Survey effortnext section
Sampling techniques varied in efficiency. All species were detected during either opportunistic or transect searching and we consider this to be the most efficient survey methods. Pitfall trapping contributed the fewest number of specimens and proved ineffective at capturing amphibians, which are often proficient climbers or strong jumpers, enabling them to escape; the technique was more useful for the detection of fossorial reptile species, although all species we detected through pitfall trapping were also identified through other sampling techniques. However, in 2000 they enabled the discovery of the so far only known specimen of Pseudioacontias menamainty. Automated acoustic recording allowed for the rapid detection of amphibian species within a habitat; however as amphibian vocalisations are usually limited to the breeding period (Glaw and Vences, 2007) seasonally in-active species will have been missed. Existing audio reference recordings were required to correctly pair a vocalisation to a species during the analysis stage, thus automated acoustic recording must be used in tandem with other sampling methods to avoid missing the vocalisations of undescribed or unknown taxa. Thus, in our opinion this technique is more suited to habitat surveys for areas where most species are already known, rather than species inventories in limited surveyed areas. Despite providing no unique species records compared to the other techniques, unlike pitfall trapping, it required minimal field effort and enabled the expansion of monitoring to areas that may otherwise have been missed. We thus consider it to be a very useful tool for herpetological surveys.
Species composition of the Sahamalaza Peninsula
Although surveys always depend on contingency, it is likely that a significant proportion of Sahamalaza’s amphibian and reptile fauna have been detected, when considering all herpetological survey work of the area. The detection of three new taxa (Boophis ankarafensis, Stumpffia sp. aff. pygmaea Ca “Sahamalaza”, Geckolepis humbolti) unnoticed during the previous surveys, highlights the efficiency of using an integrative approach to species identification. The detection of several species (almost 20%) in the final few weeks of the wet season, together with the detection of species missed during a previous survey by Andreone et al. (2001), highlights the necessity of conducting herpetological surveys over extended periods for areas with strong seasonal differences.
The presence of species representative of the drier biomes of West Madagascar (e.g. Aglyptodactylus securifer, Blommersia sp. Ca05 (UCS), Heterixalus luteostriatus, H. tricolor, Laliostoma labrosum, Blaesodactylus ambonihazo, Oplurus cuvieri, Madascincus stumpffi and Zonosaurus laticaudatus concurrent with species representative of the rainforests of Sambirano region to the north (e.g. Boophis brachychir, B. jaegeri, B. tephraeomystax, Gephyromantis pseudosasper, Mantella ebenaui, Mantidactylus ulcerosus, Stumpffia gimmeli, Alluaudina bellyi, Brookesia stumpffi, B. minima, Ebenavia inunguis, Ithycyphus perineti, Phelsuma laticauda, P. vanheygeni, Paroedura oviceps, P. stumpffi, Uroplatus henkeli and U. ebenaui) confirms that Sahamalaza’s intermediate climate supports a transitional fauna between these two biomes.
The two forests of Anabohazo and Ankarafa show broadly similar species compositions with a few notable differences (Table 2). Four amphibian and ten reptile species were recorded in Anabohazo Forest but not in Ankarafa, while two amphibian and eight reptile species were found in Ankarafa but not in Anabohazo. It is likely that some of these differences only reflect bias in survey effort between the two locations. For example, several of the Gekkonidae detected from Ankarafa and not in Anabohazo: Blaesodactylus ambonihazo, E. inunguis, Phelsuma sp. aff. quadriocellata and P. vanheygeni, were likely missed due to the shorter time spent surveying this area, coupled with their infrequent to rare encounter rates. On the other hand, the species recorded in Anabohazo but not in Ankarafa are prone to have been missed, due to the positive bias in the sampling period in Ankarafa. However, the two forests fragments differ in size, habitat quality and geography and so some differences in species composition might be due to this. The two fragments are separated from one another by around 20 km of savannah and scrubland, potentially isolating many of the forest-dependent species. The far-ranging calls of G. pseudoasper were extremely conspicuous throughout Anabohazo yet entirely absent from Ankarafa. This difference cannot be attributed to season as surveys in Ankarafa took place immediately before and after the sampling period in Anabohazo. Anabohazo marks the most southerly extent of this species range (Glaw and Vences, 2007) and it is possible that the climate or geography of Ankarafa make it unsuitable for G. pseudoasper. The recently described Boophis ankarafensis was only found along perennial lotic streams in Ankarafa, a hydrological feature that is entirely absent within Anabohazo, which may explain its potential absence from here. This factor likely accounts for the non record of the helmeted turtle Pelomedusa subrufa.
All species documented from Ankarafa Forest represent new records from this locality. Several species were recorded from Sahamalaza for the first time and represent important range extensions. For Boophis tsilomaro, Cophyla berara and Blaesodactylus ambonihazo we provide the first distribution record outside of their respective type localities. In the case of Boophis tsilomaro and Cophyla berara the range expension is still limited to the Sahamalaza Peninsula. Yet, it is worthnoting that in C. berara we observe a genetic distance of 1% between the two known populations of Berara and Ankarafa. The record of Blaesodactylus ambonihazo represents a significant increase in its distributional range (extended northward by over 200 km; Bauer et al., 2011; Ikeuchi and Mori, 2014). The recent formal description of this species (Bauer et al., 2011), along with those of B. victori Ineich et al., 2016 and B. microtuberculatus Jono et al., 2015, together with the confirmed sympatry of B. victori with B. sakalava (Grandidier, 1867), and of B. microtuberculatus with B. boivini Duméril, 1856 (Jono et al., 2015; Ineich et al., 2016), further highlights the importance of applying a taxonomically integrative approach, and the need to reassess previously known localities and providing new genetic data. The distribution of B. ambonihazo may extend to other dry forests fragments in north-western Madagascar, but due to the genera’s apparent requirements for areas of relatively low disturbance containing at least some large trees, its distribution is likely to be severely fragmented (Ineich et al., 2016).
The species Acrantophis madagascariensis and Crocodylus niloticus are reported from Sahamalaza for the first time. Unfortunately, Acrantophis madagascariensis was only recorded in the 2011-2012 expedition and no genetic data are available on this record. Only a single specimen of C. niloticus was sighted, of which the tail-end was seen slipping into the water of the Vavan’aneno River in Antafiabe; local people attested the presence of multiple specimens within the area but note that the largest individuals have been lost to hunting. The presence of the snake Ithycyphus perineti, gecko Ebenavia inunguis, chameleon Brookesia minima and treefrog Boophys brachychir within Sahamalaza extend their ranges over 100 km south along Madagascar’s west coast from Nosy Be (Glaw and Vences, 2007). The population of the Ebenavia inunguis sampled in Ankarafa belong to the Clade Cb (sensu Hawlitschek et al., 2017) as the popuation from Nosy Be, that is the type locality of this taxon. These two populations have a genetic distance of 4% at the analysed COI fragment, and thus far this represent the only other record for this taxon outside of Nosy Be. Brookesia minima was known at least from Nosy Be and Manongarivo and the population sampled in this study has a genetic distance of 5% with the population from Manongarivo. Similarily, B. brachychir was already reported for Nosy Be, Manongarivo, Forêt d’Ambre and near Antsiranana. This record thus represent the southern most new distribution.
The presence of Heterixalus tricolor confirms the species’ distribution between Nosy Be and Ankarafantsika (Glaw and Vences, 2007). The presence of the turtle Pelomedusa subrufa extends their range over 200 km northeast of a record from Mahajanga (Iverson, 1992; Glaw and Vences, 2007; Petzold et al., 2014), placing this population at the northern edge of their projected distribution (Boycott and Bourquin, 2008), although no genetic distance was observed between the P. subrufa sequences of the newly reported population and the available sequences in Genbank.
We treated Phelsuma kochi as synonymous with P. madagascariensis, recorded by Andreone et al. (2001) following molecular identification; however, a photo from the earlier survey period resembles P. grandis Gray, 1870, known from the Sambirano region to the north, and it remains possible that the species occurs in sympatry with P. kochi. The occurance of P. kochi extends their range over 200 km northeast of Ankarafantsika (Mori et al., 2006; Glaw et al., 2011) and the genetic distance between the population from Sahamalaza and Tsingy de Bemaraha is of 7% at the analysed COI fragment.
This survey documents the first record of Phelsuma laticauda from Sahamalaza, a species known from a number of locations across northern Madagascar (Gelach et al., 2011). The presence of Phelsuma vanheygeni increases their known range of about 50 km south beyond the Ampasindava peninsula, where the species was classified as Endangered due to their restricted range (Randrianantoandro et al., 2011). The presence of Phelsuma sp. aff. quadriocellata marks their only documented occurrence in north western Madagascar and a significant distance from the populations known from Eastern Madagascar (Glaw and Vences, 2007). Furthermore, individual’s from Sahamalaza occurred at heights of 150-170 m asl, significantly lower than the mid-elevation areas of 720-1350 m asl where the species is generally reported in the East (Glaw and Vences, 2011). The individuals encountered were found residing in Pandanus screw palms, a trait shared with Phelsuma quadriocellata (Peters 1883), however it is unknown whether they are truly conspecific as genetic data are not available. Their rare encounter rate from Sahamalaza may mean they have been missed by other surveys and indicate the species occurs between these distant sites; alternatively, they may belong to a different P. species. The species may be synonymous with Phelsuma cf. quadriocellata reported from Nosy Be (Andreone et al., 2003). The record of Madascincus stumpffi in Sahamalaza, similar with the record from Marojejy, mark the southernost distributional record for the species, but the population from Sahamalaza have a genetic distrace of 9% at the analysed COI fragment if compared with the population of M. stumpffi of Forest d’Ambre.
The presence of the frog Blommersia sp. Ca05 (UCS) marks a range increase of over 300 km beyond Tsingy de Bemaraha. Populations are also known from Isalo, Makay and Kirindy, while recent records of a Blommersia species from Mariarano and Mitsinjo near the Besiboka delta may also be attributed to B. sp. Ca05 (Rakotoarison et al., 2015), potentially indicating the species is widely distributed along the Madagascar’s west coast. Finally, we report a new record of the recently resurrected Geckolepis humbolti which, in Madagascar, was until now known only in the Tsingy de Bemaraha. The newly reported population of Geckolepis humbolti from Sahamalaza has a genetic distance of 8-9% with the populations from the Comoros and Mayotte.
Two new species of amphibians (B. ankarafensis and Stumpffia sp. aff. pygmaea Ca “Sahamalaza” (UCS)) identified in this survey and four species identified in prior surveys (Boophis tsilomaro, Cophyla berara, Rhombophryne sp. and Platypelis sp.) may represent local endemics as they have not been detected in other surveys of Northwest Madagascar (e.g. Nosy Be, Manongarivo, Tsaratanana, Benavony), in some cases, despite their prominent and distinctive calls (Vences et al., 2005b, 2010b; Glaw and Vences, 2007). The failure to detect neither Platypelis sp. nor Rhombophryne sp. during the most recent surveys mean that further effort should be invested in the area, as representatives of these genera are sometimes very difficult to detect. The population of Lygodactylus tolampyae from Sahamalaza was already known, however this population has a high genetic divergence with the other population of this species for which genetic data are available. A more in depth taxonomic revision of this genus is needed to apply this name to a specific taxon, until then it will not be possible to assess the taxonomic identification of the Lygodactylus tolampyae population from Sahamalaza. However, this might represent a new microendemic species of retile along with the previously identified and highly elusive skink Pseudoacontias menamainty. All this points towards the Sahamalaza peninsula being an important centre of microendemicity.
The new species of treefrog, Boophis ankarafensis, was described following the results of this survey (Penny et al., 2014). The species is only known from the banks of perennial streams in intact forest vegetation in Ankarafa Forest and has been classified as Critically Endangered on the IUCN Red List. The new candidate species, Stumpffia sp. aff. pygmaea Ca “Sahamalaza” (UCS), still awaits formal description but molecular data found only a 92-93% match (p-distance transformed into percent; at the analysed 16S fragment) with S. pygmaea and their taxonomic distinctness seems therefore to be granted. The species produces inconspicuous calls from within leaf-litter which are difficult to locate, thus the species may have been missed during surveys outside the peninsula. On the contary, this is such a small amphibian species that dispersal capacities might be very low. Our survey expands the range of Boophis tsilomaro beyond their type locality of Berara. The detection of B. tsilomaro from Anketsakely, a fragment of forest within Anabohazo, contributes only a marginal increase in range, and the species is confined to an area of less than 5 km2, qualifying it as Critically Endangered. The species’ absence from Ankarafa Forest, the only other significant area of forest on the peninsula, reinforces the importance of protecting all remaining areas of natural habitat in Sahamalaza, as populations may be reliant on particular conditions.
This survey expands the range of Cophyla berara beyond their type locality of Berara: a fragment of primary forest in Anabohazo (Vences et al., 2005b). We document the species throughout the fragments of Ankarafa Forest, the surroundings of Antafiabe village and the fragment of Anketsakely in Anabohazo Forest. These locations are no greater than 20 km distant from the type locality, yet mark an important extension to the distribution of this species and indicate multiple populations exist. Furthermore, C. berara were found in abundance in low quality secondary forest, a habitat common throughout the peninsula. Secondary tracts of regenerating forest are one of the most common forest types in Ankarafa and past land clearances have created a matrix of interlinked forest fragments surrounded by large thickets of bamboo. C. berara were extremely abundant in these forest edge habitats, and in interior sections where bamboo were present, a habit also reported in C. maharipeo (Rakotoarison et al., 2015). This association is likely due to their breeding habitat of laying spawn inside water-filled segments of bamboo. The species was detected in all surveyed forest fragments, including isolated sections of heavily degraded forest that had experienced recent burning; callers were also found perched on scorched leaves and branches. Thus, this species seems to be adapted to disturbed forest, and is less likely to experience severe decline in the immediate future. However, its long-term viability in these small isolated forest fragments is unknown and even with these new range extensions, it is still known from just three areas within the Sahamalaza Peninsula, which itself totals around 26000 hectares. There appears to be limited gene flow between populations in Ankarafa and Anababohazo and molecular analyses show they have already slightly diverged, with two fix substitutions at the analysed mitochondrial 16S fragment (Penny et al., 2016). Although the species appears relatively well adapted to disturbed forest, it is still a forest-dependent species and at risk from future habitat destruction.
Threats and conservation
Forest on the peninsula continues to be exploited by the human populations. Fire has already destroyed Analavory Forest, leaving Ankarafa and Anabohazo the largest areas of intact forest in Sahamalaza. These two locations are subject to high levels of forest clearance to make way for crop cultivation and pastureland (Penny et al., 2014, 2016). Furthermore, fires lit in the dry season to rejuvenate grazing land frequently spread out of control and burn adjacent areas of intact forest. Selective logging of tropical hardwoods and small-scale quarrying were also observed in Anabohazo Forest, although currently this is still the more intact of the two remaining forests. If actions, such as those outlined by a recently published conservation action plan on the amphibians of Sahamalaza (Penny et al., 2016) are not promptly implemented, then all the peninsula’s forest dwelling herpetofauna will suffer serious population declines and the local endemics will be pushed towards extinction.