A Bat on the Brink? a Range-Wide Survey of the Critically Endangered Livingstone's Fruit Bat Pteropus Livingstonii

A bat on the brink? A range-wide survey of the Critically Endangered Livingstone's fruit bat Pteropus livingstonii

B RONWEN M. DANIEL,KATHLEEN E. GREEN,HUGH D OULTON D ANIEL M OHAMED S ALIM,ISHAKA S AID,MICHAEL H UDSON J EFF S. DAWSON,RICHARD P. YOUNG and A MELAID H OUMADI

Abstract The Livingstone’s fruit bat Pteropus livingstonii Keywords Comoros, Comorian flying fox, Critically is endemic to the small islands of Anjouan and Mohéli in Endangered, deforestation, old world fruit bat, population the Comoros archipelago, Indian Ocean. The species is survey, Petropus livingstonii, western Indian Ocean under threat from anthropogenic pressure on the little that remains of its forest habitat, now restricted to the islands’ upper elevations and steepest slopes. We report the results of the most comprehensive survey of this species to Introduction date, and present recommendations for ongoing field conservation efforts and monitoring. Morning counts were ruit bats of the genus Pteropus are recognized as being conducted at roost sites in the wet and dry seasons during Fimportant pollinators and seed dispersers (Cox et al.,    –. Habitat structure around the roosting sites was ; Fujita & Tuttle, ; Pierson & Rainey, ), especially characterized and roost numbers compared, to investigate on isolated island systems with a low abundance of pollinator   the potential effect of habitat loss and degradation. We esti- fauna (Cox et al., ; Carroll & Feistner, ; Allen-Wardell   mate the population to comprise c. , individuals distrib- et al., ; Mohd-Azlan et al., ), and in fragmented sys-   uted across  roosts on the two islands. All occupied tems (Bollen & Van Elsacker, ; Nyhagen et al., ). The roosting sites were restricted to a narrow altitudinal range, loss of fruit bat populations, particularly of endemic species, is and roosting populations in agroforestry areas were smaller of conservation concern as it is thought to lead to the loss of than those found in degraded and undisturbed forest. Only endemic plants and a decrease in plant biodiversity   one of the  roosts on Anjouan was found in undisturbed, (Allen-Wardell et al., ; Mohd-Azlan et al., ;Bollen    old-growth forest with no nearby signs of clearance for et al., ; Nyhagen et al., ;Jenkinsetal., ). agriculture or landslides following tree-felling upslope. Many Pteropus species are under threat, mainly from Following a suspected severe population decline as a result habitat loss and hunting, both for subsistence use and for    of widespread and long-term forest loss Livingstone’s fruit sale (Jenkins et al., ; Epstein et al., ). Of the  bat has been recategorized as Critically Endangered on the Pteropus species on the IUCN Red List (IUCN, ), four IUCN Red List. have already gone extinct, five are Critically Endangered, five are Endangered and  are Vulnerable. Many Pteropus species are limited to islands in the Pacific and Indian Oceans, where typhoons and cyclones also pose a threat to the integrity of the remaining forest fragments (Pierson & BRONWEN M. DANIEL,MICHAEL HUDSON*, JEFF S. DAWSON and RICHARD P. YOUNG† Rainey, ; Carroll & Feistner, ; Mickleburgh et al., (Corresponding author), Durrell Wildlife Conservation Trust, Les Augres ;Powell&Wehnelt,). As ecologically specialized is- Manor, Trinity, Jersey, Channel Islands, JE3 5BP, UK E-mail [email protected] land endemics with a preference for undisturbed old-growth   KATHLEEN E. GREEN and HUGH DOULTON‡ Bristol Zoological Society, Bristol Zoo forest (Brooke, ; Mildenstein et al., ; Nyhagen et al., Gardens, Clifton, Bristol, UK ), many are particularly vulnerable to habitat loss. With a

DANIEL MOHAMED SALIM,ISHAKA SAID and AMELAID HOUMADI§ Dahari, Hombo– low reproductive rate of only one young per female per Mutsamudu, Anjouan, Union des Comores annum, Pteropus are limited in their ability to recover from *Also at: Durrell Institute of Conservation and Ecology, University of Kent, population declines (Pierson & Rainey, ). Canterbury, Kent, UK ’ †Also at: Department of Life Sciences, Imperial College London, Silwood Park At the western limit of the genus distribution, the west- Campus, Ascot, Berkshire, UK ern Indian Ocean is home to an assemblage of eight Pteropus ‡Also at: Dahari, Hombo–Mutsamudu, Anjouan, Union des Comores species (Carroll & Feistner, ; Mickleburgh et al., ). §The original version of this article was published with an author omitted from The rarest of these is thought to be the Livingstone’s fruit bat the list of authors. A notice detailing this has been published and the error rec- tified in the online and print PDF and HTML copies. Pteropus livingstonii, which is found only on the islands of Received  November . Revision requested  February . Anjouan and Mohéli in the Comoros archipelago Accepted  May . First published online  November . (Anjouan, Mohéli and Grande Comore, which comprise

the Union of Comoros, plus the French Departement of Mayotte). The species was first described by the Victorian-era explorer David Livingstone, who shot an indi- vidual on Anjouan in  (Cheke & Dahl, ). It is the only Pteropus species in the western Indian Ocean that is not threatened by hunting (Carroll & Feistner, ; Entwistle, ; Jenkins et al., ; Mickleburgh et al., ), but its population is suspected to have undergone severe declines as a result of widespread deforestation since the s (Cheke & Dahl, ; Mickleburgh et al., ; Reason & Trewhella, ), which continues today. It was estimated that forest on the Comoro Islands was lost at a rate of .% per year during –, the highest rate of deforestation of any country (FAO, ). The main causes of natural forest loss are the cutting of large trees for construction wood and the extension of the agricultural frontier, driven by un- productive agricultural systems, a high population growth rate, poverty, a lack of economic alternatives and uncertain  land tenure. Only c.  and  km of undisturbed old- growth forest remains on Anjouan and Mohéli respectively, and this is largely restricted to higher elevations and steep slopes (Green, a). The status of Livingstone’s fruit bat was recently reassessed, and as a result of this habitat loss the species is now categorized as Critically Endangered on the IUCN Red List (Sewall et al., ). A number of surveys of Livingstone’s fruit bat have been conducted since the early s by counting at known day- time roosts, with more intensive survey effort on Anjouan than Mohéli. In  a field expedition conducted the first FIG. 1 Distribution of Livingstone’s fruit bat Pteropus livingstonii survey of the species by counting bats at all known long-term roosts (numbers as in Table ) on (a) Anjouan and (b) Mohéli roosts (Trewhella et al., ). The population was estimated islands in the Comoros archipelago, in the western Indian to be as few as  individuals, found at  roost sites. Ocean. Coverage of undisturbed and degraded forest is based on a land cover map by Green (a). Following the survey a long-term community-led monitoring programme was set up, in which local people, supported most comprehensive population estimate for the species, by technicians from the national NGO Action Comores, based on a – survey of all known roosts as well as conducted monthly counts at known roosts, and searches previously unknown roosts. We characterize the condition for unknown roosting sites. By ,  active roosts were of the forest around each roost and compare this to roost known to exist across Anjouan and Mohéli and were be- size to investigate the potential effects of habitat loss and lieved to support c. , bats (Sewall et al., ). degradation on the populations at roost sites. Based on However, the reliability of these counts conducted by com- our findings we make a series of recommendations for con- munity enumerators with varying levels of training and poor servation efforts and population monitoring. access to survey equipment has not yet been assessed. Despite its threatened status Livingstone’s fruit bat re- ceives no legal protection at national level and there are cur- Study area rently no long-term forest conservation measures in place on either Anjouan or Mohéli. Forest conservation initiatives The work was conducted on the islands of Anjouan and are being developed, including the planned designation of Mohéli in the Comoros archipelago, in the Western Indian  the first terrestrial protected area in the Comoros (UNDP, Ocean (Fig. ). Anjouan is both the larger ( km ) and high- ), and the development of community-conserved er island, reaching an elevation of , m, and is characterized  areas by a local NGO, Dahari. The locations and sizes of cur- by steep ravines. Mohéli ( km ) reaches only  mand rent roosts, and proximate pressures on each roost site, need consists of a single mountainous ridge. Both islands are domi- to be quantified and the data made available to conservation nated by agricultural and agroforesty areas but Mohéli retains planners to ensure these conservation initiatives take into a greater percentage of its original forest cover (Green, a), account the needs of Livingstone’s fruit bat. We report the partly because it has a smaller human population, of c. ,.

Anjouan has a population of c. ,,anditshighpopula- the mean value taken. Canopy diameter was measured by − tion density of .  people km has contributed to high two team members standing directly under opposite edges levels of deforestation. of the canopy and measuring the distance with a tape meas- The annual climatic cycle in the Comoros consists of a cool- ure. Tree trunk diameter at breast height was measured er, drier season during May–October and a warm, humid sea- using a tape measure. The Comorian names of tree species son during November–April. To sample these two annual and any anthropogenic uses were identified by the climatic extremes, and for the data to be directly comparable Comorian field technicians and local guides. to previous surveys (Granek, ;Sewalletal.,), field The structure of the understorey vegetation was sampled work was carried out on Anjouan during the wet season in at eight locations at  and  m north, south, east and west December  and January , and repeated in January of the main roost tree. The location, list of plant species, and , and dry season counts were conducted across both description of the ground substrate (rocky, loose soil or fer- Anjouan and Mohéli in June and July . Limited project re- tile soil) within a radius of  m of each sample point was also sources meant it was not possible to conduct a wet-season recorded. Canopy cover was measured at each point using a count on Mohéli. spherical densitometer (Lemmon, ). Observations of any trees felled for planks, clearance for agriculture, or damage caused by cyclones or floods were also recorded. Methods The land cover type (based on land cover maps produced by Green, a) in which the roosts were located was iden- A team of one scientist and four field technicians surveyed all tified using the geographical information system ArcGIS known roost sites on each island in as short a time-frame as v. .. (ESRI, Redlands, USA). The land cover types were un- possible to minimize error caused by double counting as a re- disturbed forest (old-growth forest with closed canopy and sult of inter-roost movements; the mean duration of the three little trace of human impact); degraded forest (mainly closed main survey periods was  days. This included a total of  canopy consisting of native species but with abundant signs sites on Anjouan identified previously (Sewall et al., )as of human pressure, such as selective logging and/or clearing having been used historically by Livingstone’s fruit bats for of understorey vegetation; some presence of non-native trees roosting, and six on Mohéli (Table ). Local reports of previ- or crops); agroforestry (agroforestry species with natural or ously unidentified roosts were investigated by intensive non-natural understorey); and plantation (few native and/ searches of relevant areas of forest. Roost counts were carried or non-native trees present but dominated by crops). out hourly during .–. from pre-identified vantage Percentage canopy cover at the roost site, diameter at breast points, using binoculars. Vantage points were typically lo- height of the roost trees, elevation and land cover type were cated c. – m from the roost, and slightly above or included as candidate variables in the analysis. All statistical level with the crown of the trees used for roosting. Each sur- analyses were conducted in Rv... (R Development Core veyor conducted an independent count, and the results were Team, ). An initial Poisson generalized linear model was compared after the count was finished. The final number was run to investigate the relationship between the environmen- agreed following cross-checking of counts. The species of each tal variables and mean wet-season roost size in the  survey tree used for roosting was recorded (during a survey of the years. Wet-season roost counts were considered to be a su- roost site after the count was completed) along with the num- perior estimate of population size to dry-season counts, as ber of Seychelles fruit bats Pteropus seychellensis comorensis fruit bats are less dispersed throughout the landscape in (which are readily distinguishable by their smaller size) to in- the wetter part of the year (Mickleburgh et al., ; vestigate the extent of mixing at the long-term roosts of Jenkins et al., ). Goodness-of-fit tests revealed an over- Livingstone’s fruit bats. Counts were not carried out during dispersion parameter of . for the Poisson generalized lin- prolongedspellsofrainormistasthisaffectedvisibility. ear model, and therefore the standard errors were corrected Roost sites were defined as one or more trees at a discrete using a quasi-Poisson generalized linear model in which the location (typically within  m of each other) that were used variance was the dispersion parameter multiplied by the by at least one bat for roosting. To quantify the canopy mean of the dependent variable. Model selection was carried structure and the composition of tree species at roost sites, out using stepwise backward deletion of the least significant the  nearest large trees (diameter at breast height .  cm) variable until all variables retained were significant at to the main roosting tree (the tree containing most bats) α = .. were identified and measured. If the number of roosting The Extent of Occurrence (EoO) and Area of Occupancy trees was greater than , all trees that contained roosting (AoO) of the species were calculated to provide two metrics bats were sampled. The location of each tree was recorded of its geographical range, to guide reassessment for the with a global positioning system. To measure the dimen- IUCN Red List. The former was estimated by drawing a sions of the trees, the total height and height of the first minimum convex polygon (MCP) around the outermost branch were estimated by eye by each team member and known active roosts on both islands. AoO was calculated

Oryx, 2017, 51(4), 742–751 © 2016 Fauna & Flora International doi:10.1017/S0030605316000521 Downloaded from https://www.cambridge.org/core. IP address: 170.106.35.234, on 26 Sep 2021 at 23:19:59, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0030605316000521 https://www.cambridge.org/core/terms Downloaded from Oryx ’  07 14,742 51(4), 2017, , TABLE 1 Survey results from roost sites of Livingstone s fruit bats Pteropus livingstonii on the islands of Anjouan and Mohéli in the Comoros archipelago (Fig. ), with the land cover type, https://www.cambridge.org/core altitude, presence of the Seychelles fruit bat Pteropus seychellensis comorensis, and land uses and impacts within a  m radius.

Roost count (Number of trees used for roosting) Wet season Dry season Wet season Land cover Altitude Seychelles – 751 Roost name 2011–2012 2011–2012 2012–2013 type (m) fruit bat* Land uses & impacts within 100 m © . https://doi.org/10.1017/S0030605316000521 Anjouan 06Fua&FoaItrainldoi:10.1017/S0030605316000521 International Flora & Fauna 2016 1. Adda (Chava) 28 (2) 18 (2) 41 (3) Degraded 999 Extensive agriculture & felling for construction forest . IPaddress: 2. Bazimini 75 (7) 37 (5) 56 (6) Degraded 771 . 10 Encroachment of agriculture (Hamogne-Panga) forest 3. Dindri (Dremelani) 6 (2) 0 (0) 6 (2) Agroforestry 762 . 200 Extensive felling for firewood, construction & agricul-

170.106.35.234 ture since 2000; trees sparse; loose soil 4. Hombo (Mratsini/ 87 (3) 164 (3) 146 (3) Undisturbed 812 Undisturbed forest; steep Massakini) forest 5. Hombo (Mkirini) 0 0 Agriculture 369 Roost abandoned; agricultural fields

, on 6. Kangani (Kove) 29 (2) 12 (2) 25 (2) Degraded 616 Extensive agriculture & felling for construction

26 Sep2021 at23:19:59 forest 7. Kangani (Bwejou) 3 (2) 5 (3) 3 (2) Agroforestry 645 Extensive agriculture, bare soil & few remaining large trees 8. Limbi (Trondroni) 69 (5) 46 (3) 53 (4) Undisturbed 977 Mostly undisturbed forest; some banana planting forest 9. Lingoni (Yimere) Present Present Undisturbed 741 Undisturbed forest; steep; major landslide in 2007

, subjectto theCambridgeCore termsofuse,available at forest 10. Moya (Makini) 77 (7) 73 (14) 75 (5) Degraded 626 Encroachment of agriculture; landslides observed in forest May 2012 11. Mpage (Massakini) 53 (6) 37 (5) 128 (10) Degraded 750 Clearance for agriculture started in May 2012 forest 12. Mromaji (Bandra) 34 (3) 31 (3) 32 (4) Agroforestry 452 . 200 Extensive agriculture 13. Outsa 19 (2) 22 (2) 33 (2) Agroforestry 615 . 90 Extensive agriculture; in 2011 a road was constructed (Dindri-Jumpbe) within 200 m of roost 14. Outsa (Hadzoum) 0 0 Agriculture 698 Roost abandoned; agricultural field 15. Nindri-Kowet 67 (9) 18 (3) 81 (3) Degraded 610 Extensive agriculture; roost trees felled in high winds in forest May 2012 16. Ouzini (Moihajou) 105 (13) 72 (14) 88 (13) Degraded 953 Landslide in May 2012 forest

17. Ouzini (Havoundre) 0 0 Agriculture 803 Roost abandoned; agricultural field 745 brink? the on bat A 18. Salamani 97 (5) 40 (14) 59 (5) Degraded 797 1 Extensive clearance for agriculture & felling for con- (Mroni-Papane) forest struction wood since 1995 19. Salamani 9 (2) 15 (2) Degraded 1,078 First located in 2009; some clearance for agriculture & (Maveyajou) forest felling for construction wood Total 749 584 841 746 B. M. Daniel et al.

as the surface area of undisturbed and degraded forest and agroforestry within the MCP located between the altitudes of the lowest and highest currently active roosts (–, m on Anjouan; – m on Mohéli).

Results

On Anjouan  of the  historical roost sites identified in the surveys of – (Sewall et al., ) were found to be still active. Ouzini-Havoundre appeared to be abandoned following extensive recent tree-felling (Fig. ) and no individuals were found at Hombo and Outsa-Hadzoum; the sites are now agricultural fields. No bats were found at the Dindri roost during the dry season but six individuals were counted during both wet-season surveys. The Lingoni roost, previ- site nearby, which shows evidence of recent agriculture ously reported as being one of the largest (Sewall et al., ), was visited but accurate counts were not possible as no clear vantage point could be accessed safely. Following reports by local farmers, a small, previously unknown roost site was located at Salamani Maveyajou. We therefore consider Seychelles fruit bat* Land uses & impacts within 100 m that there are currently  active roosts on Anjouan. In the wet season of ,  individuals were recorded at Anjouan roosts (not including the Lingoni roost; Table ). 582466354360223 Undisturbed forest; bats moved from the original roost 339 Extensive agriculture Encroachment of agriculture Undisturbed forest Encroachment of agriculture Possibly only a feeding site; encroachment of agriculture

Altitude (m) During the following dry season  individuals were recorded, including nine at the newly discovered roost in Salamani. In the wet season of ,  individuals were recorded across the Anjouan roosts (not including the Lingoni roost), including  at the newly discovered roost, confirm- Land cover type forest forest forest forest forest forest ing that the newly identified site is used in both seasons (Table ). The altitudinal range of occupied roosts on Anjouan was –, m(Table ), and the four abandoned roosts were all located at – m. 2013 – All five roost sites on Mohéli identified during –   Wet season 2012 were found to be occupied in (Fig. ); however, the roost at Fomboni was a few hundred metres from the historically recorded location. Following local reports, three individuals

3)4)6)12) Undisturbed Degraded 1) Degraded Undisturbed were found Undisturbed at a sixth site in Ouallah (Mkirini), but further . . . . .

2012 monitoring is required to confirm that this is a roost rather – 37 ( 25 ( 45 (5) Undisturbed 280 than a feeding site. A seventh roost was reported locally but

Dry season 2011 confirmation of its existence was not possible. All Mohéli roosts were found at altitudes of – m. A total of  individuals were recorded in the dry season; no wet-season counts were conducted. Summing the Anjouan  wet- 2012

– season counts and the Mohéli dry season counts (Table ) we calculate the minimum global Livingstone’s fruit bat Roost count (Number of trees used for roosting) Wet season 2011 population to be , individuals. The mean size of roosting populations in the dry season was . ± SE . on Mohéli and . ± SE . on Anjouan (Table ). ’ .) We estimated the EoO of Livingstone s fruit bat to be  . km (an MCP around all occupied roosts). Within Cont the EoO the surface area of potentially occupied habitat,  and therefore the AoO, was estimated to be . km in (Hanakulemba) (Hassera-Ndrengé) (Kidogo-Basse)   Mohéli 1. Fomboni (Mlédjélé)2. Ndrondroni 20 ( Roost name 3. Nioumachoi (Banda)4. Ouallah 5. Ouallah 6. Ouallah (Mkirini)Total 150 (   3 ( Table 1 ( *Minimum number recorded at roost site total ( . km on Anjouan, . km on Mohéli).

individuals at  long-term roosts and one previously unknown roost on Anjouan, and  individuals at five roosts on Mohéli during the dry season. The largest roosts on Anjouan were found at sites within a narrow altitudinal range of – m, with three roosts located at lower altitudes found to be abandoned. Given the extent and intensity of the survey, alongside a contemporaneous major biodiversity assessment of Anjouan’s forest (Green, b), we consider it unlikely that we missed any Livingstone’s fruit bat roosts supporting more than a few individuals on this island. However, resource and logistical constraints meant the survey had a number of limitations. It was not possible to conduct simultaneous counts at all the roosts on each island, and therefore we cannot rule out the possibility that bats FIG. 2 Boxplot of Livingstone’s fruit bat’s wet-season roost size in undisturbed forest, degraded forest and agroforestry areas. were moving between roosts within the survey period and Mean roost size is significantly lower in agroforestry areas than were double-counted or missed entirely, although these in undisturbed (P = .) and degraded forest (P = .), with sources of bias are likely to balance each other out. On no significant difference between the latter two (P = .). Anjouan a landslide rendered the Lingoni roost inaccessible and no counts were possible. Bat numbers at this roost were In the analysis of environmental correlates of roost previously reported to have been high (Sewall et al., ), size, only land cover type was retained in the final model with  counted in  (Granek, ) and  in      (F(,) = . ,P= . ), with evidence of a significantly (Sewall et al., ). Assuming the Lingoni roost population   lower wet-season roost size in agroforestry areas (mean = . has not changed markedly since then we are likely to have ±      ±   SE . ,n= ) than in undisturbed (mean = . SE . , underestimated the Anjouan population by c.  individuals.      ±   n= ;P= . ) and degraded forest (mean = . SE . , Counts on Mohéli were only possible during the dry sea-     n= ;P= . ; Fig. ). Although the mean roost size in un- son, when roost populations on Anjouan were on average % disturbed forest was greater than in degraded forest, there lower than in the wet season. This is in line with previous ob-   was no evidence of a significant difference (P = . ). servations of this species (Granek, ) and seasonal changes All but one occupied roost site on Anjouan, Hombo- in colony size of other Pteropus species (Mickleburgh et al., Mrantsini, had signs of vegetation clearance or significant ;Jenkinsetal.,),andisattributedtotheloweravail-  erosion (Table ). Severe weather caused landslides or ability of fruit during the dry season resulting in bats roosting brought down roost trees at three roost sites during the sur- more frequently at temporary sites nearer to food resources. If vey period alone and signs of recent felling of trees for agri- we adjust the Anjouan population estimate to take into ac-  culture and/or construction wood were evident at of the count the missing Lingoni count and increase the Mohéli   occupied sites (Plate ). count by % to allow for seasonal differences, we estimate Only four tree species were used for roosting on Mohéli, the global population of Livingstone’sfruitbattobec.,   compared with on Anjouan (Table ). Ficus exasperata individuals. was the only species used as a roost tree on both islands; This is substantially lower than numbers of any other it was the most frequently used species for roosting and Indian Ocean fruit bat for which robust population data was found at the greatest number of sites. Larger numbers exist, and possibly the lowest of any fruit bat globally. The of F. exasperata roost trees were found mainly at elevations Critically Endangered Rodrigues fruit bat Pteropus rodricen- ,   m. Of all tree species used for roosting, % were na- sis has the next lowest known population size of c. , tive or endemic to the Comoros, with the hard wood individuals (V. Tatayah, pers. comm.). The Vulnerable Gyrostipula comoriensis being the most frequently used for Aldabra fruit bat Pteropus aldabrensis is thought to be low  roosting on Anjouan (n = trees). Albizia glaberrima was in number but has not been subject to rigorous surveys and the only introduced species used for roosting on Anjouan. therefore its population size is uncertain. It is difficult to re- Two introduced species were used on Mohéli: Albizia late the current population estimate to those published pre-  lebbeck was the most frequently used (n = ) at three sites, viously (Reason & Trewhella, ; Trewhella et al., ;  and Terminalia catappa was used only at one site (Table ). Granek, ; Sewall et al., ), to quantify population trends in this species. In particular, the survey intensity Discussion and number of roost sites visited has varied between studies. Prior to this study the most recent published estimate of We report the most comprehensive survey to date of , individuals (Sewall et al., ), based on surveys con- Livingstone’s fruit bat, in which we recorded – ducted in , was obtained using a team of local

PLATE 1 Livingstone’s fruit bat Pteropus livingstonii roost sites on the islands of Anjouan and Mohéli in the Comoros archipelago (Fig. ). (a) View of the Moya roost from the survey point; (b) undisturbed forest at the Fomboni roost; (c) landslide at the Ouzini-Moihajou roost; (d) unstable soil resulting from recent land clearance for agriculture at the Mpage roost.

community members with variable levels of training and ac- by severe weather, as edge effects such as increased air tem- cess to survey equipment, and the reliability of these counts perature, reduced canopy cover and elevated tree mortality has never been assessed. The current study used a small num- increase with proximity to the forest boundaries (Laurance ber of well-trained and experienced observers, who con- et al., ). On Anjouan the remaining natural forest is ducted simultaneous counts at sites, and therefore the fragmented and found only on steep slopes at high eleva- estimate reported here is likely to be the most reliable to date. tions, and therefore we expect further damage to roosting However, given the ongoing habitat loss and degradation sites in the coming years. Furthermore, all  tree species over the previous decade, our estimate of , bats probably used for roosting are currently exploited by local people does not indicate that the population has remained un- for firewood, animal forage, medicinal uses and producing changed across this time period. Deforestation rates in the planks. Eleven of these are hard woods, of which eight are Comoros were the highest of any country during – known to be felled for planks used in building construction,  (FAO, ). Three roosts on Anjouan have been lost and the emerging use of chain-saws and a commercial mar- since the early s, following tree felling and clearance for ket on Anjouan appear to be driving widespread tree cutting agriculture. We found evidence of agricultural encroach- (H. Doulton, pers. comm.). ment, under-planting of crops, tree-cutting or major land- The only environmental variable associated with size of slides following clearance of vegetation within  m of all roost populations was land cover, with roosts located in but one of the  occupied roosts on Anjouan. During the agroforestry areas being on average – times smaller than study heavy rains resulted in flooding and landslides, in degraded and undisturbed forests. A previous study which swept away at least two trees used for roosting and (Granek, ) found that the size of bat roosts was positive- opened the canopy at three roost sites. Only the ly associated with increasing canopy cover, presence of a Hombo-Mratsini roost appears to be under no immediate watercourse, steepness of slopes and a south-east aspect, threat from vegetation clearance or landslides as it is situ- suggesting they prefer to roost at sites with higher humidity ated on a well-forested slope that is too steep for agriculture. and sheltered from the midday sun and stronger winds. The Fragmented forests are more vulnerable to damage caused newly discovered Anjouan roost was found at an altitude of

TABLE 2 Tree species used for roosting by Livingstone’s fruit bats on the islands of Anjouan and Mohéli in the Comoros archipelago (Fig. ), with no. of roost trees, no. of roost sites, geographical status, human uses, and whether the species is known to be used by the bats for feeding.

No. of roost No. of roost Geographical Used for Tree species trees sites Island status Human uses feeding Ficus exasperata 23 12 Anjouan & Native Animal forage, firewood Yes Mohéli Gyrostipula 15 6 Anjouan Native Planks, firewood comoriensis Gambeya spp. 10 6 Anjouan Native Planks (high demand) Ficus lutea 10 6 Anjouan Native Animal forage, firewood Yes Nuxia pseudodentata 10 8 Anjouan Endemic Planks, firewood Albizia lebbeck 9 3 Mohéli Introduced Animal forage, firewood Ocotea comoriensis 6 3 Anjouan Endemic Planks (high demand), Yes firewood Weinmannia 5 4 Anjouan Endemic Planks (high demand), Yes comoriensis firewood Brachylaena 5 5 Anjouan Native Firewood ramiflora Ficus pirifolia 3 3 Anjouan Native Animal forage, medicinal Yes Terminalia catappa 3 1 Mohéli Introduced Animal forage, firewood Yes Antidesma 2 1 Anjouan Native Planks hildebrandtii Albizia glaberrima 2 2 Anjouan Introduced Animal forage, firewood Khaya comorensis 2 2 Mohéli Endemic Planks, medicinal Yes Cryptocarya spp. 1 1 Anjouan Native Planks

, m but all others are located below , m despite the Efforts are now underway by the NGO Dahari to con- majority of remaining undisturbed forest lying above this serve the southern Moya forests of Anjouan to protect and elevation; this suggests there is an altitudinal limit to roost- restore water catchments and biodiversity. The project ing, although the mechanism for this is not clear. intervention zone covers seven roosts of – bats. Given the suspected severe population decline as a result of Alongside this initiative are government-led plans to estab- widespread and long-term forest loss, with these threats con- lish the first terrestrial protected area on Anjouan, around tinuing even around long-term roosting sites, Livingstone’s the Ntringui forest block, and to extend the reach of the fruit bat has now been recategorized as Critically Endangered Mohéli Marine Park to forest habitat (UNDP, ). on the IUCN Red List (Sewall et al., ). The species is in ur- However, given the vulnerability of the remaining roost sites gent need of conservation measures. In the western Indian to erosion, the apparent fidelity of the bat to these long-term Ocean and beyond there have been some notable conservation roosts, and the lack of remaining suitable habitat for roosts, successes in the recovery of fruit bats; for example, the Pemba local-scale action to protect these sites is urgent. This is in fruit bat Pteropus voeltzkowi population in Tanzania was re- line with previous recommendations by Sewall et al. () duced to only a few hundred individuals by the early s, fol- for local-scale forest reserves. The challenge, however, will lowing severe habitat fragmentation and hunting for food be to design a locally acceptable and sustainable mechanism (Mickleburgh et al., ). Since the establishment of a that will both protect and restore roost sites as well as offset community-based conservation programme enforcing habitat the opportunity costs to local impoverished farmers. protection and a reduction in hunting in the mid s A payment for ecosystem services scheme to protect bat (Entwistle, ), the population had increased to c. , roosts may form part of the solution and is being trialled by by the end of  (Juma, ). The Rodrigues fruit bat has local NGO Dahari. They are currently signing agreements recovered from just a few hundred individuals in the sto with landholders around three pilot roost sites, with the perhaps . , today thanks to long-term forest restoration aim of protecting the roost trees from felling and gradually work and a moratorium on hunting (Powell & Wehnelt, ). reforesting the surrounding area with native trees to provide Recovery of Livingstone’s fruit bat in a similar way will be chal- food and shade. In return the landholders will be supported lenging considering that hunting bans have played an import- by Dahari to improve agricultural revenues from their other ant role in these successful recoveries of other Pteropus species plots through provision of improved varieties of banana and (Mickleburgh et al., ; Robinson et al., ) and that hunt- taro to plant under the native forest cover. In the longer ing is not a threat in the Comoros. term the aim is for the initiative to be financed through

ecotourism and contributions from partner zoos holding CARROLL, J.B. & FEISTNER, A.T.C. () Conservation of western the species for in situ conservation. Indian Ocean fruit bats. In Biogéographie de Madagascar (ed. W. R. – However, identifying specific conservation recommen- Louren), pp. . Institut Français de Recherche Scientifique pour le Developpement en Cooper, Paris, France. dations beyond protection of important roost sites is ham- CHEKE, A.S. & DAHL, J.F. () The status of bats on western Indian pered by a lack of knowledge of the distribution of other key Ocean islands, with special reference to Pteropus. Mammalia, , ecological resources, such as the main foraging sites. Future –. study of the population dynamics of Livingstone’s fruit bats CONDER,P.() Percy Island flying fox. In The Mammals of  − is necessary to understand the degree of movement between Australia, rd edition (eds S. Van Dyck & R. Strahan), pp. . Reed New Holland, Sydney, Australia. roosts and islands. Most importantly the feeding ecology of COX, P.A., ELMQVIST, T., PIERSON, E.D. & RAINEY, W.E. () Flying the species needs to be investigated further than initial stud- foxes as strong interactors in south Pacific island ecosystems: a ies (Trewhella et al., ; Sewall, ) to understand the conservation hypothesis. Conservation Biology, , –. availability of food resources and to identify important feed- ENTWISTLE,A.() Community-based protection successful for the  – ing sites requiring protection. This work would also help to Pemba flying fox. Oryx, , . EPSTEIN, J.H., PRAKASH, V., SMITH, C.S., DASZAK, P., MCLAUGHLIN, elucidate the bat’s role in the ecosystem as a pollinator and A.B., MEEHAN, G. et al. () Henipavirus infection in fruit bats seed disperser, and therefore its potentially high-value role (Pteropus giganteus), India. Emerging Infectious Diseases, , in forest dynamics. –. FAO () Évaluation des Ressources Forestieres Modiales . Rapport National: Comores. Report #FRA/. Food and Acknowledgements Agriculture Organization of the United Nations, Rome, Italy. FUJITA, M.S. & TUTTLE, M.D. () Flying foxes (Chiroptera: We are grateful to the following funders who supported the Pteropodidae): threatened animals of key ecological and economic field surveys: the UK Government’sDarwinInitiative, importance. Conservation Biology, , –.  l’Agence Française de Développement, The Global GRANEK,E.( ) Conservation of Pteropus livingstonii based on roost site habitat characteristics on Anjouan and Moheli, Comoros Environment Facility, and the United Nations Development Islands. Biological Conservation, , –. Programme through the Programme of Work on Protected GREEN,K.(a) Land Cover Mapping of the Comoros Islands: Areas. We also thank our partner institutions that supported Methods and Results. Engagement Communautaire pour le this work, notably the Comorian government, the University Développement Durable, Bristol Conservation and Science of the Comoros, and the Centre National de Documentation Foundation, Durrell Wildlife Conservation Trust. GREEN,K.(b) Terrestrial Biodiversity Mapping of the Comoro et de Recherche Scientifique. Islands: Methods and Results. Engagement Communautaire pour le Développement Durable, Bristol Conservation and Science Foundation, Durrell Wildlife Conservation Trust. Author contributions IUCN () The IUCN Red List of Threatened Species . Http:// www.iucnredlist.org [accessed  November ]. BD led the data collection, data analysis and drafting of the JENKINS, R.K.B., RACEY, P.A., ANDRIAFIDISON, D., RAZAFINDRAKOTO, article. KG conducted data collection and analysis and made N., RAZAFIMAHATRATRA, E., RABEARIVELO,A.etal.() Not comments on the article. HD helped draft and edit the art- rare, but threatened: the endemic Madagascar flying fox Pteropus icle. DS and IS advised on survey design and collected field rufus in a fragmented landscape. Oryx, , –.  data. MH and JD supported data analysis and provided JUMA,J.( ) Conservation of Indigenous Forest and Endemic Species on Pemba Island. Unpublished report by Fauna & Flora comments on the article. RY oversaw the design and deliv- International, Nairobi, Kenya. ery of the study and drafted and edited the article. LAURANCE, W.F., LOVEJOY, T.E., VASCONCELOS, H.L., BRUNA, E.M., DIDHAM, R.K., STOUFFER, P.C. et al. () Ecosystem decay of Amazonian forest fragments: a -year investigation. Conservation References Biology, , –. LEMMON, P.E. () A spherical densiometer for estimating forest ALLEN-WARDELL, G., BERNHARDT, P., BITNER, R., BURQUEZ, A., overstory density. Forest Science, , –. BUCHMANN, S., CANE, J. et al. () The potential consequences of MICKLEBURGH, S.P., HUTSON, A.M. & RACEY, P.A. () Old World pollinator declines on the conservation of biodiversity and stability Fruit Bats: An Action Plan for their Conservation. IUCN, Gland, of food crop yields. Conservation Biology, , –. Switzerland. BOLLEN,A.&VAN ELSACKER,L.() Feeding ecology of Pteropus MILDENSTEIN, T.L., STIER,S.C.,NUEVO-DIEGO,C.E.&MILLS,L.S. rufus (Pteropodidae) in the littoral forest of Sainte Luce, SE () Habitat selection of endangered and endemic large flying-foxes Madagascar. Acta Chiropterologica, , –. in Subic Bay, Philippines. Biological Conservation, , –. BOLLEN, A., VAN ELSACKER,L.&GANZHORN, J.U. () Relations MOHD-AZLAN, J., ZUBAID,A.&KUNZ, T.H. () Distribution, between fruits and disperser assemblages in a Malagasy littoral relative abundance, and conservation status of the large flying fox, forest: a community-level approach. Journal of Tropical Ecology, , Pteropus vampyrus, in Peninsular Malaysia: a preliminary –. assessment. Acta Chiropterologica, , –. BROOKE, A.P. () Population status and behaviours of the Samoan NYHAGEN, D.F., TURNBULL, S.D., OLESEN, J.M. & JONES, C.G. () flying fox (Pteropus samoensis) on Tutuila Island, American Samoa. An investigation into the role of the Mauritian flying fox, Pteropus Journal of Zoology, , –. niger, in forest regeneration. Biological Conservation, , –.

PIERSON, E.D. & RAINEY, W.E. () The biology of flying foxes of the livingstonii) in the Federal Islamic Republic (RFI) on the Comores. genus Pteropus: a review. In Pacific Island Flying Foxes: Proceedings Phelsuma, , –. of an International Conservation Conference (eds D.E. Wilson & TREWHELLA, W.J., RODRÍGUEZ-CLARK, K.M., DAVIES, J.G., REASON, G.L. Graham), pp. –. United States Fish and Wildlife Service, P.F. & W RAY,S.() Sympatric fruit bat species (Chiroptera: Washington, DC, USA. Pteropodidae) in the Comoro Islands (Western Indian Ocean): POWELL, V.J. & WEHNELT, S.C. () A new estimate of the diurnality, feeding interactions and their conservation implications. population size of the Critically Endangered Rodrigues fruit bat Acta Chiropterologica, , –. Pteropus rodricensis. Oryx, , –. TREWHELLA, W.J., RODRÍGUEZ-CLARK, K.M., CORP, N., ENTWISTLE, RDEVELOPMENT CORE TEAM () R: A Language and Environment A., GARRETT, S.R.T., GRANEK, E. et al. () Environmental for Statistical Computing. The R Foundation for Statistical education as a component of multidisciplinary conservation Computing, Vienna, Austria. programs: lessons from conservation initiatives for Critically REASON,P.F.&TREWHELLA, W.J. () The status of Pteropus Endangered fruit bats in the Western Indian Ocean. Conservation livingstonii in the Comores. Oryx, , –. Biology, , –. ROBINSON, J.E., BELL, D.J., SALEH, F.M., SULEIMAN, A.A. & BARR,I. UNDP () Human Development Report . The Rise of the South: () Recovery of the Vulnerable Pemba flying fox Pteropus Human Progress in a Diverse World. United Nations Development voeltzkowi: population and conservation status. Oryx, , –. Programme, New York, USA. SEWALL, B.J. () Fruit bat ecology and conservation in the Comoros Islands. MSc thesis. University of Minnesota, Minneapolis, USA. SEWALL, B.J., GRANEK, E.F., MOUTUI, M.F.E., TREWHELLA, W.J., REASON, P.F., RODRÍGUEZ-CLARK, K.M. et al. () Conservation Action Plan for Livingstone’s Flying Fox: A Strategy for an Biographical sketches Endangered Species, a Diverse Forest, and the Comorian People. Union of the Comoros. BRONWEN DANIEL and KATHLEEN GREEN were involved in a major SEWALL, B.J., FREESTONE, A.L., MOUTUI, M.F., TOILIBOU, N., SAÏD, I., terrestrial biodiversity assessment of the Comoros, including surveying TOUMANI, S.M. et al. () Reorienting systematic conservation Livingstone’s fruit bat. HUGH DOULTON is Technical Director of the assessment for effective conservation planning. Conservation Comorian NGO Dahari, for whom ISHAKA SAID and DANIEL SALIM Biology, , –. work as ecological technicians. They have worked together on eco- SEWALL, B.J., YOUNG, R., TREWHELLA, W.J., RODRÍGUEZ-CLARK, logical research in the Comoros for the past  years and are currently K.M. & GRANEK, E.F. () Pteropus livingstonii. The IUCN Red implementing a major Critical Ecosystem Partnership Fund pro- List of Threatened Species : e.TA. Http://dx.doi. gramme to identify conservation priorities on Anjouan, as well as pi- org/./IUCN.UK.-.RLTS.TA.en [accessed loting a payment for ecosystem services approach to protecting key  October ]. roost sites of Livingstone’s fruit bat. MIKE HUDSON and JEFF TREWHELLA, W.J., REASON, P.F., CLARK, K.M. & GARRETT, S.R.T. DAWSON are conservation scientists. RICHARD YOUNG was the lead () The current status of Livingstone’s flying fox (Pteropus scientist on the Comoros biodiversity assessment.