Journal of Biogeography (J. Biogeogr.) (2010) 37, 506–519 ORIGINAL ARTICLE Evidence for drought and forest declines during the recent megafaunal extinctions in Madagascar Malika Virah-Sawmy1*, Katherine J. Willis1,2 and Lindsey Gillson3 1 Oxford Long-term Ecology Laboratory, Oxford University Centre for the Environment, South Parks Road, Oxford OX1 3QY, UK, 2 Department of Biology, University of Bergen, N-5007 Bergen, Norway and 3Plant Conservation Unit, University of Cape Town, Rondebosch 7701, South Africa ABSTRACT Aim There remains some uncertainty concerning the causes of extinctions of Madagascar’s megafauna. One hypothesis is that they were caused by overhunting by humans. A second hypothesis is that their extinction was caused by both environmental change and hunting. This paper systematically addresses the second hypothesis through examination of two new pollen records from southeastern Madagascar alongside other published records across the island. Location South-eastern Madagascar. Methods We reconstructed past vegetation and fire dynamics over the past 6000 years at two sites in south-eastern Madagascar (Ste-Luce) using fossil pollen and charcoal contained in sedimentary sequences. We investigated drivers of vegetation changes and how these, in turn, influenced faunal species in the southeast, using published climatic, archaeological and faunal records. Further, we also used published records to provide a synthesis of environmental changes on the whole island. Results Vegetation reconstructions indicate that the mosaic vegetation in the region of Ste-Luce was highly dynamic in response to climatic changes. The open woodland, surrounding the littoral forest, transformed into an ericoid grassland between c. 5800 and 5200 cal. yr bp, possibly in response to a moderate drought recorded during this period. The littoral forest was more stable between c. 5100 and 1000 cal. yr bp, with only some minor compositional changes c. 2800 cal. yr bp and between c. 1900 and 1000 cal. yr bp. Significant forest decline, however, is observed at c. 950 cal. yr bp, coinciding with a drought and a marine surge. A comparison of these results with a synthesis of published vegetation records across the island shows asynchronous vegetation changes in response to various droughts during the Holocene, except for the 950 cal. yr bp drought event, with evidence of widespread vegetation transformations and fires across the island. *Correspondence: Malika Virah-Sawmy, Oxford Long-term Ecology Laboratory, Oxford University Centre for the Environment, South Parks Road, Oxford OX1 3QY, UK. E-mail: malika.virahsawmy@ouce.ox.ac.uk or malikavs@gmail.com 506 Main conclusions Pronounced climatic desiccation between 1200 and 700 cal. yr bp may have been the slow driver framing and triggering vegetation transformations and decline in megafaunal populations. In addition, hunting by drought-impacted human inhabitants and competition with newly introduced cattle would have amplified the impacts on megafaunal populations, leading to numerous extinctions in this period. Keywords Aridity, climate change, drought, extinction, fire, lemurs, Madagascar, palaeoecology, palaeontology, sea-level rise. www.blackwellpublishing.com/jbi doi:10.1111/j.1365-2699.2009.02203.x ª 2009 Blackwell Publishing Ltd Long-term ecology of south-eastern Madagascar INTRODUCTION The fossil record suggests that Madagascar’s ‘megafauna’ (large-bodied animals) experienced dramatic losses in the late Holocene. In particular, radiocarbon dating of fossil bones indicates that giant tortoises, elephant birds, hippopotami and giant lemurs were last recorded from Madagascar between 1400 and 600 cal. yr bp (Dewar, 2003; Burney et al., 2004), followed by the large lemur species Palaeopropithecus ingens at 500 cal. yr bp, and finally a species of hippo that went extinct as recently as 100 years ago (Burney et al., 2004). Prior to these extinctions, there is also evidence for a drastic decrease in spores of species of the coprophilous fungus Sporormiella at c. 1934 cal. yr bp at Belo-sur-Mer and 1620 cal. yr bp at Ambolisatra in the arid coastal south-west (Burney et al., 2003); these spores are typically associated with megafaunal dung and are interpreted as being indicative of a reduction in the density of large grazers and browsers on the landscape. Overall, systematic dating of sediments and bones indicates that there was a period of at least a millennium (between 2300 and 1400 cal. yr bp) when people and most megafauna coexisted, but that some change occurred between 1400 and 500 cal. yr bp when these animals were finally driven to extinction (Table 1). The contenders for these extinctions have included ecological transformations through anthropogenic activities, including over-hunting, fires, the introduction of cattle, diseases, or climatic/environmental change (Dewar, 2003; Godfrey & Jungers, 2003; Burney et al., 2004; Godfrey et al., 2004; Perez et al., 2005; Godfrey & Irwin, 2007). Of the proposed anthropogenic mechanisms, the most widely accepted hypothesis is that overhunting by humans played a major role in the extinctions. Evidence to support this hypothesis includes numerous examples of cut-bone marks on extinct fauna found at several palaeontological sites (swamps, caves and pits) since the onset of early settlement (MacPhee & Burney, 1991; Dewar, 2003; Burney et al., 2004; Perez et al., 2005). However, there is a question mark over whether this represents the large megafauna being hunted to extinction due to ever-increasing food requirements for human subsistence, because there is little evidence at archaeological sites of bones with cut-marks (Dewar, 2003; Burney et al., 2004; Perez et al., 2005). Palaeontologists working in Madagascar have therefore suggested that the extinction event may have been caused by a complex set of interactions, but that hunting was the final trigger for the megafaunal extinction (e.g. Burney et al., 2004). The alternative hypothesis that extinctions in Madagascar were triggered or driven by climate/environmental change has yet to be addressed systematically. In this paper, therefore, we address this second hypothesis through detailed vegetation reconstructions from two sites in south-east Madagascar in the region of Ste-Luce (Fig. 1). The south-east is particularly interesting because palaeontological excavations from the region have indicated extensive extinction and extirpations (Burney et al., 2008). Further, it has been demonstrated that mammalian species that survived in the south-east have Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd disjunct populations (Goodman & Ramanamanjato, 2007), possibly indicating drastic landscape modifications. Therefore, the aim of this study was to build a record of vegetation changes from local to landscape scale in the south-east in order to understand how regional processes operating over large spatial areas (for example climate change) and local processes (for example human disturbance, herbivory, fire) may have influenced the vegetation and faunal species in this region. This paper presents the results from this study and views these new records in the light of previously published reconstructions in order to address the hypothesis that late Holocene extinctions in Madagascar were triggered or driven by environmental/climatic change. There are now a number of published sites with reconstructed fossil pollen records that can be used to decipher late Holocene environmental change on the island in relation to these extinctions (Fig. 1a) as follows: (1) two coastal peat basins in Mandena (Virah-Sawmy et al., 2009a); (2) a small basin at Ambolisatra in the arid coastal south-west (Burney, 1993); (3) Lake Mitsinjo in the dry coastal north-west (Matsumoto & Burney, 1994); and (4) Lake Tritrivakely from the cooler central highlands (Burney, 1987a,b). These four regions encompass the range of bioclimatic zones in Madagascar and their associated vegetation: spiny bushland in the arid regions, deciduous forest in the drier zones, open grassland in the cooler central highlands, and the humid rain forests and grasslands in the east (Fig. 1a). These five palaeoecological records, together with two new records presented here, are compared in order to provide a synthesis of environmental changes in Madagascar, especially during the time of rapid faunal extinction between 1400 and 500 cal. yr bp. Specifically, we address the following questions. First, at the landscape scale: (1) what were the drivers of vegetation change in the habitats of Ste-Luce, and were they synchronous with other habitats in the south-east? and (2) how did vegetation change influence faunal species? Second, at the regional scale: (3) what were the drivers of vegetation change in other parts of Madagascar and again, were they synchronous? and (4) what insight can we gain from these environmental records in understanding the late Holocene megafaunal extinction event in Madagascar? MATERIALS AND METHODS Study site We extracted sedimentary sequences from two sites in the region of Ste-Luce, south-east Madagascar: a littoral forest fragment and an ericoid grassland (Fig. 1b). In this region the major vegetation formations are littoral and coastal forests interspersed with ericoid grasslands (Fig. 1b). The littoral forest is an ecosystem of national conservation priority due to its high plant endemism and its small extent. Three villages, Ambandrika, Ampanasatomboky and Ste-Luce (Manafiafy), occur in close proximity to the largest littoral forest remnants: fragments S6 (147 ha), S7 (198 ha), S8 (129 ha), S9 (377 ha) 507 14 Event – published date Location Literature cited C dates bp (material analysed); n is number of samples analysed Range for cal. yr bp Earliest evidence of human butchering on megafauna (e.g. Palaeopropithecus ingens) Latest dated occurrence for: Palaeopropithecus ingens (sloth lemur) Latest dated occurrence for: Archaeolemur sp. Archaeolemur cf. majori Archaeolemur cf. edwarsii (‘monkey’ lemur) Latest dated occurrence for: Megaladapis edwardsi Megaladapis grandidieri Taolambiby (south-west) Godfrey & Jungers (2003) 2325 ± 43 (bone) 2305–2371 Ankilitelo Cave (south-west) Burney et al. (2004) 510 ± 80 (bone), n = 4 478–586 Anjohikely Cave (north-west) Belo-sur-Mer (south-west) Anjohikely Cave (north-west) Burney et al. (1997) Burney (1999) Burney et al. (1997) 830 ± 60 (dung), n = 3 1370 ± 40 (bone), n = 3 2060 ± 60 (bone), n = 2 683–797 1271–131 1968–2098 Ankilitelo (south-west) Ampasambazimba Wetland (central highlands) Belo-sur-mer (south-west) Site unknown Simons (1997) Tattersall (1973) 630 ± 50 (bone), n = 2 1035 ± 50 (bone), n = 3 567–657 904–984 Burney (1999) Burney et al. (2004) 2140 ± 50 (talus), n = 2 1591 ± 60 (tooth), n = 2 2050–2214 1406–1548 Belo-sur-Mer (south-west) Burney et al. (2004) 1413 ± 80 (not specified), n = 3 1258–1376 Ampasambazimba (central highlands) Tsiandroina (south-west) Burney et al. (2004) 1410 ± 40 (bone), n = 2 1216–1334 Burney et al. (2004) 1694 ± 40 (bone), n = 3 1539–1643 Amboulitsate Paleolake/ Wetland Nossi (south-west) Belo-sur-Mer (coastal south-west) Ambolisatra (coastal southwest) Ambolisatra (coastal south-west) Tritrivakely (central highlands) Mitsinjo (coastal north-west) Burleigh & Arnold (1986) 750 ± 370 (bony carapace), n = 2 Outside calculation range Burleigh & Arnold (1986) Burney et al. (2003) 1250 ± 50 (bony carapace), n = 2 1990 ± 50 (sediment) 1119–1253 1880–1988 Burney (1993) c. 2915 ± 170 (bone) 2846–3286 Burney (1993) c. 1890 ± 90 (sediment) 1726–1930 Burney (1987a) 960 ± 90 (sediment) 782–966 Matsumoto & Burney (1994) No 14C date during the transition. Age–depth model indicate transition at 1000 cal. yr bp 824–1002 (using error margin for closest 14C date) Megaladapis madagascariensis Megaladapis sp. (large-bodied ‘koala’ lemur) Latest dated occurrence for: Hadropithecus stenognathus (‘monkey’ lemur) Latest dated occurrence for: Mesopropithecus pithecoides Mesopropithecus globiceps (sloth lemur) Latest dated occurrence for: Geochelone abrupta Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd Geochelone grandidieri Decline in Sporormiella spores* Ecological transition from arid woodland/ bushland to woody palm savanna Ecological transition from woody palm savanna to Didieraceae–Euphorbiaceae spiny bushland Ecological transition from woody savanna to grass-dominated savanna Ecological transition from woody savanna to grass-dominated savanna M. Virah-Sawmy et al. 508 Table 1 Published dates of important ecological events in Madagascar. Note that latest dated occurrences are presented only for taxa with substantial fossil dating (n ‡ 2, but see Burney et al., 2004, for more details) as the margins of error are presumed here to be smaller for well dated taxa. Note also that new findings may lead to earlier dates in the latest dated occurrences for some species. Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd Table 1 Continued 14 Location Literature cited Ecological transition from littoral forest to ericoid grassland Mandena matrix (coastal south-east) Virah-Sawmy et al. (2009a) Ecological transition from littoral forest to ericoid grassland Mandena littoral forest, M15 (coastal south-east) Virah-Sawmy et al. (2009a) Climatic desiccation at 5800 cal. yr bp using diatoms as proxy Mandena matrix (coastal south-east) Virah-Sawmy et al. (2009a) Climatic desiccation at 4600 cal. yr bp using diatoms as proxy Mandena matrix (coastal south-east) Virah-Sawmy et al. (2009a) Climatic desiccation at 3200 cal. yr bp using diatoms as proxy Mandena matrix (coastal south-east) Virah-Sawmy et al. (2009a) Climatic desiccation at 950 cal. yr bp using diatoms as proxy Mandena matrix (coastal south-east) Virah-Sawmy et al. (2009a) C dates bp (material analysed); n is number of samples analysed No 14C date during the transition. Age–depth model indicate transition at 1000 cal. yr bp No 14C date during the transition. Age–depth model indicate transition at 950 cal. yr bp No 14C date during the transition. Age–depth model indicate lowest water levels at 5800 cal. yr bp No 14C date during the transition. Age–depth model indicate lowest water levels at 4600cal. yr bp (between 4900 and 4500) No 14C date during the transition. Age–depth model indicate lowest water levels at 3200 cal. yr bp (between 3400 and 3000) No 14C date during the transition. Age–depth model indicate lowest water levels at 950 cal. yr bp (between 1200–700) *No apparent decline in Sporormiella at Tritrivakely (central highlands), Kavitaha (central highlands), Benavony (north-west), Amparihibe (north-west). Range for cal. yr bp 925–975 (using error margin for closest 14C date) 929–989 (using error margin for closest 14C date) 5760–5840 (using error margin for closest 14C date) 4570–4630 (using error margin for closest 14C date) 3170–3230 (using error margin for closest 14C date) 925–975 (using error margin for closest 14C date) 509 Long-term ecology of south-eastern Madagascar Event – published date M. Virah-Sawmy et al. Figure 1 Map of Madagascar showing the two different scales of study. (a) Bioclimate zones modified by Schatz (2000) from Cornet (1974). 1–4 refers to the four palaeoecological sites in Madagascar: 1, two coastal peat basins in Mandena and another two in Ste-Luce in the transitional south-east; 2, a small basin at Ambolisatra in the arid coastal south-west; 3, Lake Mitsinjo in the dry coastal north-west and 4, Lake Tritrivakely from the cooler central highlands. (b) Distribution of vegetation in the south-east showing three subtypes of littoral forest (Ste-Luce, Mandena and Petriky), transitional forests (Grand Lavosoa and Andohahela) and humid forests along the Anosyenne and Vohimena Mountains (modified from Watson et al., 2004). and S17 (237 ha) (these names have been given to the fragments in regional conservation zoning). The sedimentary sequences were retrieved from two small and closed peat basins of no more than 10 m radius, with the aim of gaining temporal records reflecting local vegetation (Fig. 1b). The first sequence was cored in a small basin no more than 5 m radius in the interior of Ste-Luce forest fragment S9 (S9-fragment) (24.77703 S, 047.17328 E) at an elevation of c. 19 m a.s.l. (Fig. 1b). Currently, the vegetation surrounding this basin is a closed forest canopy with the most abundant trees being Anthocleista longifolia, Asteropeia micraster, Dracaena spp., Homalium spp., Symphonia verrucosa, Pandanus sp. and Uapaca louvelii. Most of the littoral forest species are fire-sensitive. The soil around the basin is moist pseudo podzols above a sandy mineralized substrate. S9-fragment is about 2.5 km from the coast. The second basin, 10 m radius and adjacent to forest fragments S8 and S9, was cored in the open vegetation described as ericoid grassland – this formation forms the matrix vegetation that surrounds the littoral forests (Fig. 1b). This site (SM-matrix) neighbours pineapple plantations (24.77809 S, 047.16027 E) (Fig. 1b). The basin is at an elevation of 14 m a.s.l and the surrounding vegetation is composed of Erica spp., Hibbertia coriacea, Myrica (Morella) 510 spathulata and Asteraceae, with small patches of littoral forest. In contrast to littoral forest, the ericoid grassland is composed of fire-prone species and regularly experiences burning. The soil around the peat basin is moist sandy soil, possibly of marine or aeolian origin. The two basins are c. 1 km apart and are fed by groundwater as part of a dune aquifer. The depth of the groundwater is highly variable in this region and is perhaps the most important determining factor that influences the local vegetation type (Virah-Sawmy et al., 2009b). Analyses summarizing past changes in functional groups, diversity and compositional turnover at these two sites (in addition to the two sites in Mandena) have been presented by Virah-Sawmy et al. (2009b). Here, we present detailed pollen diagrams from these two sites. Chronology and resolution The chronology of the sedimentary sequences from the S9fragment was established on four calibrated accelerator mass spectrometry (AMS) 14C dates and that of the SM-matrix was based on five AMS 14C dates. The chronologies have been presented by Virah-Sawmy et al. (2009b). 14C dates indicate that organic material has accumulated in the basins from Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd Long-term ecology of south-eastern Madagascar Figure 2 Age–depth model with error margins of sedimentary sequences from S9-fragment and SM-matrix, south-eastern Madagascar, using linear interpolation. Modified from Virah-Sawmy et al. (2009b). c. 5130 cal. yr bp at S9-fragment, and 5905 cal. yr bp at SM-matrix (Fig. 2). Linear interpolation was the best fit for the age–depth models (Fig. 2). Standard deviations of the 14 C dates and the error margins for both age models were found to be relatively small (Fig. 2). Linear interpolation indicated that pollen and charcoal samples were analysed at a high temporal resolution, representing intervals of between 50 and 100 years for the sequence for S9-fragment. For SMmatrix, age–depth modelling indicates that the pollen and charcoal counts were undertaken at intervals of c. 50 years between 5905 and 3962 cal. yr bp and every c. 260 years for the period following 3962 cal. yr bp, as this was a period of slow accumulation. We used the model of Fairbanks et al. (2005) to convert dates published in yr bp to cal. yr bp (http://radiocarbon. ldeo.columbia.edu/research/radcarbcal.htm). Pollen analysis Sample preparation for pollen analysis was carried out following standard protocols (Bennett & Willis, 2001). A minimum of 400 terrestrial pollen grains (excluding Cyperaceae) were counted per sample in order to ensure a statistically significant sample size (Bennett & Willis, 2001). Pollen grains were identified using reference material held at the Oxford Long-term Ecology Laboratory and using published keys and plates of pollen grains of Madagascan species. Charcoal analysis Samples for microfossil charcoal (< 150 lm) analysis were prepared as part of the pollen-analytical protocol (Whitlock & Larsen, 2001). Microcharcoal concentration in each sample was determined using the point-count method (Clark, 1988). Macrocharcoal (> 150 lm) data from the two sites have been presented by Virah-Sawmy et al. (2009b). The macrocharcoal results were contrasted with the microcharcoal data (Figs 3 & 4). Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd Numerical analyses The pollen data were converted to percentage data by expressing the value for each pollen type counted in a sample as a percentage of the sum of all terrestrial pollen, excluding pteridophytes’ spores, aquatics and Cyperaceae. The percentage data were then plotted in a pollen diagram against age using psimpoll 4.25 (Bennett, 2005). Cyperaceae and pteridophytes were plotted as a percentage of the total sum of pollen and spores. The same program was used to plot micro- and macrocharcoal results (expressed as cm2 cm)3 and particles cm)3, respectively). Zonation by optimal splitting was applied to the data using psimpoll 4.25 to describe zone boundaries where there are statistically significant differences in the assemblage, representing thresholds of change. The determination of the number of significant zones was calculated using a brokenstick model (Bennett, 1996). RESULTS Pollen S9-fragment Zonation indicates that there were three significant zone transitions and therefore four main vegetation assemblages in the sequence from S9-fragment (Fig. 3). These four zones occurred between c. 5200 and 2800 cal. yr bp, c. 2800 and 1900 cal. yr bp, c. 1900 and 1000 cal. yr bp, and c. 1000 cal. yr bp and present time (Fig. 3). Zone S1 (c. 5200–2800 cal. yr bp). Between c. 5200 and 2800 cal. yr bp, forest pioneers, Myrica (Morella) [possibly Myrica spathulata (Mirb.) Verdc. & Polhill as it is a dominant species in south-east Madagascar] and Macaranga type 2 (there are currently only two Macaranga species in the littoral forest and this type is possibly the forest pioneer Macaranga perrieri Leandri) were dominant components of the pollen assemblage. 511 M. Virah-Sawmy et al. Figure 3 Changes in selected taxa and charcoal concentrations at S9-fragment, south-eastern Madagascar. S1–S4 refer to statistically significant compositional zones using psimpoll 4.25 (P > 0.01). In addition, species of Araliaceae and Meliacaeae, and other taxa such as Psychotria, Syzygium and Uapaca, were also important components of the vegetation. This composition indicates open thicket-type forest during this period. Zone S2 (c. 2800–1900 cal. yr bp). Between c. 2800 and 1900 cal. yr bp, a more dense and moist forest became established. Forest pioneers decreased in abundance, and humid rain forest species such as Symphonia [possibly Symphonia verrucosa (fasciculata) (Hils. & Bojer ex Planch. & Triana) Benth. & Hook. as it the only species within this genera in the littoral forest], Myrsinaceae, Dypsis type 1 [possibly Dypsis saintelucei Beentje as it is a littoral forest endemic and is one of the most abundant palm species in Ste-Luce littoral forests (in Cadotte et al., 2002)], Uapaca type 2 and a cf. Vitex/Fabaceae type became dominant. It is indicative of a moist period in the south-east. Zone S3 (c. 1900–1000 cal. yr bp). The forest remained dense and closed between c. 1900 and 1000 cal. yr bp. However, a few forest species, Dypsis type 1, Uapaca type 2 and a cf. Vitex/ Fabaceae type, which were dominant in the previous period, experienced population contraction. By contrast, there were abrupt and rapid population expansions of forest pioneers; Myrica and Macaranga type 1 (possibly the weedy shrub Macaranga obovata Boiv. ex Baillon) increased abruptly between c. 2000 and 1700 cal. yr bp, indicative of some 512 significant disturbances. This was followed by an abrupt population expansion of Pandanus species between c. 1700 and 1400 cal. yr bp. The forest pioneer, Macaranga type 1, increased rapidly again between c. 1100 and 950 cal. yr bp, indicative of further disturbances. Zone S4 (c. 950 cal. yr bp to present time). At c. 950 cal. yr bp, there was the most significant ecological transition, from a closed littoral forest to a Myrica bushland; all forest species declined abruptly except for Myrica, which underwent rapid population expansion for a few centuries until c. 600 cal. yr bp. The period from c. 600 cal. yr bp to the present is marked by the recovery of most forest species dominant in the previous zone. In addition, some previously uncommon species expanded at S9 during this period, for example another species of Dypsis type 2 [possibly Dypsis lutescens (H. Wendl.) Beentje & J. Dransf. as it associated with disturbed forest and is also a dominant palm tree in the littoral forest (in Cadotte et al., 2002)], Olea and Erica sp. SM-matrix There are four zones in terms of significantly different forest assemblages over the past 5900 years at SM-matrix (Fig. 4). These four zones occurred between c. 5900 and 5200 cal. yr bp, c. 5200 and 4800 cal. yr bp, c. 4800 and 3300 cal. yr bp, and c. 3300 cal. yr bp and the present time (Fig. 4). Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd Long-term ecology of south-eastern Madagascar Figure 4 Changes in selected taxa and charcoal concentrations at SM-matrix, south-eastern Madagascar. A1–A4 refer to statistically significant compositional zones using psimpoll 4.25 (P > 0.01). Note that the two sites have differed consistently in composition throughout time. Zone A1 (c. 5900–5200 cal. yr bp). The first zone between c. 5900 and 5200 cal. yr bp represents a forested period. During this time, a moist and open forest composed of a subset of taxa from S9-fragment, for example Uapaca type 1, Syzygium, Symphonia and Macaranga type 2 dominated the forest. Grasses, sedges and herbs such as a Solanum type were also abundant during this period. Zone A2 (c. 5200–4800 cal. yr bp). Between c. 5200 and 4800 cal. yr bp, most forest species experienced population declines. Fire-sensitive taxa such as Symphonia and Uapaca declined from c. 5800 cal. yr bp. Their population decline was then followed with the decline of the remnant forest taxa, for example Syzygium, Pandanus and Macaranga type 2 a few centuries later at c. 5200 cal. yr. bp, and Myrica and Erica assumed dominance from forest taxa at c. 5200 cal. yr bp. Zone A3 (c. 4800–3300 cal. yr bp). From c. 4800 cal. yr bp, the vegetation became established as an open ericoid–Myrica grassland. In this period, Myrica was dominant over Erica in terms of pollen abundance. Other taxa associated with this formation are Hibbertia coriacea (Pers.) Baillon, Vaccinium (another Ericaceae), and Pteridium. Several littoral forest taxa, for example Pandanus, Syzygium, Uapaca, Arialiaceae, Symphonia, Meliaceae and Macaranga, apparently grew at low densities. The only anomaly observed in the zone is the Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd short-lived but rapid increase of a species of Rapanea, between c. 4850 and 4750 cal. yr bp. Only one species of Rapanea currently occurs in the littoral forest, Rapanea erythroxyloides (Thouars) Mez, and it is presumed to be this species. Zone A4 (c. 3300 cal. yr bp to present time). Around c. 3300 cal. yr bp there was an increase in dominance of Erica relative to Myrica, while the percentage abundance of other species remained unchanged. The increase in Sphagnum in the peat basin at c. 3300 cal. yr bp confirms the acidic nature of ericoid-dominated vegetation. The vegetation remained stable to the present time. Charcoal S9-fragment Microcharcoal deposits, which give a regional fire signal, indicate that fires were extremely rare in S9-fragment, except for very small peaks at c. 600 cal. yr bp and between c. 300 and 100 cal. yr bp. SM-matrix Microcharcoal deposits at SM-matrix indicate that fires became more prevalent in the surrounding area from 513 M. Virah-Sawmy et al. c. 3300 cal. yr bp, coinciding with when Erica gained dominance over Myrica in the pollen record. DISCUSSION Drivers of vegetation change in south-east Madagascar From the palaeoecological records presented here, there is evidence that the vegetation in the S9-fragment and in the adjacent presently open habitat of the SM-matrix has been highly dynamic; both were originally forested, yet one of these sites (SM-matrix) is now an ericoid grassland, and the S9fragment is now of a different composition of species from that seen in the mid-Holocene (Figs 3 & 4). Compositional changes in the two adjacent sites, however, occurred asynchronously, which suggests that different processes were responsible for these transformations. Here we contrast vegetation changes in both sites against archaeological, charcoal and palaeoclimatic records. At SM-matrix the transformation of the open Uapaca/ Myrica forest to ericoid/Myrica grassland between c. 5800 and 5200 cal. yr bp pre-dates evidence for human presence on the island by several millennia (summary in Virah-Sawmy et al., 2009b). Instead, this transformation coincides with the moderate drought event observed in the south-east diatom records between c. 5900 and 5600 cal. yr bp (lowest water levels at 5800 cal. yr bp) (Virah-Sawmy et al., 2009a). This drought may have triggered a series of cascading impacts that eventually led to the emergence of ericoid/Myrica grassland. It appears that this drought event first led to the decline of some forest dominants, for example drought-sensitive trees such as Symphonia and Uapaca at c. 5800 cal. yr bp, and was then either followed by, or resulted in, an increase in the incidence of fires between c. 5600 and 5400 cal. yr bp, possibly through the accumulation of dead biomass (Fig. 4). These changes eventually culminated in the decline of most remnant forest species, such as Syzygium, Pandanus and Macaranga, and the emergence of ericoid/Myrica grassland by c. 5200 cal. yr (Fig. 4). Another drought between c. 3300 and 3000 cal. yr bp (Virah-Sawmy et al., 2009a) coincides with the period when Erica gained dominance from Myrica and fire became more prevalent in the system (Fig. 4). It is interesting to note that the SM-matrix has remained an ericoid grassland up to the present time, both in the absence of local human influences for nearly four millennia in the region (up to 1200 cal. yr bp) and despite the return of moister climatic conditions. This strongly suggests that ericoid grasslands in the south-east are in a selfsustaining, stable state maintained by fire, and are insensitive to the return of climatic conditions that are more favourable to woodlands. The littoral forest of the S9-fragment has also been highly dynamic, but its compositional changes appear to have been in response to different climatic drivers. The end of the moderate drought at c. 3000 cal. yr bp (Virah-Sawmy et al., 2009a) may have triggered the transition from forest thickets of Meliaceae 514 and Macaranga type 2 to a community dominated by moist rain forest species; this humid community remained stable until c. 1900 cal. yr bp (Fig. 3). It appears that some forest species (e.g. Dypsis type 1 and Uapaca type 2) declined at c. 1900 cal. yr in the absence of fires, possibly indicating a period of climatic disturbances and desiccation. There are further significant vegetation disturbances, again in the absence of fires, between c. 1900 and 950 cal. yr (Fig. 3). These events may have been caused by episodic marine surges as evidenced in geochemical changes at Mandena, a site located south of Ste-Luce (Virah-Sawmy et al., 2009a). The geochemical records from Mandena indicate marine inundations on the coastal system due to a sea-level high of +2 to 3 m between c. 2300 and 900 cal. yr bp (Battistini et al., 1976; Camoin et al., 1997, 2004; Virah-Sawmy et al., 2009a). These inundations, as evidenced by changes in the Sr/Ca ratio (an indicator of salinity) from coastal soils, occurred intermittently at c. 2200, 1600–1700, 1400 (highest intensity), 1100–1200 and 950 cal. yr bp (Virah-Sawmy et al., 2009a). An additional proxy that can be used to indicate anoxic conditions, for example due to inundations, is low phosphorus (Engstrom & Wright, 1984). In the S9-fragment there is a significant decline in soil phosphorus levels between c. 2200 and 1300 cal. yr bp (Virah-Sawmy et al., 2009b), which also suggests marine inundation during this period. Changes in the pollen records of the S9-fragment indicate that the vegetation was affected by these marine transgressions. For example, there is the rapid expansion of forest pioneers, Myrica and Maranga type 1, between c. 2000 and 1700 cal. yr bp and between c. 1100 and 950 cal. yr bp. Further, a rapid increase of the screw-pine Pandanus, a taxon found either on freshwater or saline waterlogged soils, between c. 1700 and 1400 cal. yr bp, also indicates anoxic conditions. Overall, it appears that some species in the S9 forest fragment responded rapidly to episodic marine surges. Nevertheless, forest cover was maintained until 1000 cal. yr bp. Synchronous with two forested sites at Mandena (VirahSawmy et al., 2009a,b), there was a rapid forest decline in the S9-fragment between 1000 and 900 cal. yr bp. These rapid changes occurred in the absence of fires (Fig. 3). These landscape-scale forest declines coincided with both a marine surge and severe climatic desiccation (Virah-Sawmy et al., 2009a). These findings suggest that the littoral forests and woodlands of south-eastern Madagascar were highly sensitive to the combined impact of water and salt stresses, in contrast to the highly tolerant ericoid grassland. Myrica bushlands that expanded in the S9-fragment immediately following the 950 cal. yr climatic event declined as dense littoral forest became re-established at c. 600 cal. yr bp. The pollen records indicate that the S9-fragment presently has an assemblage similar to that of the previous pollen zone: this suggests a system that can recover and a high degree of resilience to environmental disturbances (Virah-Sawmy et al., 2009b). In summary, results from these two new sequences (S9fragment and SM-matrix) in south-eastern Madagascar suggest that an environmental change occurred between 1000 and Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd Long-term ecology of south-eastern Madagascar 900 cal. yr bp that, by extension, caused a dramatic decline in forest extent along the south-eastern coast of Madagascar, and the landscape became more open, dominated by Myrica or ericoid bushland. The environmental trigger for this change appears to have been a combination of sea-level rise and aridity. The next questions to ask, therefore, are: (1) how do these records of rapid and dramatic change in the forests of south-east Madagascar relate temporally to what is known about faunal extinction in the south-east of Madagascar? and (2) how widespread was the aridification event of 950 cal. yr bp? Faunal extinction in south-east Madagascar Animal subfossil collections from Andrahomana Cave in the south-east (Fig. 1b) include a number of extinct taxa as well as several extant taxa that lived in this region in the mid- to late Holocene (Burney et al., 2008). Among the collections of extinct taxa are the large primates Megaladapis edwardsi and Hadropithecus stenognathus, which are important elements of the subfossil fauna of the arid south and south-west. Other large and more wide-ranging primates, such as Archaeolemur majori and Pachylemur insignis, are also well represented. Other extinct species include the giant jumping rat, Hypogeomys australis; elephant birds, Aepyornis spp.; the carnivore Cryptoprocta spelea; the pygmy hippo Hippopotamus lemerlei; and the giant land tortoise Geochelone grandidieri (Burney et al., 2008). Subfossils of extant taxa (but now extinct in the south-east) include species that are presently confined to either dry or humid habitats in Madagascar (Burney et al., 2008). Among those typically restricted to dry habitats are the subfossil of the lemur Lemur catta, the rodent Macrotarsomys bastardi, and a bird species, Numida meleagris. Species presently confined to humid habitats include the subfossil of the primate Avahi laniger, the rodent Nesomys rufus and the tenrec Microgale principula. This rich past faunal assemblage of both dry- and wet-adapted taxa in the south-east suggests that both humid and dry forests would once have been more extensive than their present distribution in the region. Provisional radiocarbon dating for the south-east also suggests that, as in the rest of the island, megafaunal extinction occurred in the late Holocene. The last dates of occurrence for various species with substantial fossil dating indicate that the faunal extinction event in Madagascar occurred between 1400 and 500 cal. yr bp (Table 1). Evidence from three sites in south-eastern Madagascar for a regional-scale transformation of forested vegetation to more open bushland occurring at around 950 cal. yr bp and temporally coinciding with evidence for large-scale faunal extinction leads us to hypothesize that the faunal extinction event between 1400 and 500 cal. yr bp in Madagascar was driven by natural environmental change, in particular an aridification event (between 1200 and 700 cal. yr bp) coupled with high sea-level rise (Virah-Sawmy et al., 2009a). But how common or widespread was the effect of Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd this climatic change, and how did it affect different ecosystems within Madagascar? How common or widespread was the 950 cal. yr aridification event? Two previously published diatom records in Madagascar (from the south-east and from the central highlands) indicate that the two most severe Holocene droughts lasting several centuries occurred at c. 4600 and 950 cal. yr bp (Gasse & Van Campo, 1998; Virah-Sawmy et al., 2009a). There have also been two moderate droughts at c. 5800 and 3200 cal. yr bp, and short arid intervals between c. 1900 and 1300 cal. yr bp (Virah-Sawmy et al., 2009a). Pollen analyses from sites across the island indicate both synchronous and asynchronous responses to these Holocene droughts. For example, the 5800 cal. yr drought resulted in vegetation transition at only one site (out of four) in the south-east (at SM-matrix), while the 3200 and 1900 cal. yr bp droughts appear to have affected vegetation primarily in the south-west and to a lesser extent in the south-east (Fig. 5). Interestingly, the vegetation transition at 1900 cal. yr bp in the south-west also coincides with evidence for a drastic decrease in spores of species of the coprophilous fungus Sporormiella (hence megafauna) in sedimentary sequences from the south-west, but this decline is not apparent in other regions of the island (see Burney et al., 2003), where there were also no apparent forest declines. Vegetation across the island remained relatively unperturbed during one of the most severe droughts in Madagascar at 4600 cal. yr, yet the drought at 950 cal. yr is concordant with widespread vegetation changes across the island – in the south-east (Virah-Sawmy et al., 2009a; this paper), the northwest and the central highlands (Burney, 1987a,b; Matsumoto & Burney, 1994). These ecological shifts between 1200 and 900 cal. yr bp have been interpreted previously as anthropogenically driven, through the evidence of abrupt charcoal peaks (Burney, 1987a,b; Matsumoto & Burney, 1994). However, humid forest decline in S9-fragment at St-Luce and M15 fragment in Mandena at 950 cal. yr bp (Virah-Sawmy et al., 2009a) occurred in the absence of charcoal peaks. By contrast, fires could have been ignited naturally during the arid interval at 950 cal. yr bp in fire-prone savannas of the north-west and the central highlands. With respect to the two most pronounced droughts of the Holocene, it is possible that they differed in their prevailing temperatures, and that the 4600 cal. yr bp drought occurred during cooler conditions in comparison with the 950 cal. yr bp drought event. Gasse & Van Campo (1998) interpreted changes in diatom and pollen assemblages in the central highlands around 1000 cal. yr bp as a temperature rise. Warmer conditions during the 950 cal. yr bp drought may explain the greater frequency of fires and hence vegetation transformations. The 950 cal. yr bp drought has been linked to global phenomena such as solar forcing and ocean circulations, which significantly lowered lake levels in most 515 M. Virah-Sawmy et al. Figure 5 Chronologies of megafaunal extinction in Madagascar in relation to climatic and cultural changes, vegetation transformations and fire events documented from seven sedimentary sequences in three broad bioclimate zones in Madagascar. Extinction window based on Burney et al. (2004). of East Africa (Halfman et al., 1994; Verschuren et al., 2000; Alin & Cohen, 2003; Russell & Johnson, 2005). The drivers of other Holocene droughts in Madagascar are not particularly well known; whether these droughts were widespread on the island or were local may have depended on the climatic drivers – that is, whether they were generated by El Niño/ Southern Oscillation (ENSO) events, changes in sea-surface temperatures, intertropical convergent zones or solar forcing. For example, currently ENSO events lead to droughts in southern Madagascar, while other regions of the island experience rainfall higher than average (Jury, 2003; Ingram & Dawson, 2005). Did the environmental changes at 950 cal. yr BP cause the megafaunal extinction in Madagascar, and if so, how? A comparison of vegetation changes across the island highlights a series of ecological transformations around 950 cal. yr bp (Fig. 5). The drought interval around 950 cal. yr was synchronous with: (1) significant and abrupt vegetation transitions across the island, (2) a number of equally abrupt 516 fire events in fire-prone vegetation (ericoid grasslands and savannas), and (3) the introduction and expansion of livestock on the island from 1000 cal. yr bp [these were possibly introduced from Africa and Comoros around 1031 cal. yr bp (Burney et al., 2003), with clearer evidence in the form of bones that they expanded across the island by 850 cal. yr bp (Dewar, 2003)], culminating a few centuries later with the complete extinction of most megafauna. Systematic and abrupt vegetation changes around 950 cal. yr bp occurred both in the absence and presence of abrupt fire peaks (Fig. 5), suggesting evidence for an overriding regional factor framing these ecosystem changes: climatic desiccation. The combination of events fits well with the synergistic hypothesis of extinction [‘human impacts played a role, but to differing extents in various regions; factors interacting to multiply effects; some amplification by background climate change’ (Burney et al., 2004)]. However, it appears that pronounced desiccation could have been the overriding factor that led to widespread fires and vegetation transformations. As a consequence, much primary forest/woodland habitat would have been reduced during the 950 cal. yr bp drought event, while more open grasslands expanded. We may infer that some Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd Long-term ecology of south-eastern Madagascar animals would have experienced range expansion while others saw their habitats shrink or disappear. In addition, new competition may have arisen between the megafauna and newly introduced cattle. Habitat change and competition would have affected populations of megafauna more than other species, as these animals would have required larger habitats with more resources to maintain themselves. Their slower reproductive rates may also have made them more prone to extinction as a result of changing environments (Johnson, 2002). Furthermore, the pronounced 950 cal. yr bp drought would also have affected human populations, as droughts do today in Madagascar, due to crop failures [the first carbonized rice records date to 850 cal. yr bp, but it has been suggested that rice was introduced between 2000 and 1500 cal. yr bp by the early Southeast Asian settlers (Wright & Rakotoarisoa, 2003)], and would have necessitated increasing reliance on bushmeat for survival. Overall, hunting, competition and droughtinduced fires, superimposed on habitat changes, may have caused a tipping point for the megafauna. If the first settlers were not present, megafaunal population numbers might have recovered after an unstable period of adjusting to the new climate and habitats. Conversely, if climate had not already reduced the population sizes of megafauna, the relatively small number of people then on the island (Wright & Rakotoarisoa, 2003) may not have been sufficient to cause these extinctions. While there is a great need for further palaeoecological and palaeontological investigations on the island, current evidence suggests that the role of climatic influences on extinction events on larger land masses, such as Madagascar, Australia, and North and South America, should not be discounted (Wroe et al., 2006). ACKNOWLEDGEMENTS Funding for this research was provided by Oxford University Centre for the Environment, the Wingate Foundation, a Rufford Small Grant, Oxford University’s Environment Change Institute, and a Jesus College Research Grant. M.V.S. would like to thank James Millett and Terry Dawson for support at the beginning of this project and Keith Bennett for his invaluable help on psimpoll. We are grateful for permission and assistance in the field from the Ministère de l’Energie et des Mines, the University of Antananarivo, Missouri Botanical Garden and QIT Madagascar Minerals. We are thankful to the Daubeny Herbarium of the University of Oxford and the Royal Botanical Gardens, Kew, for providing reference materials. NERC Radiocarbon Laboratory provided radiocarbon dates through grant allocation 1217-0407. Finally, we thank two anonymous referees and the editors for extremely helpful comments. REFERENCES Alin, S.R. & Cohen, A.S. (2003) Lake-level history of Lake Tanganyika, East Africa, for the past 2500 years based on Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd ostracode-inferred water-depth reconstruction. Palaeogeography, Palaeoclimatology, Palaeoecology, 199, 31–49. Battistini, R., Delibrias, G. & Laborel, J. (1976) Datations des niveaux holocènes á Madagascar. Comptes réndu sommaire de la Société géologique, France, 6, 284–285. Bennett, K.D. (1996) Determination of the number of zones in a biostratigraphical sequence. New Phytologist, 132, 155–170. Bennett, K.D. (2005) Documentation for psimpoll 4.25 and pscomb 1.03. C programs for plotting and analysing pollen data. Quaternary Geology, Department of Earth Sciences, University of Uppsala, Uppsala. Available at: http://www.chrono. qub.ac.uk/psimpoll/psimpoll-faq.html (accessed 7 July 2008). Bennett, K.D. & Willis, K.J. (2001) Pollen. Tracking environmental change using lake sediments (ed. by J.P. Smol, H.J.B. Birks and W.M. Last), pp. 5–30. Kluwer Academic, Dordrecht. Burleigh, R. & Arnold, E.N. (1986) Age and dietary differences of recently extinct Indian Ocean tortoises (Geochelone s. lat.) revealed by carbon isotope analysis. Proceedings of the Royal Society of London Series B, Biological Sciences, 227, 137–144. Burney, D.A. (1987a) Late Holocene vegetation change in central Madagascar. Quaternary Research, 28, 130–143. Burney, D.A. (1987b) Pre-settlement vegetation changes at Lake Tritrivakely, Madagascar. Palaeoecology of Africa and the Surrounding Islands, 18, 357–381. Burney, D.A. (1993) Late Holocene environmental changes in arid southwestern Madagascar. Quaternary Research, 40, 98– 106. Burney, D.A. (1999) Rates, patterns, and processes of landscape transformation and extinction in Madagascar. Extinctions in near time: causes, contexts, and consequences (ed. by R.D.E. MacPhee), pp. 145–164. Kluwer/Plenum, New York. Burney, D.A., James, H.F., Grady, F.V., Rafamantanantsoa, J.-G., Ramilisonina, Wright, H.T. & Cowart, J.B. (1997) Environmental change, extinction and human activity: evidence from caves in NW Madagascar. Journal of Biogeography, 24, 755–767. Burney, D.A., Robinson, G.S. & Burney, L.P. (2003) Sporormiella and the late Holocene extinctions in Madagascar. Proceedings of the National Academy of Sciences USA, 100, 10800–10805. Burney, D.A., Burney, L.P., Godfrey, L.R., Jungers, W.L., Goodman, S.M., Wright, H.T. & Jull, A.J.T. (2004) A chronology for late prehistoric Madagascar. Journal of Human Evolution, 47, 25–63. Burney, D.A., Vasey, N., Godfrey, L.R., Ramilisonina, Jungers, W.L., Ramarolahy, M. & Raharivony, L. (2008) New findings at Andrahomana Cave, southeastern Madagascar. Journal of Cave and Karst Studies, 70, 13–24. Cadotte, M.W., Franck, R., Reza, L. & Lovett-Doust, J. (2002) Tree and shrub diversity and abundance in fragmented littoral forest of southeastern Madagascar. Biodiversity and Conservation, 11, 1417–1436. Camoin, G.F., Colonna, M., Montaggioni, L.F., Casanova, J., Faure, G. & Thomassin, B.A. (1997) Holocene sea level 517 M. Virah-Sawmy et al. changes and reef development in the southwestern Indian Ocean. Coral Reefs, 16, 247–259. Camoin, G.F., Montaggioni, L.F. & Braithwaite, C.J.R. (2004) Late glacial to post glacial sea levels in the western Indian Ocean. Marine Geology, 206, 119–146. Clark, J.S. (1988) Particle motion and the theory of charcoal analysis: source area, transport, deposition, and sampling. Quaternary Research, 30, 67–80. Cornet, A. (1974) Essai de cartographie bioclimatique à Madagascar. ORSTOM, Paris. Dewar, R.E. (2003) Relationship between human ecological pressure and the vertebrate extinctions. The natural history of Madagascar (ed. by S.M. Goodman and J.P. Benstead), pp. 103–111. University of Chicago Press, Chicago. Engstrom, D.R. & Wright, H.E. (1984) Chemical stratigraphy of lake sediments as a record of environmental change. Lake sediments and environmental history (ed. by E.Y. Haworth and J.W.G. Lund), pp. 11–69. University of Minnesota Press, Minneapolis, MN. Fairbanks, R.G., Mortlock, R.A., Li Cao, T.-C.C., Thomas, A.P., Guilderson, P., Fairbanks, T.W. & Bloom, A.B. (2005) Marine radiocarbon calibration curve spanning 0 to 50,000 years B.P. based on paired 230Th/234U/238U and 14C dates on pristine corals. Quaternary Science Reviews, 24, 1781–1796. Gasse, F. & Van Campo, E. (1998) A 40,000-yr pollen and diatom record from Lake Tritrivakely, Madagascar, in the southern tropics. Quaternary Research, 49, 299–311. Godfrey, L. & Irwin, M. (2007) The evolution of extinction risk: past and present anthropogenic impacts on the primate communities of Madagascar. Folia Primatologica, 78, 405– 419. Godfrey, L.R. & Jungers, W.L. (2003) The extinct sloth lemurs of Madagascar. Evolutionary Anthropology: Issues, News, and Reviews, 12, 252–263. Godfrey, L.R., Semprebon, G.M., Jungers, W.L., Sutherland, M.R., Simons, E.L. & Solounias, N. (2004) Dental use wear in extinct lemurs: evidence of diet and niche differentiation. Journal of Human Evolution, 47, 145–169. Goodman, S.M. & Ramanamanjato, J.-B. (2007) A perspective on the paleo-ecology and biogeography of extreme southeastern Madagascar, with special reference to animals. Biodiversity, ecology and conservation of littoral ecosystems in southeastern Madagascar, Tolagnaro (ed. by J.U. Ganzhorn, S.M. Goodman and M. Vincelette), pp. 25–48. Smithsonian Press, Washington, DC. Halfman, J.D., Johnson, T.C. & Finney, B. (1994) New AMS dates, stratigraphic correlations and decadal climate cycles for the past 4 ka at Lake Turkana, Kenya. Palaeogeography, Palaeoclimatology, Palaeoecology, 111, 83–98. Ingram, J.C. & Dawson, T.P. (2005) Climate change impacts and vegetation response on the island of Madagascar. Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences, 363, 55–59. Johnson, C.N. (2002) Determinants of loss of mammal species during the Late Quaternary ‘megafauna’ extinctions: life 518 history and ecology, but not body size. Proceedings of the Royal Society B: Biological Sciences, 269, 2221–2227. Jury, M.R. (2003) Climate: the climate of Madagascar. The natural history of Madagascar (ed. by S.M. Goodman and J.P. Benstead), pp. 75–87. University of Chicago Press, Chicago. MacPhee, R.D.E. & Burney, D.A. (1991) Dating of modified femora of extinct dwarf Hippopotamus from southern Madagascar: implications for constraining human colonization and vertebrate extinction events. Journal of Archaeological Science, 18, 695–706. Matsumoto, K. & Burney, D.A. (1994) Late Holocene environments at Lake Mitsinjo, northwestern Madagascar. The Holocene, 4, 16–24. Perez, V.R., Godfrey, L.R., Nowak-Kemp, M., Burney, D.A., Ratsimbazafy, J. & Vasey, N. (2005) Evidence of early butchery of giant lemurs in Madagascar. Journal of Human Evolution, 49, 722–742. Russell, J.M. & Johnson, T.C. (2005) A high-resolution geochemical record from Lake Edward, Uganda Congo and the timing and causes of tropical African drought during the late Holocene. Quaternary Science Reviews, 24, 1375–1389. Schatz, G. E. (2000) Endemism in the Malagasy tree flora. Diversity and endemism in Madagascar (ed. by W.R. Lourenço and S.M. Goodman), pp. 1–10. Mémoires de la Société de Biogégraphie, Paris. Simons, E.L. (1997) Lemurs: old and new. Natural change and human impacts in Madagascar (ed. by S.M. Goodman and B.D. Patterson), pp. 142–168. Smithsonian Press, Washington, DC. Tattersall, I. (1973) A note on the age of the subfossil site of Ampasambazimba, Miarinarivo Province, Malagasy Republic. American Museum Novitates, 2520, 1–6. Verschuren, D., Laird, K.J. & Cumming, B.F. (2000) Rainfall and drought in equatorial east Africa during the past 1100 years. Nature, 403, 410–413. Virah-Sawmy, M., Willis, K.J. & Gillson, L. (2009a) Threshold response of Madagascar’s littoral forest to sea-level rise. Global Ecology and Biogeography, 18, 98–110. Virah-Sawmy, M., Gillson, L. & Willis, K.J. (2009b) How does spatial heterogeneity influence resilience to climatic changes? Ecological dynamics in southeast Madagascar. Ecological Monographs, 79, 557–574. Watson, J.E.M., Whittaker, R.J. & Dawson, T. (2004) Habitat structure and proximity to forest edge affect the abundance and distribution of forest-dependent birds in tropical coastal forests of southeastern Madagascar. Biological Conservation, 120, 311–327. Whitlock, C. & Larsen, C.P.S. (2001) Charcoal as a fire proxy. Tracking environmental change using lake sediments (ed. by J.P. Smol, H.J.B. Birks and W.M. Last), pp. 75–97. Kluwer Academic, Dordrecht. Wright, H.T. & Rakotoarisoa, J.A. (2003) The rise of Malagasy societies: new developments in the archaeology of Madagascar. The natural history of Madagascar (ed. by S.M. Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd Long-term ecology of south-eastern Madagascar Goodman and J.P. Benstead), pp. 112–122. University of Chicago Press, Chicago. Wroe, S., Field, J. & Grayson, D.K. (2006) Megafaunal extinction: climate, humans and assumptions. Trends in Ecology and Evolution, 21, 61–62. BIOSKETCHES Malika Virah-Sawmy is interested in long-term ecology, conservation policy, climate-change conservation strategies, ecological restoration and invasive species management in tropical forest systems. She works as the Terrestrial Programme Coodinator for WWF Madagascar and Western Indian Ocean Programme Office. Journal of Biogeography 37, 506–519 ª 2009 Blackwell Publishing Ltd Katherine J. Willis is Professor of Long-Term Ecology at the University of Oxford. Her research interests focus on modelling the long-term relationships between vegetation dynamics and environmental change, and determining the role of past events in shaping the present-day distribution of fauna and flora. Her recent research has also focused on the applied use of long-term records in restoration ecology and biodiversity conservation. Lindsey Gillson is a lecturer in plant conservation at the University of Cape Town. Her research interests include theoretical ecology, conservation ecology and applied palaeoecology. She has research interests in Kenya, Madagascar, Mozambique, South Africa and Tanzania. Editor: Jack Williams 519