interpreting some outstanding features of the flora and

Interpreting some outstanding features of the flora and vegetation of Madagascar

Peter J. Grubb

Department of Plant Sciences, University of Cambridge, UK

Received: 14 March 2003 · Revised version accepted: 25 April 2003

Abstract

Six features are covered. (1) The high endemism, which is not discussed in detail, is all-perva-sive, and has resulted from the isolation of Madagascar from Africa some 125 million yearsago and their present separation by 430 km. (2) The great richness in plant species (especiallyrelative to Africa), seen particularly in the families of woody species in the wetter vegetation-types, involves both sympatry and allopatry within genera, and is explicable in terms of muchless extreme drying out than in Africa during the Pleistocene and effective ‘species-pumping’rather than mass extinctions during that period. (3) The abundance and species-richness ofpalms, pandans, tree-ferns, bamboos, and certain families of dicot trees (notably Lauraceae,Monimiaceae, Myrsinaceae and Myristicaceae) in the lowland rain forests also appears to bea result of both past and present wetness of the climate, while it is hypothesized that the lowstature of most lowland rain forests, paucity of large-girth trees, and small size and sparsity ofbroad-leaved herbs, are a result of most rain forest soils being old and relatively nutrient-poor. (4) Within the dry evergreen forest region where rainfall is moderate (900–1600 mmyr–1) a sub-set of trees with fire-resistant bark seems to have evolved at sites prone to frequentground fires, some perhaps spreading out of adjacent palm savanna on seasonally floodedsites. (5) Both the evolution of thicket rather than grassy woodland in the driest areas(300–600 mm yr–1), and the abundance of evergreen trees and shrubs on ordinary soils – notconfined to run-on sites – are explicable in terms of there being a finite chance of rainthroughout the year rather than one short wet season, coupled with relatively high values forair humidity throughout the year. The same factors probably explain the abundance and vari-ety of succulents in the thicket; they are found throughout and not just on rocks. (6) Concern-ing physical defence against herbivores, the rain forests, dry evergreen forests and deciduousforests all show a complete lack of plants with physiognomic features plausibly related tobrowsing by extinct giant birds (a strong contrast with New Zealand), but in the semi-decidu-ous thicket there are many tiny-leaved, mostly non-spiny shrubs and small trees, whose densebranching and impenetrability have plausibly evolved as a defence against browsing by ele-phant birds. The Didiereaceae of the thicket are spiny (unlike members of the same family inAfrica), and are giant analogues of the ‘ocotillo’ (Fouquieria splendens) in western NorthAmerica rather than of Cactaceae; their spines appear to be protecting the leaves more thanthe stems against arboreal primates, spine length paralleling leaf length.

Key words: evergreenness, fire, Madagascar, plant defences, rain forest, savanna, soil fertility,species-richness, succulents, thicket

1433-8319/03/6/01-02-125 $ 15.00/0

Corresponding author: Peter J. Grubb, Department of Plant Sciences, University of Cambridge, Downing Street, Cambridge CB2 3EA, UK;e-mail: [email protected]

Vol. 6/1,2, pp. 125–146© Urban & Fischer Verlag, 2003http://www.urbanfischer.de/journals/ppees

Perspectives in Plant Ecology,Evolution and Systematics

Introduction

The flora and vegetation of Madagascar are outstand-ing in a number of ways. Two features of the flora arealmost all-pervasive: high endemicity and greatspecies-richness. Otherwise the vegetation has manyoutstanding features which are likely to make uniqueimpressions on individual plant ecologists. Four sets offeatures impressed me particularly during a visit inNovember 2001, and are treated in this paper. Firstly,the lowland rain forests are distinctive in being no-tably rich in palms, pandans, bamboos and tree-ferns,while also being of relatively low stature, lacking inlarge-girth trees, and poor in large-leaved monocotherbs. Secondly, while there is no natural savanna(apart from possible palm savanna in seasonally wa-terlogged hollows) the dry evergreen forest region in-cludes a small set of trees highly adapted to fire. Third-ly, the nearly-natural vegetation of the driest part isthicket, not grassy woodland, and has a higher inci-dence of evergreens than the dry deciduous forest; italso has an abundance and great diversity of succu-lents as well as eleven species of Didiereaceae, strangetall spiny plants which are analogues of the ocotillo(Fouquieria). Fourthly, until a few hundred years agothere were large flightless birds which apparently atethe vegetative parts of plants and were the equivalentof the similarly extinct moas in New Zealand ratherthan the still-extant, fruit-eating cassowaries of north-eastern Australia and Papua New Guinea; the extentto which plants were defended against the birds hasnot been made clear. Contrasting with the presence ofgiant herbivorous birds is the total absence of the her-bivorous mammal flora associated with grassland andwoodland (‘savanna’) in Africa. Also, remarkably, hu-mans reached Madagascar only in the last 2000 years.

In my experience most reasonably well-read andtravelled plant ecologists are ignorant of most of thepoints made in the last paragraph, apart from the highendemicity which was spelled out in the classical liter-ature (Good 1953). That issue has been covered re-cently by Schatz (2001) and in the volume edited byLourenço & Goodman (2000), and I shall not discussit in this paper. The great richness in species was em-phasized by Guillaumet & Mangenot (1975), and therichness of the flora in general was explained by Leroy(1978) in terms of much less severe drying out in thegeological past than occurred in Africa, but I believe itis still not well known, and its extent certainly as-tounded me when I visited the island. I shall provide anumerical comparison with eastern Africa for onefunctional group of plants to drive home the point.

The ignorance of the flora and vegetation of Mada-gascar on the part of most of the world’s plant ecolo-gists has resulted to some degree no doubt from the

difficulty and expense of getting there relative toAfrica in the second half of the 20th century, but I be-lieve it reflects also the lack of coverage in major treat-ments of the world’s vegetation from the time ofSchimper (1903) onward. In Walter’s opus magnum,Die Vegetation der Erde (1964, English translation1971), there is very little mention of Madagascar.Sadly the superb account of the flora and vegetationby Koechlin et al. (1974), summarizing the results of along history of research, was not widely read in theEnglish-speaking world, partly because it was pub-lished relatively obscurely and partly because it was inFrench. The overview in English by Koechlin (1972)appeared in an expensive book of limited circulation.More accessible summaries in English were given byWhite (1983) and Guillaumet (1984). Lowry et al.(1997) have provided an invaluable historical analysisof descriptive and interpretive work on the vegetationof the island; with the web version of this paper(http://www.mobot.org/MOBOT/Madagasc/vegmad1.html) there is an extended bibliography. The twovolumes by Rauh (1995–1998) contain very numerouscolour photos of landscapes, vegetation-types andplants, with emphasis on the drier areas. There hasbeen widespread concern among conservation biolo-gists about the conversion of vast areas of forest intograssland, the accompanying soil erosion, and the ac-tual or potential loss of many species of plants and an-imals, and recent summaries of those issues are givenin the volume edited by Goodman & Patterson(1997).

I have chosen to write about the flora and vegeta-tion of Madagascar because they were the topics I dis-cussed with Tim Whitmore when I was last able to seehim in January 2002, soon after I had visited the is-land. The mixture of issues in phytogeography andecology fascinated him. I fully accept that they repre-sent my personal response to the vegetation, and thatother topics are equally deserving of attention, such asthe ecophysiology of the very various drought-resis-tant plants on the many rock outcrops, or the differ-ences between Madagascar and Africa or other partsof the tropics in pollination, seed predation and seeddispersal.

I begin with a brief account of the physical featuresof the island relevant to the distributions of four majorvegetation-types. Then I treat in turn features of inter-est in (1) the rain forests and high-altitude thickets, (2)the dry evergreen forests and fire-adapted woodlands,and (3) the deciduous forests and semi-evergreen thick-ets. I conclude with sections on physical defences andsome outstanding puzzles. My accounts of the variousvegetation-types are based on those of Koechlin et al.(1974), but are supported by personal experience of allthe major types described apart from the high-altitude

126 P.J. Grubb

Perspectives in Plant Ecology, Evolution and Systematics (2003) 6, 125–146

thickets. Koechlin et al. (1974) also described the re-maining vegetation-types such as those on rock out-crops, mangroves, salt marshes and other wetlands.Nomenclature follows Flore de Madagascar et des Co-mores (Humbert et al. 1936–2002), except for legumes(Du Puy 2001) and palms (Dransfield & Beentje 1995),unless otherwise indicated. Assignment to families fol-lows Stevens (2002), except that for convenience I re-tain the Didiereaceae for four Madagsacan genera andthree African genera.

Physical features of the island

Madagascar is about 1570 km from north to southand – at its widest – about 560 km from east to west(Fig. 1). The closest point is about 430 km from Africa(Krause et al. 1997). The island lies between 12°30′Sand about 25°32′S; in Africa this span runs fromnorthern Mozambique to northern South Africa.There is a spine of mountains running along the east-ern side of the island, with much land above 900 m al-titude, but little over 2000 m. This spine intercepts thesouth-easterly winds which are the most frequent. As aresult, shown in Fig. 1, the eastern strip is very wet(2400 to >3600 mm yr–1). Mean annual rainfall de-clines sharply going westward across the mountains(2400 down to 1600 mm yr–1), and it falls furtheracross the lower lying land toward the west coast, es-pecially toward the south-west (1600 down to 600 mm yr–1). A relatively narrow strip on the south-western and southern sides is yet drier (600–300 mmyr–1). At sea level the mean temperature of the coldestmonth is >25 °C at the northern tip, >20 °C in thesouth-eastern corner, and c. 20 °C in the dry south andsouth-west (Donque 1972; Koechlin 1972). The meanof the hottest month is 25–28.5 °C along the easternside from north to south, and 27–28 °C in the drysouth and south-west.

Recent authors have adopted a simplified version ofthe bioclimatic regions defined by Cornet (1974), asshown in Fig. 1 of Schatz (2001). The ‘humid’ regionapproximates to that with >1600 mm yr–1 and no pro-nounced dry season, the ‘subhumid’ region to that with800–2000 mm yr–1 and a marked dry season, the ‘dry’region to that with 800–1600 mm yr–1 and a strongerdry season, and the ‘subarid’ to that with <800 mm yr–1

and a significant possibility of rain at any time of year.The ‘montane’ region is effectively that above 2000 maltitude, and does not correspond to the normal usageof ‘montane’ in the world literature on mountainforests (cf. Grubb 1977; Whitmore 1985).

The core of the island from north to south is madeof Precambrian metamorphosed basement rock, inter-rupted by large intrusions of granite; there is a narrow

sedimentary plain to the east, and a wide one to thewest (Brenon 1972). There is much limestone on thewestern side, and this crosses from west to east in thenorth. There is a characteristic ‘ruiniform’ sandstonein the Isalo National Park south-west of the centre.Volcanic deposits are scattered along the spine of theisland, ranging from Cretaceous to Pliocene/Pleis-tocene, being younger toward the north. Quaternarydunes are extensive along the S and SW coasts; theolder sands are partially decalcified and reddened. Avaluable simplified map of the geology is given by DuPuy (2001). Most of the soils in the wetter part of theisland are derived from acidic rocks and long leached,and therefore strongly acidic and relatively poor inavailable nutrients. The soils derived from the sands

Flora and vegetation of Madagascar 127


Fig. 1. Isohyets for annual rainfall (cm) in Madagascar (from map 3 ofKoechlin et al. 1974).

on the narrow eastern plain (carrying ‘littoral forest’)are especially poor. Those along the central strip of theisland, suffering seasonal drought, have degeneratedas a result of forest clearance and dominance by grass,and the worst areas show severe erosion.

Rain forests and high-altitude thickets

It is generally believed that before humans invaded theisland the ‘zonal’ vegetation of the humid region and agood deal of the eastern part of the subhumid regionwas composed of rain forests. In the remaining vegeta-tion the trends with altitude are those found generallyon wet tropical mountains (cf. Grubb 1977). As usual,the transition from lowland to lower montane rainforest is gradual (Lowry et al. 1997); the transitionfrom ‘medium-altitude mossy forest’ (lower montanerain forest) to ‘high-altitude lichen forest’ (upper mon-tane rain forest) is placed by Koechlin (1972) ataround 1300 m, which is the same as in NE Queens-land at the same range of latitude (Tracey 1982). Atabout 1800–2000 m the montane forest merges intohigh-altitude thicket (Guillaumet 1983); in the broad-est sense the thicket is reminiscent of the ‘subalpinerain forest’ of the high mountains of Papua NewGuinea (Grubb & Stevens 1985), being quite species-rich, with most of the woody plants having scleromor-phic leaves, some of them ericoid. It seems that thiszone in Madagascar is a good deal more species-richthan the corresponding Ericaceae-dominated commu-nity of the higher East African mountains (cf. Lind &Morrison 1974).

The great species-richness of the flora

I confine myself to trees, treelets and shrubs ‘at least(4–)5 m tall, and/or with at least one vertical stem at-taining 5 cm in diameter at breast height’. Schatz(2001) provides numbers of species per genus for suchplants in his Generic Tree Flora. I have accepted his as-signment of genera to families except where more re-cent research has suggested better arrangements(Stevens 2002 and references cited by him). The result isa list of 110 families. Of these five are endemic, andthree are widespread non-endemic families not found ineastern Africa (Chloranthaceae, Elaeocarpaceae andWinteraceae). Of the 102 non-endemic families that arepresent in Madagascar and eastern Africa, I consideronly the 56 found exclusively or primarily in the humidand subhumid areas, and I omit the 18 found only orprimarily in the subhumid, dry and subarid areas, andthe 28 which are shared fairly evenly between the wet-ter and drier regions. I add the Gramineae-Bambuseae,

which are found mainly in the humid and subhumidareas (Dransfield 2000, 2002), and which were not cov-ered by Schatz (2001).

I compare the numbers of genera and species in therelevant families in Madagascar with the numbers inthose families in what I call ‘eastern Africa’, i.e. thearea stretching from the northern border of Kenya(5°N) to Port Elizabeth in the Eastern Cape of SouthAfrica (34°S), covered by the Flora of Tropical EastAfrica (Kenya, Uganda and Tanzania; Turrill et al.1952–2002), the Flora Zambeziaca (Malawi, Zambia,Zimbabwe, Mozambique, Botswana and the CapriviStrip of Namibia; Exell et al. 1960–2002) and TreeFloras of South Africa by Palgrave (1988) and vanWyk & van Wyk (1997). I have included only thoseSouth African species occurring east of 26°E, whichpasses close to Port Elizabeth. A notable feature ofmany genera in Madagascar is the difficulty of decid-ing just how many species to recognize (Koechlin et al.1974), but the summary produced by Schatz (2001) istreated here as a proper basis for comparison withAfrica.

Of the 57 families found primarily or only in thehumid-subhumid regions in Madagascar, almost al-ways in evergreen forest, 28 have ten or more indige-nous species on the island, and are shown in Table 1.All of the 29 families with <10 species in Madagascarare relatively small on a world scale.

For three of the 28 families in Table 1 we cannotmake a full comparison because either the TropicalEast Africa account has not been published (Eri-caceae) or the Zambeziaca account (Lamiaceae andThymelaeaceae). Thymelaeaceae have fewer species inMadagascar than in eastern Africa. All of the 25 fami-lies for which we have full information have morespecies in Madagascar than in eastern Africa, al-though the difference is marginal in five cases wherethe quotient Madagascar/Africa is <1.5 (Annonaceae,Icacinaceae, Pittosporaceae, Rhizophoraceae andScrophulariaceae). Most spectacular are the quotientsfor the Monimiaceae (43), Cunoniaceae (20), Pan-danaceae (17), Arecaceae (14), Myrsinaceae (8.8) andLauraceae (8.5). On a world scale all of these familiesare most speciose in high-rainfall areas. The medianquotient is 3.2. In contrast, 8/26 families have fewergenera in Madagascar, and the median quotient fornumbers of genera is 1.0. The explanation offered byLeroy (1978) for the greater richness of the flora ingeneral in Madagascar than in Africa – that it sufferedmuch less drying out during the Pleistocene, Plioceneand probably late Miocene – must surely be part ofthe explanation of the particular feature of richness inspecies per genus.

Examination of the recently published floras forpalms (Dransfield & Beentje 1995) and legumes (Du

128 P.J. Grubb


Puy 2001), and the check-list of orchids (Du Puy et al.1999), as well as the older and less complete accountsof many families in the Flore de Madagascar showsthat there is a great deal of allopatry in the present-daydistributions of congeners. However, there are alsostriking instances of sympatry, for example 13 speciesof Diospyros on a single 0.1-ha plot (Gentry 1993).The situation is like that in the part of the NorthernHemisphere apparently least dried out in the Pleis-tocene, i.e. eastern Asia, where moderate drying outand isolation of populations rather than their annihila-tion, has led to ‘species-pumping’ (Grubb 1987; Lath-am & Rickleffs 1993).

Although the evolution of species-richness in Mada-gascar is most striking in the wetter parts, many generain the drier parts also show the evolution of numerous,largely allopatric species. Adansonia, Aloe, Euphorbiaand Pachypodium, all with distinctive growth forms, aregood examples. Moreover, three of the four genera ofwoody plants with the most species in Madagascar spanthe drier and wetter parts of the island (Schatz 2001):Diospyros c. 125 spp., Croton c. 150 spp. and Dombeyac. 180 spp. (Dypsis with 137 spp., almost confined to thehumid and subhumid parts, is the fourth). Thus moder-ate drying out seems to have favoured speciation in thedrier as well as the wetter areas.



Table 1. Families of trees found in Madagascar (MAD) wholly or predominantly in evergreen forests in the humid-subhumid regions, and having ten or morespecies on the island: numbers of species and genera compared with those in SE Africa (SEA, Flora Zambesiaca area plus eastern South Africa), Tropical EastAfrica (TEA, the area of the Flora) and all eastern Africa (AEA). Definitions and sources given in the text.

Plant families Biogeographical units––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––Number of species Number of genera–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––– ––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––MAD SEA TEA AEA MAD SEA TEA AEA

PTEROPSIDA

Cyatheaceae 40 4 13 13 1 1 1 1

MONOCOTS

Arecaceae 170 9 9 12 13 5 6 7Pandanaceae 85 1 4 5 1 1 1 1Poaceae (tribe Bambuseae) 32 4 5 6 11 3 5 5

DICOTS

Annonaceae 62 30 48 56 6 11 24 24Araliaceae 66 16 15 24 3 4 3 4Bignoniaceae 61–65 7 8 10 8 6 5 7Clusiaceae 74–80 14 18 22 5 3 5 5Cunoniaceae 40 2 0 2 1 2 0 2Ericaceae 41–50 21 – – 3 3 – –Hamamelidaceae 12 3 3 5 1 1 1 1Hypericaceae 27 4 9 10 3 3 3 4Icacinaceae 10 4 4 7 4 3 3 4Lamiaceae 138 – 90 – 6 – 4 –Lauraceae 111 12 5 13 5 3 3 3Lecythidaceae 13 1 3 3 2 1 2 2Melastomataceae 94 3 21 21 3 2 4 4Memecylaceae 87 8 26 31 2 1 1 1Monimiaceae 43 1 1 1 3 1 1 1Myristicaceae 10 2 3 3 3 2 3 3Myrsinaceae 106 8 7 12 4 3 4 4Myrtaceae 58 7 18 21 2 2 2 2Pittosporaceae 11 4 6 8 1 1 1 1Rhizophoraceae 16–21 11 10 14 6 4 4 4Salicaceae 107 35 30 42 10 12 12 13Sapotaceae 82–86 29 45 52 9 12 13 13Scrophulariaceae 10 6 5 7 3 2 2 2Thymelaeaceae 14–16 >10* 11 >17* 5 >5* 5 7

*Based on data for eastern South Africa, and in the Flora of Tropical East Africa account; no account for the Zambesiaca region is yet available.–, no information available.

Structure and physiognomy of the lowlandrain forests

Following on from the previous section, we see thatthere are four growth forms whose abundance in thelowland rain forests of Madagascar is plausibly to beexplained in terms of past and present wetness: palms,pandans, bamboos and tree-ferns. Both palms andpandans are not only abundant but also strikingly di-verse in mature size and in form; in contrast, in largeareas of Tropical East Africa there are very few speciesof palms and pandans, and they are confined to topo-graphically determined wet sites. In Madagascar sever-al bamboo species are widespread and abundant in thelowlands, although there are certainly more species inthe montane forests (Dransfield 2000); in Africa bam-boos – bar one species (Lind & Morrison 1974) – areknown only in montane forests. Likewise tree-fernsare confined to montane forests in eastern Africa; theother parts of the world with abundant tree-ferns inthe lowlands have especially high rainfall, such as theSolomon Islands.

Koechlin et al. (1974) emphasized three features ofthe lowland rain forest which I hypothesize to be anevolutionary result of low availability of limiting min-eral nutrients in most of the soils. The largest trees arenot very tall (few >30 m) and have relatively smallgirths (few much >1 m). As expected in a forest withnot many large-girth trees, the stem density is notably

high (Gentry 1993; Lowry et al. 1997). The third fea-ture is a paucity of tall, broad-leaved monocot herbs.The paucity of these herbs in terms of biomass (mostspecies confined to gaps and not >0.5 m tall) is paral-leled by a paucity of species, compared with TropicalEast Africa (Table 2), a striking contrast with the rich-ness in species of woody plants documented in Table 1.

In Africa, Australia and South America forestheight commonly parallels soil fertility (Hall & Swaine1976; Tracey 1982; Coomes & Grubb 1996; Richards1996). The position in SE Asia is confused by thesuper-plants of the Dipterocarpaceae which make verytall forests on some poor soils, but otherwise the cor-relation is also found there (Whitmore 1985). Vigor-ous growth and diversity of herbs are commonly asso-ciated with higher soil fertility in both tropical andtemperate vegetation (Grubb 1987; Coomes & Grubb1996). Determination of foliar nutrient concentrationscould be useful first test of the nutrient-shortage hy-pothesis.

Madagascan rain forests are notable for the abun-dance and diversity of Dracaena species. There is someparallel in Africa, but the extremes are greater inMadagascar, as illustrated by the undescribed speciesfound in forest on rocky ridges on the Masoala Penin-sula with little branched stems up to c. 12 m tall andleaves c. 2 m long and 10 cm wide (Fig. 2).

Damage by cyclones is a major feature in the dy-namics of the forests, and one notable plant to benefitfrom the gaps is the ‘Travellers’ Palm’, Ravenala

130 P.J. Grubb


Table 2. Comparison of species-richness in families of broad-leaved perennial monocotyledonous herbs found in lowland rain forests in Madagascar and inTropical East Africa. Information has been taken from Flore de Madagascar et Comores, Flora of Tropical East Africa, Thistleton-Dyer (1902) and Agnew (1974).

Plant families Number of species (genus names)–––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––––Madagascar Tropical East Africa

Anthericaceae 1 (Chlorophytum) 3 (Chlorophytum)

Araceae 4 (Arophyton) 19 (Amorphophallus 3, Anchomanes 2, Arisaema 5, Callopsis 1,Gonatopus 4, Stylochiton 3, Zamioculcas 1)1

Commelinaceae 4 (Commelina 2, Pollia 1, Pseudoparis 1)2 11 (Aneilema 4, Commelina 4, Floscopa 1, Palisota 1, Pollia 1)

Costaceae 0 7 (Costus)Dioscoreaceae 1 (Tacca) 1 (Tacca)

Marantaceae 2 (Halopegia, Marantochloa) 5 (Ataenidia, Marantochloa, Megaphrynum, Sarcophrynum, Trachyphrynum)

Musaceae 03 1 (Ensete)

Poaceae (tribes Olyreae, Phareae, 2 (Leptaspis, Olyra) 4 (Leptaspis, Maltebrunia, Olyra, Streptogyna)Streptogyneae & Oryzeae)

Zingiberaceae 5 (Alautandra)4, 5 18 (Aframomum 15, Renealmia 3)

1 All but the Callopsis and Zamioculcas die back seasonally, so are found in rain forests with a dry season.2 Other species in these genera, and five other genera are present in Madagascar, but are found at open, non-forest sites.3 Ensete perrieri is confined to deciduous forests.4 Plus 2–3 undescribed species (G. Schatz, pers. comm., March 2003).5 Afromomum angustifolium and Hedychium peregrinum are treated as being introduced.

madagascariensis (Strelitiziaceae) with leaves likethose of a banana plant and a woody stem up to c. 30m tall (Hladik et al. 2000). Its seeds, a few mm acrossand hard-coated, persist in the soil awaiting gap-for-mation (Koechlin 1972).

Dry evergreen forests and fire-adaptedwoodlands

It is commonly believed that before the intervention ofhumans dry evergreen forest occupied much of thewestern part of the subhumid region, though this hasnow been deforested singularly completely over vastareas. Most remnants of this vegetation-type occur onthe western scarp of the central mountains (above c.800 m), but there are a few within the central moun-tainous area. The stands that appear to be least dis-

turbed consist of small-trunked trees rarely >12 m tallwith small leaves – mostly microphyll sensu Raunkiaer(Thomasson 1977). The canopy lets through a gooddeal of light to the undergrowth of shrubs and sub-shrubs, often dominated by Erica and Helichrysum.There are few climbers and very few epiphytes.

The version of the ‘dry evergreen forest’ encoun-tered most often now is in fact a woodland, i.e. thetrees are at some distance from each other (Fig. 3). Thetrees belong to a small sub-set of species in the forest.The overwhelming dominant is usually the tapia tree,Uapaca bojeri (Phyllanthaceae, formerly Euphor-biaceae). Like the four Asteropeia spp. and the decidu-ous Dicoma incana also found in this kind of wood-land, it has thick, fissured, fire-resistant bark (Fig. 4).There is often little other than grass beneath the trees.A few extensive stands of Uapaca are found well with-in the major mountainous area, e.g. on both east- andwest-facing slopes near the main north-south road forabout 13 km some 20 km south of Antisrabe. The per-sistence of such stands is possibly due to the use of asilkworm (Boroceras madagascariensis) which lives onthe Uapaca.

Like the drier facies of the montane rain forests andhigh-altitude thickets to the east, the dry evergreen for-est is flammable after dry spells, and is likely to havebeen set alight by lightning strikes, at a guess every50–100 years. There are certainly remains of charcoalin peat formed well before the arrival of humans1500–2000 years ago (Burney 1997). Presumablymany of the species sprouted after occasional fire.

What is striking about the Uapaca and Asteropeia isthat they are resistant to frequent ground fires ratherthan periodic crown fires, i.e. they are like Pinus palus-tris rather than like P. rigida among pines (cf. Burns &Honkala 1990). Uapaca and Asteropeia do resprout as



Fig. 2. Lowland rain forest on a rocky ridge on the Masoala Peninsula in NEMadagascar, showing an unnamed Dracaena species with unbranched trunkand huge leaves.

Fig. 3. Dry evergreen woodland dominated by Uapaca bojeri on shallowsoils over sandstone ion the Isalo National Park; the major grass is Isalushumbertii.

juveniles, and multi-stemmed trees are seen, but theyare not recorded as resprouting from the crowns. I sawcrowns of Uapaca that had been partially burnt andkilled in the previous year, and these had not resprout-ed. It seems likely to me that these trees evolved as thedominants of sites within the dry evergreen forest land-scape that were subject to relatively frequent ground-level fires. One possibility is that Uapaca and its associ-ates were dominant on slopes immediately above natu-ral savannas with the palm Bismarckia nobilis in hol-lows where the soils are waterlogged in the wet season,but the grasses dry out thoroughly in the dry season.This scenario is suggested by what can be seen at pre-sent immediately south-west of the Isalo National Park(Fig. 5). Nowadays the Bismarckia savanna here isprobably burnt yearly, and it has almost certainly lostsome woody elements, just as can be seen still happen-ing in Mauritia flexuosa savanna at remote sites in thenorthern Amazon (Coomes & Grubb 1996).

The replacement of relatively species-rich dry ever-green forest by species-poor Uapaca woodland overwide areas parallels the replacement of Marquesia-dominated dry evergreen forest in Zambia by Brachys-tegia woodland as a result of burning (Trapnell 1959;Lawton 1978). The replacement of Uapaca woodlandby grassland under a regime of annual burning is alsoparalleled in Zambia.

Deciduous forests and semi-deciduous thickets

Deciduous forests

The received view is that before humans arrived decidu-ous forests occupied most soils in the whole of the ‘dry’region, the northern part of the ‘subarid’ and some frac-tion of western part of the ‘subhumid’ region. Today

132 P.J. Grubb


Fig. 4. The base of a forked-trunk Uapaca bojeri with thick, fissured barkand a basal burn-scar, providing evidence of recovery by sprouting.

Fig. 6. A line of ‘coniferoid’ Pandanus along a seasonal water course per-sisting in a landscape where deciduous forest on soils not subject to periodicinundation has been totally replaced by grassland through frequent burning.

Fig. 5. Frequently burned savanna with the palm Bismarckia nobilis, a littlesouth-west of the Isalo National Park, showing in the distance Uapaca bojeriwoodland on slopes immediately uphill of the savanna.

remnants occur in a discontinuous band along the west-ern side of the island from the northern end, and mergeinto semi-deciduous thicket in the south. Most of theland they are believed to have covered formerly is nowoccupied by grassland with or without scattered treesor shrubs (‘savanna’). In the gently rolling country westof the Isalo National Park, we see along the courses ofseasonal streams and rivers that run through an other-wise deforested landscape lines of amazing ‘coniferoid’Pandanus (section Acanthostyla), with a crown shapelike that of a Christmas tree, Picea abies (Figs. 6 & 7;see also Stone 1970, and Rauh 1995–1998, p. 74). Ap-parently a trunk structure that involves the phloem inbeing deeply buried inside, away from the transienthigh temperatures of grass fires, enables the tree to sur-vive such fires in the manner seen commonly for palms,also for Ravenala (Koechlin et al. 1974) and sometimesfor tree-ferns, as in Papua New Guinea (cf. cover ofGrubb & Stevens 1985).

The deciduous forests have disappeared most com-pletely from the clayey soils, and the most extensiveremnants are on limestone, with others on sand. Theremnant forests vary greatly in height (8–30 m), de-pending on soil depth and degree of access to subsoilwater, but in most areas the canopy is generally 12–15 m tall, with occasional emergents. They are lessrich in species per unit area in small plots than the rainforests (Abraham et al. 1996). Although the forests areoverwhelmingly deciduous, some non-succulent ever-greens are to be found virtually throughout. The dom-inant leaf size is microphyll sensu Raunkiaer (Thomas-son 1977). Climbers, some thick-stemmed, can becommon, but vascular epiphytes are rare (mostly or-chids). On limestone particularly there are distinctiveherbs that die back in the dry season, e.g. in Amor-phophallus, Begonia, Carlophyton, Colletogyne andDorstenia. Even the leaves of the giant herb Enseteperrieri die back at that time.

Bottle-trees and other kinds of succulent arewidespread, providing in fact a complete overlap atthe generic level with the thicket (Table 3 below). Thetallest of the Adansonia species, A. grandidieri whichreaches 30 m, occurs over large areas, and is some-times left when the rest of the forest is cleared. In thesouthern forests, there are two species of the highlydistinctive Didiereaceae, otherwise confined to thesemi-deciduous thicket.

Deciduous forests extend southward into the thick-et area, forming gallery forests along rivers andstreams; Tamarindus indica is often common, and maybe native.

Semi-deciduous thickets

This is the most distinctively Madagascan of all vege-tation-types on the island (Figs. 8 & 9). It is normally2–6 m tall, occasionally 10 m, and reduced to <1 m onthe wind-blasted Cap Ste Marie (southern tip of the is-land). It is characterized by impenetrability, resultingin general from the close spacing of the woody plantsand in particular from the presence of many shrubsand small trees with closely spaced branches and inmost cases very small leaves, and by the lack of a herblayer, whether of grasses or forbs. The denselybranched shrubs are discussed in detail toward the endof this article (p. 146). I insist on the term ‘semi-decid-uous thicket’ rather than ‘deciduous thicket’ used bysome recent authors, because the community containsan appreciably higher density of evergreens (whetherwith succulent or scleromorphic leaves) than the de-ciduous forest.

There is a huge variety of succulents ranging fromleaf-succulents such as Aloe and Kalanchoe spp. viatall ‘bottle trees’ such as Adansonia spp. to vines with



Fig. 7. The crown of an isolated ‘coniferoid’ pandan.

very swollen stems and/or roots such as Adenia andCyphostemma spp. Particularly characteristic are thetree Euphorbia spp. with broad, much-branchedcrowns composed of relatively slender, unarmed,cylindrical succulent stems, quite different from thevarious ‘candelabra’ Euphorbia spp. of Africa withtheir fat, ridged and spiny shoots. Keraudren (1961)suggested that the stems being slimmer in the Mada-gascan species reflected the less drying conditions, butthat cannot be right as most of the shrubby Euphorbiaspp. found in very dry semi-desert or desert in Namib-ia (cf. p. 146) have relatively slender shoots. The vari-ety in the form of the shoots and trunk bark of Mada-gascan Euphorbia spp. has been documented by Hum-bert (1927), Leandri (1966) and Rabesandratana(1984). It is important to emphasize that the succu-lents of the thicket occur on plain flat or sloping

ground, not just on sites with fissured rock. That is abig contrast with the large variety of succulents foundunder 250–550 mm yr–1 in northern Namibia; forbeautiful accurate paintings of many succulents in situon rocky hills in that area see Craven & Marais(1992).

Most characteristic of the semi-deciduous thicketare the Didiereaceae, composed of four genera andeleven species in Madagascar, and regarded as endemicuntil 2000 when three genera of tall-growing Portula-caceae in S, C and NE Africa were transferred; forconvenience I follow this usage, although Stevens(2002) has absorbed the Didiereaceae in this widersense into his Portulacaceae. Two species transgressinto deciduous forest, two species commonly reach aheight of only 1.5–2.0 m, and two have muchbranched crowns, but they are essentially a family of

134 P.J. Grubb


Table 3. Comparison of the genera present as succulents or as related specialized life-forms in three semi-arid areas. Genera in underlined type are represent-ed in only one of the three areas.

Thicket in S & SW Madagascar Thicket in NE Kenya Woodland and rock outcrops in N Namibia1

Emergents with very spiny non-succulent stemsAlluaudia, Alluaudiopsis, Decaryia, Didierea None, but non-emergent, non-spiny None, but non-emergent, non-spiny ‘Didiereaceae’

‘Didiereaceae’ present (Calyptrotheca) present at semi-desert edge to north and south (Ceraria)

Trees with succulent ultimate shootsEuphorbia (shoots slender, unarmed) Euphorbia (‘candelabra’ type; shoots thick, E. tirucalli (shoots slender; unarmed);

ridged, spiny) E. virosa with thick ridged spiny stems to 2 m

Bottle-treesAdansonia, Commiphora, Cyphostemma, Adansonia, Cyphostemma, Delonix, Moringa, Adansonia, Cyphostemma, Moringa, Delonix, Moringa, Operculicarya decaryi, Sesamothamnus, Sterculia Pachypodium, Sesamothamnus, SterculiaPachypodium2

‘Marginal bottle trees’Givotia, Gyrocarpus, Jatropha, Givotia, Gyrocarpus, Jatropha GyrocarpusOperculicarya hyphaenoides, Uncarina

Stem succulents <2 m tallAdenia, Euphorbia, Sarcostemma, Senecio3, Adenia, Euphorbia, Sarcostemma, Senecio3 Duvalia, Euphorbia, Hoodia, Sarcostemma,Stapelianthus Senecio3, Stapelia

Leaf succulentsAloe, Kalanchoe, Xerosicyos4 Aloe, Kalanchoe, Sansevieria Aloe, Crassula, Kalanchoe, Sansevieria

Root or rhizome succulentsAdenia, Ceropegia, Corallocarpus, Adenia, Cephalopentandra, Ceropegia, Adenia, Coccinia, Corallocarpus, Cyphostemma,Cyphostemma, Dioscorea, Dolichos, Coccinia, Corallocarpus, Cyclantheropsis, Dioscorea, Kedrostis, Momordica, TrochomeriaKarimbolea, Kedrostis, Odosicyos, Cyphostemma, Dioscorea, Dolichos,Peponium, Seyrigia, Toxocarpus, Gerrardanthus, Kedrostis, Momordica,Trochomeriopsis Peponium, Pyrenacantha, Trochomeria

1 Districts ETO, GR, OTJ, OU and OVA of Merxmüller (1966–1970).2 Uncarina, confamilial with Sesamothamnus, comes in next group.3 Formerly separated as Kleinia or Notonia.4 Within Madagascar Sansevieria, as traditionally defined, is native only in the NW; the genus is included in Dracaena by Stevens (2002).

tall, minimally-branched plants which emerge abovemost other plants in the community. The primaryshoots of most Madagascan species develop countlesspersistent spines, very firm and sharp (Fig. 10); theirmorphological nature is not entirely certain (Stevens2002). All along the major axes, even when thesereach 10 cm thick, there are groups of small, some-what succulent leaves at the bases of the spines; theseleaves fall in the driest periods. The major axes are notlike those of Cactaceae; they have a central core ofxylem which is either solid (e.g. Alluaudia procera; P.J.Grubb, pers. observ.; see also Rauh 1995, p. 59) orcomposed of plates of tissue alternating with air spaces(e.g. Didierea madagascariensis in Cliché 79 of Koech-lin et al. 1974; Alluaudia adscendens on p. 70 of Rauh1998). The bark is thick, say 2.0–2.5 cm thick on astem of diameter 10 cm (P.J. Grubb, pers. observ. onAlluaudia comosa and A. dumosa; Rauh 1995, p. 59

for A. procera, and Rauh 1998, pp. 54 & 70 for A. ad-scendens and Didierea madagascariensis). Functional-ly these Didiereacae are most closely related to theFouquieriaceae, best known through the ocotillo(Fouquieria splendens), but composed of elevenspecies in Mexico and the USA with varying heightsand degrees of trunk development (Hendrikson 1972).I have traced no published description of the root sys-tems of Didiereaceae, but that of Fouquieria splendensis shallow and wide (Cannon 1911); within the Mada-gascan thicket Commiphora monstruosa and Oper-culicarya spp. are shallow-rooted (Koechlin et al.1974, p. 269). While the Didiereaceae often dominatethe appearance of the thicket, it is not clear how farthe plants effect ecological dominance over plants ofother growth-forms, e.g. by root competition.

Climbers are common, and appear to be especiallyabundant alongside tracks. Among the commonest



Fig. 8. A path through tall semi-deciduous thicket at Berenty Private Re-serve; Alluaudia procera (Didiereaceae) is abundant. Note dense layer ofshrubs and small trees to the left and right of the path.

Fig. 9. Short semi-deciduous thicket on the Mahafaly limestone plateauwith emergent Alluaudia comosa (Didiereaceae).

climbers Xerosicyos danguyi (Cucurbitaceae) hastough thickly succulent round leaves 4–5 cm in diame-ter, and Folotsia grandiflora (Apocynaceae) has succu-lent stems up to 2.5 cm thick. Climbers with thick suc-culent upright trunks (e.g. Cyphostemma corniferum)or boulder-like stems (Adenia firingalavensis), occurespecially at rocky sites, where others have swollenroots (species in the same genera plus Dioscorea,Dolichos and Toxocarpus). One of the most spectacu-lar is Odosicyos bosseri Keraudren (Cucurbitaceae)with a swollen root some 50 cm across, the top ofwhich may be level with the soil surface, emergent orbelow ground (Rauh 1995, p. 117).

Epiphytes are, in general, rare. Ground-dwellingpoikilohydres are also rare, though Xerophyta da-sylirioides occurs on some rocky sites, and the mat-forming Selaginella nivea can be common on the soilwhere not trampled by people or cattle.

There is appreciable variation in structure andphysiognomy of the stand, but it is convenient to em-phasize two main facies. On gneiss and on redPliocene sands in the less dry part (400–600 mm yr–1),which lies to the south-east, toward Tolanaro, thecanopy is generally c. 6 m tall (Fig. 8); emergentspecies of Alluaudia (Didiereaceae) dominate the phys-iognomy, stem-succulent Euphorbia (especially E. pla-giantha and E. stenoclada) are also abundant, andthere is a sprinkling of bottle-trees represented byAdansonia za, Moringa drouardii and an occasionalPachypodium, a few tall aloes (especially Aloevaombe), and various leaf succulents and evergreensamidst a wide array of shrubs and climbers. There arealso what I call ‘marginal bottle trees’, listed as bottletrees by Koechlin et al. (1974) but hardly convincing(notably Gyrocarpus americanus), and a few tall de-ciduous trees that transgress from the deciduous forestand have no obvious morphological adaptations ex-

cept perhaps slightly smaller leaves or leaflets thanmost deciduous forest trees (e.g. Cedrelopsis grevei,Rutaceae, until recently Ptaeroxylaceae).

In contrast, the thicket on the flat-bedded limestoneof the low, undulating Mahafaly Plateau in the some-what drier area (c. 300–400 mm yr–1) on the south-western side of the island, inland from Toliara, ismostly only c. 2 m tall (Fig. 9). From the air it appearsthat relatively undisturbed thicket remains in a longnorth-south belt. For most of us our detailed impres-sion is from the road from Toliara to Isalo, which bi-sects the area. It is notable for its lack of the type ofDidiereacaeae described so far; instead Alluaudia co-mosa, a species with a densely branched, roundedcrown, and spines much longer than the leaves, ischaracteristic though found at low density and onlylocally (Fig. 9). Tall succulent-stemmed Euphorbia(wholly or mainly E. onoclada) and tall Aloe are un-common. The occasional bottle trees are of a differentAdansonia (A. rubristipa, formerly A. fony), and ofDelonix floribunda (wrongly named D. ‘adanso-nioides’ in the literature; cf. Du Puy 2001). One majormarginal bottle tree, Jatropha mahafaliensis, is locallyvery common.

As shown in Table 3, the genera of succulent or re-lated plants overlap hugely with those of the thicketfound in the Somalia-Masai region of White (1983), apoint emphasized by Thomasson & Thomasson(1991). More surprising is the huge overlap in genericcontent between the succulent plants in the Madagas-can thicket and those in the totally different-lookinglandscape (woodland plus succulents on rock out-crops) found under 250–550 mm yr–1 in northernNamibia, also shown in Table 3.

Among the non-succulent shrubs and small trees,the overlap between the thicket of southern Madagas-car and that of north-eastern Kenya is much lessstrong. Considering the two areas at least 93 generaare involved (Appendices 1 and 2). Only 22 are incommon, 47 are found only in the Madagsacan thick-et and 24 only in the Kenyan. Also striking is the factthat certain genera are hugely more speciose in onearea than the other. Most notably Croton has 26species in the Madagascan thicket and only four in theKenyan, while Commiphora has only eight species inthe Madagascan and 37 in the Kenyan. In terms ofecological dominance, the big differences between thetwo areas concern Acacia and Commiphora. In Kenyathe dominance by these genera is so marked that thecommunity is commonly referred to as the Acacia-Commiphora bushland or thicket. In the Madagascanthicket Acacia is not at all obvious, while Commipho-ra is abundant but not as dominant as it appears to bein the Kenyan thicket.

136 P.J. Grubb


Fig. 10. A shoot of Alluaudia procera, showing the short shoots developedas firm, sharp spines, and leaves similar in length to the spines.

then; they can endure much longer dry seasons as aroutine. Once perennial grasses are dominant, they notonly encourage fire which cuts back young woodyplants (and selects for resprouting ability) but also ex-acerbate the water shortage in the dry season by their‘wasteful’ usage in the wet season (Walter 1971;Knoop & Walker 1985).

Abundance of evergreen trees and shrubs in the thicket

Approximately 17% (33 out of 189) of species of non-succulent shrubs and small trees are evergreen (Ap-pendix 1). Judging by the account of Koechlin et al.(1974), the proportion is much higher than in the de-ciduous forest to the north, but I cannot make a pre-cise comparison. There is certainly a contrast with themuch lower proportion in the deciduous thicket of NEKenya: 8.2% (12/147; Appendix 2). Several of the ev-ergreens in the Madagascan thicket are quite local andor uncommon, but others are widespread and locallyabundant (Flore).

The evergreen leaves vary greatly in size, texture andpresentation, as is nicely shown by three common ever-green tree species (4–6 m tall) in the thicket on BerentyPrivate Reserve near Amboasary in southern Madagas-car (about 60 km WSW of Tolanaro): Salvadora angus-tifolia (Salvadoraceae, Brassicales), Maerua filiformis(Brassicaceae, Capparoideae) and Mundulea stenophyl-la (Fabaceae: Faboideae). The leaves of the Salvadoraand Maerua are strikingly narrow compared with thoseof S. persica and M. schinzii, which are so widespreadat rocky sites in Namibia (photos and maps in van Wyk& van Wyk 1997). Indeed, M. filiformis has cylindricphyllodes, the first leaves on a shoot being diminutivebut showing the ovatiform blade that is usual in thegenus (Koechlin et al. 1974, p. 273). The leaves of theSalvadora (which are slightly succulent) are pendent,

Why thicket and not grassy woodland in the driest area?

Koechlin (1972) provided the beginnings of an answer.He suggested that thicket is favoured by (a) the signifi-cant chance of rain all through the year, and (b) thenotably high relative humidity all the year. The meanat 1000 hours is c. 60% (Koechlin et al. 1974, p. 244);in contrast, in the Sahel and Kalahari, it sinks to muchlower values in the rainless months (Müller 1982).White (1983, pp. 47–48) pointed out that the onlyother area in Africa-cum-Madagascar to have thicketas the climatically determined vegetation-type is his‘Somalia-Masai’ region which has two wet seasonseach year rather than a significant chance of rain allthrough the year. That area is a little drier than thesouth and south-west of Madagascar (200–500 mm yr–1

rather than 300–600 mm yr–1), and includes semi-desert in its driest part. Other areas with the samerange of total annual rainfall (250–550 mm yr–1), bothnorth and south of the Equator, have a single short,well-defined wet season and 3–4 months with virtuallyno chance of rain. In these regions some kind of grassywoodland is the climatically determined vegetation-type, and thicket is confined to rocky hills, where mostgrasses cannot flourish. In Fig. 11 two typical climatediagrams for the Madagascan thicket are comparedwith one for the Kenyan thicket and one for the wood-land area of northern Namibia.

Neither Koechlin et al. (1974) nor White (1983)provided an explicit mechanism, but I suggest the fol-lowing. Firstly, the trees exacerbate the dryness of anydry spell by extracting water from the topsoil, and sec-ondly the treelets and shrubs (like grasses) are general-ly shallower-rooted than the trees and more dependenton topsoil water. By hypothesis they cannot survivevery long totally rainless spells, especially if low hu-midities prevail. In contrast, grasses have the ability todie back in the dry season and need almost no water



Fig. 11. Climate diagrams for Tsihombe (about 150 km WSW of Tolanaro, cf. Fig. 1) and Toliara (Tulear) in the south and south-west of Madagascar, respec-tively, Garissa in the thicket region of north-eastern Kenya, and Tsumeb in northern Namibia (from Walter & Lieth 1960–1964).

and thus reminiscent of a narrow-leaved eucalypt; thoseof the Maerua are held at various angles from 45° up to45° down, and are thus reminiscent of Acacia aneurawhich covers so much of central Australia. The leafletsof the Mundulea are parallel-sided, fibrous and mostlysub-erect, reminiscent of various Myrtaceae in the scle-rophyll vegetation of Australia.

How are so many evergreens able to flourish in thethicket despite the mean annual rainfall being muchlower than in the more thoroughly deciduous forest? –White (1983) suggested that the responsible factorsare the same as favour the development of thicketrather than woodland – the chance of rainfall beingspread through the year, and the high relative humidi-ty – and I agree.

The same prominence of evergreens in an otherwisedeciduous community, coupled with a finite chance of rain throughout the year and a mean of about 400 mm yr–1, is seen on the Peninsula de Paraguaná onthe northern coast of Venezuela (c. 12°N) for which apreliminary description is given by Matteuci (1987).There we find several evergreen species of Capparis,and one each of Guaiacum (Zygophyllaceae) andJacquinia (Theophrastaceae). Both in south-westernMadagascar, and on the north-western coast ofVenezuela, the evergreen trees grow on ‘zonal’ soils,and they are not concentrated – as are species ofMaerua and Salvadora in northern and central Namib-ia – along water courses or among rocks or on land re-ceiving water that drains off large rocks.

It is notable than in the thicket of the lowlands ofthe Masai-Somalian region, where the humidity is notkept especially high by a maritime influence, ever-greens are generally lacking; they become prominentonly in the mountains where evaporative loss is re-duced (White 1983). The prominence of evergreens onthe north-western coast of Venezuela near sea levelmay reflect in part a strong cooling effect by the north-east trade winds; severely wind-pruned trees are acharacteristic sight on exposed shores.

It is also possible that direct uptake of water by theshoots is involved. Koechlin et al. (1974) andRabesandratana (1984) emphasized the abundance ofdew-formation in the spiny thicket, a clear result ofthe prevailing high humidity. Koechlin et al. (1974, p.276) mention specifically Croton and Grewia, as re-covering turgidity after uptake from dew; both haverather soft leaves and are deciduous. Diaz (1999) hasshown unequivocally for Croton heliaster on thenorth-western coast of Venezuela that there is biologi-cally significant direct uptake of water by leafy shootsfrom very light rains there. Whether the evergreens,which are likely to have less permeable cuticles, cantake up water from dew is another question.

Abundance of succulents in the thicket

My hypothesis is that the factors which have led to theevolution of thicket rather than grassy woodland, andallow evergreens to be relatively abundant, also ac-count for the abundance of succulents. However, it isnecessary to consider this idea critically in relation toWalter’s interpretation of the conditions favouring anabundance and/or variety of succulents, and the needfor its correction. Although I criticize his theory, I ac-cept his use of the term ‘desert’ for areas where eitherthere is no perennial plant present (‘extreme desert’) or(more commonly) the perennial vegetation is restrictedto sites with exceptionally favourable water supply(typically dry river beds), and his use of ‘semi-desert’for vegetation with a diffuse cover of perennial woodyplants or grasses over the landscape, even it if amountsto only 1–5%.

The parts of the world to which Walter (1964,1971) drew attention to as having abundant and/orvaried succulents all have either a significant chance ofrainfall spread through the whole year, e.g. much ofsouth-western USA, the northern coast of South Amer-ica, and north-western South Africa, or two rainy sea-sons a year, e.g. the ‘caatinga’ of NE Brazil, part ofnorth-eastern Tanzania and the island of Socotra offthe Horn of Africa. He noted that, contrary to the ex-pectation of many lay people and beginning studentswho see the cactus as the ultimate xerophyte, succu-lents are generally poorly developed in deserts and thedriest semi-deserts. They are absent from the desertand semi-desert communities of almost the whole ofthe largest dry area on earth today, the Sahara. His ex-planation (Walter 1971, p. 318–319) was based on theproperties of the juvenile; he emphasized that whilethe adult succulents can have enormous water storesrelative to their mass, and lose water very slowly dur-ing prolonged drought, that is not true of seedlings ofthe same species with their much higher surface-to-volume ratios. Moreover, the very slow growth of theseedlings keeps them vulnerable for a long time. Forthat reason supposedly they cannot regenerate wherethe annual amount of rainfall is always very small, orthere are several absolutely dry months.

Observations in Namibia and Chile cause us to re-ject this approach. In Namibia succulent Euphorbiaspp. with shoots of the order of 1 cm diameter formmound-like shrubs 2–4 m across which dominatesemi-desert (with c. 125–150 mm yr–1) over huge areasof gently undulating more-or-less flat-bedded rockycountry along the eastern side of the Namib Desert (E. damarana Leach in the north, and E. gregariaMarloth in the south; Craven & Marais 1992). Anoth-er species (E. gummifera Boiss.), found on sand receiv-ing water running off outcropping rocks, forms some

138 P.J. Grubb


of the last stands of perennial vegetation against theextreme desert (P.J. Grubb & P.J. Carrick, pers. ob-serv.). Yet another species (E. virosa Willd. with thick,ridged, spiny stems) is almost the only species in the‘accidental’ vegetation (very sparse and composed ofhaphazardly scattered individuals) on the steep coarse-ly rocky slopes of the extremely dry Fish RiverCanyon. It is also true that cacti are found in very dryplaces in the Atacama Desert (Gulmon et al. 1979;Rundel et al. 1991). I should emphasize that in thecase of neither the Namib nor the Atacama am I con-cerned with the various plants enabled to live by thefog rolling from the sea, near to the coast. I am con-cerned with succulents living outside the influence ofthe fog.

Also important is the fact that most of northernNamibia has a negligible chance of rain for 3–4 monthsevery year (Walter & Lieth 1960–1964), and yet therewe find succulent representatives of at least ten families(Apocynaceae, Asphodelaceae, Asteraceae, Crassu-laceae, Euphorbiaceae, Malvaceae, Moringaceae, Ped-aliaceae, Ruscaceae and Vitaceae; cf. Table 3).

A different perspective on the regeneration of suc-culents is needed, and is provided by Jordan & Nobel(1979). In a laboratory experiment with very youngseedlings of Agave deserti they determined the impactsof the amount of water given for establishment, andthe severity of a subsequent dry period. Using a mathe-matical model to relate the water supply in the lab tometeorological data of a conventional kind, they fore-cast that this species could not have established newindividuals at their field site in the last 16 years. In factthe youngest plants were 17 years old. The generalpoint is that establishment of such plants occurs onlyrarely, on the order of every 10–100 years, and the oc-currence of exceptionally wet years which make estab-lishment of seedlings possible is the single most impor-tant feature of the long-term rainfall regime for theplants concerned, not the distribution of rainfall in anaverage year.

Although the local dominance of a few species ofsucculent in fog-free deserts in Namibia and Chileshows that they can live under extremely low totalrainfalls, and the variety of succulents in northernNamibia with an absolute dry season of 3–4 monthsshows that they can tolerate long dry periods on a reg-ular basis, it is reasonable to suppose – along withWalter (1971) – that both a higher total rainfall (up toa point) and a wider spread of rainfall through theyear will favour succulents. In that case, the latter fac-tors are likely to constitute the chief explanation forthe diversity as well as abundance of succulents in thesemi-deciduous thicket of Madagascar, and their oc-currence in the main vegetation-type rather than beingrestricted to rocky sites.

White (1983) emphasized that the density of succu-lents is not as high in the thicket of the Somali-Masairegion as in that of Madagascar, and that some of thespecies are clearly more abundant where the rainfall ishigher. Perhaps both a lower rainfall in general in theSomali-Masai thicket, and a lower mean relative hu-midity, are the key factors.

Physical defences against herbivores

Vegetation-types other than the semi-deciduousthicket

I deal here particularly with spines and with verydensely and finely branched shrubs or small trees. Myparticular interest is to look for plants which were de-fended against the now-extinct elephant birds, andwhich might be analogues for ‘divaricating shrubs’and other specialized plant types seen in New Zealandwhere giant flightless birds (moas) browsed until a fewhundred years ago (Atkinson & Greenwood 1989).Divaricating shrubs have notably wiry or rigid stems,which branch at a wide angle and are often densely in-terlaced; usually their leaves are very small relative tothose of non-divaricating relatives. In most species theshoots are not spiny; 15 families are involved.

The warm temperate and cool temperate rainforests of New Zealand contain numerous divaricatingshrubs, especially on more fertile soils and at edges,and also other unusual physiognomic types apparentlydefended against tall browsing birds. The tropical rainforests of Madagascar lack such plants completely. Inmany parts of the tropics a large proportion of thespiny plants in rain forests is made up of palms, pan-dans and tree-ferns; I suggested earlier that increasedinvestment in defence by such plants can be under-stood in terms of the one shoot apex not being replace-able (Grubb 1992). In Madagascar the palms are no-tably non-spiny. Dransfield & Beentje (1995) drew at-tention to this point, and to the contrast with the Sey-chelles where many rain forest palms are spiny; theyhypothesized that the key difference between Mada-gascar and the Seychelles was the lack of herbivorousgiant tortoises on the former. In contrast to the palms,the pandans in the Madagascan rain forest all havespiny leaf margins. Eleven of the 40 tree-fern species(of which many are found only in montane forest orhigh-altitude thicket) are known to have spines on thetrunks (3) or petioles (7) or both (1), and the figuremay be higher as the trunks and petioles of severalspecies are unknown (Flore); the percentage of specieswith spines may be very similar to that in Malesia(36%; Grubb 1992). There are a few shade-tolerantdicots with spiny leaves, e.g. some Plagioscyphus spp.



(Sapindaceae), for which no rationale is apparent (cf. Grubb 1992).

The position in the deciduous forests of Madagas-car is also as in the tropics generally. That is to say,apart from a few palms and pandans, spines are al-most confined to evergreen dicots which clearly sufferan increased risk of being eaten while the deciduousplants are leafless. Some such as Diospyros aculeataand Ludia dracaenoides have a spine at each leaf apex,while others such as Euonymopsis humbertii and E.longipes (Celastraceae) and Rinorea spinosa (Vio-laceae) have leaves like the European holly. Just as inthe case of Ilex in the Northern Hemisphere (Grubb1992), Euonymopsis and Rinorea are represented inwholly evergreen forests by non-spiny species (E. acu-tifolia and E. obcuneata, and 14 Rinorea species). InMadagascan deciduous forest there also a few spinystem-succulents, notably Pachypodium rutenber-gianum, and on rocks within the forest Euphorbiaankarensis and E. pachypodioides.

Semi-deciduous thicket

It is remarkable that so very few of the many succu-lents in the thicket are spiny. All of the abundant tallEuphorbia spp. with succulent stems bar E. stenocla-da, all the convincing bottle trees bar Pachypodium(Adansonia, Combretum, Delonix, Moringa, Opercu-laria decaryi), and all the marginal bottle trees (Givo-tia, Gyrocarpus, Jatropha, Opercularia hyphaenoides,Uncarina) lack spines. Similarly the leaf-succulentKalanchoe spp., the low-growing stem-succulentsSenecio (Kleinia) and Sarcostemma, and the manyclimbers with succulent leaves, stems or roots all lackspines. Just a very few lower-growing Euphorbia spp.carry spines, as do the Aloe spp. (well illustrated byRauh 1995–1998). Perhaps the few stem and leaf suc-culents with spines are particularly attractive as a re-sult of being especially fat, water-rich and lacking instrong defences of any other sort; alternatively theymay simply carry features which they evolved early on,and still show in Africa as well as Madagascar.

I suggest that the spines of the Didiereaceae are pri-marily protective of the leaves, not the stems. The evi-dence is that the length of the spines is related to thelength of the leaves (Fig. 12). At one extreme there isDidierea madagascariensis which has long leaves andspines, while at the other there is Alluaudia dumosawhich has leaves only in the first year of the life of astem and is thereafter leafless with very short spines ornone at all. In between we have Alluaudiopsis fihere-nensis with middle-length leaves and spines. Thespecies with leaves in the shorter range (10–20 mmlong) show a variety of relative lengths of the spines,but three of the eight species fall close to the line de-

fined by the species with longest, middle-sized and noleaves. My hypothesis is consistent with the fact thatthere are no spines on the tall stem-succulent Euphor-bia spp., only one of which (E. antso) produces anyleaves at all, and they soon fall (good photos in Rauh1998, pp. 44 & 46). As 9/11 of the Didiereaceae in theMadagascan thicket reach mature heights of 4–6 m ormore (Rauh 1963), it seems likely that the protectivespines have evolved as a response to lemurs ratherthan elephant birds. The situation among the elevenspecies of the Fouquieriaceae is quite different; thedata of Hendrickson (1972) show no correlation be-tween spine length and leaf length.

The incidence of spines on non-succulent shrubsand small trees in the Madagascan semi-deciduousthicket is modest: 12% of all species (22/189).Thomasson (1975), using data from substantial as-semblages of species in each of three kinds of thicket(on white sand, red sand and limestone) found a verysimilar mean (13%). The incidence is much higher inthe thicket of NE Kenya: 31% (46/147), and the dif-ference is statistically significant (chi-square contin-gency table: χ2 = 18.6, P < 0.001). The difference be-tween Madagascan and Kenyan thickets is shownclearly within particular taxa: 1/8 vs. 19/37 spiny inCommiphora, 0/28 vs. 5/12 in Euphorbiaceae, and3/17 vs. 12/13 in Fabaceae-Mimosoideae. I return tothis difference in spininess shortly.

140 P.J. Grubb


Fig. 12. The relationship between mean spine length and mean leaf lengthon short shoots in the eleven Madagascan species of Didiereaceae, based onthe data of Rauh (1963); one species (Alluaudia dumosa) has no leaves onshort shoots.

The thickets of the two areas are alike in thatamong the non-succulent shrubs and small trees thereis a very high proportion of species with very smallleaves (leptophylls and nanophylls): 62% (118/189) inMadagascar, and 53% (78/147) in Kenya (details inAppendices 1 and 2). For the Madagascan thicket,leaving aside the genera Acacia and Dichrostachys,which have leptophylls or nanophylls under any cli-mate, and the closely related endemic genus Alantsilo-dendron, which always has leptophylls, the value isstill high at 59% (100/170). The comparable figure forthe Kenyan thicket is 49% (66/134). Thomasson(1975) found a mean of 56% for all shrubs and smalltrees in SW Madagascar, and no significant differencebetween thickets on white sand, red sand and lime-stone.

The possession of especially small leaves is associat-ed with the development of impenetrable bushycrowns composed of densely branching shoots. Onmany species most of the leaves are borne on veryshort side shoots, and the leaf-clumps are scattered allthrough the crown. White (1983) noted the abundanceof ‘small narrow leaves’, but in fact most are obovate(Fig. 13) – a good shape for minimizing mutual shad-ing in mini-rosettes of semi-erect lvs produced on shortlateral shoots. Because, as stated in one of Corner’srules, twig diameter is closely related to mean area oflamina on the leaves borne (Corner 1949), it is possi-ble to argue that selection has worked primarily onleaf size or on density of branching (dense branchingneeding slender twigs), or indeed on a combination ofsmall leaf size and slender twigs.

If we compare leaf (or leaflet) sizes for the thicket-forming species in Madagascar with congeners in apart of Africa with a similar total annual rainfall and astriking similarity in genera present but woodland in-stead of thicket, i.e. northern Namibia, we see just

how distinct the two areas are. In all of eight genera,for which comparisons may be made (see Appendices1 and 3), the Madagascan species have smaller leaves.

The classical physiological explanation for havingespecially small leaves would be the advantage of cool-ing more easily (the still air layer on smaller leavesbeing thinner) and of reducing the problem of supply-ing water to the lamina edge from the base duringtimes of extreme water shortage. Some support for thisapproach comes from considering species within gen-era along a rainfall gradient in Namibia, comparingthose from rocky hills at the edge of the true desert (atabout 100–125 mm yr–1) with those in the woodlandlandscape (at 250–550 mm yr–1). Of the seven Com-miphora species found on rocky sites fringing thedesert four have notably tiny leaflets: C. kraeuseliana3.6 mm2, C.dinteri 12 mm2, C. virgata 35 mm2 and C. saxicola 160 mm2 (cf. 170–1500 mm2 in tenwoodland species; Appendix 3). Of the eight Grewiaspp. in northern Namibia the only one going intosemi-desert (G. tenax) has the smallest leaves (100 vs.450–2800 mm2; Appendix 3).

It is hard to consider extreme lack of water as theexplanation of very small leaves in S and SW Mada-gascar in view of the perspective put forward abovefor abundance of evergreens and succulents. Bearing inmind the fact that in New Zealand the evolution ofvery small leaves has manifestly occurred in relation tosome factor other than water shortage, most plausiblyherbivory, I suggest that in the Madagascan thicketsmall leaf size has evolved, along with dense branch-ing, as a defence against ground-based herbivores,specifically elephant birds. Only a minority have di-varicate branching in the strict sense (a kind of pseu-do-dichotomy), e.g. Terminalia divaricata and severalspecies of Croton, most notably C. ambovombensisand C menarandrae, and only a small minority have



Fig. 13. Examples of leaf form on short shoots of shrub and small-tree species of the semi-deciduous thicket: (a) Commiphora simplicifolia, (b) Diospyroshumbertiana, and (c) Terminalia divaricata (based on drawings in Flore de Madagascar et Comores, Humbert et al. 1936–2002). The Scale = 10 mm.

rainfall in the tropics during the Cretaceous beforeMadagascar and Africa separated (Vakhrameev 1991).We have only to assume that all the relevant generaevolved early on. However, the received view is that bythe early Tertiary wet climates were strongly dominant(Wolfe 1985). Where did the present-day dry-climategenera live on Madagascar then?

The flora and vegetation of Madagascar are notonly of exceptional interest in themselves, but throwup serious problems for understanding of plant life onAfrica and in the world as a whole.

Acknowledgements. I am deeply indebted to John Dransfieldfor organizing our visit to so many parts of Madagascar, andto George Schatz for frank and constructive criticism of adraft of this paper. Phil Hulme, Johannes Kollmann, JeremyMidgley, Mike Swaine and Ian Turner also provided veryhelpful critiques. I thank very warmly Johannes Kollmannfor making the figures.

References

Abraham JP, Benja R, Randrianasolo M, Ganzhorn JU,Jeannoda V & Leigh EG (1996) Tree diversity on smallplots in Madagascar: a preliminary review. Revue d’Ecologie (Terre & Vie) 51: 93–116.

Agnew ADQ (1974) Upland Kenya Wild Flowers. OxfordUniversity Press, London.

Atkinson IAE & Greenwood RM (1989) Relationships be-tween moas and plants. New Zealand Journal of EcologySuppl. 12: 67–101.

Brenon P (1972) The geology of Madagascar. Biogeographyand Ecology of Madagascar (eds. R Battistini & GRichard-Vindard), pp. 27–86. Junk, The Hague.

Burney DA (1997) Theories and facts regarding Holoceneenvironmental change before and after human coloniza-tion. Natural Change and Human Impact in Madagascar(eds. SM Goodman & BD Patterson), pp. 93–123. Smith-sonian Institution Press, Washington.

Burns RM & Honkala BH (1990) Silvics of North America.Vol. 1. Conifers. Agricultural Handbook 654. Forest Ser-vice, US Department of Agriculture, Washington DC.

Cain SA, Castro GM de O, Pires JM & Da Silva NT (1956)Applications of some phytosociological techniques toBrazilian rain forest. American Journal of Botany 43:911–941.

Cannon WA (1911) The Root Habit of Desert Plants. Carne-gie Institution, Washington.

Capuron R (1969) Révision des Sapindacées de Madagascaret des Comores. Mémoires du Museum National D’His-toire Naturelle, Série B, Botanique 19: 1–189.

Coomes DA & Grubb PJ (1996) Amazonian caatinga andrelated communities at La Esmeralda, Venezuela: foreststructure, physiognomy and floristics, and control by soilfactors. Vegetatio 122: 167–191.

Corner EJH (1949) The Durian Theory or the origin of themodern tree. Annals of Botany New Series 13: 367–414.

zig-zag growth of the primary shoots though this habit– when combined with long spines – can appear veryprotective of the leaves, e.g. in Ximenia perrieri.

The evolution of similarly bushy impenetrableshrubs in S and SW Madagascar and in the Somalia-Masai region shows that the design cannot haveevolved specifically against elephant birds. However, Isuggest that dense bushiness has evolved as a defenceagainst medium-sized ground-based herbivores,whether birds or mammals. Moreover, the distinctivefeature of the Madagascan shrubs, relative those of theSomalia-Masai region is the low density of spines,which in turn parallels the nature of the great majorityof divaricating shrubs in New Zealand, widely be-lieved to have evolved in response to browsing bymoas. The much higher incidence of spiny plants inthe Somalia-Masai region is consistent with the de-fence of these being against mammals with softmouths rather than hard beaks.

Some outstanding puzzles

In my discussion of the Madagascan semi-deciduousthicket I emphasized the high proportion of specieswith very small leaves. Equally impressive is the verywide range in lamina area, up to two orders of magni-tude, found within genera with 2–26 species in theMadagascan thicket, or 2–37 in the Kenyan deciduousthicket (Appendices 1 & 2). Does this variation in leafsize contribute in some way to the maintenance of co-existence? – This question parallels the query over thesignificance of the huge range in nitrogen concentra-tion, and many related properties, found elsewherewithin a single functional group in vegetation rangingfrom semi-desert to rain forest (Grubb 2002).

The abundance and diversity of three types of plantin the Madagascan rain forests – palms, pandans andbamboos – is commonly attributed to the more securewater supply enjoyed by plants on Madagascar in thepast. Yet all three groups are represented in the driestvegetation-type on the island: Ravenea xerophila, Pan-danus aridus St John and an undescribed bamboo (S. Dransfield, pers. comm., March 2003). How havethese groups not managed better in Africa with its pastand present dry climates? Did the chance of producinga drought-tolerant species depend on there being alarge reservoir of biotypes in ‘safe’ moist-climateareas?

Finally, there is a huge commonality at the genericlevel between Madagascar and Africa in the plants ofdry-climate vegetation (Table 3, Appendices 1–3). Inone respect, this observation is readily compatiblewith the received record of vegetational history be-cause there is believed to have been a wide range of

142 P.J. Grubb


Cornet A (1974) Essai de cartographie bioclimatique àMadagascar. Notice explicative 55. ORSTOM, Paris.

Craven P & Marais C (1992) Damaraland Flora. GamsbergMacmillan, Windhoek.

Dale IR & Greenway PJ (1961) Kenya Trees and Shrubs.Buchanan’s Kenya Estates, London.

Diaz M (1999) Ecophysiology of very dry tropical foresttrees: effects of soil characteristics, rainfall and irrigation.PhD dissertation, University of Cambridge.

Donque G (1972) The climatology of Madagascar. Biogeog-raphy and Ecology in Madagascar (eds. R Battistini & GRichard-Vindard), pp. 87–144. Junk, The Hague.

Dransfield J & Beentje H (1995) The Palms of Madagascar.Royal Botanic Gardens, Kew, and the International PalmSociety, Lawrence, Kansas.

Dransfield S (2000) Woody bamboos (Gramineae-Bambu-soideae) of Madagascar. Grasses: Systematics and Evolu-tion (eds. SWL Jacobs & J Everett), pp. 43–50. CSIRO,Melbourne.

Dransfield S (2002) Sirochloa, a new bamboo genus fromMadagascar (Poaceae-Bambusoideae). Kew Bulletin 57:963–970.

Du Puy DJ (ed.) (2001) The Leguminosae of Madagascar.Royal Botanic Gardens, Kew.

Du Puy DJ, Cribb P, Bosser J, Hermans J & Hermans C(1999) The Orchids of Madagascar: an Annotated Check-list. Royal Botanic Gardens, Kew.

Exell AW, Wild H, Fernandes A, Brenan JPM, Launert E,Pope GV & Martins ES (eds.) (1960–2002) Flora Zambe-siaca. Crown Agents, London, and Royal Botanic Gar-dens, Kew.

Gentry AH (1993) Diversity and floristic composition oflowland tropical forest in Africa and South America. Bio-logical Relationships between Africa and South America(ed. P Goldblatt), pp. 500–547. Yale University Press,New Haven.

Good R (1953) The Geography of the Flowering Plants.Longmans Green, London.

Goodman SM & Patterson BD (eds.) (1997) NaturalChange and Human Impact in Madagascar. SmithsonianInstitution, Washington DC.

Grubb PJ (1977) The control of forest growth and distribu-tion on wet tropical mountains: with special reference tomineral nutrition. Annual Review of Ecology & Systemat-ics 8: 83–107.

Grubb PJ (1987) Global trends in species-richness in terres-trial vegetation: a view from the northern hemisphere.Organization of Communities Past and Present (eds. JHRGee & PS Giller) (Symposia of the British Ecological Soci-ety 27), pp. 99–118. Blackwell Scientific Publications,Oxford.

Grubb PJ (1992) A positive distrust in simplicity: lessonsfrom plant defences and from competition among plantsand among animals. Journal of Ecology 80: 585–610.

Grubb PJ (2002) Leaf form and function – towards a radicalnew approach. New Phytologist 155: 317–320.

Grubb PJ & Stevens PF (1985) The forests of the FatimaBasin and Mt Kerigomna, Papua New Guinea, with a re-view of montane and subalpine forests elsewhere inPapuasia. Department of Biogeography & Geomorpholgy

Publication BG/5. Australian National University Press,Canberra.

Guillaumet JL (1983) Forêts et fourrés de montagne àMadagascar. Candollea 38: 481–502.

Guillaumet JL (1984) The vegetation: an extraordinary di-versity. Key Environments: Madagascar (ed. A Jolly, POberle & R Albignac), pp. 27–54. Pergamon, Oxford.

Guillaumet JL & Mangenot G (1975) Aspects de la spécia-tion dans la flore malgache. Boissiera 24: 119–123.

Gulmon SL, Rundel PW, Ehleringer JR & Mooney HA(1979) Spatial relationships and competition in a Chileandesert cactus. Oecologia 44: 40–43.

Hall JB & Swaine MD (1976) Classification and ecology ofclosed canopy forest in Ghana. Journal of Ecology 64:913–952.

Hendrikson J (1972) A taxonomic revision of the Fouquieri-aceae. Aliso 7: 439–538.

Hladik A, Blanc P, Dumetz N, Jeannoda V, RabenandrianaN & Hladik CM (2000) Données sur la repartition géo-graphique du genre Ravenala et sur son rôle dans la dy-namique forestière à Madagascar. Diversité et Endémismeà Madagascar (eds. WR Lourenço & SM Goodman), pp.1–9. Mémoires de la Société de Biogéographie, Paris.

Humbert H (1927) La destruction d’une flore insulaire par lefeu. Principaux aspects de la végétation à Madagascar.Documents photographiques et notices. Mémoires del’Académie Malgache 5: 1–79.

Humbert H et al. (eds.) (1936–2002) Flore de Madagascar etdes Comores. Muséum National d’Histoire Naturelle, Paris.

Jordan PW & Nobel PS (1979) Infrequent establishment ofseedlings of Agave deserti (Agavaceae) in the northwest-ern Sonoran Desert. American Journal of Botany 66:1079–1084.

Keraudren M (1961) Quelques aspects des formationsxérophiles du sud de la Republique Malgache. Bulletin dela Société Botanique de France 108: 73–79.

Knoop WT & Walker BH (1985) Interactions of woody andherbaceous vegetation in a southern African savanna.Journal of Ecology 73: 235–253.

Koechlin J (1972) Flora and vegetation of Madagascar. Bio-geography and Ecology in Madagascar (eds. R Battistini& G Richard-Vindard), pp. 145–190. Junk, The Hague.

Koechlin J, Guillaumet JL & Morat P (1974) Flore et Végé-tation de Madagascar. Cramer, Vaduz.

Krause DW, Hartman JH & Wells NA (1997) Late Creta-ceous Vertebrates from Madagascar. Implications for bi-otic change in deep time. Natural Change and Human Im-pact in Madagascar (eds. SM Goodman & BD Patterson),pp. 93–123. Smithsonian Institution Press, Washington.

Latham RE & Rickleffs RE (1993) Continental comparisonsof temperate-zone tree species diversity. Species Diversityin Ecological Communities (eds. RE Rickleffs & DSchluter), pp. 294–314. Chicago University Press, Chicago.

Lawton RM (1978) A study of the dynamic ecology of Zam-bian vegetation. Journal of Ecology 66: 175–198.

Leandri J (1966) Observations sur l’Euphorbia onocladaDrake et sur quelques Euphorbes coralliformes malgach-es. Adansonia Nouvelle Série 6: 331–350.

Leroy JF (1976) Essais de taxonomie syncrétique. I. Etudesur les Meliaceae de Madagascar. Adansonia NouvelleSérie 16: 167–203.



Stevens PF (2002) Angiosperm Phylogeny Website. Version3, June 2002. http://www.mobot.org/MOBOT/APweb

Stone BC (1970) Morphological studies in Pandanaceae. II.The ‘coniferoid’ habit in Pandanus sect. Acanthostyla.Bulletin of the Torrey Botanical Club 97: 144–149.

Thistleton-Dyer WJ (ed.) (1902) Flora of Tropical Africa.Vol. 8. Reeve, London.

Thomasson M (1975) Essai sur la physiognomie de la végé-tation des environs de Tuléar (sud-ouest Malgache). Bul-letin du Museum National D’Histoire Naturelle, ParisSérie 3, Ecologie Générale 22: 1–27.

Thomasson M (1977) La forêt dense sclerophylle de mon-tagne du Tsiafajavona (Madagascar). Adansonia, Série 2,16: 487–492.

Thomasson M & Thomasson G (1991) Essai sur la flore dusud-ouest Malgache: originalité, affinitités et origines.Bulletin du Museum National D’Histoire Naturelle, ParisSection B, Adansonia Série 4, 13: 71–89.

Tracey JG (1982) The Vegetation of the Humid Tropical Re-gion of North Queensland. CSIRO, Melbourne.

Trapnell CG (1959) Ecological results of woodland burningexperiments in Northern Rhodesia. Journal of Ecology47: 129–168.

Turrill WB, Milne-Redhead E, Hubbard CE, Polhill RM &Beentje HJ (eds.) (1952–2002) Flora of Tropical EastAfrica. Crown Agents, London, and Balkema, Rotterdam.

Vakhrameev VA (1991) Jurassic and Cretaceous Floras andClimates of the Earth. Cambridge University Press, Cam-bridge.

Van Wyk B & van Wyk P (1997) Field Guide to the Trees ofSouthern Africa. Struik, Cape Town.

Walter H (1964) Die Vegetation der Erde in ökophysiologis-cher Betrachtung. Vol. 1. Die tropischen und subtropi-schen Zonen. Fischer, Jena.

Walter H (1971) Ecology of Tropical and Subtropical Vege-tation. Oliver & Boyd, Edinburgh.

Walter H & Lieth H (1960–1964) Klimadiagramm Weltatlas.Vols. 1 & 2. Fischer, Jena.

White F (1983) The Vegetation Map of Africa, a DescriptiveMemoir to Accompany the UNESCO/AETFAT/UNSOVegetation Map of Africa. Natural Resources Research20, UNESCO, Paris.

Whitmore TC (1985) Tropical Rain Forests of the Far East.Clarendon Press, Oxford.

Wolfe JA (1985) Distribution of major vegetational types dur-ing the Tertiary. The Carbon Cycle and Atmospheric CO2:Natural Variations Archean to Present (eds. ET Sundquist& WS Broecker), pp. 357–375. Geophysical Monograph32. American Geophysical Union, Washington DC.

Leroy JF (1978) Composition, origin and affinities of theMadagascan vascular flora. Annals of the MissouriBotanical Garden 65: 535–589.

Lind EM & Morrison MES (1974) East African Vegetation.Longman, London.

Lourenço WR & Goodman SM (eds.) (2000) Diversity andEndemism in Madagascar. Mémoires de la Société deBiogéographie, Paris.

Lourteig A (1960) Une Lythracée dioïque: Capuronia mada-gascariensis gen. nov., sp. nov. de Madagascar. ComptesRendus Hebdomadaire des Séances de l’Académie des Sci-ences, Paris 251: 1033–1034.

Lowry PP, Schatz GE & Phillipson PB (1997) The classifica-tion of natural and anthropogenic vegetation in Madagas-car. Natural Change and Human Impact in Madagascar(eds. SM Goodman & BD Patterson), pp. 93–123. Smith-sonian Institution Press, Washington.

Matteuci S (1987) The vegetation of Falcón State,Venezuela. Vegetatio 70: 67–91.

Merxmüller H (ed.) (1966–1970) Prodromus einer Floravon Südwestafrika. Cramer, Vaduz.

Müller MJ (1982) Selected Climatic Data for a Global Set ofStandard Stations for Vegetation Science. Junk, TheHague.

Palgrave KC (1988) Trees of Southern Africa. Struik, CapeTown.

Rabesandratana R (1984) Flora of the Malagasy Southwest.Key Environments: Madagascar (eds. A Jolly, P Oberle &R Albignac), pp. 55–74. Pergamon, Oxford.

Rauh W (1963) Didiereaceae. Flore de Madagascar et desComores, 121. Museum National d’Histoire Naturelle,Paris.

Rauh W (1995–1998) Succulent and Xerophytic Plants ofMadagascar. Vols. 1 & 2. Strawberry Press, Mill Valley.

Raunkiaer C (1934) The Life-forms of Plants and StatisticalPlant Geography. Clarendon Press, Oxford.

Richards PW (1996) The Tropical Rain Forest. CambridgeUniversity Press.

Rundel PW, Dillon MO, Palma B, Mooney HA, Gulmon SL& Ehleringer JR (1991) Phytogeography and ecology ofthe coastal Atacama and Peruvian Deserts. Aliso 13:1–49.

Schatz GE (2001) Generic Tree Flora of Madagascar. RoyalBotanic Gardens, Kew, and Missouri Botanical Garden, StLouis.

Schimper AFW (1903) Plant-geography upon a Physiologi-cal Basis. Clarendon, Oxford.

144 P.J. Grubb




Appendix 1. Check-list of species of non-succulent shrubs (>1 m tall) and small trees (not normally >6 m tall in thicket) in the Semi-deciduous Thicket ofsouthern and south-western Madagascar.

Within each functional group species are ordered firstly by family, then by genus in alphabetical order, and then by increasing leaf or leaflet size. Definite bottletrees and ‘marginal bottle trees’ (cf. Table 3) are excluded. Compiled from the Flore de Madagascar et des Comores, and information given by Lourteig (1960),Capuron (1969), Koechlin et al. (1974), Leroy (1976), Rabesandratana (1984), Dransfield & Beentje (1995), Du Puy (2001), Schatz (2001), Mr A. Radcliffe-Smith (Flore de Madagascar et Comores, in press; pers. comm., March 2003) for Croton, three papers quoted by Schatz (2001) for Grewia, and my own obser-vations in Herb. Kew for Cordia, Pandanus, Turraea and the remaining Grewia spp.The number after each species is my best estimate of the mean area of a leaf or leaflet blade (mm2) based on arithmetic means of usual upper and lower limitson length and breadth given in publications cited; for most species, area has been calculated as length x breadth x 0.67 (following Cain et al. 1956), but for lin-ear leaves I have used simply length x breadth, and for nearly round leaves I have used π r2 (all values rounded off to two significant figures; *, with stipularspines; **, with axillary spines or spines terminating short shoots or thorns; ***, with a spine or spines on lamina).

Evergreen species (33)1

Arecaceae, Ravenea xerophila 6200; Asteraceae, Dicoma carbonaria 92, Psiadia coarctata 670, Vernonia antandroy 28, V. sublutea var. sublutea 75; Bignoni-aceae, Phyllarthron bernerianum 220 (phyllode segment); Brassicaceae-Capparoideae, Boscia madagascariensis 650, B. longifolia 770 (phyllode), Capparissepiaria* 310, Maerua filiformis 100, Maerua humbertii 340, Thylachium sumangui 270, T. pouponii 340, T. seyrigii 450, T. monophyllum 2000; Buxaceae,Buxus madagascariensis ssp. xerophyta 110; Celastraceae, ‘Elaeodendron’ humbertii [really a new genus, fide Schatz 2001] 240 (phyllode), Euonymopsislongipes*** 2100, Salvadoropsis arenicola 330; Ebenaceae, Diospyros humbertiana 20, D. intricata 44, D. aculeata ssp. meridionalis*** 110; Fabaceae-Cae-salpinoideae, Baudouinia rouxevillei 14, Brenierea insignis 300 (cladode); Fabaceae-Papilionoideae, Mundulea stenophylla 69; Ochnaceae, O. greveanum1200; Opiliaceae, Rhopalopilia perrieri 200; Pandanaceae, Pandanus aridus*** 24000; Rhamnaceae, Scutia myrtina** 190; Salvadoraceae, Azima tetracan-tha** 300, Salvadora angustifolia 500; Sapotaceae, Capurodendron androyense 45, C. mandrarense 1200.

Mimosoid Fabaceae (17, all deciduous)A group usually leptophyllous or nanophyllous in any climate.Acacia bellula* 2, A. viguieri* 20; Albizia divaricata 3, A. masikororum, A. commiphoroides 3.7, A. atakataka** 6.7, A. tulearensis 190; Alantsilodendronglomeratum 0.3, A. pilosum 1, A. ramosum 1, A. alluaudianum 3, A. decaryanum 3.5, A. brevipes 4.8, A. humbertii 5.2, A. mahafalense 15, Dichrostachysdumetaria 0.3, D. venosa 3.1.

Other deciduous species (139)Acanthaceae, Achyrocalyx decaryi 24, Barleria brevituba* 25, B. decaryi 50, B. her 140, B. alluaudii* 250; Asteraceae, Brachylaena microphylla 600, Vernoniaswinglei 40, V. mahafaly 130, V. cloiselii 400, V. pectoralis 770, V. seyrigii 6700, V. mandrarensis 7200; Bignoniaceae, Phylloctenium decaryanum** 28,Rhigozum madagascariense 16, Stereospermum nematocarpon 160; Boraginaceae, Cordia mairei 230, C. sinensis 520, C. subcordata 2200; Burseraceae,Commiphora humbertii 7.7, C. sinuata 52, C. brevicalyx 55, C. simplicifolia** 56; C. orbicularis 190, C. lamii 310, C. marchandii 620; Celastraceae, Gym-nosporia polyacantha** 220, G. trigyna var. macrocarpa 630; Combretaceae: Terminalia ulexoides** 13, T. divaricata 36, T. disjuncta 52, T. seyrigii 160, T.subserrata 450, T. taliala 530; Ebenaceae, Diospyros manampetsae 25; D. latispathulata 140, D. saklavarum ssp. tulearanum 1700; Erythroxylaceae, Erythroxy-lum xerophilum 2.7, E. seyrigii 150, E. pervillei 200; Euphorbiaceae, Croton menarandrae 15, C. inops 22, C. kimosorum 32, C. elliottianus 40, C. sakamalien-sis 44, C. manampetsae 48, Croton new sp. 1 54, C. androiensis 55, C. aubrevilecta 65, C. adabolavensis 74, C. ambovombensis 100, C. chauvetiae 110, C.cotoneaster 110, C. geayi 110, C. peltieri 120, C. miarensis 130, C. boiteaui 190, C. salviformis 190, C. antanosiensis 200, C. tranomarensis 280, C. decaryi360, C. meridionalis 420, C. cassinoides 570, C. greveanus 580, C. leandri 790, C. mavoravina 1100, Croton new sp. 2 1900, C. barorum 2200; Fabaceae-Caesalpinioideae, Baudouinia fluggeiformis 530, Bauhinia grandididieri 31, B. xerophyta 93, B. morondavensis 180, Lemuropisum edule 13, Senna meridion-alis 12, S. leandri 23, S. bosseri 30, S. viguierella 140, S. anthoxantha 230; Tetrapterocarpon geayi 120; Fabaceae-Papilionoideae, Chadsia grevei 140, Dalber-gia xerophila 5.4, Mundulea micrantha 84, M. antanossarum 190; Lamiaceae, Capitanopsis angustifolia 70, C. cloiselii 420, Clerodendrum subtruncatum 32,C. humberti 64, C. perrieri 95, C. emirnense 290, C. boivinii 370, C. manombense 1200, Karomia angustifolia 27, K. microphylla 140, Premna humbertii 490;Loganiaceae, Strychnos decussata 450; Lythraceae, Capuronia madagascariensis** 33; Malpighiaceae, Acridocarpus excelsus ssp. parvifolius 160; Malvaceae,Grewia obdeltoidea Capuron 28, G. erythroxyloides 36, G. clavata Baker 87, G. microcyclea 93, G. mahafaliensis Capuron 100, G. tulearensis 150, G. an-droyensis Capuron 160, G. trinervata Baker 220, G. fransiscana Capuron 440, G. andramparo 480, G. grandididieri 500, G. betulifolia Boivin 620, C. humber-tii 870, C. lavanalensis 1500, G. grevei 1800, G. meridionalis 1900, Megistostegium microphyllum 34, M. nodulosum 230, M. perrieri 490; Meliaceae, Capuro-nianthus mahafaliensis 630, Neobeguea mahafaliensis 350, Turraea rhombifolia Baker 59; Olacaceae, Ximenia perrieri* 150; Polygalaceae, Polygala humbertii45, P. compressa 50, P. heteranthera 74, P. submonoica 220; Rhamnaceae, Colubrina humbertii 12, C. alluaudii 130, C. sphaerocarpa 1200, Zizyphus made-cassus* 800; Salicaceae: Calantica decaryana 160; Sapindaceae, Allophyllus cobbe ‘crataegifolius’ and ‘dissectus’ 170, Dodonaea viscosa var. linearis 230, Ery-throphysa aesculina 560; Scrophulariaceae, Androya decaryi 240; Solanaceae, Lycium acutifolium** 37, Solanum batoides** 22, S. bumeliaefolium** 120, S.tuliaraea 150, S. heinianum 240, S. betroka 280, S. croatii** 890.

1I have omitted Capparis chrysomeia as found in gallery forest rather than thicket (P. J. Grubb, pers. observ.), and Gymnosporia linearis as typical of cleared landrather than thicket (Koechlin et al. 1974).

146 P.J. Grubb


Appendix 2. Check-list of species of shrubs (>1 m tall) and small trees (not normally >6 m tall in thicket) in the Bushland and Thicket of the NorthernProvince of Kenya.

Compiled from the Flora of Tropical East Africa with help from Dale & Greenway (1961) and White (1983). Conventions as in Appendix 1. The two subspeciesof Commiphora holtziana are so different in leaf size that they have been treated as different species. The extent of evergreenness in Maerua spp. is unclearfrom the published account. This list is certainly incomplete as there is no published account of the Polygala or Solanum spp. of the area; there is no modern ac-count of the Acanthaceae present, but all those I have traced at Kew are <1 m tall.

Evergreens (12)Brassicaceae-Capparoideae, Boscia coriacea 720, Maerua sessiliflora 12, M. kaessneri 38, M. endlickii 50; Celastraceae, Elaeodendron aquifolium*** 1600;Ebenaceae, Diopyros scabra 130, D. wajirensis** 130, D. kanurii 500; Salvadoraceae, Dobera loranthifolia 1000, D. glabra 1100, Salvadora persica 840;Sapotaceae, Manilkara mochisia 510.

Mimosoid Fabaceae (13, all deciduous)Acacia tortilis ssp. spirocarpa* 0.5, A. edgeworthii* 1.3, A. reficiens* 1.9, A. bussei* 2.2, A. turnbulliana* 2.2, A. stuhlmannii* 3.1, A. horrida* 3.4, A. paolii*4.6, A. nubica* 6.9, A. condylocarpa* 61, A. mellifera* 79, Albizzia amara ssp. amara 4.0, Dichrostachys cinerea ssp. cinerea* 3.6.

Other species (118)Amaranthaceae, Sericomopsis hildebrandtii 400, S. pallida 520; Anacardiaceae, Lannea alata 24, L. greenwayi 340, L. triphylla 600, L. malifolia 1100, L. rivae2200; Apocynaceae, Strophanthus mirabilis 130, Wrightia demartiniana 350; Asteraceae, Vernonia cinerascens 75, V. wakefieldii 1840; Boraginaceae, Bourre-ria lyciacea 140, B. teitensis 300, Cordia ellenbeckii 180, C. quercifolia 660, C. longipetiolata 1100, C. crenata ssp. meridionalis 1400, C. monoica 1600; Bras-sicaceae-Capparoideae, Boscia minimifolia 67, Cadaba glandulosa 190, C. farinosa 290, C. mirabilis 400, Maerua glauca 210, M. crassifolia 220; Burseraceae,Boswellia neglecta 32, B. microphylla 34, B. rivae 52, Commiphora confusa** 12, C. holtziana ssp. microphylla 19, C. velutina 21, C. incisa** 25, C. pseu-dopaolii 42, C. oddurensis** 45, C. ellenbeckii** 52, C. gracilispina** 54, C. tenuis** 77, C. chaetocarpa** 90, C. corrugata 90, C. habessinica ssp.habessinica** 100, C. campestris ssp. campestris** 110, C. kua** 120, C. sarandensis ssp. moyaleensis 120, Commiphora sp. B 120, C. holtziana ssp.holtziana 140, C. myrrha** 140, C. schimperi** 160, C. kataf 170, C. boranensis 180, C. bruceae** 180, C. ogadensis 180, C. sennii** 190, C. rostrata**200, C. swynnertonii** 220, C. africana** 240, C. alaticaulis 240, C. longipedicellata 250, C. cyclophylla 290, C. milbraedii 400, C. danduensis** 470, C. cili-ata 520, C. edulis 980, C. erosa 1000, C. samharensis** ssp. terebintha 1200, C. unilobata 1200; Celastraceae, Gymnosporia (Maytenus) putterlickioides**750; Combretaceae, Combretum contractum 570, C. aculeatum 1000, Terminalia parvula** 14, T. polycarpa 22; Euphorbiaceae, Croton somalensis 330, C.menyarthii 640, C. pseudopulchellus 730, C. dichogamus 1400, Jatropha parvifolia* 95, J. rivae* 95, J. arguta* 100, J. oblanceolata 210, J. pelargonifolia720, J. ellenbeckii* 840, J. dichtar* 880; Fabaceae-Caesalpininioideae, Bauhinia taitensis 140, Cassia longiracemosa 96, C. baccarinii 130, C. singueana 600,C. abbreviata ssp. kassneri 630, C. ruspoli 1600, Dalbergia eremicola** 15, D. microphylla 16, D. vaccinioides 120, D. commiphoroides** 160, Lonchocarpusbussei 770, Platycelyphum voense 1200; Lamiaceae, Clerodendrum robecchii 40, C. makanjanum 650, Premna oligotricha 330, P. resinosa 630; Malpighi-aceae, Acridocarpus glaucescens 220, Caucanthus albidus 140, Triaspis erlangeri 540; Malvaceae, Grewia pedunculata 240, G. kakothamnus 250, G. ery-thraea 270, G. nematopus 270, G. penicillata 310, G. tenax 350, G. tristis 350, G. lilacina 360; G. villosa 5700; Meliaceae, Melia volkensii 430, Turraea parvi-folia 23, T. kokwaroana 680, T. barbata 800, T. cornucopia 990; Rhamnaceae, Zizyphus hamur* 46; Rubiaceae, Hymenodictyon parvifolium 920; Rutaceae,Zanthoxylum chalybeum** 1300; Sapindaceae, Bottegoa insignis 16, Haplocoelum foliolosum 150; Solanaceae, Lycium europaeum** 36; Verbenaceae, Lan-tana viburnoides 310, L. humuliformis 350; Zygophyllaceae, Balanites orbicularis** 200.

Appendix 3. Species of the dry woodland landscape of northern Namibia in the same genera as shrubs or small trees in the Madagascan semi-deciduous thicket.Compiled from Palgrave (1988); calculation of lamina area as in Appendix 1.

Bignoniaceae, Rhigozum brevispinosum 38; Burseraceae, Commiphora multijuga 170, C. oblanceolata 200, C. tenuipetiolata 310, C. pyracanthoides 350, C.merkeri 440, C. angolensis 450, C. africana 480, C. glaucescens 540, C. mollis 750, C. crenato-serrata 1500; Celastraceae, Gymnosporia senegalensis 600, G.buxifolia 670; Combretaceae, Terminalia prunioides 250, T. sericea 1500, T. brachystemma 3200; Euphorbiaceae, Croton menyarthii 670, C. gratissimus 1500;Fabaceae, Bauhinia petersiana 2300; Malvaceae, Grewia tenax 100, G. flava 450, G. retinervis 500, G. bicolor 600, G. flavescens 1400, G. schinzii 1900, G.villosa 2800; Olacaceae, Ximenia americana 300, X. caffra 600.

Species of Commiphora from rocky sites fringing full desert for comparison: C. kraeuseliana 3.6, C. dinteri 12, C. virgata 35, C. saxicola 160, C. wildii 480, C.discolor 500, C. anacardifolia 2100.