The crowned sifaka (Propithecus coronatus) is Endangered. It has a large but highly fragmented distribution; its known range extends from the Betsiboka River in the north of Madagascar, to the Mahavavy River in the north-west, and down to the Tsiribihina River in the south-west. The species lives in forest habitats that are highly and increasingly fragmented and are continuously suffering perturbations and destruction. In order to carry out effective conservation measures targeting P. coronatus, its conservation status needs to be updated so that measures can be taken before anthropogenic or natural environmental changes lead to the extirpation of the species in most of its forests. We (i) identified forest fragments where the species is still present and (ii) using the line-transect “Distance” sampling method, estimated the population size and density in the principal remaining forest fragments in the northern part of its range, including both protected and unprotected areas. We visited most of the forests in the northern part of its range in order to update the current area of occupancy, and to rate the state of its forests using a qualitative “forest quality index.” Our survey results have shown that (i) a large number of forests have disappeared or decreased in size in the last 10 years, and (ii) population densities vary considerably among forest fragments (ranging from 49 to 309 individuals per km2), with some very high densities in forests located along the Mahavavy River and in the Antrema area. Their abundance in the area surveyed is likely to be between 4,226 and 36,672 individuals, and most probably above 10,000. It is difficult to extrapolate from these estimates to the total abundance across the species' entire range, but we estimate that it is likely to be large, probably between 130,000 and 220,000 individuals. Unfortunately, many field observations suggest that its populations continue to decline at a high rate due to habitat loss and hunting, and we argue for the re-evaluation of the conservation status from Endangered A2cd to Endangered A4acd, and the need to survey the rest of the range of P. coronatus.
Crowned sifakas (Propithecus coronatus) are diurnal lemurs, inhabiting mainly dry deciduous forests and mangroves (Petter and Andriatsarafara 1987). Neither the distribution of P. coronatus nor its total population size are well known (Mittermeier et al. 2010). Its distribution was first shown to encompass the north-west of Madagascar between the Betsiboka (which separates it from Coquerel's sifaka Pwcoquereli) and Mahavavy (where it is believed to hybridize with Decken’s sifaka, P. deckenii) rivers (Kaudern 1915). Nevertheless, early work from the 1929–1931 Archbold Expedition reported the presence of melanistic individuals in the Bongolava population of P. deckenii (Tattersall 1986), which might be attributed to P. coronatus, and Paulian (1953) also reported the presence of P. coronatus near Tsiroanomandidy, south of the Manambolo River (in Wilmé et al. 2006). Later, Petter and Andriatsafara (1987) reported the past presence of P. coronatus further to the east, in the Ambohitantely Special Reserve, but this record may have arisen from an error in translation (Rakotonirina et al. this issue) leaving unresolved its past presence in the region east of the Ikopa River (Rakotonirina et al. this issue). Thalmann and Rakotoarison (1994) reported its occurrence to the south of the Manambolo River (south of the putative distribution of P. deckenii). All these studies thus suggested that P. coronatus might have a distribution much larger than was previously thought. In particular, these authors proposed for the first time a geographic range that would include the inter-river systems (IRS) between the Betsiboka and Mahavavy rivers in the north-west (corresponding to IRS1 and IRS4 in Fig. 1) and between the Manambolo and Tsiribihina rivers in the central-west. This distribution would thus surround that of P deckenii, which would then be restricted to the IRS2 and IRS3 areas (Fig. 1) with some contact zones along the main rivers and in the Bongolava region. This hypothetical geographic range has not yet been entirely validated because of the remoteness of these regions, and problems of security when visiting them. Razafindramanana and Rasamimanana (2010) extended the species' range in Dabolava and Miandrivazo, to the south of the Mahajilo River, suggesting as a result that the Mania River should be the northern limit of P. verreauxi and should correspond to the extreme southern limit of P. coronatus. This supported the suggestions of Thalmann and Rakotoarison (1994) and Wilmé and Callmander (2006). Finally Rakotonirina et at. (this issue) recently conducted a widespread survey and confirmed the presence of P. coronatus to the west of the Ikopa River and to the north of the Mahajilo River, again confirming previous hypotheses concerning its range limits.
While most presence-absence studies of P. coronatus have been carried out in the north-west (Curtis et at. 1998; Müller et al. 2000 in Katsepy, Anjamena, and Anaborengy), it should be noted that some fragments of the southern area of the IRS1, for example, Andranovelona/Madirovalo (Rasoloharijaona et al. 2005), Madirolavo (change to: Sussman 1977 in Wilmé et al. 2006) have been visited. Most of the forest fragments of this extended and putative geographic range, however, have not yet been surveyed (Fig. 1), and only a few studies have been carried out to estimate P. coronatus population densities. Moreover, these studies have produced very different figures.
To our knowledge, three studies have estimated P. coronatus densities in Anjamena. Müller (1997) estimated a very high density of 543 ind/km2, based on home range size. Curtis et al. (1998) reported density estimates of 32 ind/km2 (a value 17 times smaller than that of Müller 1997), whereas Müller et al. (2000) reported densities of 173 ind/km2. In Katsepy, Curtis et al. (1998) found densities of 5 ind/km2, whereas Pichon et al. (2010), using long-term survey data, estimated a minimum of 300 ind/km2 in the Badrala forest of the Antrema Forest Station close to Katsepy.
Due to their matriarchal social system, with groups usually composed of two to eight individuals, and their relatively long generation time (probably between 6 and 15 years, based on data from Propithecus verreauxi; Richard et al. 2002; Lawler et al. 2007), the recovery of small isolated populations is likely to be difficult both from a demographic and genetic point of view. The lack of consistent density estimates, together with the limited number of studies and regions surveyed, and the huge threats imposed on primates and sifakas across Madagascar, led to the classification of P. coronatus as Critically Endangered in 1996 (Baillie and Groombridge 1996, in IUCN 2008). Its status was updated in 2008 and considered Endangered A2cd (IUCN 2008) as it was thought to have undergone a reduction of more than 50% over the past three generations or 30 years (assuming a generation length of 10 years), due primarily to a decline in area and quality of habitat within its known range, but also due to hunting. Its status as Endangered was reaffirmed in a Red Listing Workshop held in Antananarivo, Madagascar, in 2012.
In this study, we collected presence-absence survey results for 70 forest fragments and 12 corridors and gallery forests around 19 sites in the north-west of the Betsiboka-Mahavavy IRS. Furthermore, we provide here new density estimates of P. coronatus for six localities. Using density and available GIS data on forest cover, we estimated the area of occupancy and the total population size of the species. Finally, we discuss a possible update to the conservation status of P. coronatus.
Presence-absence survey and human impact assessment
To detect the presence of P. coronatus and compute a “quality index” of the forest fragments in the northern part of its range, we visited 70 forest fragments and 12 forest corridors, in the vicinities of 19 sites (villages or chief towns) between the Betsiboka and Mahavavy rivers during two field seasons: from July to October 2009, and April to July 2010 (Table 1; Fig. 2).
The first surveys were in three protected areas. We surveyed the forests neighboring the Kingany, Boeny Ampasy, Boeny Aranta, Antsilaiza, Anaborengy, Ambohibary, Antani-malandy and Ankarahara localities in the Mahavavy-Kinkony complex managed by the Malagasy NGO Asity. The forests around Kingany and Boeny Ampasy are mainly small, dry semi-deciduous, forest fragments. The forest fragments surrounding the Mataitromby locality are in the Bombetoka-Belemboka protected area managed by the Malagasy NGO Fanamby. Forests visited around Antrema, Katsepy, Masokohamena and Ambanjabe are in the Antrema Forest Station protected area, which contains three of the typical north-western ecosystems (dry semi-deciduous forest, mangrove swamp, savanna), which suffer moderate anthropogenic pressure. We also visited unprotected forest fragments around Androhibe, Ankarabato and Antsalatsala.
In each survey site, we walked slowly and quietly in the forests or on its edges in order to detect the presence of P. coronatus. When a group of sifaka was found, its size and composition was estimated and GPS coordinates were recorded. The general human impact on the forest was estimated qualitatively: fire residues, logging, evidence of poaching, forest clearing and charcoal ovens were registered when observed.
Population densities and total species abundance
To study variation in the density of P. coronatus among forest patches, we carried out line transect sampling surveys, following the distance sampling methodology (Peres 1999; Buckland et al. 2001), in six of the principal forest fragments in the northern part of its range: Ambohibary, Antsilaiza, Antsoherikely, Antanimalandy, Ankarahara, located along the Mahavavy River (Fig. 2); and Antrema in the north along the Mozambique Channel. The selection of these forests was motivated mainly by their size and characteristics, on the basis that distance sampling methods require a minimum number of observations to provide reliable estimates. Density surveys were thus not performed in small fragments and corridors. All were of lowland dry semi-deciduous forest, and suffered different levels of human disturbance.
Coordinates, characteristics and number of visited forests during crowned sifaka surveys.
Our census took place during the dry season; from July to October 2009 (four months) and April to July 2010 (four months). Between three and five line-transects were randomly delineated at each site; 20 line-transects in total. GPS coordinates were recorded every 20 m along each transect. The transects ranged from 800 m to 3,500 m in length. They were surveyed 6–10 times during 3–5 days by three 2-member teams, to achieve at least 40 observations, as recommended by Peres (1999). Every day, one team member changed teams and transects to avoid observational biases among teams and to ensure that at least one team member had already walked that transect (Quéméré et al. 2010).
On seeing a sifaka group we collected the following data: date, time, transect number, group size, group spread, and sighting distance and angle to the center of the group (to compute perpendicular sighting distance). We then estimated sifaka densities (ind/km2) in each fragment using the DISTANCE 6.0 software (Thomas et al. 2010). In this method, the surveyed area corresponded to the product of the total survey effort per fragment (km) and the effective sighting width (ESW). The ESW is estimated using a calculation of the decreasing probability of seeing an animal as a function of its distance from the transect. Various functions can be used to model this probability and estimate ESW. Here we tested the uniform, hazard rate, and half normal models with cosine, polynomial and Hermite adjustments and compared them using the Akaike Information Criterion (AIC) as recommended by Buckland et al. (2001).
The global P. coronatus distribution was determined using all available P. coronatus observations (Sussman 1977; Tattersall 1986; Petter and Andriatsarafana 1987; Thalmann and Rakotoarison 1994; Curtis et al. 1998; Müller et al. 2000; Thalmann et al. 2002; Rasoloharijaona et al. 2005; Wilmé et al. 2006; Razafindramanana and Rasamimanana 2010; Rakotonirina et al. this issue). Combined, the published data argue for a wide-ranging distribution of the species delimitated by the Betsiboka and Ikopa rivers in the east, by the Mahavavy and Manambolo rivers in the west, and by the Tsiribihina and Mania rivers in the south (Fig. 1). Its occurrence between the Mahavavy and the Manambolo, and between the Mania and the Ikopa rivers has been defined approximately, without clear observational data, and needs thus to be confirmed. Melanistic variants of P. deckenii in Ambohijanahary and Kasijy have not been taken into account to delimit the probable range of P. coronatus (see Rakotonirina et al. this issue). As the report of the past presence of crowned sifaka in Ambohitantely now appears to have been a translation error (Rakotonirina et al. this issue), and as these authors (this issue) reported their absence in Bekirobo, the area between the Ikopa and the Betsiboka rivers was not included. The extent of suitable habitat across the range of P. coronatus was calculated using forest classification from the Madagascar Vegetation Mapping Project data (available online at < http://www.kew.org/ gis/projects/mad_veg/datasets.html>; Moat and Smith 2007) from 1999 and 2000 satellite images, and MEFT-USAID-CI (2009) from 2005 satellite images classification on ArcMap software (ESRI). To estimate the total species abundance, we multiplied the minimum and average density estimates by the area of occupancy, which was obtained using the two different GIS (Geographic Information System) data sets (Moat and Smith 2007; MEFT-USAID-CI 2009).
In all, 331 discrete social groups were sighted during the presence-absence surveys. They comprised a total of 1,234 individuals (adults only) with an average group size of 3.6 (Table 2) during a 169 day × people survey effort. Ninety-eight groups had newborn offspring (29.6% of the groups seen). Sifakas were not found in the forest fragment surrounding the village of Anaborengy, and around localities in the south-east of the survey area, from Ankarabato to Antsalatsala. Time spent in the latter area was short but this is known to be a “Dahalo” (Zebu thief) area and most of the forests were burned or burning during our visit (August 2010). Few observations were made in Anjamena and in the Boeny-Kingany region. Most forests had been cleared around Anjamena, and we therefore spent little time in this area. The Boeny-Kingany region is composed of small and highly fragmented forests. Most of those neighboring Boeny Ampasy had been burned a few years before, and the sifaka encounter rate in the regenerating fragments was lower than one group per day. To our surprise, we found large numbers of sifaka groups in the Mataitromby and Androhibe forest corridors along small rivers located between the Betsiboka and Mahavavy rivers.
Population density and size
We surveyed a total of 220 km and made 444 sightings of social groups comprising 1,753 individuals (note that these numbers do not always correspond to distinct groups or individuals, as each transect was repeated more than once). In agreement with surveys elsewhere (Plumptre and Reynolds 1994) including those for golden-crowned sifakas (Propithecus tattersalli) (Quéméré et al. 2010), the hazard-rate model was identified as the best fit for our data in all fragments. We gathered enough observations to accurately compute density estimates for five of the six survey sites (Table 3). Only 21 social groups (71 individuals) were observed in Antanimalandy despite four days of census and a total of 12.71 km of survey effort. This limited number of observations (as a comparison, 156 groups corresponding to 281 individuals were observed in Antsoherikely in 2009 for 52.13 km surveyed, i.e., twice as many per km) did not allow us to accurately compute the ESW (Buckland et al. 2001). Nevertheless, we provide an estimate of sifaka density using the ESW estimated by Distance on the basis that the ESW estimates were unlikely to be much greater or smaller than for other sites. Results for Antanimalandy should, however, be regarded with caution; indeed confidence intervals for this forest are very large (9–714) and little informative. Density estimates in all fragments range from 46 ind/km2 in Ankarahara to 255 ind/km2 in Salmona et al. Antrema (Table 3). We found lower ESW and higher densities for the census carried out in Antsoherikely in 2009 (ESW = 21.4 and D = 309) than for the censuses made in 2010 (ESW = 33.2, D = 75). This discrepancy was surprising at first but a closer look at the data suggests that the higher ESW values in 2010 result from several observations of groups located on the edges of neighboring fragments (i.e., across open habitat) at distances of 30 to 70 m. When the results obtained for the same transect in 2009 and 2010 were compared, they were actually very similar, suggesting that the high density results are still valid here (data not shown).
Propithecus coronatus presence-absence survey results and main threats observed.
Propithecus coronatus density data.
Estimated area of occupancy of the crowned sifaka and the total population size.
The two different GIS datasets showed some discrepancies. We found that the 1999–2000 Kew Garden GIS (Moat and Smith 2007) dataset tended to overestimate forest areas, based on our field observations. This could be due to the fact that data were obtained from 1999 and 2000 Landsat images. On the other hand, the 2005 MEFT-USAID-CI (2009) GIS dataset was found, based on our field observations, to under-estimate the size of forests in several cases. Consequently, we used both GIS datasets in order to provide low and high population size estimates.
Using the lowest density estimate (Ankarahara, 46 ind/km2) and applying it to the available habitat calculated using USAID GIS data of 2005 (2,353 km2), indicates a minimum estimate of the total number of P. coronatus across the whole forested area of 115,325 individuals (Table 4). If we use the average value of density (171 ind/km2), we obtain 402,460 individuals (Table 4). When performing the estimation with the Kew Garden GIS data of 1999–2000, we obtain population sizes of 203,285 and 709,422 individuals, using the minimum and average densities, respectively. These values should be regarded with caution as there are many uncertainties regarding the calculations, but they probably represent the best available estimates for the global abundance of P. coronatus, at the time that the GIS data sets were built. Given that the highest densities may be due to the concentration of P. coronatus individuals in the remaining forests, as a consequence of forest loss, a likely figure is possibly closer to our lower estimate; around 100,000 sifakas.
Population density and size
The density of P. coronatus was shown to vary among fragments, from a low of 46 ind/km2 to a high of 309 ind/km2. Previous studies published by different authors found values that were even more variable, with values of 5 ind/km2 to 500 ind/km2 (Curtis et al. 1998 and Müller 1997, respectively). How all these values compare is difficult to say since they were produced using different methods. The discrepancies between our density estimates and those of Curtis (1998), who found 5 ind/km2 and 32 ind/km2 in Katsepy and Anjamena, respectively, could be explained by the different methodologies. Some other previous studies appear to produce more reasonable estimates. For instance, Müller et al. (2000) found density values similar to ours in the same area. Moreover, our results for Antrema (255 ind/km2) are on the same order as those estimated by Pichon et al. (2010; >300 ind/km2) also in Antrema. It is worth noting that we combined the Badrala forest (surveyed by Pichon et al. 2010) together with a forest located east of Antrema next to the Katsepy lighthouse. When we performed the distance analysis using only the Badrala forest, we also obtained a density estimate of 350 ind/km2, thereby confirming consistency between Pichon et al. (2010) and our results. The fact that our study was performed in different habitats, some of which were suitable for sifakas (Antsilaiza, Antsoherikely, Antrema-Badrala) and others much less so (i.e., secondary, degraded, or partly cleared forest; Ankarahara, Antsoherikely, Antrema-Katsepy), suggests that our estimates are not major overestimates of population densities for P. coronatus.
When we compare our estimates with those published for other sifaka species, we also find that our results fit reasonably well (Table 5). If we exclude the case of P. perrieri, one of the most endangered primates of the world with a density of 3.11 ind/km2 (Banks et al. 2007), the densities published for P. verreauxi (Kelley et al. 2007) and P. tattersalli (Quéméré et al. 2010; Table 5) are also of the same order as those obtained here.
Altogether this suggests that estimates of abundance can be reasonably drawn from our density calculations. As we see below, there are, however, many uncertainties, which still require some caution. By extrapolating our results to the likely range of P. coronatus, we found that the total abundance probably ranged between 115,325 and 402,425 individuals. Even if we limit ourselves to the six surveyed localities, which represent only a small part of the total geographic range of the species, we find a total of 4,226 to 14,747 individuals with the USAID data and between 10,508 and 36,672 individuals with the Kew Garden data. Moreover, the presence-absence survey showed that there were at least 1,234 independent individuals.
To estimate the area of occupancy of crowned sifakas we took into account all identified forests. This could lead to an overestimation of the total population size since some may not be large enough to host crowned sifaka. The environmental conditions also vary considerably between the northern and southern parts of this broad geographic range. Total size estimates are only based on north-western density estimates and could thus be biased towards lower or higher density in the southern region.
Total population size estimates between 100,000 and 400,000 appear to be very high, but if we compare them to the recent estimates of P. tattersalli (>11,000 and probably around 18,000 individuals, Quéméré et al. 2010; Table 5), a Critically Endangered sifaka with an area of occupancy less than one tenth that of P. coronatus, the new estimates appear more reasonable. Nevertheless, it is important that they should be confirmed by field work carried out in the regions that have until now been little visited. Until then, it might be more reasonable to first consider the estimates which correspond to the region that we have actually studied, i.e., a total abundance between approximately 4,000 and 36,000 individuals.
Most of the remaining geographic range of P. coronatus is to a large extent unexplored and lacks even basic data on the presence or absence of populations. Moreover, the areas not explored here are mostly unprotected, with the exception of the recently established community-based conservation program around the Dabolava-Miandrivazo region. The Betsiboka-Ikopa region remains to be more carefully surveyed in order to determine if the species was ever or is still present. Furthermore, the range of P. coronatus remains to be clarified between the Manambolo and Tsiribihina rivers, where only one study has reported its presence (Thalmann and Rakotoarison 1994) but where P verreauxi was also surprisingly reported (ZICOMA 1998, in Wilmé et al. 2006).
Sifaka density and population size estimates in the literature.
Conservation status update for crowned sifaka (Propithecus coronatus).
We should also stress here that most of those areas are located in “dangerous” zones due to the presence of “Dahalo” (Zebu thieves) groups and are furthermore difficult to reach. Given that these regions may harbor the majority of the species' population, the figures presented here could be over-optimistic if it was found in the future that most crowned sifakas have actually been hunted or burnt along with the forests. A long-term conservation strategy incorporating extended monitoring will require surveys to be carried out in these regions, but safety concerns may make such surveys difficult to implement in the near future.
Implication for conservation
During our survey many blowpipe darts were found in the forests neighboring the Mahavavy River. One of these darts was spotted with blood and had white hair stuck on it, thus strongly suggesting that it was used for hunting sifaka. While local populations do not normally hunt sifakas as they are protected by local taboos (“fady”), some people do not necessarily adopt this taboo and may still hunt them. In Antanimalandy, local people repeatedly mentioned the events of the 2008 dry season, when a group of a dozen of hunters came to their village, hired guides and hunted dozens of sifakas every day during one week in the Ankarahara area. This may explain the low densities estimated there in comparison to other neighboring localities (Ambohibary and Antsoherikely). In Mataitromby, sifaka hunting also occurred in 2008 and was apparently and surprisingly carried out by armed forces. Between Ankarabato and Antsalatsala, we found no inhabited villages due to the presence of “Dahalo,” and the remaining forest fragments had been burned or were burning during our visit. More recently, in 2010, the two sifaka groups living next to the Katsepy lighthouse were hunted by “unknown” military men (Peace Corps pers. comm.). Finally, the Tsiamarakely and Tsiamarabe forests in the south of Boeny Aranta had burned a few years before our 2010 visit and were no longer suitable for crowned sifaka and we only found a few individuals. These observations are important in the way that they are testimonies of the threats against crowned sifaka populations and the voluntary or involuntary ignorance of existing regulations.
Finally, using the bibliographical data, our surveys, and the Kew Garden and USAID GIS data we estimated the crowned sifaka area of occupancy to be between 4,493 km2 and 2,690 km2, respectively, hence confirming that the species should be maintained as Endangered on the basis of the IUCN B1 criterion. We argue, however, that the conservation status of crowned sifakas should be modified from “En A2cd” to “En A4acd” (Table 6). Indeed the current A2cd status is based mainly on assumptions that are difficult to verify (i.e., a reduction of the population size of 50% in the last 10 years or 3 generations). Given that the deforestation rate across the known area of occupancy of P. coronatus was of ∼11% between 1990 and 2005 (calculated using CI/USAID deforestation analysis, MEFT, USAID, CI. 2009), and the long generation time recently suggested by Lawler et al. (2007) for P. verreauxi, this suggests that the population probably decreased by 20–30% in the last 3 generations. Nevertheless the Endangered A4 status is also warranted if a species is suspected to have undergone a reduction of 50% considering both recent past and present rates. Considering that both deforestation (for charcoal production and timber export) and hunting rates have significantly increased after the 2009 political events, it seems reasonable to suspect that crowned sifaka's populations have unfortunately undergone and will undergo a decline of more than 50% in the ongoing three generations. This proposition (the change from “En A2cd” to “En A4acd”) was recently presented and approved at the 2012 IUCN/SSC Primate Specialist Group Lemur Red-Listing and Conservation-Planning Workshop held in Antananarivo in July 2012.
Our results suggest that there are more P. coronatus individuals across the whole geographic range than was previously thought. Although the exact number is difficult to estimate with certainty, it seems reasonable to indicate that it is probably above 10,000 in the northern part of its distribution and possibly around (and larger than) 100,000 across the species' range. While these figures are higher than originally expected, it is important to note that they represent an extrapolation across the crowned sifaka distribution range on the basis of results obtained in the northern part of its range, using only forest cover and ignoring possible changes in forest cover related to climate change. Finally, using the bibliographical data, survey and GIS data we argue for a change of conservation status criteria for the crowned sifaka from “En A2cd” to “En A4acd”.
We thank CAFF/CORE, the ‘Direction Générale des Environnement et Forêts’ and Madagascar National Parks/ANGAP for giving us permission to conduct this study. Financial support was provided by the Fundação para a Ciência e a Tecnologia [FCT, ref. PTDC/BIA-BEC/100176/2008, and SFRH/BD/64875/2009], the Institut Français de la Biodiversité, Programme Biodiversité de l’Océan Indien [ref. CD-AOOI-07-003], the GDRI Madagascar, the «Laboratoire d’Excellence (LABEX)» entitled TULIP (ANR -10-LABX-41), the Instituto Gulbenkian de Ciência, and the “Optimus Alive!” Biodiversity grant. The field work was possible thanks to the continuous support of the University of Mahajanga, the Département de Biologie Animale et Ecologie, the Fanamby NGO [in particular S. Rajaobelina and B.D. Razafindrakoto]. We also acknowledge R. Rabarisoa and C. Nirina from Asity Madagascar, and C.A. Gauthier and V. Vavindraza from the Antrema SFUM for allowing us to visit their conservation sites. We also thank J. Razafindramanana for actively involving us in this project and M. Irwin, C. Schwitzer, T. King and A. B. Rylands for the constructive reviews on the manuscript and R. Zaranaina and A. Finomana who actively took part in the field work. This study was made in accordance with the laws of the countries of Portugal, France and Madagascar.