande (1995) and Lynch et al. (1995) mention that populations with small effective sizes have reduced ability of genome purging, i.e. the removal of slightly deleterious alleles by natural selection, which could lead to their extinction through ‘mutational meltdown’. Byers & Waller (1999) conclude that purging is an inconsistent force within plant populations. Young & Boyle (2000) mention that to assess short-term population viability, examination of the interactions between inbreeding, heterozygosity, fitness, and demography is needed. Young & Boyle (2000) also indicate the potential danger of outbreeding depression where fragmentation leads to breakdown of local populations, while geneflow between populations is maintained.

Kindt & Lengkeek (1999) described that for areas with increasing deforestation, conservation of tree species on-farm may offer the best option for some species. Where conditions in agroecosystems vary from those in forest ecosystems, natural selection may favour different gene complexes. Domestication may also select other genotypes than natural selection, e.g. trees with smaller competitive ability (shorter stems) but higher production (more fruit). Conservation-through-use will therefore probably not lead to equivalent genetic conservation as by in situ conservation. Where natural habitats disappear, however, this may be the only feasible species conservation strategy.

Kindt et al. (Chapter 4 - 9) have described methods of analysing species diversity to diversify agroecosystems, which could lead to larger stability and productivity of these systems (provided the ecological conditions for a positive link between diversity, and stability and productivity, are satisfied). Methods that focus on increasing the evenness in species abundance distributions could lead to larger population sizes of less frequent species – increasing the abundance of less frequent species would thus not only have an ecological-functioning, but also a genetic diversity (avoidance of inbreeding depression, conservation) advantage. Minimum population sizes for less frequent species versus tree abundances required by farmers further advocate for beta-diversity approaches (i.e. not to attempt to grow each species everywhere, but to segregate species compositions among villages and farms). These also conform to ecological considerations on the asymptotic relationship between diversity and species stability and productivity (Kindt et al. – Chapter 2, 4 & 5).


200

Table 12.1. Parameters for each tree used during randomizations Para-

meter Explanation Randomisation ‡

Aget Age of the individual in season s Ages+1=Ages+1; Ages+1=0 if Ages+1 > Agemax Agemax Maximum age of the individual Predetermined or Age0=INT(ran*(Agemax+1)) Agemat Maturity; age at which individual trees can flower or produce

cuttings Predetermined

Flow Chance of flowering (0-1) in each season Tree flowers if Flow = Ran and Ages = Agemat FFt and

MFt

Presence (1) or absence (0) of female (FF) or male flowers (MF) on the individual

Calculated for each tree with Ages=0 in season s

FFs/m and MFs/m

Number of other trees with female (FF) and male (MF) flowers in the same subpopulation (s) and in other subpopulations of the metapopulation (m)

Calculated for each tree with Ages=0 in season s

Pt Chance of pollination from the same tree Pollen from the same mother tree if Ran < Po Ps Pt MFt / (Ps (Pt MFt + MFs) + Pm MFm)

Ps Chance of pollination from the same subpopulation Pollen from the same subpopulation if Ran < Po (Ps (Pt MFt + MFs) / (Ps (Pt MFt + MFs) + Pm MFm)

Pm Chance of pollination from other subpopulations Pollen from the metapopulation if Ran < Po (Ps (Pt MFt + MFs) + Pm MFm) / (Ps (Pt MFt + MFs) + Pm MFm)

Po Chance of pollination from outside the metapopulation or mutation (0-1)

Pollen with new allele identity if Ran = Po (Ps (Pt MFt + MFs) + Pm MFm) / (Ps (Pt MFt + MFs) + Pm MFm)

Ws Chance that the wildling originates from the same subpopulation Wildling from the same subpopulation if Ran < Wo Ws (FFt + FFs) / (Ws (FFt + FFs) + FFm)

Wo Chance that the wildling originates from outside the metapopulation or chance of mutation (0-1)

Wildling from the metapopulation if Ran < Wo (Ws (FFt + FFs) + FFm) / (Ws (FFt + FFs) + FFm)

FM Chance that a new tree is female, if the species is dioecious Tree is female if Ran > 1 / (1 + FM) ‡ Ran: random number ∈[0,1[ that was calculated for each comparison, except for choice of pollen and wildling origins

Table 12.2. Actual and farmer-preferred (Pref.) forms and origins of germplasm expressed in percentages of the total number of trees. Some farmers did not have preference for particular forms or origins of particular species (No. pref.)

Form Origin Own farm Neighbours Outside village - Total

Actual Pref. Actual Pref. Actual Pref. No pref. Actual Pref. Cutting 7.3 34.3 24.0 4.7 7.1 0.1 0.2 38.4 39.3 Seedling 2.5 7.8 5.9 0.8 15.5 4.6 0.6 23.9 13.7 Wildling 19.1 16.7 0.7 0.2 0.1 0.0 0.0 19.8 17.0 Seedling or wildling 0.0 2.1 0.0 0.0 0.0 0.0 0.0 0.0 2.1 Seed 10.4 24.9 6.0 1.8 1.1 0.9 0.0 17.5 27.6 Fruit 0.1 0.2 0.1 0.1 0.2 0.0 0.0 0.4 0.3 Total 39.3 86.0 36.7 7.6 23.9 5.6 0.8 100 100

CHAPTER 12

201

Table 12.3. Results for random mating and eleven scenarios of non-random mating for metapopulations of five subpopulations of size 10 (100,000 permutations). Average values are provided, with the range of minimum-maximum values in between brackets. (Parameters: see text)

Scena-rio

Parameters different from random mating

Hom (%) Simpson (%)

Ne 2H0 2-H2 (%)

2-H8 (%)

E8

0 None 39.3 (0-100)

39.9 (13-100)

50.3 (1-178)

4.0 (1-10)

37.7 (13-100)

49.5 (18-100)

-0.99

1 Agemax=4, 5 generations of 10 trees with Age0 =0, 1, 2, 3 or 4

23.6 (0-100)

24.9 (10-100)

89.4 (1-238)

6.8 (1-14)

23.8 (10-100)

35.2 (13-100)

-1.27

2 Flow=0.5 64.2 (8-100)

64.6 (18-100)

24.8 (1-127)

2.1 (1-6)

61.1 (18-100)

70.2 (24-100)

-0.57

3 FM=4 55.8 (2-100)

56.9 (9-100)

30.5 (1-279)

2.6 (1-34)

53.7 (9-100)

64.0 (16-100)

-0.72

4 Pt=20 52.7 (12-100)

44.4 (15-100)

43.6 (1-159)

3.5 (1-9)

41.9 (15-100)

53.4 (19-100)

-0.91

5 Pt=Ps=20 74.1 (30-100)

52.4 (16-100)

34.6 (1-143)

2.9 (1-8)

49.4 (16-100)

60.2 (18-100)

-0.78

6 Pt=Ps=Ws=20 74.2 (30-100)

49.5 (15-100)

37.6 (1-153)

3.1 (1-9)

46.6 (15-100)

57.7 (19-100)

-0.82

7 Pm=0 43 (2-100)

41.1 (14-100)

48.4 (1-171)

3.9 (1-10)

38.8 (14-100)

50.5 (17-100)

-0.96

8 Pt=0 38.4 (0-100)

39.6 (13-100)

50.8 (1-181)

4.0 (1-9)

37.4 (13-100)

49.2 (16-100)

-0.99

9 Seedlings all from the same subpopulation

38.5 (2-100)

39.2 (14-100)

51.6 (1-171)

4.1 (1-10)

37 (14-100)

48.7 (17-100)

-0.99

10 Po=0.99 31.4 (0-100)

32.1 (10-100)

66.1 (1-244)

6.7 (1-18)

30.1 (10-100)

42.9 (13-100)

-1.53

11 50% wildlings and 50% cuttings

38.5 (2-100)

39.7 (13-100)

50.7 (1-180)

4.0 (1-10)

37.5 (13-100)

49.2 (17-100)

-0.98

12 Combination 1+2+6+9 78.3 (40-100)

34.3 (11-100)

60.9 (1-223)

5.4 (1-14)

32.4 (11-100)

44.5 (16-100)

-1.26


202

0 10 20 30 40 50 60 70 80 90 100 110 120 130140 150 160 170

Species rank

1

10

100

1,000

10,000

100,000

Ab

un

dan

ce0 10 20 30 40 50 60 70 80 90 100

Percentage of species

0

10

20

30

40

50

Ab

un

dan

ce

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Den

sity

(tre

es/h

a)

abundance (left axis)abundance (right axis)density

Figure 12.1. Rank-abundance and rank-density curves for the 175 species encountered on farms.

0 1 2 3 4 5Log10(abundance)

0

1

2

3

4

log

10 (n

orm

aliz

ed d

ensi

ty+1

)within farmoutside farm and within villageoutside village

1 10 100 1,000 10,000 100,000

Abundance

1

10

100

1,000

10,000

No

rmal

ized

den

sity

+1

Figure 12.2. Normalized densities of for all species for distance categories of the same farm, the same village, and other villages. Normalized densities larger than 1 indicate aggregation within the distance category. The size of the symbol corresponds to the

number of species with the same abundance and normalized density.

CHAPTER 12

203

-3 -2 -1 0 1 2 3

Log10(density)

-3

-2

-1

0

1

2

Lo

g10

(P/(

1-P

))

0.001 0.010 0.100 1.000 10.000 100.000 1,000.000

Density (trees/ha)

0.001

0.010

0.100

1.000

10.000

100.000P

/(1-P

)

Figure 12.3. Density-abundance relationships for all species. The size of the symbol corresponds to the number of species with

the same density and abundance. Robust linear regression (S-PLUS 2000) (full line, dotted lines 95% confidence intervals) explained 95% of variation.

0123456789101112

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

H

1

2

4

8

2H

Figure 12.4. Rényi diversity profiles (H) for Wright-Fisher populations of size 50 (full lines), and various scenarios of non-

random mating (dotted lines) calculated for various scale parameter values (alpha, inf.: 8). The details of the various scenarios are detailed in table 12.3.

1 10 100 1,000 10,000 100,0000.0

0.1

0.2

0.3dominated by seeds or seedlingsdominated by wildlingsdominated by cuttingsno dominant form

N>9: 45.1

N<10: 1.7

a

1 10 100 1,000 10,000 100,0000.0

0.1

0.2

0.3dominated by seeds, seedlings or wildlingsdominated by cuttingsno dominant form

N>9: 26.8

N<10: 0.5

b

1 10 100 1,000 10,000 100,0000.0

0.1

0.2


N>9: 53.7

N<10: 2.7

c

1 10 100 1,000 10,000 100,0000.0

0.1

0.2

0.3

N<10: 5.5

N>9: 31.7

d

Figure 12.5. The average proportion of homozygous trees (vertical axis) encountered within each village during simulations of 50 seasons (1000 randomizations for species with abundance < 1000, 100 for other species) plotted against species abundance (horizontal axis). Lines are Friedman’s variable span smoother (S-PLUS) for species dominated by various forms of germplasm, except the dotted-dashed reference line, which corresponds to the reciprocal of abundance. The arrows indicate the percentage of metapopulations of size < 10 and = 10 above the reference line. a: scenario that uses the

preferred forms and origins for every species and farm, b: scenario where germplasm is collected from the old tree, c: scenario where germplasm is collected from the same farm, d: scenario where germplasm is collected at random within the village.

1 10 100 1,000 10,000 100,0000.0

0.1

0.2

0.3

0.4

0.5dominated by seeds or seedlingsdominated by wildlingsdominated by cuttingsno dominant form

N>9: 86.6

N<10: 17.1

a

1 10 100 1,000 10,000 100,0000.0

0.1

0.2

0.3

0.4


N>9: 58.5

N<10: 18.0

b

1 10 100 1,000 10,000 100,0000.0

0.1

0.2

0.3

0.4

0.5dominated by seeds, seedlings or wildlingsdominated by cuttingsno dominant formN<10: 21.3

N>9: 90.2

c

1 10 100 1,000 10,000 100,0000.0

0.1

0.2

0.3

0.4

0.5

N<10: 31.1

N>9: 98.2

d

Figure 12.6. The chance of selecting (with replacement) the same allele twice from the metapopulation (horizontal axis) calculated from simulations of 50 seasons (1000 randomizations for species with abundance < 1000, 100 for other species) plotted against species abundance (horizontal axis). Lines, arrows, and a-d as in figure 12.5.

1 2 3 4 5 6 7 8 9 10 11 12 13 14

1

2

3

4

5

6

7

8

9

10

11

122 4 8 16 32 64 128 256 512 1024 2048 4096 8192 16384

2

4

8

16

32

64

128

256

512

1024

2048

4096

dominated by seeds or seedlingsdominatd by wildlingsdominated by cuttingsno dominant type

a

dominated by seed, seedlings or wildllingsdominated by cuttingsno dominant form

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

1

2

3

4

5

6

7

8

9

10

11

12

132 4 8 16 32 64 128 256 512 1024 2048 4096 8192 1638432768

2

4

8

16

32

64

128

256

512

1024

2048

4096

8192

b

dominated by seed, seedlings or wildllingsdominated by cuttingsno dominant form

1 2 3 4 5 6 7 8 9 10 11 12 13 14

1

2

3

4

5

6

7

8

9

10

11

122 4 8 16 32 64 128 256 512 1024 2048 4096 8192 16384

2

4

8

16

32

64

128

256

512

1024

2048

4096

c

1 2 3 4 5 6 7 8 9 10 11 12 13 14

1

2

3

4

5

6

7

8

9

10

11

122 4 8 16 32 64 128 256 512 1024 2048 4096 8192 16384

2

4

8

16

32

64

128

256

512

1024

2048

4096

d

Figure 12.7. The number of alleles (lower horizontal axis: log2 scale) and reciprocal of the proportion of the dominant allele (left horizontal axis: log2 scale) within the metapopulation calculated from simulations of 50 seasons (1000 randomizations for species with abundance < 1000, 100 for other species). The size of the symbols corresponds to the abundance of the species. a-d as in figure 12.5.

Page not included!

CHAPTER 13 COMPARING SPECIES RICHNESS AND EVENNESS CONTRIBUTIONS TO ON-FARM TREE DIVERSITY FOR DATASETS WITH VARYING SAMPLE SIZES FROM KENYA, UGANDA, CAMEROON, AND NIGERIA WITH

RANDOMIZED DIVERSITY PROFILES R KINDT, A DEGRANDE, L TURYOMURUGYENDO, C MBOSSO, P VAN DAMME & AJ SIMONS PAPER PRESENTED AT IUFRO CONFERENCE ON FOREST BIOMETRY, MODELLING AND INFORMATION SCIENCE, 26 - 29 JUNE, UNIVERSITY OF GREENWICH.

(URL: HTTP://CMS1.GRE.AC.UK/CONFERENCES/IUFRO/PROCEEDINGS/)

Most often in natural and agricultural systems, species counts are provided as the measure of diversity. Continuing this logic, diversification equates to adding more species. Species diversity, however, is both a function of the number of species, and the evenness in distribution of abundances of species. Furthermore, diversity is often a function of sample size. In this article, we propose an approach and software that enables to distinguish between the contributions of richness, evenness, and sample size to diversity. The method is based on the Rényi diversity profile that can be dissociated into contributions of species richness, the evenness in proportion of the dominant species and the evenness in the proportions of the other species. The method we developed further included calculating the Rényi diversity profiles associated with random site sequences (100,000 in the study we present here) for a fixed total number of sites, fixed total area sampled or fixed number of individuals. Subsequently, the average diversity profile, associated 95% confidence intervals, minimum and maximum values, and the probabilities that profile values are smaller or larger than predetermined values can be calculated. This information thus enables comparisons of richness and evenness for systems for which species presence and abundance information was obtained with different sample sizes. This approach was tested with data sets documenting the complete tree species composition of 201 farms from western Kenya, 80 farms from western Uganda, and 39 farms from Cameroon. Interpolated comparisons based on the same number of farms, trees, or area indicated Cameroon to be the location with most evenly distributed species. Average species richness was highest in Kenya on a per area basis, yet lowest and intermediate on a per farm basis and per tree abundance basis, respectively. Implications for sampling in diversity surveys and tree diversity promotion activities are discussed.

COMPARING DIVERSITY FROM UNEQUAL SAMPLES

208

13.1. Introduction

Most often in natural and agricultural systems, species counts are provided as the measure of diversity. Continuing this logic, diversification equates to adding more species. Species diversity, however, is both a function of the number of species, and the evenness in distribution of abundances of species (Magurran 1988; Purvis & Hector 2000). Furthermore, diversity is often a function of sample size. In this article, we propose an approach and software that enables to distinguish between the contributions of richness, evenness, and sample size to diversity.

The method is demonstrated with data sets documenting on-farm tree species diversity in study sites from Cameroon, Kenya, Nigeria, and Uganda. In these and other tropical agroecosystems, the International Centre for Research in Agroforestry (ICRAF), as part of its strategy for integrated natural resource management, is working together with its partner organisations to ensure that resource-poor farmers and the global community continue to benefit from agroforestry tree species (ICRAF 2000; Weber et al. in prep.).

One of the purposes of agroforestry tree domestication – through one focus on landscape diversity and the other on genetic diversity within priority species – is the enhancement of stability and productivity of agro-ecosystems by diversifying on-farm tree species composition. Information on landscape-level diversity as provided by the methods presented here will assist in targeting tree domestication activities, and will provide benchmark information so that the impact of these interventions on diversity can be measured.

Species providing fruit were selected as priority species for domestication in Cameroon and Nigeria (Franzel et al. 1996), whereas special attention is also given to these species in eastern Africa since they provide cash income to farmers (“high-value trees”). Parallel investigations therefore focused on the exclusive diversity of fruit trees.


Diversity profiles (4.2.4), accumulation patterns of diversity profiles (4.2.5)

13.3. Results

13.3.1. Calculation of Average Values and Determination of 95% Intervals From 100,000 Random Site Sequences

Figures 13.1 – 13.2 show the values that were obtained at various scales for the diversity and evenness profile values calculated for Kenya at Fref = 39 based on 100,000 random sequences.

Random sequences yielded diversity and evenness profile values that most frequently fell within the 0.01 interval that corresponded to the average value. Histogrammes were symmetrical compared to the average value. The width of histogrammes was positively correlated with the scale parameter value.

The lower and upper confidence interval values for H0 and ln(E8 ) reported in tables 13.2 and 13.3 confirm the symmetrical nature of the interval, and their increasing width with scale parameter value. Since Aref and Nref could not be exactly matched within each farm sequence, intervals were wider. However, the average abundance and area for Aref and Nref -based subsampling were very close to the reference values.

CHAPTER 13

209

In four series of random farm sequences (U1, U3, U4 for Nref, and U3 for Aref, table 13.2), cases occurred where the algorithm selected no farms as this choice provided the closest match to the reference sample size. In these cases, no profile could be calculated, while the average profile value was based on the valid sequences.

Table 13.4 shows the subsample sizes (Fref) that corresponded to the maximum and minimum values of H0 and ln(E8 ) for comparisons between countries on equal Aref and Nref (maximum and minimum values were not included in this table, but can be inferred from figures 13.2-13.3). The table gives an indication of strong ß-diversity (i.e. differences in species composition among farms) influences on the diversity of the subsample. The number of farms associated with the highest H0 (species richness) was substantially larger than the number of farms associated with the lowest H0. This pattern indicates that a subsample of similar area or abundance, but constructed from more farms, contains more species. Results for ln(E8 ) did not yield a consistent pattern. In some cases, the dominant species was less evenly distributed on a smaller number of farms, while in other cases the reverse pattern could be observed. Uganda (Nref) and Cameroon (Aref and Nref) are the only cases where the dominant species was less evenly distributed on a larger number of farms.

13.3.2. Comparisons between Countries of Overall Diversity

In the following sections, we demonstrate some of the conclusions that can be drawn from average diversity and evenness profiles and their associated confidence intervals on Fref, Aref and Nref. In some cases, the large number of profiles presented, and the large overlaps between confidence intervals, did not result in clear figures, although their interpretation – that diversity was similar in several surveys – is straightforward. Figures 13.3 - 13.6 demonstrate, probably too extensively, that our method allows for immediate graphical interpretation of differences in diversity and evenness by investigating whether profiles of survey x were above, similar or below profiles of survey y.

Figure 13.3 and table 13.2 provide a comparison in diversity and evenness between Cameroon, Kenya, and Uganda.

Comparisons based on Fref showed Cameroon to be richest (largest H0 and S), and Kenya (after comparing it with Uganda only, dotted lines in figure 13.3a) to be least rich. Comparing richness based on Nref (figure 13.3b) showed Cameroon to be richest, and Uganda to be lowest in richness. On Aref (figure 13.3c), Kenya was richest, while confidence intervals for Cameroon and Uganda overlapped.

When comparing the values of H8 (figure 13.3a-c), which indicates the proportion of the dominant species, Cameroon had the largest values for Fref and Aref, while values for the three countries overlapped for Nref. In direct comparisons of Kenya and Uganda, Kenya had larger H8 values. For comparisons of ln(E8 ) (figure 13.3d-f, table 13.2), the evenness in the frequency of the dominant species, Kenya and Uganda had lower values than Cameroon for Aref, Uganda had lower values than Cameroon for Fref, while other cases had overlapping intervals. In direct comparisons between Kenya and Uganda, Uganda had the lowest values of ln(E8 ) on Fref and Nref.

For some scale parameter values, there was no overlap of the confidence intervals of Kenya and Uganda with the profiles of Cameroon. The pattern indicated that the other species (not the dominant species) in Cameroon were more evenly distributed. The more even distribution of other species in Cameroon is also indicated by the smaller ln(E1)/ln(E8 ) ratio (table 13.2). On Fref and Nref, the evenness profile of Kenya was above the profile for Uganda, indicating a more even distribution of species in the first country. The average evenness profiles for Kenya and Uganda were intersecting for Aref, indicating that the dominant species was more evenly distributed in


210

Kenya, while the other species were less evenly distributed in this country – however, the confidence intervals for the evenness profiles overlapped.

13.3.3 Comparisons between Villages of Overall Diversity

Figure 13.4 and table 13.2 provide a comparison between the diversity and evenness between villages in Cameroon, Kenya, and Uganda. The figure shows that many profiles are intersecting, and that many confidence intervals are overlapping. It is possible that larger sample sizes would have revealed clearer differentiation among villages, which stresses the importance of ensuring that sample sizes are large enough.

The general pattern that can be observed across the three countries, however, is that the dominant species was more evenly distributed (ln(E8 ) larger) in villages closer to forests, and that the dominant species was less evenly distributed in villages most distant to forests. This analysis was confirmed by comparisons of villages within the same country only, using the village with the smallest sample size within each country as reference, for which we did not include data in this chapter. However, the pattern of the gradient towards forests could not be observed for each village for each reference basis. For instance, the village at intermediate distance to the forest in Uganda had the most evenly distributed dominant species on Fref, although the two villages close to the forest had larger values than the two villages farthest away from the forest. For the Ugandan villages, however, the trend corresponding to distance from the forest could be observed for the evenness of the other species (ln(E1)/ln(E8 ) in table 13.2), as the village at intermediate distance had values lower than those corresponding to the villages close to the forest, and larger than those corresponding to the villages farthest away.

13.3.4. Comparison between Countries of Fruit Tree Diversity

Figure 13.5 and table 13.3 provide a comparison between the diversity and evenness of fruit trees between Cameroon, Kenya, and Uganda.

Comparisons based on Fref showed Cameroon to be richest and Kenya to be the least rich. For Nref, Kenya was less rich than Nigeria, while the confidence interval for Cameroon overlapped with the H0 value for Nigeria. Nigeria was richest on Aref. In direct comparisons between Cameroon and Kenya, the average richness of Cameroon was higher than the richness in Kenya, but the confidence intervals overlapped for the area reference.

For H8 , Nigeria had the largest, and Kenya the lowest values at each of the three reference bases. This pattern could also be observed for ln(E8 ) for Nigeria, except for an overlap with the confidence interval for Cameroon for Aref (figure 13.5c, table 13.3). Values of ln(E1)/ln(E8 ) (table 13.3) show that the other species are most evenly distributed in Nigeria and least evenly distributed in Kenya, regardless of the reference basis used.

13.3.5 Comparisons between Villages of Fruit Tree Diversity

Figure 13.6 and table 13.3 provide a comparison between fruit tree diversity and evenness between villages in Cameroon, Kenya, and Nigeria. As for figure 13.2, the small subsample sizes resulted in many overlapping confidence intervals.

The trend for all trees of increasing evenness with increasing proximity to forests could not be confirmed for fruit trees. For Kenya, the village farthest away from the forest had the most evenly distributed dominant species. In Cameroon, the intermediate village had the highest ln(E8 ). In Nigeria, the village furthest away from the forest had a more evenly distributed species than the village at intermediate distance.

CHAPTER 13

211

13.4. Discussion

13.4.1. The Method

As diversity profiles and evenness profiles generally show a sample size-dependent accumulation pattern (influenced by the underlying species abundance distribution and species frequency distribution among sites, Kindt et al. - Chapter 7), comparisons of diversity between various studies are only meaningful if based on the same sample size to neutralize the influence of sample size on diversity (Magurran 1988; Rennolls & Laumonier 1999; Gotelli & Colwell 2001). Moreover, the influence of sample size depends on the scale of the Rényi diversity profile (Kindt et al. - Chapter 7).

Since multilocational surveys typically collect information from a different total area and/or a different total number of individuals (i.e. as average richness per area will be different for each location), both Aref and Nref can be used as a reference basis to make comparisons between surveys or to make recommendations for sampling effort that would result in significant differentiation among locations (Gotelli & Colwell 2001; Moreno & Halffter 2001). A third meaningful reference basis for our analysis was Fref, as average area and abundance of farms varied among the surveys that we compared. Typically, biodiversity surveys use sample units of fixed area or abundance. We sampled entire farms of varying size, since we were interested in exploring possibilities for diversification of these farms, while farms were the logical sample units for participatory research. (Calculating accumulation curves based on numbers of accumulated sites, rather than accumulated area sampled, also allows for exact calculations of the average expected species richness (Kindt et al. - Chapter 6).)

We demonstrated that the randomization procedure of 100,000 random site sequences resulted in very similar average Nref and Aref based subsample sizes allowing for meaningful comparisons among datasets. Obviously, comparisons based on the same number of sites always used the same sample size. As histogrammes were symmetrical among the average, the calculated average value corresponds to the value that is most frequently obtained, while confidence interval limits (although we calculated both confidence limits independently) will be symmetrical. The logarithm in the formula for the Rényi diversity index thus resulted in a more symmetrical range of profile values. Since species richness for a particular randomization can only result in discrete values, we also calculated species richness directly, before calculating histogrammes. The results (figures 13.1 & 13.2) indicate that a very similar histogramme (and confidence interval limits) is obtained than by constructing the histogramme based on logarithmic species richness.

We observed increasing ranges of diversity and evenness profile values associated with increasing scale. Similar observations on the variation of the Shannon (scale 1) and Simpson (scale 2) diversity indices resulted in recommendations to use the Shannon diversity index for comparisons among sites, as this index would be more sample-size efficient in yielding significant differences (Magnussen & Boyle 1995; Gimaret-Carpentier et al. 1998). However, where the evenness in the species abundance distribution differs among sites (a factor eliminated by choosing a species abundance distribution model as in Magnussen & Boyle 1995), comparisons based on the Simpson index may be more efficient (Kindt et al. - Chapter 7). For example, for a comparison between two surveys where the diversity and evenness profile become progressively steeper with sample size in one compared to the other survey, a significant difference may be detected sooner at scale 2 than at scale 1. Especially where species abundance distributions are not known a priori, we recommend analyses based on accumulated Rényi diversity profiles, also because calculation of several profile values does not pose major computational difficulties (moreover, the programs used for the calculations in this paper may be obtained for free from the authors).

Calculation of Rényi diversity and evenness profiles allows separating the influence of species richness and the evenness on diversity. Our methodology of calculating average profiles,


212

therefore, allows for separation of the effects of three factors that influence diversity: richness, evenness, and sample size.

13.4.2. Applying the Method

As our focus in this paper was on the methodology of comparing species richness and evenness among studies with varying sample sizes, we will only briefly discuss the application of the methodology to the design of interventions seeking to increase agrobiodiversity through management of interspecific tree diversity on farms – although this was the purpose of the surveys.

Domestication research conducted by ICRAF and its partners seeks to increase the stability and productivity of agroecosystems (ICRAF 2000; Weber et al. in prep.). Various conditions for a positive relationship have been documented, so that on average, and with an asymptotic relationship, larger diversity in species characteristics will result in larger stability of ecosystems, where trade-offs exist between characteristics of species and their performance under subsets of existing environmental conditions, and where environmental characteristics have spatio-temporal variation (Kindt et al. - Chapters 2, 5 & 6). In this context, increasing the diversity of species’ groups of lower diversity will resort in larger effects on ecosystem functioning, than similar increases of groups of larger diversity. The logarithmic scale in the calculation of the Rényi diversity and evenness profiles conforms to the decreasing importance of diversification at larger initial diversity levels, what justifies our choice of this particular diversity series for our explorations.

We used a similar approach as for the determination of functional ecological groups, but determined species groups based on the role (uses) each species had within the farm, as documented through ethnobotanical surveys that accompanied the species inventories. Our method allows ranking different use-groups (such as fruit discussed in this paper), so that interventions could focus on groups of lower diversity, lower richness and/or lower evenness. Targeting diversification efforts on species-groups of lower diversity is meaningful, since we expect, on average, larger effects on stability and productivity within these groups.

It should be obvious, however, that the choice of a use-group to target should not be only determined by the diversity of this group, and that priorities of farmers should be included as well (Kindt et al. - Chapter 5). For instance, increasing diversity may not be the preferred management option for every farmer.

Once a use-group has been selected for diversification, even if it is not the least diverse, our method still allows to make a choice on the balance between targeting richness (i.e. adding new species) or targeting evenness (i.e. making species present more evenly distributed). In addition, the method can be used to measure diversity before and after interventions – for instance where surveys before and after interventions do not sample the same number of farms. The method could, for instance, reveal that interventions that focused on introduction of more trees resulted in increases of diversity on Aref, but decreases on Nref.

Some indication of beta-diversity influences was provided as subsamples of a similar area or abundance that consisted of a larger number of farms had larger richness. The approach that we documented, however, is an alpha-diversity-analysis approach, what means that we compare diversity among various systems, while ignoring differences in species compositions of these systems. The approach can be used to detect the most diverse system, to select the system with the largest richness, or to identify the system with the largest evenness. However, the combination of two less diverse systems may contain more species than the combination of two more diverse systems, if the less diverse systems share fewer species than the more diverse systems. A beta-diversity approach that considers differences in species composition (such as

CHAPTER 13

213

ordination or clustering methods based on ecological distance measures) could complement the alpha-diversity analysis. Beta-diversity analysis could, for instance, provide information on potential species that could be moved from one system to another system to increase the richness of the latter system.

Analysis of differences in diversity among areas allows prioritization of activities. However, an understanding of the underlying factors for differences among diversity may be crucial to select priority areas. For instance, we observed a relationship between proximity to forests on the evenness of the dominant species. However, as proximity to forest is correlated with other factors such as altitude, rainfall, and population density, we could not separate the influence of proximity to forests from these co-varying factors. (Correlation among explanatory variables poses general problems in regression models, Shaw & Mitchell-Olds 1993, Legendre & Legendre 1998 pp. 518-521. Statistical separation between effects requires the presence of various levels of one factor at particular levels of the other factor. The design of our survey did not address separate between various factors that could cause the relationship between distance to forest and on-farm biodiversity, but only investigated differences among locations.) We were able to document differences in diversity among Cameroonian, Kenyan, Nigerian, and Ugandan locations. Our results could be interpreted in terms of the global biodiversity conservation value of one location over the other. In this context, we want to stress the importance of studying differences in composition among locations, investigating genetic diversity and reproductive ecology of component species, and sampling species diversity in natural ecosystems adjacent to the agroecosystems that we sampled, to evaluate their conservation value. The methodology presented will probably prove more useful to design interventions within landscapes, rather than prioritizing among landscapes.


214

Table 13.1. Information on proximity to forests and sample sizes of villages Country Village name ‡ Village reference

for other tables Proximity to

forest Farms Abundance of all

trees Abundance of

fruit trees Area surveyed

(ha) Cameroon Chopfarm C1 Close 14 NA 635 30.49 Elig-Nkouma C2 Intermediate 20 3358 1632 53.14 Nko'ovos C3 Close 19 2924 1436 114.99 Makenene C4 Distant 19 NA 3310 100.02 Kenya Ebuchiebe K1 Distant 50 18542 1514 26.18 Shimutu K2 Close 51 25107 3108 47.65 Mutambi K3 Intermediate 50 34591 1400 21.46 Madidi K4 Intermediate 50 23898 3246 20.43 Nigeria Ilile N1 Intermediate 20 NA 960 13.44 Uguwaji N2 Distant 20 NA 888 16.10 Uganda Butoboore U1 Intermediate 16 24036 NA 25.03 Kashasha U2 Close 16 8533 NA 76.67 Kagarama U3 Distant 16 23385 NA 166.70 Nangara U4 Distant 16 20129 NA 88.58 Kitojo U5 Close 16 9358 NA 62.41 ‡ : In Uganda, farms were grouped in parishes, which are groups of villages – for simplicity in this chapter, we referrred to the “parishes” of this country as “villages”

CHAPTER 13

215

Table 13.2. Results for species richness (H0), evenness in the distribution of the dominant species (E8 ), and evenness in the distribution of other species (lnE1/lnE8 ) for comparisons between countries and between villages of overall species diversity for the same number of farms, same number of trees or same area sampled. LCL and UCL indicate the lower and upper 95% confidence interval limits calculated from 100,000 random farm sequences. NA indicates that the confidence interval was not calculated, as the respective country or village had the smallest sample size (the reference sample size). Country / Average H0 E8 lnE1 / Permu- Village or Parish

sample size ‡

Average LCL UCL Average LCL UCL lnE8 tations

Cameroon 39 4.77 NA NA -2.80 NA NA 0.42 100,000 12495 4.77 NA NA -2.80 NA NA 0.42 100,000 115.71 4.71 4.66 4.76 -2.75 -2.87 -2.62 0.41 100,000

C2 16 4.47 4.43 4.50 -2.56 -2.65 -2.42 0.38 100,000 2924.19 4.49 4.45 4.52 -2.57 -2.65 -2.49 0.39 100,000 20.44 4.33 4.20 4.44 -2.45 -2.69 -2.15 0.37 100,000

C3 16 4.29 4.24 4.32 -2.32 -2.44 -2.15 0.42 100,000 2924 4.32 NA NA -2.34 NA NA 0.42 100,000 20.46 3.86 3.61 4.07 -1.99 -2.35 -1.57 0.57 100,000

Kenya 39 4.60 4.45 4.75 -2.99 -3.42 -2.63 0.64 100,000 12495.46 4.43 4.21 4.62 -2.88 -3.38 -2.46 0.63 100,000 115.72 5.16 NA NA -3.39 NA NA 0.70 100,000

K1 16 4.03 3.85 4.18 -3.25 -3.55 -2.84 0.65 100,000 2923.98 3.75 3.49 3.96 -2.95 -3.33 -2.42 0.65 100,000 20.42 4.33 4.26 4.39 -3.57 -3.68 -3.42 0.66 100,000

K2 16 4.07 3.89 4.25 -2.29 -2.69 -1.97 0.64 100,000 2923.21 3.67 3.40 3.94 -2.20 -2.75 -1.70 0.59 100,000 20.46 4.18 4.02 4.34 -2.34 -2.63 -2.07 0.65 100,000

K3 16 4.13 3.93 4.31 -3.09 -3.58 -2.53 0.66 100,000 2933.82 3.57 2.94 3.99 -3.53 -3.22 -2.07 0.76 100,000 20.47 4.54 4.51 4.56 -3.54 -3.62 -3.38 0.67 100,000

K4 16 4.33 4.07 4.53 -2.71 -3.10 -2.32 0.64 100,000 2924.32 3.91 3.55 4.24 -2.45 -3.08 -1.93 0.61 100,000 20.43 4.71 NA NA -3.04 NA NA 1.02 100,000

Uganda 39 4.80 4.70 4.90 -3.73 -4.02 -3.38 0.66 100,000 12499.29 4.38 3.89 4.67 -3.42 -4.00 -2.71 0.66 100,000 115.70 4.62 4.34 4.81 -3.61 -4.04 -3.15 0.66 100,000

U1 16 4.32 NA NA -2.94 NA NA 0.63 100,000 2554.85 3.43 2.39 3.99 -2.56 -3.31 -1.59 0.63 93,754 20.45 4.23 4.09 4.31 -2.83 -3.10 -2.43 0.65 100,000

U2 16 4.52 NA NA -3.21 NA NA 0.46 100,000 2924.83 4.15 4.00 4.28 -2.84 -3.28 -2.41 0.46 100,000 20.42 4.03 3.68 4.28 -2.73 -3.28 -2.13 0.47 100,000

U3 16 4.43 NA NA -3.83 NA NA 0.76 100,000 2619.11 3.38 2.56 4.18 -2.69 -3.43 -1.78 0.67 93,715 17.12 3.41 2.39 4.18 -2.67 -3.45 -1.45 0.66 93,715

U4 16 4.14 NA NA -3.76 NA NA 0.75 100,000 2483.29 3.34 2.39 3.79 -2.63 -3.27 -1.55 0.69 93,836 20.50 3.51 2.89 3.88 -2.93 -3.57 -2.09 0.73 100,000

U5 16 4.44 NA NA -3.30 NA NA 0.52 100,000 2923.72 4.03 3.82 4.22 -2.86 -3.33 -2.31 0.52 100,000 20.41 4.04 3.82 4.21 -2.88 -3.34 -2.29 0.51 100,000

‡ upper row: number of farms, middle row: abundance, bottom row: area (ha)


216

Table 13.3. Results for species richness (H0), evenness in the distribution of the dominant species (E8 ), and evenness in the distribution of other species (lnE1/lnE8 ) for comparisons between countries and between villages of fruit species diversity for the same number of farms, same number of trees or same area sampled. LCL and UCL indicate the lower and upper 95% confidence interval limits calculated from 100,000 random farm sequences. NA indicates that the confidence interval was not calculated, as the respective country or village had the smallest sample size (the reference sample size). All 100,000 random farm sequences yielded valid results. Country / Average H0 E8 lnE1/lnE8 Village sample size ‡ Average LCL UCL Average LCL UCL Cameroon 40 3.38 3.25 3.50 -2.57 -2.74 -2.38 0.58

1847.98 3.20 2.94 3.37 -2.38 -2.67 -2.06 0.57 29.51 2.89 2.48 3.18 -2.05 -2.50 -1.54 0.55

C1 14 2.83 NA NA -2.00 NA NA 0.47 635 2.83 NA NA -2.00 NA NA 0.47 13.44 2.63 2.39 2.78 -1.76 -2.05 -1.44 0.44

C2 14 2.81 2.70 2.90 -1.60 -1.74 -1.47 0.61 635.11 2.65 2.39 2.84 -1.48 -1.68 -1.21 0.57 13.43 2.51 2.19 2.71 -1.36 -1.60 -1.09 0.84

C3 14 3.20 3.09 3.26 -2.03 -2.20 -1.84 0.58 635.24 3.06 2.83 3.22 -1.92 -2.22 -1.54 0.58 13.49 2.51 1.79 3.00 -1.49 -2.10 -0.98 0.52

C4 14 2.81 2.70 2.90 -2.32 -2.47 -2.15 0.60 636.12 2.46 2.07 2.71 -1.96 -2.28 -1.46 0.57 13.23 2.37 1.94 2.64 -1.87 -2.26 -1.26 0.56

Kenya 40 2.78 2.56 3.00 -2.44 -2.69 -2.14 0.70 1847.86 2.78 2.48 3.00 -2.45 -2.68 -2.19 0.70 29.55 2.85 2.63 3.05 -2.51 -2.74 -2.26 0.70

K1 14 2.32 2.07 2.49 -1.78 -2.07 -1.34 0.53 635.23 2.40 2.19 2.49 -1.88 -2.08 -1.64 0.55 13.46 2.43 2.30 2.49 -1.91 -2.09 -1.68 0.55

K2 14 2.43 2.07 2.78 -2.20 -2.57 -1.75 0.75 635.40 2.30 1.79 2.78 -2.08 -2.46 -1.60 0.75 13.44 2.44 2.07 2.78 -2.21 -2.57 -1.78 0.75

K3 14 2.42 2.07 2.64 -1.76 -2.23 -1.24 0.59 635.20 2.56 2.30 2.78 -1.89 -2.29 -1.42 0.62 13.43 2.66 2.48 2.78 -2.00 -2.30 -1.64 0.61

K4 14 2.49 2.19 2.78 -2.22 -2.52 -1.77 0.70 634.56 2.38 1.94 2.71 -2.12 -2.42 -1.74 0.70 13.46 2.73 2.56 2.84 -2.49 -2.63 -2.30 0.73

Nigeria 40 3.18 NA NA -1.83 NA NA 0.40 1848 3.18 NA NA -1.83 NA NA 0.40 29.54 3.18 NA NA -1.83 NA NA 0.40

N1 14 3.04 2.94 3.10 -1.35 -1.52 -1.18 0.36 635.13 3.04 2.94 3.10 -1.34 -1.52 -1.15 0.37 13.44 3.09 NA NA -1.40 NA NA 1.63

N2 14 2.88 2.70 2.95 -2.09 -2.19 -1.93 0.46 635.32 2.88 2.70 2.95 -2.09 -2.19 -1.93 0.46 13.43 2.91 2.77 2.94 -2.11 -2.19 -1.96 0.46

‡ upper row: number of farms, middle row: abundance, bottom row: area (ha)

CHAPTER 13

217

Table 13.4. Sample sizes (on area or abundance reference) and number of farms that corresponded to minimum and maximum values for species richness (H0) and for evenness in the distribution of the dominant species (E8 ). Measure Country Reference Parameter Minimum Sites Maximum Sites All species Cameroon Area (ha) H0 115.05 19 114.21 28 E8 114.04 23 115.22 26 Kenya Abundance H0 12054 14 12480 22 E8 12503 14 12512 22 Uganda Abundance H0 9597 2 12586 24 E8 11992 11 11202 2 Area (ha) H0 115.48 6 115.34 30 E8 114.13 11 102.34 26 Fruit species Cameroon Abundance H0 1847 13 1855 25 E8 1928 20 1749 11 Area (ha) H0 19.08 2 28.10 11 E8 30.50 9 23.06 3 Kenya Abundance H0 1809 24 1839 53 E8 1884 25 1837 49 Area (ha) H0 29.39 43 29.16 46 E8 29.38 43 28.99 55


218

0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 5500

Number of permutations

0

1

2

3

4

5H

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

EX

P(H

)

n=39

3

inf.10

Figure 13.1. The average diversity profile calculated for subsamples with all species of 39 Kenyan farms, and the range in values obtained in 100,000 random site sequences. The vertical axes show the obtained diversity profile values, the lower horizontal axis the number of permutations (random site sequences) with values ∈[H,H+0.01[ corresponding to the histogrammes (circles) (only

scales 0, 0.1, 0.25, 0.5, 0.75, 1, 2, 3, 10 and 8 were included for clarity), the upper horizontal axis the scale parameter values corresponding to the average diversity profile (squares) (inf.= 8). The histogramme with crosses (scale 0) was calculated based on

species richness (S), not on ln(S) as for H0.

0 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 2600027000

Number of permutations

-4

-3

-2

-1

0

ln(E

)

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0.018

0.030

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E

n=39

3

inf.

10

Figure 13.2. The average evenness profile calculated for subsamples of 39 Kenyan farms containing all species, and the range in

values obtained in 100,000 random site sequences. The vertical axes show the obtained evenness profile values, the lower horizontal axis the number of permutations (random site sequences) with values ∈[ln(E),ln(E)+0.01[ corresponding to the

histogrammes (circles) (only scales 0.1, 0.25, 0.5, 0.75, 1, 2, 3, 10 and 8 were included for clarity), and the upper horizontal axis the scale parameter values corresponding to the average diversity profile (squares).

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

6H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

244.7

403.4

EX

P(H

)

CameroonKenyaUganda

a 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

6

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

244.7

403.4

EX

P(H

)

CameroonKenyaUganda

b 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

6

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

244.7

403.4

EX

P(H

)

CameroonKenyaUganda

c

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-5

-4

-3

-2

-1

0

ln(E

)

0.007

0.011

0.018

0.030

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E

CameroonKenyaUganda

d 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-5

-4

-3

-2

-1

0

ln(E

)

0.007

0.011

0.018

0.030

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E

CameroonKenyaUganda

e 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-5

-4

-3

-2

-1

0

ln(E

)

0.007

0.011

0.018

0.030

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E

CameroonKenyaUganda

f

Figure 13.3. Comparisons between countries of diversity and evenness of all trees. Full lines: comparison of Cameroon, Kenya and Uganda, Dotted lines: comparisons of remaining two countries after removal of reference country of full lines (see table 13.1), a: comparisons of diversity based on the same number of farms, b: comparisons of diversity based on the same number of trees, c:

comparisons of diversity based on the same area sampled, d-f: respective comparisons of evenness. Boxes show 95% confidence intervals, while lines represent the range (minimum-maximum) obtained from 100,000 permutations. Some intervals and ranges are not represented at the actual scale parameter value to aid visual comparisons.

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

EX

P(H

)

CameroonKenyaUganda

a 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

EX

P(H

)

CameroonKenyaUganda

b 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

EX

P(H

)

CameroonKenyaUganda

c

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-4

-3

-2

-1

0

ln(E

)

0.018

0.030

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E

CameroonKenyaUganda

d 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-4

-3

-2

-1

0

ln(E

)

0.018

0.030

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E

CameroonKenyaUganda

e 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-4

-3

-2

-1

0

ln(E

)

0.018

0.030

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E

CameroonKenyaUganda

f

Figure 13.4. Comparisons between villages of diversity and evenness of all trees. Full lines: villages close to forests, dotted lines: villages distant from forests, dashed lines: villages at intermediate positions, a: comparisons of diversity based on the same number of farms, b: comparisons of diversity based on the same number of trees, c: comparisons of diversity based on the same area sampled, d-f: respective comparisons of evenness. Boxes show 95% confidence intervals, while lines represent the range (minimum-maximum) obtained from 100,000 permutations. Some intervals and ranges

are not represented at the actual scale parameter value to aid visual comparisons.

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

EX

P(H

)

CameroonKenyaNigeria

a 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

EX

P(H

)


b 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

EX

P(H

)


c

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-3

-2

-1

0

ln(E

)

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E


d 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-3

-2

-1

0

ln(E

)

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E


e 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-3

-2

-1

0

ln(E

)

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E


f

Figure 13.5. Comparisons between countries of diversity and evenness of fruit trees. Full lines: comparison of Cameroon, Kenya and Nigeria, Dotted lines: comparisons of remaining two countries after removal of reference country of full lines (see table 13.1), a: comparisons of diversity based on the same number of farms, b: comparisons of diversity based on the same number of trees, c: comparisons of diversity based on the same area sampled, d-f: respective comparisons of evenness. Boxes show 95% confidence intervals, while lines represent the range (minimum-maximum)

obtained from 100,000 permutations. Some intervals and ranges are not represented at the actual scale parameter value to aid visual comparisons.

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

EX

P(H

)


a 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

EX

P(H

)


b 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

EX

P(H

)


c

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-3

-2

-1

0

ln(E

)

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E


d 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-3

-2

-1

0

ln(E

)

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E


e 0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

-3

-2

-1

0

lnIE

)

0.050

0.082

0.135

0.223

0.368

0.607

1.000

E


f

Figure 13.6. Comparisons between villages of diversity and evenness of fruit trees. Full lines: villages close to forests, dotted lines: villages distant from forests, dashed lines: villages at intermediate positions, a: comparisons of diversity based on the same number of farms, b: comparisons of diversity based on the same number of trees, c: comparisons of diversity based on the same area sampled, d-f: respective comparisons of evenness. Boxes show 95% confidence intervals, while lines represent the range (minimum-maximum) obtained from 100,000 permutations. Some intervals and ranges

are not represented at the actual scale parameter value to aid visual comparisons.

CHAPTER 14 SPECIES DIVERSITY ON FARMS IN CAMEROONIAN, KENYAN AND UGANDAN LANDSCAPES R KINDT, JC WEBER, AG LENGKEEK, JM BOFFA, A DEGRANDE, JG JOURGET, D VAN OIJEN, C MBOSSO, P VAN DAMME & AJ SIMONS

IN PREPARATION FOR A BOOK ON ‘AGROFORESTRY AND BIODIVERSITY CONSERVATION IN TROPICAL LANDSCAPES’

Agroforestry is aimed at diversifying and sustaining agricultural landscapes for increased social, economic and environmental benefits. We calculated species accumulation curves (documenting the relationship between sample scale and species richness) and diversity profiles (decomposing diversity in richness and evenness) for all species and species belonging to the dominant on-farm niches and functions of trees on-farm in each of 4 landscapes of tropical Africa. Important differences could be observed between functions and niches, that allow to target diversification efforts to groupings of species of lower diversity. Ecological research demonstrated that diversification of a group of low diversity will have larger effect on ecosystem functioning than the same level of diversification of a group of higher diversity. Targeting diversification towards groups of low diversity could also be more relevant to limit risks of non-production. Species were not distributed at random over villages, which allows for increases in their richness by distributing species more randomly in the landscape. Distance to forest, wealth, farm size and family size were positively linked with species richness within a use-group in a minimum of 3 landscapes each. In western and central Kenya, >70% of trees were planted, whereas in Cameroon and central Uganda <50% of trees were planted. Exotic species had higher abundance than their species proportion. Population sizes of indigenous species were, therefore, small so that village-level genetic diversity management efforts may be necessary to ensure that tree populations are sustained.

TREE SPECIES DIVERSITY IN SEVERAL LANDSCAPES

224

14.1. Introduction

The species-area relationship of Arrhenius (1921) describes that when a larger area is sampled, more species will be encountered. We investigated scaling patterns of tree species richness within four African agroecosystems. In these managed ecosystems, species presence either indicates that trees were planted, or that trees that regenerated naturally were protected. Species-area scaling patterns imply that information on the average species richness on a farm poorly represents diversity at larger scales in the landscape. Since we focused on species diversity in agroforestry systems, we only measured on-farm diversity. In farmed sections of landscapes, however, we distinguished between different on-farm niches for trees provided by various types of land-use.

Although many equate diversity to species richness (the number of species encountered), diversity is a function of the number of species, and the evenness in distribution of species’ abundances (Magurran 1988 pp. 7-8; Purvis & Hector 2000). We used a method that discriminates between both facets of diversity, and investigated scaling patterns of each facet. The reason that we distinguished between both facets was not just to document diversity more accurately. Discriminating between both facets was also related to the purpose of our surveys of investigating the possibility of agroecosystem diversification.

One of the purposes of agroforestry tree domestication is to enhance the stability and productivity of agroecosystems by diversifying the composition of tree species (species richness and evenness of species abundances) on farms (ICRAF 2000). In connection to the focus on agroecosystem productivity, we grouped species according to their uses documented by ethnobotanical surveys that complemented the biodiversity surveys, and explored ways of diversifying composition within each use. Simons (in prep.) distinguished between possible interventions on the tree component of agroecosystems through domestication, including “replacement” of a tree by a tree of the same species, “substitution” of a tree by a tree of a different species, or “addition” of new trees of the same or new species. Besides documenting the distribution of tree diversity in the investigated landscapes, the information that we collected can steer interventions towards greater impacts on agroecosystem diversity. By discriminating between richness and evenness, interventions can be designed that tackle enhancement of each aspect separately.

By grouping species according to their uses on each farm, and by including characteristics of each farm as explanatory variables in analyses, we tested the hypothesis that diversity varied among use-groups and among farms. Farm characteristics that were included in the analyses for each landscape were the gender of the household head, the wealth ranking of the household and the proximity of the farm to forests. When the hypothesis is accepted, domestication efforts can focus on groups or types of farms with lower diversity or on groups with lower diversity of pre-determined focal types of farmers (for instance poorer or female-headed households).

Our results show how many indigenous species are present on-farms and ways in which landscapes can be diversified. Integrating indigenous tree species in agroecosystems is only one condition for their long-term conservation within these systems, however. The primary objective of our surveys was to explore options for diversification of agroecosystems – and therefore not to investigate the design of (ecological) approaches to conservation of indigenous or endemic tree species. It is obvious that further research on aspects of individual species such as their reproductive ecology, genetic diversity, and landscape-level metapopulation dynamics should be investigated to address their long-term conservation (e.g. Hanski 1999; Young et al. 2000; Palmer et al. 2001).

CHAPTER 14

225


Formulation of use-group and niche matrices (4.1), species accumulation curves (4.2.2), randomization tests on the influence of sample size on species richness (4.2.3), diversity profiles (4.2.4), accumulation patterns of diversity profiles (4.2.5), multiple, stepwise and partial regression analysis (4.4.3), analysis of species composition (4.5)

14.3. Results

14.3.1. Overall Species Diversity in the Landscapes

Only the Rényi profile accumulation surfaces for all species are presented below, while for trees arranged in niches or use-groups (see below), only species accumulation patterns and diversity profiles for the complete sample are provided.

Figure 14.1 shows the overall patterns between richness, evenness, and sample size in the four landscapes. Some of the statistics that are provided below were calculated directly from the data used to create figure 14.1 – the resolution of the figure does not allow for calculations that have the same precision. Figure 14.1 indicates that overall species richness (=exp(H0) for all farms) is relatively high in each landscape, and has the lowest value in Cameroon with 119 and the highest value in Meru with 294 species. When comparing for the same number of farms (n=35, which is the sample size for Meru), however, western Kenya becomes the landscape with the lowest accumulated richness of 96.5 species.

Species richness accumulates in a downwards convex manner in the ln(sample size)-ln(richness) space, so that the same difference in sample size at larger sample sizes corresponds to smaller increases in richness. The convex pattern implies that a two-parameter S=aNz model (N: number of farms) would not describe species accumulation accurately since such model would result in a straight line in the log-log space (i.e. logS=loga+zlogN). Figure 14.1 further indicates that landscapes mainly differ in alpha diversity (the average number of species of one farm), whereas the shapes of the species accumulation curves are relatively similar in the four sites. Average richness of a farm ranges from 16.6 in western Kenya to 53.2 in Meru.

The differences of the profile values at scale 0 and at scale 8 indicate that each landscape contains a dominant species. The dominant species has the largest frequency (indicated by the highest value of H8 ) in Meru and the lowest value in Mabira. The respective percentage of total abundance for the dominant species is 12.6% in Mabira and 32.0% in Meru. The other species (excluding the dominant species) are least evenly distributed in western Kenya since it has the largest ln(E1)/ln(E8 ). Distribution in evenness of other species, expressed in percentage decline of the profile from scale 0? 1, ranks landscapes from lowest to highest from western Kenya (70%), Meru (51%), Mabira (46%) to Cameroon (41%).

Figure 14.1 shows that an increasing a corresponds to decreased accumulation of Ha with sample size. This pattern results in diversity profiles that become steeper (an increasing H0-H8 ) with sample size. Especially for Meru, where accumulation of H8 is minimal, this indicates that the dominant species maintains approximately the same percentage of abundance. With increasing species richness with increased sample size, this indicates decreasing evenness in the distribution of the dominant species.

Figures 14.2-14.5 repeat some features of figure 14.1. Figures 14.2 and 14.4 also show the accumulation pattern of overall species richness, while figures 14.3 and 14.5 show the Rényi diversity profile for all farms and all species.


226

Table 14.1 provides some of the statistics that were mentioned above, the overall species richness and the average richness. Table 14.1 further shows the large number of botanical families (Cronquist 1981) to which these species belong, ranging from 42 in Cameroon to 64 in Meru. The large total species richness is mainly formed by indigenous species since only 12.6% (Cameroon) to 27.9% (Meru) of species are exotics. Exotic species however constitute a proportionally larger percentage of the total abundance, ranging from 29.9% (Cameroon) to 72.3% (western Kenya). The Kenyan sites differ from the others in the larger percentage of planted trees (71.1% and 80.2% versus 48.2% and 37.4%).

We consulted the 1997 IUCN Red List of Threatened Plants (Walter & Gillett 1998 at http://www.unep-wcmc.org) to check the IUCN threat category of the species that we encountered. Of the 237 species included in the list for Kenya, one rare species (Euphorbia friesiorum) and two vulnerable species (Milletia tanaensis and Vitex keniensis) were encountered in Meru. Vitex keniensis was also encountered in western Kenya. None of the 15 species listed for Uganda nor the 87 species listed for Cameroon were encountered.

Table 14.2 gives indications about the spatial distribution of species diversity and species composition. The fact that proximity-based richness is below the 95% interval for randomly-clustered richness means that it is unlikely that species are clustered at random within villages or axes. The larger values for the 95% interval indicate that random distribution results in higher average richness. Spatial clusters are therefore more similar in species composition, which results in their lower average richness. The RDA significance levels show that there is a significant difference in species composition among spatial clusters. This finding confirms that species are not distributed at random in the landscape. The low percentage of explained variation by spatial location indicates differences in species composition among farms within the same spatial cluster.

14.3.2. Species Diversity in Various Sections of the Landscape

Similar analyses as documented above for total species richness were performed for trees that occurred in various sections of the landscape, listed here as on-farm niches.

Similar to total richness of farms, species accumulation curves had a downward convex pattern (figure 14.2). For some curves, curvature was minimal in the log-log space, such as internal boundaries in Mabira, and crop boundaries in Meru. Richest were the cocoa gardens in Cameroon, cropland in Mabira, external boundaries in Meru and homesteads and cropland in western Kenya. Lower richness in some niches was linked to lower frequency of these niches – for example, homegardens only occurred on 53.8% of farms in Cameroon and woodlots only on 45.7% of farms in Meru. In general, niches with higher alpha diversity also had higher total species richness. However, some accumulation curves intersected in the various landscapes, with the exception of Mabira where none crossed. Examples of this phenomenon are homegardens and fallows in Cameroon, external boundaries, woodlots and fallows in western Kenya, and fallows and coffee fields in Meru. In these cases, groups with higher richness of the average farm had lower total species richness, which indicates strong sample-size influences on species richness.

Figure 14.3 demonstrates that species were not distributed evenly in any niche, since an even distribution would have been reflected by horizontal profiles corresponding to the niches. The many intersections of profiles indicate that many niches can not be ranked as more or less diverse – a niche that is richer often is less evenly distributed. The most evenly distributed niches are those with fewest species, such as homegardens in Cameroon, crop contours in western Kenya and crop boundaries in Meru. Coffee fields in Meru are most strongly dominated by a single species.

CHAPTER 14

227

Table 14.1 shows that some differences exist between niches in the percentage of exotic species they contain. Homegardens and homesteads (the homestead area is the space around houses that can be distinguished from other niches such as cropland or boundaries, but that is not necessarily dense enough to be a homegarden – we did not use a fixed criterion to discriminate between homegardens or homesteads, however) contain the largest percentage of exotic species in Cameroon, Mabira and Meru. In western Kenya, differences in percentages are not as explicit, although homesteads are also among the niches with larger proportions of exotic species. In Cameroon and Mabira, homegardens and homesteads also formed the niche with largest percentage of exotic and planted trees. In Meru, coffee gardens had the overall largest percentage of planted trees, although also in homegardens more than 60% of trees were planted. In western Kenya also, more than 60% than trees were planted in homesteads. In this landscape, however, the other niches, with the exception of fallows, had a larger percentage of planted trees.

Tests on the spatial distribution of species indicated that species distribution could be random in some niches, for example homegardens in Cameroon, fallows in Mabira, Meru, and western Kenya (table 14.2). The RDA significance indicated non-random differences in species composition among those niches where proximity-based richness was outside the 95% randomly grouped richness interval. The homestead area in Mabira formed an exception, however.

14.3.3. Species Diversity Distributed over Uses

Since we grouped trees in particular use-groups based on the ethnobotanical information collected on each farm, diversity of use-groups was not only the result of the distribution of species over farms, but also of the distribution of uses over species. When calculating diversity and accumulation patterns of a particular use, we only considered those trees on a farm for which the household indicated the particular use.

Many species have several uses, therefore the sum of species totals of use-groups exceeds the species total. Only the most frequent use-groups were analyzed. The total number of use-groups distinguished were 17 in Cameroon, 51 in Mabira, 60 in western Kenya, and 62 in Meru. In Cameroon, use categories were pre-defined. In the other surveys, use-groups that occurred only on one farm were the most frequent with respectively 17, 19, and 18 groups.

Figure 14.4 shows the accumulation of species within each use. Although, in general, the use-groups with larger alpha diversity had larger total species richness, more intersections among accumulation curves can be observed than for niches (see above). This is an indication of stronger differences in beta diversity of use-groups. The larger differences in species composition of farms of some groups (e.g. tools compared to fruit in Cameroon) result in higher richness at larger scale, whereas they had smaller richness at smaller scale. The fruit group had large alpha diversity and relatively low beta diversity in the four landscapes. Not all groups show the downwards convex pattern with increasing sample scale that was observed for total species richness and niches. Some groups instead show a downwards concave pattern at larger sample size – examples include vegetables in Cameroon, stakes in Mabira, beverage and fodder in western Kenya, and cash-generating trees in Meru.

Within the four landscapes, firewood was the use-group with largest average richness and total richness, which indicates that many species have firewood as a primary or secondary function. The difference between total species richness and species richness of firewood is largest in Meru (294 versus 121), an indication that in the other landscapes a larger percentage of species that are present is used for firewood. Use-groups with low total species richness indicate more specialized use categories. Specialized use categories, such as species that have very hairy leaves that are used to clean utensils (Ficus exasperata in Mabira), species with leaves that can be eaten as vegetables (Markhamia tomentosa, Piptadeniastrum africanum and Triplochiton scleroxylon in Cameroon), or species


228

used as beverage (Anacardium occidentale, Camellia sinensis, Coffea arabica and Theobroma cacao in western Kenya), had low species richness. The more general service functions shade, ornamental, boundary demarcation, and soil fertility improvement never had fewer than 10 species.

Figure 14.5 shows the diversity of the use-groups at the complete survey scale. The many intersections show the complex pattern where many use-groups that are richer, have a less even species distribution. It is therefore impossible to rank most use-groups in terms of diversity. Some use-groups can however be distinguished clearly with lower diversity: these groups are especially vegetables in Cameroon, beverage in western Kenya and cash in Meru.

Table 14.1 shows that groups of larger total species richness were rarely dominated by few botanical families, except for animal traps (15 species) in Meru which is composed of only two families. In some use-groups, nearly half or more of the species are exotic. These groups are fruit and ornamental in Mabira, cash, fruit (or nut), and ornamental in Meru, and fruit, construction, and beverage in western Kenya. In terms of abundance, many use-groups join the groups mentioned above – examples are fruit in Cameroon, stakes in Mabira and construction in Meru. In many use-groups, more than half of the trees have been planted. The general pattern that can be observed again is that groups with more planted trees contain more exotic species. However, exceptions exist, such as stimuli in Cameroon that contain no exotic species, but where 90.1% of trees are planted. Other examples are fodder in Mabira, and plant support in Meru.

The proximity-based richness was contained within the 95% interval for randomly-clustered richness for some groups, but not for others, indicating that only for some groups species were not distributed at random within villages or axes (table 14.2). Where proximity-based richness was lower than the lower interval limit, the RDA significance also indicated differences in species composition among spatial clusters. Mabira shows some particular cases – medicine, construction, shade, and ornamental – where the proximity-based richness is larger than the 95% interval. This is an indication of a more even distribution of species over the axes than a random distribution. In these cases, RDA permutations confirmed that there were no significant differences in species composition among the axes. In some cases, for example timber in Mabira or fruits or nuts in Meru, RDA explained a significant amount of variation of ecological distances among farms, whereas average species richness could result from random species distribution. This pattern probably reflects compositional differences in terms of proportional abundance rather than caused by presence or absence of particular species.

14.3.4. Multiple Regression Results

Table 14.1 indicates differences in average abundance between the various landscapes. Meru stands out with a much larger number of trees per farm. One contributing factor to the difference are the fact that coffee was included in the inventory, whereas cocoa was not counted in Cameroon (in this country, cocoa-gardens was one of the niches that was distinguished, but cocoa-trees within the niche were not inventoried). Another explanatory factor is the fact that farm sizes in Meru are larger than in western Kenya, whereas coffee and tea were not as frequent (these were listed under the beverage use-group). Both Kenyan sites had a larger number of boundary markers.

Regression results (table 14.3) show that use-group and farm characteristics explained 57 – 81% of variation in species richness, and 53 – 79% in abundance. The significance of the partial regression coefficients (which measure the rate of change in the response variable if all other explanatory variables remain constant, Legendre & Legendre 1998 p. 518) indicate that significant differences exist among the various use-groups for average richness (alpha diversity).

Although partial regression results indicated that the variation explained by farm characteristics alone was very low (< 5%), many household characteristics had associated regression coefficients

CHAPTER 14

229

that were significantly different from zero. Farm size, where this characteristic was measured, had a positive influence on diversity and abundance. Distance to forest had a positive influence on richness, with the exception of Meru, and a positive influence on abundance in Cameroon and Mabira. Wealth (housing type and number of cattle are wealth indicators in western Kenya) had positive influence on richness and abundance for all surveys – if we include the factor ‘off-farm employment’ as additional wealth indicator for Meru and use more flexible significance levels (P < 0.07). The size of the family and species richness were positively correlated in Mabira, western Kenya, and Cameroon (at more flexible significance levels for the latter survey).

Other characteristics of farms showed less consistent patterns across the four surveys. Male-headed farmers had more species in western Kenya, fewer species in Cameroon, while there was no significant influence in the other surveys. The age of the household head was only correlated with species richness in a significant way in western Kenya. Some other characteristics were only measured in one survey. In Cameroon, households that were indigenous to the village had more species on their farms. In Mabira, farmers that were using the forest had more trees and more species on their farms.

14.4. Discussion

14.4.1. Are Tree Populations on Farms Large Enough?

Our results indicate that a large number of tree species can be found on farms, of which most are indigenous. When considering tree abundance, many trees turn out to be exotic. In the Kenyan landscapes, although the percentage of indigenous species is larger than the percentage of exotic species, a larger percentage of trees are exotic. This pattern indicates that, although farmers are protecting and actively planting some indigenous trees on their farms, a larger proportion of exotic species are planted. Further research is required to determine if these differences reflect (a) the abundance of indigenous species desired by farmers, or (b) expectations by farmers that indigenous species will regenerate naturally although current levels of natural regeneration of indigenous species are below the desired abundance.

Farmers do not manage species – they manage individual trees or populations of trees. There are two numbers to consider in inventories for genetic resource management – the census number and the effective number of individuals (Baradat et al. 1995). There may be 50 individuals of a particular species on a particular farm in a census, but not all of these will equally contribute in terms of pollen, ovules and seeds to the next generation, due to differences in sexual maturity, flowering phenology, fecundity, pollinator foraging distance, and other reasons. If farmers plan to manage trees for sustainable production, then the effective number of trees on farm should be maintained at least 50 trees to ensure that most genetic diversity is maintained over time (O’Neill et al. 2001).

The fact that the abundance of many indigenous species was rather low – in Cameroon, Mabira, western Kenya and Meru, respectively 39%, 53%, 63% and 47% of indigenous species had fewer than 10 trees in the surveys – stresses the importance of evaluating effective population sizes of tree species. An assessment is necessary of the proportion of indigenous species that are currently composed of sink populations – (meta)populations of “living dead” that are not able to reproduce rapidly enough to avoid their extinction. Although most tropical species are self-incompatible (Boshier 2000), effective population sizes may be too small to avoid “mutational meltdown” preceding extinction (Lynch et al. 1995; Lande 1995).

Species that are grown in densities of lower than 1 tree ha -1 do not necessarily have high risk of genetic erosion – most canopy trees of tropical rain forests have densities lower than 1 tree ha -1


230

(Chase et al. 1996). Hamrick & Nason (2000) mention that pollen flow can be quite extensive (>25%) over distances of one kilometre. Young & Boyle (2000) indicate that pollen flow can be high in fragmented populations, provided vectors can pass non-forest habitat. Young & Merriam (1994) and White et al. (2002) showed that fragmentation could actually lead to an increase in pollination distances. Chase et al. (1996) found that isolated trees could act as stepping stones for geneflow among populations. Stacy et al. (1996) reported the combined effects of subpopulation size and species aggregation, so that plants in small clusters received more pollen from outside than plants occurring in larger clusters or in more even distributions. Young & Boyle (2000) indicated the potential danger of outbreeding depression where fragmentation had lead to breakdown of local populations while geneflow between populations is maintained, however. Overharvesting of trees reduces census numbers and may lead to lowering of genetic diversity and inbreeding depression (Murawski et al. 1994; Dayandandan et al. 1999).

Cain et al. (2000) pointed out that the patchy nature of many landscapes makes long distance seed dispersal of many spatially isolated species a necessary, although unusual, event. Reay & Norton (1999) and Galindo-González et al. (2000) report that dispersal processes were present in restoration sites and pastures, respectively. However, Hanski (1999) mentions that a significant amount of habitat is often unoccupied, indicating limitations in movement of many species. Dalling et al. (1998) and Hubbell et al. (1999) report that dispersal limitations in tropical forests result in their higher species richness. Cordeiro & Howe (2001) found that recruitment of animal-dispersed trees was three times smaller in small forest fragments (< 9 ha) than in larger fragments of three times their size, whereas recruitment of wind- and gravity dispersed trees was unaffected. As most tropical trees bear fruit adapted for animal dispersal, these authors, therefore, expect tree recruitment limitations for most species following forest fragmentation. Aldrich & Hamrick (1998) indicated further complexities as they found that 68% of seedlings in forest remnants originated from remnant adults in surrounding pastures, creating a genetic bottleneck.

Using a modified approach as used by Condit et al. (2000) showed that most species in western Kenya were aggregated within farms and within villages (Kindt et al. – Chapter 12). Whether current abundance and distribution of indigenous tree species within a matrix of farmland and natural ecosystems leads to effective population sizes of more or fewer than 50 trees is difficult to assess, however. Information on the reproductive ecology of many tropical tree species is very scant (Alvarez-Buylla et al. 1996; Boshier 2000). Metapopulation models that simulated spatially-realistic geneflow - which were expected to provide more accurate predictions of genetic diversity dynamics - suffered from lack of information on geneflow, current levels of genetic diversity and species distribution in other sections of the landscape (Kindt 2001c; Kindt et al. – Chapter 12.). More species-specific information is thus required to assess how many species are composed of sink populations only. For example, nine trees were recorded of the vulnerable Milletia tanaensis. We however do not know whether the trees that belonged to two separate villages were connected by geneflow to trees in other sections of the landscape. The two other species that were listed in the IUCN Red List, Euphorbia friesiorum and Vitex keniensis had respective Meru census abundance of 289 and 597 trees, but we have no information on potential earlier population bottlenecks of these populations.

Increasing the effective number of trees on farm for a large number of species is only possible in regions where farmers have adequate land holdings and/or tree planting density is high. In this case, farmers could maintain high inter- and intra-specific diversity on their farms. In western Kenya and other regions where farms are too small to accommodate an increase in the effective number of trees of many species, community-level management of tree breeding populations may be essential. For example, a particular farmer or group of neighbouring farmers could specialise in the management of trees of the same species, while other farmers could manage trees of other species. If neighbouring farmers manage trees of the same species, and the trees on neighbouring farms are able to mate with each other, then the total number of trees within the community may

CHAPTER 14

231

be sufficient to maintain the productive genetic diversity. In addition, farmers could co-ordinate community-based collections of a large number of selected trees and exchange the germplasm within and among communities to complement conservation-through-use (Kindt & Lengkeek 1999; O’Neill et al. 2001).

Local collection strategies, however, may not be appropriate for outcrossing species with very low tree density on farm: in this case, cross pollination may not be possible, so there would be no seed production. In this case, the reintroduction of high-quality germplasm from external seed sources would be recommended if farmers attach sufficient value to the species. Continuous interventions of injecting diverse germplasm in the agroecosystem necessary to conserve populations of indigenous species could be termed as narrow-sense “domestication of the landscape” in analogy with crop species, where narrow-sense domestication indicates the need of human interventions for survival in the agroecosystem (Kindt et al. – Chapter 5). An alternative reintroduction scheme would involve planting trees as pollination corridors among remnant populations to facilitate gene flow, and increase effective population sizes within the landscape. Clearly such management approaches would require a certain level of socio-economic integration and technical co-ordination within the community: farmers must come together, agree upon the preferred tree species for their community, and then ensure that there is a sufficient number of high-quality inter-mating trees at the community level.

14.4.2. Do African Farmers Want Tree Diversity?

The fact that farmers prefer certain species and only maintain other species in low abundance does not mean that they are unwilling to foster diversity. In western Kenya, farmers were requested to rank species by preference, and also asked which species they desired on their farms, using participatory methods including drawings of “ideal farms” (Kindt et al. – Chapter 11). In the survey, although exotic species often were preferred for particular use-groups (e.g. Eucalyptus spp. for construction and firewood, Persea americana for fruit), farmers expressed the desire to maintain a variety of indigenous species on their farms for these uses. However, some indigenous species were preferred for other uses (e.g. Sesbania sesban for soil fertility improvement, Warburgia ugandensis for medicine). Therefore, although many indigenous species regenerated naturally and were not highly preferred in western Kenya, farmers desired their presence. The various explanations that farmers provided for preferring diversity included statements of the advantage of complementary use characteristics, continuous supply, or exploitation of various niches. Examples of complementary characteristics were strong poles and flexible branches for construction, higher efficacy of medicines when used in mixtures, and fast versus more robust growth for boundary marking or timber. Continuous supply of products was often mentioned for fruit, firewood, and charcoal – groups that all relate to daily-needed food. Several farmers, when asked about the reasons of diversity within a particular use-group, also mentioned risk avoidance through diversification. Occasionally, the need to conserve species for children or grandchildren (although often only one individual tree was protected), or the human duty of providing fruit to local fauna were mentioned. Lengkeek et al. (in prep.) obtained similar indications that farmers opted for diversity. Experiments that introduced new species to farmers resulted in substitution of the dominant (and exotic) species Grevillea robusta by other species. Participatory research in Meru also indicated that they would prefer one individual of three species to three trees of a single species.

Figure 14.8 shows that there is a positive relationship between the number of species on a farm, and the number of uses on the farm in all the surveys. Such pattern suggest that farmers are interested in growing species for many purposes, and that additional species are grown to provide new purposes. Thus, farmers are not only willing to grow diversity within important functions of those trees on farm, but also among important functions of those trees. Figure 14.8 also


232

documents by the ranges in uses at a fixed level of species richness that some farms have low redundancy (few species per use, on average), while other farms have high redundancy. Farms with low redundancy could specifically be targeted in the design of domestication projects.

Whereas farmers wanted diversity mainly for differentiation among and within product or service characteristics, ecological surveys, experiments, and models have demonstrated that there is a conditional positive relationship between ecosystem diversity, and ecosystem stability and productivity (Kindt et al. – Chapter 2). This relationship is based on heterogeneity and complementarity in species’ and environmental characteristics. Since this relationship is asymptotic, diversification will result in effects that are more significant where initial diversity is lower (for example by increasing richness from 1 to 2 species versus 16 to 17 species, which involves addition of a single species in each case). Diversification as a risk-avoidance strategy in consideration of market dynamics or the likelihood of pest and disease occurrences, will also result in larger effects for systems with lesser diversity. Our method of using diversity profiles, therefore, not only documents the diversity of species grouped according to use or niche, but also allows prioritising groups for diversification.

14.4.3. Distribution of Species in Landscapes

We listed above groups with the lowest diversity, while other groups with low diversity profiles can be seen in the figures. The finding of the largest difference between total species richness and species richness for firewood which was observed for the survey with the largest total species richness (Meru) probably indicates that the asymptotic relationship between diversity and functionality of diversity is experienced by farmers as well. To potentially limit disease or pest threats (where threats are family- rather than species-specific; diversifying species does not necessarily reduce their threat where new species are also hosts, Schroth et al. 2000), analyses could be repeated where trees are categorized into botanical families. Groups with high species richness often contain many families. Selection of groups with low species richness thus also captures those groups of low family richness.

The intersections in diversity profiles show that some groups have relatively low richness, while other groups have lower evenness, and that richness and evenness are not necessarily positively correlated. Diversification strategies could therefore target the lower facet of diversity. For example, in a group with low richness but large evenness, a new species can be introduced while maintaining a similar abundance distribution. Likewise, in a group with low evenness and medium or high richness, some of the trees of the dominant species can be substituted with other species already belonging to the group. Alternatively, more trees can be added for the less abundant species. Species substitutions and additions conform to two of the four potential interventions of domestication projects (Simons in prep.), while investigations of diversity profiles as documented here show how such interventions can be planned. It is possible that such interventions would alter the proportions of planted versus protected seedlings and trees, especially if diversity would be enhanced through additions or replacements of indigenous species that are planted to a lower extent now. Cases of intersecting profiles, as observed, mean that usage of a single diversity index (e.g. Shannon or Simpson) will result in a particular diversity ranking that may be different when another index would have been used. Diversity profiles have the advantage over single diversity indices of allowing for a more complete characterisation of diversity – and therefore a well-informed design of diversification interventions.

Our results indicated spatial patterns in the distribution of diversity of many use-groups or niches. In many cases, a more random distribution of species (and knowledge on species’ uses) would result in higher average richness of spatial clusters. Such results indicate the potential of diversification without introducing new species in the landscape. Species that are dominant in one cluster and not in another could be prime species to introduce in clusters where they have

CHAPTER 14

233

low abundance, since such species would already have demonstrated their fitness for a particular purpose in some location(s). Species accumulation patterns also provide information on the possibility of enhancing diversity by modifying the distribution of species that are already present in the landscape. Especially where alpha diversity is low and beta diversity high, the distribution of species of lower frequencies over a larger number of farms would substantially increase the alpha diversity of the average farm. In contrast, systems with high alpha and low beta diversity have a more limited scope for diversification with species that are already present.

The regression results that indicated significantly differences in alpha diversity among use-groups confirm that some groups can be selected with lower average alpha diversity. However, within each use-group, farms will show a range of low to higher diversity, and therefore diversification may not be relevant for each farm. The low percentage of variation that was explained by characteristics of farmers show that there is limited scope of selecting farms of lower diversity by targeting specific types of farms (e.g. female-headed households). We did not investigate interactions between use-groups and characteristics of households since the total number of interactions was very large. Analysis of the influence of household characteristics on diversity and abundance of use-groups separately for western Kenyan households indicated interaction effects, for example. In contrast to the overall trend (table 14.3), wealthier households had fewer trees in the construction, fruit, and medicine use-groups, for example. Percentages of explained variation were low as well for the models for separate use-group, however (Kindt et al. – Chapter 8).

What needs to be investigated together with farmers are the reasons why some species occur in lower frequencies now – maybe they have limited fitness for a particular use, or few farmers know how to use the species, or farmers do not have access to germplasm of the particular species, or their low frequency could even be caused by over-use. It is obvious that efforts to increase the frequency of species should consider farmers’ perceptions and limitations. Where species’ performance is currently limiting farmers’ adoption of a species, genetic improvement for particular on-farm uses may contribute to wider adoption, but this can be an expensive and time-consuming process.

The accumulation patterns of species richness and diversity profiles indicate sample scale-dependency of diversity. The existence of accumulation patterns documents that it would be meaningless to make statements such as “the agroecosystem around Mabira harbours 30% of the tree species of Mabira forests” without making a reference to the (equal) sample size on which the comparison is made. Groups with intersecting accumulation patterns especially demonstrate scale-dependency since, in these cases, diversity rank orders will reverse from small to large sample scale. These patterns indicate the need to base comparisons of diversity on the same sample scale. As we were mainly interested in differences in species’ groups for the same survey, it was not necessary to interpolate Rényi diversity profiles within surveys. When comparing diversity across surveys, however, diversity will need to be interpolated to the same sample size. Kindt et al. (Chapter 13) demonstrated how average diversity profiles and ranges can be constructed for equal sample sizes based on site number, area, or tree abundance, and how choice of sample reference influences diversity ranking.

We present comparisons among countries based on sample sizes of the number of accumulated farms and the number of accumulated trees in figures 14.9 and 14.10, respectively. For random accumulation of farms, figure 14.9 was derived from figure 14.8 by multiplying the number of farms with the average number of trees of a farm for that location. This is the standard procedure for changing the reference of samples to individuals, since N × average abundance number of trees are expected for N randomly-accumulated sites (Gotelli & Colwell 2001; Kindt et al. – Chapter 6). For each location, two rarefaction curves were calculated: based on randomly-accumulated farms (as previously in this chapter), and based on randomly-accumulated individual trees. The average number of accumulated species for random accumulation of trees was calculated by individual-based rarefaction calculated from the abundance of each component


234

species (Sanders (1968) corrected by Hurlbert (1971)) using Kindt (2001b). (Results from individual-based rarefaction differ with the abundance reference of Chapter 13, which was calculated by randomly accumulating farms until the fixed abundance reference was approximated. Rarefaction implies that each tree was selected at random.) The fact that the individual-based curves are positioned above the sample-based curves implies spatial aggregation of species within sites (Gotelli & Colwell 2001) which corresponds to our earlier findings. It can be observed that central Kenya has most species on basis of the accumulated number of farms (figure 14.9), while central Uganda has most species on basis of the accumulated number of individuals (figure 14.10). Cameroon seems to be the location with the strongest non-linear pattern of curves in the log-log space, as (visual) extrapolation of the curves suggests that this would be the location with the smallest species richness if more farms or trees would be sampled.

The downwards convex pattern of most species accumulation curves was documented earlier for Kenyan use-groups (Kindt et al. – Chapter 6), while smaller slopes of the accumulation curves at increasing scale were reported for other surveys as well (Plotkin et al. 2000; May & Stumpf 2000; Crawley & Harral 2001). The scale-dependent values of the slope-parameter zN of log-log species accumulation patterns are presented in figure 14.6 for niches, and figure 14.7 for use-groups. It is interesting to observe that the slope-parameters for all species had values in between 0.6 and 0.7 for 2 accumulated farms, and had values around 0.5 for 40 accumulated farms in each location. The scale-dependent pattern means that, on a log-log scale, species richness will increase in a less than linear way, with largest increments for the smaller sample sizes. Used in an extrapolative manner, this means that smaller species richness is expected for the complete area than would be associated with a straight log-log accumulation curve. The convex pattern also indicates that fragmentation of the area would result in smaller species losses initially – although larger losses would be expected later through ecosystem relaxation (Rosenzweig 1999, 2001; Gonzalez 2000; Kindt et al. – Chapter 6).

14.4.4. Biodiversity Conservation in African Agroecosystems

In general, our results demonstrate that farmers cultivate relatively high diversity of trees on farms, especially when scaled-up from the individual farm to the village and larger spatial areas. Our results also indicate areas where tree species composition could be diversified, including intra-specific diversity – although this aspect of diversity was not targeted specifically in our surveys. Based on the information that we collected and some of the analyses that we reported in this chapter, efforts to “domesticate landscapes” are currently underway in the areas that we surveyed. Although we do not expect that farmers will conserve all indigenous species that were historically present in the landscapes that we investigated, we are hopeful that ongoing research conducted together with farmers will demonstrate that a substantial percentage of tree species can be conserved-through-use in an agroecosystem while also contributing through diversity to the wellbeing of those local farmers that manage these systems. Especially in areas where forests are under threat of fragmentation and extinction, conservation-through-use may offer the best conservation approach.

We do not want to undervalue the need to protect remaining forest ecosystems in the landscapes that we investigated, however. Although only three species were included in the IUCN Red List, more species could be threatened as deforestation progresses. Some of these species could not be useful to farmers or suited to ecological conditions of agroecosystems and could therefore only be conserved in forests. Evolutionary forces may be different in agroecosystems and conservation-through-use may therefore not be equivalent to in situ conservation (nor as ex situ conservation for that matter). In fragmented landscapes, farms may provide corridors that provide a necessary link for conservation of tree species in otherwise isolated forest fragments –

CHAPTER 14

235

trees may be needed both in agroecosystems and in remaining forest ecosystems to enable survival of the species.

Where forests are fragmented or gone entirely in a landscape, trees in agroforestry systems may offer suitable habitats for other organisms. Safford & Jones (1998) for example reported that restoration of native vegetation is not always the most effective conservation method of animal species, but that certain exotic species can be essential. On the other hand, some agri-environmental schemes may not be effective in conserving plant and animal species (Kleijn et al. 2001). Investigations of bird species composition on farms and in Mabira forest showed significant differences in their composition (Boffa et al. in prep.). Some farmers in the western Kenya survey explained that they established some fruit trees to feed bird species as they felt it was their human duty to do so. Similar to the need to investigate whether tree species are “living dead”, case-specific studies are therefore needed to investigate whether associated species are or could be conserved in mixed landscapes.

The results of the Biodiversity Conservation Network that studied the hypothesis that people will conserve natural habitats if they can benefit financially from community-based enterprises that depended on them showed that they only lead to conservation under a limited set of conditions (Salafsky et al. 2001). There was a weak association between enterprise success and conservation success, but a strong association between local involvement in the enterprise and conservation success. Enterprises that are not linked to biodiversity that are easier to be implemented and more profitable may actually be more effective. This study indicates that conflicts may exist between conservation of local biodiversity and livelihood strategies of local people.

Balmford et al. (2001) and Huston et al. (2001) indicate that in sub-Saharan Africa, biodiversity conservation and human needs may indeed result in conflicts since biodiversity and human population density are positively correlated. McNeely & Scherr (2001) describe that, since 1.1 billion people live in the 25 global biodiversity hotspots identified by Myers et al. (2001), a new type of agriculture is needed that leads to increased food security and conservation gains. Their report provides examples of innovative landscape management strategies that successfully combined both objectives by applying ecoagriculture strategies. These six strategies include enhancing wildlife habitat on farms and corridors that link uncultivated spaces in the landscape (Strategy 2); establishing protected areas near farming areas (Strategy 3) and mimicking natural habitats by integrating productive perennial plants (Strategy 4). Although we did not collect data that enable us at this point to establish that management of tree species in the four landscapes could be described as effective ecoagriculture, we did document here that farmers managed their ecosystems in ways similar to some ecoagriculture examples, especially for Strategy 4. We can therefore be carefully optimistic that some biodiversity can be conserved, although this hypothesis needs to be effectively tested through whole-landscape research, especially on metapopulation dynamics of flora and fauna in the landscape matrices.


236

Table 14.1. Farm frequencies, alpha diversity (average farm diversity), richness of botanical species and families, abundance on farms where the group is present and proportions of dominant species (dom.) and exotic and planted species and trees. Survey Group (all, niche or use) % farms Alpha

diver-sity

Species total (ST)

% exotic

species

Family total

Abun-dance

per farm

% abun-dance dom.

% abun-dance

exotics

% abun-dance

planted

% abun-dance

planted exotics

Cameroon All 100.0 29.5 119 12.6 42 171.8 13.8 29.9 48.2 28.2 (n=39) Cocoa 92.3 24.9 116 11.2 41 147.8 13.8 26.3 46.7 24.8 Homegarden 53.8 3.1 23 47.8 18 18.1 21.6 78.4 90.0 74.5 Foodcrop 41.0 3.2 56 16.1 31 36.9 15.6 35.2 51.8 34.6 Fallow 23.1 2.7 52 19.2 30 45.4 12.5 23.2 25.4 20.3 Firewood 100.0 21.7 109 9.2 39 128.7 18.0 33.6 46.8 31.7 Fruit 100.0 8.1 30 36.7 18 84.8 27.6 59.0 90.6 56.1 Medicine (human) 100.0 13.6 84 9.5 37 88.3 16.0 30.8 54.6 29.8 Construction 97.4 10.5 76 3.9 35 46.3 18.4 1.2 12.1 1.2 Spices 97.4 2.0 18 11.1 12 9.8 55.0 0.5 33.8 0.3 Tools 84.6 3.1 40 12.5 23 14.9 19.1 10.0 15.7 10.0 Soil fertility 76.9 1.4 28 7.1 16 13.2 44.7 4.3 18.4 4.0 Stimuli 64.1 1.0 5 0.0 4 5.6 82.3 0.0 90.1 0.0 Gums 51.3 0.7 11 18.2 10 14.0 45.0 22.9 55.7 16.1 Drugs 30.8 0.6 13 0.0 10 10.0 45.0 0.0 46.7 0.0 Fodder / animal medicine 30.8 0.5 15 26.7 13 12.8 27.9 53.2 64.9 48.1 Shade 25.6 0.8 19 10.5 15 16.8 16.7 23.8 37.5 23.8 Vegetables 23.1 0.2 3 0.0 3 3.2 89.7 0.0 0.0 0.0 Mabira All 100.0 27.9 249 23.3 62 210.7 12.6 31.7 37.4 22.6 (n=105) Cropland 99.0 20.1 202 21.3 55 122.3 13.3 29.0 31.6 18.1 External boundary 85.7 4.6 100 21.0 35 37.6 25.7 40.0 64.0 33.8 Homestead 81.9 5.3 120 36.7 41 21.7 15.7 56.8 70.7 49.4 Fallow 21.9 2.5 94 14.9 37 88.5 8.0 11.4 8.1 5.3 Internal boundary 13.3 0.6 46 15.2 25 28.9 38.0 44.9 56.3 42.0 Firewood 100.0 15.1 187 17.6 46 118.3 15.5 20.3 24.1 11.4 Fruit 100.0 5.5 24 58.3 17 31.4 28.2 95.2 61.2 59.1 Medicine 92.4 3.6 79 24.1 29 25.2 27.8 27.5 24.4 19.1 Construction 89.5 2.4 70 9.5 30 41.7 59.5 15.9 25.2 10.9 Timber 89.5 3.8 71 19.7 27 33.2 18.5 4.4 22.1 3.6 Shade 80.0 3.0 93 33.3 36 24.5 32.0 16.9 46.7 10.3 Boundary demarcation 78.1 1.4 19 36.8 12 32.3 43.2 51.0 67.9 42.9 Soil fertility 60.0 1.0 32 21.9 12 17.5 22.2 28.7 55.4 27.4 Leaves for cleaning utensils 41.0 0.4 1 0.0 1 4.2 100.0 0.0 1.1 0.0 Charcoal 36.2 1.1 58 12.1 22 24.7 20.7 7.0 8.5 4.8 Fodder 33.3 0.6 16 31.3 6 19.5 55.3 28.2 72.4 24.0 Ornamental 24.8 0.4 31 48.4 18 4.3 20.7 47.8 78.8 45.1 Stakes 24.8 0.3 6 16.7 3 46.1 93.4 93.6 91.0 89.2 western All 100.0 16.6 175 22.9 49 508.1 16.9 72.3 80.2 62.6 Kenya Cropland 98.5 6.5 105 27.6 37 107.1 56.9 71.8 87.9 67.2 (n=201) External boundary 95.0 4.6 64 31.3 31 211.8 30.8 70.8 87.9 63.8 Homestead 89.6 5.0 107 29.0 34 24.6 22.6 73.4 67.5 57.1 Woodlot 74.6 2.9 72 29.2 32 152.2 65.1 79.6 72.2 68.0 Internal boundary 36.3 0.9 43 25.6 26 65.5 16.4 54.8 90.8 48.4 Fallow 20.9 1.3 81 27.2 35 113.8 41.5 55.3 14.7 9.1 Crop contours 11.9 0.2 23 43.5 12 34.6 19.5 63.0 46.7 41.7 Firewood 100.0 15.2 156 25.6 45 476.7 18.0 75.9 79.4 65.7 Fruit 100.0 4.7 25 60.0 14 46.1 72.6 99.2 21.8 21.5 Boundary demarcation 98.0 2.8 34 38.2 20 217.6 30.9 73.1 92.6 66.3 Construction 98.0 1.9 20 50.0 11 111.6 78.4 80.3 82.9 79.3 Furniture 97.5 3.9 49 28.6 23 142.2 61.9 92.9 93.3 90.3 Shade 82.6 3.0 84 34.5 29 28.7 13.0 55.6 58.2 38.7 Soil fertility 67.7 1.0 27 44.4 18 87.8 52.8 63.7 89.7 63.6 Medicine 55.7 1.2 58 12.1 22 23.3 66.4 10.5 19.6 10.4 Ornamental 47.8 1.0 53 35.8 23 55.0 50.9 56.2 94.8 52.3 Charcoal 39.8 0.6 27 37.0 13 42.6 49.6 57.5 60.2 54.1 Beverage 30.8 0.4 4 75.0 4 309.4 80.3 80.3 100.0 80.3 Fodder 23.4 0.3 7 57.1 4 9.7 42.1 44.3 54.6 41.6 Meru All 100.0 53.2 294 27.9 64 1798.5 32.0 61.8 71.1 59.4 (n=35) Cropland 100.0 17.9 136 26.5 49 160.6 17.2 49.7 61.1 36.3 External boundary 100.0 26.6 198 20.7 54 692.2 10.0 47.7 60.4 37.2 Homegarden/homestead 94.3 8.7 107 53.3 42 28.3 20.2 63.7 77.4 59.6 Coffee garden 91.4 10.7 100 20.0 44 677.9 71.1 94.8 93.4 92.1 Internal boundary 88.6 8.2 114 34.2 41 186.4 15.4 76.9 80.9 71.0 Fallow 68.6 6.7 117 20.5 45 114.1 13.8 40.0 29.9 24.4 Crop boundary 54.3 1.9 36 36.1 20 17.9 19.1 72.7 72.4 64.2 Woodlot 45.7 5.3 103 23.3 46 93.0 11.3 31.9 51.1 27.3

CHAPTER 14

237

Table 14.1 (cont.d) Survey Group (all, niche or use) % farms Alpha

diver-sity

Species total (ST)

% exotic

species

Family total

Abun-dance

per farm

% abun-dance dom.

% abun-dance

exotics

% abun-dance

planted

% abun-dance

planted exotics

Meru Firewood 100.0 16.1 121 23.1 45 534.1 35.1 70.1 67.4 59.0 (n=35) Fodder 100.0 9.1 87 17.2 39 263.7 28.3 48.3 51.5 38.1 Fruits or nuts 100.0 9.8 53 49.1 28 106.2 13.9 71.9 46.0 40.0 Timber 100.0 8.1 54 24.1 29 191.3 45.7 65.7 62.8 51.3 Boundary demarcation 97.1 7.2 73 26.0 47 508.4 26.6 51.4 81.0 55.5 Cash 97.1 2.6 8 50.0 6 637.2 93.3 94.7 99.8 94.5 Medicine 97.1 7.1 96 14.6 40 200.6 14.7 23.0 45.3 17.1 Construction 91.4 4.2 49 26.5 27 126.2 28.5 63.1 55.9 47.3 Ornamental 82.9 3.6 51 70.6 24 46.1 34.1 86.9 98.2 88.5 Plant support 62.9 1.2 16 12.5 11 122.7 40.9 37.1 91.3 37.1 Animal traps 51.4 0.9 15 0.0 2 15.0 34.1 0.0 13.7 0.0 Charcoal 45.7 1.1 20 10.0 14 27.8 24.0 6.1 9.4 4.3 Tool handles 45.7 0.6 9 0.0 8 25.2 42.9 0.0 34.7 0.0 Shade 42.9 0.6 13 53.8 11 7.9 49.6 79.8 65.5 53.8

Table 14.2. Information on the spatial distribution of species richness and species composition for all species encountered in various landscapes, and for species arranged in niches or use-groups. Calculation of random, proximity-based and randomly-clustered richness are based on the smallest spatial cluster in each landscape (s, see Chapter 3), for calculation see methods section. RDA: results from Redundancy Analysis, with percentage of explained variation and significance based on 9,999 permutation tests. Survey Group Random Proximity-

based Randomly-clustered

(95% interval) RDA

% and significance Cameroon All 105.7 83.7 106.8 (103.5-109.5) 16.9 0.0001 (n=39) Cocoa 101.5 79.6 101.5 (97.9-104.5) 15.3 0.0001 (s=19) Homegarden 18.0 17.9 18.0 (16.5-19.1) 4.8 0.0691 Foodcrop 36.9 32.9 36.9 (31.9-39.7) 5.3 0.0249 Fallow 34.8 32.4 34.8 (27.4-38.8) 2.4 0.5017 Firewood 95.1 73.6 95.1 (91.9-97.6) 12.9 0.0001 Fruit 25.7 23.4 25.7 (24.4-27.0) 12.5 0.0001 Medicine (human) 67.5 57.6 67.5 (64.3-70.2) 8.5 0.0001 Construction 60.0 47.6 60.0 (56.6-62.6) 13.2 0.0001 Spices 11.9 10.4 11.9 (10.5-12.6) 7.9 0.0106 Tools 27.9 23.5 27.9 (25.4-30.0) 8.7 0.0001 Soil fertility 17.1 15.3 17.0 (15.4-18.6) 7.0 0.0030 Stimuli 4.3 4.0 4.3 (3.5-4.6) 6.9 0.0625 Gums 7.0 5.9 7.0 (5.9-8.1) 2.4 0.3842 Drugs 8.0 7.4 8.0 (6.8-9.0) 4.5 0.0345 Fodder / animal medicine 8.5 8.3 8.5 (7.0-9.1) 3.0 0.3207 Shade 11.3 10.7 11.3 (9.3-12.6) 3.9 0.0744 Vegetables 2.0 1.5 2.0 (1.4-2.1) 16.2 0.0077 Mabira All 141.9 129.8 141.9 (138.7-144.8) 9.5 0.0001 (n=105) Cropland 108.4 98.6 108.4 (105.5-110.8) 7.8 0.0001 (s=21) External boundary 44.8 43.0 44.8 (42.7-46.4) 8.3 0.0001 Homestead 50.3 48.0 50.3 (48.5-52.0) 4.5 0.1079 Fallow 38.0 37.2 38.0 (34.7-40.6) 4.0 0.3017 Internal boundary 11.8 12.4 11.8 (10.9-12.6) 3.3 0.8450 Firewood 97.9 93.0 97.9 (95.1-100.4) 7.9 0.0001 Fruit 17.3 17.4 17.3 (16.3-18.2) 6.1 0.0077 Medicine 34.8 37.0 34.8 (33.3-36.2) 4.5 0.1177 Timber 30.8 30.2 30.8 (29.3-32.0) 7.7 0.0002 Construction 25.6 27.0 25.6 (24.1-26.8) 5.1 0.1139 Shade 36.8 39.2 36.8 (35.1-38.4) 4.2 0.2633 Boundary demarcation 7.6 7.8 7.6 (7.1-8.0) 5.8 0.0798 Soil fertility 12.8 12.4 12.8 (11.7-13.6) 3.8 0.4481 Leaves for cleaning utensils 1.0 1.0 1.0 (0.9-1.0) 2.5 0.6060 Charcoal 19.1 19.8 19.1 (17.9-20.2) 3.9 0.4323 Fodder 6.4 6.8 6.4 (5.7-7.0) 4.8 0.2041 Ornamental 7.8 9.0 7.8 (7.1-8.2) 3.8 0.4827 Stakes 2.2 2.2 2.2 (1.9-2.2) 4.0 0.4036


238

Table 14.2 (cont.d) Survey Group Random Proximity-

based Randomly-clustered

(95% interval) RDA

% and significance western All 109.3 93.4 109.2 (106.9-111.5) 15.1 0.0001 Kenya Cropland 61.9 55.7 61.9 (59.9-63.8) 9.6 0.0001 (n=201) External boundary 40.5 33.9 40.5 (38.9-42.1) 14.3 0.0001 (s=50) Homestead 61.7 54.4 61.7 (59.7-63.6) 5.8 0.0001 Woodlot 37.6 34.2 37.6 (36.0-39.1) 3.1 0.0056 Internal boundary 22.6 20.0 22.7 (21.2-24.0) 4.5 0.0001 Fallow 37.6 35.2 37.6 (34.9-39.8) 2.7 0.0103 Crop contours 8.8 8.0 8.8 (7.9-9.6) 2.6 0.0059 Firewood 97.6 85.6 97.6 (95.5-99.6) 15.5 0.0001 Fruit 17.2 16.2 17.2 (16.4-17.8) 7.7 0.0001 Boundary demarcation 20.9 16.0 20.9 (19.7-22.0) 20.9 0.0001 Construction 9.4 9.0 9.4 (8.4-10.3) 5.4 0.0015 Furniture 28.9 25.7 28.9 (27.7-30.0) 9.8 0.0001 Shade 46.7 40.9 46.7 (44.9-48.3) 5.3 0.0001 Soil fertility 13.1 10.5 13.1 (12.2-14.0) 10.7 0.0001 Medicine 25.6 22.4 25.6 (24.2-26.8) 4.0 0.0001 Ornamental 23.9 19.7 23.9 (22.5-25.3) 5.3 0.0001 Charcoal 12.7 11.7 12.7 (11.7-13.8) 4.9 0.0006 Beverage 2.5 2.2 2.5 (2.4-2.6) 24.6 0.0001 Fodder 4.6 3.7 4.6 (4.2-4.8) 7.4 0.0001 Meru All 186.4 58.7 186.4 (175.0-196.4) 20.5 0.0001 (n=35) Cropland 75.5 70.4 75.5 (68.4-81.4) 14.3 0.0001 (s=10) External boundary 117.0 102.2 117.0 (107.8-124.8) 18.4 0.0001 Homegarden/homestead 53.2 50.5 53.2 (47.5-57.9) 8.6 0.0053 Coffee garden 54.6 46.7 54.6 (47.7-60.1) 14.8 0.0001 Internal boundary 55.8 50.9 55.8 (48.0-61.8) 10.6 0.0002 Fallow 50.7 46.5 50.8 (36.7-59.5) 10.2 0.0016 Crop boundary 14.8 15.5 14.8 (11.7-17.3) 5.9 0.4266 Woodlot 42.8 39.1 42.8 (31.2-50.8) 10.0 0.0055 Firewood 71.7 62.6 71.7 (66.6-76.2) 14.4 0.0001 Fruits or nuts 29.1 28.1 29.1 (26.3-31.7) 17.0 0.0001 Fodder 47.3 42.8 47.3 (42.2-51.5) 18.9 0.0001 Timber 30.8 27.8 30.8 (27.9-33.3) 16.1 0.0001 Boundary demarcation 34.1 31.8 34.2 (27.9-37.9) 17.7 0.0001 Cash 5.1 5.1 5.1 (4.3-5.6) 9.8 0.0927 Medicine 46.6 40.8 46.6 (39.4-51.3) 12.1 0.0001 Construction 24.2 23.1 24.2 (20.7-27) 11.0 0.0008 Ornamental 23.5 22.5 23.5 (19.8-26.3) 7.4 0.0904 Plant support 6.5 5.6 6.5 (4.7-7.8) 21.1 0.0019 Animal traps 6.4 6.2 6.4 (4.9-7.7) 10.7 0.0207 Charcoal 8.7 7.7 8.7 (6.4-10.6) 16.5 0.0001 Tool handles 4.0 3.4 4.0 (3.0-4.8) 26.5 0.0001 Shade 4.8 4.6 4.8 (3.6-5.8) 8.2 0.0892

CHAPTER 14

239

Table 14.3. Variation explained of use-group species richness (S) and abundance (N) of individual farms by listed explanatory variables of multiple regression models for each survey. Variables are binary (0/1) or ranged within the 0-1 interval (*). Partial regression results document total variation and variation explained by use-groups and by other explanatory variables. Coeff.: coefficient for the respective explanatory variable in the regression model, Sign.: significance (based on 99,999 permutations) Survey Explanatory variables ln(S+1) ln(N+1) Coeff. Sign. Coeff. Sign. Cameroon (Intercept) -0.44 0.01107 -0.93 0.00853 (n=39) Fodder -0.02 0.84396 0.14 0.53144 Firewood 2.74 0.00001 4.06 0.00001 R2 p(F) Spices 0.71 0.00001 1.37 0.00001 Richness (S) Fertility 0.46 0.00002 1.03 0.00001 All 0.81 0 Vegetables -0.15 0.17337 -0.28 0.19898 Use-groups 0.79 0 Fruit 1.85 0.00001 3.54 0.00001 Other 0.03 0.13 Gums 0.13 0.23529 0.52 0.01840

Construction 1.97 0.00001 2.82 0.00001 Abundance (N) Medicine 2.24 0.00001 3.56 0.00001 All 0.72 0 Shade 0.02 0.82689 0.06 0.78350 Use-groups 0.68 0 Tools 0.91 0.00001 1.45 0.00001 Other 0.04 7E-3 Stimuli 0.26 0.01798 0.40 0.06974

Distance from forest 0.24 0.00005 0.53 0.00001 Farm size* 0.04 0.00001 0.10 0.00001 Wealth rank* 0.05 0.00183 0.08 0.02254 Male-headed -0.22 0.01326 -0.19 0.27759 Age of head 0.00 0.49231 0.00 0.73203 Education level* 0.04 0.35120 0.16 0.06433 Indigenous to village 0.26 0.00503 0.28 0.13327 Size of household* 0.01 0.05940 0.01 0.28891 Use of hired labour 0.08 0.15032 0.10 0.32401

Mabira (intercept) 0.02 0.81101 0.00 0.99035 (n=97) Boundary 0.57 0.00001 1.60 0.00001 Charcoal 0.16 0.09556 0.30 0.09326 R2 p(F) Firewood 2.50 0.00001 3.53 0.00001 Richness (S) Fodder 0.05 0.64500 0.19 0.28069 All 0.57 0 Fruit 1.70 0.00001 2.42 0.00001 Use-groups 0.53 0 Medicine 1.16 0.00001 1.88 0.00001 Other 0.04 1E-6 Ornamental -0.12 0.22013 -0.24 0.16446

Construction 1.00 0.00001 2.19 0.00001 Abundance (N) Shade 0.93 0.00001 1.46 0.00001 All 0.47 0 Soil fertility 0.39 0.00009 0.81 0.00002 Use-groups 0.43 0 Timber 1.18 0.00001 1.85 0.00001 Other 0.04 7E-8 Male -0.07 0.12118 -0.12 0.17414 Female de jure -0.12 0.06879 -0.08 0.51562 Forest user 0.19 0.00084 0.35 0.00096 Wealth* 0.29 0.00001 0.49 0.00001 Years head* 0.09 0.42414 0.43 0.04136 Maximum age head or partner* 0.01 0.93748 -0.22 0.30149 Number of children* 0.49 0.00007 0.69 0.00143 Close to forest 0.09 0.07240 0.15 0.09480 Far from forest 0.17 0.00066 0.38 0.00007


240

Table 14.3 (cont.d) Survey Explanatory variables ln(S+1) ln(N+1) Coeff. Sign. Coeff. Sign. western Kenya (intercept) -0.20 0.00072 0.35 0.06338 (n=183) Boundary demarcation 1.00 0.00001 3.47 0.00001 Charcoal 0.11 0.02267 -0.19 0.20373 R2 p(F) Construction 0.80 0.00001 2.68 0.00001 Richness (S) Firewood 2.47 0.00001 4.55 0.00001 All 0.70 0 Fodder -0.07 0.14105 -0.89 0.00001 Use-groups 0.69 0 Fruit 1.45 0.00001 1.93 0.00001 Other 0.02 1E-4 Timber 1.23 0.00001 2.92 0.00001 Medicine 0.37 0.00001 -0.22 0.14834 Abundance (N) Ornamental 0.27 0.00001 -0.08 0.60078 All 0.59 0 Shade 0.90 0.00001 0.59 0.00005 Use-groups 0.56 0 Soil fertility improvement 0.33 0.00001 0.44 0.00368 Other 0.02 1E-6 Distance to Kakamega forest* 0.13 0.00002 0.15 0.10614 Male headed household 0.11 0.00009 0.27 0.00199 Female headed household (de jure) 0.14 0.00021 0.34 0.00224 Farm size* 0.18 0.00384 0.86 0.00002 Permanent house 0.00 0.92120 0.18 0.11070

Thatch-roofed house -0.05 0.02753 -0.26 0.00095 Number of cattle* 0.12 0.07939 0.35 0.09167 Years head* 0.08 0.12928 0.02 0.89883 Age household head* 0.24 0.00012 0.45 0.02425 Number of children* 0.10 0.03889 0.18 0.25009 Education level* 0.15 0.00026 0.33 0.00959 Meru (intercept) 0.23 0.16650 0.62 0.15725 (n=32) Boundary demarcation 1.38 0.00001 4.24 0.00001 Cash 0.72 0.00001 4.40 0.00001 R2 p(F) Charcoal 0.03 0.83089 -0.01 0.97264 Richness (S) Construction materials 0.88 0.00001 2.18 0.00001 All 0.69 0 Fodder 1.63 0.00001 3.63 0.00001 Use-groups 0.65 0 Fruit or nut 1.82 0.00001 3.00 0.00001 Other 0.04 0.04 Fuelwood 2.22 0.00001 4.32 0.00001

Medicine 1.33 0.00001 2.62 0.00001 Abundance (N) Ornamental 0.66 0.00001 0.98 0.00909 All 0.61 0 Plant support 0.16 0.25396 1.44 0.00017 Use-groups 0.58 0 Shade -0.18 0.20136 -0.56 0.12606 Other 0.04 0.07 Timber 1.61 0.00001 3.63 0.00001

Tool handles -0.12 0.37165 -0.04 0.92463 Distance from forest* -0.08 0.19412 0.02 0.92094 Farm size* 0.23 0.00076 0.45 0.01471 Male-headed household -0.07 0.30296 -0.01 0.93782 Age of trial participant* 0.09 0.16638 0.17 0.32241 Education level of trial participant* -0.29 0.17702 0.26 0.64882

Wealth rank* 0.22 0.06931 0.32 0.32145 Income generation from farm 0.07 0.60959 -0.43 0.21145 Off-farm employment 0.30 0.00906 0.57 0.06308 Number of children* 0.00 0.98099 0.30 0.40115

Figure 14.1. Accumulation patterns for the Rényi diversity profile for all species. Horizontal axes represent the Rényi scale parameter a and the randomly accumulated number of farms, the vertical axis the Rényi diversity profile Ha. The bold line corresponds to the average diversity profile corresponding to 35 farms, which is the smallest sample size of the four surveys (Meru). Alpha values of 4, 4.5

and 5 in the figure correspond to scale parameter a values of 10, 100 and 8. Upper left: Cameroon, upper right: Mabira, lower left: western Kenya , lower right: Meru.

1 10

Number of farms

1

10

100

Nu

mb

er o

f sp

ecie

scocoahomegardensfallowsfoodcropsall

1 10 100

Number of farms

1

10

100

Num

ber

of s

peci

es

intern. bound.homesteadcroplandfallowextern. bound.all

1 10 100

Number of farms

1

10

100

Nu

mb

er o

f sp

ecie

s

intern. bound.homesteadwoodlotcroplandcrop contoursfallowextern. bound.soil fertility

1 10

Number of farms

1

10

100

Num

ber

of s

peci

es

intern. bound.crop bound.homesteadwoodlotfallowcoffeecroplandextern. bound.all

Figure 14.2. Species accumulation curves for trees in various on-farm niches. Upper left: Cameroon, upper right: Mabira, lower left: western Kenya , lower right: Meru.

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

EX

P(H

)

cocoafoodcrophomegardenfallowall

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

6

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

244.7

403.4

EX

P(H

)

internal boundaryhomesteadcroplandfallowexternal boundaryall

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

6

H

internal boundarycroplandwoodlotexternal boundaryhomesteadcrop contoursfallowall

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

244.7

403.4

EX

P(H

)

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

6

H

internal boundarycoffeecroplandcrop boundarywoodlotfallowexternal boundaryhomesteadall

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

244.7

403.4

EX

P(H

)

Figure 14.3. Rényi diversity profiles for trees in various on-farm niches for the complete sample. Upper left: Cameroon, upper right: Mabira, lower left: western Kenya , lower right: Meru.

1 10

Number of farms

1

10

100

Nu

mb

er o

f sp

ecie

sspicesfruitfirewoodmedicinestimuliconstructionshadesoil fertilitytoolsgumsfodderdrugsvegetablesall

1 10 100

Number of farms

1

10

100

Num

ber

of s

peci

es

boundaryfruitfirewoodmedicinetimberconstructionshadesoil fertilityleavescharcoalfodderornamentalstakesall

1 10 100

Number of farms

1

10

100

Nu

mb

er o

f sp

ecie

s

boundaryfruitfirewoodmedicinetimberconstructionshadesoil fertilitybeveragecharcoalfodderornamentalall

1 10

Number of farms

1

10

100

Num

ber

of s

peci

es

boundaryfruit/nutfirewoodmedicinetimberconstructionshadecashplant supportcharcoalfodderanimal trapsornamentalstool handlesall

Figure 14.4. Species accumulation curves for trees in various use-groups. Upper left: Cameroon, upper right: Mabira, lower left: western Kenya , lower right: Meru.

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

EX

P(H

)

spicesfruitfirewoodmedicinestimuliconstructionshadesoil fertilitytoolsgumsfodderdrugsvegetablesall

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 NaN

alpha

0

1

2

3

4

5

6

H

1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

244.7

403.4

EX

P(H

)

boundaryfruitfirewoodmedicinetimberconstructionshadesoil fertilityleavescharcoalfodderornamentalstakesall

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

6

H


1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

244.7

403.4

EX

P(H

)

0 0.1 0.25 0.5 0.75 1 1.25 1.5 1.75 2 2.5 3 10 100 inf.

alpha

0

1

2

3

4

5

6

H


1.0

1.6

2.7

4.5

7.4

12.2

20.1

33.1

54.6

90.0

148.4

244.7

403.4

EX

P(H

)

Figure 14.5. Rényi diversity profiles for trees in various use-groups for the complete sample. Upper left: Cameroon, upper right: Mabira, lower left: western Kenya , lower right: Meru.

10

Number of farms

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Nu

mb

er o

f sp

ecie

scocoahomegardensfallowsfoodcropsall

10 100

Number of farms

0.4

0.5

0.6

0.7

0.8

0.9

1.0

z


10 100

Number of farms

0.4

0.5

0.6

0.7

0.8

0.9

1.0

Nu

mb

er o

f sp

ecie

s

intern. bound.homesteadwoodlotcroplandcrop contoursfallowextern. bound.soil fertility

10

Number of farms

0.4

0.5

0.6

0.7

0.8

0.9

1.0

z

intern. bound.crop bound.homesteadwoodlotfallowcoffeecroplandextern. bound.all

Figure 14.6. Accumulation patterns of the parameter z of the S=aNz model for trees in various niches. Upper left: Cameroon, upper right: Mabira, lower left: western Kenya , lower right: Meru.

10

Number of farms

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

zspicesfruitfirewoodmedicinestimuliconstructionshadesoil fertilitytoolsgumsfodderdrugsvegetablesall

10 100

Number of farms

0.4

0.5

0.6

0.7

0.8

0.9

1.0

z


10 100

Number of farms

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

z


10

Number of farms

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1.0

z


Figure 14.7. Accumulation patterns of the parameter z of the S=aNz model for trees in various use-groups.. Upper left: Cameroon, upper right: Mabira, lower left: western Kenya , lower right: Meru.

0 5 10 15 20 25 30 35 40 45 50 55


0

5

10

15

20

0

5

10

15

20

Nu

mb

er o

f use

s p

er h

ou

seh

old

0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75


0

5

10

15

20

25

0

5

10

15

20

25

Nu

mb

er o

f use

s p

er h

ou

seh

old

0 5 10 15 20 25 30 35 40 45 50


0

5

10

15

20

25

30

0

5

10

15

20

25

30

Nu

mb

er o

f use

s p

er h

ou

seh

old

0 10 20 30 40 50 60 70 80 90 100


0

5

10

15

20

25

30

35

0

5

10

15

20

25

30

35

Nu

mb

er o

f use

s p

er h

ou

seh

old

Figure 14.8. The relationship between the number of species and the number of uses on a farm. Upper left: Cameroon, upper right: Mabira, lower left: western Kenya , lower right: Meru.

CHAPTER 14

1 10 100

Number of farms

10

100

Nu

mb

er o

f sp

ecie

s

central Ugandacentral KenyaCameroonwestern Kenya

random accumulation of treesrandom accumulation of farms

Figure 14.9. The average accumulation of species in the four locations when trees or farms are accumulated at random. The

reference basis of the same number of accumulated farms is used.

100 1,000 10,000 100,000

Number of trees

10

100

Nu

mb

er o

f sp

ecie

s

central Ugandacentral KenyaCameroonwestern Kenya

random accumulation of treesrandom accumulation of farms

Figure 14.10. The average accumulation of species in the four locations when trees or farms are accumulated at random. The

reference basis of the same number of accumulated trees is used.

CHAPTER 15 GENERAL CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE RESEARCH AND DEVELOPMENT

ACTIVITIES

R KINDT

CONCLUSIONS AND RECOMMENDATIONS

252

In this chapter, the hypotheses that were formulated in Chapter 1 are revisited. Some of the new tools that were developed to test some of the hypotheses are discussed in further detail. These tools could be listed as results from the investigations of this thesis. Based on the results of the previous chapters, some recommendations for future research and development activities in the research areas are formulated.

15.1. General Conclusions

As we indicated in Chapter 1, a large part of this thesis was devoted to ranking subsections of the surveyed landscapes from low to high diversity. Based on diversity rankings, development activities could focus on subsections of lower diversity, as conform ecological theory and modelling larger consequences can be expected on ecosystem functioning by enhancing diversity of these subsections. We did not alter the diversity of the agroecosystems in which we conducted the investigations, and did not measure consequences of changes in diversity, for example on the stability of an agroecosystem. We even did not measure ecosystem properties of stability and productivity of the agroecosystems in which the surveys were conducted. The results and conclusions for this thesis thus reflect exploratory research as it was concerned with the (statistical) analysis of patterns observed in the investigated systems. The nature of the research was thus not experimental. Therefore, the results do not reflect investigations of the consequences of experimental modifications of systems, which allow for investigations of cause-result relationships (and not merely correlation relationships following the assumption that no other factors affected the systems that are investigated before and after the experimental modifications).

We compensated the exploratory nature of the investigations somewhat by discussing experimental and modelling results (a model could be interpreted as a set of rules for an imaginary system that allows for simulation experiments of cause-result patterns) of diversity-stability and diversity-productivity relationships in Chapter 2. Chapter 2 further documented that analysis of the diversity-productivity relationship in simple grassland ecosystems had recently lead to controversies regarding the statistical methodology used. Methods that are more suitable and more ad hoc were only documented as recently as in a July 2001 issue of Nature (Loreau & Hector 2001). The most recent review on the subject provided in an October 2001 issue of Science also indicated that further ecological research is needed, even stressing the need for additional statistical methods given the large number of species combinations that could be tested (Loreau et al. 2001). Cottingham et al. (2001) presented a critique of methodological limitations to extrapolation of past studies of the richness – temporal variability relationship, including limitations in experimental design, lack of realism of theoretical models, and problems with statistical analysis. Another recent review stated that “even the most recent reviews of biodiversity assessment have not discussed the sampling issues we address in this review in relation to the measurement and comparison of species richness” (Gotelli & Colwell 2001). It is also noteworthy that many ecological experiments were conducted in grassland ecosystems (Chapter 2) or other types of less complex ecosystems (e.g. Naeem et al. 1994; Cardinale et al. 2002), whereas agroforestry systems should probably be measured over substantially longer time scales given the perennial nature of trees. Even a grassland experiment investigating the diversity – productivity relationship took 7 years to yield substantial results (Tilman et al. 2001b). Thus, experimental investigation of diversity (and by consequence, diversification as an increase from one diversity level to the next) is not an easy task, and future research could benefit from new ecological/statistical analysis methods that still need to be developed.

Within ICRAF - the World Agroforestry Centre, diversification of agroecosystems is presented as a pathway towards increments in productivity and stability of these systems, for which there is an

CHAPTER 15

253

ecological basis as documented in Chapter 2. This thesis can be framed within this diversification theme, as it provides detailed information on the diversity that is currently present in these systems, i.e. the baseline by which subsequently the degree of diversification (degree of modification of experimental research) can be measured. The research could be portrayed as simply addressing diversification of agroecosystems which is an objective persecuted by the institute, assuming that diversification will automatically lead to increased social, economic and environmental benefits as documented in the Leakey (1996) definition of agroforestry. At the same time, by analyzing differences in diversity we explored the scene for potential future research that would effectively test the underlying hypothesis of a diversity-ecosystem behaviour relationship, which was not tested by our investigations. This hypothesis could be formulated as ‘Diversification of a section of lower diversity of an ecosystem will positively affect agroecosystem p roductivity and stability to a higher degree than the same degree of diversification of a section of higher diversity’ . Our results allow setting up experiments in sections of low and high diversity discovered by the investigations reported here, and for experimentally sound testing of the diversification-ecosystem behaviour hypothesis.

An essential reason for the existence of a centre with a worldwide mandate for research in agroforestry is to investigate hypotheses such as the Leakey (1996) definition of agroforestry. As one ICRAF scientist recently remarked (T. Tomich, pers. comm.), the role of ICRAF is not so much to state that “trees are good” but to document “what trees are good for”.

In summary, the investigations of this thesis were concerned with documenting the current levels of diversity in ecosystems, and of differences in diversify among subsections of these ecosystems, within a theme of diversification towards better ecosystem functioning, which forms a first phase towards experimental testing of such relationship.

The hypotheses that are listed hereafter were provided in Chapter 1. They all reflect the exploratory nature of this thesis. Hypothesis 22 forms the exception, documenting the results of experiments in modeled metapopulations, which were populations of tree populations, where each subpopulation occurred on one farm. The chapters in which these hypotheses were tested, and whether the hypothesis was accepted or rejected, are indicated. The consequences of accepting or rejecting a hypothesis for designing management options for tree species diversity on farms was already provided in Chapter 1. For brevity, the consequences of groups of hypotheses are provided below.


? Hypothesis accepted in Chapter 5 (figure 5.5) & 14 (figure 14.8)

? The relationship is not valid for each individual farm, however, as at the same richness level some farms have many uses and other farms only have a few uses.

H2 The number of uses of a species in the entire survey equals the number of uses of a species on each farm

? Hypothesis rejected in Chapter 5 (figure 5.2).

H3 Recording information only on the main use of each species on a farm provides an adequate representation of the distribution of diversity of species of similar on-farm use

? Hypothesis rejected in Chapter 5 (table 5.1 - 5.3).


? Hypothesis rejected in Chapter 5 (table 5.2), 6 (table 6.1, figure 6.1) & 14 (table 14.1, figure 14.2 & 14.4)


254


? Hypothesis rejected in Chapter 5 (table 5.2), 6 (figure 6.1) & 14 (table 14.1, figure 14.2 - 14.5)


? Hypothesis rejected in Chapter 5 (table 5.2), 7 (figure 7.1, table 7.1) & 14 (figure 14.3 & 14.5)

H7 Species density, the average number of trees per species per ha, is the same for each grouping of species of similar on-farm use

? Hypothesis rejected in Chapter 5 (table 5.3)


? Hypothesis rejected in Chapter 6 (figure 6.2) & 14 (figure 14.6 & 14.7)




? Hypothesis accepted in Chapter 6 (table 6.2) & 14 (table 14.2)








? Hypothesis accepted in Chapter 8 (table 8.4) & 14 (table 14.3)

? Percentages of variation that were explained were low, however. This indicates that many farms of a certain type deviate from the average diversity or abundance of their type.

H15 The same area of land, consisting of many or few randomly-accumulated farms, has the same diversity and abundance

? Hypothesis rejected in Chapter 8 (table 8.5)


? Hypothesis accepted in Chapter 9 (table 9.2), 10 (table 10.4) & 14 (table 14.2)

? Percentages of variation that were explained were low, however. This indicates that many farms of a certain type deviate from the average composition of their type.

CHAPTER 15

255

H17 Relationships between farm characteristics and the abundance of a particular species corresponds to the relationships between farm characteristics and the relative species composition of the farm

? Hypothesis accepted in Chapter 9 (figures 9.1 – 9.14, table 9.3) & 10 (figures 10.1 – 10.7, table 10.5).

H18 The diversity of species that is present on farms is desired by farmers

? Hypothesis accepted in Chapter 11 (figure 11.1 – 11.3, table 11.1)

? A fortiori, we recorded that farmers wanted to establish more diversity on their farms than currently present.

H19 For a particular use, farmers have reasons to maintain more than one species

? Hypothesis accepted in Chapter 11 (table 11.2)

? Not every farmer wanted to maintain several species for each use, however.

H20 For a particular use, farmers have reasons not to maintain species in perfectly even distributions


H21 Species are aggregated within farms and villages

? Hypothesis accepted in Chapter 12 (figure 12.2)

H22 Changes in the way in which germplasm is collected will affect genetic diversity

? Metapopulation simulations with information on the farmer-preferred origin and form of germplasm showed minimal consequences on genetic diversity from different scenarios of germplasm acquisition (figures 12.5 – 12.7).

? Simulations were hampered, however, as information on species-specific reproductive ecology and genetic diversity currently present (and possibly affected by past population bottlenecks) was lacking. Comparisons were also based on simulations of 50 subsequent flowering seasons, which may be too short to result in significant differences, although such time frame was chosen as it was a realistic planning horizon for discussions with farmers.

H23 Farmers prefer to collect most trees from the own farm


H24 The diversity rank-order of surveys depends on whether comparisons are based on an equal sample size of the same number of sites, the same number of trees, or the same area

? Hypothesis accepted in Chapter 13 (figures 13.3 – 13.6)

H25 Distance from forests influences diversity of villages

? Hypothesis accepted in Chapter 13 (figure 13.4)

? Distance from forests mainly affected diversity of villages through a reduction in the evenness of the dominant species.

H26 The diversity rank-order of gamma diversity of surveys depends on whether comparisons are based on an equal sample size of the same number of sites or the same number of trees

? Hypothesis accepted in Chapter 14 (figure 14.9 & 14.10)

H27 A larger percentage of on-farm trees are exotic than the percentage of exotic species on-farm



256

As indicated in Chapter 1, we investigated subsections by first forming their respective farms × species matrix and subsequently calculating some statistics that allowed ranking these matrices from low to high diversity. When development activities are planned, criteria need to be determined that specify how a landscape (and the trees occurring in this landscape) is subdivided, before subsections can be compared. Various options for potential criteria were provided throughout this thesis: (a) geographical position (village); (b) uses of the trees; (c) on-farm niche of the trees; (d) characteristics of the farm; or (e) combinations of the above.

Hypotheses 4-6, 9, 11-12, 14, 16, and 25 were all concerned with ranking subsections of the landscapes in diversity. Hypothesis 7, 14, and 16 (based on Hypothesis 17) focused the investigations on differences in abundance. We will concentrate our discussion here on differences in diversity, however. Results on abundance can be interpreted in a similar way when planning interventions. These hypotheses were all accepted or rejected to the effect that subsections of substantially lower diversity could be identified. Within the framework of selecting priority subsections for diversification, these are important findings. That there is a plethora of hypotheses that yielded significant differences indicates that future research or development has many opportunities. Hypotheses 4-7, for example, ranked subsections based on uses, so uses of lower diversity could be targeted. Subsections based on niches were investigated and ranked in an analogous way in Chapter 14, although we did not formulate niche-based hypotheses equivalent to Hypotheses 4-7. In Chapter 14, farm types of lower diversity were listed, that could be targeted for diversification. As discussed in that chapter, due to low percentages of variation that are explained, diversification will not be relevant for each farm of the type with the lower average diversity, although diversification will be relevant for the majority of farms of this type. Alternatively, diversification may be relevant for some farms of the types of higher average diversity, but it will not be relevant for the majority of these farms. Therefore, low percentages of variation that are explained do not completely rule out that farm typologies are used to target interventions. Moreover, some development projects may specifically target some types of farmers, most typically potentially socially disadvantaged clients such as poorer farmers or female farmers.

In addition, the methodology allows comparing the diversity of the same subsection at two different points in time, which could be useful when measuring the impact of projects that aim at diversifying a specific landscape. The methods that we presented can also easily be modified for comparisons with a predetermined desired diversity level for each subsection. Although we did not provide such hypotheses in Chapter 1, we did use some diversity threshold levels for a number of comparisons, however. For example, we used threshold level 2 for species richness, and exp(Shannon) and reciprocal Berger-Parker diversity indices in Chapter 8.

The existence of clear rankings in diversity among subsections implicitly proves that substantial diversity is already present in the landscapes. More specifically, this is the case when considering subsections of higher diversity. The fact that logarithmic measuring scales were used for the comparisons (e.g. figures 14.2-14.5) emphasises this finding. In analogy to the analysis by Myers et al. (2001) that determined global hotspots of high endemic species richness within habitats that had lost a large percentage of their original extent, we therefore provided information on tree diversity ‘hotspots’ and ‘coldspots’ in the investigated landscapes. Initially, we had not expected to record overall gamma diversities larger than 100 species. Consequently, we had expected that all species accumulation curves for uses would reach a saturation level, for example at 20 species when more than 50 farms were randomly sampled (as documented in Chapter 6, saturation levels will not be reached for random accumulation patterns when one of the component species only occurs on one farm).

The findings of high overall gamma diversity and high gamma diversity in some subsections have important consequences for the planning of diversification, as diversification is probably not relevant for ‘hotspots’. Findings of the occurrence of both ‘hotspots’ and ‘coldspots’ in a

CHAPTER 15

257

landscape indicate the potential benefit of targeting diversification efforts by focusing only on ‘coldspots’. In other words, the procedure of investigating various subsections in the landscape as performed in the various chapters where the hypotheses were tested was a procedure that yielded useful results for further planning and for other areas (see below). Prominent ‘coldspots’ and prominent ‘hotspots’ could be the subsections in which diversification experiments – to test the diversification hypothesis listed above – could be conducted, following the practice of exploiting the variation of the survey area through stratification rather than randomization. To test differences in the effects from diversification in relationship to the original diversity of systems, results that are more significant are expected for a linear relationship between the explanatory and the response variable if the experiment is conducted at low and high levels of the explanatory variable. Many sites of intermediate original diversity will not significantly contribute to statistically prove a linear relationship.

Hypotheses 2-3, 8-9, 13, 17, 24, and 26 dealt with sampling issues. More specifically, these hypotheses dealt with the question whether some data collection or analyses methods were redundant, i.e. whether too much information had been collected or whether several analysis methods lead to the same results. These hypotheses are, therefore, relevant for the planning of future surveys, under the condition, however, that diversity occurs in similar patterns in the investigated landscapes. Except for hypothesis 17, the results indicated no redundancy. For example, the results for Hypotheses 2 & 3 showed that information was necessary on the uses of each species on each farm in order to accurately compare diversity among uses in a landscape. Hypotheses 24 & 26 are examples that indicated that sample size, either measured as the number of farms sampled, the number of trees sampled, or the size of the area sampled, influenced the outcome of diversity rankings. The consequence of such findings is that comparisons in diversity should always specify which reference was used for the comparisons, for example 50 ha, 10,000 trees, or 50 samples. In the same way as deciding on how a landscape is subdivided as discussed for the first group of hypotheses, development projects will need to decide what the most relevant sample size and reference basis are for their purposes. In analogy, the consequence of rejecting Hypothesis 13 is that projects will need to choose whether uses of lower richness, or uses of lower evenness, are selected. A hybrid approach is also possible, for example to select the two subsections of the landscape of lowest richness, and the two subsections of the landscape of lowest evenness. Diversity profiles offer another type of hybrid approach, if for each profile value a threshold level is specified. For example, in figure 4.1, the profile corresponding to system A could be chosen as a threshold profile that would select each diversity profile where not all profile values are higher than the values of system A.

The distribution of species over the landscape was addressed by Hypotheses 10, 15, and 21. The results showed that species are not distributed randomly over the landscape. These results have consequences for the planning of diversification. For example, random distribution of species over farms would result in higher diversity of the average village, and potentially of a more stable ecosystem in that village. The importance of this finding is that no new species need to be introduced in the landscape, and even that no more trees need to be established. To increase species richness of villages it would suffice to change the species composition of some farms of village x with that of farms from the other villages and vice versa. In terms of the farms × species matrix, swapping rows between villages would suffice – and such was exactly the procedure used in the randomisation tests of Chapter 6 & 14. Such procedure does not change the total sum of the cells of the matrix, and ‘transfers’ of species compositions, therefore, do not alter the total number of trees in the landscape. Thus, simple random distribution of species over villages would result in their higher richness. It is even possible to obtain an average richness for villages above the average richness for random distributions of farms: the random sequences above the average provide such examples. The aggregation of conspecific trees over farms also implies that species can be distributed more randomly, which would result in higher average richness of


258

farms. However, less aggregation of species and thus fewer conspecific trees in the direct surroundings of each individual tree may have consequences for genetic erosion of the species (see below).

Possibly more complicated results were that equal-area aggregations of more farms contained more species than aggregations of fewer farms. Thus, for the same area, smaller farms aggregated more species than larger farms. This could actually be quite positive news for on-farm conservation of species, as more species would be harboured in a fragmented landscape, which is projected based on population growth estimations. Already, a positive relationship was documented between population density and the number of trees based on aerial photography (Chapter 8). Some authors (e.g. Arnold 1995) argued that very small farm sizes lead to extensification of land use and more trees, since households needed to depend on income earned from employment off-farm, so that labour became the main limited resource on farms. Now, we also documented a relationship between the degree of fragmentation and diversity. Although our results showed that species diversity was not accidental (see below), this relationship has a temporal dynamic as it is theoretically possible that farmers with small landholdings will realise later that they prefer to specialise in fewer tree species (the information that we obtained on the trees that farmers wanted to maintain on their farms was not based on a long term planning horizon). Specialisation does not necessarily imply that fewer species will be encountered at higher level in the landscape, for instance in villages, as each farmer may specialise in different species. Formulated in ecological terms: alpha diversity may decrease, but beta diversity may increase to the extent that gamma diversity becomes greater than the original level. The true relationship between fragmentation and species diversity will, therefore, only be unequivocally documented by sampling at various points in time and at various spatial scales. Our results can provide the baseline information for such comparisons.

The remaining hypotheses referred to local management of inter- and intraspecific diversity on farms. The most important hypotheses in this context are Hypotheses 1, 18, & 19. The pattern that becomes evident is that farmers prefer to have species that provide several uses on their farms, and to have several species that provide the same use. As documented above, the degree of diversity among uses is variable (we could distinguish ‘hotspots’ and ‘coldspots’), but the findings indicate that diversity is locally appreciated. Thus, introductions of new species may be relatively easily accepted by farmers if such introductions correspond to similar reasons that they have for desiring and thus maintaining and/or planning different species of the same use. The most important reason that farmers provided was that species had complementary characteristics. For example, one species provided strong poles, another species flexible branches, one species bore fruits in January, another species in July. Farmers explained that one species provided quick timber, while another species had more valuable timber that took a longer time to mature. Mixtures of medicine from various species had a higher chance of eliminating a disease when a precise diagnosis could not be made – and many more examples were provided in Chapter 11. The consequence of the ‘complementarity’ pattern is that farmers may more easily accept new species that are considered more complementary to the species that were established before. In other words, some types of ‘optimal mixtures’ could be identified. This is an interesting analogue to ecological research that documents that diversity in species’ traits is more important than species richness per se (Chapter 2). Thus, a group consisting of many species that provide a similar product or characteristic may not be as attractive as a smaller group where each species has a characteristic trait for the product or service.

Risk elimination strategies were not mentioned as frequently, but the information that several species were grown “because they all provide good products” gave indications in this direction. The pattern documented in Hypothesis 1 suggests that introductions of species that provide new uses may be especially successful. In a way, such pattern corresponds to the pattern of complementarity, but not within uses now, but among uses.

CHAPTER 15

259

In summary, farmers’ appreciation of species diversity was linked to the appreciation of the diversity in products and services that were provided by species diversity. We did not find any evidence for the contrasting scenario where farmers wanted to specialise in a single product or service, and not much evidence of farmers that wanted to specialise in a particular species within a specific use. Thus, farmers wanted to grow tree diversity. Such pattern could easily complement diversification efforts with the goal of improving the ecological functioning of agroecosystems, if species with complementarity use and ecological traits would be introduced or more widely distributed in the landscape.

Acceptance of Hypothesis 23 could have important consequences for genetic diversity of tree species on farms. However, we could not realistically test the consequences for this farmer preference, since information on reproductive ecology of the component species was scarce. Therefore, it was also not possible to test whether alternative methods of collection of germplasm (e.g. bulking all the seed produced in a village) would have an impact on genetic diversity, and our research does not allow to formulate recommendations. Collecting seed from the old tree that is being replaced and establish it on the same farm, as often practised by farmers now, will lead to inbreeding quickly in case selfing is tolerated by the species. In such situation, pollen and ovules will very frequently originate from the same tree, with a high chance for homozygosity, especially after several generations. For species that do not tolerate selfing, however, collecting from the old tree may be a practice that reduces genetic drift, i.e. the omission of genetic material from some parents in the next generation due to stochastic sampling. Therefore, information whether species are selfing or outcrossing is crucial to test the various scenarios. When genetic diversity that is present now in the landscape is already very low (for example in case the metapopulation went through a population bottleneck previously), the various sampling schemes of germplasm will also have minimal results as they are all concerned with the loss of genetic diversity. Lack of information on reproductive ecology and genetic diversity is a general problem for tropical tree species. Information on genetic diversity may be more readily available in the future if new devices (potentially a future generation of “lab-on-a-disc” technology as described by Meldrum 2000) would be developed that allowed assessing genetic diversity directly in the field rather than in a molecular laboratory. Our software will then allow making predictions that are more accurate of how genetic diversity would change over time for each species.

14.2. Some New Tools for the Analysis of Diversity of a Landscape

To test some of the hypotheses listed above, some new tools were developed. These tools could benefit diversity research in other agroecosystems, or even natural ecosystems. Although the aim of the investigations in this thesis was not to develop new tools but to describe patterns of diversity in the landscapes where the surveys were conducted, these tools can be listed as results from the conducted research and are therefore listed here.

The fact that more species will be recorded when a larger area and more individuals are sampled has been observed for several decades – Arrhenius for example published his model for the relationship between area and species richness in 1921. As described in Chapter 6 (and documented in a recent review article by Gotelli & Colwell 2001), it was thought that randomized species accumulation curves for samples could only be calculated by randomization procedures that calculated the average species richness from 500 or more random sequences of samples. The formula based on the hypergeometric distribution that was introduced in Chapter 4 is novel as randomization procedures are not necessary to obtain the average species richness for a specific number of randomly selected samples. However, such randomization procedures have their value in determining the range in species richness values that is obtained for a specified sample size.


260

Average species richness is only influenced by the number of sites on which each species occurs, whereas the range in species richness for particular sequences is an indication of aggregation of species within particular sites. We thus disagree with Gotelli & Colwell (2001) who stated that “Because the sample-based rarefaction curve depends on the spatial distribution of individuals as well as the size and placement of samples, it cannot be derived theoretically. (...) Thus, computations require Monte Carlo resampling in which samples are randomly accumulated in many iterations”. A wider use of the formula of Chapter 6 can, therefore, benefit future ecological investigations which are concerned with the relationship between sample size and species richness, be it for natural or anthropogenic ecosystems, and thus beyond purposes of planning for agroecosystem diversification as provided in this thesis. Software was developed that calculates species accumulation from sites × species matrices using the new formula (Kindt 2001a).

The average richness of villages was compared with the richness of ‘artificial villages’ in Chapter 6. These artificial villages were constructed by randomly selecting the farms for each village. These types of randomizations had not been performed earlier. The procedure introduced here tested whether the average richness of villages could have been caused by a random distribution of farms. The procedure is an example of null model analysis as described by Gotelli (2001), defined as investigations whether a random process could result in the same pattern as observed for the real data. What we actually investigated was whether compositional differences among villages resulted in lower or higher richness than expected by random chance. In contrast, the ordination analysis of Chapter 9 & 10 investigated whether compositional differences existed among villages, but did not provide any insights whether the average richness of villages was low or high. This procedure could be used for other ecosystem studies that deal with the question whether spatial distribution patterns resulted in high or low average richness of spatial clusters. Such procedure actually provides an approach that resembles the analysis of species aggregation (Chapter 12), the difference being that not the aggregation of trees of the same species is being investigated this time, but the aggregation of species richness. Software was developed that calculated average species richness for random clusters of sites from sites × species matrices (Kindt 2001h).

In Chapter 7, we derived evenness profiles that provided a measure only of the evenness of each system. Although the derivation of evenness profiles from diversity profiles was mathematically easy, this procedure had not been presented earlier. Hayek & Buzas (1997 p. 379) claimed that their decomposition of the Shannon index into richness and evenness provided researchers for the first time with a simple way to examine evenness separately from richness. We showed in Chapter 7 that their approach was not complete in characterising evenness, and therefore did not allow to rank systems in terms of evenness (i.e. evenness profiles may intersect).

The calculation of accumulation patterns for the Rényi diversity profile was a new procedure. Combined with the decomposition of the diversity profile in statistics that either describe richness or evenness, this procedure allowed for discrimination between the contribution of sample size, richness, and evenness to the diversity of a system. Various measures of equal sample sizes of different systems can be used, either the same number of sample sites, the same area of the sample sites, or the same number of individuals counted. This procedure allows for meaningful comparisons of diversity, richness, and evenness among surveys that differ in sample sizes, by eliminating the effects of differences in sample size on the comparisons. Given the fact that the average density of individuals per sample usually differs among surveys, using the same number of individuals or the same area to base comparisons on are expected to yield different results for the majority of surveys. Thus, our approach would be suitable for most surveys to investigate whether the sampling base of individuals or area would yield similar results. These sampling issues had been described earlier for comparisons of species richness among surveys (e.g. Chazdon et al. 1999; Gotelli & Colwell 2001). Alternatively, the study of the pattern within a

CHAPTER 15

261

single survey of the influence of sample size on species richness has obtained a lot of past and recent attention (Chapter 6). Our approach allows expanding the analyses for diversity and evenness, either to compare results of different surveys at the same sample size, or to investigate the relationship between sample size and diversity. It is, therefore, likely that other biodiversity studies would benefit from the approach.

In addition, we documented in Chapter 7 that the pattern of accumulation of the Rényi profile provides information on the most likely species abundance distribution model for the systems studied. Since it is often difficult to select the abundance distribution model with best fit to the observed data, our approach may offer an alternative method of selecting the best model. Various programmes were written that calculate accumulation patterns for the Reyni series and Rényi evenness profile from a sites × species. Programmes differ in the sample base that is used for comparisons (Kindt 2001d-g).

A feature that we did not use in this thesis, but that was included in the software that was developed to calculate accumulation patterns for the Rényi profile, was that the software ‘remembered’ the random sequence of sites that yielded the highest and the lowest profile values. As documented in Chapter 7, diversity and evenness profiles are two-dimensional diagrams that represent the profile value on the vertical axis against the scale parameter value on the horizontal axis. It is necessary to construct a two-dimensional diagramme, as a single diversity statistic will not suffice to provide all information on the diversity of as system. The software can, therefore, be used to make a selection of sites that could be shown to contain relatively high diversity. To select the sites, the decision will need to be made at which scale of the Rényi diversity or evenness profile maximal values have to be obtained, and which reference basis is used (see the previous paragraph). Simulations showed that selections of sites that had the highest richness were not necessarily the same sites that had the highest diversity at other scales. Such procedure could be used to select a subset of sites that would maximise the diversity contained in them, which could be very useful for the planning of biodiversity conservation. Similar software are for example WORLDMAP, documented in Balmford et al. (2001), or CPLEX, described in Rodrigues & Gaston (2001), which are based on selection of sites with the highest number of endemic species.

Although we did not obtain significant differences for the scenarios that we investigated, the metapopulation model that we developed could be also useful for some other analyses. The model allows calculating genetic diversity parameters for metapopulations consisting of individuals of varying ages with geneflow within and among different generations and subpopulations. For each individual tree, parameters describe how germplasm (either seed, wildlings, or cuttings) and pollen is sampled from the entire genepool when a new individual is established at the same location. Other parameters describe the maximum age, the chance of producing flowers in a given season, and the chances that the individual is male or female for a dioecious species. The model therefore allows for a large range of simulations of genetic diversity dynamics, and to our knowledge, such software was previously not available. Potential improvements of the model that would result in simulations that are more realistic were discussed in Chapter 12, but these would require additional parameters, which will have to be calibrated in the field.

14.3. Recommendations for Future Research and Development Activities

A first recommendation that we want to make is that researchers have to be aware that the concept and the measurement of diversity can be dissociated into richness and evenness, and that measurement of diversity is a tricky endeavour as its value depends on sample size.


262

One basic problem that we encountered is that species identity can not always be established in the field, especially when species do not have reproductive organs (flowers or fruits). We therefore recommend the practice of collecting herbarium specimens, to become familiarised with the flora of a target area before starting any data collection on farmers’ fields, and to include a botanist in the data collection team. A good checklist of vernacular names versus botanical names may also be very useful in the field. Researchers should be aware, however, that the same local name may be used for many botanical names (or vice versa), that farmers may not know all the vernacular names of the species on their farms (a problem which we often encountered in western Kenya), or that some farmers could misidentify species on their farm.

The ‘toolbox’ that was developed (Chapter 4, section 15.2) could be used in other areas as it will provide a more accurate description of their diversity. The methods allow measuring the impact of projects by comparing diversity before and after interventions. In addition, as documented in section 15.1, the methodology may be used to select subsections of lower diversity in an area that can be targeted by diversification initiatives.

As previous recommendations relate more to starting research in new areas where exploratory research may be especially useful, the following recommendations relate to future research in the same areas where the analyses for this thesis were conducted.

As documented in section 15.1, the research documented in this thesis was exploratory research that allows a next stage which could be more experimental that could test the hypothesis that ‘diversification of a section of lower diversity of an ecosystem will positively affect agroecosystem productivity and stability to a higher degree than the same degree of diversification of a section of higher diversity’ . Information on hotspots and coldspots in the systems allows for a stratified set-up of experiments that would be the most efficient in testing the hypothesis. By testing this hypothesis, one will implicitly test the hypothesis that diversity leads to enhanced agroecosystem productivity and stability.

A related research and practical development problem to enhancing diversity relates to the species composition of more diversified system i.e. the identities of the species that diversify the system. As documented in this thesis, these could either be species that already occur in the landscape, or be species that are currently absent in the landscape. Below, the discussion will be focused more on interventions that could provide some information on the identities of these species. As documented in Section 15.1, both from an ecological or ethnobotanical perspective, complementarity with the existing species’ pool is the most important feature to consider when selecting additional species. Put more practically, adding a new species to a mixture will be meaningless if this species has exactly the same ecological or ethnobotanical features as species that are already present. Ecological models (Chapter 2) indicate that particular mixtures of species of higher diversity may result in lower productivity and stability of the ecosystem than mixtures of lower diversity that offer a wider range in species’ characteristics.

No farmer was using all the species that were encountered in their landscape. Not every farmer that had the same species on his or her farm was using this particular species for exactly the same purposes. We could not establish whether these differences resulted from redundancy for a specific use (as other species were better on farms where the particular species was not used for that purpose), missing information in our dataset (maybe a farmer forgot to mention the particular purpose for the particular species – a long-term survey that would attempt to monitor how farmers use species could yield better results, but would be very time-intensive), or from differences in lack of knowledge. We did find occasional evidence for knowledge limitations as some farmers mentioned that their father was using the species but that they did not know for what purpose. One farmer indicated that neighbours collected bark from a tree on his farm for medicinal purposes, but then these neighbours were refusing to communicate information on the specific ailment and treatment. To establish current levels of knowledge, each farmer could be interviewed about the uses of species that do not occur on his or her farm, to check the

CHAPTER 15

263

relationship between the fact that they do not establish a particular species and lacking knowledge on uses of that species. Research projects could then test whether wider dissemination of information on potential uses of species would result in wider use of species that are already present on farms and wider distribution of species in the landscape (i.e. the establishment of new trees, but not new species in the landscape).

A strong recommendation to make is that future diversification initiatives that distribute species should ensure that at the same time information on their potential uses is distributed. At the same time, since farmers may not know the local name or the botanical name of each species, diversification initiatives should provide the correct names for each species and maybe some information how a species can be recognized. Future research could test the hypothesis that lack of knowledge limits species diversity on-farms by measuring the consequences of spreading brochures with information on the uses of species, either directly to farmers, to key farmers, or to extension officers. Mass media could feature certain species. Another approach that ICRAF is currently exploring is introducing agroforestry as a topic into the curricula of primary schools, and such approach could provide information on uses of specific species to the “farmers of the future”. Alternative pathways to spread information on species’ uses more widely could be to organise visits of farmers to forests, arboreta, or tree nurseries and present the uses of the encountered species. Other methods could be to arrange for farmer exchange visits where farmers that are local tree specialists would explain about the uses of species on their farm to other farmers, or to arrive at a village-level documentation system where farmers deposit information on uses of species.

A participatory mapping exercise for a village where each species is mapped and information is provided on the uses may also serve the purpose of distribution of knowledge about uses of tree species more widely. Mapping the positions of individual trees could also prove useful to ensure that germplasm is collected from many trees so that effective population size would be large enough to prevent genetic erosion (see Chapter 12). Information on the locations of conspecific trees could especially be useful for farmers that set up a tree nursery, so that they can guarantee that the trees that they produce have a wide genetic base, especially when the species is maintained at low densities (see below).

Future diversification initiatives could borrow some of the methodologies that have been developed for participatory assessment of biodiversity. Hellier et al. (1999) for example used semi-structured interviews, transect walks, participatory mapping, analysis of maps and aerial photographs. People’s Biodiversity Registers (Thomas et al. n.d.) provide another example. Other participatory methods are listed by Lawrence & Ambrose-Oji (2001), whereas the strengths and weaknesses of these methods were discussed during an European Tropical Forest Research Network internet workshop convened by the Environmental Change Institute, University of Oxford, 7 - 25 January 2002 (http://www.etfrn.org/etfrn/workshop/biodiversity/index.html). Lawrence & Ambrose-Oji (1999) showed that there was a need for approaches that explore local non-utilitarian values of biodiversity. Similarly, information could be assembled and distributed to farmers on the non-use values of trees species (see also Nunes et al. 2001), although collection and dissemination of such data may be more complicated.

It may be necessary to safeguard information to ensure that intellectual property rights of local people are protected. Safeguarding intellectual property rights does not necessarily mean that the private sector is not given access to the information. The medicinal and seed industries are, for example, provided access to the People’s Biodiversity Registers a fair cost (Lawrence & Ambrose-Oji 2001). Care should also be taken that wider knowledge on uses of species would not lead to the reverse effect of over-exploitation of trees in the landscape, rather than increased tree diversity. In addition, information on potential uses of species might have spill-over effects to natural ecosystems that could suffer higher exploitation in consequence. Information on uses of species that also occur in natural environments could, however, also lead to more appreciation


264

of the value of these environments, either in terms of their direct use or in terms of their indirect use (e.g. as source of genetic resources, see below).

A future research project could look at the reasons why some farmers are not willing to communicate about the uses of a species, and whether this is related to fears that the trees would be over-harvested or some produce would be stolen by other farmers. Some evidence for distrust among farmers in western Kenya was documented as some farmers mentioned that they preferred to cultivate medicinal trees in the homestead area for fear of theft. Some farmers even mentioned that the purpose of some species was “to scare away snakes and thieves”, but we did not establish whether these fears were directed to tree products. Another reason why farmers may not be willing to share information with other farmers would be that they are afraid of loosing market share for their tree products. In case farmers fear to be disadvantaged by sharing information with other farmers, then diversification initiatives should address these fears before information is spread more widely on the uses and the locations of trees. Development initiatives should probably adopt an adaptive management process to test assumptions, adapt and learn from the results, such as the “learning portfolios” developed by the Biodiversity Conservation Network (Salafsky et al. 2001).

Lawrence & Ambrose-Oji (1999) pointed out that there were methodological gaps in obtaining information on local values along the genes-species-habitats-landscape continuum. The discussion of the local value of trees and tree diversity at higher spatial scales is related to methodological issues with measuring use and non-use values (see above). It also touches on the fact that the final users of tree products (or services) are not always the farmers that grow the trees. Farmers may not always be aware of the range of species that other users appreciate, and this could limit the number of species that they grow. On the other side, off-farm users may not be aware of all the species that are available from farmers’ fields. Wider dissemination at the demand side of potential uses of species could thus also result in promoting the cultivation of more trees and more species. Diversification initiatives that would herald expectations of other users could also embark in participatory marketing research, as for example initiated by ICRAF in central Kenya. Farmers are not always aware of the qualities and quantities that are required by the market, and often lack market intelligence and organization to enable them to negotiate better prices. One example that was documented in central Kenya was that most farmers were not aware of the standard length of poles of Grevillea robusta trees required by timber mills. Some farmers could therefore have doubled their income in some instances simply by delivering a pole that was 0.5 m longer – as then two standard-sized poles could have been obtained at the timber mill. Eco-labelling schemes may open global markets for farmers. For these markets, information on potential customers and potential suppliers of products could also result in extended cultivation of tree diversity on farms.

Examples that the final users of services may are not exclusively the farmers are C-sequestration to mitigate global climatic change, air, and water purification in peri-urban areas, provision of pollination to crop plantations, or biodiversity conservation. Databases on tree abundance and diversity could be maintained at village and aggregated at higher administrative levels to obtain funding for C-sequestration (Simons, unpublished results). Given the current global biodiversity crisis (Chapter 2), farmers could be rewarded for conserving biodiversity on their farms. Interestingly in this context, the Biodiversity Conservation Network that studied the hypothesis that people will conserve natural habitats if they can benefit financially from community-based enterprises that depended on them, found that the 37 enterprises that they studied only lead to conservation under a limited set of conditions. There was a weak association between enterprise success and conservation success, but a strong association between local involvement in the enterprise and conservation success. Salafsky et al. (2001) therefore concluded that by establishing an enterprise, project staff members of conservation organisations gain access to a community, so that community members may be more receptive to conservation ideas, which would have the

CHAPTER 15

265

largest influence on conservation success. Enterprises that are not linked to biodiversity that are easier to be implemented and more profitable may actually be more effective. On the other hand, McNeely & Scherr (2001) provided examples of innovative landscape management strategies that successfully combined objectives of increasing local food security and biodiversity conservation by applying ecoagriculture strategies (see Chapter 14). The challenge indeed is whether and how objectives of farmers and other stakeholders can be combined.

As mentioned above, complementarity among species may lead both towards better ecological functioning of agroecosystems and towards increased farmer interest in growing additional species on the farm. In addition, species that vary in their responses to environmental heterogeneity lead to increased productivity and stability under environmental fluctuations. Recent research on the ecological role of diversity in grasslands revealed that new tools are needed to investigate what the role of particular species is in a particular mixture (see section 15.1). Therefore, future diversification initiatives should be aware of the evolutions in these techniques. Agroforesty systems have the disadvantage that they often do not allow for short-term field experiments. Testing systems in a model environment offers a solution to this problem, and this was the reason that models such as WaNuLCAS (water, nutrient and light capture in agroforestry systems) (Van Noordwijk & Lusiana 1999) were developed. Such models, which could be combined with the methodology of Loreau & Hector (2001) to study effects on productivity or that proposed by Cottingham et al. (2001) to study effects on stability, could shed light on the traits of species that would result in the largest compatibility of species. Analysis of such multispecies models that need many equations could potentially be simplified by “mean-field approximations” (e.g. McKane et al. 2000) or “moment approximation methods” (e.g. Norberg et al. 2001). Field observations and, if means permit, experiments could investigate the relationship between environmental changes and responses of individual species, so that species could be classified in terms of their ‘environmental niche’. Ordination techniques as demonstrated in Chapters 9 & 10 could be used for the statistical analysis. Species could then be introduced in landscapes where their environmental niche is left unoccupied. However, it has to be remarked here that not all future environmental changes can be predicted. Some researchers mention the unpredictability of global change as a reason to maintain as much biodiversity as possible as no single species can be classified as being absolutely redundant (see Chapter 2).

At present, no information is available to be disseminated to farmers on the performance of each potential species under each type of environmental condition or in each possible mixture. It can thus not always be established which species would be the best ‘compatible partners’ for the trees that are already present on farms. In addition, in conditions where farmers are using a particular species for a specific purpose, their management of this species may not the optimal method to result in the best possible product or service of the species. For example, Van Mele (2000) documented that pest management of fruit trees by farmers was not optimal in Vietnam. Another example is that pruning schemes of trees used for timber is not optimal in Kenya and Uganda (Lengkeek pers. comm.). For this reason, tree management was listed as one of the iterative steps of tree domestication (see Chapter 1). Thus, diversification initiatives may embark on participatory testing of new species with the objective of both increasing the general knowledge on cultivation of the species, and to increase local knowledge on uses of a species.

Future diversification initiatives should also ensure that species are tested over their intraspecific range of traits. Therefore, evaluations should not be based on single trees but preferentially on conspecific trees from different populations. An individual tree occupies a larger section of the spatio-temporal space that is available on farms than individual crop species. Landscapes also become more fragmented with time. Therefore, future research could follow a scheme in which not every tree species is tested on each farm, but where each farmer tests a subset of species. Information exchange mechanisms could be set up at the village level by which farmers would communicate about the performance of the species on their farms, and learn about the


266

performance of other species on other farms. Such scheme would ensure that information on species’ uses is more widely distributed. Such scheme could complement initiatives, as documented above, of sharing information about uses (including management aspects) of species that are already present.

Initiatives that experiment with new species in a village should make it clear to farmers that they are research projects and not only development projects. Some newly introduced species may perform poorly, and this should be explained to farmers that avoid that their expectations of success are too high which could eventually lead to failure of the entire initiative. Some farmers are already experimenting with new species (especially timber species that take a long time to mature, see Chapter 11). Therefore, future research on the performance of species could explain to complement on-farm testing that was already going on. To arrive at a synthesis of scientific ( i.e. experimentally and statistically verified) and farmers’ knowledge, a differentiation could be made between knowledge aspects that are easily observed in the field, and knowledge aspects that are more difficult to observe (Van Mele 2000 pp. 41-43). Farmer training could focus on gaps in farmers’ knowledge or misperceptions related to topics that are difficult to observe. For instance to improve the management of citrus mites, which are very small and highly mobile creatures, farmers require a better knowledge on the diagnostics and ecology of this pest. This in turn requires improving their observation skills using hand lenses. Visualization is an essential step in the learning process (Van Mele 2000 pp. 135-136). Ecological complementarity among species could be another example of an aspect that is more difficult to observe, especially when considering the experimental and analytical problems of such research that were mentioned above. “Ethnobotanical complementarity”, where farmers appreciate species diversity for the range among and within products that are provided, could be investigated more easily through participatory research with greater involvement in measurement and evaluation by farmers.

The research on new species, or new uses of species already present in the landscape, should be conducted within the context of providing farmers with more options than they currently perceive or which are currently available to them. In a risk-limitation strategy, increased awareness of alternatives will probably lead to increasing species diversity. Information exchange on the uses of species that are already present, and information generation and dissemination on the uses of new species, should include a component of the complementarity of species to show how mixtures of species could benefit the farmer. However, one cannot always predict the consequences of research or development activities. As discussed in Chapter 11, the farm household will be the only party that can effectively conduct whole-farm planning, as households are the only units that can integrate all strategic, practical, and technical elements of the entire farm.

In the jungle-rubber agroforests of Sumatra, rubber is grown in association with many indigenous tree species that are tolerated by the farmers. The idea was to increase the profitability of this system and, therefore, the sustainability of biodiversity conservation in this system, by introducing clonal rubber varieties with superior production. However, when obtaining clonal rubber varieties, farmers preferred to modify the jungle rubber systems into monoculture rubber plantations. This showed that where one species is much more profitable than the other species, diversification might not be a realistic option for farmers, even when such less diverse system could be much riskier to fail to produce in the event of disease, pest, or market problems for the particular species. Researchers should, therefore, be aware that domestication of a species could lead to simplification of farming systems. Mere expectations of farmers that a domesticated species will be much more profitable could already result in alterations in planning by farmers. Information on jungle-rubber agroforests was discussed during a Workshop on farmer knowledge and management of complex agroforests: can profitability and biodiversity conservation be combined? Is there a need for 'environmental service benefit transfer' mechanisms? 3rd-6th September 2001, Muara Bungo, Sumatra.

CHAPTER 15

267

Besides investigating whether landscape-level diversification of interspecific diversity will lead to improvements in the functioning of the agroecosystem, diversification initiatives may also look at intraspecific diversity at the landscape level. The study of genetic diversity may be one area where knowledge aspects are not so easily observed by farmers, and where scientific research could provide a synergy with local generation of knowledge (see above). As discussed in Section 15.1, we were not able to make realistic investigations of the dynamics of genetic diversity since information was lacking on the genetic diversity that occurs at present in tree populations, and on the reproductive ecology of these species. Future projects may, therefore, embark on molecular analysis and characterisation of reproductive ecology. Programmes such as the trees-in- metapopulations simulation software developed for this thesis (Section 15.2) could then more realistically predict levels of genetic erosion. As indicated in Chapter 14, these simulations may indicate that in fragmented landscapes, farmers should agree on cultivating the same tree species on their farms, as this could be the only way in which sustainable populations are maintained on their farms. Such agreements would not necessarily mean that fewer species would occur in the landscape, as different farmers could concentrate on different species. Similarly, such investigations could indicate that some species could only be conserved within a matrix of remnant forests and agroforestry systems, as forests or agroforestry systems alone would not be sufficient. The latter investigations would indicate the value of forests as sources of germplasm or of pollination vectors, which could be used as an argument of their preservation. Obviously, to allow these investigations, some samples will have to be taken in forests.

When tree populations exist both within forests and outside of forests, investigation of geneflow between such populations deserves special attention. The study of Aldrich & Hamrick (1998), for example, demonstrated that trees in pastureland resulted in a genetic bottleneck (a faster decline in genetic diversity than expected from population size) for trees in forests since pollinators preferred to collect fruits from trees in pastureland, while preferring shelter in remnant forests. What could be investigated also could be the consequence of harvesting individual trees of superior growth first. Such practice could result in negative selection for growth in case more germplasm is obtained from the remnant trees of slower growth. Where species occur both on farmland and in forests, the effects of selective harvesting on genetic diversity should be investigated for the entire landscape, together with the consequences of various germplasm collection methods, to be able to formulate recommendations on optimal collection methods that minimise genetic erosion. Potentially, the advice could be given to farmers to obtain seed for a new tree generation from natural populations only, or to only establish seed from trees of superior characteristics.

Studies of genetic diversity could indicate whether conservation-through-use is a realistic option for particular species. Given the fact that many tree species have recalcitrant seeds i.e. seeds that do not maintain their viability when stored (a problem for diversification initiatives that wish to distribute seed of these species as well), these species are not appropriate for conservation ex situ. In addition, ex situ conservation is also difficult for tree species with orthodox seed that can be stored in a genebank, as rejuvenation of genebank accessions requires much more land and much more time than the rejuvenation for accessions of crop species given the differences in spatio-temporal space required individuals between tree or crop species. Pollination may be a problem during the rejuvenation of tree accessions. For long term conservation with the purpose of reintroduction of the species in habitats where the species became extinct, associated species such as N2-fixing microsymbionts, mycorrhizal species, pollination vectors and seed dispersal vectors, should be conserved as well. Bush & Rivera (2001), for example, state that neotropical rain forest floras are primarily zoophilous, and that zoophily is often a product of coevolution of the plant and its pollinator.

Brown (2000) lists ten advantages of in situ crop conservation, including continuing evolutionary processes, greater allelic richness, and conservation linked with use. Swanson & Goeschl (2000)


268

describe the main advantage of in situ conservation in generating new information on better performing varieties because the best varieties at a given point in time will depreciate over time, while ex situ techniques allow for the best use of existing information. It is possible that tree populations will undergo evolutionary changes because different conditions exist in agroecosystems as opposed to forest ecosystems, in which case conservation on farms will not conserve the same genetic diversity as in situ conservation. Since evolutionary forces may be different in anthropogenic ecosystems than in natural ecosystems, Simons (in prep.) differentiated between in situ (in natural ecosystems, particularly forest ecosystems), ex situ (particularly long-term storage as seed in a genebank) and circa situ conservation (on farms in agroecosystems juxtaposed to the remaining natural habitats of the species). Where forests are under serious threat, conservation-through-use in agroecosystems may offer the only practical conservation method, even if some evolutionary forces are different in these systems. As other evolutionary forces may be similar in adjacent forest and agroecosystems, some proportion of local genetic adaptation could be conserved circa situ. In case an initiative would target wider establishments of local species, then the use of germplasm from locally-adapted rather than remote populations is recommended (although some genetic material from remote populations could have superior production and be suited to local conditions, but range-wide collection and evaluation as conducted by ICRAF for priority species (Chapter 1) are needed to evaluate various populations).

Investigations of geneflow in between agroecosystems and forest ecosystems, as described above, could show that simultaneous conservation in agroecosystems and forest ecosystems is necessary to prevent extinction of a tree species in the fragmented landscape. Such investigations could be especially important for species with higher levels of endemism – species that only occur in a specific (fragmented) forest ecosystem and possibly in the adjacent agroecosystem. As we demonstrated in this thesis, farmers may be interested to maintain many species on their farms. Given the small population sizes of many species (Chapter 12 & 14), these current population sizes may not ensure that the species can be sustained. Medicinal species, where one tree could produce enough medicine for several farmers, could be an example of species that would only be maintained in small densities in the landscape (Lengkeek, unpublished results). Therefore, species that are only maintained in small abundance may require continuous inputs of new genes to prevent their extinction in an agroecosystem. Similar to the strict definition of domestication where human interventions are needed for a species to reproduce, such situations could be defined as ‘strict tree domestication of the landscape’.

Research projects could investigate whether and to what extent human interventions are needed to allow small tree populations to survive in an agroecosystem. Possibly, such investigations could indicate an analogy with community-conservation of national parks, where some scientists argue that “local peoples should not carry the sole responsibility for the political viability of protected areas” (Chicchón 2000). Other stakeholders could be approached for financial or technical support for such continuous injections of new germplasm, if this would be the only mode of survival of some species in fragmented landscapes. Also similar to the discussion of conservation of natural habitats would be the statement that “care must be exercised whenever research or monitoring activities are promoted at the possible expense of important conservation actions” since “measuring is not protecting” and “much of the current emphasis is on watching problems proceed rather than trying to halt them” (Sheil 2001).

In summary, we documented in this thesis that wider distribution of tree species within tropical agroecosystems would lead to increased on-farm diversity. Future research could focus on the identity of tree species to feature in diversification initiatives, whether current population levels allow sustaining particular tree species in agroecosystems, whether conservation-through-use is relevant for biodiversity conservation at the global level, whether diversification would correspond to reasons for local or higher-level management of tree species diversity, and whether diversity would lead to higher productivity and stability of these ecosystems.

LITERATURE LIST

LITERATURE LIST

270

Aldrich PR & Hamrick JL (1998) Reproductive dominance of pasture trees in a fragmented tropical forest mosaic. Science 281: 103-105.

Alvarez-Buylla ER, Garcia-Barrios R, Lara-Moreno C & Martinez-Ramos M (1996) Demographic and genetic models in conservation biology: applications and perspectives for tropical rain forest tree species. Annual Reviews of Ecology and Systematics 27: 387-421.

Anderson MJ & Legendre P (1999) An empirical comparison of permutation methods for tests of partial regression coefficients in a linear model. Journal of Statistical Computation and Simulation 62: 271-303.

Arnold JEM & Dewees PA eds. (1995) Tree management in farmer strategies: responses to agricultural intensification. New York: Oxford University Press, 292 pp.

Arnold JEM (1995) Retrospect and prospect, pp. 271-287. In: Arnold JEM & Dewees PA eds. Tree management in farmer strategies: responses to agricultural intensification. New York: Oxford University Press.

Arrhenius O (1921) Species and area. Journal of Ecology 9: 95-99.

Austerlitz F, Mariette S, Machon N, Gouyon P-H & Godelle B (2000) Effects of colonization processes on genetic diversity: differences between annual plants and tree species. Genetics 154: 1309-1321.

Ayensu E, van Claasen DR, Collins M, Dearing A, Fresco L, Gadgil M, Gitay H, Glaser G, Juma C, Krebs J, Lenton R, Lubchenco J, McNeely JA, Mooney HA, Pinstrup-Andersen P, Ramos M, Raven P, Reid WV, Samper C, Sarukhán J, Schei P, Tundisi JG, Watson RT, Guanhua X & Zakri AH (1999) International ecosystem assessment. Science 286: 685-686.

Balmford A, Moore JL, Brooks T, Burgess N, Hansen LA, Williams P & Rahbek C (2001) Conservation Conflicts Across Africa. Science 291: 2616-2619.

Baradat P, Adams WT & Müller-Starck G (1995) Population genetics and genetic conservation of forest trees. Amsterdam: SPB Academic Publishing.

Bedigian D & Harlan JR (1983) Nuba agriculture and ethnobotany, with particular reference to sesame and sorghum. Economic Botany 37(4): 384-395.

Beentje HJ (1994) Kenya trees, shrubs and lianas. Nairobi: National Museums of Kenya, 722 pp.

Berendse F (1994) Ecosystem stability, competition and nutrient cycling, pp. 409-431. In: Schulze E-D & Mooney HA eds. Biodiversity and ecosystem function. Berlin: Springer-Verlag, 525 pp.

Bockstael N, Costanza R, Strand I, Boynton W, Bell K & Wainger L (1995) Ecological economic modeling and valuation of ecosystems. Ecological Economics 14: 143-159.

Bolker BM, Pacala SW, Bazzaz FA, Canham CD & Levin A (1995) Species diversity and carbon dioxide fertilization: conclusions from a temperate forest model. Global Change Biology 1: 373-381.

Borcard D, Legendre P & Drapeau P (1992) Partialling out the spatial component of ecological variation. Ecology 73(3): 1045-1055.

Boshier DH (2000) Mating systems, pp. 63-79. In: Young A Boshier D & Boyle T eds. Forest Conservation Genetics. Collingwood: CSIRO Publishing & Wallingford: CABI Publishing.

Boster JS (1985) Selection for perceptual distinctiveness: evidence from Aguaruna cultivars of Manihot esculenta. Economic Botany 39(3): 310-325.

Bradley PN (1991) Woodfuel, women and woodlots, volume 1. London and Basingstoke: Macmillan Education Ltd, 338 pp.

Bradley PN, Chavangi N & Van Gelder A (1985) Development research and energy planning in Kenya. Ambio 14 (4-5): 228-236.

LITERATURE LIST

271

Brodie AW, Labarta-Chavarri RA & Weber JC (1997) Tree germplasm management and use on-farm in the Peruvian Amazon: a case study from the Ucayali region, Peru. London: Overseas Development Institute & Nairobi: ICRAF, 69 pp.

Brown AHD (2000) The genetic structure of crop landraces and the challenge to conserve them in situ on farms, pp. 29-48. In: Brush SB ed. Genes in the field: on-farm conservation of crop diversity. Boca Raton: IPGRI, IDRC and Lewis Publishers.

Bush MB & Rivera R (2001) Reproductive ecology and pollen representation among neotropical trees. Global Ecology and Biogeography 10: 359-367.

Byers DL & Waller DM (1999) Do plant populations purge their genetic load? Effects of population size and mating history on inbreeding depression. Annual Reviews of Ecology and Systematics 30: 479-513.

Caballero A (1994) Developments in the prediction of effective population size. Heredity 73: 657-679.

Cain ML, Milligan BG & Strand AE (2000) Long-distance seed dispersal in plant populations. American Journal of Botany 87(9): 1217-1227.

Cardinale BJ, Palmer MA & Collins SL (2002) Species diversity enhances ecosystem functioning through interspecific facilitation. Nature 415: 426-429.

Carter SE & Crowley EL (forthcoming) Land use change in western Kenya, 1900-1990. In: Lynam J ed. Diversity and dynamics: understanding spatial and temporal variation in African agricultural systems.

Chapin FS III, Sala OE, Burke IC, Grime JP, Hooper DU, Lauenroth WK, Lombard A, Mooney HA, Mosier AR, Naeem S, Pacala SW, Roy J, Steffen WL & Tilman D (1998) Ecosystem consequences of changing biodiversity: experimental evidence and a research agenda for the future. Bioscience 48: 45-52.

Chapin FS III, Walker BH, Hobbs RJ, Hooper DU, Lawton JH, Sala OE & Tilman D (1997) Biotic control over the functioning of ecosystems. Science 277: 500-504.

Chapin FS III, Zavaleta ES, Eviner VT, Naylor RL, Vitousek PM, Reynolds HL, Hooper DU, Lavorel S, Sala OE, Hobbie SE, Mack MC & Diaz S (2000) Consequences of changing biodiversity. Nature 405: 234-242.

Chase MR, Moller C, Kesseli R & Bawa KS (1996) Distant gene flow in tropical trees. Nature 383: 398-399.

Chazdon RL, Colwell RK, Denslow JS, Kobe RK & Hubbell SP (1999) Tropical tree richness and resource-based niches. Science 285: 1459.

Chicchón A (2000) Conservation theory meets practice. Conservation Biology 14(5): 1368-1369.

Clark JS, Beckage B, Camill P, Cleveland B, HilleRisLambers J, Lichter J McLachlan J, Mohan J & Wyckoff P (1999) Interpreting recruitment limitation in forests. American Journal of Botany 86(1): 1-16.

Clawson D (1985) Harvest security and intraspecific diversity in traditional tropical agriculture. Economic Botany 39(1): 56-67.

Cleveland DA (1993) Is variety more than the spice of life? Diversity, stability and sustainable agriculture. Culture and Agriculture Bulletin 45-46: 2-7.

Colasanti RL, Hunt R & Askew AP (2001) A self-assembling model of resource dynamics and plant growth incorporating plant functional types. Functional Ecology 15: 676-687.

Colfer CJP (1991) Indigenous rice production and the subtleties of culture change: an example from Borneo. Agriculture and Human Values 8 (1&2): 67-84.

Collins SL, Glenn SM & Gibson DJ (1995) Experimental analysis of intermediate disturbance and initial floristic composition: decoupling cause and effect. Ecology 76(4): 486-492.

LITERATURE LIST

272

Colwell RK (1997) EstimateS: Statistical estimation of species richness and shared species from samples. Version 5. User’s Guide and application. URL http://viceroy.eeb.uconn.edu/estimates.

Condit R, Ashton PS, Baker P, Bunyavejchewin S, Gunatilleke S, Gunatilleke N, Hubbell SP, Foster RB, Itoh A, LaFrankie JV, Lee HS, Losos E, Manokaran N, Sukumar R & Yamakura T (2000) Spatial patterns in the distribution of tropical tree species. Science 288: 1414-1418.

Condit R, Hubbell SP, Lafrankie JV, Sukumar R, Manokaran N, Foster RB & Ashton PS (1996) Species-area and species-individual relationships for tropical trees: a comparison of three 50-ha plots. Journal of ecology 84: 549-562.

Cordeiro NJ & Howe HF (2001) Low recruitment of trees dispersed by animals in African forest fragments. Conservation Biology 15(6): 1733-1741.

Costanza R, d’Arge R, de Groot R, Farber S, Grasso M, Hannon B, Limburg K, Naeem S, O’Neill RV, Paruelo J, Raskin RG, Sutton P & van den Belt M (1997) The value of the world’s ecosystem services and natural capital. Nature 387: 253-260.

Cottingham KL, Brown BL & Lennon JT (2001) Biodiversity may regulate the temporal variability of ecological systems. Ecology Letters 4: 72-85.

Cox K, Hearne D, Kauffman C & Robertson E with Helsel J & Smith B (1999) Kakamega Forest – a forest of the people. URL http://www.earlham.edu/~biol/kakamega/index.htm.

Crawley MJ & Harral JE (2001) Scale dependence in plant biodiversity. Science 291: 864-868.

Cromwell E & van Oosterhout S (2000) On-farm conservation of crop diversity: policy and institutional lessons from Zimbabwe, pp. 217-238., pp. 165-191. In: Brush SB ed. Genes in the field: on-farm conservation of crop diversity. Boca Raton: IPGRI, IDRC and Lewis Publishers.

Cronquist A (1981) An integrated system of classification of flowering plants. Columbia University Press, New York, 1262 pp.

Curran LM, Caniago I, Paoli GD, Astianti D, Kusneti M, Leighton M, Nirarita CE & Haeruman H (1999) Impact of El Niño and logging on canopy tree recruitment in Borneo. Science 286: 2184-2188.

Dalling JW, Hubbell SP & Silvera K (1998) Seed dispersal, seedling establishment and gap partitioning among tropical pioneer trees. Journal of Ecology 86(4): 674-689.

David S & Swinkels R (1994) Socio-economic characteristics of households engaged in agroforestry technology testing in Western Kenya. AFRENA Report No. 78, Nairobi: ICRAF, 33 pp.

Dawson IK & Powell W (1999) Genetic variation in the Afromontane tree Prunus africana, an endangered medicinal species. Molecular Ecology 8: 151-156.

Dawson IK, Waugh R, Simons AJ & Powell W (1997) Simple sequence repeats provide a direct estimate of pollen-mediated gene dispersal in the tropical tree Gliricidia sepium. Molecular Ecology 6: 179-183.

Dayandandan S, Dole J, Bawa KS & Kesseli R (1999) Population structure delineated with microsatellite markers in fragmented populations of a tropical tree, Carapa guianensis (Meliaceae). Molecular Ecology 8: 1585-1592.

De Leo GA & Levin S (1997) The multifaceted aspects of ecosystem integrity. Conservation Ecology (online) 1. URL

http://www.consecol.org/vol1/iss1/art3

Den Biggelaar C (1995) When a woman is a man: inter- and intra-gender differences in tree tenure and planting in Rwanda. Agroforestry Today 7(2): 13-15.

Deng H-W & Fu Y-X (1998) Conditions for positive and negative correlations between fitness and heterozygosity in equilibrium populations. Genetics 148: 1333-1340.

LITERATURE LIST

273

Dewees PA (1995) Farmer responses to tree scarcity: the case of woodfuel, pp. 174-197. In: Arnold JEM & Dewees PA eds. Tree management in farmer strategies: responses to agricultural intensification. New York: Oxford University Press.

Dover MJ & Talbot LM (1987) To feed the earth: agro-ecology for sustainable development. New Delhi: Mohan Primlani for Oxford and IBH Publishing Co. Pvt. Ltd, 96 pp.

Dynesius M & Jansson R (2000) Evolutionary consequences of changes in species’ geographical distributions driven by Milankovitch climate oscillations. Proceedings of the National Academy of Sciences 97(16): 9115-9120.

Elton CS (1958) The ecology of invasions by animals and plants. London: Methuen, pp. 145-153. Cited in: Tilman D (1996) Biodiversity: population versus ecosystem stability. Ecology 77: 350-63.

Epperson BK (1999) Gene genealogies in geographically structure populations. Genetics 152: 797-806.

Fonseca CR & Ganade G (2001) Species functional redundancy, random extinctions and the stability of ecosystems. Journal of Ecology 89: 118-125.

Frank DA & McNaughton SJ (1991) Stability increases with diversity in plant communities: empirical evidence from the 1988 Yellowstone drought. Oikos 62: 360-362.

Franzel S, Jaenicke H & Janssen W (1996) Choosing the right trees: setting priorities for multipurpose tree improvement. ISNAR Research Report No. 8. The Hague: International Service for National Agricultural Research, 87 pp.

Galindo-González J, Guevara S & Sosa VJ (2000) Bat- and bird-generated seed rains in isolated trees in pastures in a tropical rainforest. Conservation Biology 14: 1693-1703.

Gaston KJ (2000) Global patterns in biodiversity. Nature 405: 220-227.

Gavrilets S (1999) A dynamical theory of speciation on holey adaptive landscapes. The American naturalist 154: 1-22.

Gimaret-Carpentier C, Pelissier R, Pascal J P & Houllier F (1998) Sampling strategies for the assessment of tree species diversity. Journal of Vegetation Science 9(2): 161-172.

Gitay H, Wilson JB & Lee WG (1996) Species redundancy: a redundant concept? Journal of Ecology 84: 121-124.

Givnish TJ (1999) On the causes of gradients in tropical tree diversity. Journal of Ecology 87: 193-210.

Gonzalez A (2000) Community relaxation in fragmented landscapes: the relation between species richness, area and age. Ecology Letters 3: 441-448.

Gotelli NJ & Colwell RK (2001) Quantifying biodiversity: procedures and pitfalls in the measurement and comparison of species richness. Ecology Letters 4: 379-391.

Gotelli NJ (2001) Research frontiers in null model analysis. Global Ecology & Biogeography 10: 337-343.

Griffiths D (1999) On investigating local-regional species richness relationships. Journal of Animal Ecology 68: 1051-1055.

Guinand Y (1996) Impact assessment study of two wings agroforestry groups, Kabale district, Uganda. AFRENA Report No. 101, Nairobi: ICRAF, 37 pp.

Hamrick JL & Murawski DA (1990) The breeding structure of tropical tree populations. Plant Species Biology 5: 157-165.

Hamrick JL & Nason JD (2000) Gene flow in forest trees, pp. 81-90. In: Young A Boshier D & Boyle T eds. Forest Conservation Genetics. Collingwood: CSIRO Publishing & Wallingford: CABI Publishing.

Hamrick JL, Godt MJ & Sherman-Broyles SL (1992) Factors influencing levels of genetic diversity in woody plant species. New Forests 6: 95-124.

LITERATURE LIST

274

Hanski I & Gyllenberg M (1997) Uniting two general patterns in the distribution of species. Science 275: 397-400.

Hanski I (1998) Metapopulation dynamics. Nature 396: 41-49.

Hanski I (1999) Metapopulation ecology. New York: Oxford University Press, 313 pp.

Harlan JR (1975) Crops and Man. Madison: American Society of Agronomy, 295 pp.

Harrington L (1996) Diversity by design: conserving biological diversity through more productive and sustainable agroecosystems. Paper presented at Biodiversity and Sustainable Agriculture, a workshop arranged by the Swedish Scientific Council on Biological Diversity, 14 – 17 August, Ekenas, Sweden. URL http://www.cgiar.org/cimmyt.

Harte J, Kinzig A & Green J (1999) Self-similarity in the distribution and abundance of species. Science 284: 334-336.

Haugerud A & Collison M (1990) Plants, genes and people: improving the relevance of plant breeding in Africa. Experimental Agriculture 26: 67-84.

Hayek L-A C & Buzas MA (1997) Surveying natural populations. New York: Columbia University Press, 563 pp.

Hector A, Schmid B, Beierkuhnlein C, Caldeira C, Diemer M, Dimitrakopoulos PG, Finn JA, Freitas H, Giller PS, Good J, Harris R, Högberg P, Huss-Danell K, Joshi J, Jumpponen A, Körner C, Leadley PW, Loreau M, Minns A, Mulder CP, O'Donovan G, Otway SJ, Pereira JS, Prinz A, Read DJ, Scherer-Lorenzen M, Schulze E-D, Siamantziouras AD, Spehn EM, Terry AC, Troumbis AY, Woodward FI, Yachi S & Lawton J (1999) Plant diversity and productivity in European grasslands. Science 286: 1123-1127.

Hellier A, Newton AC & Ochoa SG (2001) Use of indigenous knowledge for rapidly assessing trends in biodiversity: a case study from Chiapas, Mexico. Biodiversity and Conservation 8(7): 869-889.

Hoekstra DA (1988) Summary of the zonal agroforestry potentials and research across landuse systems in the highlands of Eastern and Central Africa. AFRENA Report No. 5, Nairobi: ICRAF, 30 pp.

Holmes B (1998) Life support. New Scientist 159(2147): 30-34.

Holmgren D (2001) Revegetation planning for farm forestry. The Overstory No. 88. Holualoa: Permanent Agriculture Resources. URL: http://www.overstory.org

Holmgren P, Masakha EJ & Sjoholm H (1994) Not all African land is being degraded: a recent survey of trees on farms in Kenya reveals rapidly increasing forest resources. Ambio 23(7): 390-395.

Holsinger KE (2000) Reproductive systems and evolution in vascular plants. Proceedings of the National Academy of Sciences 97(13): 7037-7042.

Hood G (2000) Poptools Windows DLL. URL http://www.dwe.csiro.au/vbc/poptools/.

Hooper DU & Vitousek PM (1998) Effects of plant composition and diversity on nutrient cycling. Ecological Monographs 68: 121-149.

Hu X-S & Ennos RA (1999) Impacts of seed and pollen flow on population genetic structure for plant genomes with three contrasting modes of inheritance. Genetics 152: 441-450.

Hubbell SP, Foster RB, O’Brien ST, Harms KE, Condit R, Wechsler B, Wright SJ & Loo de Lao S (1999) Light-gap disturbances, recruitment limitation and tree diversity in a neotropical forest. Science 283: 554-557.

Hurlbert SH (1971) The nonconcept of species diversity: a critique and alternative parameters. Ecology 52: 577-586.

LITERATURE LIST

275

Huston MA, Aarssen LW, Austin MP, Cade BS, Fridley JD, Garnier E, Grime JP, Hodgson J, Lauenroth WK, Thompson K, Vandermeer JH, Wardle DA, Hector A, Schmid B, Beierkuhnlein C, Caldeira MC, Diemer M, Dimitrakopoulos PG, Finn JA, Freitas H, Giller PS, Good J, Harris R, Högberg P, Huss-Danell K, Joshi J, Jumpponen A, Körner C, Leadley PW, Loreau M, Minns A, Mulder CPH, O'Donovan G, Otway SJ, Pereira JS, Prinz A, Read DJ, Scherer-Lorenzen M, Schulze E-D, Siamantziouras A-SD, Spehn E, Terry AC, Troumbis AY, Woodward FI, Yachi S & Lawton JH (2000) No consistent effect of plant diversity on productivity. Science 289: 1255a.

Huston MA, Balmford A, Moore J, Brooks T, Burgess N, Hansen LA, Lovett JC, Tokumine S, Williams P, Woodward FI & Rahbek C (2001). People and biodiversity in Africa. Science 293: 1591-1592.

Hutchinson GE (1959) Homage to Santa Rosalia, or why are there so many kinds of animals? The American Naturalist 93: 145-159.

ICRAF (1997) ICRAF Medium-Term Plan 1998-2000. Nairobi: ICRAF, 77 pp.

ICRAF (2000) Paths to prosperity through agroforestry. ICRAF’s corporate strategy 2001-2010. Nairobi: ICRAF, iii+43 pp.

Ipalei G (1999). Kakamega Forest National Reserve. Nairobi: Kenya Wildlife Services. URL http://www.kenya-wildlife-service.org/kakamega.htm

Ives AR, Gross K & Klug JL (1999) Stability and variability in competitive communities. Science 286: 542-544.

Ives AR, Klug JL & Gross K (2000) Stability and species richness in complex communities. Ecology Letters 3(5): 399-411.

Jaetzhold R & Schmidt H (1983) Farm management handbook of Kenya. Ministry of Agriculture, Kenya and GTZ, Rossdorf

Jarvis D & van Hodgkin T (2000) Farmer decision making and genetic diversity: linking multidisciplinary research to implementation on-farm, pp. 261-278. In: Brush SB ed. Genes in the field: on-farm conservation of crop diversity. Boca Raton: IPGRI, IDRC and Lewis Publishers.

Jarvis DI, Myer L, Klemick H, Guarino L, Smale M, Brown AHD, Sadiki M, Sthapit B & Hodgkin T (2000) A training guide for in situ conservation on farm. Version 1. Rome: International Plant Genetic Resources Institute, 161 pp.

Johnson KH, Vogt KA, Clark HJ, Schmitz OJ & Vogt DJ (1996) Biodiversity and the productivity and stability of ecosystems. Trends in ecology and evolution 11: 372-377.

Kaiser J (2000a) Rift over biodiversity divides ecologists. Science 289: 1282-1283.

Kaiser J (2000b) When do many species matter? Science 289: 1283.

Kaj I & Lascoux M (1999) Probability of identity by descent in metapopulations. Genetics 152: 1217-1228.

Kelly RD & Walker BH (1976) The effect of different forms of land use on the ecology of a semi-arid region in South-Eastern Rhodesia. Journal of ecology 64: 553-576.

Kindt R & Lengkeek AG (1999) Tree diversity on farm – use it or lose it. Papers presented at the national workshop on agricultural biodiversity conservation, 27-29 January 1999, Nairobi (Kenya). Intermediate Technology Development Group, Kenya: pp. 75-85.

Kindt R (1997) Local perceptions on tree propagation and domestication - results from a survey in western Kenya. AFRENA Report No. 115, Nairobi: ICRAF, 45 pp.

Kindt R (2001a) ExactS. Program to calculate the exact average species richness associated with random site sequences. International Centre for Research in Agroforestry, Nairobi.

Kindt R (2001b) ExactSN. Program to calculate the exact average species richness associated with random sequences of individuals. International Centre for Research in Agroforestry, Nairobi.

LITERATURE LIST

276

Kindt R (2001c) MetaPop. Individual-tree-based simulation program of genetic diversity dynamics in metapopulations International Centre for Research in Agroforestry, Nairobi.

Kindt R (2001d) RenyiAccum. Program to calculate average values and ranges for the Rényi diversity series associated with random site sequences. International Centre for Research in Agroforestry, Nairobi.

Kindt R (2001e) RenyiRef. Program to calculate average values, ranges, and confidence intervals for Rényi diversity and evenness profiles at fixed subsample sizes based on random site sequences. International Centre for Research in Agroforestry, Nairobi.

Kindt R (2001f) RenyiRefAbun. Program to calculate average values, ranges, and confidence intervals for Rényi diversity and evenness profiles for fixed abundance-based subsamples based on random site sequences. International Centre for Research in Agroforestry, Nairobi.

Kindt R (2001g) RenyiRefSize. Program to calculate average values, ranges, and confidence intervals for Rényi diversity and evenness profiles for fixed area-based subsamples based on random site sequences. International Centre for Research in Agroforestry, Nairobi.

Kindt R (2001h) SpeciesRichClus. Program to calculate average values, ranges, and confidence intervals for species richness for sites that are randomly placed in clusters of equal or unequal size. International Centre for Research in Agroforestry, Nairobi.

Kindt R (2001i) SpeciesRichRef. Program to calculate average values, ranges, and confidence intervals for species richness at fixed subsample sizes based on random site sequences. International Centre for Research in Agroforestry, Nairobi.

Kindt R, Degrande A, Turyomurugyendo L, Mbosso C, Van Damme P & Simons AJ (2001) Comparing species richness and evenness contributions to on-farm tree diversity for data sets with varying sample sizes from Kenya, Uganda, Cameroon, and Nigeria with randomized diversity profiles. Paper presented at IUFRO conference on forest biometry, modelling and information science, 26 - 29 June, University of Greenwich.

URL http://cms1.gre.ac.uk/conferences/iufro/proceedings/

Kirchner JW & Weil A (2000) Delayed biological recovery from extinctions throughout the fossil record. Nature 404: 177-180.

Kirchner JW (2002) Evolutionary speed limits inferred from the fossil record. Nature 415: 65-68.

Kleijn D, Berendse F, Smit R & Gilissen N (2001) Agri-environment schemes do not effectively protect biodiversity in Dutch agricultural landscapes. Nature 413: 723-725.

Kohyama T (1991) Simulating stationary size distribution of trees in rain forests. Annals of botany 68: 173-180.

Kress WJ & Beach JH (1994) Flowering plant reproductive systems, pp. 161-182. In: McDade LA, Bawa KS, Hespenheide HA & Hartshorn GS eds. La Selva, ecology and natural history of a neotropical rain forest. Chicago: University of Chicago Press.

Lande R (1995) Mutation and conservation. Conservation Biology 9: 782-791.

Lauriks (1996) Onderzoek naar veranderingen in landgebruik in West-Kenya met behulp van luchtfoto-interpretatie. Gent: Faculteit landbouwkundige en toegepaste biologische wetenschappen, 125 pp.

Lauriks R, De Wulf R, Carter SE & Niang A (1999) A methodology for the description of border hedges and the analysis of variables influencing their distribution: a case study in western Kenya. Agroforestry Systems 44: 69-86.

Lawrence A & Ambrose-Oji B (1999) Participatory evaluation of biodiversity: a literature review. Bangor: University of Wales, 73 pp.

Lawrence A & Ambrose-Oji B (2001) Participatory assessment, monitoring and evaluation of biodiversity: the art and the science. Draft background paper as an input to the European Tropical Forest Research Network Workshop. URL http://www.etfrn.org/etfrn/workshop/biodiversity/index.html

LITERATURE LIST

277

Leakey RRB (1996) Definition of agroforestry revisited. Agroforestry Today 8: 5-7.

Legendre P & Anderson MJ (1999) Distance-based redundancy analysis: testing multispecies responses in multifactorial ecological experiments. Ecological Monographs 69 (1): 1-24.

Legendre P & Gallagher ED (2001) Ecologically meaningful transformations for ordination of species data. Oecologia 129: 271-280.

Legendre P & Legendre L (1998) Numerical ecology. Second English edition. Amsterdam: Elsevier Science BV, 853 pp.

Legendre P (1999a) Program for multiple linear regression with permutation test. User’s notes. Département de sciences biologiques, Université de Montréal., 5 pp. URL http://www.fas.umontreal.ca/biol/legendre

Legendre P (1999b) Program for transformation of frequency data. User’s notes. Département de sciences biologiques, Université de Montréal, 2pp. URL: http://www.fas.umontreal.ca/biol/legendre/

Levine JM (2000) Species diversity and biological invasions: relating local process to community pattern. Science 288: 852-854.

Lomolino MV (2000) Ecology’s most general, yet protean pattern: the species-area relationship. Journal of Biogeography 27: 17-26.

Long J, Cromwell E & Gold K (2000) On-farm management of crop diversity: an introductory bibliography. London: Overseas Development Institute for ITDG, 42 pp.

Loreau M & Hector A (2001) Partitioning selection and complementarity in biodiversity experiments. Nature 412: 72-76.

Loreau M (1998) Biodiversity and ecosystem functioning: a mechanistic model. Proceedings of the National Academy of Sciences 95 (10): 5632-5636.

Loreau M (2000) Are communities saturated? On the relationship between α, β and γ diversity. Ecology Letters 3: 73-76.

Loreau M, Naeem S, Inchausti P, Bengtsson J, Grime JP, Hector A, Hooper DU, Huston MA, Raffaelli D, Schmid B, Tilman D & Wardle DA (2001) Biodiversity and ecosystem functioning: current knowledge and future challenges. Science 294: 804-807.

Losos JB & Schluter D (2000) Analysis of an evolutionary species-area relationship. Science 408: 847-850.

Lowe AJ, Gillies ACM, Wilson J & Dawson IK (2000) Conservation genetics of bush mango from central/west Africa: implications from random amplified polymorphic DNA analysis. Molecular Ecology 9: 831-841.

Lynch M, Conery J & Burger R (1995) Mutation accumulation and the extinction of small populations. American Naturalist 146: 489-518.

Mabberley DJ (1997) The plant-book: a portable dictionary of the vascular plants, 2nd edn. Cambridge University Press, Cambridge, 858 pp.

MacDicken KG, Wolf GV & Briscoe CB (1991) Standard research methods for multipurpose trees and shrubs. Arlington: Winrock International Institute for Agricultural Development, 92 pp.

Maddux RD, Athreya K, Harte J, Kinzig A & Green J (1999) On the distribution and abundance of species. Science 286: 1647a.

Magnussen S & Boyle TJB (1995) Estimating sample size for inference about the Shannon-Weaver and the Simpson indices of species diversity. Forest Ecology and Management 78: 71-84.

Magurran AE (1988) Ecological diversity and its measurement. London: Croom Helm Limited, 179 pp.

LITERATURE LIST

278

Mahy G, Bruederle LP, Connors B, Van Hofwegen M & Vorsa N (2000) Allozyme evidence for genetic autopolyploidy and high genetic diversity in tetraploid cranberry, Vaccinium oxycoccos (Ericaceae). American Journal of Botany 87(12): 1882-1889.

Makarenkov V & Legendre P (1998) Program Polynomial RdaCca. Département de sciences biologiques, Université de Montréal, 8 pp. URL: http://www.fas.umontreal.ca/biol/legendre/.

Makarenkov V & Legendre P (1999) Une méthode d’ analyse canonique non-linéaire et son application à des données biologiques. Mathématiques, Informatique et Sciences Humaines. 147: 135-147.

Makarenkov V & Legendre P (2002) Nonlinear redundancy analysis and canonical correspondence analysis based on polynomial regression. Ecology 83(4): 1146-1161.

May RM & Stumpf MPH (2000) Species-area relations in tropical forests. Science 290: 2084-2086.

May RM (1972) Will a large complex system be stable? Nature 238: 413-414.

May RM (1975) Patterns of species abundance and diversity, pp. 81-120. In: Cody ML & Diamond JM eds. Ecology and the evolution of communities. Cambridge and London: The Belknap Press of Harvard University Press.

McAleece N, Lambshead J, Patterson G & Gage J (1997) Biodiversity Professional. The Natural History Museum & The Scottish Association for Marine Science.

URL: http://www.nhm.ac.uk/zoology/bdpro

McCann KS (2000) The diversity-stability debate. Nature 405: 228-233.

McCune B & Mefford MJ (1999) PC-ORD for Windows. Multivariate Analysis of Ecological Data version 4.13. Oregon: MjM Software.

McKane A, Alonso D & Solé RV (2000) A mean field stochastic theory for species-rich assembled communities. Physical Review E 62(6): 8466-8484.

McNaughton SJ (1994) Biodiversity and function of grazing ecosystems, pp. 361-383. In: Schulze E-D & Mooney HA, eds. Biodiversity and ecosystem function. Berlin: Springer-Verlag, 525 pp.

McNeely JA & Scherr SJ (2001) Common ground, common future. How ecoagriculture can help feed the world and save wild biodiversity. IUCN - the World Conservation Union & Future Harvest, Gland & Washington, 26 pp.

Meldrum D (2000) Automation for genomics, part two: sequencers, microarrays, and future trends. Genome Research 10: 1288-1303.

Merrick L (1990) Crop genetic diversity and its conservation in traditional agroecosystems, pp. 3-11. In: Altieri MA & Hecht SB. Agroecology and small farm development. Boston: CRC Press.

Miller JS & Venable DL (2000) Polyploidy and the evolution of gender dimorphism in plants. Science 289: 2335-2338.

Momose K, Takakazu Y, Nagamitsu T, Kato M, Nagamasu H, Sakai S, Harrison RD, Itioka T, Hamid AA & Inoue T (1998) Pollination biology in a lowland dipterocarp forest in Sarawak, Malaysia. I. Characteristics of the plant-pollinator community in a lowland dipterocarp forest. American Journal of Botany 85(10): 1477-1501.

Moreno CE & Halffter G (2001) On the measure of sampling effort used in species accumulation curves. Journal of Applied Ecology 38: 487-490.

Mortimore M, Harris FMA & Turner B (1999) Implications of land use change for the production of plant biomass in densely populated Sahelo-Sudanian shrub-grasslands in north-east Nigeria. Global Ecology and Biogeography 8: 243-256.

Murawski DA, Gunatilleke IAUN & Bawa KS (1994) The effects of selective logging and inbreeding on Shorea megistrophylla (Dipterocarpaceae) from Sri Lanka. Conservation Biology 8, 997-1002.

LITERATURE LIST

279

Myers N, Mittermeier RA, Mittermeier CG, Da Fonseca GAB & Kent J (2001). Biodiversity hotspots for conservation priorities. Nature 403: 853-858.

Naeem S, Thompson LJ, Lawler SP, Lawton JH & Woodfin RM (1994) Declining biodiversity can alter the performance of ecosystems. Nature 368: 734-737.

Nagylaki T (1998) The expected number of heterozygous sites in a subdivided population. Genetics 149: 1599-1604.

Nevo E (2001) Inaugural Article: Evolution of genome-phenome diversity under environmental stress. Proceedings of the National Academy of Sciences 98(11): 6233-6240.

Nijs I & Roy J (2000) How important are species richness, species evenness and interspecific differences to productivity? A mathematical model. Oikos 88: 57-66.

Norberg J, Swaney DP, Dushoff J, Lin J, Casagrandi R & Levin SA (2001) Phenotypic diversity and ecosystem functioning in changing environments: a theoretical framework. Proceedings of the National Academy of Sciences 98(20): 11376-11381.

Novacek MJ & Cleland EE (2001) The current biodiversity extinction event: Scenarios for mitigation and recovery. Proceedings of the National Academy of Sciences 98(10): 5466-5470.

Nunes PALD & van den Berg JCGM (2001) Economic valuation of biodiversity: sense or nonsense? Ecological Economics 39 (2): 203-222.

O’Neill GA, Dawson IK, Sotelo-Montes C, Guarino L, Current D, Guariguata M, & Weber, JC (2001) Strategies for genetic conservation of trees in the Peruvian Amazon basin. Biodiversity and Conservation 10(6): 837-850.

Økland RH (1996) Are ordination and constrained ordination alternative or complementary strategies in general ecological studies? Journal of Vegetation Science 7: 289-292.

Økland RH (1999) On the variation explained by ordination and constrained ordination axes. Journal of Vegetation Science 10: 131-136.

Oksanen J (1996) Is the humped relationship between species richness and biomass and artefact due to plot size? Journal of Ecology 84: 293-295.

Ostling A, Harte J, Green J & Condit R (2000) Self-similarity and clustering in the spatial distribution of species. Science 290: 671a.

Pachepsky E, Crawford JW, Bown JL & Squire G (2001) Towards a general theory of biodiversity. Nature 410: 923-926.

Palmer J, Harwood C & Van Wyk G (2001). Internally commissioned external review – ICRAF Tree Domestication Programme. Nairobi: ICRAF, 39 pp.

Patel SH, Pinckney TC & Jaeger WK (1995) Smallholder wood production and population pressure in East Africa: evidence of an environmental Kuznets curve? Land-Economics 71(4): 516-530.

Phillips OL, Hall P, Gentry AH, Sawyer SA & Vásquez R (1994) Dynamics and species richness of tropical rain forests. Proceedings of the National Academy of Sciences 91: 2805-2809.

Pimm SL (1984) The complexity and stability of ecosystems. Nature 307: 321-326.

Pimm SL, Russell GJ, Gittleman JL & Brooks TM (1995) The future of biodiversity. Science 269: 347-350.

Plotkin JB, Potts MD, Yu DW, Bunyavejchewin S, Condit R, Foster R, Hubbell S, LaFrankie J, Manokaran N, Seng LH, Sukumar R, Nowak MA & Ashton PS (2000) Predicting species diversity in tropical forests. Proceedings of the National Academy of Sciences 97 (20): 10850-10854.

Ponce ER, Ponce LB & Maurillo LA (1991) Preferred characteristics of multipurpose tree species: a case study with lowland and upland farmers in Leyte, Philippines. Forestry/Fuelwood Research and Development Project Report No. 17, Arlington: Winrock International Institute for Agricultural Development, 35 pp.

LITERATURE LIST

280

Price GR (1970) Selection and covariance. Nature 227: 520-521.

Purvis A & Hector A (2000) Getting the measure of biodiversity. Nature 405: 212-218.

Rao CR (1995) A review of canonical coordinates and an alternative to correspondence analysis using Hellinger distance. Quaderns d’ Estadística i Investigació Operativa 19: 23-63.

Reay SD & Norton DA (1999) Assessing the success of restoration plantings in a temperate New Zealand Forest. Restoration Ecology 7: 298-308.

Reice SR (1994) Nonequilibrium determinants of biological community structure. American Scientist 82: 424-435.

Reich PB, Walters MB & Ellsworth DS (1997) From tropics to tundra: global convergence in plant functioning. Proceedings of the National Academy of Sciences 94: 13730-13734.

Rennolls K & Laumonier Y (1999) Tree Species-Area and Species-Diameter Relationships at Three Lowland Rain Forest Sites in Sumatra. Journal of Tropical Forest Science 11: 784-800.

Rennolls K & Laumonier Y (2000) Species Diversity Structure Analysis at Two Sites in the Tropical Rain Forest of Sumatra. Journal of Tropical Ecology 16: 253-270.

Richards P (1993) Culture and community values in the selection and maintenance of African rice. Paper presented at the conference on intellectual property rights and indigenous knowledge, Granlibakken, Lake Tahoe, 5-10 October 1993.

Ritchie ME & Olff H (1999) Spatial scaling laws yield a synthetic theory of biodiversity. Nature 400: 557-560.

Rodrigues ASL & Gaston KJ (2001) How large do reserve networks need to be? Ecology Letters 4: 602-609.

Rodriguez MA & Gomez-Sal A (1994) Stability may decrease with diversity in grassland communities: empirical evidence from the 1986 Cantabrian Mountains (Spain) drought. Oikos 71: 177-180.

Rosenzweig ML (1999) Heeding the warning of biodiversity’s basic law. Science 284: 276-277.

Rosenzweig ML (2001) Loss of speciation rate will impoverish future diversity. Proceedings of the National Academy of Sciences 98(10): 5404-5410.

Rousseau D, Van Hecke P, Nijssen D & Bogaert J (1999) The relationship between diversity profiles, evenness and species richness based on partial ordering. Environmental and Ecological Statistics 6: 211-223.

Russell JR, Weber JC, Booth A, Powell W, Sotelo-Montes C & Dawson IK (1999) Genetic variation of Calycophyllum spruceanum in the Peruvian Amazon Basin, revealed by amplified fragment length polymorphism (AFLP) analysis. Molecular Ecology 8: 199-204.

Saccheri IJ, Wilson IJ, Nichols RA, Bruford MW & Brakefield PM (1999) Inbreeding of bottlenecked butterfly populations: estimation using the likelihood of changes in marker allele frequencies. Genetics 151: 1053-1063.

Safford RJ & Jones CG (1998) Strategies for land-bird conservation on Mauritius. Conservation Biology 12(1): 169-176.

Salafsky N, Cauley H, Balachander G, Cordes B, Parks J, Margoluis C, Bhatt S, Encarnacion C, Russell D & Margoluis R (2001) A systematic test of an enterprise strategy for community-based biodiversity conservation. Conservation Biology 15(6): 1585-1595.

Sanders HL (1968) Marine benthic diversity: a comparative study. The American Naturalist 102: 243-282.

Scheffer M, Carpenter S, Foley JA, Folke C & Walker B (2001) Catastrophic shifts in ecosystems. Nature 413: 591-596.

Scherr SJ (1995) Meeting household needs: farmer tree-growing strategies in western Kenya, pp. 271-287. In: Arnold JEM and Dewees PA eds. Tree management in farmer strategies: responses to agricultural intensification. New York: Oxford University Press.

LITERATURE LIST

281

Schroth G, Krauss U, Gasparotto L, Duarte Aguilar JA, Vohland K (2000) Pests and diseases in agroforestry systems of the humid tropics. Agroforestry Systems 50(3): 199-241.

Schulze E-D & Mooney HA (1994) Ecosystem function of biodiversity: a summary, pp. 497-510. In: Schulze E-D & Mooney HA eds. Biodiversity and ecosystem function. Berlin: Springer-Verlag.

Seydack AHW (1995) An unconventional approach to timber yield regulation for multi-aged, multispecies forests. I. Fundamental considerations. Forest Ecology and management 77: 139-153.

Seydack AHW, Vermeulen WJ, Heyns HE, Durrheim GP, Vermeulen C, Willems D, Ferguson MA, Huisamen J & Roth J (1995) An unconventional approach to timber yield regulation for multi-aged, multispecies forests. II. Application to a South African forest. Forest Ecology and management 77: 155-168.

Shaw RG & Mitchell-Olds T (1993) Anova for unbalanced data: an overview. Ecology 74(6): 1638-1645.

Sheil D (2001) Conservation and biodiversity monitoring in the tropics: realities, priorities, and distractions. Conservation Biology 15(4): 1179-1182.

Shepherd KD, Ndufa JK, Ohlsson E, Sjogren H & Swinkels R (1997) Adoption potential of hedgerow intercropping in maize-based cropping systems in the highlands of western Kenya. 1. Background and agronomic evaluation. Experimental Agriculture 33 (2): 197-209.

Shigeta M (1990) Folk in-situ conservation of ensete (Ensete ventricosum (Welw.) EE Cheesman): towards the interpretation of indigenous agricultural science of the Ari, Southwestern Ethiopia. African Study Monographs 10(3): 93-97.

Simons AJ (1996) Delivery of improvement of agroforestry trees (Keynote presentation), pp. 391-400. In: Dieters MJ, Matheson AC, Nikles DG, Harwood CE & Walker SM eds. Tree Improvement for Sustainable Forestry. IUFRO Conference, 27 October - 1 November 1996, Caloundra, Australia.

Simons AJ (1997) Tree domestication – better trees for rural prosperity. Agroforestry Today 9(2): 4-6.

Simons AJ, MacQueen DJ & Stewart JL (1994) Strategic concepts in the domestication of non-industrial trees, pp. 91-102. In: Leakey RRB & Newton AC eds. Tropical trees: the potential for domestication and the rebuilding of forest resources. London: HMSO.

Simons AJ, Sanchez PA, Konuche PKA, Place F & Kindt R The Future of Trees is on Farm: High value, high impact. In preparation.

Smith RS & Rushton SP (1994) The effects of grazing on the vegetation of mesotrophic (meadow) grassland in Northern England. Journal of Applied Ecology 31: 13-24.

S-PLUS 2000 (1999) Professional Edition for Windows, Release 1. MathSoft, Inc.

Stacy EA, Hamrick JL, Nason JD, Hubbell SP, Foster RB & Condit R (1996) Pollen dispersal in low-density populations of three neotropical tree species. The American Naturalist 148: 275-298.

Stewart JL & Salazar R (1992) A review of measurement options for multipurpose trees. Agroforestry Systems 19: 173-183.

Swanson T & Goeschl T (2000) Optimal genetic resource conservation: in situ and ex situ, pp. 165-191. In: Brush SB ed. Genes in the field: on-farm conservation of crop diversity. Boca Raton: IPGRI, IDRC and Lewis Publishers.

Swift MJ, Vandermeer J, Ramakrishnan PS, Anderson JM, Ong CK & Hawkins BA (1996) Biodiversity and Agroecosystem Function, pp. 261-298. In: Mooney, HA; Cushman, JH; Medina, E; Sala, OE & Schulze, E-D, eds. Functional roles of biodiversity: a global perspective. Chichester: John Wiley & Sons Ltd.

Ter Braak CJF & Smilauer P (1998) CANOCO reference manual and user’s guide to Canoco for windows: software for canonical community ordination (version 4). Ithaca: Microcomputer Power, 352 pp.

LITERATURE LIST

282

Ter Braak CJF (1986) Canonical correspondence analysis: a new eigenvector technique for multivariate direct gradient analysis. Ecology 67: 1167-1179.

Thijssen HJC, Murithi FM, Nyaata OZ, Mwangi JN, Aiyelaagbe IOO & Mugendi DN (1993) Report of an ethnobotanical survey of woody perennials in the coffee zone of Embu District, Kenya. AFRENA Report No. 62, Nairobi: ICRAF, 32 pp.

Thomas DW, Rao PRS, Gokhale Y, Ghate U & Gadgil M (n.d.) A methodology manual for documenting people’s priorities for biodiversity and conservation. URL http://ces.iisc.ernet.in/hpg/cesmg/pew/srusti.html.

Thompson JN (1999) The evolution of species interactions. Science 284: 2116-2118.

Thrupp LA (1998) Cultivating diversity. Agrobiodiversity and food security. Washington: World Resources Institute, 72 pp.

Tilman D & Downing JA (1994) Biodiversity and stability in grassland. Nature 367: 363-365.

Tilman D & Lehman JA (2001) Human-caused environmental change: Impacts on plant diversity and evolution. Proceedings of the National Academy of Sciences 98(10): 5433-5440.

Tilman D (1994) Community diversity and succession: the roles of competition, dispersal and habitat modification, pp. 327-342. In: Schulze E-D & Mooney HA. Biodiversity and ecosystem function. Berlin: Springer-Verlag.

Tilman D (1996) Biodiversity: population versus ecosystem stability. Ecology 77: 350-63.

Tilman D (1999a) Diversity by default. Science 283: 495-496.

Tilman D (1999b) Global environmental impacts of agricultural expansion: the need for sustainable and efficient practices. Proceedings of the National Academy of Sciences 96: 5995-6000.

Tilman D (2000) Causes, consequences and ethics of biodiversity. Nature 405: 208-211.

Tilman D (2001) An evolutionary approach to ecosystem functioning. Proceedings of the National Academy of Sciences 98(20): 11979-11980.

Tilman D, Fargione J, Wolff B, D’Antonio C, Dobson A, Howarth R, Schindler D, Schlesinger WH, Simberloff D & Swackhamer D (2001a) Forecasting agriculturally driven global environmental change. Science 292: 281-284.

Tilman D, Knops J, Wedin D, Reich P, Ritchie M & Siemann E (1997a) The influence of functional diversity and composition on ecosystem processes. Science 277: 1300-1302.

Tilman D, Lehman CL & Kareiva P (1997b) Population dynamics in spatial habitats, pp. 3-20. In: Tilman D & Kareiva P eds. Spatial Ecology. The role of space in population dynamics and interspecific competition. Princeton: Princeton University Press.

Tilman D, Lehman CL & Thomson KT (1997c) Plant diversity and ecosystem productivity: theoretical considerations. Proceedings of the National Academy of Sciences 94: 1857-1861.

Tilman D, Reich PB, Knops J, Wedin D, Mielke T & Lehman C (2001b) Diversity and productivity in a long-term grassland experiment. Science 294: 843-845.

Tilman D, Wedin D & Knops J (1996) Productivity and sustainability influenced by biodiversity in grassland ecosystems. Nature 379: 718-720.

Tóthmérész B (1995) Comparison of different methods for diversity ordering. Journal of Vegetation Science 6: 283-290.

Trenbath BR (1994) Biomass productivity of mixtures. Advances in Agronomy 26: 177-210.

Trenbath BR (1999) Multispecies cropping systems in India. Agroforestry Systems 45: 81-107.

Van der Heide WM, Thripp R & de Boef WS (1996) Local crop development: an annotated bibliography. Rome: International Plant Genetic Resources Institute, 153 pp.

LITERATURE LIST

283

Van der Maarel E (1993) Some remarks on disturbance and its relations to diversity and stability. Journal of Vegetation Science 4: 733-736.

Van Mele P (2000) Evaluating farmers’ knowledge, perceptions and practices: a case study of pest management of fruit farmers in the Mekong Delta, Vietnam. Doctoral Thesis Wageningen University, the Netherlands, 229 pp.

Van Noordwijk M & Lusiana B (1999) WaNuLCAS, a model of water, nutrient and light capture in agroforestry systems. Agroforestry Systems 43: 217-242.

Van Noordwijk M & Ong CK (1999) Can the ecosystem mimic hypotheses be applied to farms in African savannahs? Agroforestry Systems 45: 131-158.

Van Noordwijk M, Dijksterhuis G & Van Keulen H (1994) Risk management in crop production and fertilizer use with uncertain rainfall: how many eggs in which baskets. Netherlands Journal of Agricultural Science 42: 249-269.

Van Noordwijk M, Tomich TP De Foresta H & Michon G (1997) To segregate or to integrate? The question of balance between production and biodiversity conservation in complex agroforestry systems. Agroforestry Today 9(1): 6-9.

Vitalis R & Couvet D (2001) Estimation of effective population size and migration rate from one- and two-locus identity measures. Genetics 157: 911-925.

Voss J (1992) Conserving and increasing on-farm genetic diversity: farmer management of varietal bean mixtures in Central Africa, pp. 34-51. In: Moock JL & Rhoades RE eds. Tree management in farmer strategies: responses to agricultural intensification. Ithaka and London: Cornell University Press.

Waide RB, Willig MR, Steiner, CF, Mittelbach G, Gough L, Dodson SI, Juday GP & Parmenter R (1999) The relationship between productivity and species richness. Annual Review of Ecology and Systematics 30: 257-300.

Walker B (1995) Conserving biological diversity through ecosystem resilience. Conservation Biology 9: 747-752.

Walter KS & Gillett HJ (1998) 1997 IUCN Red List of Threatened Plants. Compiled by the World Conservation Monitoring Centre. IUCN - The World Conservation Union, Gland and Cambridge. URL http://www.unep-wcmc.org

Wang J (1997) Effective size and F-statistics of subdivided populations. I. Monoecious speciees with partial selfing. Genetics 146: 1453-1463.

Warner K (1993) Patterns of farmer tree growing in Eastern Africa: a socio-economic analysis. Tropical Forestry Papers No. 27, Oxford: Oxford Forestry Institute and Nairobi: ICRAF.

Warner K (1995) Patterns of tree growing by farmers in eastern Africa, pp. 90-137. In: Arnold JEM & Dewees PA eds. Tree management in farmer strategies: responses to agricultural intensification. New York: Oxford University Press.

Weithoff G, Walz N & Gaedke U (2001) The intermediate disturbance hypothesis – species diversity or functional diversity. Journal of Plankton Research 23(10): 1147-1155.

Western D (2001) Human-modified ecosystems and future evolution. Proceedings of the National Academy of Sciences 98(10): 5458-5465.

White GM, Boshier DH & Powell W (2002) Increased pollen flow counteracts fragmentation in a tropical dry forest: an example from Swietenia humilis Zuccarini. Proceedings of the National Academy of Sciences 99: 2038-2042.

Whitlock MC & Barton NH (1997) The effective size of a subdivided population. Genetics 146: 427-441.

Wills C, Condit R, Foster RB & Hubbell SP (1997) Strong density- and diversity-related effects help to maintain tree species diversity in a neotropical forest. Proceedings of the National Academy of Sciences 94: 1252-1257.

LITERATURE LIST

284

Woodruff DS (2001) Declines of biomes and biotas and the future of evolution. Proceedings of the National Academy of Sciences 98(10): 5471-5476.

Wu R, Gallo-Meagher M, Littell RC & Zeng Z-B (2001) A general polyploid model for analyzing gene segregation in outcrossing tetraploid species. Genetics 159: 869-882.

Yachi S & Loreau M (1999) Biodiversity and ecosystem productivity in a fluctuating environment: the insurance hypothesis. Proceedings of the National Academy of Sciences 96 (4): 1463-1468.

Yeh FC (2000) Population genetics, pp. 21-37. In: Young A, Boshier D & Boyle T eds. Forest Conservation Genetics. Collingwood: CSIRO Publishing & Wallingford: CABI Publishing.

Young AG & Boyle TJ (2000) Forest fragmentation, pp. 123-134. In: Young A Boshier D & Boyle T eds. Forest Conservation Genetics. Collingwood: CSIRO Publishing & Wallingford: CABI Publishing.

Young AG & Merriam HG (1994) Effects of forest fragmentation on the spatial genetic structure of Acer saccharum Marsh. (sugar maple) populations. Heredity 72: 201-208.

Zhu Y, Chen H, Fan H, Wang Y, Li Y, Chen J, Fan J, Yang S, Hu L, Leung H, Mew TW, Teng PS, Wang Z & Mundt CC (2000) Genetic diversity and disease control in rice. Nature 406: 718-722.

APPENDIX I MATRICES WITH INFORMATION ON TREE SPECIES IDENTITIES, FARM CHARACTERISTICS AND SPECIES

DISTRIBUTION OVER FARMS IN WESTERN KENYA

APPENDIX I

286

Table I.1. Species encountered on farms in western Kenya with farm frequencies of the most frequent use groups. (unid.: unidentified species)

No.

Spec

ies

Fam

ily

Beve

rage

Boun

dary

Cha

rcoa

l

Con

stru

ctio

n

Fire

woo

d

Fodd

er

Frui

t

Tim

ber

Med

icin

e

Orn

amen

tal

Shad

e

Soil

fert

ility

1 Acacia abyssinica Fabaceae (Mimosoideae) 0 1 1 0 2 0 0 0 0 0 0 0 2 Acacia lahai Fabaceae (Mimosoideae) 0 0 1 0 2 0 0 0 1 0 1 0 3 Acacia mearnsii Fabaceae (Mimosoideae) 0 0 1 0 2 0 0 0 0 0 0 0 4 Acacia persiciflora Fabaceae (Mimosoideae) 0 0 0 0 0 0 0 0 0 0 0 0 5 Acanthus pubescens Acanthaceae 0 0 0 0 1 0 0 0 0 0 0 0 6 Acacia seyal Fabaceae (Mimosoideae) 0 0 0 0 0 0 1 0 1 0 0 0 7 Acrocarpus fraxinifolius Fabaceae (Caesalpinoideae) 0 0 1 0 13 0 0 13 0 0 2 0 8 Afrosersalisia cerasifera Sapotaceae 0 0 0 0 4 0 2 1 0 1 1 0 9 Agave sisalana Agavaceae 0 7 0 0 0 0 0 0 0 0 0 0 10 Albizia coriaria Fabaceae (Mimosoideae) 0 0 2 2 26 0 0 4 12 0 9 0 11 Albizia grandibracteata Fabaceae (Mimosoideae) 0 0 0 0 6 0 0 1 1 0 2 0 12 Albizia gummifera Fabaceae (Mimosoideae) 0 0 0 0 4 0 0 0 0 0 1 0 13 Aleurites moluccana Euphorbiaceae 0 0 0 0 2 0 2 0 0 0 1 0 14 Allophylus ferrugineus Sapindaceae 0 0 0 0 2 0 0 0 1 0 0 0 15 Anacardium occidentale Anacardiaceae 1 0 0 0 1 0 0 0 0 0 0 0 16 Annona muricata Annonaceae 0 0 0 0 1 0 1 0 0 0 0 0 17 Annona senegalensis Annonaceae 0 0 0 0 1 0 1 0 0 0 0 0 18 Antiaris toxicaria Moraceae 0 0 0 0 3 0 0 1 1 0 0 0 19 Artocarpus heterophyllus Moraceae 0 0 0 0 5 0 5 0 0 0 1 1 20 Bambusa vulgaris Poaceae (Bambusoideae) 0 0 0 0 2 0 0 0 0 0 0 0 21 Bersama abyssinica Melianthaceae 0 0 0 0 1 0 0 0 1 0 0 0 22 Bischofia javanica Euphorbiaceae 0 0 0 0 29 0 0 29 0 3 29 0 23 Bridelia micrantha Euphorbiaceae 0 7 8 5 86 0 0 2 2 0 1 0 24 Buddleja davidii Loganiaceae 0 55 0 0 55 0 0 0 0 24 0 0 25 Caesalpinia decapetala Fabaceae (Caesalpinioideae) 0 5 0 0 2 0 0 0 0 0 0 0 26 Cajanus cajan Fabaceae (Papilionoideae) 0 0 0 0 16 0 0 0 0 0 0 2 27 Calliandra calothyrsus Fabaceae (Mimosoideae) 0 0 0 0 12 12 0 0 0 0 1 2 28 Callistemon citrinus Myrtaceae 0 0 0 0 3 0 0 0 0 3 2 0 29 Camellia sinensis Theaceae 24 0 0 0 24 0 0 0 0 0 0 4 30 Carica papaya Caricaceae 0 0 0 0 0 0 129 0 0 1 5 8 31 Casimiroa edulis Rutaceae 0 0 0 0 19 0 19 1 0 1 5 1 32 Casuarina equisetifolia Casuarinaceae 0 0 0 0 4 0 0 2 0 1 4 0 33 Cassia spectabilis Fabaceae (Caesalpinioideae) 0 1 2 1 19 0 0 2 0 2 7 0 34 Chaetacme aristata Ulmaceae 0 0 0 0 1 0 0 0 0 0 0 0 35 Citrus limon Rutaceae 0 0 0 0 15 0 15 0 0 0 3 0 36 Citrus sinensis Rutaceae 0 0 1 0 12 0 12 0 0 0 2 0 37 Theobroma cacao Sterculiaceae 1 0 0 0 0 0 0 0 0 0 0 0 38 Coffea arabica Rubiaceae 48 0 0 0 48 0 0 0 0 0 1 8 39 Combretum molle Combretaceae 0 0 0 0 11 0 0 0 2 0 1 1 40 Cordia africana Boraginaceae 0 0 0 0 4 0 0 3 0 0 1 0 41 Croton macrostachyus Euphorbiaceae 0 2 7 1 138 0 0 27 19 4 41 1 42 Croton megalocarpus Euphorbiaceae 0 4 2 2 44 0 0 44 2 6 44 0 43 Cupressus lusitanica Cupressaceae 0 100 2 5 100 0 0 100 0 20 2 0 44 Cyphomandra crassicaulis Solanaceae 0 0 0 0 7 0 7 0 0 0 0 0 45 Delonix regia Fabaceae (Caesalpinioideae) 0 0 0 1 5 0 0 1 0 1 4 0 46 Dovyalis caffra Flacourtiaceae 0 9 0 0 9 0 0 0 0 0 0 0 47 Dracaena dracocephala Dracaenaceae 0 0 0 0 1 0 0 0 0 0 1 0 48 Dracaena fragrans Dracaenaceae 0 66 0 0 8 0 0 0 2 3 1 0 49 Dracaena steudneri Dracaenaceae 0 0 0 0 0 0 0 0 0 2 0 0 50 Entada abyssinica Fabaceae (Mimosoideae) 0 0 0 0 18 0 0 5 5 2 9 0 51 Eriobotrya japonica Rosaceae 0 0 2 0 103 0 103 1 0 0 23 0 52 Erythrina abyssinica Fabaceae (Papilionoideae) 0 2 0 0 2 0 0 0 2 0 0 1 53 Erythrococca atrovirens Euphorbiaceae 0 0 0 0 2 0 0 0 1 0 0 0 54 Erythrina caffra or excelsa Fabaceae (Papilionoideae) 0 9 0 0 9 1 0 0 0 1 0 0 55 Erythrina excelsa Fabaceae (Papilionoideae) 0 0 0 0 2 0 0 0 0 0 1 1 56 Eucalyptus spp. Myrtaceae 0 1 9 190 190 0 0 190 1 0 13 0 57 Euphorbia tirucalli Euphorbiaceae 0 124 0 0 124 0 0 0 0 0 0 0 58 unid. (fruit, thorny) - 0 0 0 0 1 0 1 0 0 0 0 0 59 Ficus benjamina Moraceae 0 0 0 0 1 0 0 0 0 0 1 0

APPENDIX I

287

No.

Spec

ies

Fam

ily

Beve

rage

Boun

dary

Cha

rcoa

l

Con

stru

ctio

n

Fire

woo

d

Fodd

er

Frui

t

Tim

ber

Med

icin

e

Orn

amen

tal

Shad

e

Soil

fert

ility

60 Ficus exasperata Moraceae 0 0 0 0 1 0 0 1 0 0 0 0 61 Ficus sycomorus Moraceae 0 0 0 0 4 0 0 1 0 0 0 0 62 Ficus thonningii Moraceae 0 2 1 0 18 0 0 3 1 1 7 0 63 Ficus verruculosa Moraceae 0 0 0 0 18 0 0 6 0 0 7 1 64 Funtumia africana Apocynaceae 0 0 1 0 14 0 0 6 1 0 3 1 65 Grewia bicolor Tiliaceae 0 0 0 0 1 0 0 0 0 0 0 0 66 Grevillea robusta Proteaceae 0 0 0 4 64 0 0 64 0 3 7 3 67 Harungana madagascariensis Clusiaceae/Guttiferae 0 0 0 3 85 0 0 0 37 1 3 0 68 Hymenodictyon floribundum Rubiaceae 0 0 0 0 1 0 0 0 0 0 0 0 69 Indigofera arrecta Fabaceae (Papilionoideae) 0 0 0 0 8 0 0 0 2 0 0 0 70 Indigofera trita Fabaceae (Papilionoideae) 0 0 0 0 2 0 0 0 2 0 0 0 71 Inga edulis Fabaceae (Mimosoideae) 0 0 0 0 1 0 2 0 0 0 0 0 72 Jacaranda mimosifolia Bignoniaceae 0 0 1 1 29 0 0 12 0 8 29 0 73 Jasminium abyssinicum Oleaceae 0 0 1 0 1 0 0 1 0 0 0 0 74 Kigelia africana Bignoniaceae 0 0 0 0 0 0 0 0 0 0 0 0 75 Lantana camara Verbenaceae 0 102 0 0 102 1 0 0 0 1 2 2 76 Leucas calostachys Lamiaceae/Labiatae 0 0 0 0 6 0 0 0 4 0 0 0 77 Leucaena leucocephala Fabaceae (Mimosoideae) 0 1 0 1 17 17 0 0 0 1 2 4 78 unid. (like Schinus molle) - 0 0 0 0 1 0 0 1 0 1 1 0 79 unid. (shiny leaves, probably Ficus) - 0 0 0 0 2 0 0 0 1 0 0 0 80 unid. - 0 0 0 0 1 0 0 0 0 1 1 0 81 Vernonia myriantha Asteraceae/Compositae 0 0 0 0 1 0 0 0 1 0 0 0 82 unid. - 0 0 0 0 1 0 0 0 0 0 1 0 83 unid. - 0 0 0 0 2 0 0 0 2 0 1 0 84 unid. (flowers) - 0 2 0 0 2 0 0 0 0 3 1 0 85 unid. - 0 0 0 0 1 0 0 0 1 0 0 0 86 unid. (probably Fabaceae) - 0 2 1 0 7 0 0 1 0 3 3 0 87 Bougainvillea sp. Nyctaginaceae 0 0 0 0 1 0 0 0 0 1 0 0 88 unid. - 0 0 0 0 1 0 0 0 0 0 0 0 89 unid. (flowers) - 0 0 0 0 1 0 0 0 0 1 1 0 90 unid. - 0 0 0 0 0 0 0 0 0 0 0 0 91 unid. (flowers, yellowish leaves) - 0 0 0 0 0 0 0 0 0 1 1 0 92 unid. (thorny) - 0 0 0 0 1 0 0 0 0 0 0 0 93 unid. (thorny) - 0 0 0 0 1 0 0 0 1 0 0 0 94 unid. - 0 0 0 0 0 0 0 0 1 0 0 0 95 unid. - 0 0 0 0 1 0 0 0 1 0 0 0 96 unid (shrubby) - 0 0 0 0 1 0 0 0 1 0 0 0 97 unid. (red) - 0 0 0 0 0 0 0 0 1 0 0 0 98 unid. (shiny leaves) - 0 0 0 0 1 0 0 0 0 0 0 0 99 unid. (flowers) - 0 0 0 0 1 0 0 0 0 1 1 0 100 unid. - 0 0 0 0 1 0 0 0 0 0 0 0 101 unid. - 0 0 0 0 1 0 0 0 0 1 0 0 102 unid. (thorny, propagated by leaf

cuttings) - 0 1 0 0 0 0 0 0 0 1 0 0

103 unid. - 0 0 0 0 1 0 0 0 0 0 0 0 104 Aningeria altissima Sapotaceae 0 0 0 0 1 0 0 0 0 0 0 0 105 unid. (probably Oleaceae) - 0 0 0 0 0 0 0 0 0 0 0 0 106 unid. - 0 0 0 0 1 0 0 0 0 0 0 0 107 unid. Sapindaceae 0 0 0 0 4 0 0 0 0 0 0 0 108 unid. - 0 0 0 0 2 0 0 1 0 2 0 0 109 unid. - 0 0 0 0 1 0 0 0 1 0 0 0 110 unid. - 0 0 0 0 1 0 0 0 0 0 0 0 111 unid. (maybe Parkia) - 0 0 0 0 1 0 0 1 0 0 0 0 112 unid. (Cactaceae or Euphorbiaceae) - 0 1 0 0 1 0 0 0 0 0 0 1 113 Pavetta ternifolia Rubiaceae 0 0 0 0 4 0 0 0 0 0 0 0 114 unid. ("nightrose") - 0 6 0 0 38 0 0 0 0 18 13 0 115 unid. - 0 0 0 0 10 0 10 0 0 0 5 0 116 unid. (maybe Albizia) - 0 0 0 0 1 0 0 0 0 1 1 0 117 unid. - 0 0 0 0 1 0 0 0 0 0 0 0 118 unid. - 0 0 0 0 1 0 0 0 0 0 1 0 119 unid. - 0 0 0 0 0 0 0 0 1 0 0 0 120 unid. - 0 0 0 0 1 0 0 1 0 0 1 0

APPENDIX I

288

No.

Spec

ies

Fam

ily

Beve

rage

Boun

dary

Cha

rcoa

l

Con

stru

ctio

n

Fire

woo

d

Fodd

er

Frui

t

Tim

ber

Med

icin

e

Orn

amen

tal

Shad

e

Soil

fert

ility

121 Parinari curatellifolia Chrysobalanaceae 0 0 0 0 1 0 0 0 0 0 1 0 122 unid. - 0 0 0 0 1 0 0 0 1 0 0 0 123 unid. - 0 0 0 0 1 0 0 0 1 0 0 0 124 unid. (red flowers, no thorns) - 0 0 0 0 2 0 0 0 1 2 1 0 125 unid. - 0 0 0 0 1 0 0 0 0 0 1 0 126 unid. (maybe Bischofia) - 0 0 0 0 2 0 0 0 0 0 2 0 127 unid. (flowers) - 0 1 0 0 5 0 0 0 0 4 2 0 128 Macadamia tetraphylla Proteaceae 0 0 0 0 1 0 1 0 0 0 0 0 129 Maesopsis eminii Rhamnaceae 0 0 1 0 20 0 0 20 0 0 5 0 130 Malus spp. Rosaceae 0 0 0 0 1 0 1 0 0 0 0 0 131 Mangifera indica Anacardiaceae 0 0 4 0 151 0 151 0 4 2 63 0 132 Markhamia lutea Bignoniaceae 0 5 54 162 162 0 0 38 0 2 11 0 133 Azadirachta indica Meliaceae 0 0 0 0 43 0 0 0 43 0 7 0 134 Morus alba Moraceae 0 0 0 0 9 0 9 0 0 1 0 0 135 Olea capensis Oleaceae 0 0 0 0 1 0 0 1 0 0 0 0 136 Parkia eminii Fabaceae (Mimosoideae) 0 0 0 0 2 0 0 2 0 0 0 0 137 unid. (Parkia) Fabaceae (Mimosoideae) 0 0 0 0 0 0 0 0 0 1 0 0 138 Persea americana Lauraceae 0 1 0 0 192 7 192 4 0 0 21 1 139 Periploca linearifolia Asclepiadaceae 0 0 0 0 2 0 0 0 2 0 0 0 140 Phoenix reclinata Arecaceae/Palmae 0 0 0 0 0 0 0 0 0 0 0 0 141 Piliostigma thonningii Fabaceae (Caesalpinioideae) 0 0 0 0 9 0 4 0 1 1 3 0 142 Pinus patula Pinaceae 0 0 0 1 12 0 0 12 0 5 6 0 143 Plectranthus barbatus Lamiaceae/Labiatae 0 0 0 0 0 0 0 0 0 0 0 0 144 Plumeria rubra Apocynaceae 0 0 0 0 0 0 0 0 1 2 2 0 145 Polyscias fulva Araliaceae 0 2 0 0 16 0 0 16 0 0 3 0 146 Premna angolensis Verbenaceae 0 0 0 0 2 0 0 0 1 0 1 0 147 Prunus africana Rosaceae 0 0 2 0 14 0 0 7 1 0 2 0 148 Psidium guajava Myrtaceae 0 11 5 2 183 0 183 0 0 0 21 0 149 Ricinus communis Euphorbiaceae 0 0 0 0 31 0 0 0 0 2 1 15 150 Sapium ellipticum Euphorbiaceae 0 0 2 1 16 0 0 1 2 0 0 0 151 Schinus molle Anacardiaceae 0 0 0 0 0 0 0 0 1 0 0 0 152 Senna didymobotrya Fabaceae (Caesalpinioideae) 0 0 0 0 18 0 0 0 18 1 0 2 153 Sesbania macrantha Fabaceae (Papilionoideae) 0 0 0 0 9 1 0 0 0 0 0 3 154 Sesbania sesban Fabaceae (Papilionoideae) 0 2 0 3 113 14 0 0 2 0 4 113 155 Solanum aculeastrum Solanaceae 0 11 0 0 11 0 0 0 5 1 0 0 156 Spathodea campanulata Bignoniaceae 0 1 8 1 55 0 0 20 6 8 17 1 157 Steganotaenia araliacea Umbelliferae 0 0 0 0 4 0 0 0 11 0 3 0 158 Syzygium cumini Myrtaceae 0 0 2 2 71 0 71 71 1 0 20 1 159 Tamarindus indica Fabaceae (Caesalpinioideae) 0 0 0 0 1 0 1 0 0 0 1 0 160 Teclea nobilis Rutaceae 0 5 2 1 33 0 0 0 14 1 9 0 161 Tephrosia nyikensis Fabaceae (Papilionoideae) 0 0 0 0 1 0 0 0 0 0 0 0 162 Tephrosia vogelii Fabaceae (Papilionoideae) 0 0 0 0 23 0 0 0 0 0 0 10 163 Terminalia mantaly Combretaceae 0 0 0 0 40 0 0 3 0 40 40 0 164 Terminalia mollis Combretaceae 0 0 0 0 1 0 0 1 0 0 0 0 165 Thevetia peruviana Apocynaceae 0 1 0 0 8 0 0 0 0 7 5 0 166 Tithonia diversifolia Asteraceae/Compositae 0 21 0 0 21 0 0 0 1 0 0 7 167 Trema orientalis Ulmaceae 0 0 0 0 4 0 0 1 0 0 0 0 168 Trichilia dregeana Meliaceae 0 0 0 0 1 0 0 1 0 0 1 0 169 Trichilia emetica Meliaceae 0 0 0 0 5 0 0 2 3 0 1 0 170 Vangueria apiculata Rubiaceae 0 0 0 0 22 0 21 0 0 0 9 0 171 Vernonia amygdalina Asteraceae/Compositae 0 0 0 0 7 0 0 0 2 0 0 1 172 Vernonia auriculifera Asteraceae/Compositae 0 0 0 0 1 0 0 0 1 0 0 0 173 Vernonia holstii Asteraceae/Compositae 0 0 0 0 1 0 0 0 0 0 0 0 174 Vitex keniensis Verbenaceae 0 0 0 0 6 0 0 4 2 1 0 0 175 Zanthoxylum gillettii Rutaceae 0 0 0 0 45 0 0 45 13 2 22 0

APPENDIX I

289

Table I.2. Characteristics of farms and households (see tables 3.2 and 3.3) No. Village Type of

household head

Farm size (acres)

Type of house

Number of crossbred cattle


Number of years household head

Maximum age of head or partner

Number of children

Level of schooling

1* Ebuchiebe female de facto 0.5 permanent 0 4 17 38 4 6 2* Ebuchiebe male 2 thatch roof 0 1 3 33 5 5 3* Ebuchiebe female de facto 2 permanent 0 2 27 43 6 4 4* Ebuchiebe male 0.75 thatch roof 0 1 29 49 3 8 5* Ebuchiebe male 2 thatch roof 0 0 7 46 5 6 6* Ebuchiebe male 0.5 iron roof 0 0 22 40 5 7 7* Ebuchiebe male 0.5 iron roof 0 3 15 47 6 5 8* Ebuchiebe female de facto 1 permanent 0 0 21 51 6 7 9* Ebuchiebe male 1 iron roof 0 1 23 55 10 7 10* Ebuchiebe male 2 iron roof 0 4 14 76 7 1 11* Ebuchiebe male 1 iron roof 0 4 37 73 7 2 12* Ebuchiebe female de jure 1 thatch roof 0 0 2 30 4 7 13* Ebuchiebe female de facto 1.5 permanent 0 2 8 34 4 10 14* Ebuchiebe male 1 thatch roof 0 2 44 44 0 2 15* Ebuchiebe female de facto 0.75 permanent 0 2 19 45 7 6 16* Ebuchiebe male 0.5 iron roof 0 0 15 59 0 1 17* Ebuchiebe female de jure 0.75 iron roof 0 1 1 68 8 5 18* Ebuchiebe male 0.5 iron roof 0 2 23 56 6 5 19* Ebuchiebe female de facto 0.5 iron roof 0 0 25 48 7 6 20* Ebuchiebe female de jure 0.75 iron roof 0 0 2 29 3 8 21* Ebuchiebe male 1 iron roof 0 5 48 70 7 0 22* Ebuchiebe female de jure 1.5 iron roof 0 0 24 54 8 1 23* Ebuchiebe male 1.5 thatch roof 0 0 12 26 1 5 24* Ebuchiebe male 1.5 iron roof 0 3 3 55 9 3 25* Ebuchiebe male 5 permanent 4 2 50 73 5 10 26* Ebuchiebe female de jure 4 iron roof 0 0 13 62 7 1 27* Ebuchiebe male 1 iron roof 0 0 10 38 1 7 28* Ebuchiebe male 1.5 iron roof 0 2 18 50 5 4 29* Ebuchiebe male 1 thatch roof 0 2 1 45 5 10 30* Ebuchiebe male 4 iron roof 0 4 10 66 0 5 31* Ebuchiebe female de facto 1 iron roof 0 2 30 55 7 3 32* Ebuchiebe male 1.5 iron roof 0 0 5 61 4 4 33* Ebuchiebe female de jure 3.5 iron roof 0 2 6 53 5 0 34* Ebuchiebe female de jure 1 thatch roof 0 0 34 52 3 3 35* Ebuchiebe female de facto 0.75 iron roof 0 0 8 50 3 3 36* Ebuchiebe male 1 thatch roof 0 0 1 40 2 6 37* Ebuchiebe female de facto 1 iron roof 0 0 1 34 4 7 38* Ebuchiebe male 1 iron roof 0 3 23 55 8 8 39* Ebuchiebe female de facto 0.75 permanent 0 4 16 44 4 12 40* Ebuchiebe female de facto 1.5 iron roof 0 1 55 58 7 4 41* Ebuchiebe female de facto 1 iron roof 0 1 18 61 5 2 42* Ebuchiebe male - iron roof 0 4 52 70 1 0 43* Ebuchiebe male 0.5 iron roof 0 0 1 46 7 4 44* Ebuchiebe male 1 thatch roof 0 0 9 38 3 6 45* Ebuchiebe male 1 thatch roof 0 0 0 48 7 8 46 Ebuchiebe female de facto 0.5 thatch roof 0 0 1 27 2 8 47* Ebuchiebe male 2 thatch roof 3 0 53 53 6 3 48* Ebuchiebe male 0.25 thatch roof 0 0 1 19 1 4 49* Ebuchiebe female de facto 1 iron roof 0 2 14 67 4 0 50* Ebuchiebe female de jure - thatch roof 0 0 28 79 1 1 51* Shimutu male 4 thatch roof 0 1 23 76 0 0 52* Shimutu male 1.5 permanent 0 1 52 52 4 8 53* Shimutu female de jure 2 iron roof 0 1 0 59 7 0 54 Shimutu male 2.5 iron roof 0 1 58 70 5 3 55 Shimutu male 1.5 thatch roof 0 1 3 36 3 12 56 Shimutu male 2 iron roof 0 5 31 81 2 0 57 Shimutu male 2 thatch roof 0 0 37 51 0 4 58* Shimutu male 3 iron roof 0 8 35 78 6 0 59 Shimutu male 2 thatch roof 0 0 40 40 5 4 60 Shimutu male 1.5 iron roof 0 2 1 50 8 7 61* Shimutu female de facto 1.5 thatch roof 0 0 29 52 7 8 62* Shimutu male 1.5 iron roof 0 0 13 68 0 4 63 Shimutu male 3 iron roof 0 6 60 60 3 0 64 Shimutu male 5 iron roof 0 10 - 78 0 - 65* Shimutu female de jure 1.5 thatch roof 0 0 - 78 2 0 66 Shimutu male 2.5 thatch roof 0 0 28 - 4 0 67* Shimutu male 2.5 iron roof 0 0 15 33 1 7 68 Shimutu female de jure 4 iron roof 0 2 43 59 2 0 69 Shimutu male 0.5 iron roof 0 4 8 43 7 6 70* Shimutu male 3 iron roof 0 1 15 72 2 3 71* Shimutu male 2 thatch roof 0 5 18 63 3 4

APPENDIX I

290

No. Village Type of household head

Farm size (acres)

Type of house





Number of children

Level of schooling

72* Shimutu female de facto 2 thatch roof 0 1 - - 2 0 73* Shimutu male 2 iron roof 0 0 19 56 1 0 74* Shimutu female de jure 1 thatch roof 0 0 2 42 4 0 75 Shimutu male 3 thatch roof 0 1 4 64 7 12 76 Shimutu male 1 thatch roof 0 0 22 48 6 4 77* Shimutu male 2 iron roof 0 0 22 86 5 0 78* Shimutu male 2.5 iron roof 0 0 7 34 3 7 79 Shimutu male 2.5 iron roof 0 3 17 37 5 2 80* Shimutu male 3 iron roof 0 5 31 74 4 0 81 Shimutu female de jure 3 iron roof 0 1 45 65 1 0 82 Shimutu male 2 iron roof 0 2 13 31 3 3 83 Shimutu male 2 iron roof 0 2 8 44 5 0 84* Shimutu male 3 iron roof 0 3 13 50 1 7 85 Shimutu male 2.5 iron roof 0 0 18 - 5 4 86 Shimutu male 2.5 iron roof 1 1 39 66 7 0 87* Shimutu male 3.5 permanent 0 10 16 62 4 4 88 Shimutu male 3 iron roof 0 0 2 54 8 4 89 Shimutu female de jure 3.5 thatch roof 0 3 61 - 2 0 90 Shimutu male 2 thatch roof 0 4 5 59 2 5 91* Shimutu female de jure 2.5 thatch roof 0 0 35 58 3 0 92 Shimutu male 2.5 iron roof 0 2 9 41 1 5 93* Shimutu male 2 iron roof 0 2 78 82 1 8 94* Shimutu male 2.5 thatch roof 0 3 38 57 6 5 95* Shimutu female de facto 2 iron roof 0 1 14 40 5 2 96 Shimutu male 2 thatch roof 0 1 38 54 1 0 97* Shimutu female de facto 2 thatch roof 0 0 24 48 6 4 98 Shimutu male 1 iron roof 0 0 0 41 1 12 99 Shimutu female de facto 1.5 iron roof 0 3 23 50 4 0 100 Shimutu male 2.5 permanent 0 1 21 71 4 3 101 Shimutu male 3.5 permanent 0 10 16 62 4 4 102 Mutambi male 5 permanent 1 6 45 61 8 10 103 Mutambi male 1.5 iron roof 0 0 45 45 3 12 104 Mutambi male 1.5 iron roof 0 2 11 52 8 10 105 Mutambi female de facto 1.5 iron roof 0 2 19 55 7 6 106 Mutambi female de jure 0.5 thatch roof 0 2 15 - 2 4 107* Mutambi male 0.5 thatch roof 0 0 69 70 0 0 108 Mutambi female de jure 0.75 iron roof 0 3 29 50 2 8 109 Mutambi male 0.5 thatch roof 0 1 10 30 4 0 110 Mutambi male 0.5 iron roof 0 2 0 44 6 7 111* Mutambi female de facto 0.75 iron roof 2 1 - 45 4 - 112 Mutambi male 1.5 iron roof 2 0 2 36 3 10 113* Mutambi male 0.5 iron roof 0 4 21 71 4 4 114 Mutambi male 1.75 iron roof 0 1 5 41 5 5 115* Mutambi male 0.5 thatch roof 0 1 18 - 1 0 116 Mutambi female de jure 1.5 iron roof 0 1 40 58 4 6 117 Mutambi female de jure 1 iron roof 0 3 47 56 2 1 118 Mutambi male 0.75 iron roof 0 0 15 54 1 8 119 Mutambi male 0.75 iron roof 0 4 4 61 1 12 120 Mutambi female de facto 0.5 iron roof 0 2 14 47 5 7 121* Mutambi male 0.75 iron roof 0 3 4 70 7 5 122 Mutambi male 5.5 iron roof 1 0 23 54 0 0 123* Mutambi male 1 permanent 1 1 3 63 5 11 124 Mutambi female de jure 0.5 iron roof 0 3 45 68 1 0 125 Mutambi female de jure 1 iron roof 0 2 55 74 5 2 126 Mutambi male 0.25 iron roof 0 3 26 49 6 6 127 Mutambi male 0.75 permanent 1 2 33 83 1 0 128 Mutambi male 1.25 permanent 2 0 19 44 9 12 129 Mutambi female de facto 1 iron roof 0 3 34 59 11 4 130 Mutambi male 0.75 thatch roof 0 0 33 58 1 8 131 Mutambi male 1.25 iron roof 1 2 0 45 8 8 132 Mutambi female de jure 1 thatch roof 0 1 49 68 2 0 133 Mutambi male - iron roof 0 2 10 45 9 0 134 Mutambi male 1.5 iron roof 0 3 14 77 2 0 135 Mutambi female de jure 0.25 iron roof 0 2 33 49 4 7 136 Mutambi female de jure 1.25 permanent 0 0 - 80 1 0 137 Mutambi female de facto 0.75 iron roof 0 3 3 25 2 4 138 Mutambi male 0.75 thatch roof 0 2 2 64 5 0 139 Mutambi female de facto 0.25 iron roof 1 0 17 39 8 7 140 Mutambi female de jure 0.25 iron roof 0 3 52 71 0 2 141 Mutambi male 0.5 permanent 0 0 15 58 7 8 142 Mutambi female de jure 1.5 thatch roof 0 2 44 55 3 4 143 Mutambi female de jure 0.5 thatch roof 0 2 29 80 1 0 144 Mutambi female de facto 1 permanent 2 0 19 46 5 11

APPENDIX I

291

No. Village Type of household head

Farm size (acres)

Type of house





Number of children

Level of schooling

145 Mutambi female de facto 0.25 iron roof 0 1 8 34 2 9 146 Mutambi female de facto 0.25 thatch roof 0 0 14 42 5 6 147 Mutambi female de jure 0.25 iron roof 0 0 49 68 1 1 148 Mutambi male 0.25 iron roof 0 0 33 67 1 10 149 Mutambi female de facto 3 permanent 4 2 28 61 7 1 150 Mutambi male 1 iron roof 0 0 40 59 5 12 151 Mutambi male 2.25 iron roof 1 2 7 63 12 8 152 Madidi male 0.25 thatch roof 0 2 9 65 8 3 153 Madidi female de facto 0.25 iron roof 0 2 17 39 8 6 154 Madidi male 1 iron roof 0 0 14 55 0 7 155 Madidi female de jure 0.75 iron roof 0 4 43 60 4 4 156 Madidi male 0.25 thatch roof 0 1 4 30 1 12 157 Madidi male 1 thatch roof 0 0 1 29 0 7 158 Madidi male 0.5 iron roof 0 1 10 34 5 10 159 Madidi male 3 permanent 0 3 20 78 0 8 160 Madidi male 0.25 iron roof 0 1 11 72 3 3 161 Madidi male 1.5 iron roof 0 2 44 44 5 7 162 Madidi male 2.5 iron roof 0 4 9 66 4 10 163 Madidi male 0.5 iron roof 0 3 8 81 0 4 164 Madidi male 0.25 thatch roof 0 0 3 42 5 10 165 Madidi female de facto 0.75 iron roof 0 2 23 52 4 7 166 Madidi male 0.5 thatch roof 0 3 28 68 5 2 167 Madidi male 1 iron roof 0 0 3 30 4 5 168 Madidi female de facto 1 iron roof 0 2 25 53 8 0 169 Madidi male 1 iron roof 0 1 2 52 4 0 170 Madidi female de facto 0.5 iron roof 0 1 24 47 4 7 171 Madidi female de jure 2 iron roof 0 2 24 43 6 7 172 Madidi female de facto 0.5 thatch roof 0 1 10 38 5 8 173 Madidi male 2 permanent 5 1 1 60 5 10 174 Madidi male 0.5 iron roof 0 2 16 49 3 7 175 Madidi male 0.5 iron roof 0 2 1 67 6 8 176 Madidi male 2 iron roof 0 1 18 73 2 6 177 Madidi male 2 iron roof 0 2 8 45 3 12 178 Madidi female de facto 0.5 iron roof 0 1 13 39 3 5 179 Madidi male 1.5 thatch roof 0 1 57 57 8 8 180 Madidi female de jure 2 iron roof 0 0 66 82 1 0 181 Madidi female de facto 0.5 iron roof 0 1 26 49 6 7 182 Madidi male 2 thatch roof 0 4 24 65 3 2 183 Madidi female de jure 1.5 iron roof 0 1 47 63 6 1 184 Madidi male 1.25 iron roof 0 1 - 61 3 4 185 Madidi female de facto 0.75 iron roof 0 1 5 44 6 7 186 Madidi female de jure 0.5 iron roof 0 0 34 51 7 6 187 Madidi female de facto 0.75 iron roof 0 4 30 54 2 4 188 Madidi female de facto 0.25 thatch roof 0 2 43 63 3 0 189 Madidi male 1 iron roof 0 3 51 73 5 2 190 Madidi female de facto 0.5 iron roof 0 2 32 53 6 2 191 Madidi male 0.5 thatch roof 0 0 - 38 3 4 192 Madidi male 1 iron roof 0 2 52 54 2 7 193 Madidi female de jure 0.25 thatch roof 0 0 11 73 4 0 194 Madidi female de facto 0.75 iron roof 0 2 14 43 5 10 195 Madidi female de jure 3 thatch roof 0 0 - 51 6 0 196 Madidi male 1.25 iron roof 0 2 3 77 2 0 197 Madidi male 0.5 iron roof 0 3 13 37 5 10 198 Madidi male 1 thatch roof 0 0 4 37 4 10 199 Madidi female de facto 1 iron roof - - 26 49 9 4 200 Madidi female de facto 1 iron roof 0 0 1 40 3 10 201 Madidi female de facto 1.5 iron roof 0 2 3 44 1 6 * = included in second survey (see Chapter 3)

Next pages: Table I.3. Cross-tabulation of farms (rows, codes from Table I.2) and species (columns, codes from Table I.1). Cell values indicate abundance of each species on each farm.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 60 0 0 4 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 2 1 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 18 5 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0 0 0 2 0 0 2 0 0 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 3 0 0 9 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 1 0 0 0 1

10 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 17 0 0 11 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 5 0 0 0 0 1 12 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 13 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 14 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 15 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 4 0 4 0 0 0 16 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 1 0 0 0 0 17 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 1 0 0 18 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 4 3 0 21 0 0 19 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 21 0 0 0 0 0 0 1 0 0 2 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 22 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 1 23 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 2 0 0 0 0 0 24 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 1 0 0 0 0 25 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 2 2 0 0 0 0 0 0 0 0 0 0 0 0 26 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 27 0 0 0 0 0 0 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 28 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 29 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 2 0 0 0 0 0 30 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 0 4 0 1 0 0 0 31 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 2 0 0 0 32 0 0 0 0 0 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 1 0 0 33 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 10 0 0 0 0 0 0 3 0 0 10 0 0 34 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 9 0 0 0 0 0 0 0 0 0 0 0 0 35 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 36 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 37 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 2 2 0 0 0 0 0 0 6 0 0 27 0 0 38 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 1 2 0 1 0 0 39 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 1 0 0 0 0 40 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 41 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 5 0 0 0 0 0 42 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 43 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 44 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 2 1 0 0 0 1 45 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 47 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 48 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 49 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 11 0 0 0 0 0 0 3 2 0 0 0 0 51 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 1 11 0 0 0 0 0 0 14 0 0 0 0 0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

52 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 2 53 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 54 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 2 0 0 0 0 0 55 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 56 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 0 10 0 0 0 0 0 57 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 2 4 0 0 0 0 0 0 0 0 0 0 0 0 58 0 0 17 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 3 0 0 0 0 0 59 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 60 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 61 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 2 0 0 0 0 0 62 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 1000 3 0 0 0 0 0 63 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 64 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 3 0 0 0 0 0 65 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 66 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 0 1 0 0 0 0 0 67 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50 0 0 0 0 0 0 1 0 0 0 0 0 68 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 69 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 70 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 71 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 4 0 0 0 0 0 0 3 0 0 0 0 0 72 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 73 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 4 0 0 0 0 0 74 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 75 0 0 0 0 0 0 0 2 0 1 0 0 0 0 0 0 0 0 0 0 0 1 3 0 0 0 0 0 0 1 0 0 0 0 0 76 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 10 0 0 0 0 1 77 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 78 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 79 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 11 0 0 0 0 0 0 4 0 0 0 0 0 80 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 19 0 0 3 0 0 0 0 0 81 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 0 1 0 0 0 0 0 82 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 5 0 0 0 0 0 83 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 84 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 85 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 0 0 0 0 0 0 0 0 86 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 4 0 0 0 0 0 87 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 11 0 0 0 0 0 0 0 0 0 0 0 0 88 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 300 0 0 0 0 1 0 89 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 0 0 1 0 0 1 0 0 0 0 0 90 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 21 0 0 0 0 0 0 2 0 0 0 0 0 91 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 1 0 0 1 92 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 100 0 0 0 0 0 0 0 4 4 0 0 0 0 93 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 94 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 49 0 0 0 0 0 95 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1000 2 0 0 0 0 0 96 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 97 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 4 0 0 0 0 0 98 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 99 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0

100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 101 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 102 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 1 0 0 0 0 3000 0 0 0 0 0 0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

103 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 104 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 3 1 0 0 0 0 0 0 0 0 0 0 0 105 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 50 0 0 0 0 0 3 1 0 0 0 0 106 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 107 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 108 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 210 0 0 0 0 0 3 0 0 0 0 0 109 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 50 0 0 0 0 0 20 0 0 0 0 0 110 0 0 0 0 0 0 0 0 0 1 3 0 0 0 0 0 0 0 0 0 0 0 2 90 4 0 0 0 25 154 0 0 0 0 0 111 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 50 0 0 0 0 0 1 0 0 0 0 0 112 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 22 0 0 0 0 2000 1 0 0 0 0 0 113 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 6 0 0 0 0 0 114 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 40 0 0 0 0 0 2 0 0 0 0 0 115 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 116 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 21 0 0 0 1 0 0 3 0 0 1 0 0 117 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 12 0 0 0 0 0 3 0 0 0 0 0 118 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 119 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 500 0 0 0 0 0 1 0 0 0 0 0 120 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 22 0 0 0 0 0 0 2 0 0 0 0 0 121 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 122 0 0 0 0 0 0 0 0 0 2 3 0 0 0 0 0 0 0 0 0 10 0 50 0 0 0 0 0 200 0 0 0 0 0 0 123 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 124 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 16 0 0 12 0 0 125 0 0 0 0 0 0 0 0 0 2 1 0 0 0 0 0 0 0 0 0 0 0 1 100 0 0 0 0 1000 1 0 0 0 0 1 126 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 200 0 0 0 0 200 1 0 0 0 0 0 127 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1000 1 0 0 0 0 0 128 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 5 0 50 0 0 0 0 0 0 0 0 0 0 129 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 60 0 0 0 0 850 1 0 0 0 0 0 130 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 5 8 0 0 0 0 1500 4 0 0 0 0 0 131 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 74 0 0 0 0 1000 1 0 0 0 0 0 132 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 30 0 0 0 0 0 0 0 0 0 0 0 133 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 60 0 0 0 0 0 2 0 0 1 0 0 134 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0 1 0 0 0 0 0 135 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 200 0 0 0 0 0 0 0 0 0 0 0 136 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 200 0 0 0 0 0 0 137 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 100 0 0 0 0 200 0 0 0 0 0 0 138 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 8 102 0 2 0 0 0 2 0 0 0 0 0 139 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 90 0 0 0 0 0 0 0 0 0 0 0 140 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 30 0 0 0 0 0 3 0 1 0 0 1 141 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 30 0 0 0 0 0 1 0 0 0 0 0 142 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 143 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 3 0 0 0 0 0 144 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 300 0 0 0 0 0 0 145 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 16 0 1 0 0 0 0 0 0 0 0 0 146 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 147 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 148 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 200 2 0 0 0 0 0 149 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 30 0 0 0 0 1000 0 0 0 0 0 0 150 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 13 0 0 0 0 0 0 0 0 0 0 0 0 151 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 5 50 0 1 0 0 0 4 0 0 1 0 0 152 0 0 0 0 0 0 2 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 2 0 0 2 0 0 0 2 2 0 0 0 1 153 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 0 0 0 0 0 0 4 0 0 0 0 0

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

154 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 3 0 0 0 1 0 0 0 0 155 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 3 0 0 0 0 1 156 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 157 0 0 0 0 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 50 0 0 10 0 0 0 1 0 0 1 0 0 158 0 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 159 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 18 0 0 0 0 0 0 0 1 0 1 0 0 160 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 3 0 0 0 1 0 0 0 0 1 161 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 2 0 0 0 0 0 0 1 0 0 0 0 0 162 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 320 1 0 0 0 0 0 163 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 90 5 0 0 0 0 0 164 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 35 0 0 0 0 0 0 0 0 0 0 0 165 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 18 0 0 0 0 0 8 0 0 0 0 0 166 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 11 0 0 0 0 2 5 0 0 0 0 0 167 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 100 0 0 0 0 0 0 0 0 0 0 1 168 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50 0 0 0 0 0 1 0 0 0 0 0 169 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 250 0 0 0 0 0 0 0 0 0 0 0 170 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 400 0 0 0 0 0 0 0 0 0 0 0 171 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 320 0 0 30 1 0 0 1 1 0 0 0 172 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 6 0 0 0 5 0 0 0 0 0 173 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 174 0 0 0 0 0 0 0 0 20 1 0 0 0 0 0 0 0 0 0 0 0 0 0 100 0 0 0 0 4 0 0 0 0 0 1 175 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 100 0 0 0 0 0 0 0 0 2 0 0 176 0 0 0 0 0 0 0 0 40 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 6 0 0 0 0 0 177 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 150 0 10 0 0 0 3 1 0 0 0 0 178 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 1 0 0 0 0 0 179 0 0 0 0 0 0 0 0 30 1 0 0 0 0 0 0 0 0 0 0 0 1 2 0 0 0 2 0 0 4 1 0 0 0 0 180 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 100 0 6 0 0 0 1 2 0 0 0 0 181 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 90 0 0 0 0 0 0 0 0 0 0 0 182 1 1 1 0 0 0 0 0 200 1 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 12 0 0 3 0 0 0 0 0 0 183 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 2 0 0 184 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 1 0 0 1 0 0 3 0 0 0 0 0 185 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 16 2 0 0 0 0 2 0 0 0 0 0 186 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 150 0 0 0 0 0 3 0 0 0 0 0 187 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 20 0 0 0 0 0 2 0 0 0 0 0 188 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 189 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 45 6 0 0 0 0 2 0 0 0 0 0 190 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 6 0 0 0 0 0 191 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 1 0 0 0 0 0 192 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 15 0 1 0 3 0 0 0 0 0 0 0 193 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50 0 0 9 0 0 4 0 0 0 0 0 194 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 30 0 36 0 0 0 1 0 0 0 0 0 195 100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 196 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 197 0 0 0 0 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 1 0 36 0 0 0 0 0 0 0 0 0 0 0 198 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 12 0 60 0 0 0 2 0 0 0 0 0 199 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 101 0 0 1 0 0 0 0 0 1 0 0 200 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 201 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0

36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

1 0 0 0 0 0 1 0 7 0 0 0 0 0 0 1 0 0 0 0 0 31 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 16 100 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 4 0 0 0 0 10 392 0 0 0 0 0 0 0 0 2 0 0 0 0 4 1 0 0 0 0 5 0 3 0 0 0 0 0 0 0 0 0 0 0 0 111 326 0 0 0 0 0 0 0 0 2 1 0 0 0 5 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 3 0 0 0 0 127 121 0 0 0 0 0 0 0 0 4 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 12 130 0 0 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 0 5 0 0 0 1 0 1 0 0 0 3 0 0 0 0 33 1040 0 0 0 0 0 2 0 0 1 0 0 0 0 8 0 0 0 0 0 0 0 39 0 0 0 0 0 0 0 0 2 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 9 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 190 500 0 0 0 0 0 0 0 0 0 0 0 0 0

10 0 0 0 1 0 0 0 1 0 0 0 0 0 0 5 0 0 0 0 0 143 254 0 0 0 0 1 0 0 0 1 0 0 0 0 11 4 0 0 0 0 2 0 14 0 0 0 0 1 0 0 2 0 0 0 0 59 234 0 0 0 0 0 0 0 0 11 0 0 0 0 12 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 64 500 0 0 0 0 0 0 0 0 5 0 0 0 0 13 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 89 404 0 0 0 0 0 0 0 0 0 0 0 0 0 14 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 19 360 0 0 0 0 0 0 0 0 0 0 0 0 0 15 0 0 0 0 0 2 1 1 0 0 0 0 0 0 0 0 0 0 0 0 194 25 0 0 0 0 0 0 0 0 0 0 0 0 0 16 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 27 60 0 0 0 0 0 0 0 0 4 0 0 0 0 17 0 0 0 0 0 3 0 0 0 0 0 0 0 0 1 2 0 0 0 0 7 30 0 0 0 0 0 1 0 0 13 1 0 0 0 18 0 0 0 0 0 1 1 3 0 0 0 0 0 0 0 1 0 0 0 0 72 1 0 0 1 0 0 0 0 0 8 1 0 0 0 19 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 60 0 0 0 0 0 0 0 0 0 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 93 230 0 0 0 0 0 0 0 0 0 0 0 0 0 21 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 1 0 0 0 0 223 100 0 0 0 0 1 1 0 0 1 0 0 0 0 22 0 0 0 0 0 2 0 1 0 3 0 0 0 0 1 6 0 0 0 0 11 30 0 0 0 0 0 0 0 0 1 0 0 0 0 23 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 9 300 0 0 0 0 0 0 0 0 1 1 0 0 0 24 0 0 0 0 0 4 2 22 0 0 0 0 0 0 0 1 0 0 0 0 120 520 0 0 0 0 0 0 0 0 1 0 0 0 0 25 0 0 0 0 0 5 0 2 0 0 1 0 3 0 0 0 1 0 0 2 116 61 0 0 0 0 1 0 0 0 3 0 0 0 0 26 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 3 0 0 0 56 50 0 0 0 0 0 0 0 0 0 0 0 0 0 27 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 51 501 0 0 0 0 0 1 0 0 16 0 0 0 0 28 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 41 1 0 0 0 0 0 0 0 0 0 0 0 0 0 29 0 0 0 0 0 2 0 0 0 0 0 0 1 0 1 2 0 0 0 0 12 400 0 0 0 0 0 0 0 0 6 0 0 20 0 30 0 0 0 2 0 1 12 202 1 0 0 0 0 0 0 2 1 0 0 0 85 501 0 1 0 0 0 0 0 0 1 0 0 0 0 31 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 87 201 0 0 0 0 0 0 0 0 0 0 0 0 0 32 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 45 1 0 0 0 0 0 1 0 0 1 0 0 0 0 33 0 0 0 0 0 0 0 2 0 0 0 0 60 0 0 1 1 0 0 0 35 360 0 0 0 0 0 0 0 0 0 2 0 0 0 34 0 0 0 0 0 2 0 0 0 0 0 0 1 0 2 0 0 0 0 0 2 35 0 0 0 0 0 0 0 0 4 1 0 0 0 35 0 0 0 0 0 11 1 0 0 0 0 0 0 0 0 0 0 0 0 0 26 200 0 0 0 0 0 0 0 0 2 0 0 0 0 36 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 12 0 0 0 0 0 0 0 0 0 1 0 0 0 0 37 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 3 0 0 1 17 60 0 0 0 0 0 0 0 0 0 0 0 0 0 38 0 0 0 0 0 3 0 0 0 0 0 0 0 0 1 1 0 0 0 0 126 1 0 0 0 0 0 1 0 0 5 0 0 0 0 39 1 0 0 0 0 0 0 503 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 1 0 0 0 0 40 0 0 0 1 0 1 0 0 0 0 0 0 0 0 1 2 0 0 0 0 28 50 0 0 0 0 0 1 0 0 0 0 0 0 0 41 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 16 9 0 0 0 0 0 0 0 0 0 2 0 0 0 42 0 0 0 0 0 3 4 0 0 0 0 0 0 0 0 1 0 0 0 0 42 80 0 0 0 0 1 1 0 0 0 0 0 0 0 43 0 0 0 0 0 0 5 0 0 0 0 0 0 0 1 2 0 0 0 0 40 100 0 0 0 0 0 0 0 0 0 0 0 0 0 44 0 0 0 0 0 1 0 4 0 0 0 0 0 0 0 3 0 0 0 0 108 50 0 0 0 0 0 0 0 0 26 0 0 0 0 45 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 26 46 0 0 0 0 0 0 0 0 0 0 0 0 0 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 80 0 0 0 0 0 0 0 0 0 0 0 0 0 47 0 0 0 0 0 3 1 0 0 0 0 0 0 0 0 0 0 0 0 0 98 40 0 0 0 0 0 0 0 0 1 5 0 0 0 48 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 61 0 0 0 0 0 0 0 0 1 0 0 0 0 49 0 0 0 0 0 0 0 20 0 0 0 0 0 0 0 0 1 0 0 0 60 45 0 0 0 0 0 3 0 0 0 0 0 0 0 50 0 0 0 1 0 6 0 0 0 0 0 0 0 0 0 4 1 0 0 0 50 20 0 0 0 0 0 0 22 0 9 0 0 0 0 51 0 0 0 10 0 47 0 1 0 0 0 0 300 0 1 3 0 0 0 0 63 20 0 0 1 0 0 0 0 0 5 33 0 0 0

36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

52 0 0 1 0 0 23 1 3 0 0 0 0 86 0 0 1 0 0 0 0 65 80 0 0 0 0 0 0 0 0 1 1 0 0 0 53 0 0 15 0 0 3 0 1 0 0 0 0 400 1 0 0 0 0 0 0 51 50 0 0 0 0 0 0 1 0 0 1 0 0 0 54 0 0 0 0 0 11 0 23 0 0 0 0 0 0 0 0 0 0 0 0 57 0 0 0 0 0 0 0 0 0 0 0 0 0 0 55 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 1 0 0 0 0 31 0 0 0 0 0 0 0 0 0 0 31 0 0 0 56 0 0 0 0 0 19 0 1 0 0 0 0 30 0 0 2 0 0 0 0 67 0 0 0 0 0 0 0 0 0 1 7 0 0 0 57 0 0 0 0 0 1 0 0 0 0 0 0 100 0 0 0 0 0 0 0 50 1 0 0 0 0 0 0 0 0 5 8 0 0 0 58 0 0 0 0 0 16 0 2 0 0 0 0 30 0 0 0 0 0 0 0 21 80 0 0 0 0 0 0 0 0 0 7 0 0 0 59 0 0 8 0 0 14 0 3 0 0 0 0 0 0 0 1 0 0 0 0 86 50 0 0 0 0 0 0 0 0 5 14 0 0 0 60 0 0 0 0 0 11 0 11 0 0 0 0 200 0 0 1 0 0 0 0 75 0 0 0 0 0 0 0 0 0 0 1 0 0 0 61 0 0 0 0 0 46 0 17 0 0 0 0 200 0 0 1 0 0 0 0 264 0 0 0 0 0 1 0 1 0 5 3 0 0 0 62 0 0 0 0 0 26 20 7 0 0 1 0 50 0 0 0 0 0 0 0 74 50 0 0 0 0 0 0 1 0 1 0 0 0 0 63 0 0 0 0 0 39 0 0 0 0 20 0 0 0 0 0 0 0 0 0 291 0 0 0 0 0 0 0 1 0 0 30 0 0 0 64 0 0 0 0 0 33 0 0 0 0 0 0 20 0 0 0 0 0 0 0 44 0 1 0 0 0 0 0 0 0 0 8 0 0 0 65 0 0 0 0 0 42 0 0 0 0 0 0 200 2 0 0 0 0 0 0 9 100 0 0 0 0 0 0 0 0 0 3 0 0 0 66 0 0 0 0 0 3 0 2 0 0 0 0 0 0 0 1 0 0 0 0 16 0 0 0 0 0 0 0 0 0 0 4 0 0 0 67 0 0 0 0 1 4 0 1 0 0 0 0 50 0 0 2 0 0 0 0 20 0 0 0 0 0 1 0 1 0 1 40 0 0 0 68 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 3 0 0 0 0 346 0 0 0 0 0 0 0 3 0 1 40 0 0 0 69 0 0 0 0 0 2 0 100 0 0 0 0 0 0 0 0 0 0 0 0 13 30 0 0 0 0 0 0 0 0 0 4 0 0 0 70 0 0 0 0 0 38 0 16 0 0 0 0 0 0 0 2 0 0 0 0 31 1 0 0 0 0 0 0 1 0 1 0 0 0 0 71 0 0 3 0 0 9 0 11 0 0 0 0 1 0 0 0 0 0 0 0 22 0 0 0 0 0 0 0 0 0 0 1 0 0 0 72 0 0 0 0 0 71 0 1 0 0 0 0 200 0 0 1 0 0 0 0 24 100 0 0 0 0 0 0 0 0 3 4 0 0 0 73 0 0 2 0 0 11 0 101 0 0 0 0 0 0 0 0 0 0 0 0 57 0 0 0 0 0 0 0 0 0 31 0 0 0 0 74 0 0 0 0 0 63 0 2 0 1 0 0 0 0 0 0 0 0 0 0 172 7 0 0 0 0 2 0 0 0 0 0 0 0 0 75 0 0 200 0 0 21 0 34 0 0 0 0 100 0 0 0 0 0 0 0 108 130 0 0 0 0 0 0 1 0 38 1 0 0 0 76 0 0 0 0 2 47 0 1 0 0 0 0 0 0 0 6 0 0 0 0 136 0 0 0 0 0 0 0 0 0 3 0 0 0 0 77 0 0 0 0 0 28 0 17 0 0 0 0 0 0 0 0 0 0 0 0 54 0 0 0 0 0 0 0 0 0 0 0 0 0 0 78 0 0 0 0 0 9 0 0 0 0 0 0 20 0 0 0 1 0 0 0 12 0 0 0 0 0 0 0 0 0 0 0 0 0 0 79 0 0 0 0 0 3 0 8 1 0 0 0 100 0 0 2 0 0 0 0 36 0 0 0 0 0 0 0 0 0 0 1 0 0 0 80 0 0 0 0 0 149 0 48 0 0 0 0 100 0 0 3 0 0 0 0 151 30 0 0 0 1 0 0 0 0 8 10 0 0 0 81 0 0 0 0 0 47 0 4 0 0 4 0 0 0 0 2 1 0 0 0 158 0 0 0 0 1 0 0 0 0 1 12 0 0 0 82 0 0 0 0 0 5 0 0 5 0 0 0 0 0 0 1 0 0 0 0 43 0 0 0 0 0 0 0 0 0 0 0 0 0 0 83 0 0 0 0 0 11 0 0 0 0 0 0 200 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 84 0 0 0 0 0 42 0 40 0 0 0 0 0 0 0 0 0 0 0 0 79 0 0 0 0 0 0 0 0 0 0 0 0 0 0 85 0 0 0 0 0 23 0 0 0 0 4 0 50 0 0 0 0 0 0 0 33 0 0 0 0 0 0 0 0 0 0 6 0 0 0 86 0 0 0 0 0 24 0 200 0 0 0 0 100 0 0 1 0 0 0 0 211 0 0 0 0 0 0 0 0 0 6 1 0 0 0 87 0 0 16 0 0 35 0 126 0 0 0 0 0 0 0 0 0 0 0 0 141 0 0 0 0 1 0 0 0 0 0 0 0 0 0 88 0 0 0 0 0 20 0 34 0 0 0 0 0 0 0 1 0 0 0 0 22 0 0 0 0 0 0 0 0 0 1 0 0 0 0 89 0 0 0 0 0 21 0 6 0 0 0 0 50 0 0 0 0 0 0 0 56 0 0 0 0 0 0 0 0 0 0 27 0 0 0 90 0 0 0 0 0 15 0 0 0 0 0 0 50 0 0 1 1 0 0 0 14 0 0 0 0 0 1 0 0 0 0 16 0 0 0 91 0 0 0 0 0 42 0 1 0 0 0 0 100 1 0 1 0 0 0 0 72 0 0 0 0 0 0 0 0 0 0 8 0 0 0 92 1 0 0 0 0 10 0 13 0 1 0 0 0 0 0 2 0 0 0 0 37 0 0 0 0 0 0 0 0 0 0 11 0 0 0 93 0 0 0 0 0 36 0 0 0 0 0 0 9 2 0 3 0 0 0 0 2 0 0 0 0 0 3 0 0 0 0 0 0 0 0 94 0 0 0 0 1 12 0 9 0 0 0 0 0 0 0 0 0 0 0 0 56 230 0 0 0 0 0 0 0 0 0 0 0 0 0 95 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 96 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 31 6 0 0 0 0 0 0 0 0 0 8 0 0 0 97 0 0 0 0 0 32 0 0 0 0 6 0 55 0 0 1 0 0 0 0 10 0 0 0 0 0 1 0 0 0 0 0 0 0 0 98 0 0 0 0 0 13 1 0 0 0 0 0 0 0 0 0 0 0 0 0 64 0 0 0 0 0 0 0 0 0 0 0 0 0 0 99 0 0 0 0 0 77 0 12 0 0 0 0 0 0 0 0 1 0 0 0 36 30 0 0 0 0 0 0 0 0 0 1 0 0 0

100 0 0 0 0 0 16 0 102 1 0 0 0 20 0 0 0 0 0 0 0 37 0 0 0 0 1 0 0 0 0 0 1 0 0 0 101 0 0 0 0 0 9 0 87 0 0 0 0 0 0 0 0 0 0 0 0 42 0 0 0 0 0 0 0 0 0 0 0 0 0 0 102 0 0 100 0 0 20 0 73 0 0 0 0 0 0 0 0 2 0 0 0 83 0 0 0 0 0 0 0 0 0 0 212 0 0 0

36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

103 0 0 0 0 0 41 0 0 0 0 0 0 0 0 0 0 0 0 0 0 419 0 0 0 0 0 0 0 0 0 0 700 0 0 0 104 0 0 0 0 0 5 0 0 0 0 0 0 50 0 0 2 1 0 0 0 111 0 0 0 0 0 0 0 1 1 0 5 0 2 0 105 0 0 32 0 0 1 0 0 0 0 0 0 300 0 0 0 0 0 0 0 150 20 0 0 0 0 0 1 0 0 0 17 0 0 0 106 0 0 8 0 0 2 0 0 0 0 0 0 100 0 0 1 0 0 0 0 62 30 0 0 0 0 0 0 0 0 0 0 0 0 0 107 0 0 0 0 0 1 0 0 0 0 0 0 26 0 0 3 0 0 0 0 44 42 0 0 0 0 0 0 0 0 0 1 0 0 0 108 0 0 32 0 0 0 0 0 0 0 0 0 10 0 0 1 0 0 0 0 2 4 0 0 0 0 0 0 0 0 0 0 0 0 0 109 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 30 0 0 0 0 0 0 0 0 0 1 0 0 0 110 0 0 0 0 0 1 0 2 0 0 0 0 40 0 0 0 0 0 0 0 29 30 0 0 0 0 0 0 0 0 0 1 0 0 0 111 0 0 0 0 0 3 1 0 0 0 0 0 0 0 0 0 0 0 0 0 284 60 0 0 0 0 0 0 0 0 0 0 0 0 0 112 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 2 0 613 24 0 0 0 0 0 0 0 0 0 12 0 0 0 113 3 0 99 0 0 3 0 60 0 0 0 0 0 0 0 2 0 0 0 0 45 31 0 0 0 0 0 0 0 0 0 0 0 0 0 114 1 0 7 0 0 9 1 0 0 0 0 0 29 0 0 0 0 0 0 0 148 101 0 0 0 0 0 0 0 1 0 14 0 0 1 115 0 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 116 0 0 65 1 0 9 0 96 0 0 0 0 80 0 0 1 0 0 0 0 189 50 0 0 0 0 0 0 0 0 1 15 0 0 0 117 0 0 5 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 118 0 0 5 0 0 16 0 0 0 0 0 0 0 0 0 1 0 0 0 0 2 107 0 0 0 0 0 0 0 0 0 0 0 0 0 119 0 0 0 0 0 1 0 3 0 0 0 0 10 0 0 5 0 0 0 0 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 120 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 64 100 0 0 0 0 0 0 0 0 0 0 0 0 0 121 0 0 50 0 0 0 0 0 0 0 0 0 20 0 0 0 0 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 122 0 0 2 0 0 20 0 0 0 0 10 0 0 0 0 2 1 10 0 0 289 0 0 0 0 0 0 1 2 0 0 50 0 0 2 123 0 0 50 0 0 3 1 470 0 0 0 0 0 0 0 1 0 0 0 0 52 60 0 0 0 0 0 0 0 0 0 0 0 0 0 124 0 0 50 0 0 20 2 0 0 0 0 0 0 0 0 2 0 0 0 0 97 50 0 0 0 0 0 0 0 0 0 20 0 0 0 125 1 0 60 0 0 9 1 200 0 0 0 0 0 0 0 2 3 0 0 0 70 0 0 0 0 0 0 1 0 0 0 5 0 0 0 126 0 1 12 0 0 1 7 0 0 0 0 0 0 0 0 2 0 0 0 0 11 100 0 0 0 0 1 0 0 0 0 0 0 0 0 127 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 128 0 0 1400 0 0 2 1 60 0 0 0 0 0 0 0 4 24 0 0 0 255 30 0 0 0 0 0 0 0 0 0 71 0 0 0 129 0 0 0 0 0 2 0 0 1 0 0 0 0 0 0 5 0 0 0 0 76 0 0 0 0 0 1 0 0 0 1 36 0 0 0 130 0 0 20 0 0 11 0 0 0 0 0 0 0 0 0 0 0 0 0 0 11 100 0 0 0 0 0 0 0 0 0 10 1 0 0 131 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 78 0 0 0 0 0 0 0 0 0 0 8 0 0 0 132 0 0 70 0 0 2 0 130 0 0 0 0 0 0 0 0 0 0 1 0 144 0 0 0 0 0 0 0 0 0 0 30 0 0 0 133 0 0 0 0 0 1 2 1 0 0 0 0 200 0 0 3 0 0 0 0 524 0 0 0 0 0 0 0 0 0 0 0 0 0 0 134 0 0 150 0 0 6 0 0 0 0 0 0 0 0 0 1 0 0 0 0 342 0 0 0 0 0 0 1 0 0 0 10 0 3 0 135 0 0 15 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 16 6 0 0 0 0 0 0 0 0 0 0 0 0 0 136 0 0 70 0 0 6 3 0 0 0 1 0 0 0 0 0 0 0 0 0 8 30 0 0 0 0 0 0 0 0 0 0 0 0 0 137 0 0 16 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 42 30 0 0 0 0 0 0 0 0 0 2 0 0 0 138 0 0 0 0 0 4 16 0 0 0 0 0 0 0 0 10 0 0 8 0 151 50 0 0 0 0 1 0 0 0 0 12 0 0 0 139 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 2 20 0 0 0 0 0 0 0 0 0 1 0 0 0 140 0 0 200 0 0 0 1 202 0 0 0 0 0 0 0 0 0 0 0 0 12 1 0 0 0 0 0 0 0 0 0 0 0 0 0 141 0 0 40 0 0 3 3 49 0 0 0 0 0 0 0 1 0 0 0 0 45 20 0 0 0 0 1 0 0 0 0 3 0 0 0 142 0 0 70 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 265 0 0 0 0 0 0 2 0 0 0 80 0 0 0 143 0 0 57 0 0 4 0 0 0 0 0 0 60 0 0 0 0 0 0 0 24 20 0 0 0 0 0 0 0 0 0 1 0 0 0 144 0 0 100 0 0 6 0 500 0 0 0 0 0 0 0 1 0 0 0 0 15 25 0 0 0 0 0 0 0 0 4 0 0 0 0 145 0 0 27 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 30 0 0 0 0 0 0 0 0 0 0 0 0 0 146 0 0 40 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 147 0 0 21 0 0 0 0 500 0 0 0 0 0 0 0 0 0 0 0 0 37 0 0 0 0 0 0 0 0 0 0 0 0 0 0 148 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 1 0 0 0 0 2 0 0 0 0 0 0 0 0 0 1 0 0 0 0 149 0 0 0 1 0 2 0 60 0 0 0 0 0 0 0 0 0 0 0 0 14 24 0 0 0 0 0 0 0 0 0 0 0 0 0 150 0 0 500 1 0 12 3 0 0 0 0 0 0 0 0 42 2 20 0 0 1036 10 0 0 0 0 0 0 0 0 0 60 0 0 0 151 0 0 3 0 0 0 0 100 1 0 0 0 0 0 0 2 0 0 0 0 410 24 0 0 0 0 0 0 0 0 3 42 0 0 0 152 0 0 0 0 0 0 0 12 0 0 0 0 0 0 0 1 0 0 0 0 62 0 0 0 0 0 0 0 0 0 2 0 0 0 0 153 0 0 0 0 0 14 0 70 0 0 0 0 40 0 1 0 0 0 0 0 244 20 0 0 0 0 0 0 0 0 0 16 0 2 0

36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

154 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 60 50 0 0 0 0 2 0 0 0 0 142 0 0 0 155 0 0 6 0 0 0 0 200 0 0 0 0 0 0 2 2 0 0 0 0 20 4 0 0 0 0 0 0 0 0 1 0 0 0 0 156 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 16 0 0 0 0 0 0 0 0 0 0 0 0 0 0 157 0 0 0 0 0 50 1 9 0 0 0 0 0 0 0 1 0 0 0 0 182 0 0 0 0 0 3 0 0 0 3 174 0 0 0 158 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 1 0 0 0 0 38 0 0 0 0 0 0 0 0 0 0 34 0 0 0 159 1 0 40 13 1 24 22 220 0 0 2 0 1 0 1 22 0 0 48 0 114 0 0 0 0 1 2 0 5 0 5 164 0 0 2 160 0 0 0 0 0 11 0 300 0 0 0 0 0 0 0 0 0 0 0 0 91 100 0 0 0 0 0 0 0 0 0 0 0 0 0 161 0 0 0 0 0 18 15 0 0 0 0 0 0 0 0 3 0 0 0 0 130 0 0 0 0 0 0 0 0 0 0 4 0 0 0 162 0 0 3 0 0 26 3 46 0 0 0 0 200 0 1 2 3 0 0 0 372 60 0 0 0 0 0 0 1 0 1 60 0 0 0 163 0 0 0 0 0 6 2 0 0 0 0 0 0 0 1 1 0 0 0 0 149 100 0 0 0 0 1 0 0 0 0 20 0 0 0 164 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 286 0 0 0 0 0 0 0 0 0 0 0 0 0 165 0 0 0 0 0 5 0 20 0 0 0 0 50 0 0 4 0 0 0 0 57 50 0 0 0 0 0 0 0 0 0 0 0 0 0 166 0 0 0 0 0 0 0 0 0 0 0 0 300 0 0 1 0 0 0 0 80 100 0 0 0 0 0 0 0 0 1 0 0 0 0 167 0 0 0 0 0 1 0 0 0 0 0 0 80 0 0 1 0 0 0 0 176 200 0 0 0 0 0 0 0 0 0 3 0 2 0 168 0 0 0 0 0 0 14 54 0 0 0 0 300 0 0 2 0 0 0 0 300 0 0 0 0 0 0 1 0 0 0 1 0 0 0 169 0 0 0 0 0 0 2 40 0 0 0 0 200 0 0 0 0 0 0 0 12 140 0 0 0 0 0 0 0 0 0 0 0 0 0 170 0 0 0 0 0 3 4 200 0 0 0 0 100 0 0 0 0 0 0 0 18 0 0 0 0 0 0 0 0 0 0 0 0 0 0 171 2 0 0 0 0 0 7 60 0 0 0 0 50 0 0 1 0 0 0 0 33 107 0 0 0 0 0 1 0 0 10 6 0 0 0 172 0 0 16 0 0 0 3 49 0 0 0 0 100 0 0 1 0 0 0 0 52 0 0 0 0 0 0 0 0 0 0 0 0 1 0 173 3 0 0 0 0 0 0 275 0 0 0 0 20 0 0 2 0 0 0 0 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 174 0 0 70 0 0 0 0 15 0 0 0 0 0 0 0 4 0 0 0 0 141 100 0 0 0 0 0 0 0 0 0 0 0 0 0 175 0 0 0 0 0 1 0 8 0 0 0 0 200 0 0 1 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 176 0 0 0 0 0 0 0 60 0 0 0 0 0 0 0 2 0 0 12 0 249 100 0 0 0 0 0 0 0 0 0 2 0 0 0 177 0 0 0 0 0 0 0 121 0 0 0 0 0 0 0 0 0 0 0 0 24 50 0 0 0 0 0 0 0 0 0 8 0 10 0 178 0 0 0 0 0 0 0 400 0 0 0 0 0 0 0 0 0 0 0 0 19 80 0 0 0 0 0 0 0 0 0 0 0 0 0 179 0 0 3 0 0 1 11 33 0 0 0 0 50 0 1 3 0 0 0 0 14 80 0 0 0 0 0 0 2 0 3 0 0 0 0 180 0 0 3 0 0 11 0 0 0 0 0 0 20 0 0 3 0 0 4 0 24 0 0 0 0 0 0 0 0 0 0 0 0 0 0 181 0 0 0 0 0 0 0 0 0 0 0 0 80 0 0 1 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 182 2 0 0 4 0 44 9 97 0 0 0 0 0 0 0 13 0 0 0 0 676 0 0 0 0 1 1 1 0 0 3 122 0 1 0 183 0 0 0 0 0 4 0 71 0 0 0 0 100 0 0 6 0 0 2 0 81 0 0 0 0 0 0 0 0 0 0 1 0 0 0 184 0 0 0 0 0 6 1 200 0 0 0 0 0 0 0 1 0 0 30 0 109 50 0 0 0 0 0 0 0 0 1 8 0 0 0 185 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 0 0 0 140 0 0 0 0 0 0 0 0 0 0 0 0 0 186 0 0 0 0 0 0 0 0 0 0 0 0 100 0 0 1 0 0 0 0 21 80 0 0 0 0 0 0 0 0 0 0 0 0 0 187 0 0 0 0 0 4 3 1 0 0 0 0 0 0 0 0 0 0 0 0 96 36 0 0 0 0 0 0 0 0 0 0 0 0 0 188 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 189 0 0 0 0 0 0 0 97 0 0 0 0 75 0 0 0 0 0 0 0 36 100 0 0 0 0 0 0 0 0 1 0 0 0 0 190 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 2 10 0 0 0 0 0 0 0 0 0 0 0 0 0 191 0 0 0 0 0 0 1 0 0 0 0 0 50 0 0 1 0 0 0 0 90 51 0 0 0 0 0 0 0 0 0 0 0 0 0 192 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 1 0 0 0 0 33 100 0 0 0 0 0 0 0 0 0 0 0 0 0 193 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 194 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 195 0 0 0 0 0 2 0 25 0 0 0 0 0 0 0 1 0 0 0 0 79 0 0 0 0 0 0 1 0 0 0 160 0 0 0 196 0 0 0 0 0 0 1 400 0 0 0 0 0 0 0 1 0 0 0 0 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 197 0 0 0 0 0 2 1 100 0 0 0 0 0 0 0 0 0 0 6 0 302 0 0 0 0 0 0 0 0 0 5 0 0 0 0 198 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 60 0 0 0 0 0 0 0 0 0 0 0 0 0 0 199 0 0 0 0 0 3 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 200 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 201 0 0 0 0 0 1 0 220 0 0 0 0 0 0 0 0 0 0 0 0 198 10 0 0 0 0 0 0 0 0 1 1 0 0 0

71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105

1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 0 5 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 9 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 11 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 12 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 13 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 14 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 15 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 16 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 17 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 18 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 19 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 21 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 22 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 23 0 0 0 0 80 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 24 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 26 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 27 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 28 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 29 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 30 0 2 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 31 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 32 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 33 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 34 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 35 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 36 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 37 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 38 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 39 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 40 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 41 0 0 0 0 100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 42 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 43 0 0 0 0 200 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 44 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 45 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 46 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 47 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 48 0 0 0 0 15 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 49 0 0 0 0 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50 0 0 0 0 250 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 51 0 5 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105

52 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 53 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 54 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 55 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 56 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 57 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 58 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 59 0 0 0 0 60 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 60 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 61 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 62 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 63 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 64 0 0 0 0 90 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 65 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 66 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 67 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 68 0 0 0 0 150 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 69 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 70 0 1 0 0 100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 71 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 72 0 0 0 0 200 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 73 0 2 0 0 100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 74 0 0 0 0 150 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 75 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 76 0 0 0 0 250 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 77 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 78 0 0 0 0 300 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 79 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 80 0 0 0 0 100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 81 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 82 0 0 0 0 100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 83 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 84 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 85 0 0 0 0 100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 86 0 0 0 0 30 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 87 0 1 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 88 1 0 0 0 200 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 89 0 0 0 0 251 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 90 0 0 0 0 100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 91 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 92 0 2 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 93 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 94 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 95 0 0 0 0 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 96 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 97 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 98 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 99 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 101 0 0 0 0 28 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 102 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105

103 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 104 0 0 0 1 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 105 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 106 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 107 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 108 0 0 0 0 56 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 109 0 0 0 0 40 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 110 0 0 0 0 106 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 111 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 112 0 0 0 0 215 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 113 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 114 0 1 0 0 37 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 115 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 116 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 117 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 118 0 7 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 119 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 120 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 121 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 122 0 0 0 0 30 0 0 0 10 1 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 123 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 124 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 125 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 126 0 1 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 127 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 128 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 129 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 130 0 0 0 0 24 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 131 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 132 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 133 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 134 0 1 0 0 10 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 135 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 136 0 0 0 0 40 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 137 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 138 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 139 0 0 0 0 100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 140 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 141 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 142 0 0 0 0 13 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 143 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 144 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 145 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 146 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 147 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 148 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 149 0 1 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 150 0 0 0 0 10 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 151 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 152 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 153 0 0 0 0 36 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105

154 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 155 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 156 0 0 0 0 114 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 157 0 0 0 0 0 0 5 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 158 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 159 0 1 0 0 30 0 1 0 0 0 0 0 0 0 0 13 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 160 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 161 0 0 0 0 0 0 10 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 162 0 0 0 0 80 0 9 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 163 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 164 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 165 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 166 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 167 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 168 0 0 0 0 50 1 4 0 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 169 0 0 0 0 70 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 170 0 0 0 0 40 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 171 0 0 0 0 50 0 4 0 0 0 0 0 0 0 0 5 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 172 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 173 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 174 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 175 0 0 0 0 200 0 0 0 0 0 0 0 0 107 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 176 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 177 0 0 0 0 110 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 178 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 179 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 180 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 181 0 0 0 0 100 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 182 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 1 0 183 0 0 0 0 6 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 184 0 0 0 0 50 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50 1 0 0 0 0 0 185 0 0 0 0 204 0 0 0 0 0 0 0 0 80 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 186 0 0 0 0 110 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 4 1 0 0 0 0 0 0 0 0 187 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 188 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 189 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 190 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 191 0 0 0 0 100 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 192 0 0 0 0 71 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 193 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 194 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 195 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 196 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 197 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 198 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 199 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 200 0 0 0 0 30 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 201 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140

1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 1 0 1 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 102 0 0 0 0 0 3 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 81 0 0 0 0 0 6 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 19 0 0 0 0 0 3 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 0 15 270 7 0 0 0 0 9 0 0 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 3 0 0 0 0 0 3 0 0 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 0 0 0 0 0 4 0 0 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 23 0 0 0 0 0 1 0 0 9 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 2 20 2 0 0 0 0 2 0 0

10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 20 1 0 0 0 0 2 0 0 11 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 13 2 0 0 0 0 2 0 0 12 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 4 0 0 0 0 0 2 0 0 13 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 28 0 0 0 0 0 0 0 0 14 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 7 0 0 0 0 0 1 0 0 15 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 43 3 0 0 0 0 1 0 0 16 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 15 0 0 0 0 0 1 0 0 17 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 25 0 0 0 4 0 2 0 0 18 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 11 0 0 0 0 0 15 0 0 19 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 11 0 0 0 0 0 1 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 7 0 0 0 0 0 0 0 0 21 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 23 2 0 0 0 0 1 0 0 22 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 13 6 0 0 0 5 0 3 0 0 23 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 22 0 0 0 0 0 1 0 0 24 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 31 1 0 0 0 0 1 0 0 25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 156 0 0 0 1 0 3 0 0 26 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 1 0 0 0 0 2 0 0 27 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 4 0 1 0 0 0 10 0 0 28 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 28 0 0 0 0 0 6 0 0 29 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 5 1 0 0 0 0 0 4 0 0 30 0 1 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 80 0 0 0 14 0 4 0 0 31 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 42 2 0 0 0 0 4 0 0 32 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 61 0 0 0 0 0 8 0 0 33 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 48 0 0 0 0 0 5 0 0 34 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 38 0 0 0 0 0 2 0 0 35 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 26 0 0 0 0 0 4 0 0 36 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 44 0 0 0 0 0 1 0 0 37 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 12 0 0 0 0 0 2 0 0 38 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 3 0 0 0 0 1 0 0 39 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 3 0 0 0 0 0 1 0 0 40 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 45 0 0 0 0 0 17 0 0 41 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 15 0 0 0 0 0 1 0 0 42 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 18 0 0 0 0 0 3 0 0 43 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 45 0 0 0 0 0 6 0 0 44 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 10 1 0 0 0 0 3 0 0 45 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 4 0 0 46 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 47 0 1 0 0 0 1 52 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 2 2 0 0 0 0 0 1 0 0 48 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 48 0 0 0 0 0 0 0 0 49 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 38 0 0 0 0 0 1 0 0 50 0 2 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 15 0 1 0 0 0 4 0 0 51 0 0 0 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 83 0 1 0 0 0 4 0 0

106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140

52 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 31 0 0 0 0 0 6 0 0 53 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 2 27 1 0 0 0 0 1 0 0 54 0 0 0 0 0 0 0 0 101 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 11 0 0 0 0 0 3 0 0 55 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 1 0 0 0 0 0 0 2 0 0 56 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 12 0 0 57 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 0 0 0 0 0 4 0 0 58 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 4 104 0 0 0 0 0 2 0 0 59 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 30 2 0 0 0 0 3 0 0 60 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 61 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 63 0 0 0 0 0 8 0 0 62 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 13 0 0 0 0 0 0 0 0 63 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 10 0 0 0 0 0 9 0 0 64 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 42 0 0 0 0 1 5 0 0 65 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 55 0 0 0 0 0 3 0 0 66 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 60 0 0 0 0 0 0 0 0 67 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 30 0 0 0 0 0 1 0 0 68 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 70 0 0 0 0 0 4 0 0 69 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 70 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 1 8 1 0 0 0 0 4 0 0 71 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 20 0 0 0 0 0 2 0 0 72 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 2 97 0 0 0 0 0 10 0 0 73 0 0 0 0 0 0 0 0 50 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0 26 0 0 74 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 0 2 0 0 75 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 36 0 0 0 0 0 1 0 0 76 0 0 0 0 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 2 14 0 0 0 0 0 11 0 0 77 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 4 0 0 0 0 0 40 0 0 78 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 1 0 0 0 0 5 0 0 79 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 1 0 0 0 0 0 7 0 0 80 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 1 145 0 0 0 0 0 2 0 0 81 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 6 0 0 0 0 0 2 0 0 82 0 0 0 0 0 0 0 0 200 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 1 0 0 83 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 14 0 0 84 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 0 3 0 0 85 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 3 0 0 86 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 44 0 0 0 0 0 2 0 0 87 0 0 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 2 65 0 0 0 0 0 2 0 0 88 0 0 0 0 0 0 0 0 0 0 2 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 30 0 0 0 0 0 5 0 0 89 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 1 59 1 0 0 0 0 4 0 0 90 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 3 0 0 91 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 89 0 0 0 0 0 11 0 0 92 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 5 2 13 1 3 5 0 0 0 14 0 0 93 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 5 0 0 0 0 0 3 0 0 94 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 1 7 0 0 0 0 0 1 0 0 95 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 9 0 0 0 0 0 4 0 0 96 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 5 0 0 97 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 6 0 0 0 0 0 3 0 0 98 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 2 0 0 0 0 0 2 0 0 99 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 14 0 0 0 0 0 6 0 0

100 0 0 0 0 0 0 0 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 193 0 1 0 0 0 0 0 0 101 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 1 0 0 0 0 1 0 0 102 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 8 0 0 0 0 0 1 0 0

106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140

103 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 13 0 0 0 0 0 2 0 0 104 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 3 4 0 0 0 0 0 2 0 0 105 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 5 0 0 106 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 8 0 0 0 0 0 5 0 0 107 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 14 0 0 108 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 8 0 0 0 0 0 3 0 0 109 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 2 0 0 0 0 0 3 0 0 110 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 111 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 3 0 0 112 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 1 0 0 113 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 3 0 0 114 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 24 0 0 0 0 0 6 0 0 115 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 6 0 0 116 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 1 0 0 0 0 0 0 0 2 13 2 0 0 0 0 5 0 0 117 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 11 0 0 0 0 0 3 0 0 118 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 3 0 0 119 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 3 2 2 0 0 0 10 0 0 120 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 11 0 0 0 0 0 6 0 0 121 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 7 1 0 0 0 0 10 0 0 122 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 4 0 0 0 0 0 5 0 0 123 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 1 0 0 0 0 2 0 0 124 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 1 0 0 0 0 0 5 0 0 125 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 12 1 0 0 0 0 6 0 0 126 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 10 0 0 127 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 0 0 0 0 0 2 0 0 128 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 23 0 0 0 0 0 4 0 0 129 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 4 0 0 0 0 10 0 0 130 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 5 0 0 131 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 3 7 1 0 0 0 0 3 0 0 132 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 6 0 0 0 0 0 8 0 0 133 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 2 0 0 0 0 2 0 0 134 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 16 1 0 0 0 0 2 0 0 135 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 2 0 0 136 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 1 0 0 137 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 138 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 14 1 0 0 0 0 3 0 0 139 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 5 0 0 140 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 24 0 0 0 0 0 1 0 0 141 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 7 0 0 142 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 10 10 0 0 0 0 2 0 0 143 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 0 0 0 0 0 3 0 0 144 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 2 0 0 145 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 1 0 0 146 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 147 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 2 1 0 0 0 0 1 0 0 148 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 149 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 1 0 0 150 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 12 0 0 0 0 0 2 0 0 151 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 1 1 0 0 0 0 5 0 0 152 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 6 0 0 0 0 0 0 0 0 153 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 4 1 0 0 0 0 6 0 0

106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140

154 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 22 0 0 0 0 0 0 0 0 155 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 3 22 0 0 0 0 0 2 0 0 156 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 2 3 0 0 0 0 0 4 0 0 157 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 8 10 0 0 0 0 0 21 0 0 158 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 2 0 0 159 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 4 220 1 0 0 0 0 3 0 3 160 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 0 2 0 0 0 0 1 0 0 161 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 74 0 0 0 0 0 7 0 0 162 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 8 2 1 0 0 0 0 8 0 0 163 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 8 2 0 0 0 0 1 0 0 164 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 2 0 0 165 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 3 0 0 166 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 2 0 0 0 17 0 0 167 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 2 0 0 0 0 0 3 0 0 168 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 6 0 0 0 0 0 9 0 0 169 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 1 0 0 0 0 0 1 0 0 170 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 1 0 1 0 0 0 0 5 0 0 171 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 14 1 0 0 0 0 4 0 0 172 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 10 0 0 0 0 0 13 0 0 173 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 11 0 0 0 0 0 2 0 0 174 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 16 0 0 0 0 0 5 0 0 175 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 176 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 2 0 0 0 0 0 2 0 0 177 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 3 0 0 0 0 0 3 0 0 178 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 0 179 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 4 7 0 2 1 0 0 11 0 1 180 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 1 10 0 0 0 0 0 2 1 0 181 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 6 0 0 182 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 8 27 0 0 0 0 0 3 0 3 183 0 0 0 0 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 8 0 2 0 0 0 11 0 0 184 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 2 0 0 0 0 5 0 0 185 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 4 0 0 0 0 0 5 0 0 186 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 14 0 0 0 0 0 3 1 0 187 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 30 0 0 0 0 0 6 0 0 188 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 189 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 190 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 6 0 0 0 0 0 2 0 0 191 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 7 0 0 192 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 1 4 0 0 0 0 4 0 0 193 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 194 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 195 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 62 0 0 0 0 0 3 0 3 196 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 4 0 0 197 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 1 0 0 198 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 199 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 6 0 0 0 0 0 4 0 0 200 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0 1 0 0 201 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 35 0 0 0 0 0 3 0 0

141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175

1 0 7 0 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 20 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 25 0 2 0 0 0 11 0 3 0 2 0 0 0 0 0 0 36 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 1 0 0 0 0 0 2 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 5 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 6 0 45 0 0 0 0 0 7 0 0 0 0 0 0 0 2 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 7 0 0 0 0 0 0 0 2 0 1 0 0 4 0 0 1 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 8 0 0 0 0 0 0 0 33 0 0 0 0 0 4 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 9 0 0 0 0 0 0 0 2 0 0 0 0 0 12 0 1 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0

10 0 0 0 0 0 0 0 1 0 0 0 0 0 66 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 11 0 11 0 0 0 0 0 3 0 0 0 0 0 241 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 12 0 0 0 0 0 0 0 3 0 0 0 0 1 5 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 13 0 1 0 0 0 0 0 1 0 0 0 0 0 20 0 2 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 14 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 15 3 0 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 16 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 17 0 0 0 0 0 0 0 1 0 1 0 0 0 10 0 2 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 18 0 0 0 0 0 0 0 10 0 0 0 0 0 51 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 19 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 21 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 22 0 1 0 0 0 0 0 2 0 0 2 0 0 13 0 2 0 3 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 23 0 0 0 0 0 0 0 32 0 0 0 0 0 7 0 0 0 0 0 0 0 0 0 0 40 0 0 0 0 1 0 0 0 0 0 24 0 0 0 0 0 0 0 3 0 0 0 0 0 15 0 2 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 25 0 0 0 1 0 0 0 40 0 0 0 0 0 28 0 3 0 5 0 0 0 0 2 0 0 0 1 0 0 0 0 0 0 0 0 26 0 0 0 0 0 0 0 36 0 0 0 0 0 2 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 27 0 0 0 0 0 0 0 0 0 0 0 0 0 8 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 28 0 0 0 0 0 0 0 30 0 0 0 0 0 20 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 29 0 0 0 0 0 0 0 2 1 0 0 0 0 6 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 30 0 3 0 0 0 2 0 36 0 0 0 0 2 52 0 1 0 3 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 31 0 0 0 0 0 0 0 32 0 0 0 0 4 30 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 32 0 0 0 0 0 0 0 75 0 0 0 0 0 30 0 3 0 31 0 0 0 0 0 0 0 0 1 0 0 1 0 0 0 0 0 33 0 0 0 0 0 0 0 135 2 1 0 1 0 34 0 1 0 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 34 0 0 0 0 0 0 0 2 6 0 0 0 0 10 0 0 0 4 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 35 0 0 0 0 0 0 0 14 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 36 0 0 0 0 0 0 0 25 2 0 0 0 0 308 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 37 0 0 0 0 0 0 0 16 0 0 0 0 0 19 0 2 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 38 0 0 0 0 0 0 0 23 0 0 0 0 0 0 0 2 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 39 0 0 0 0 0 0 0 2 0 0 0 0 0 3 0 1 0 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 40 0 0 0 0 0 0 0 8 0 0 0 0 0 1 0 1 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 41 0 0 0 0 0 0 0 20 0 0 0 0 0 7 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 42 0 0 0 0 0 1 0 43 0 0 0 0 0 6 0 2 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 43 0 0 0 0 0 0 0 42 0 0 0 0 0 4 0 1 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 44 0 0 0 0 0 0 0 0 2 0 0 0 0 20 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 45 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 46 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 47 0 0 0 0 0 0 0 27 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 48 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 49 0 0 0 0 0 0 0 103 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 50 0 0 0 0 0 0 0 40 0 0 0 0 0 20 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 51 1 0 0 0 0 0 0 6 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0

141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175

52 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 53 0 0 0 0 0 0 2 11 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 5 54 0 0 0 0 0 0 0 15 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 200 0 0 0 1 0 0 0 0 0 55 0 0 0 0 0 0 0 50 0 0 0 0 0 0 0 0 0 0 0 0 0 70 0 0 0 0 0 0 0 0 0 0 0 1 0 56 0 0 0 0 0 0 1 35 0 0 0 0 0 15 0 0 0 0 0 0 0 0 0 0 0 250 0 0 0 0 0 0 0 0 0 57 0 0 0 0 0 0 0 18 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 58 0 0 0 0 0 0 0 67 0 0 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 30 0 3 0 0 0 0 0 0 0 59 0 0 0 0 0 0 1 82 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 100 0 0 1 1 0 0 0 0 0 60 0 0 0 0 0 0 0 100 0 0 0 0 0 0 0 0 0 0 0 0 0 17 0 0 0 200 0 0 0 0 0 0 0 0 0 61 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 12 0 0 0 0 0 0 0 0 0 0 0 0 0 62 0 0 0 0 0 0 0 10 1 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 200 0 0 0 0 0 0 0 0 3 63 0 0 0 0 0 0 1 59 0 0 0 0 0 5 0 0 0 0 0 0 0 6 0 0 0 0 0 0 0 1 0 0 0 0 4 64 0 0 0 0 0 0 1 52 0 0 0 0 0 202 436 0 0 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 0 0 1 65 0 0 0 0 0 0 1 10 0 0 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 66 0 0 0 0 0 0 0 63 0 1 0 0 0 1 0 0 0 0 0 0 0 12 0 0 0 0 0 0 0 0 2 0 0 0 0 67 0 0 0 0 0 0 4 341 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 106 0 0 0 0 0 0 0 0 0 68 0 0 0 0 0 0 0 89 0 0 0 0 0 0 0 1 0 0 0 0 0 6 0 0 0 100 0 0 0 0 0 0 0 0 0 69 0 0 0 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 70 0 0 0 0 0 0 0 27 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 71 0 0 0 0 0 0 0 10 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 1 72 0 0 0 0 0 0 0 35 4 0 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 73 0 0 0 0 0 0 0 0 1 0 0 0 0 100 0 1 0 0 0 0 0 0 0 0 0 70 0 0 0 0 0 0 0 0 0 74 0 0 0 0 0 0 0 8 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 75 0 0 0 0 0 0 1 51 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 76 0 0 0 0 0 0 0 3 0 0 0 4 0 2 0 0 0 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 0 0 0 77 0 0 0 0 0 0 0 15 0 0 0 6 0 6 1 0 0 0 0 0 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 78 0 0 0 0 0 0 0 6 0 0 0 0 0 4 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 79 0 0 0 0 0 0 0 7 0 0 0 0 0 4 0 0 0 0 0 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 80 0 0 0 0 0 0 0 110 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 100 0 0 0 0 0 0 0 0 0 81 0 0 0 0 0 0 0 40 0 1 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 100 0 0 0 0 0 0 0 0 0 82 0 0 0 0 0 0 0 10 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 150 0 0 0 0 0 0 0 0 0 83 0 0 0 0 0 0 0 62 0 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 0 400 0 0 0 0 0 0 0 0 0 84 0 0 0 0 0 0 0 29 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 85 0 0 0 0 0 0 0 15 0 0 0 0 0 3 0 0 0 0 0 0 0 10 0 0 0 100 0 0 0 0 0 0 0 0 0 86 0 0 0 0 0 0 1 76 2 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 1 87 0 0 0 0 0 0 1 12 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 88 0 0 0 0 0 0 0 4 0 1 0 0 0 37 0 1 0 1 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 1 0 89 0 0 0 0 0 0 1 364 0 5 0 0 0 0 0 1 0 2 0 0 0 106 0 0 0 0 0 0 1 0 0 0 0 1 2 90 0 0 0 0 0 0 0 230 0 3 0 0 0 11 0 1 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 91 0 0 0 0 0 0 0 139 0 0 0 0 0 1 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 92 0 0 0 0 0 0 0 41 0 0 0 0 0 6 0 0 0 0 0 0 0 10 5 0 0 0 0 0 0 0 0 0 0 0 4 93 0 0 0 0 0 0 0 5 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 94 0 0 0 0 0 0 0 4 9 0 0 0 0 0 1 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 95 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 96 0 0 0 0 0 0 0 63 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 97 0 0 0 0 0 0 0 29 0 0 0 0 0 0 4 0 0 0 0 0 0 0 1 0 0 100 0 0 0 0 0 0 0 0 0 98 0 0 0 0 0 0 0 6 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 99 0 0 0 0 0 0 0 83 5 1 0 0 0 51 0 1 0 0 0 0 0 4 0 0 0 200 0 0 0 0 0 0 0 0 0

100 0 0 0 0 0 0 0 7 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 60 0 0 0 0 0 0 0 0 0 101 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 102 0 0 0 0 3 0 0 47 0 0 0 1 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3

141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175

103 0 0 0 0 0 0 0 8 0 0 0 13 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 104 0 0 0 0 1 0 0 7 0 2 0 0 0 0 0 0 0 2 0 1 0 0 4 0 0 0 0 0 0 0 1 0 0 0 3 105 0 0 0 0 0 0 0 3 0 0 0 0 0 11 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 106 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 107 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 108 0 0 0 0 0 0 0 6 0 0 0 0 0 0 0 0 2 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 109 0 0 0 0 0 0 0 4 0 0 0 0 0 8 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 110 0 0 20 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 111 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 112 0 0 0 0 0 0 0 9 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 10 0 0 0 1 0 0 0 0 0 113 0 0 0 0 1 0 0 7 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 114 0 0 0 0 0 0 0 49 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 115 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 116 1 0 0 0 1 0 0 4 0 0 0 2 0 28 0 2 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 117 0 0 0 0 0 0 0 6 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 118 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 119 0 0 0 0 0 0 0 21 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 120 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 121 0 0 0 0 0 0 0 13 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 122 0 0 0 0 3 0 0 1 1 1 0 0 0 10 0 10 0 0 0 0 0 0 0 0 0 0 0 0 0 11 0 0 0 0 14 123 0 0 0 0 0 0 0 1 0 0 0 0 0 30 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 124 0 0 0 0 0 0 0 17 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 125 0 1 0 0 0 0 0 12 0 0 0 1 0 3 0 1 0 0 0 1 0 0 1 0 0 0 0 0 0 2 0 0 0 0 0 126 0 0 0 0 1 0 0 4 2 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 127 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 128 0 0 0 0 0 0 0 14 0 0 0 1 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 129 0 0 0 0 1 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 130 0 0 0 0 0 0 0 6 0 0 0 0 0 0 0 1 0 0 0 0 0 0 2 0 0 0 0 0 2 0 0 0 0 0 1 131 0 0 0 0 0 0 0 4 1 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 2 132 0 0 0 0 0 0 0 24 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 133 0 0 0 0 0 0 0 11 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 3 134 0 0 0 0 0 0 0 21 0 0 0 0 0 48 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 135 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 136 0 0 0 0 0 0 0 20 1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 6 0 0 0 0 0 137 0 0 0 0 0 0 0 9 2 0 0 0 0 8 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 138 0 0 0 0 3 0 0 6 1 0 0 0 0 2 0 0 0 0 0 9 0 0 0 0 0 0 0 0 0 0 0 0 0 0 9 139 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 140 1 1 0 0 0 0 0 1 1 0 0 1 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 3 141 0 0 0 0 0 0 0 10 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 142 0 0 0 0 0 0 0 81 0 0 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 143 0 0 0 0 0 0 0 79 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 144 0 0 0 0 0 0 0 3 3 0 0 0 0 11 0 0 0 0 0 0 0 2 1 0 0 1 0 0 0 0 0 0 0 0 0 145 0 0 0 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 40 0 0 0 0 0 0 0 0 1 146 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 147 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 148 0 0 0 0 0 0 0 0 3 0 0 0 0 7 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 149 0 0 0 0 0 0 0 180 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 6 0 0 0 0 0 0 0 0 0 150 0 0 0 0 0 0 0 12 0 0 0 0 0 4 0 0 0 0 0 1 0 0 1 0 0 0 0 0 0 6 0 0 0 0 13 151 0 0 0 0 0 0 0 13 0 0 0 3 0 5 0 0 1 1 0 0 0 0 2 0 0 0 1 0 0 0 0 0 0 0 0 152 1 0 0 0 0 0 0 50 0 0 0 0 60 20 0 0 1 4 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 153 0 0 0 0 0 0 0 25 0 0 0 0 0 6 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175

154 0 0 0 0 0 0 0 352 0 0 0 0 0 20 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 155 0 2 0 0 0 0 0 8 0 0 0 0 0 2 0 2 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 156 0 0 0 0 0 0 0 0 0 0 0 0 0 23 0 0 0 2 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 157 0 0 0 0 0 0 0 134 0 0 0 0 0 3 0 4 0 26 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 158 0 0 0 0 0 0 0 4 0 0 0 0 0 6 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 159 3 1 0 0 1 0 0 327 0 0 0 2 0 6 0 1 1 25 0 9 0 0 0 0 0 0 0 0 0 4 0 0 0 0 0 160 0 0 0 0 0 0 0 34 0 0 0 0 0 0 0 0 0 2 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 161 0 0 0 0 2 0 0 39 0 0 0 0 0 56 0 2 0 8 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 162 0 0 0 0 0 0 0 193 12 0 0 0 0 16 0 1 0 1 0 1 0 0 0 0 1 0 0 0 0 1 0 0 0 0 0 163 0 0 0 0 3 0 0 301 1 0 0 0 0 0 0 0 0 0 0 2 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 164 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 165 0 0 0 0 0 0 0 12 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 166 1 0 0 0 0 0 0 6 0 0 0 0 0 4 0 0 0 1 0 2 0 0 2 0 0 0 0 0 0 0 0 0 0 0 1 167 0 0 0 0 0 0 0 8 0 0 0 1 0 14 0 1 0 0 0 0 0 20 1 0 0 0 0 0 0 0 0 0 0 0 0 168 0 0 0 0 0 0 0 1 3 0 0 0 0 1 1 0 0 2 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 169 0 0 0 0 0 0 0 15 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 170 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 171 0 0 0 1 0 0 0 80 0 0 0 0 0 0 0 0 1 1 0 2 0 0 0 0 7 0 0 0 1 0 0 0 0 0 1 172 0 0 0 0 1 0 0 7 1 0 0 0 0 1 6 0 0 0 0 1 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 173 0 0 0 0 0 0 0 4 0 0 0 0 0 0 0 0 0 0 0 1 0 0 4 0 0 0 0 0 0 0 0 0 0 0 1 174 0 0 0 0 0 0 0 5 0 0 0 0 1 10 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 1 175 0 0 0 1 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 176 0 0 0 0 1 0 0 30 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 177 0 0 0 0 0 0 0 117 0 0 0 0 0 10 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 178 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 3 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 179 0 3 0 0 0 0 0 0 2 2 0 0 0 0 2 0 0 1 0 6 0 0 1 0 0 0 0 0 0 0 1 1 0 0 3 180 0 0 0 0 0 0 0 35 6 0 0 0 0 30 0 0 0 1 0 0 0 6 0 0 0 0 0 0 0 0 1 0 1 0 4 181 0 0 0 0 0 0 0 3 0 0 0 0 0 2 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 182 0 3 0 0 1 0 0 114 1 1 0 0 0 20 0 1 1 21 0 7 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 183 1 0 0 0 0 0 0 25 0 0 0 0 0 0 0 0 0 1 0 4 0 0 2 0 0 0 0 0 0 1 0 0 0 0 0 184 2 0 0 0 0 0 0 15 0 0 0 0 0 2 0 2 1 0 1 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 185 0 0 0 0 0 0 0 20 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 186 0 0 0 0 1 0 0 4 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 187 0 0 0 0 0 0 0 4 0 0 0 0 0 3 0 1 0 2 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 188 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 189 0 0 0 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 4 0 0 0 0 0 0 0 0 0 0 190 0 0 0 1 0 0 0 1 0 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 191 0 0 0 0 0 0 0 1 0 0 0 0 0 5 0 0 1 10 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 192 0 0 0 0 0 0 0 5 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 193 0 0 0 0 0 0 0 2 0 0 0 0 1 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 194 0 0 0 0 0 0 0 100 7 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 195 0 0 0 0 1 0 0 343 0 0 0 0 0 0 0 0 0 10 0 0 0 0 2 0 0 0 0 0 0 0 0 0 0 0 0 196 0 0 0 0 0 0 0 1 0 0 0 0 0 2 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 197 0 0 0 0 0 0 0 55 0 0 0 0 0 1 0 0 0 0 0 0 0 0 2 0 1 0 0 0 0 0 0 0 0 0 0 198 0 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 199 0 0 0 0 0 0 0 24 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 200 0 0 0 0 0 0 0 25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 201 0 0 0 0 0 0 0 23 0 0 0 0 0 8 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 3 0

APPENDIX II QUESTIONNAIRES USED IN THE WESTERN KENYAN

SURVEYS

APPENDIX II

314

1. Farmer Interview I

1.1. Sampling

Interviews will be done in 4 villages, 50 households per village.

Stratification of interviews: try to include

• Minimum 10 - maximum 30 households with farm < 2 acres

• Minimum 10 - maximum 30 households with farm 2 - 4 acres

• Minimum 10 - maximum 30 households with farm > 4 acres

• Minimum 20 - maximum 30 female-headed households

• Minimum 20 - maximum 30 male-headed households

• For the female-headed households: try to interview mainly those women who have decision power over trees to plant

1.2. Preparations

Explain to farmers the aim of the survey: to understand better which trees farmers plant, in order to find out where ICRAF could possibly assist with its research capacity (as one programme of ICRAF deals specifically with trees). Also explain that the survey is not a one-off event, but that follow-up will happen, possibly small interventions. The purpose of the interviews is to understand what is happening on farms - good information will enable ICRAF to target its research better.

Ask to interview the household head, together with anybody who has knowledge on trees planted or retained in the farm. If the household head is not present, ask when he/she is there and try to make an appointment.

Try to make appointments with all farmers before the interview starts. The best approach would probably be to identify farmers on basis of the stratification some days before the interview, and to make an appointment for the day to interview.

1.3. Section 1 and 2 (Interview identification and household data)

Fill in section 1 and 2 before starting the walk around the farm.

The sampling unit is the household and the farmland that the household manages.

• Question 1/5, in case of female head (husband absent), note the name of the husband here

• Question 2/1 refers to the household type. Female headed household (husband absent) means that the husband has a job somewhere else and is most of the time absent from the farm doing his job, not that the husband is away for the particular time of the interview.

• Question 2/5 refers to the age structure of the family. Indicate for every child the gender (male or female) and the year in which born or the age.

• If the farm contains a tree nursery that supplies trees to other farmers, indicate this under question 2/7 as distance being 0 km and under name: supply from own farm. This farm can then later be visited for interviews about the supply side.

APPENDIX II

315

1.4. Section 3

The purpose of section 3 is to collect information on numbers of species on farm and their age structure. The information is collected while walking around the farm.

1.4.1. Flow of the interview

Information has to be recorded by niche and by species generation.

A species generation is defined as a group of trees of the same species, growing in the same niche and established at the same time. Every row under section 3 refers to one species generation.

The protocol for section 3 is proposed as follows:

• Point to a certain species growing in the farm

• Ask the farmer about the age or planting date of this tree

• Ask the farmer whether more trees of the same species were planted at the same time in the niche

• Indicate the number of trees in the generation under #.

• Record all information on the generation (as described below)

• Try to record information in a systematic way in order to let the interview progress rapidly – try to record information niche by niche and species by species.

1.4.2. Information to be recorded

If differences in information occur, then indicate between brackets the number of trees for which the information is valid (e.g. form seedling (5), seed (7); replace by same (6), citrus (6); origin own farm (3), neighbours (1))

• Question 3/1. Record the local name and/or scientific name. In doubt, put a questionmark behind the name. Ask for variety or whether grafted for fruit trees. Also ask for fruit trees whether those would ideally be replaced by another/a variety under question 6/1

• Question 3/2. Niches are listed at the bottom of the page. If the niche is not listed, enter in full next to TF.

• Question 3/3.

• Length is the length of the longest stem – this has to be estimated.

• Circumference is measured at 30 cm above ground. For shrubs, measure the 3 stems with the biggest diameter. For coppiced or pollarded trees: try to measure under the place where the tree was coppiced, and where the stem seems most homogeneous (the reason is that we are trying to link age of the tree with its diameter and height).

• Take at least one measurement for every group. When obvious differences in diameter or height occur within a generation, then take the extreme measurements (and measurement for the biggest tree, one for the smallest tree).

• Question 3/4. Retained means that the tree started growing naturally in the farm and was left to grow at the place it emerged. Record the year or the age of the tree, otherwise indicate that

APPENDIX II

316

the farmer does not know this (a “/” might indicate that the farmer did not know). Management refers to coppicing, pollarding or top pruning — events that may have altered the height/diameter of trees. Coppicing means cutting the tree close to the ground, pollarding means cutting the tree at about 1.5 – 2 m depth, top pruning means cutting the tree higher up. Indicate what type of management was done and when it was done, for example coppiced, 1997. In case not all trees of the generation were managed, indicate the number between brackets as with the other categories, for example coppiced, 1997 (3).

• Question 3/5. Be precise in recording the origin or form of the generation. Indicate the (approximate) distance to the farm. For forms, differentiate between wildlings and seedlings – wildlings regenerated naturally, seedlings were planted as seeds.

• Question 3/6. Replacements after removal: tell the farmer to think of the ideal case where all species are available in sufficient numbers, in all possible forms and from all origins and at no cost. Ask for fruit trees whether those would ideally be replaced by a specific variety. Also indicate if the farmer does not know when to replace or by what to replace.

1.5. Section 4

In contrast to section 3, information is recorded here species by species. Where information on section 3 was recorded while walking around the farm, section 4 can be collected at one place in the farm.

This section is preferably completed at another date than sections 1-3 as copying origins and forms from section 3 takes some time while the farmer is waiting, and also because it is better not to let individual interviews take too long.

Question 4/2 and 4/3. Origins and forms.

• Copy all origins and forms from section 3, and when there are several, ask the farmer to rank these in order of preference imagining the situation that germplasm is available in all possible forms and from all possible origins.

• Ask whether another form or origin would have been better (again in the situation that germplasm is available in all possible forms and from all possible origins), and if yes, indicate which one this is.

• Give the reason for the preference – also in the case that only one origin or form was given, still the reason why this origin or form was the preferred one has to be recorded.

Question 4/4. Purpose on farm.

• Ask for the main purpose of a tree of a species on the farm, and ask whether this main purpose is the same for all the trees of this species. Two scenarios are possible:

• Only one main purpose: fill in one row for the species.

• Several main purposes for the same species. Write under # the number of trees for each main purpose and complete the information for every main purpose. Ask the farmer how these trees are different. Is it their age, their form, just a decision by the farmer to use one tree for this purpose, another tree for another purpose?

• For timber, specify for which use: for construction, for furniture?

APPENDIX II

317

Question 4/5. Products.

• Record the information for the main product. The main product is not in all cases the main purpose for which a tree is growing. For example, the main purpose for growing a tree may be providing shade, while the main product could be firewood. In this case record information on the firewood product.

1.6. Section 5

Copy all species for the same main purpose from section 4 (if time is limited, copying can be done back at base, and information collected at another stage).

• Ask the farmer to rank all species (of course if only one species occurs for a main purpose, no ranking needs to be done).

• Ask for the reason of the ranking. If one species is better, in which aspects is it better? (Ask for details).

1.7. Back at base

• Try to photocopy and safeguard (send to Nairobi) information as fast as possible

• Keep track of stratification – distribution of household characteristics listed in section 2.

Farmer interview I Photocopy and safeguard information as soon as possible.

Please put any comments on any sections on the back page.

1. Interview identification

1. Interview number _ / _ _ (village/household)

2. Interview date _ _ / _ _ / 199 _ (day/month/year)

3. Village

4. Name household head

5. Name absent husband (for female headed household)

6. GPS positions (WGS 84) 0 º _ _ ‘ _ _ N ; 3 _ º _ _ ‘ _ _ E ; 1 _ _ _ m asl

7. Name(s) of respondent(s)

8. Respondent(s) is/are (circle) Husband Wife Child(ren)

9. Name(s) of interviewer(s)

2. Household data

1. Household head (circle) Male ; Female (no husband) ; Female (husband absent)

2. Size of landholding

(indicate h (ha) or a (acres))

Main farm: _ _ . _ Other plots: _ _ . _

Homestead: _ _ . _ Woodlot: _ _ . _ Fallow: _ _ . _

3. Housing type (circle) Permanent Iron roof Thatch roof

4. Number of cattle Crossbreds: _ _ Local: _ _

5. Age structure of family (year) / indicate M(ale)/F(emale) for children

Husband: _ _ Wife: _ _ Wife2: _ _

Resident children:

6. Other information on household head

Years being head: _ _

Schooling (highest):

Previous off-farm job:

7. Nearest tree nursery Distance (km):

Name (name owner):

3. Tree inventory data Interview number : _ / _ _ (village/household) # 1. Tree names / 2. Niche 3. Dimensions 4. Establishment 5. Origin / Form 6. Removing

Local (language):

Scientific:

Var/grafted:

Length(m):

C30(cm):

Planted / Retained

When / Age:

Whom:

Origin:

Distance (km):

Form:

When / Age / Not known / Other:

Replace by: Same / Other (var) / Not:

Niche : HS ; CM ; CC ; IB ; EB ; WL ; TF When managed last:

Local (language):

Scientific:

Var/grafted:

Length(m):

C30(cm):

Planted / Retained

When / Age:

Whom:

Origin:

Distance (km):

Form:


Replace by: same / other (var):


Local (language):

Scientific:

Var/grafted:

Length(m):

C30(cm):

Planted / Retained

When / Age:

Whom:

Origin:

Distance (km):

Form:




Local (language):

Scientific:

Var/grafted:

Length(m):

C30(cm):

Planted / Retained

When / Age:

Whom:

Origin:

Distance (km):

Form:




Local (language):

Scientific:

Var/grafted:

Length(m):

C30(cm):

Planted / Retained

When / Age:

Whom:

Origin:

Distance (km):

Form:




Local (language):

Scientific:

Var/grafted:

Length(m):

C30(cm):

Planted / Retained

When / Age:

Whom:

Origin:

Distance (km):

Form:




Niches: Homestead, Crops-Mixed, Crops-Contours, Internal Boundaries, External Boundaries, Woodlot, Tree Fallow; Dimensions: circumference measured at 30 cm above ground level, length of longest stem

4. Species data Interview number : _ / _ _ (village/household) # 1. Species name 2. Origins 3. Forms 4. Purpose on farm 5. Products

Local (language):

Scientific:

Origins (rank):

Other better:

Forms (rank):

Other better:

Main purpose:

Other purposes:

Main product:

When last collected:

Who uses the product:

Was the product sold: yes / no / some

Reason preference:

Reason preference:

Local (language):

Scientific:

Origins (rank):

Other better:

Forms (rank):

Other better:

Main purpose:

Other purposes:

Main product:




Reason preference:

Reason preference:

Local (language):

Scientific:

Origins (rank):

Other better:

Forms (rank):

Other better:

Main purpose:

Other purposes:

Main product:




Reason preference:

Reason preference:

Local (language):

Scientific:

Origins (rank):

Other better:

Forms (rank):

Other better:

Main purpose:

Other purposes:

Main product:




Reason preference:

Reason preference:

If several main purposes for individual trees of the same species, fill in several rows and ask for reasons. Indicate the number of trees under “#” for every purpose in case several purposes exist.

5. Preference data Interview number : _ / _ _ (village/household) 1. Main purpose 2. Species 3. Rank 4. Reason for ranking (highest/lowest)

If more than 7 species in a block, use two blocks

APPENDIX II

322

2. Farmer Interview II

2.1. Sampling

In Ebuchiebe, every household will be visited again. For the other villages, the subsample of households to be visited will be communicated.

2.2. Preparations before interviews start

The first part of the interview is based on drawings of the farm:

Drawing 1. This drawing should show the farm as it looked like at the time of the interview.

Drawing 2. After discussions, this drawing should show the farm as the farmer would like it to be.

The drawings are meant to be a tool to guide the discussions on the species and their numbers wanted in different niches only; any alterations can be made to the proposed procedure below as long as it allows for the discussions and the results can be transferred to the table of section 2.

Add to both drawings the species (local names) to the correct niches, with the numbers of trees. For drawing 2, the information that is entered at first is the information on replacements. The numbers of trees can be entered with the number of trees on the farm between brackets and the number of trees desired next to it, e.g. “Linera (200) 300”. Write “(0)” if the species did not occur in that niche before, e.g. “Linera (0) 200”. Underline the species on drawing 1 that also occur on drawing 2 (these are the species that are replaced).

All necessary information to produce the maps is provided in printouts for every individual farm.

Niches have the same abbreviations as in questionnaire I (HS = Homestead, CM = Crops-Mixed, CC = Crops-Contours, IB = Internal Boundaries, EB = External Boundaries, WL = Woodlot, TF = Tree Fallow). Details and comments on niches provide extra information on the trees and the niches in which they grow – this information should also be incorporated in the drawings where possible. For example, in some cases the farmer explained that the species would be moved to another niche – in such cases, indicate on drawing 2 that the tree is wanted in that niche. In other cases, the farmer provided extra information on a niche, e.g. HS next to EB or CM near HS. For those examples, include that species in the homestead or in the CM, but write the extra information next to the information on the number of trees, e.g. “Linera (200) 300 next to HS”.

2.3. Section 1 (Interview identification)

• Fill in questions 1/1, 1/2, 1/3 and 1/7 before the interview starts (probably when preparing the drawings for the interviews).

• Question 1/7. Missing information from the previous questionnaire on household information is provided on the printout also containing information on trees present and replaced. Fill in the item that is missing (e.g. farmsize, years that head was in charge, ...).

• Question 1/7. One part of “missing” information are GPS readings. Many of those readings were not very accurate, and maybe this can be contributed to the fact that the readings were not in 3-D (also altitude). Take new readings at every part, making sure that the GPS goes into 3-D readings.

APPENDIX II

323

• Questions 1/4, 1/5, 1/6 and 1/8 should be filled in when the interview starts.

2.4. Section 2 (Ideal species composition desired)

The purpose of section 2 is to complement drawing 2 so that it finally looks like the farm that respondents would like to have ideally.

The proposed protocol for section 2 is as follows:

• Explain the nature of each drawing to the farmer. Explain clearly that the drawings are about species and niches. Ask whether the drawings are clear and provide extra explanations if necessary.

• Explain that the maps are a means of getting an idea of the ideal species composition in the mind of the farmer, in terms of species wanted, the numbers of trees wanted for every species, and the niches in which these trees would be established. The farmer should picture the situation that enough material is available for any species of interest, so that he or she can establish any species of choice in any number. Also, explain that this information is needed to be able to help him/her in the end.

• Stress that the drawing shows a number of potential niches, all with trees in them for the moment. Explain that he/she is completely free on the number of trees in the niche. Not all niches have to be used. The drawing is only to explain the nature of the niche.

• First, discuss the species for which farmers indicated that they wanted to replace them (species already indicated on drawing 2). Try to cycle through the species in a systematic way (maybe by niche).

• Confirm whether the species is really wanted for the particular niches.

• Would the species be wanted in other niches? If so, add the species to those niches. If the species already occurred in that niche according to drawing 1, then copy the number too.

• Ask for every niche whether more or fewer trees are wanted. Base the questions on the number of trees indicated to be replaced (if available), otherwise on number of trees present, otherwise ask for the number of trees wanted.

• If more or fewer trees are wanted, probe for an as precise number as possible. It is very possible that the farmer has no very clear idea about the exact number of trees wanted, but would be interested in roughly twice as much trees, or only half the present number. It could also be possible that farmers would rather indicate a range than one exact number. Note these rough estimates.

• Note for every species that is wanted the number of trees wanted. This could be done by an arrow, e.g. “Linera (200) 300 → 350”; “Linera (200) 300 → double” or “Linera (200) 300 → double to triple”.

• In case the farmer provides any details on the location in the niche, note this information too (e.g. “ Linera (200) 300 → 400 in bananas”).

• Then discuss the species for which the farmers had not indicated that they wanted them replaced (species not underlined on drawing 1).

• Confirm that the species will not be replaced for that particular niche.

APPENDIX II

324

• Check whether the species would not be wanted in other niches.

• In case the answer of any of above questions is positive, add the species to drawing 2, and discuss the species further as species that had been indicated before as being replaced. Copy the number of trees on the farm to drawing 2 as well.

• Ask whether the farmer would like to add any other species to the farm.

• In case other species are wanted, then add these to drawing 1. The format of noting would be like “Linera (0) →“

• Finally ask the farmer on basis of drawing 2 whether this reflects the species and numbers wanted by the farmer. Make changes if necessary.

• When the farmer agrees with the information from drawing 2, the information can be copied into the table. (It could be possible to leave the drawing with the farmer overnight so that he/she could think about it more). This table should represent all information from drawing 2. Maybe this copying could be done back at base.

• The information should also be copied into sections 3 and 4. Note that questions are only asked for species for which different numbers are wanted than originally on the farm. One reason for this is to avoid asking questions about numbers about which the farmers have no ideas (some interviews could therefore be quite short). Another reason is that specific questions are asked about the reasons of the differences between the number of trees wanted and the number of trees present.

2.5. Section 3 (Species for which more trees are wanted)

Questions 3/1, 3/2 and 3/3 should have been entered before starting section 4.

• Question 3/1. Indicate both scientific (binomial, Latin) as local name for the species.

• Question 3/2 and 3/3. Calculate the totals for the species over all niches.

• Question 3/2. Enter “0” if the species is new in that niche or if the species is new on the farm.

Questions 3/4 and 3/5 probe (=try to understand as good as possible) for the reasons why more trees are wanted, and why this higher number was not established earlier.

Note that the farmer should be allowed at any point to change the number of trees wanted. In that case, revisit section 2.

• Question 3/4. Probe why the farmer wants more trees. Try to ask the question as specific as possible; if the farmer indicated that he/she wanted 15 trees, ask why he/she wanted 15 trees.

• Question 3/5. Probe why not more trees were established earlier, as this higher number is desired. Especially ask for details if the farmer explains that the problem was germplasm-related. Was it a specific form of germplasm that was lacking, or was material wanted from a specific source or in a specific form. Would material be selected in any other way?

Questions 3/6, 3/7 and 3/8 and 3/9 ask questions on the practicalities of how the higher number of trees will be established.

APPENDIX II

325

• Question 3/6. When will the higher number of trees be established? Be prepared to note ranges in time (e.g. between now and 2005 or 2 trees next year, 2 trees in 2002).

• Question 3/7. How will the higher number of trees be established? Note that details on form and origin (e.g. transplanting wildlings from the own farm, buying hybrid seedlings from Maseno, …)

• Question 3/8. What are the ratio of the amount of germplasm that is needed and the number of trees that is wanted? In case the amount of germplasm is higher, ask for the reason. This question, together with question 3/9 probes for losses, such as low germination, losses after transplanting. For example, the ratio of germplasm/trees would be “2” if only half the amount of seeds obtained would be expected to germinate. Enter “1” if the same amount of germplasm is needed as the number of trees.

• Question 3/9. Probe for the reason if the amount of germplasm needed is larger than the number of trees needed.

2.6. Section 4 (Species for which fewer trees are wanted)

Questions 4/1, 4/2 and 4/3 should have been entered before starting section 4.

• Question 4/1: Indicate both scientific (binomial, Latin) as local name for the species.

• Question 4/2 and 4/3. Calculate the totals for the species over all niches.

• Question 4/3: enter 0 if the species is not wanted anymore in that niche.

Questions 4/4 and 4/5 probe for the reasons why fewer trees are wanted, and why this smaller number was established earlier.

Note that the farmer should be allowed at any point to change the number of trees wanted. In that case, revisit section 2.

• Question 4/4. Probe why the farmer wants fewer trees. Try to ask the question as specific as possible; if the farmer indicated that he wanted 15 trees, ask why he/she wanted 15 trees. If the farmer does not want the tree anymore, ask why here.

• Question 4/5. Probe why more trees were established earlier, as this higher number is not wanted.

Questions 4/6 and 4/7 investigate whether a species that is not wanted anymore would be retained if it regenerated naturally.

Only ask these questions for species that are not wanted anymore.

• Question 4/6. Would the species be retained if it occurred naturally in that niche? Circle the correct answer.

• Question 4/7. Probe for the reasons that it would be retained, although not being wanted anymore.

APPENDIX II

326

2.7. Section 5 (Use group data)

Every page refers to one use group. Note the species that are wanted on the farm.

Information on the use group of the species can be obtained from an additional printout. For new species on the farm, ask the farmer about their uses at the start of this section and add those species to the respective groups.

Note that it is possible at any point to add new species to the functional group. In this case, go back to section 2 and complete the information on the new species in section 3.

2.7.1. Group and species

Questions 5/1, 5/2, 5/3 and 5/4 should have been entered before starting section 4.

• Question 5/1. Indicate the use group. In case more than eight species belong to the use group, then use the additional sheet for the group. Indicate the number of pages between brackets, e.g. 1/2 and 2/2 or 1/3, 2/3 and 3/3.

• Question 5/2. List the scientific name. Beware confusions with local names.

• Question 5/3. Copy from drawing 2 or sections 4 and 5.

• Question 5/4. Copy from the additional information. Beware that only the ranking for the primary purpose will be listed, other rankings should be left blank.

Questions 5/5 asks for more details for the individual species in the functional groups.

• Question 5/5. This question deals with the total rank of species in the functional group; this is the rank of all species providing the product or service, while 5/4 only listed the ranking of species for which this purpose was a primary purpose. When asking question 5/5, the information already provided for the rankings of the primary purposes should be used (by asking questions as “Is there any species better than species X ranking first for the purpose as a primary purpose”, “Are any species worst than the worst than species Y ranking lowest for the purpose?”). If all species were ranked under 5/4, then do not ask question 5/5.

2.7.2. Number of species/group

The questions on the number of species per group probe for reasons for the diversity within the group. All questions prompt for discussions with the farmers on the meaning of the different number of species in the group, the number of trees and what characteristics are wanted of the group. Although very open-ended in nature, answers to these questions will be very important in exploring the management of tree species diversity on farm.

(Note that the discussions are complicated as many species belong to several groups. If possible, try to differentiate between the diversity wanted for the group alone and the diversity caused by having species with several purposes. Discuss the diversity within a group anyway with farmers, and if they reply for one group that the tree species are used for another group, then note this (as any reaction of the farmers should be noted, also “farmer does not know” or “the number of species does not really matter”). )

• Question 5/6. Ask the farmer why the farmer wants that number of species in the group. The question could also be asked as why species a, species b and species c are wanted as they all provide a similar product. What is the reason? Are there several reasons? Why are not fewer or not more species planted in that group? Why is not just one species planted to provide

APPENDIX II

327

the product? Also discuss if there is only one species in the group – then ask why only one species is planted, and not two or more.

• Question 5/7. Do the numbers in which the species occur in the group matter? The question probes whether it makes any difference that there are 30 trees of species a, and 70 trees of species b; or 10 trees of species a, and 90 trees of species b. Try to ask the question using the actual numbers of trees in the group. (Note: it might be confusing as trees belong to several groups, and the numbers might reflect to other uses they have; but ask this question anyway).

• Question 5/8. This question probes for the minimum characteristics of a species to be wanted on the farm for that purpose. It is seen as a reaction to answers to question 5/6 as “only those species produce enough”, or “only those species produce good enough products”. In these cases, what would the minimum characteristics be for another species to be included?

2.7.3. Productivity of the group when trees established in desired numbers

Questions 5/9, 5/10 and 5/11 check whether the productivity of the group will be adequate, more or not enough for what the family needs. Note that the productivity must be analyzed on basis of the species and numbers wanted for the group (as noted under 5/3).

• Question 5/9. Circle the correct response.

• Question 5/10. If the production would be more or less, give the ratio of what is needed versus what will be produced. The ratio should not be calculated, but can be just put as the ratio, e.g. 3/2.

• Question 5/11. Ask for the reasons for differences. Why is the production not high enough? How is enough material obtained for the family? It is also possible that not enough will be obtained: why is that so? Alternatively, why is more produced on the farm?

2.7.4. Generations

So far, discussions on groups were only on numbers, but not on the age structure of trees. The purpose of this section is to find out whether the farmer envisages different generations of trees on the farm, or whether the age structure of the trees does not matter very much.

• Question 5/12. Circle the correct response.

• Question 5/13. In case different generations are wanted, indicate here the desired age gap between generations.

2.8. Back at base

• Try to photocopy and safeguard (send to Nairobi) information as fast as possible

• Keep track of stratification – distribution of household characteristics listed in section 2.

Farmer interview II Photocopy and safeguard information as soon as possible.

Please put any comments on any sections on the back page.

1. Interview identification 1. Interview number _ / _ _ (village/household)

2. Village

3. Name household head

4. Interview date _ _ / _ _ / 1999 (day/month/year)

5. Name respondent(s)

6. Respondent(s) is/are (circle) Husband Wife Child(ren)

Missing information

7. Item 8. New information

GPS (3D) 0 º _ _ ‘ _ _ N ; 3 _ º _ _ ‘ _ _ E ; 1 _ _ _ m asl

2. Ideal species composition desired

Discuss on basis of the figures provided (figure one with number of trees present, figure two with number of replacing trees) for all niches. Make additions to figure two until it looks like the farm that the farmers would like to have. In case farmers do not have precise ideas about replacements, just indicate rough estimates (e.g. one third more, only half) of the number of trees to farmers would like to add or remove.

When information has been collected, complete the table for the species that are wanted

2. Species and niches – number of trees wanted Interview number : _ / _ _ (village/household) 1. Species 2. Niche 3. # now 4. # wanted 1. Species 2. Niche 3. # now 4. # wanted

3. Species for which more trees are wanted Interview number : _ / _ _ (village/household) 1. Species 2. # now

3. # wanted 4. Why # wanted desired? 5. Why was the (# wanted) (or species) not established

earlier? (if problem with germplasm: record details such as form or origin)

6. When will it be established? 7. How would the (# wanted)

be established?

8. Is the germplasm needed equal to the number of trees needed, or is it bigger?

If bigger, how much? 9. If there is a difference, why?

Scientific: Now: Reason: Reason: When:

How:

Germplasm/trees =

Reason:

Local: Wanted:


How:

Germplasm/trees =

Reason:

Local: Wanted:


How:

Germplasm/trees =

Reason:

Local: Wanted:


How:

Germplasm/trees =

Reason:

Local: Wanted:

4. Species for which fewer trees are wanted Interview number : _ / _ _ (village/household) 1. Species 2. # now

3. # wanted 4. Why # wanted desired in that niche? 5. Why were more/# now established earlier? If the species is not wanted anymore

6. Would it be retained or would wildlings be transplanted if it regenerated naturally?

7. Why as it is not wanted? Scientific: Now:

Reason: Reason: Still retained: yes no

Reason:

Local: Wanted:

Scientific: Now:


Reason:

Local: Wanted:

Scientific: Now:


Reason:

Local: Wanted:

Scientific: Now:


Reason:

Local: Wanted:

5. Use group data Interview number : _ / _ _ (village/household)

Group and species 1. Group ( / )

2. Species wanted 3. Number wanted 4. Rank 5. Total rank

Number of species/group 6. Why this number of species in the group?

7. Any reason for the numbers of individual species?

8. What are the minimum characteristics for another species to be included?

Productivity of the group when trees established in desired numbers 9. Will the production be similar, more or less than what the family needs for its own consumption?

10. What will the ratio be of the amount produced to the amount needed?

11. What is the reason for the differences?

Similar More

produced/needed =

What is done with the rest?

Less

Needed/produced =

How is enough obtained (and if not enough obtained, why are not more trees established)?

Generations 12. Are different generations of trees wanted in the group? Yes No

13. If yes, what would be the desired gap between generations? Generation gap:

Download - tree desertification planning for african agroecosystem

Top Related