This article is downloaded from

http://researchoutput.csu.edu.au

It is the paper published as: Author: L. Morin, A. M. Reid, N. M. Sims-Chilton, Y. M. Buckley, K. Dhileepan, G. T. Hastwell, T. L. Nordblom and S. Ragu Title: Review of approaches to evaluate the effectiveness of weed biological control agents Journal: Biological Control: theory and applications in pest management ISSN: 1049-9644 Year: 2009 Volume: 51 Issue: 1 Pages: Jan-15 Abstract: We review key issues, available approaches and analyses to encourage and assist practitioners to develop sound plans to evaluate the effectiveness of weed biological control agents at various phases throughout a program. Assessing the effectiveness of prospective agents before release assists the selection process, while post-release evaluation aims to determine the extent that agents are alleviating the ecological, social and economic impacts of the weeds. Information gathered on weed impacts prior to the initiation of a biological control program is necessary to provide baseline data and devise performance targets against which the program can subsequently be evaluated. Detailed data on weed populations, associated plant communities and, in some instances ecosystem processes collected at representative sites in the introduced range several years before the release of agents can be compared with similar data collected later to assess agent effectiveness. Laboratory, glasshouse and field studies are typically used to assess agent effectiveness. While some approaches used for field studies may be influenced by confounding factors, manipulative experiments where agents are excluded (or included) using chemicals or cages are more robust but time-consuming and expensive to implement. Demographic modeling and benefit-€“cost analyses are increasingly being used to complement other studies. There is an obvious need for more investment in long-term post-release evaluation of agent effectiveness to rigorously document outcomes of biological control programs. Author Address: tom.nordblom@industry.nsw.gov.au URL: http://dx.doi.org/10.1016/j.biocontrol.2009.05.017 www.elsevier.com/locate/ybcon http://researchoutput.csu.edu.au/R/-?func=dbin-jump-full&object_id=19407&local_base=GEN01-CSU01 http://unilinc20.unilinc.edu.au:80/F/?func=direct&doc_number=001083156&local_base=L25XX CRO Number: 19407

<Louise.Morin@csiro.au>

27/05/2009 07:49 AM

To <Adele.Reid@csiro.au>, <Nikki.Sims-Chilton@dpi.qld.gov.au>, <y.buckley@uq.edu.au>,

cc

bcc

Subject FW: BCON-08-467R1 - Final Decision

Good morning all,

See below.

CheersLouise

Dr Louise Morin CSIRO Entomology GPO Box 1700 Canberra ACT 2601 Phone: +61 (02) 6246 4355 Fax: +61 (02) 6246 4362

-----Original Message-----From: ees.bcon.0.1425b.44102af3@eesmail.elsevier.com [mailto:ees.bcon.0.1425b.44102af3@eesmail.elsevier.com] On Behalf Of BIO-CON (ELS)Sent: Wednesday, 27 May 2009 5:58 AMTo: Morin, Louise (Entomology, Black Mountain)Subject: BCON-08-467R1 - Final Decision

Ms. No.: BCON-08-467R1Title: Review of approaches to evaluate the effectiveness of weed biological control agentsCorresponding Author: Dr. Louise MorinAuthors: Adele M Reid; Nikki M Sims-Chilton; Yvonne M Buckley; K Dhileepan; Graeme T Hastwell; Tom L Nordblom; S Raghu

Dear Dr. Morin,

I am pleased to inform you that your manuscript referenced above has been accepted for publication in Biological Control.

Many thanks for submitting your fine paper to Biological Control. I look forward to receiving additional papers from you in the future.

With kind regards,

John H. HoffmannEditorBiological Control

E-mail: bio-con@elsevier.com

Entomology

Clunies Ross Street, ACTON ACT 2601

GPO Box 1700, Canberra ACT 2601

Telephone : 02 6246 4001 Facsimile : 02 6246 4000

www.csiro.au

ABN 41 687 119 230

Dr John HoffmannEditorBiological Control Elsevier

11 May 2009

Dear John,

Re. BCON-08-467 revised version

Please find enclosed the revised manuscript now entitled “Review of approaches to evaluate the effectiveness of weed biological control agents” by L. Morin, A.M. Reid, N.M. Sims-Chilton, Y.M. Buckley, K. Dhileepan, G.T. Hastwell, T.L. Nordblom, and S. Raghu. We have responded to all points raised by the two reviewers in a separate file (attached) and made changes accordingly in the revised manuscript. I have also attached a copy of the revised version where all changes are shown.

Looking forward to hear back from you.

Yours sincerely,

Dr Louise MorinResearch ScientistLouise.morin@csiro.au+61 2 6246 4355

Cover Letter

1

Review of approaches to evaluate the effectiveness of weed 1

biological control agents2

3

L. Morin1,2, A.M. Reid1,2, N.M. Sims-Chilton1,3,5, Y.M. Buckley1,3,4, K. Dhileepan1,5, G.T. 4

Hastwell5,8, T.L. Nordblom6, S. Raghu75

6

1 Cooperative Research Centre for Australian Weed Management7

2 CSIRO Entomology, GPO Box 1700, Canberra, ACT 2601, Australia8

3 The University of Queensland, School of Biological Sciences, St Lucia, Queensland 4072, 9

Australia10

4 CSIRO Sustainable Ecosystems, 306 Carmody Road, St Lucia, Queensland 4067, Australia11

5 Department of Employment, Economic Development and Innovation, Biosecurity 12

Queensland, Alan Fletcher Research Station, PO Box 36, Sherwood, Queensland 4076, 13

Australia 14

6 E.H. Graham Centre for Agricultural Innovation (Charles Sturt University and NSW 15

Department of Primary Industries), Wagga Wagga, NSW 2650, Australia16

7 Arid Zone Research Institute, Northern Territory Government, PO Box 8760, Alice Springs, 17

Northern Territory 0871, Australia18

8 Present address: School of Natural Sciences, University of Western Sydney, Locked Bag 19

1797, Penrith DC, NSW 1797, Australia 20

21

Corresponding author: louise.morin@csiro.au, Phone: +61 2 6246 4355, Fax: +61 2 6246 22

4362 23

* Manuscript

2

Abstract1

By reviewing key issues, available approaches and analyses, we aim to encourage and assist 2

practitioners in developing sound plans to evaluate the effectiveness of agents in adversely 3

affecting their target weeds at various phases throughout a biological control program. 4

Assessing the effectiveness of prospective agents before release assists the selection process, 5

while post-release evaluation aims to determine the extent that agents are alleviating the 6

ecological, social and economic impacts of the weeds. Information gathered on these weed7

impacts prior to the initiation of a biological control program is necessary to provide baseline 8

data and devise performance targets against which the program can subsequently be 9

evaluated. Detailed data on weed populations, associated plant communities and, in some 10

instances ecosystem processes collected at representative sites in the introduced range several 11

years before the release of agents can be compared with similar data collected later to assess 12

agent effectiveness. Laboratory, glasshouse and field studies are typically used to assess agent13

effectiveness. While some approaches used for field studies may be influenced by 14

confounding factors, manipulative experiments where agents are excluded (or included) using 15

chemicals or cages are more robust but time-consuming and expensive to implement.16

Demographic modelling and benefit-cost analyses are increasingly being used to complement 17

other studies. There is an obvious need for more investment in long-term post-release 18

evaluation of agent effectiveness to rigorously document outcomes of biological control 19

programs.20

21

Key words: Weed biological control; Invasive plants; Impacts; Benefit-cost analysis; 22

Demographic modelling; Evaluation; Monitoring 23

24

25

3

1. Introduction1

2

Classical weed biological control programs are increasingly criticized for their lack of 3

rigorous evaluation of the ultimate outcomes of deliberate introductions of exotic organisms 4

(McEvoy and Coombs., 2000; Wajnberg et al., 2001; Louda et al., 2003; Pearson and 5

Callaway, 2003, 2004; Thomas et al., 2004; Thomas and Reid, 2007; Carson et al., 2008; 6

Müller-Schärer and Schaffner, 2008). In a typical program, most resources are invested in 7

surveying the native range of the invader for natural enemies, but limited time and energy are 8

spent assessing the effectiveness of candidate biological control agents in suppressing the 9

target weed (Sheppard, 2003; McClay and Balciunas, 2005; Morin et al., 2006; Raghu et al., 10

2006b). In contrast, extensive resources are allocated to experimentally demonstrating that the 11

candidate agents do not pose an unacceptable risk to non-target plant species (Evans, 2000; 12

Briese, 2005). Once agents are approved for release by regulatory agencies, funds are then 13

directed to mass-culture, release and monitor their establishment (e.g. Briese and McLaren, 14

1997; Clark et al., 2001; Smith et al., 2009). Evaluating the effectiveness of agents in 15

reducing weed populations and the response of associated plant communities to the reduction 16

or removal of the weed is often neglected or omitted altogether due to inadequate resources 17

and funding (McClay, 1995; Blossey and Skinner, 2000; Syrett et al., 2000; Thomas and 18

Reid, 2007; Carson et al., 2008). Furthermore, the long-term economic and social outcomes 19

from biological control programs are generally poorly documented (Syrett et al., 2000; Jetter, 20

2005).21

22

There is growing pressure on practitioners to undertake research to identify, prior to their 23

release, those agents that are most likely to be effective (Holt and Hochberg, 2001; Pearson 24

and Callway, 2003; McClay and Balciunas, 2005; van Klinken and Raghu, 2006). This is due 25

4

to the costs involved in testing and releasing ineffective agents, and the need to minimize the 1

number of deliberate introductions of exotic organisms to address public concerns about risks 2

of off-target effects (Ewel et al., 1999; Louda et al., 2003; Willis et al., 2003; Willis and 3

Memmott, 2005; Pearson and Callaway, 2003). Despite these pressures, there are 4

disincentives to undertake pre-release evaluation of agent effectiveness because of the 5

additional time and resources required and the potential likelihood of rejecting agents that 6

could be effective in the introduced range (McClay and Balciunas, 2005).7

8

Pre-release modelling and experimental studies to evaluate the effectiveness of prospective 9

agents can provide some indication of their potential to adversely affect key growth 10

parameters of the target weed and thus assist the agent selection and prioritisation process 11

(Sheppard, 2003; McClay and Balciunas, 2005; Morin et al., 2006; Raghu et al., 2006b, 12

2007). They will however, never fully predict how a candidate agent will perform on weed 13

populations in the introduced range, when faced with a new set of environmental conditions, 14

particularly if the agent suffers from high rates of parasitism or predation (McFadyen and 15

Spafford Jacob, 2003; Broughton and Pemberton, 2008). Pre-release experimental studies to 16

demonstrate the effectiveness of candidate agents are either performed in the field in the 17

native range of the target weed (e.g. Brun et al., 1995; Briese et al., 2003; Goolsby et al., 18

2004; Gerber et al., 2007a, b) or on individual plants under controlled conditions, often in 19

conjunction with host-specificity trials (e.g. Hasan and Aracil, 1991; Shishkoff and Bruckart, 20

1996; Balciunas and Smith, 2006; Baars et al., 2007). A priori predictions of agent efficacy 21

however, have so far not often been explicitly tested by quantifying effectiveness in the field 22

after release. 23

24

5

On the other hand, demonstrating that an agent introduced to a new country is effective in 1

reducing populations of a target weed across its range and over the long-term is a challenging 2

and non-trivial task. Not only do researchers need to determine whether the agent has 3

adversely affected the weed, they also need to show that the observed suppression is greater 4

than would be expected given the underlying spatio-temporal variability of the biological 5

system and abiotic conditions (McClay, 1995; Carson et al., 2008). Therefore, pre-release 6

measurements of target weed populations and associated plant communities are valuable to 7

provide baseline data against which the biological control program can be subsequently 8

evaluated (Evans and Landis, 2007; Yates and Murphy, 2008). To date, the majority of post-9

release evaluation studies have focused on subjective assessments of agent establishment and 10

performance at the individual plant level (McClay, 1995; Dhileepan, 2003a; Thomas and 11

Reid, 2007). Few studies have quantified the effects of agents in reducing weed populations 12

(e.g. Hoffmann and Moran, 1998; Sheppard et al., 2001; Buckley et al., 2003a, b; Dhileepan, 13

2003b; Story et al., 2006), fewer have attempted to quantitatively measure the response of 14

associated plant communities to the removal of the weed (e.g. Huffaker and Kennett, 1959; 15

McEvoy et al., 1991; Hunt-Joshi et al., 2004; Paynter, 2005; Barton et al., 2007; Dhileepan, 16

2007) and none, as far as we are aware have monitored changes in animal and microbial 17

communities and ecosystem processes (e.g. nutrient cycles, hydrological cycles). Assessing 18

the degree to which weed management programs in general are contributing to the 19

conservation of native communities and ecosystem processes is increasingly being performed 20

to determine if programs lead to the stable recovery of ecosystems (e.g. Erskine Ogden and 21

Rejmánek, 2005; Galatowitsch and Richardson, 2005; Gosper et al., 2006; Harms and 22

Hiebert, 2006; Blanchard and Holmes, 2008; French and Buckley, 2008).23

24

6

There are several reasons to invest resources in measuring the effectiveness of agents in 1

suppressing the target weed in the introduced range: 1) it can provide the necessary 2

information to fine-tune a biological control program’s strategy and improve future programs, 3

2) it may inform practitioners that other agents, which affect different life stages of the target 4

weed, are needed to exert additional stresses on the weed at both individual plant, and 5

population levels, and 3) it may indicate that altering other management practices would 6

increase agent effectiveness. Most importantly, evaluating the effectiveness of agents using a 7

range of ecological, economic and social indicators provides data to assess whether and to 8

what extent the agents have successfully suppressed weed populations and contributed to 9

aesthetic, recreational, human health, biodiversity and economic benefits (Syrett et al., 2000). 10

Such rigorous assessments can then be used to justify continued investment in biological 11

control research and implementation. 12

13

By reviewing key issues, available approaches and analyses, we aim to encourage and assist 14

practitioners in developing sound plans to evaluate the effectiveness of agents throughout the 15

lifespan of weed biological control programs. While we recognize that the majority of16

biological control practitioners are aware of the benefits of selecting highly effective agents 17

and measuring their effectiveness after release, there are still many programs that do not 18

include any pre- and post-release assessments of agent effectiveness. This review’s main goal 19

is to present a compilation of the various approaches, including their strengths and 20

weaknesses, which can be used for assessing agent effectiveness and not a summary of the 21

effectiveness of agents from different systems. Specifically we discuss: 1) the need to assess 22

ecological, economic and social impacts of the target weed before initiating a biological 23

control program, 2) the importance of setting performance targets at the start of a program and 24

gathering adequate baseline ecological data for future comparison, 3) the types of data 25

7

typically collected to assess agent effectiveness and 4) the experimental, modelling and 1

economic approaches that have been used to evaluate the effectiveness of agents pre- and 2

post-release. We will not discuss the evaluation of possible direct effects of biological control 3

agents on non-target plants as these were reviewed in detail by Barton (2004) and Willis et al. 4

(2003). While we have included examples and references from many different countries to 5

provide a global perspective to the review, we emphasize certain Australian examples that we 6

are well acquainted with to illustrate particular points.7

8

2. Ecological, economic and social impacts of the target weed9

10

Data on the ecological and socio-economic impacts of a target weed prior to implementing a 11

biological control program provide crucial information to use later for comparison when 12

evaluating the effectiveness of introduced agents. Such information also helps determine 13

whether the investment in a program is justified in the first place. It can also assist in 14

identifying potential conflicts of interest, such as when the target weed is of economic benefit 15

to some industries or individuals (Delfosse, 1989). If data show that the weed is a symptom 16

rather than the cause of the degradation of natural or agricultural ecosystems (MacDougall 17

and Turkington, 2005), a biological control program may not necessarily deliver economic or 18

ecological benefits unless the underlying causes of the initial invasion are addressed (Hulme 19

2006). 20

21

The ecological damage caused by a weed can be assessed through field studies that correlate 22

weed abundance with native plant abundance and diversity, and ecosystem processes, or that 23

measure native community composition in plots with weeds removed and weed-infested plots 24

(Grice, 2004; MacDougall and Turkington, 2005; Hulme and Bremner, 2006). Information on 25

8

the current economic damage caused by the weed and potential adverse impacts should the 1

species spread beyond its present distribution will also provide essential data for future 2

comparison (Leistritz et al., 1992; Walker et al., 2005; Kaiser, 2006; Ireson et al., 2007; Juliá 3

et al., 2007). There is however, often a scarcity of useful quantitative economic data of weed 4

impacts in the literature, and therefore consultation with key stakeholder groups is frequently 5

necessary to obtain realistic estimates (Eiswerth et al., 2005; Page and Lacey, 2006).6

Information on the reduction in agricultural output (yield) due to a weed and financial costs of 7

its control is generally readily available for many intensive agriculture industries (Sinden et 8

al., 2004; O’Donovan et al., 2005), but weed biological control has rarely been contemplated 9

for such systems. In contrast, there is generally limited information on the impact of weeds on 10

pasture or rangeland production and the subsequent reduction in carrying capacity (e.g. for 11

livestock) (Auld et al., 1987). Estimating the financial costs of a weed invading natural 12

ecosystems is a recurrent challenge due to lack of data on the value of ecosystem and 13

recreational services that are lost due to the invasion (Sinden et al., 2004; Eiswerth et al., 14

2005; Kaiser, 2005; Sinden and Griffith, 2007). Consequently, economic analyses of 15

environmental weeds often only include estimates of control costs. 16

17

Consultation with health centres and review of medical studies can be used to reveal social 18

costs of some weeds. For example, Bass et al. (2000) found that pollen from common 19

ragweed (Ambrosia artemisiifolia L.) caused severe hay fever and asthma. Similarly, pollen 20

and plant dust of parthenium (Parthenium hysterophorus L.) result in allergenic respiratory 21

reactions and contact dermatitis (Khosla and Sobti, 1979; McFadyen 1995). An increase in 22

numbers of patients suffering from vector-borne tropical diseases has been attributed to the 23

expansion of water hyacinth (Eichhornia crassipes (Mart.) Solms) infestations in Lake 24

Victoria, Africa (Epstein, 1998). In this case, dense weed populations have provided a refuge 25

9

for breeding of Anopheles spp. mosquitoes and the snail Biomphalaria sudanica Martens, 1

vectors of malaria and schistosomiasis, respectively (Navarro and Phiri, 2000; Plummer, 2

2005). 3

4

The scarcity of relevant quantitative information on the ecological, economic and social 5

impacts of a weed often encountered before the initiation of a biological control program is a 6

challenge for practitioners. They can partially circumvent this difficulty by encouraging other 7

researchers interested in broader issues relating to plant invasion mechanisms to use their 8

target weed as a model system for obtaining relevant ecological and economic impact data. 9

10

3. Setting performance targets and gathering detailed baseline 11

ecological data 12

13

Biological control programs benefit from having well-defined performance targets. Setting 14

weed management outcomes to aim for provides quantitative benchmarks against which 15

achievements can be subsequently evaluated (van Klinken and Raghu, 2006; Thomas and 16

Reid, 2007). These targets are crucial to assist the selection of an appropriate candidate agent 17

to increase likelihood of achieving the desirable level of control of the target weed. They also 18

provide the basis to select the design, scale and duration of experiments to measure agent 19

effectiveness after release and to determine the type, timing and frequency of data collection 20

(Vos et al., 2000; Hobbs, 2003; Roni et al., 2005).21

22

Performance targets may include a quantitative estimate of 1) the required decline in 23

propagule production to reduce weed spread, 2) the level of control necessary to make the 24

weed more amenable to other control measures and/or, 3) the necessary suppression in 25

10

populations over a defined timeframe to reduce the weed’s economic or ecological impacts to 1

an acceptable level. For example, reducing the percentage cover or biomass of the target weed 2

per unit area below 50% for 10 continuous years may be deemed necessary to obtain a 3

satisfactory increase in desirable vegetation at infested sites. Setting such a target implies that 4

there is a functional relationship between weed abundance or density and recovery of 5

desirable vegetation and/or agricultural productivity. Knowing the shape of these impact-6

density curves allows an observed level of decline in the weed to be equated with a degree of 7

economic or ecological benefit and can considerably improve management decisions 8

(Yokomizo et al. 2009). However, this requires considerable knowledge of the impacts of the 9

weed and the level of control necessary to ameliorate them (Eiswerth et al., 2005; Jetter, 10

2005; Kaiser, 2006). Unfortunately, this information is rarely available for environmental 11

weeds, and consequently there is great uncertainty associated with performance target 12

estimates (Jetter, 2005; Hulme, 2006; Thomas and Reid, 2007). 13

14

Detailed baseline data on selected populations of the target weed across its introduced range 15

collected soon after the inception of a biological control program and well before any agents 16

are released will later facilitate assessment of agent effectiveness (Thomas and Reid, 2007; 17

Carson et al., 2008). Weed demographic data are also helpful to parameterize models to 18

predict agent efficacy and assist with agent selection. Data collected over multiple, 19

consecutive years before the first agent is released should provide a realistic indication of 20

spatio-temporal fluctuations in plant population parameters (Underwood, 1993). Pre-release 21

baseline ecological data can also help narrow down the number of different measurements 22

that will have to be collected following the release of agents, based on relationships 23

established between various weed parameters (e.g. Story et al., 2001; Swirepik and Smyth, 24

2003), providing the agents do not alter these relationships. 25

11

1

Measuring key parameters of plant communities co-existing with the weed will enable 2

subsequent assessment of their response following a reduction in the weed population due to 3

biological control (e.g. Lesica and Hanna, 2004, Schooler et al., 2006b). It may also be 4

beneficial in some instances to gather baseline data on ecosystem processes at weed-infested 5

sites (Vitousek, 1990; Denslow and D’Antonio, 2005; Hulme, 2006). It is too often assumed 6

that associated plant, animal and microbial communities and ecosystem processes will 7

automatically recover as a result of a local reduction in the size or density of weed 8

populations (Thomas and Reid, 2007; Carson et al., 2008). Changes in weed populations are 9

unfortunately not necessarily positively or linearly related to impacts on the ecosystems.10

Biological control practitioners need to be aware of the wider ecosystem implications of 11

releasing agents and need to promote a whole-system weed management approach so that 12

weeds managed by biological control are not replaced by other weeds (Müller-Schärer and 13

Frantzen 1996; Hulme, 2006; Buckley et al., 2007; Buckley, 2008; Carson et al., 2008).14

15

4. Types of data collected to assess agent effectiveness 16

17

Agent-related measurements18

Several different types of measurements can be made to quantify damage caused by agents at 19

different densities on individual plants. Typically for insect agents, density is assessed by 20

counting the number of individuals of each life-stage on plants. This information is valuable 21

to understand trends in the agent populations, phenology and dispersal patterns. To quantify 22

feeding damage, measurements may include portions of leaf area removed, proportion of 23

leaves exhibiting agent damage, severity of the damage per unit size/biomass, litterfall, 24

number of stem tips damaged, gall density, distribution of galls on the plant, number of 25

12

oviposition marks or loss of photosynthetic tissue (e.g. Dhileepan and McFadyen, 2001; 1

Goolsby et al., 2004; Paynter, 2006; Gerber et al., 2007a, b, 2008). Such measurements may 2

not be straightforward for some agents, due to difficulties in sampling (e.g. large trees) or the 3

nature of insect feeding mode (e.g. endophagous or below-ground).4

5

Damage assessment for pathogen agents also depends on their life-history characteristics. For 6

example, a visual assessment of gall numbers per tree was performed for the gall-forming rust 7

pathogen Uromycladium tepperianum (Sacc.) on Port Jackson wattle (Acacia saligna (Labill.) 8

H.L.Wendl) in South Africa (Wood and Morris, 2007). In contrast, for pathogens infecting 9

leaves incidence (e.g. percentage of foliage or number of leaves with necrotic lesions or rust 10

pustules) and severity (e.g. portions of leaf area infected or number of lesions per leaf) of 11

disease symptoms on individual plants are typically assessed (e.g. Baudoin et al., 1993; Brun 12

et al., 1995). Disease incidence can also be determined at the population level by estimating 13

percent cover or counting the number of plants (or shoots), within a defined area, that are 14

infected by the pathogen (e.g. Morin et al., 2002; Dhileepan et al., 2006). These 15

measurements not only provide a quantitative assessment of damage but also give a measure 16

of the pathogen population density. Pictorial keys can be developed to standardize and 17

streamline visual assessments of severity of disease symptoms on leaves or stems (James, 18

1971; Stonehouse, 1994). 19

20

Quantification of agent incidence and damage is not in itself informative if it is not related to 21

effects on weed life-history parameters, taking into account density-dependent relationships 22

and the ability of plants to compensate for damage (e.g. Buckley et al., 2005b; Schooler and 23

McEvoy, 2006; Gerber et al., 2007a, b, 2008). Such relationships are essential to predict if the 24

13

agent will have strong effects on weed abundance and spread (McEvoy and Coombs, 1999; 1

Shea, 2004; Davis et al., 2006).2

3

Weed-related measurements4

Measurements on the weed are ideally taken at the individual plant and population levels. 5

Growth (number, size and biomass of above- and below-ground parts) and reproductive6

(number and biomass of flowers, fruits and seeds) parameters, and survival are typically 7

measured for individual plants. At the weed population level, measurements may include 8

weed density and cover, stand size and age structure, seedling recruitment and survival, viable 9

seed bank density and spread. The specific measurements collected will depend on the weed-10

agent system. For example, the number of viable seeds produced by the weed will be 11

measured if a seed-feeding agent is anticipated for release, but it is important to verify if this 12

parameter affects the magnitude of the weed’s seedbank and its recruitment, and weed 13

population dynamics (McEvoy et al., 1991; Kriticos et al., 1999; Buckley et al., 2005b). 14

Gathering fecundity data may also be important for other types of agents since reduction in 15

plant growth may indirectly reduce fecundity. The weed’s growth form, life history 16

characteristics and ecological context are also considered when designing protocols to 17

measure change in populations, so that relevant data can be collected to a high level of 18

resolution. For example, artificial trellises had to be used to support climbing shoots of the 19

environmental weed bridal creeper (Asparagus asparagoides (L.) Druce), on which most fruit 20

production occurs, to collect data on reproductive output (Stansbury et al., 2007). Gathering 21

weed demographic information across broad spatial scales is necessary to test whether 22

different populations respond similarly to agents (Shea and Kelly 1998; Buckley et al. 2003a;23

Davis et al. 2006).24

25

14

Ecosystem-related measurements1

The response of an ecosystem following reduction of a dominant weed population by 2

biological control agents can be assessed by measuring plant community parameters such as 3

diversity and abundance of native and weedy plant species, and productivity in 4

agricultural/forestry systems (e.g. pasture carrying capacity, crop yield). Measurements of 5

other characteristics of the ecosystem, such as species richness and abundance of beneficial 6

arthropods and microbes, soil nutrient levels and water quality, or ecosystem processes, such 7

as rate of nutrient cycling and decomposition, and changes in hydrological cycles may also be 8

considered (Vitousek, 1990; Denslow and D’Antonio, 2005; Hulme, 2006). 9

10

5. Laboratory and glasshouse studies 11

12

Laboratory or glasshouse experiments that simulate possible agent damage on the target weed 13

can be performed ahead of any agent selection, to identify the types of damage that will 14

severely affect individual plants (e.g. Schat and Blossey, 2005; Schooler et al., 2006a; 15

Dhileepan et al., 2009). For example, Raghu et al. (2006a) used glasshouse simulated 16

herbivory experiments to assess how damage to various plant parts affected growth of the 17

invasive liana cat’s claw creeper (Macfadyena unguis-cati (L.) A. H. Gentry). They found that 18

artificial defoliation caused the greatest reduction in above and below-ground biomass. In 19

contrast, direct damage to below-ground organs only had a negligible effect on overall plant 20

productivity and even triggered compensatory responses by the plant. These results suggested 21

that specialist herbivores in the leaf-feeding guild capable of removing over 50% of the leaf 22

tissue would be desirable for the biological control of this species. Results of simulated 23

experiments however, are of limited value if the simulated damage does not compare with 24

15

that of actual agent damage observed in the laboratory, glasshouse or field (Hjalten, 2004; 1

Lehtilä and Boalt, 2004; Raghu and Dhileepan, 2005).2

3

In situations where field studies cannot be carried out in the native range, laboratory or 4

glasshouse experiments are typically performed to demonstrate the potential effectiveness of 5

prospective agents (e.g. Shishkoff and Bruckart, 1996; Evans and Bruzzese, 2003; Smith, 6

2005). These experiments enable detailed measurements of key weed growth parameters 7

affected by the candidate agent, including those such as below-ground biomass that are often 8

difficult to measure under field conditions (Morin et al., 2002; Raghu et al., 2006a) and 9

provide useful data for parameterizing models (Buckley et al., 2005b). They also provide 10

opportunities to test the influence of multiple agents, plant competition or different abiotic 11

factors on agent effectiveness (e.g. Willis et al., 1998; Callaway et al., 1999; Heard and 12

Winterton, 2000). An indication of the effects of an agent on pair-wise competitive 13

interactions between a weed and co-occurring desirable plant species can be obtained from 14

pot experiments (e.g. Groves and Williams, 1975). 15

16

The drawback of laboratory and glasshouse experiments is that they are of limited value for 17

long-lived trees and shrubs (Wirf, 2006). They also tend to concentrate on measuring effects 18

of agents on individual plants and early life-stages under a prescribed set of environmental 19

conditions, and consequently can generate exaggerated or unrealistic results (McClay, 1995; 20

Raghu et al., 2006b). Conversely, there are examples of agents, such as the rust fungus 21

Puccinia xanthii Schw. on Noogoora burr (Xanthium occidentale L.) and the mealybug 22

Hypogeococcus pungens (Granara de Willink) on harrisia cactus (Harrisia martini (Labour.) 23

Britton), that were predicted to be ineffective in controlling populations based on results of 24

laboratory experiments (Julien et al., 1979; McFadyen, 1979), but which built up and 25

16

maintained large populations after release, and effectively controlled their target weeds 1

(McFadyen and Tomley, 1981; Chippendale, 1995; Morin et al., 1996). These cases highlight 2

the importance of combining results of laboratory experiments with that of field studies in the 3

native range and demographic modelling, to generate more robust predictions of agent 4

efficacy.5

6

6. Field studies7

8

There are several approaches that have been used to assess the effectiveness of prospective 9

agents in the field in the native range or after their release in the introduced range. With all 10

approaches except for manipulative experiments, the effects of agents may not be easily 11

separated from those caused by a variety of confounding variables. For example, abiotic or 12

biotic differences in sites and/or years can influence both agents and weeds, and the 13

competition or facilitation exerted by the associated plant communities on the weeds 14

(McClay, 1995; Müller-Schärer and Schaffner, 2008). To account for these possible 15

confounding factors and ascertain that observed changes are primarily due to the agents, field 16

studies are conducted over several years and at many representative sites across different 17

climatic regions and habitats (e.g. Swirepik and Smyth, 2003; McConnachie et al., 2004; 18

Denoth and Myers, 2005). Although wide fluctuations in plant populations between years are 19

less common for perennial weeds, capturing variability over several years is still important 20

because it may take considerable time for these weeds to respond to stresses exerted by the 21

agents (Buckley et al., 2004). In many instances, firm evidence of the effectiveness of the 22

agent in damaging the weed and affecting weed populations can only be obtained by carrying 23

out different types of field studies, including the more robust manipulative experiments, 24

17

coupled with laboratory or glasshouse studies and demographic modelling, whenever possible 1

(Briese, 2006). 2

3

Correlative studies4

Correlative field studies aim to determine the relationship between agent densities/damage 5

levels and the weed growth and reproductive performance. They may be carried out in the 6

native range of the target weed to demonstrate the effectiveness of specific natural enemies in 7

debilitating the weed (e.g. Brun et al., 1995; Wood, 2002; Briese et al., 2003; Hafliger et al.,8

2005). For example, Briese (2000) demonstrated a strong relationship between attack levels 9

by the seed weevil Larinus latus Herbst and seed destruction of Scotch thistle (Onopordum 10

spp.) by surveying 46 populations in the native range. He conceded however, that the effect of 11

reduced seed production due to the agent on the weed population dynamics may be negligible 12

since Scotch thistle has a long-lived seed bank. This case highlights the importance of 13

including measurements at the weed population level to more accurately predict agent 14

efficacy. 15

16

Many agents in the native range may be host limited (bottom up effect), making it difficult to 17

assess their effectiveness in reducing plant populations. Common-garden experiments in the 18

native range, where a high density population of the target weed is artificially created 19

(equivalent to those found in the introduced range) can be used to remove this resource 20

constraint to obtain measurements of agent effectiveness that are more realistic for the 21

introduced range. Correlative studies involving target density manipulation in the native range 22

can therefore provide an opportunity to investigate the population dynamics of the candidate 23

agent and identify the key ecological mechanisms that may be limiting its effectiveness i.e. 24

bottom up and top down limitation of population growth (Blossey, 1995; Sheppard, 2003; 25

18

Sheppard, 2006). Measuring the levels of predation or parasitism affecting the agent 1

population (top down effect) may also assist in interpreting data collected on their 2

effectiveness in reducing the weed population should such agents escape this following 3

release (Gassmann, 1996). 4

5

Determining if a candidate agent affects weed populations in the native range may not be 6

relevant for assessing effectiveness if other factors such as plant competition rather than 7

natural enemies are the main limiting factors on target populations. In the introduced range, 8

where weed densities are much higher the agent may no longer be resource limited or could 9

escape its co-evolved parasites/predators. Sheppard (2003) highlighted that candidate agents 10

likely to be effective in reducing weed populations in the introduced range are generally 1) 11

widely distributed in the native range, 2) capable of attacking the most vulnerable plant stage12

identified in weed population models based on introduced range data and, 3) able to outbreak 13

and kill their host, at least in some circumstances. A retrospective analysis of past biological 14

control programs would be a first step towards testing these predictions.15

16

Correlating agent densities and damage levels to changes in growth and reproductive 17

parameters of individual plants has also been used to assess the effectiveness of released 18

agents (e.g. Swirepik and Smyth, 2003; Schooler and McEvoy, 2006; Wood and Morris, 19

2007). A more comprehensive assessment of the effectiveness of agents can be obtained by 20

extending measurements to plant population parameters (e.g. Davis et al. 2006). Hoffmann 21

and Moran (1991) for example, showed that the 98% seed reduction in Sesbania punicea 22

(Cav.) Benth.) caused by the weevil Trichapion lativentre (Beguin-Billecocq) did not lead to 23

a decline in the density of mature plants. 24

25

19

Comparing sites (or plots) with and without agents1

When agents have not been widely released or have not spread extensively by natural means, 2

it is possible to compare weed population parameters at sites or plots where agents are present 3

with others where they are absent (e.g. Dhileepan et al., 2000a; Paynter, 2005). Carson et al. 4

(2008) recommend releasing agents in randomly selected sites that are paired with non-release5

sites (control sites) in a replicated manner, stratified across relevant temporal and spatial 6

biotic and abiotic gradients. Paynter (2005) conducted such comparisons over three years to 7

quantify the effects of the stem-mining moth Carmenta mimosa Eichlin & Passoa on mimosa 8

(Mimosa pigra L.). In this study, differences in litter and seed fall, seed banks, vegetation 9

cover, size, density and age structure of mimosa stands at nine sites where the stem-mining 10

moth was present were compared with similar data taken at eight sites where the agent was 11

absent. The study found that mimosa seed rain, seed banks and seedling establishment were 12

reduced at C. mimosa-infested sites compared to that of the other sites. Such comparative 13

field studies however, are often not appropriate because many agents, in particular Diptera, 14

Lepidoptera, Hymenoptera as well as some Coleoptera insects, and rust pathogens can rapidly 15

disperse over long distances soon after their release (e.g. Cullen et al., 1973; Edwards et al., 16

1999; Cruz et al., 2006).17

18

Comparing before- and after-release data 19

Photo points. The simplest and cheapest approach to evaluate effectiveness of agents is to 20

establish fixed reference points at weed-invaded sites before or soon after the release of 21

agents and take a series of photos from those at regular intervals after release (e.g. Room et 22

al., 1981; Johnston and Lloyd, 1982; Freeman, 1992; Trujillo et al., 2001; Dhileepan, 2003a). 23

This approach is suitable when the effects of the agent(s) are visually conspicuous and a 24

dramatic decline in weed density and cover is observed. For example, comparison of photos 25

20

taken at fixed reference points of infestations of the aquatic weed red waterfern (Azolla 1

filiculoides Lamarck) before and after the introduction of the frond-feeding weevil 2

Stenopelmus rufinasus Gyllenhal in South Africa, showed the collapse of the weed mat 270 to 3

312 days after release (McConnachie et al., 2004). Similar photos of infestations of terrestrial 4

weeds can also show recovery of associated plant communities over time, although it is often 5

difficult to tell apart the individual species that have replaced the weed. 6

7

It can be difficult however, to measure modest changes in weed density from such a photo 8

series. Where appropriate, photographs of vertical views of fixed areas (e.g. quadrats) are 9

more suitable for measuring changes in density because they can be analyzed with high-10

precision image analysis software (Hall, 2001). Decrease in weed density may not be easily 11

detected in photographs if other species with similar growth habits colonize the site. In these 12

instances, additional information on types of plants present at photo points has to be collected. 13

Furthermore, photographs do not provide evidence that the decline in weed populations is due 14

to the agent(s), unless it is coupled with assessments of a priori weed abundance, agent 15

damage and density.16

17

Comparing historical and contemporary data. Published and unpublished historical 18

information, such as field ecological data and aerial or satellite photos of specific weed-19

infested sites, collected regardless of the initiation of a biological control program, can 20

provide baseline data to compare with contemporary data taken at the same sites. 21

22

Comparing historical ecological data with more recently acquired data is most useful when 23

they are of equal quality and exist for several sites. Ireson et al. (2007) recently collected data 24

on the seed bank and rosette densities of common ragwort (Senecio jacobaea L.) in fixed 25

21

quadrats at four sites in the north of the state of Tasmania, Australia where similar data had 1

been collected from 1979-1981. The data had been gathered as part of a demographic study 2

that was performed independent of the initiation of a biological control program (but 3

coinciding with the release of the ragwort flea beetle, Longitarsus flavicornis (Stephens) in 4

1979) and was never published. Seed bank and rosette densities were found to have 5

considerably declined when the historical and contemporary datasets were compared. 6

Additionally, during the contemporary survey, larval populations of the ragwort flea beetle 7

were found actively feeding on medium to large rosettes of the weed in the vicinity of three of 8

the four samples sites. Change in grazing management or use of herbicides were ruled out as 9

having contributed to the reduction of ragwort populations at these sites. 10

11

Historical maps of a weed distribution before the start of a biological control program can be 12

compared with maps assembled years after the release of agents to determine the effectiveness 13

of the program at a landscape scale (Henderson, 1999). Aerial and satellite imagery can be 14

particularly powerful by providing a long data series prior, during and after release of agents 15

over a large spatial scale (e.g. Everitt et al., 2005; Lass et al., 2005). Colour aerial 16

photography and hyperspectral remote sensing have been used to measure changes in 17

saltcedar (Tamarix spp.) distribution in western United States and to identify areas where 18

agents were most active in defoliating trees (Anderson et al., 2005; Curruthers et al., 2006). 19

The use of remote sensing however, is not necessarily applicable to all situations: it requires a 20

certain level of expertise to analyze the data and is only appropriate for a limited number of 21

weed species that can be visually separated from other background vegetation (Everitt et al., 22

1995; Lass et al., 2005).23

24

22

Stakeholder surveys. Perceptions of land managers on the prevalence of a target weed and 1

the ecological, economic and/or social problems that it causes can be surveyed before the 2

initiation of a biological control program and compared with answers obtained in subsequent 3

surveys performed years after the release of the agent(s). Alternatively, a single survey 4

performed after agents have spread extensively and are believed to be affecting the weed can 5

be used to assemble similar data, providing stakeholders have a good recollection of the 6

extent of the weed problem before the release of agents. Information obtained from more 7

general surveys may also provide some indications of the benefits of a biological control 8

program. For example, Tasmanian land managers were surveyed to list the weeds on their 9

property in order of importance and indicate the economic impact of each weed and whether 10

problems caused by the weeds had changed over the last 10 years (Ireson et al., 2007). 11

Respondents indicated a consistent decrease in the problem status of European blackberry 12

(Rubus fruticosus L. aggregate) and common ragwort over that period, two species on which 13

biological control agents were released in the late 1970’s and mid 1980’s. Such surveys are a 14

relatively inexpensive way to obtain a qualitative assessment of the status of weeds following 15

the release of agents, but may be deceptive because the change in status may have been 16

independent of the biological control program. 17

18

Comparing quantitative ecological data. Quantitative data collected on selected target weed 19

populations, co-existing plant communities and/or other ecosystem parameters for several 20

years before the release of agent(s) can be compared with similar data gathered after their 21

release (McClay, 1995; Blossey, 1999; Carson et al., 2008). Most studies published so far 22

have concentrated on monitoring changes in weed population growth and reproductive 23

parameters along with agent abundance and individual plant damage (e.g. Denoth and Myers, 24

2005; Grevstad, 2006). For example, Paynter (2006) compared litterfall data collected for two 25

23

years in the early 1980s at a site infested by mimosa in Northern Territory, Australia1

(Lonsdale, 1988), just prior to the release of agents, with similar data gathered between 2001 2

and 2003 along the original transects at the same site. Results showed that seed rain had 3

declined by approximately 60% at the edge of the weed stand, where the most severe damage 4

caused by the twig and stem-mining moths Neurostrota gunniella Busck and C. mimosa (both 5

released in 1989) was recorded. Nonetheless the recovery of competing vegetation at the edge 6

of the stand, due to the culling of feral Asiatic water buffalos (Bubalis bubalis Lydekker) in 7

the late 1980s - early 1990s, was identified as another possible contributor to the suppression 8

of mimosa recruitment and hence a decline in litterfall over time. 9

10

Changes in plant communities and other ecosystem characteristics, as a result of a decline in 11

weed populations due to biological control, are still seldom quantified (Thomas and Reid, 12

2007; Carson et al., 2008). Barton et al.’s (2007) recent study which documented the recovery 13

of native plants following the rapid decline of mistflower (Ageratina riparia (Regel) R. King 14

and H. Robinson) populations after the introduction of the fungus Entyloma ageratinae15

Barreto and Evans in New Zealand, is a good example of this type of research.16

17

Manipulative experiments18

Exclusion experiments using cages, insecticides and/or fungicides are performed to determine 19

the contribution that agents have in regulating the target weed population (e.g. Burdon et al., 20

2000; Sheppard et al., 2001; Dhileepan, 2003b; Briese et al., 2004; Goolsby et al., 2004; 21

Hunt-Joshi et al., 2004; Tipping et al., 2008). Inclusion experiments, whereby an insect agent 22

is added to cages containing the weed in the field, have also been used to measure agent 23

effectiveness in the native (e.g. Briese et al. 2002) and introduced ranges (e.g. Dhileepan et 24

al., 2000b; Dhileepan, 2003c). Inclusion experiments however, are not frequently used 25

24

because they can generate exaggerated results due to the often unrealistically high agent 1

population that develops within the cages. 2

3

Exclusion and inclusion experiments are useful to relate agent densities directly to plant 4

performance, independent of confounding external factors (McClay, 1995), and their design 5

generally allows for rigorous statistical analysis of agent effects (Farrell and Lonsdale, 1997; 6

Carson et al., 2008). They can also be used to differentiate between the effects of multiple 7

agents (James et al., 1992; Paynter et al., 2006) and measure the initial response of the co-8

existing plant community to a reduction in weed competitiveness or density, if plots/cages are 9

sufficiently large (Hunt-Joshi et al., 2004). They are however, time-consuming and expensive 10

and therefore are generally conducted on a small scale for relatively short periods of time. 11

Furthermore, researchers should be cautious in extrapolating the results from small-scale 12

exclusion studies to broader-scale predictions. Nonetheless, we believe that exclusion 13

experiments are worth conducting to obtain comparisons of agent effectiveness within sites 14

over one or several years.15

16

Carefully designed and implemented exclusion experiments are essential to account for any 17

effects the chemical or cage treatments may have on the whole system (Lonsdale and Farrell, 18

1998). These types of experiments have been used on weedy tree saplings (Tipping et al., 19

2008), but are generally not suitable for larger woody weeds and high rainfall areas because it 20

may be difficult to continuously exclude agents with cages or chemical applications (Paynter, 21

2004). They still however, give an indication of the response of the weed to a reduction in 22

agent damage. Physical barriers, such as nylon mesh used in cages, can reduce light levels and 23

evapo-transpiration rates, which in turn may alter plant performance (Hand and Keaster, 24

1967; Kidd and Jervis, 1996). The exclusion treatment can also affect interactions other than 25

25

those occurring between the weed and the agent (e.g. with pollinators, predators and 1

parasitoids), and these effects need to be taken into consideration during analysis and 2

interpretation of results (Lonsdale and Farrell, 1998; Hunt-Joshi et al., 2004; Maron et al.,3

2004). The chemical used to exclude the target agent may also have unwanted fertilizing or 4

phytotoxic effects that affect plant performance and vigour (Baudoin et al., 1993; Tipping and 5

Center, 2002; Siemann et al., 2004). Experiments that compare the performance of plants, 6

independent of agents, in caged or chemically-treated plots versus untreated plots are 7

necessary to estimate the effect of applying the exclusion treatment per se, which can be taken 8

into account when interpreting results of exclusion experiments (e.g. Adair and Holtkamp, 9

1999). 10

11

Exclusion and inclusion experiments may involve the transfer of potted plants of the target 12

weed or transplantation of standardized plants to the field to enable detailed examination of 13

responses to agents under natural conditions, while controlling for differences in soil and 14

plant size, age and genotype (e.g. Baudoin et al., 1993; Dhileepan et al., 2000b; Corn et al., 15

2006). One important advantage of such pot or planting methods is the ability to retrieve 16

entire individual plants at the end of the experiment, making it easier to examine the effect of17

the agent on below-ground biomass. The main disadvantage of transferring potted plants to 18

the field is that it makes exclusion experiments even more labour-intensive and generally 19

requires new plants to be transferred each growing season. Furthermore, this approach focuses 20

on single plant and early life-stage responses, which may lead to overestimates of agent 21

effects, if they are extrapolated to mixed-age populations. Nonetheless, differences in the 22

response of different life stages as well as between transplanted, potted and field plants can be 23

investigated experimentally. 24

25

26

7. Demographic modelling1

2

Models of plant population dynamics can assist with the selection of effective agents by 3

identifying the most vulnerable stages of the weed’s lifecycle where damage would cause 4

major adverse effects on populations (e.g. Smith et al., 1997; Buckley et al., 2005a; Raghu et 5

al., 2006b, 2007). For example, Buckley et al. (2001) simulated control strategies for scentless 6

chamomile (Tripleurospermum perforatum (Merat) M. Lainz) at a number of different life 7

history stages and found that control regimes need to integrate rigorous and continuous 8

reductions in fecundity and survival late in the growing season in order to manage this 9

species. Similarly, demographic models on garlic mustard (Alliaria petiolate (Bieb.) Cavara 10

& Grande) demonstrated that biological control agents would need to directly affect fecundity 11

and rosette-to-flowering-plant stage to adversely affect population growth (Davis et al. 2006). 12

As far as we are aware none of the predictions made by the models mentioned above have 13

been explicitly tested in the field yet. Models simulating insect herbivore attack on seeds of 14

nodding thistle (Carduus nutans L.) (Shea and Kelly, 1998) and Scotch broom (Cytisus 15

scoparius (L.) Link) (Parker, 2000) showed that losses of approximately 70% of seeds would 16

be required to decrease populations of these weeds. If the Shea and Kelly (1998) and Parker 17

(2000) models had been developed prior to release, seed feeding agents may have been 18

considered ineffective in decreasing populations and subsequently not introduced. The 19

proportional contribution of fecundity to modelled spread speed in Scotch broom is similarly 20

low (Neubert and Parker 2004). Models can therefore serve as a guide for estimating desirable 21

levels of impacts by agents. 22

23

Collection of data needed to parameterize such detailed species-specific population models 24

can be time-consuming, and the resulting model can be complex. The problem is exacerbated 25

27

if data are collected without reference to the needs of the model being used. To overcome the 1

impediment caused by lack of detailed weed demographic data, Ramula et al. (2008) recently 2

used comparative demographic models of invasive and native plants to identify general 3

targets for control, depending on plant life-history characteristics. Their study revealed that 4

growth and fecundity transitions of short-lived weeds should be prioritized as control targets, 5

while for long-lived weeds, targeting more than one demographic process, preferably survival 6

and growth, is required to reduce populations.7

8

Even well-parameterized models may not accurately predict the outcomes of a biological 9

control program as agent effectiveness is only one of many factors influencing the interaction 10

with the target weed. For example, the agents may not establish after their release, they may 11

rapidly acquire pre-adapted natural enemies that severely affect population build-up or they 12

may be unable to tolerate climatic variations in the introduced range (McFadyen and Spafford 13

Jacob 2003). Nevertheless, models constructed in the early stage of a biological control 14

program and parameterized with pre-release agent efficacy data provide crucial baseline 15

information that can be used to make predictions, which can be tested and updated as the 16

program proceeds and more data become available.17

18

Once an agent is released, models are useful in demonstrating a mechanistic connection 19

between the agent and reduced weed densities in order to exclude other explanations of 20

population decline of the weed over time (Kriticos, 2003). Population models can be 21

developed to predict the long-term dynamics of weed populations under pressure from the 22

agent and evaluate if the agent will be sufficiently effective to achieve the performance targets 23

of the program. For example, a model produced by Lonsdale et al. (1995) demonstrated that 24

continued herbivory of spinyhead sida (Sida acuta Burman f.) in the field by the chrysomelid 25

28

beetle Calligrapha pantherina Stål. should reduce the weed population to much lower 1

densities; the extent dependent on the search efficiency of the beetle. Similarly, a coupled 2

plant-herbivore model developed for Paterson’s curse (Echium plantagineum L.) predicted 3

stable coexistence of the weed and agent with the seed bank of the weed falling to less than a 4

third of its value when the herbivore was not present (Buckley et al, 2005b). This model 5

successfully predicted reductions in weed density observed in the field. 6

7

Modelling the effects of introduced agents on weed populations may emphasize the need for 8

further agents to be introduced to enhance impact on the same life stage or feeding niche or to 9

target different stages of the plant lifecycle. For example, Stuart et al. (2002) showed that the 10

recorded 23-31% reduction in viable seed production of bitou bush (Chrysanthemoides 11

monilifera ssp. rotundata (DC). T.Norl.) caused by the seed fly Mesoclanis polana Munro is 12

unlikely to affect the persistence and colonisation ability of this weed species. Further 13

modelling exploring interactions between this agent and other management techniques 14

indicated that additional agents which attack bitou bush foliage and open up the canopy are 15

required to significantly reduce the weed’s competitive ability (Kriticos et al., 2004). It is 16

noteworthy that the presence of an introduced biological control agent, even if it does not 17

reduce the weed density sufficiently, may influence the plant in such a way that the pre-18

release model of plant demography and parameters will need to be updated before using it to 19

assess the suitability of other candidate agents. Environmental conditions within the exotic 20

range that vary from those prevailing in the native range, on which the initial model was 21

based, may also influence the most appropriate stage of plant life-history to target with 22

additional biological control agents (Buckley et al., 2005b; Davis et al., 2006). 23

24

29

Models that show negligible effects of an agent on weed populations, may also point to the 1

need for integrating biological control with other management methods. Incorporating 2

multiple processes such as weed population dynamics, disturbance and existing or potential 3

biological and non-biological control tactics into models can assist in determining the long-4

term effect of the agent within a broader integrated management strategy for the target weed. 5

For example, a model for Scotch broom suggested that a substantial adverse effect of 6

biological control agents on the weed population would only occur if the disturbance rate was 7

high and fecundity and seedling survival were low; highlighting the need to integrate several 8

control tactics into the management program (Rees and Paynter, 1997). Through the use of 9

models based on two agents released in the late 1980s in Australia, Buckley et al. (2004) 10

predicted that it would take from 12-29 years to reduce the cover of mimosa from 90% to 11

<5% at sites where the biological control agents have established. Explicit modelling of 12

integrated weed management strategies also revealed that other approaches such as burning 13

and mechanical control were needed in conjunction with biological control to reduce the short 14

to medium-term detrimental impacts of mimosa. 15

16

A combination of empirical studies and models is particularly useful for evaluating the effect 17

of a recently introduced agent on target weed populations and provides the necessary data to 18

design a realistic long-term evaluation program to demonstrate whether or not the program 19

has been successful. For example, the data collected by Paynter (2005) on the effects of the 20

stem-mining moth C. mimosa successfully validated the spatial model developed by Buckley 21

et al. (2004). Through such coupled studies, biological control practitioners can be more 22

confident in the decisions they are making and subsequently develop more robust 23

management programs. Modelling based on results from empirical studies, and ideally with 24

30

exclusion experiments can also facilitate evaluation of biological control programs for which 1

limited prior data are available, even years after they were initiated.2

3

8. Benefit-cost analyses4

5

Benefit-cost analyses measure the expected or actual return on investment from a biological 6

control program, which is expressed as a benefit-cost ratio or as a net present value. A 7

benefit-cost ratio is derived by dividing the value of the losses avoided by affected industries 8

or other stakeholders (in other words what economic damage would be caused by the weed in 9

the absence of the project, including costs for control), by the research and development costs 10

of the program. A net present value is an indicator of how much cash value a project adds to 11

the value of the industry concerned. Both costs and benefits are discounted at a specified rate 12

to account for differences in the time when they were incurred; costs are up-front while13

benefits increase steadily as agents gradually affect the weed. Unpriced values for benefits 14

such as the preservation of native biodiversity and scenic amenity, improved recreational 15

access to land and water, reduction of fire hazard and decreased exposure of environment to 16

herbicides (Greer, 1995; Page and Lacey, 2006), and costs such as ‘collateral damage’ on 17

non-target plants (if applicable) (Nordblom, 2003; Jetter, 2005), are not included in the 18

analysis unless a definitive dollar value has been estimated (Auld et al., 1987; Eiswerth et al., 19

2005). Nevertheless, describing and quantifying the unpriced (priceless) environmental and 20

social benefits and costs can provide a critical complement to the economic analysis. 21

Anticipatory (ex-ante) benefit-cost studies are useful in showing the expected value of a 22

proposed biological control program. Retrospective (ex-post) benefit-cost studies are used to 23

demonstrate accomplishments of a program (Culliney, 2005).24

25

31

Benefit-cost analyses performed before or in the early stages of a biological control program 1

are valuable to predict the extent of expected effects of a candidate agent will translate into an 2

overall economic benefit. An ecological-economic model was used to demonstrate that the 3

benefits of releasing the rust fungus U. tepperianum for the biological control of Port Jackson 4

wattle in lowland fynbos ecosystems of South Africa could potentially outweigh the benefits 5

derived from harvesting the weed by the fuel wood industry (Higgins et al., 1997). Another 6

analysis showed that biological control of Scotch broom in New Zealand would result in a 7

favourable 2.9:1 benefit-cost ratio even after taking into account the potential negative 8

impacts on beekeepers, due to reduced pollen availability, and the non-target effects of the 9

agents on the fodder plant tagasaste (Chamaecytisus palmensis (Christ) Hutch.) (Jarvis et al., 10

2006). Due to the high number of unknown parameters before the start of a biological control 11

program, benefit-cost analyses may include several scenarios of projected cost and benefits 12

over time for different contingencies and uncertainty levels, with and without the program, 13

and different subjective probabilities assigned to each (Nordblom, 2003). It is noteworthy that 14

probabilistic analyses to estimate future benefits of biological control in light of agent 15

effectiveness and possible off-target effects, has only been published for the biological control 16

of Paterson’s curse in Australia (IAC, 1985; Jetter, 2005).17

18

Well-established biological control programs also use benefit-cost analyses to evaluate 19

economic impacts (e.g. Chippendale, 1995; Greer, 1995; Coombs et al., 1996; McConnachie 20

et al., 2003; van Wilgen et al., 2004; Culliney 2005). Page and Lacey (2006) recently carried 21

out a series of benefit-cost analyses to identify the role that weed biological control has 22

played in Australia. Their study revealed an outstanding overall benefit-cost ratio of 23:1, 23

based on increased production, control cost savings and health benefits. Considerable 24

limitations however, were encountered in gathering data on economic, environmental and 25

32

social impacts of the target weeds, their rate of spread and distribution, effectiveness of 1

biological control agents and research and development costs of programs. Benefit-cost 2

studies conducted in South Africa, also showed positive returns on investment in biological 3

control programs with benefit-cost ratios ranging from 8:1 for lantana (Lantana camara L.) to 4

as high as 709:1 for jointed cactus (Opuntia aurantiaca Lindl.) (van Wilgen et al., 2004). 5

Benefit-cost ratios typically increase drastically when future estimates of benefits are 6

considered. For example, the benefit-cost ratio of 2.5:1 calculated by McConnachie et al.7

(2003) for the biological control program against red waterfern in South Africa in the year 8

2000, increases to 13:1 in 2005 and 15:1 in 2010.9

10

Benefit-cost analyses are more robust when carried out after there has been ample time for 11

field impact of agents to be realized and sufficient long-term ecological data to be assembled. 12

They are also more convincing when based on accurate data on research and development 13

costs of programs and economic and social costs of the target weed and its rate of spread and 14

distribution before and after releases of agents. Comprehensive ecological data are a 15

prerequisite to support good decision-making in an economic framework (Turpie, 2004), 16

particularly where spatial and temporal differences in the interactions between the target weed 17

and agent(s) are observed or likely. 18

19

Economic analyses may also point to the need for redistribution of the introduced agents to 20

increase effectiveness across the range of the target weed and assist in identifying the 21

geographic locations to target for the highest returns on investment. Nordblom et al. (2002) 22

calculated where and how many additional releases of agents were economically justified to 23

speed up the process of agent spread and effective reduction in Paterson’s curse densities 24

across southern Australia. Expected marginal contributions in increasing the number of new 25

33

releases were simulated for 31 districts. The benefits of biological control were expressed in 1

terms of the expected value of recovered pasture productivity (due to weed population 2

suppression), keyed to estimates of loss and to historical district livestock inventories 3

converted to dry sheep equivalent feed availability levels. This conservative analysis showed 4

that 17 of these districts would offer marginal gains in productive value greater than or equal 5

to five times the marginal cost of a release. Under that criterion, some 400 new, targeted 6

releases were justified in those districts. The underlying assumption of the analysis that 7

removal of Paterson’s curse plants would lead to an increase of more beneficial plants in the 8

pasture within the annual cycle of the model was, however, not supported with data. The 9

timeframe over which weed reduction leads to an increase of desirable plants should be 10

explicitly tested and included in future economic impact models (A.W. Sheppard, pers. 11

comm.).12

13

9. Long-term evaluation14

15

Long-term evaluation of biological control agents is primarily performed to check if the 16

agents are performing as expected and to understand the factors that influence their 17

effectiveness. Continued, regular monitoring of the effects of a released agent over many 18

years (in some cases for up to 30 years) using correlative studies or by comparing before and 19

after-release ecological data is recommended: 1) because the agent may be slow to build up20

populations and cause detectable changes in weed populations and associated ecosystems, 2) 21

because short-term responses of weed populations to the agent are not always consistent with 22

long-term trends, 3) to account for possible confounding factors, and 4) to gather strong 23

evidence that the agent is responsible for the observed changes in the weed population across 24

its range (Hoffmann et al., 1998; McFadyen, 1998; Blossey, 1999). Monitoring programs 25

34

need to be designed around worst-case scenarios, because released agents may not necessarily 1

live up to efficacy predictions obtained from laboratory and glasshouse experiments and/or 2

field studies in the native range. Scarcity of resources however, often means sites are only re-3

assessed every two to three years (in particular at release sites) and up to 10 year intervals to 4

capture long-term effects (McFadyen, 1998). Streamlined approaches, such as simplification 5

of monitoring protocols and sampling regimes, can be used to minimize costs and boost 6

participation by community groups and state and local government officers, which in turn 7

increase the number of sites monitored at a regional scale (Briese and McLaren, 1997; 8

Swirepik and Smyth, 2003). Nonetheless, in instances where the effects of the agent are not 9

massive, over-simplified protocols may generate poor data, which may not be as informative 10

as fewer but higher quality data. Infrequent monitoring may also neglect to provide sufficient 11

information to disentangle the effects of variability in abiotic influences from the effects of 12

the agent on the weed. 13

14

While the detection of large effect sizes does not need elaborate sampling schemes, unless the 15

agent is only effective episodically, the collection of data with a higher level of resolution is 16

required to identify more subtle agent effects. In these instances, guidance on determining the 17

necessary number of sampling sites and the frequency and duration of sampling can be 18

obtained from long-term ecological data on the weed population dynamics and productivity. 19

Power analysis can be used to estimate the sampling effort necessary to detect a significant 20

effect of the agents (Cohen, 1988; Lenth, 2001; Quinn and Keough, 2002; but see Field et al., 21

2007). Such an analysis requires an estimate of the variance in the study population (usually 22

obtained from the literature or a pilot study) and the level of effect that is biologically 23

important as well as a set significance level (usually α=0.05) (Quinn and Keough, 2002). 24

Variance component analysis (Cox and Solomon, 2003) can be used to reveal the relative 25

35

contributions of spatial and temporal variability to the overall variance in such data (Buckley 1

et al., 2003b), so that sampling effort can be weighted towards the greater source of variance. 2

3

Unfortunately, long-term empirical data on the population dynamics of many weeds targeted 4

by biological control are scarce. In such cases, long-term regional temperature and rainfall 5

records may provide useful surrogates for agent population dynamics and plant performance 6

(Andrewartha and Birch, 1954; Birch, 1957; Chippendale, 1995; Grace, 1997). As outlined7

above, variance component analysis can then be conducted to compare the extent to which the 8

climatic parameters vary between years, and between sites within years.9

10

The inclusion of covariates, such as climatic parameters or initial plant size, into statistical 11

models can often reduce the unexplained variance and reveal agent effects that would 12

otherwise have gone undetected (Bowers and Stamp, 1993; Goolsby et al., 2000). 13

Additionally, incorporating information about agent abundance into the statistical model is 14

likely to account for at least some of the variability in weed target performance, and should 15

therefore assist in elucidating agent effects (e.g. Cochran 1957). 16

17

When to stop monitoring18

When evaluating the long-term effectiveness of biological control agents, it may be difficult 19

to know when it is appropriate to stop monitoring; particularly if a statistical analysis reveals 20

that no significant effect from the agents has been detected after years of data collection. This 21

result may be due to ineffectiveness of the agents or it might merely be a consequence of 22

erratic fluctuations in environmental conditions. In circumstances when the latter is suspected, 23

it is often tempting to continue monitoring. Although such a reaction may be understandable, 24

it is not scientifically legitimate because it violates critical assumptions that underpin the logic 25

36

of inferential statistical tests (Armitage et al., 1969; Jennison and Turnbull, 1990). Extra data 1

cannot be collected, and a supplanted dataset re-analyzed, without increasing the probability 2

of drawing the wrong conclusion. In particular, this practice may increase the chance of 3

incorrectly detecting an ‘effect’ that is not in fact present (Armitage et al., 1969). In a manner 4

akin to the multiple post hoc comparison problems (see Quinn and Keough, 2002), the chance 5

of this occurring grows rapidly for every occasion that a dataset is supplanted and re-6

analyzed.7

8

Medical scientists have developed procedures called group sequential methods for dealing 9

with similar problems in clinical trials (Jennison and Turnbull, 1990). Although the details of 10

these methods are beyond the scope of this paper, the central concept is straightforward.11

Trials are planned to be sufficiently well-replicated to detect a treatment effect of a pre-12

determined magnitude, which in the context of biological control would be the smallest agent 13

effect size that researchers want to be able to observe (Pocock, 1977; Jennison and Turnbull, 14

1989, 1990). Then one or more 'interim' significance tests are planned to be conducted at15

some point during the course of the trial. These interim significance tests are carried out using 16

a more stringent level, one that has been set at a level that will maintain the desired overall 17

level across all of the planned tests to be conducted during the trial. Should one of the 18

interim tests produce a significant result at the more stringent , the study can be terminated 19

as the treatment has been shown to be effective. For example, take the hypothetical case of a 20

biological control program where researchers are seeking 20% or more reduction in the 21

biomass of the target weed. A monitoring program is designed to collect quarterly data over a 22

three-year period, with a sample size across the period sufficient to detect a 20% effect size at 23

= 0.05. Researchers decide to analyze their accumulated data at the end of each of the three 24

years. Consulting Table 1 in Pocock (1977), they set an interim significance level of = 25

37

0.0221, which permits three analyses of their data while maintaining an overall = 0.05. The 1

agent is released, and monitoring commences. At the end of the first year, analysis of the data 2

indicates an agent effect with a probability of P = 0.0899. Consequently monitoring continues 3

into the second year. At the end of the second year, analysis of the accumulated data (i.e. data 4

from both year 1 and year 2) indicates an agent effect with a probability of P = 0.0197, which 5

is less than the of 0.0221. Hence the agent has been shown to reduce the biomass of the 6

target weed, and monitoring can cease: it is not necessary to continue monitoring into the 7

third year as originally planned. Further examination of the data reveals that while variability 8

of the data was similar to that anticipated when the sampling regime was designed, the agent 9

was reducing target biomass by considerably more than the 20% that the monitoring program 10

was designed to detect.11

12

More elaborate adaptive group sequential tests allow researchers to modify their sampling 13

design mid-trial should interim analyses indicate additional sampling would be highly 14

desirable (Bauer and Köhne, 1994; Jennison and Turnbull, 2006). These methods are very 15

complex, and best left to statisticians trained in their use. Nevertheless, they may provide the 16

only feasible approach to evaluate agent effectiveness in circumstances where it is particularly 17

difficult to estimate the appropriate sampling effort during the design stage of the monitoring 18

program.19

20

10. Conclusion21

22

Evaluating the effectiveness of agents should be an integral component of any biological 23

control program; from the early agent selection phase to post-release long-term monitoring 24

activities (Fig. 1). This paper has provided an overview of key issues and methodologies 25

38

available to predict efficacy of candidate agents before release and to measure their 1

effectiveness once released and established in the introduced range. We emphasized the need 2

to measure the effectiveness of released agents not only on individual plants, but also on weed 3

populations and to evaluate the response of associated vegetation to a reduction in the weed in 4

order to gather the information necessary to thoroughly assess whether a biological control 5

program has been a success or failure. A successful program may also contribute to the stable 6

recovery of ecosystem processes, but it has so far been generally outside the scope of 7

biological control programs to perform such detailed assessments. 8

9

Evaluating the effectiveness of candidate biological control agents before their release may 10

increase the likelihood of selecting agents that will suppress populations of the target weed in 11

the introduced range. Admittedly, additional time and resources are required to undertake 12

such assessments in the field in the native range or under laboratory or glasshouse conditions 13

and there is no guarantee that predictions of agent efficacy will be correct once it is released 14

in the introduced range. Nonetheless, efficacy assessment is a sound initial investment to 15

make so that candidate agents with the highest chance of adversely affecting the weed 16

undergo host-specificity testing, a costly phase of biological control programs. Furthermore, a 17

reduction in the establishment of ineffective agents will reduce the risk of unpredicted off-18

target effects and help sustain, and potentially increase, the profile of weed biological control 19

as an efficient management tool for troublesome invasive plants. 20

21

Sufficient resources and funding should be negotiated and set aside at the start of biological 22

control programs to support activities to evaluate agent effectiveness soon after their release 23

and establishment in the field. Additional funding should be accessible later to support long-24

term monitoring. This is unfortunately not often achieved, and practitioners have to come up 25

39

with creative schemes to obtain support to undertake long-term post-release evaluation. A 1

range of methodologies, from simple and cheap to complex and expensive is available to 2

choose from depending on resources available. While plans for short-term monitoring of 3

agent effectiveness are more likely to obtain financial support because of the lower level of 4

commitment required, it is the longer-term evaluation activities that will deliver reliable data 5

to gauge if a program has been successful at reducing weed populations sufficiently to enable 6

recovery of associated ecosystems. Such ecological data can then be fed into a benefit-cost 7

economic analysis; a powerful means to illustrate return on investment for a biological control 8

program and to make the case for sustained investment in the science and application of 9

biological control. Qualitative data on environmental and social benefits could also be 10

highlighted to stakeholders.11

12

Although excuses (valid or otherwise) will likely continue to be made for the limited 13

investment of time and resources in quantifying effectiveness of biological control agents, 14

practitioners need to understand and widely promote the importance of this component of 15

programs to influence stakeholders and funding bodies. This is vital to reliably demonstrate 16

the utility of biological control as a valuable tool in weed management.17

18

Acknowledgements19

20

We thank the Cooperative Research Centre (CRC) for Australian Weed Management and the 21

respective organisation of each co-author for their support during the preparation of this 22

paper. We would like to thank Royce Holtkamp (NSW Department of Primary Industries) and 23

Michael Day (Queensland Department of Primary Industries and Fisheries) for their 24

assistance in organizing the workshop on Impact Evaluation of Biological Control in 25

40

Brisbane, March 2006, where some of the information in this paper was discussed. We are 1

also grateful to Dr Rachel McFadyen (CRC for Australian Weed Management) and Dr Andy 2

Sheppard (CSIRO Entomology) as well as two anonymous reviewers for their valuable 3

comments on a draft of the manuscript. 4

5

References 6

7

Adair, R.J., Holtkamp, R.H., 1999. Development of a pesticide exclusion technique for 8

assessing the impact of biological control agents for Chrysanthemoides monilifera. 9

Biocontrol Science and Technology 9, 383-390.10

Anderson, G.L., Carruthers, R.I., Ge, S., Gong, P., 2005. Monitoring of invasive Tamarix 11

distribution and effects of biological control with airborne hyperspectral remote 12

sensing. International Journal of Remote Sensing 26, 2487-2489.13

Andrewartha, H., Birch, L., 1954. The Distribution and Abundance of Animals. The 14

University of Chicago Press, Chicago. 15

Armitage, P., McPherson, C., Rowe, B., 1969. Repeated significance tests on accumulating 16

data. Journal of the Royal Statistical Society: Series A 132, 235-244. 17

Auld, B.A., Menz, K.M., Tisdell, C.A., 1987. Weed Control Economics. Academic Press Inc., 18

London.19

Baars, J.R., Hill, M.P., Heystek, F., Neser, S., Urban, A.J., 2007. Biology, oviposition 20

preference and impact in quarantine of the petiole-galling weevil, Coelocephalapion 21

camarae Kissinger, a promising candidate agent for biological control of Lantana 22

camara. Biological Control 40, 187-195.23

41

Balciunas, J.K., Smith, L., 2006. Pre-release efficacy assessment, in quarantine, of a tephritid 1

gall fly being considered as a biological control agent for Cape-ivy (Delairea 2

odorata). Biological Control 39, 516-524.3

Barton, J., 2004. How good are we at predicting the field host-range of fungal pathogens used 4

for classical biological control of weeds? Biological Control 31, 99-122.5

Barton, J., Fowler, S.V., Gianotti, A.F., Winks, C.J., de Beurs, M., Arnold, G.C., Forrester, 6

G., 2007. Successful biological control of mist flower (Ageratina riparia) in New 7

Zealand: Agent establishment, impact and benefits to the native flora. Biological 8

Control 40, 370-385.9

Bass, D.J., Delpech, V., Beard, J., Bass, P., Walls, R.S., 2000. Ragweed in Australia. 10

Aerobiologia 16, 107-111.11

Baudoin, A.B.A.M., Abad, R.G., Kok, L.T., Bruckart, W.L., 1993. Field evaluation of 12

Puccinia carduorum for biological control of musk thistle. Biological Control 3, 53-13

60.14

Bauer, P., Köhne, K., 1994. Evaluation of experiments with adaptive interim analyses. 15

Biometrics 50, 1029-1041. 16

Birch, L., 1957. The role of weather in determining the distribution and abundance of 17

animals. Cold Spring Harbor Symposia on Quantitative Biology 22, 203-215. 18

Blanchard, R., Holmes, P.M., 2008. Riparian vegetation recovery after invasive alien tree 19

clearance in the fynbos biome. South African Journal of Botany 74, 421-431. 20

Blossey, B., 1995. A comparison of various approaches for evaluating potential biological 21

control agents using insects on Lythrum salicaria. Biological Control 5, 113-122.22

Blossey, B., 1999. Before, during and after: the need for long-term monitoring in invasive 23

plant species management. Biological Invasions 1, 301-311.24

42

Blossey, B., Skinner, L., 2000. Design and importance of post-release monitoring. In: 1

Spencer, N.R. (Ed.), Proceedings of the X International Symposium on Biological 2

Control of Weeds. Montana State University, Bozeman, pp. 693-706.3

Broughton, A.J., Pemberton, R.W., 2008. Efforts to establish a foliage-feeding moth, 4

Austromusotima camptozonale, against Lygodium microphyllum in Florida, considered 5

in the light of a retrospective review of establishment success of weed biocontrol 6

agents belonging to different arthropod taxa. Biological Control 47, 28-36.7

Bowers, M.D. and Stamp, N.E., 1993. Effects of plant age, genotype and herbivory on 8

Plantago performance and chemistry. Ecology 74, 1778-1791.9

Briese, D.T., 2000. Impact of the Onopordum capitulum weevil Larinus latus on seed 10

production by its host-plant. Journal of Applied Ecology 37, 238-246.11

Briese, D.T., 2005. Translating host-specificity test results into the real world: The need to 12

harmonize the yin and yang of current testing procedures. Biological Control 35, 208-13

214.14

Briese, D.T., 2006. Can an a priori strategy be developed for biological control? The case of 15

Onopordum spp. thistles in Australia. Australian Journal of Entomology 45, 317-323.16

Briese, D.T., Bottcher, S., Vitou, J., Thomann, T., 2003. Prevalence and impact of the crown 17

fly, Botanophila spinosa, on its host thistle, Onopordum acanthium, in southern 18

France. Acta Oecologia 24, 77-86.19

Briese, D.T., McLaren, D.A., 1997. Community involvement in the distribution and 20

evaluation of biological control agents: Landcare and similar groups in Australia. 21

Biocontrol News and Information 18, 39N-49N.22

Briese, D.T., Pettit, W.J., Walker, A., 2004. Evaluation of the biological control agent, Lixus 23

cardui, on Onopordum thistles: experimental studies on agent demography and 24

impact. Biological Control 31, 165-171.25

43

Briese, D.T., Thomann, T., Vitou, J., 2002. Impact of the rosette crown weevil 1

Trichosirocalus briesei on the growth and reproduction of Onopordum thistles. 2

Journal of Applied Ecology 39, 688-698.3

Brun, L.A., Sheppard, A.W., Delmotte, E., Hasan, S., Jourdan, M., 1995. Development of 4

cercospora blight epidemics and effect on the summer annual Heliotropium 5

europaeum in the field. Annals of Applied Biology 127, 137-150.6

Buckley, Y.M., 2008. The role of research for integrated management of invasive species, 7

invaded landscapes and communities. Journal of Applied Ecology 45, 397-402.8

Buckley, Y.M., Bolker, B.M., Rees, M., 2007. Disturbance, invasion and re-invasion: 9

managing the weed-shaped hole in disturbed ecosystems. Ecology Letters 10, 809-10

817.11

Buckley, Y.M., Briese, D.T., Rees, M., 2003a. Demography and management of the invasive 12

plant species Hypericum perforatum. II. Construction and use of an individual-based 13

model to predict population dynamics and the effects of management strategies. 14

Journal of Applied Ecology 40, 494-507.15

Buckley, Y.M., Briese, D.T., Rees, M., 2003b. Demography and management of the invasive 16

plant species Hypericum perforatum. I. Using multi-level mixed-effects models for 17

characterizing growth, survival and fecundity in a long-term data set. Journal of 18

Applied Ecology 40, 481-493.19

Buckley, Y.M., Brockerhoff, E., Langer, L., Ledgard, N., North, H., Rees, M., 2005a. 20

Slowing down a pine invasion despite uncertainty in demography and dispersal. 21

Journal of Applied Ecology 42, 1020-1030.22

Buckley, Y.M., Hinz, H.L., Matthies, D., Rees, M., 2001. Interactions between density-23

dependent processes, population dynamics and control of an invasive plant species, 24

Tripleurospermum perforatum (scentless chamomile). Ecology Letters 4, 551-558.25

44

Buckley, Y.M., Rees, M., Paynter, Q., Lonsdale, M., 2004. Modelling integrated weed 1

management of an invasive shrub in tropical Australia. Journal of Applied Ecology 41, 2

547-560.3

Buckley, Y.M., Rees, M., Sheppard, A.W., Smyth, M.J., 2005b. Stable coexistence of an 4

invasive plant and biocontrol agent: a parameterized coupled plant-herbivore model. 5

Journal of Applied Ecology 42, 70-79.6

Burdon, J.J., Thrall, P.H., Groves, R.H., Chaboudez, P., 2000. Biological control of Carduus 7

pycnocephalus and C. tenuiflorus using the rust fungus Puccinia cardui-pycnocephali. 8

Plant Protection Quarterly 15, 14-17.9

Callaway, R.M., DeLuca, T.H., Belliveau, W.M., 1999. Biological control herbivores may 10

increase competitive ability of the noxious weed Centaurea maculosa Ecology 80, 11

1196-1201.12

Carson, W.P., Hovick, S.M., Baumert, A.J., Bunker, D.E., Pendergast, T.H., 2008. Evaluating 13

the post-release efficacy of invasive plant biocontrol by insects: a comprehensive 14

approach. Arthropod-Plant Interactions 2, 77-86.15

Chippendale, J.F., 1995. The biological control of Noogoora burr (Xanthium occidentale) in 16

Queensland: an economic perspective. In: Delfosse, E.S., Scott, R.R., (Eds.), 17

Proceedings of the VIII International Symposium on Biological Control of Weeds. 18

CSIRO, Melbourne, pp. 185-192.19

Clark, S.E., Van Driesche, R.G., Sturdevant, N., Elkinton, J., Buonaccorsi, J.P., 2001. Effects 20

of site characteristics and release history on establishment of Agapeta zoegana21

(Lepidoptera: Cochylidae) and Cyphocleonus achates (Coleoptera: Curculionidae), 22

root-feeding herbivores of spotter knapweed, Centaurea maculosa. Biological Control 23

22, 122-130.24

45

Cochran, W.G., 1957. Analysis of Covariance: its Nature and its Uses. Biometrics 13, 261-1

281.2

Cohen, J., 1988. Statistical Power Analysis for the Behavioral Sciences, 2nd Edition. 3

Academic Press Inc., New York.4

Coombs, E.M., Radtke, H., Isaacson D.L., Snyder, S.P., 1996. Economic and regional 5

benefits from the biological control of tansy ragwort, Senecio jacobaea in Oregon. In: 6

Moran V.C., Hoffmann, J.H., (Eds.), Proceedings of the IX International Symposium 7

on Biological Control of Weeds. University of Cape Town, South Africa, pp. 489-494.8

Corn, J.G., Story, J.M., White, L.J., 2006. Impacts of the biological control agent 9

Cyphocleonus achates on spotted knapweed, Centaurea maculosa, in experimental 10

plots. Biological Control 37, 75-81.11

Cox, D.R., Solomon, P.J., 2003. Components of Variance. Chapman & Hall/CRC, Boca 12

Raton, Florida. 13

Cruz, Z.T., Muniappan, R., Reddy, G.V.P., 2006. Establishment of Cecidochares connexa14

(Diptera: Tephritidae) in Guam and its effect on the growth of Chromolaena odorata15

(Asteraceae). Annals of the Entomological Society of America 99, 845-850.16

Cullen, J.M., Kable, P.F., Catt, M., 1973. Epidemic spread of a rust imported for biological 17

control. Nature 244, 462-464.18

Culliney, T.W., 2005. Benefits of classical biological control for managing invasive 19

plants. Critical Reviews in Plant Sciences 24, 131-150.20

Curruthers, R., Anderson, G., DeLoach, J., Knight, J., Ge, S., Gong, P., 2006. Remote 21

sensing of saltcedar biological control effectiveness. In: Aguirre-Bravo, C., Pellicane, 22

P.J., Burns, D.P., Draggan, S., (Eds.), Monitoring Science and Technology 23

Symposium: Unifying Knowledge for Sustainability in the Western Hemisphere 24

46

Proceedings. U.S. Department of Agriculture, Forest Service, Fort Collins, Colorado, 1

pp. 50-56.2

Davis, A.S., Landis, D.A., Nuzzo, V., Blossey, B., Gerber, E., Hinz., H.L., 2006. 3

Demographic models inform selection of biocontrol agents for garlic mustard (Alliaria 4

petiolata). Ecological Applications 16, 2399-2410.5

Delfosse, E.S., 1989. Echium in Australia: the conflict continues. In: Delfosse, E.S., (Ed.), 6

Proceedings of the VII International Symposium on Biological Control of Weeds. 7

CSIRO, Melbourne, p. 117.8

Denoth, M., Myers, J.H., 2005. Variable success of biological control of Lythrum salicaria in 9

British Columbia. Biological Control 32, 269-279.10

Denslow, J.S., D’Antonio, C.M., 2005. After biocontrol: Assessing indirect effects of insect 11

releases. Biological Control 35, 307-318.12

Dhileepan, K. 2003a. Evaluating the effectiveness of weed biocontrol at the local scale. In: 13

Spafford Jacob, H., Briese, D.T., (Eds.), Improving the Selection, Testing and 14

Evaluation of Weed Biological Control Agents. CRC for Australian Weed 15

Management Technical Series No. 7, Glen Osmond, Australia, pp. 51-60. Retrieved 16

29 April 2009, from http://www.weedscrc.org.au/publications/technical_series.html17

Dhileepan, K., 2003b. Seasonal variation in the effectiveness of the leaf-feeding beetle 18

Zygogramma bicolorata (Coleoptera: Chrysomelidae) and stem-galling moth 19

Epiblema strenuana (Lepidoptera: Tortricidae) as biocontrol agents on the weed 20

Parthenium hysterophorus (Asteraceae). Bulletin of Entomological Research 93, 21

393-401.22

Dhileepan, K., 2003c. Current status of the stem-boring weevil Listronotus setosipennis23

(Coleoptera: Curculionidae) introduced against the weed Parthenium hysterophorus 24

(Asteraceae) in Australia. Biocontrol Science and Technology 13, 3-12.25

47

Dhileepan, K., 2007. Biological control of parthenium (Parthenium hysterophorus) in 1

Australian rangelands translates to improved pasture production. Weed Science 55, 2

497-501.3

Dhileepan, K., Florentine, S.K., Lockett, C.J., 2006. Establishment, initial impact and 4

persistence of parthenium summer rust Puccinia melampodii in north Queensland. In: 5

Preston, C., Watts, J.H., Crossman, N.D. (Eds.), Proceedings of the 15th Australian 6

Weeds Conference. Weed Management Society of South Australia, Adelaide, pp. 577-7

580.8

Dhileepan, K., Lockett, C.J., Robinson, M., Pukallus, K., 2009. Prioritising potential guild of 9

specialist herbivores as biological control agents for prickly acacia through simulated 10

herbivory. Annals of Applied Biology 154, 97-105.11

Dhileepan, K., McFadyen, R.E., 2001. Effects of gall damage by the introduced biocontrol 12

agent Epiblema strenuana (Lepidoptera: Tortricidae) on the weed Parthenium 13

hysterophorus (Asteraceae). Journal of Applied Entomology 125, 1-8.14

Dhileepan, K., Setter, S., McFadyen, R.E., 2000a. Impact of defoliation by the biocontrol 15

agent Zygogramma bicolorata on the weed Parthenium hysterophorus in Australia. 16

BioControl 45, 501-512.17

Dhileepan, K., Setter, S., McFadyen, R.E., 2000b. Response of the weed Parthenium 18

hysterophorus (Asteraceae) to defoliation by the introduced biocontrol agent 19

Zygogramma bicolorata (Coleoptera: Chrysomelidae). Biological Control 19, 9-16.20

Edwards, P.B., Holtkamp, R.H., Adair, R.J., 1999. Establishment and rapid spread of the 21

bitou seed fly Mesoclanis polana in eastern Australia. Australian Journal of 22

Entomology 38, 148-150.23

48

Eiswerth, M.E., Darden, T.D., Johnson, W.S., Agapoff, J., Harris, T.R., 2005. Input-output 1

modelling, outdoor recreation, and the economic impacts of weeds. Weed Science 53, 2

130-137.3

Epstein, P., 1998. Weeds bring disease to the East African waterways. Lancet 351, 577.4

Erskine Ogden, J.A., Rejmánek, M., 2005. Recovery of native plant communities after the 5

control of a dominant invasive plant species, Foeniculum vulgare: Implications for 6

management. Biological Conservation 125, 427-439.7

Evans, H.C., 2000. Evaluating plant pathogens for biological control of weeds: an alternative 8

view of pest risk assessment. Australasian Plant Pathology 29, 1-14.9

Evans, J.A., Landis, D.A., 2007. Pre-release monitoring of Alliaria petiolata (garlic mustard) 10

invasions and the impacts of extant natural enemies in southern Michigan forests. 11

Biological control 42, 300-307.12

Evans, K.J., Bruzzese, E., 2003. Life history of Phragmidium violaceum in relation to its 13

effectiveness as a biological control agent of European blackberry. Australasian Plant 14

Pathology 32, 231-239.15

Everitt, J.H., Anderson, G.L., Escobar, D.E., Davis, M.R., Spencer, N.R., Andrascik, R.J., 16

1995. Use of remote sensing for detecting and mapping leafy spurge (Euphorbia 17

esula). Weed Technology 9, 599-609.18

Everitt, J.H., Flores, D., Davis, M., 2005. Using remote sensing to assess biological control of 19

giant salvinia. Journal of Aquatic Plant Management 43, 76-80.20

Ewel, J.J., O’Dowd, D.J., Bergelson, J., Daehler, C.C., D’Antonio, C.M., Gomez, L.D., 21

Gordon, D.R., Hobbs, R.J., Holt, A., Hopper, K.R., Hughes, C.E., LaHart, M., Leakey, 22

R.R.B., Lee, W.G., Loope, L.L., Lorence, D.H., Louda, S.M., Lugo, A.E., McEvoy, 23

P.B., Richardson, D.M., Vitousek, P.M., 1999. Deliberate introductions of species: 24

Research needs - Benefits can be reaped, but risks are high. Bioscience 49, 619-630.25

49

Farrell, G., Lonsdale, M., 1997. Measuring the impact of biological control agents on weeds. 1

In: Julien, M., White, G. (Eds.), Biological Control of Weeds: Theory and Practical 2

Applications. Australian Centre for International Agriculture Research, Canberra, pp. 3

105-118.4

Field, S.A., O’Connor, P.J., Tyre, A.J., Possingham, H.P., 2007. Making monitoring 5

meaningful. Austral Ecology 32, 485-491.6

Freeman, D.B., 1992. Prickly pear menace in eastern Australia 1880-1940. Geographical 7

Review 82, 413-429.8

French, K., Buckley, S., 2008. The effects of the herbicide metsulfuron-methyl on litter 9

invertebrate communities in a coastal dune invaded by Chrysanthemoides monilifera10

ssp. rotundata. Weed Research 48, 266-272.11

Galatowitsch, S.M., Richardson, D.M., 2005. Riparian scrub recovery after clearing of 12

invasive alien trees in headwater streams of the Western Cape, South Africa. Biological 13

Conservation 122, 509-521.14

Gassmann, A., 1996. Classical biological control of weeds with insects: a case for 15

emphasizing agent demography. In: Moran V.C., Hoffmann, J.H., (Eds.), Proceedings 16

of the IX International Symposium on Biological Control of Weeds. University of 17

Cape Town, South Africa, pp. 171-176.18

Gerber, E. Hinz, H.L., Blossey, B., 2007a. Interaction of specialist root and shoot herbivores 19

of Alliaria petiolata and their impact on plant performance and reproduction. 20

Ecological Entomology 32, 357-365.21

Gerber, E. Hinz, H.L., Blossey, B., 2007b. Impact of the belowground herbivore and potential 22

biological control agent, Ceutorhynchus scrobicollis, on Alliaria petiolata23

performance. Biological Control 42, 355-364.24

50

Gerber, E., Hinz, H. L., Blossey, B., 2008. Pre-release impact assessment of two stem-boring 1

weevils proposed as biological control agents for Alliaria petiolata. Biological Control2

45, 360-367.3

Goolsby, J.A., Makinson, J., Purcell, M., 2000. Seasonal phenology of the gall-making fly 4

Fergusonina sp. (Diptera: Fergusoninidae) and its implications for biological control 5

of Melaleuca quinquenervia. Australian Journal of Entomology 39, 336-343.6

Goolsby, J.A., Zonneveld, R., Bourne, A. 2004. Pre-release assessment of impact on biomass 7

production of an invasive weed, Lygodium microphyllum (Lygodiaceae: Pteridophyta), 8

by a potential biological control agent, Floracarus perrepae (Acariformes: 9

Eriophyidae). Environmental Entomology 33, 997-1002.10

Gosper, C.R., Whelan, R.J., French, K. 2006. The effect of invasive plant management on the 11

rate of removal of vertebrate-dispersed fruits. Plant Ecology 184, 351-363.12

Grace, J., 1997. Plant water relations. In: Crawley, M.J. (Ed.), Plant Ecology. Blackwell, 13

London, pp. 28-50. 14

Greer, G., 1995. Economics and the biological control of weeds. In: Delfosse, E.S., Scott, 15

R.R., (Eds.), Proceedings of the VIII International Symposium on Biological Control 16

of Weeds. CSIRO, Melbourne, pp. 177-184.17

Grevstad, F.S., 2006. Ten-year impacts of the biological control agents Galerucella pusilla 18

and G. calmariensis (Coleoptera: Chrysomelidae) on purple loosestrife (Lythrum 19

salicaria) in Central New York State. Biological Control 39, 1-8. 20

Grice, A.C., 2004. Weeds and the monitoring of biodiversity in Australian rangelands. Austral 21

Ecology 29, 51-58.22

Groves, R.H., Williams, J.D., 1975. Growth of skeleton weed (Chondrilla juncea L.) as 23

affected by growth of subterranean clover (Trifolium subterraneum L.) and infection 24

51

by Puccinia chondrilla Bubak and Syd. Australian Journal of Agricultural Research 1

26, 975-983.2

Hafliger, P., Schwarzlaender, M., Blossey, B., 2005. Biology of Platycephala planifrons3

(Diptera:Chloropidae) and its potential effectiveness as biological control agent for 4

invasive Phragmites australis in North America. Biological Control 34, 302-311.5

Hall, F.S., 2001. Ground-based photographic monitoring. US Dept of Agriculture, General 6

Technical Report PNW-GTR-503. Retrieved 29 April 2009, from 7

http://www.fs.fed.us/pnw/pubs/gtr503/8

Hand, L.F., Keaster, A.J., 1967. The environment of an insect field cage. Journal of Economic 9

Entomology 60, 910-915.10

Harms, R.S., Hiebert, R.D., 2006. Vegetation response following invasive Tamarisk (Tamarix11

spp.) removal and implications for riparian restoration. Restoration Ecology 14, 461-12

472.13

Hasan, S., Aracil, E., 1991. Biology and effectiveness of Uromyces heliotropii, a potential 14

biological control agent for Heliotropium europaeum. New Phytologist 118, 559-563.15

Heard, T.A., Winterton, S.L., 2000.Interactions between nutrient status and weevil herbivory 16

in the biological control of water hyacinth. Journal of Applied Ecology 37, 117-127.17

Henderson, L., 1999. The Southern African Plant Invaders Atlas (SAPIA) and its contribution 18

to biological weed control. African Entomology Memoir 1, 159-163.19

Higgins, S.I., Azorin, E.J., Cowling, R.M., Morris, M.J., 1997. A dynamic ecological-20

economic model as a tool for conflict resolution in an invasive-alien-plant, biological 21

control and native-plant scenario. Ecological Economics 22, 141-154.22

Hjalten, J., 2004. Simulating herbivory: problems and possibilities. In: Weisser, W.W., 23

Siemann, E. (Eds.), Insects and Ecosystem Function. Springer-Verlag, Berlin, pp. 243-24

256.25

52

Hobbs, R.J., 2003. Ecological managment and restoration: assessment, setting goals and 1

measuring success. Ecological Management and Restoration 4(S1), 2-3.2

Hoffmann, J.H., Moran, V.C., 1991. Biocontrol of a perennial legume, Sesbania punicea3

using a florivorous weevil, Trichapion lativentre: weed population dynamics with a 4

scarcity of seeds. Oecologia 88, 574-576.5

Hoffmann, J.H., Moran, V.C., 1998. The population dynamics of an introduced tree Sesbania 6

punicea in South Africa in response to long-term damage caused by different 7

combinations of three species of biological control agents. Oecologia 114, 343-348.8

Hoffmann J.H., Moran V.C., Zeller D.A., 1998. Long-term population studies and the 9

development of an integrated management programme for control of Opuntia stricta 10

in Kruger National Park, South Africa. Journal of Applied Ecology 35 156-160.11

Holt, R.D., Hochberg M.E., 2001. Indirect interactions, community modules, and biological 12

control: a theoretical perspective. In: Waijnberg, E., Scott, J.K. amd Quimby, P.C. 13

(Eds.), Evaluation of Indirect Ecological Effects of Biological Control. CABI, 14

Wallingford, UK, pp. 13-37.15

Huffaker, C.B., Kennett, C.E., 1959. A ten-year study of vegetation change associated with 16

biological control of Klamath weed. Journal of Range Management 12, 69-82. 17

Hulme, P.E., 2006. Beyond control: wider implications for the management of biological 18

invasions. Journal of Applied Ecology 43, 835-847.19

Hulme, P.E., Bremner, E.T., 2006. Assessing the impact of Impatiens glandulifera on riparian 20

habitats: partitioning diversity components following species removal. Journal of 21

Applied Ecology 43, 43-50.22

Hunt-Joshi, T.R., Blossey, B., Root, R.B., 2004. Root and leaf herbivory on Lythrum 23

salicaria: Implications for plant performance and communities. Ecological 24

Applications 14, 1574-1589.25

53

IAC (Industries Assistance Commission), 1985. Biological control of Echium species 1

(including Paterson’s curse / Salvation Jane). Report No. 371, Australian Government 2

Publishing Service, Canberra.3

Ireson, J.E., Davies, J.T., Friend, D.A., Holloway, R.J., Chatterton, W.S., Van Putten, E.I., 4

McFadyen, R.E.C., 2007. Weeds of Pastures and Field Crops in Tasmania: Economic 5

Impacts and Biological Control. CRC for Australian Weed Management Technical 6

Series No. 13, Glen Osmond, Australia. Retrieved 29 April 2009, from 7

http://www.weedscrc.org.au/publications/technical_series.html8

James, W.C., 1971. An illustrated series of assessment keys for plant diseases, their 9

preparation and usage. Plant Disease Survey 51, 39-65. 10

James, R.R., McEvoy, P.B., Cox, C.S., 1992. Combining the cinnabar moth (Tyria jacobaeae) 11

and the ragwort flea beetle (Longitarsus jacobaeae) for control of ragwort (Senecio 12

jacobaea): an experimental analysis. Journal of Applied Ecology 29, 589-596.13

Jarvis, P.J., Fowler, S.V., Paynter, Q., Syrett, P., 2006. Predicting the economic benefits and 14

costs of introducing new biological control agents for Scotch broom Cytisus scoparius15

into New Zealand. Biological Control 39, 135-146.16

Jennison, C., Turnbull, B.W., 1989. Interim analyses: the repeated confidence interval 17

approach. Journal of the Royal Statistical Society: Series B 51, 305-361. 18

Jennison, C., Turnbull, B.W., 1990. Statistical approaches to interim monitoring of medical 19

trials: a review and commentary. Statistical Science 5, 299-317. 20

Jennison, C., Turnbull, B.W., 2006. Adaptive and non-adaptive group sequential tests. 21

Biometrika 93, 1-21. 22

Jetter, K., 2005. Economic framework for decision making in biological control. Biological 23

Control 35, 448-357.24

54

Johnston, W.R., Lloyd, P., 1982. Fight against cacti pests in Queensland. Queensland 1

Agricultural Journal 108, 215-221.2

Juliá, R., Holland, D.W., Guenthner, J., 2007. Assessing the economic impact of invasive 3

species: The case of yellow starthistle (Centaurea solsitialis L.) in the rangelands of 4

Idaho, USA. Journal of Environmental Management 85, 876-882.5

Julien, M.H., Broadbent, J.E., Matthews, N.C., 1979. Effects of Puccinia xanthii on Xanthium 6

strumarium [Compositae]. Entomophaga 24, 29-34.7

Kaiser, B.A., 2006. Economic impacts of non-indigenous species: Miconia and the Hawaiian 8

economy. Euphytica 148, 135-150.9

Khosla, S. N., Sobti, S. N., 1979. Parthenium - a national health hazard, its control and utility 10

- a review. Pesticides 13, 121-127.11

Kidd, N.A.C., Jervis, M.A. 1996. Population dynamics. In: Jervis M.A., Kidd, N.A.C. (Eds), 12

Insect Natural Enemies: Practical Approaches to their Study and Evaluation. Chapman 13

Hall, London, pp. 293-374.14

Kriticos, D.J., 2003. The roles of ecological models in evaluating weed biological control 15

agents and projects. In: Spafford Jacob, H., Briese, D.T., (Eds.), Improving the 16

Selection, Testing and Evaluation of Weed Biological Control Agents. CRC for 17

Australian Weed Management Technical Series No. 7, Glen Osmond, Australia, pp. 18

69-74. Retrieved 29 April 2009, from 19

http://www.weedscrc.org.au/publications/technical_series.html20

Kriticos, D., Brown, J., Radford, I., Nicholas, M., 1999. Plant population ecology and 21

biological control: Acacia nilotica as a case study. Biological Control 16, 230-239.22

Kriticos, D.J., Stuart, R.M., Ash, J.E., 2004. Exploring interactions between cultural and 23

biological control techniques: modelling bitou bush (Chrysantehmoides monilifera24

spp. rotundata) and a seed fly (Mesoclanis polana). In: Cullen, J.M., Briese, D.T., 25

55

Kriticos, D.J., Lonsdale, W.M., Morin, L., Scott, J.K. (Eds.), Proceedings of the XI 1

International Symposium on Biological Control of Weeds. CSIRO Entomology, 2

Canberra, pp. 559-566.3

Lass, L.W, Prather, T.S, Glenn, N.F, Weber, K.T, Mundt, J.T, Pettingill, J., 2005. A review of 4

remote sensing of invasive weeds and example of the early detection of spotted 5

knapweed (Centaurea maculosa) and babysbreath (Gypsophila paniculata) with a 6

hyperspectral sensor. Weed Science 53, 242-251.7

Lehtilä, K., Boalt, E., 2004. The use and usefulness of artificial herbivory in plant-herbivore 8

studies. In: Weisser, W.W., Siemann, E. (Eds.), Insects and Ecosystem Function. 9

Springer-Verlag, Berlin, pp. 257-275.10

Leistritz, F.L., Thompson, F., Leitch, J.A., 1992. Economic impact of leafy spurge in North 11

Dakota. Weed Science 40, 275-280.12

Lenth, R.V., 2001. Some practical guidelines for effective sample-size determination. The 13

American Statistician 55, 187-193.14

Lesica, P., Hanna, D., 2004. Indirect effects of biological control on plant diversity vary 15

across sites in Montana grasslands. Conservation Biology 18, 444-454. 16

Lonsdale, W.M., 1988. Litterfall in an Australian population of Mimosa pigra, an invasive 17

tropical shrub. Journal of Tropical Ecology 4, 381-392.18

Lonsdale, W.M., Farrell, G., 1998. Testing the effects on Mimosa pigra of a biological control 19

agent Neurostrota gunniella (Lepidoptera: Gracillaridae), plant competition and fungi 20

under field conditions. Biocontrol Science and Technology 8, 485-500.21

Lonsdale, W.M., Farrell, G., Wilson, C.G., 1995. Biological control of a tropical weed: A 22

population model and experiment for Sida acuta. Journal of Applied Ecology 32, 391-23

399.24

56

Louda, S.M., Pemberton, R.W., Johnson, M.T., Follett, P.A., 2003. Non-target effects – the 1

Achilles heel of biological control? Retrospective analyses to reduce risk associated 2

with biocontrol introductions. Annual Review of Entomology 48, 365-396.3

MacDougall, A.S., Turkington, R., 2005. Are invasive species the drivers or passengers of 4

change in degraded ecosystems? Ecology 86, 42-55.5

Maron, J.L., Vilà, M., Arnason, J., 2004. Loss of enemy resistance among introduced 6

populations of St. John's Wort (Hypericum perforatum). Ecology 85, 3243-3253. 7

McClay, A.S., 1995. Beyond before-and-after: experimental design and evaluation in classical 8

weed biological control. In: Delfosse, E.S., Scott R.R. (Eds.), Proceedings of the VIII 9

International Symposium on Biological Control of Weeds. CSIRO, Melbourne, pp. 10

213-219.11

McClay, A.S., Balciunas, J.K., 2005. The role of pre-release efficacy assessment in selecting 12

classical biological control agents for weeds – applying the Anna Karenina principle. 13

Biological Control 35, 197-207.14

McConnachie, A.J., Hill, M.P., Byrne, M.J., 2004. Field assessment of a frond-feeding 15

weevil, a successful biological control agent of red waterfern, Azolla filiculoides, in 16

southern Africa. Biological Control 28, 25-32.17

McConnachie, A.J., de Wit, M.P., Hill, M.P., Byrne, M.J., 2003. Economic evaluation of the 18

successful biological control of Azolla filiculoides in South Africa. Biological Control 19

29, 326-331.20

McEvoy, P., Coombs, E., 1999. Biological control of plant invaders: regional patterns, field 21

experiments, and structured population models. Ecological Applications 9, 387-401.22

McEvoy, P. B., Coombs, E. M., 2000. Why things bite back: unintended consequences of 23

biological weed control. In: Follett, P.A., Duan, J.J. (Eds.), Nontarget effects of 24

biological control. Kluwer Academic Publishers, Boston, USA, pp. 167-194.25

57

McEvoy, P., Cox, C., Coombs, E., 1991. Successful biological control of ragwort, Senecio 1

jacobae by introduced insects in Oregon. Ecological Applications 1, 430-442.2

McFadyen, R.E., 1979. Cactus mealy bug Hypogeococcus festerianua [Hem., 3

Pseudococcidae] an agent for the biological control of Eriocereus martini [Cactaceae] 4

in Australia. Entomophaga 24, 281-287.5

McFadyen, R.E., 1995. Parthenium weed and human health in Queensland. Australian Family 6

Physician 24, 1455-1458.7

McFadyen, R.E., 1998. Biological control of weeds. Annual Review of Entomology 43, 369-8

393.9

McFadyen, R., Spafford Jacob, H., 2003. Insects for the biocontrol of weeds: predicting 10

parasitism levels in the new country. In: Cullen, J.M., Briese, D.T., Kriticos, D.J., 11

Lonsdale, W.M., Morin, L., Scott, J.K. (Eds.), Proceedings of the XI International 12

Symposium on Biological Control of Weeds. CSIRO Entomology, Canberra, pp. 135-13

140.14

McFadyen, R.E., Tomley, A.J., 1981. Biological control of Harrisia cactus, Eriocereus 15

martinii, in Queensland by the mealybug Hypogeococcus festerianus. In: Delfosse, 16

E.S., (Ed.), Proceedings of the V International Symposium on Biological Control of 17

Weeds. CSIRO, Melbourne, pp. 589-594.18

Morin, L., Auld, B.A., Smith, H.E., 1996. Rust epidemics, climate and control of Xanthium 19

occidentale. In: Moran V.C., Hoffmann, J.H., (Eds.), Proceedings of the IX 20

International Symposium on Biological Control of Weeds. University of Cape Town, 21

South Africa, pp. 385-391.22

Morin, L., Evans, K.J., Sheppard, A.W., 2006. Selection of pathogen agents in weed 23

biological control: critical issues and peculiarities in relation to arthropod agents. 24

Australian Journal of Entomology 45, 348-364.25

58

Morin, L., Willis, A.J., Armstrong, J. and Kriticos, D., 2002. Spread, epidemic development 1

and impact of the bridal creeper rust in Australia: summary of results. In: Spafford 2

Jacob, H., Dodd, J., Moore, J.H. (Eds.), Proceedings of the 13th Australian Weeds 3

Conference. Plant Protection Society of WA, Perth, pp. 385-388.4

Müller-Schärer, H., Frantzen, J., 1996. An emerging system management approach for 5

biological weed control in crops: Senecio vulgaris as a research model. Weed 6

Research 36, 483-491.7

Müller-Schärer, H., Schaffner, U., 2008. Classical biological control: exploiting enemy 8

escape to manage plant invasions. Biological Invasions 10, 859-874.9

Navarro, L.A., Phiri, G., 2000. Water Hyacinth in Africa and the Middle East: A Survey of 10

Problems and Solutions. International Development Research Centre, Ottawa, Canada.11

Neubert, M.G., Parker, I.M., 2004. Projecting rates of spread for invasive species. Risk 12

Analysis 24, 817-831.13

Nordblom, T., 2003. Putting biological reality into economic assessments of biocontrol. In: 14

Spafford Jacob, H., Briese, D.T., (Eds.), Improving the Selection, Testing and 15

Evaluation of Weed Biological Control Agents. CRC for Australian Weed 16

Management Technical Series No. 7, Glen Osmond, Australia, pp. 75-85. Retrieved 17

29 April 2009, from http://www.weedscrc.org.au/publications/technical_series.html18

Nordblom, T., Smyth, M., Swirepik, A., Sheppard, A., Briese, D., 2002. Spatial economics of 19

biological control: investing in new releases of insects for earlier limitation of 20

Paterson’s curse in Australia. Agricultural Economics 27, 403-424.21

O’Donovan, J., Blackshaw, R.E., Harker, K.N., Clayton, G.W., Maurice, D.C., 2005. Field 22

evaluation of regression equations to estimate crop yield losses due to weeds. 23

Canadian Journal of Plant Science 85, 955-962.24

59

Page, A.R., Lacey, K.L., 2006. Economic Impact Assessment of Australian Weed Biological 1

Control. CRC for Australian Weed Management Technical Series No. 10, Glen 2

Osmond, Australia. Retrieved 29 April 2009, from 3

http://www.weedscrc.org.au/publications/technical_series.html4

Parker, I.M., 2000. Invasion dynamics of Cytisus scoparius: A matrix model approach. 5

Ecological Applications 10, 726-743.6

Paynter, Q., 2004. Evaluating Mimosa pigra biological control in Australia. In: Julien, M., 7

Flanagan, G., Heard, T., Hennecke, B., Paynter, Q., Wilson, C. (Eds.), Research and 8

Management of Mimosa pigra. CSIRO Entomology, Canberra, pp. 141-148. Retrieved 9

29 April 2009, from 10

http://www.ento.csiro.au/weeds/pdf/mimosa_symposium/23Paynter.pdf11

Paynter, Q., 2005. Evaluating the impact of a biological control agent Carmenta mimosa on 12

the woody wetland weed Mimosa pigra in Australia. Journal of Applied Ecology 42, 13

1054-1062.14

Paynter, Q., 2006. Evaluating the impact of biological control against Mimosa pigra in 15

Australia: Comparing litterfall before and after the introduction of biological control 16

agents. Biological Control 38, 166-173.17

Paynter, Q., Waipara, N., Peterson, P., Hona, S., Fowler, S., Gianotti, A., Wilkie, P., 2006. 18

The impact of two introduced biocontrol agents, Phytomyza vitalbae and Phoma 19

clematidina, on Clematis vitalba in New Zealand. Biological Control 36, 350-357.20

Pearson, D.E., Callaway, R.M., 2003. Indirect effects of host specific biological control 21

agents. Trends in Ecology and Evolution 18, 456-461.22

Pearson, D.E., Callaway, R.M., 2004. Response to Thomas et al.: biocontrol and indirect 23

effects. Trends in Ecology and Evolution 19, 62-63.24

60

Plummer, M.L., 2005. Impact of invasive water hyacinth (Eichhornia crassipes) on snail 1

hosts of schistosomiasis in Lake Victoria, East Africa. EcoHealth 2, 81-86.2

Pocock, S.J., 1977. Group sequential methods in the design and analysis of clinical trials. 3

Biometrika 64, 191-199. 4

Quinn, G., Keough, M.J., 2002. Experimental design and data analysis for biologists. 5

Cambridge University Press, Cambridge, UK, pp 164-166.6

Raghu, S., Dhileepan, K., 2005. The value of simulating herbivory in selecting effective 7

weed biological control agents. Biological Control 34, 265-273.8

Raghu, S., Dhileepan, K., Treviño, M., 2006a. Response of an invasive liana to simulated 9

herbivory: implications for its biological control. Acta Oecologica 29, 335-345.10

Raghu, S., Wilson, J.R., Dhileepan, K., 2006b. Refining the process of agent selection 11

through understanding plant demography and plant response to herbivory. 12

Australian Journal of Entomology 45, 308-316.13

Raghu, S., Dhileepan, K., Scanlan, J.C., 2007. Predicting risk and benefit a priori in 14

biological control of invasive plant species: A systems modelling approach. 15

Ecological Modelling 208, 247-262.16

Ramula, S., Knight, T.M., Burns, J.H., Buckley, Y.M., 2008. General guidelines for 17

invasive plant management based on comparative demography of invasive and 18

native plant populations. Journal of Applied Ecology 45, 1124-1133. 19

Rees, M., Paynter, Q., 1997. Biological control of Scotch broom: modelling the determinants 20

of abundance and the potential impact of introduced insect herbivores. Journal of 21

Applied Ecology 34, 1203-1221.22

Room, P.M., Harley, K.L.S., Forno, I.W., Sands, D.P.A., 1981. Successful biological control 23

of the floating weed salvinia. Nature 394, 78-80.24

61

Roni, P., Liermann, M., Jordan, C.E., Steel, E.A., 2005. Steps for designing a monitoring and 1

evaluation program for aquatic restoration. In: Roni, P. (Ed.), Methods for Monitoring 2

Stream and Watershed Restoration. American Fisheries Society, Bethesda, Maryland, 3

pp. 13-34.4

Schat, M., Blossey, B., 2005. Influence of natural and simulated leaf beetle herbivory on 5

biomass allocation and plant architecture of purple loosestrife (Lythrum salicaria L.). 6

Environmental Entomology 34, 906-914.7

Schooler, S.S., McEvoy, P.B., 2006. Relationship between insect density and plant damage 8

for the golden loosestrife beetle, Galerucella pusilla, on purple loosestrife (Lythrum 9

salicaria). Biological Control 36, 100-105.10

Schooler, S.S., Baron, Z., Julien, M., 2006a. Effect of actual and simulated herbivory on 11

alligator weed growth and reproduction. Biological Control 35, 74-79.12

Schooler, S.S., McEvoy, P.B., Coombs, E.M., 2006b. Negative per capita effects of purple 13

loosestrife and reed canary grass on plant diversity of wetland communities. Diversity 14

and Distributions 12, 351-363.15

Shea, K., 2004. Models for improving the targeting and implementation of biological control 16

of weeds. Weed Technology 18, 1578-1581.17

Shea, K., Kelly, D., 1998. Estimating biocontrol agent impact with matrix models: Carduus 18

nutans in New Zealand. Ecological Applications 8, 824-832.19

Sheppard, A.W., 2003. Prioritising agents based on predicted efficacy: beyond the lottery 20

approach. In: Spafford Jacob, H., Briese, D.T., (Eds.), Improving the Selection, 21

Testing and Evaluation of Weed Biological Control Agents. CRC for Australian Weed 22

Management Technical Series No. 7, Glen Osmond, Australia, pp. 11-22. Retrieved 23

29 April 2009, from http://www.weedscrc.org.au/publications/technical_series.html24

62

Sheppard, A.W., 2006. How to best select an agent for weed biological control. In: Hoddle M. 1

(Ed.), Proceedings of the Californian Conference on Biological Control – V. 2

University of California Citrus Research Center and Agricultural Experiment Station, 3

Riverside, USA. pp. 55-65. Retrieved 29 April 2009, from 4

http://www.cnr.berkeley.edu/biocon/CCBC%20V.htm5

Sheppard, A.W., Smyth, M., Swirepik, A., 2001. The impact of Mogulones larvatus and 6

pasture competition on Echium plantagineum. Journal of Applied Ecology 38, 291-7

300.8

Shishkoff, N., Bruckart, W.L., 1996. Water stress and damage caused by Puccinia jaceae on 9

two Centaurea species. Biological Control 6, 57-63.10

Siemann, E., Carson, W.P., Rogers, W.E., Wiesser, W.W., 2004. Reducing herbivory using 11

insecticides. In: Wiesser, W.W., Siemann, E. (Eds), Insects and ecosystem function. 12

Springer-Verlag, Berlin, pp. 303-328.13

Sinden, J.A., Griffith, G., 2007. Combining economic and ecological arguments to value the 14

environmental gains from control of 35 weeds in Australia. Ecological Economics 61, 15

396-408.16

Sinden, J., Jones, R., Hester, S., Odom, D., Kalisch, C., James, R., Cacho, O., 2004. The 17

Economic Impact of Weeds in Australia. CRC for Australian Weed Management 18

Technical Series No. 8, Glen Osmond, Australia. Retrieved 29 April 2009, from 19

http://www.weedscrc.org.au/publications/technical_series.html20

Smith, L., 2005. Host plant specificity and potential impact of Aceria salsolae (Acari: 21

Eriophyidae), an agent proposed for biological control of Russian thistle (Salsola 22

tragus). Biological Control 34, 83-92.23

63

Smith, E.G., De Clerck-Floate, R.A., Van Hezewijk, B.H., Moyer, J.R., Pavlik, E. 2009. Costs 1

of mass-producing the root weevil, Mogulones cruciger, a biological control agent for 2

houndstongue (Cynoglossum officinale L.). Biological Control 48, 281-286.3

Smith, L., Story, J.M., 2003. Plant size preference of Agapeta zoegana L. (Lepidoptera: 4

Tortricidae), a root-feeding biological control agent of spotted knapweed. Biological 5

Control 26, 270-278.6

Smith, M.C., Reeder, R.H., Thomas, M.B., 1997. A model to determine the potential for 7

biological control of Rottboellia cochinchinensis with the head smut Sporisorium 8

ophiuri. Journal of Applied Ecology 34, 388-398.9

Stansbury, C.D., Batchelor, K.L., Morin, L., Woodburn, T.L., Scott, J.K., 2007. Importance of 10

support for fruit production by the invasive climber, Asparagus asparagoides. Weed 11

Technology 21, 820-824.12

Stonehouse, J., 1994. Assessment of Andean bean diseases using visual keys. Plant Pathology 13

43, 519-527. 14

Story, J.M., Callan, N.W., Corn, J.G., White, L.J., 2006. Decline of spotted knapweed density 15

at two sites in western Montana with large populations of the introduced root weevil, 16

Cyphocleonus achates (Fahraeus). Biological Control 38, 227-232. 17

Story, J.M., Smith, L., Good, W.R., 2001. Relationship among growth attributes of spotted 18

knapweed (Centaurea maculosa) in Western Montana. Weed Technology 15, 750-19

761.20

Stuart, R.M., Kriticos, D.J., Ash, J.E., 2002. Modelling the biological control of bitou bush 21

(Chrysanthemoides monilifera: Asteraceae) by Mesoclanis polana (Tephritidae). In: 22

Spafford Jacob, H., Dodd, J., Moore, J.H. (Eds.), Proceedings of the 13th Australian 23

Weeds Conference. Plant Protection Society of WA, Perth, pp. 591-594.24

64

Syrett, P., Briese, D.T., Hoffmann, J.H., 2000. Success in biological control of terrestrial 1

weeds by arthropods. In: Gurr, G, Wratton, S. (Eds.), Biological Control: Measures of 2

Success. Kluwer Academic publishers, Amsterdam, pp. 189-230.3

Swirepik, A.E., Smyth, M.J., 2003. Evaluating biological control at the regional scale. In: 4

Spafford Jacob, H., Briese, D.T., (Eds.), Improving the Selection, Testing and 5

Evaluation of Weed Biological Control Agents. CRC for Australian Weed 6

Management Technical Series No. 7, Glen Osmond, Australia, pp. 61-67. Retrieved 7

29 April 2009, from http://www.weedscrc.org.au/publications/technical_series.html8

Thomas, M.B., Casula, P., Wilby, A., 2004. Biological control and indirect effects. Trends in 9

Ecology and Evolution 19, 61.10

Thomas, M.B., Reid, A.M., 2007. Are exotic natural enemies an effective way of controlling 11

invasive plants? Trends in Ecology and Evolution 22, 447-453.12

Tipping, P.W., Center, T.D., 2002. Evaluating acephate for insecticide exclusion of Oxyops 13

Vitiosa (Coleoptera: Curculionidae) from Melaleuca quinquenervia. Florida 14

Entomologist 85, 458-463.15

Tipping, P.W., Martin, M.R., Nimmo, K.R., Pierce, R.M., Smart, M.D., White, E., Madeira, 16

P.T., Center, T.D., 2008. Invasion of a West Everglades wetland by Melaleuca 17

quinquenervia countered by classical biological control. Biological Control 48, 73-78. 18

Trujillo, E.E., Kadooka, C., Tanimoto, V., Bergfeld, A., Shishido, G., Kawakami, G., 2001. 19

Effective biomass reduction of the invasive weed species banaba poka by Septoria leaf 20

spot. Plant Disease 85, 357-361.21

Turpie, J., 2004. The role of resource economics in the control of invasive alien plants in 22

South Africa. South African Journal of Science 100, 87-93.23

Underwood, A.J., 1993. The mechanics of spatially replicated sampling programmes to detect 24

environmental impacts in a variable world. Australian Journal of Ecology 18, 99-116.25

65

van Klinken, R.D., Raghu, S., 2006. A scientific approach to agent selection. Australian 1

Journal of Entomology 45, 253-258.2

van Wilgen, B.W., de Wit, M.P., Anderson, H.J., Le Maitre, D.C., Kotze, I.M., Ndala, S., 3

Brown, B., Rapholo, M.B., 2004. Costs and benefits of biological control of invasive 4

alien plants: case studies from South Africa. South African Journal of Science 100, 5

113-122.6

Vitousek, P.M., 1990. Biological invasions and ecosystem processes: towards an integration 7

of population biology and ecosystem studies. Oikos 57, 7-13. 8

Vos, P., Meelis, E., Ter Keurs, W.J., 2000. A framework for the design of ecological 9

monitoring programs as a tool for environmental and nature management. 10

Environmental Monitoring and Assessment 61, 317-344.11

Wajnberg, E., Scott, J.K., Quimby, P.C., 2001. Evaluating Indirect Ecological Effects of 12

Biological Control. CABI, Wallingford, UK.13

Walker, S.R., Taylor, I.N., Milne, G., Osten, V.A., Hoque, Z., Farquharson, R.J., 2005. A 14

survey of management and economic impact of weeds in dryland cotton cropping 15

systems of subtropical Australia. Australian Journal of Experimental Agriculture 45, 16

79-91. 17

Willis, A.J., Memmott, J., 2005. The potential for indirect effects between a weed, one of its 18

biocontrol agents and native herbivores: a food web approach. Biological Control 35, 19

299-306.20

Willis, A.J., Ash, J.E., Groves, R.H., 1998. Interactions between plant competition and 21

herbivory on the growth of Hypericum species: a comparison of glasshouse and field 22

results. Australian Journal of Botany 46, 707-721.23

Willis, A.J., Kilby, M.J., McMaster, K., Cullen, J.M., Groves, R.H., 2003. Predictability and 24

acceptability: Potential for damage to non-target plant species by biological control 25

66

agents for weeds. In: Spafford Jacob, H., Briese, D.T. (Eds.), Improving the Selection, 1

Testing and Evaluation of Weed Biological Control Agents, Perth, Western Australia, 2

Technical Series no. 7, CRC for Australian Weed Management, Glen Osmond, 3

Australia, pp. 35-49. Retrieved 29 April 2009, from 4

http://www.weedscrc.org.au/publications/technical_series.html5

Wirf , L., 2006. The effect of manual defoliation and Macaria pallidata (Geometridae) 6

herbivory on Mimosa pigra: implications for biological control. Biological Control 37, 7

346-353.8

Wood, A.R., 2002. Infection of Chrysanthemoides monilifera ssp. monilifera by the rust 9

fungus Endophyllum osteospermi is associated with a reduction in vegetative growth 10

and reproduction. Australasian Plant Pathology 31, 409-415.11

Wood, A.R., Morris, M.J., 2007. Impact of the gall-forming rust fungus Uromycladium 12

tepperianum on the invasive tree Acacia saligna in South Africa: 15 years of 13

monitoring. Biological Control 41, 68-77.14

Yates, C.N., Murphy, S.D., 2008. Observations of herbivore attack on garlic mustard (Alliaria15

petiolata) in Southwestern Ontario, Canada. Biological Invasions 10, 757-760.16

Yokomizo, H., Possingham, H.P., Thomas, M.B., Buckley, Y.M., 2009. Managing the impact 17

of invasive species: the value of knowing the density-impact curve. Ecological 18

Applications 19, 376-386.19

20

21

22

2324

67

FIGURE CAPTION1

2

Figure 1. Key steps of a weed biological control program, incorporating activities 3

contributing to the evaluation of agent effectiveness.4

5

Identify candidate agents and prioritise (based on known host range, potential

effectiveness and capacity of agents to escape parasites/predators and increase

once introduced)

Assess ecological, economic and sociological impacts of weed and its suitability for biological control

Test efficacy of most promising candidate agent (use field studies in native range,

laboratory/glasshouse experiments and/or modelling)

Perform preliminary host-specificity tests and if results are promising undertake detailed testing

Identify most susceptible life stage(s) of weed to target with biological control (use damage simulation experiments and/or modelling)

Release agent at sites across the weed range (including sites where baseline data has been collected. For agents that disperse slowly,

keep some sites agent-free to serve as control.

Monitor long-term effects of agent on weed populations and changes in associated vegetation and if necessary key ecosystem processes

(using sites where baseline data was collected)

Monitor short-term effects of agent on weed populations

(compare sites with and without the agent, perform correlative studies or

exclusion experiments)

Survey native range for natural enemies

Define performance targets for biological control program (e.g. level of control necessary to ameliorate weed impacts, to make the weed more

amenable to other control measures and/or reduce spread)

Collect baseline ecological data on selected populations of weed and associated plant

communities and if necessary key ecosystem processes in the introduced range

Monitor establishment and spread

Assist spread of agent with redistribution if necessary

Use modelling and benefit-cost analyses to demonstrate whether, and in what ways the program has been successful; quantifying where possible the rate of return on investment, as

well as social and environmental benefits

Figure