En se basant sur les éléments des dépôts subfossiles, il apparaît que les biomes de forêts sèches modernes de Madagascar, plus particulièrement ceux du quart méridional de l'île, ont subi des changements climatiques significatifs au cours du Quaternaire. Des données de dépôts palynologiques subfossiles montrent que cette région était particulièrement sèche il y a environ 3000 à 2000 ans [2], période qui coïncide avec l'extinction locale ou totale de plusieurs groupes distincts d'oiseaux et de mammifères [6], [9], [10] and [18]. Une part considérable du matériel subfossile de vertébrés utilisée dans cette étude provient de fouilles réalisées au cours d'anciennes explorations paléontologiques, alors que le contrôle stratigraphique n'était pas encore appliqué. Au cours des dernières années, des techniques de fouilles plus modernes ont été suivies dans un certain nombre de sites, qui nous permettent de présenter ici quelques-unes des nouvelles découvertes.

Les fouilles effectuées en 2003 au niveau de dépôts subfossiles de la grotte d'Andrahomana, dans l'extrême Sud-Est de Madagascar (Fig. 1), ont permis de disposer d'un matériel considérable se rapportant aux petits mammifères. La grotte, qui est située juste à l'ouest de la limite d'un écotone brusque entre un biome humide oriental et un biome sec occidental, contient des dépôts qui permettent de considérer la composition biotique locale qui régnait au Quaternaire. Parmi les ossements récupérés récemment dans la grotte, se trouvaient ceux d'un grand rongeur du genre endémique Macrotarsomys. Jusque récemment, ce genre n'était connu que de deux espèces avec M. bastardi (poids moyen adulte 24,5 g) à large distribution sur presque tout l'Ouest et le Sud, et M. ingens (poids moyen adulte 64,4 g), à distribution limitée à une petite région de forêt sèche caducifoliée du Centre-Ouest [4] and [14]. Une troisième espèce a été nommée récemment dans ce genre, M. petteri, qui n'est connue que par un seul spécimen récolté au nord de Toliara et qui pesait 105 g [11].

On the basis of evidence from subfossil deposits, the modern Malagasy dry forest biomes, particularly those in the southern quadrant of the island, have undergone significant climatic change during the Quaternary. Evidence from subfossil pollen deposits, demonstrates that approximately 3,000–2,000 years ago the region underwent considerable desiccation [2], which included, for example, the drying-up of former freshwater lakes and the disappearance of a considerable portion of the associated bird fauna [10] and either local or complete extinction of several different groups of mammals [6], [9] and [18]. Most of the subfossil vertebrate material used in these studies came from excavations made during an earlier period of paleontological exploration of the island, largely prior to 1940, when stratigraphic control was not employed. Therefore, the sequence of events involving these extinctions and climatic vicissitudes are difficult to interpret. In recent years, more modern excavation techniques have been employed at a number of sites in southern Madagascar and herein we report on some new findings that provide a further window into understanding aspects of these climatic changes.

Excavations in 2003 of the subfossil deposits in Andrahomana Cave, located in extreme southeastern Madagascar (Fig. 1), have provided considerable new mammal material, including remains of small species. The cave, located just to the west of an extraordinarily abrupt ecotone between humid (east) and dry (west) biomes, contains deposits that provide an excellent faunistic barometer of the shifts in local biotic composition during periods of the Quaternary [13] and [28]. Amongst the recently recovered bone remains from the cave is a large species of rodent belonging to the genus Macrotarsomys, which is part of an endemic radiation placed in the family Nesomyidae [17] and [19]. Until recently, the genus Macrotarsomys was composed of two species, the diminutive kangaroo-rat-like M. bastardi (mean adult mass 24.5 g), which has a broad distribution across much of western and southern Madagascar, and the notably larger M. ingens (mean adult mass 64.4 g), which is believed to be limited to a small area of dry deciduous forest near Ankarafantsika in western central Madagascar [4] and [14] (Fig. 1). Recently, a third species was named for the genus, M. petteri, known from a single specimen collected in early 2003 in the Mikea Forest north of Toliara (Fig. 1) and that had a body mass of 105 g [11]. This latter species is distinguished from M. ingens by several external pelage and cranio-dental characteristics and measurements.

Little is known about the natural history of Macrotarsomys petteri. The single specimen of an extant population was collected at Andaladomo within the Mikea Forest (Fig. 1). This region represents a zone of transition between the spiny bush of the south and the dry deciduous forest of the central west. During the 2003 biological inventory of the Mikea Forest, six sites were visited and intensively surveyed, representing a range of different habitat types [22]. The forest of Andaladomo, which rests on red sandy soils, showed notable differences to the other five sites visited with regards to structural and floristic aspects – canopy reaching 20 m, relatively high densities of terrestrial (Lissochilus) and epiphytic (Microcelia) orchids, dense patches of grass, and a local vegetational community composed of mostly non-spiny plants.

There is previous evidence of a large species of Macrotarsomys in the subfossil record of southern Madagascar. Material of members of this genus was excavated from reputed Pliocene or Lower Pleistocene deposits in a cavity in karst limestone near Lake Tsimanampetsotsa in the extreme southwest [23]. Since no age measurements were performed however, the estimated age of this material was entirely conjectural. Three different size ranges of Macrotarsomys were reported: (1) M. bastardi, which is still extant in the region, (2) an unidentified species intermediate between M. bastardi and M. ingens, and (3) a species larger than M. ingens that was considered extinct and new to science [23]. Further, there is a report of the occurrence of M. ingens in the subfossil deposits of the ‘grotte d'Ankazoabo’ [5]. Although we have not been able to examine the material collected at these sites referred to a large species of Macrotarsomys, based on the illustration in Sabatier & Legendre [23 (Fig. 4b)] and biogeographic inference, at least some of these specimens are almost certainly referable to M. petteri. The discovery of bone remains in the Andrahomana Cave deposits provides further support to this conclusion.

On the basis of current knowledge, it is presumed that all three extant species of Macrotarsomys dig their terrestrial burrows into sandy ground [4] and except by chance would not venture into a cave. Diurnal birds of prey feed on M. ingens in the Ankarafantsika region [4] and these are presumed to be the Madagascar buzzard (Buteo brachypterus) or Madagascar harrier-hawk (Polyboroides radiatus). Both of these raptors can frequent rock outcrops and cave entrances; both are broadly distributed in southeastern Madagascar [13]. At roost sites, these predators dismantle prey, which can accidentally drop to the ground uneaten, and regurgitate ingested bone remains in the form of pellets. There is also evidence that the barn owl (Tyto alba) and Boidae snakes of the genera Boa and Acrantophis feed on M. ingens [21] – both of these groups of predators can frequent caves. Bone fragments of M. ingens have been found in the scats of the native Carnivora Cryptoprocta ferox and the introduced Viverricula indica [21]. Given the relatively intact state of the Macrotarsomys specimens recovered in the Andrahomana Cave deposits, we presumed that the majority of the remains, from both excavated sites, were left behind by raptors either in the form of dropped prey remains or regurgitated pellets. The stratigraphic reversal apparent in the dates presented confirms the excavators’ impression that material in this site has been stratigraphically mixed by some process, probably redeposition.

During the course of the 2003 biological inventories of the Mikea Forest, where the holotype of M. petteri was captured, 3100 trap-nights (1 trap in place for 24 h = 1 trap-night) in total were accrued at six different sites [24]. A single specimen of this taxon was captured at Andaladomo, a forest block with distinctly more mesic vegetation than the other five sites surveyed. Without more detailed and long-term inventories in the Mikea Forest, it cannot be discerned if this species occurs elsewhere in the region. What is apparent is that at Andaladomo, this species is uncommon or at least difficult to capture. Something on the order of 15 other sites have been surveyed for small mammals in southern dry forest or spiny bush forests, using the same field techniques as in the Mikea Forest, and no evidence of an extant large species of Macrotarsomys has been found. Our assumption is that M. petteri has a very restricted range within the Mikea Forest and continued habitat destruction in this region [1] has serious implications for the long-term viability of this species.

It is conceivable that M. petteri still occurs in the extreme South of Madagascar in pockets of forest with more mesic vegetation and with sandy or relatively soft substrates – based on the needed soil type for burrow construction. To the east of Andrahomana Cave there are the sites of Grand Lavasoa and Ambatorongorongo (also known as Ambatotsirongorongo) with remnant natural forests that show mixed humid, spiny and littoral forest floristic and faunistic composition [15] and [20]. These two sites were inventoried for small mammals in early 2005 by Zafimahery Rakotomalala and Vonjy Andrianjakarivelo and, based on collected specimens (in FMNH and UADBA), there is no evidence of M. petteri occurring locally. Virtually all of the native species they captured at these sites are typical of the southern spiny bush, although within their collections there is the Tenrecidae Oryzorictes hova, a species typical of mesic habitat.

Based on current evidence, the subfossil material of M. petteri recovered from Andrahomana Cave and other sites in the southern portion of the island represent extinct populations, dating from recent geological periods when the climate was more mesic and the region had a distinctly different biota. The absence of M. petteri from AHA-F at Andrahomana, a site with shallow time depth, provides further support that this animal may have been locally extinct for several hundred years, a point corroborated by our radiocarbon dates. Given the relative rarity of M. petteri at AHA-I compared with Rattus rattus, the possibility exists that M. petteri may have been particularly vulnerable to ecological displacement or replacement by this introduced rodent during a time period when climatic desiccation was also reducing its geographic range. Extreme disproportionate numbers of Rattus rattus versus native endemic rodent taxa have been reported in extant forest rodent communities [12] and [14], and a connection has been suggested between the decline of the extant native animals and the penetration of R. rattus into even very remote forested regions of Madagascar [8]. The subfossil deposits from Andrahomana document, for the first time, this pattern of faunal turnover within a recent geological context following human settlement.

There are other parallels amongst extant Malagasy mammals to the marked range contraction found in M. petteri over the past few thousand years. The endemic rodent Hypogeomys antimena, which reaches an adult body mass of over 1 kg, once had a broad distribution in southwestern Madagascar [9]. The only extant population occurs in the Menabe region near Morondava and this species is threatened with extinction that appears to be in part associated with habitat destruction [25]. Subfossil material of this species from Ampoza (Fig. 1) was radiocarbon dated to 1350 ± 60 BP (Beta-72676) [9]. A second species, H. australis that was larger than H. antimena, has been recorded in Andrahomana Cave, but is now extinct across its former range [9]. Two radiocarbon dates are available for this taxon from the site: 4440 ± 60 BP (Beta-73370) and 1536 ± 35 BP (NZA-18996, R-28421/1) [3] and [9]. These data indicate that both taxa occurred in southern Madagascar until relatively recent times.

On the basis of these examples, different rodent species once occurred across the southern portion of the island, presumably in distinctly more mesic conditions than found today. As the region experienced aridification over the past few thousand years and local natural forest conditions changed, there was a notable contraction of their ranges. In some cases, remnant populations occur in western forests (H. antimena and M. petteri) and in at least one case a taxon is completely extirpated (H. australis). Certainly human modifications of the remaining forests in the core area of these species' ranges have added to their decline, and another important factor may have been the ecological success of introduced rodents. If the temporal scenarios postulated here are correct, the remnant populations of M. petteri and H. antimena are living relicts of a recent period when southern Madagascar had moister conditions.