Ecology and conservation of the endemic Bawean warty pig Sus verrucosus blouchi and Bawean deer Axis kuhlii E VA J OHANNA R ODE-MARGONO,HANNAH K HWAJA M ARK R ADEMAKER and G ONO S EMIADI Abstract The island of Bawean, Indonesia, is home to the island endemics is thus of paramount importance to the endemic Bawean warty pig Sus verrucosus blouchi and preservation of biodiversity in South-east Asia. Bawean deer Axis kuhlii. Despite their threatened status, The island of Bawean is a small volcanic remnant in the no long-term monitoring programme is in place for either Java Sea, separated from mainland Java for c. , years species. Using random encounter and occupancy modelling (Meijaard, ). Its most prominent endemic mammals based on , camera-trap days in and we aimed are the Critically Endangered Bawean deer Axis kuhlii and to provide population estimates and ecological data, includ- the Endangered Bawean warty pig Sus verrucosus blouchi. ing habitat preferences, for these species. For the Bawean Populations of these species are believed to be small but stable warty pig we estimate an overall population size of – (Bawean deer: –, Semiadi et al., ;Baweanwarty mature individuals and demonstrate a negative correla- pig: –, Rademaker et al., ). The Bawean warty pig tion between probability of occupancy and distance from has no protected status but the Bawean deer is protected by villages. This preference for human-modified habitat has Indonesian law (Lampiran Peraturan Pemerintah Nomor implications for human–wildlife conflict and hunting pres- Tahun & Undang-Undang No. Tahun ), is listed sure for this species. The population of the Bawean deer on CITES Appendix I (CITES, ) and is one of species could not be estimated because of the low number of en- that the Indonesian government considers to be in need of counters, but we suggest that this indicates the population special conservation attention (decree SK./IV-KKH/; is considerably smaller than previously reported. As island Ministry of Environment and Forestry, ). The primary endemics, the Bawean warty pig and Bawean deer are par- threat to the deer is thought to be the small size of the ticularly vulnerable to threats, and appropriate measures population (Semiadi et al., ), with a potentially decreased for safeguarding the species need to be taken. genetic diversity (Rahman et al., a), and hunting or stress caused by feral hunting dogs (Nursyamsi, pers. comm.; Keywords Activity pattern, Axis kuhlii, Bawean, habitat Drygala & Semiadi, pers. comm.; Rahman et al., a). preferences, occupancy modelling, population size, random Bawean warty pigs are actively hunted to protect crops from encounter modelling, Sus verrucosus blouchi damage (Nijman, ; Rademaker, ). Apart from a -month camera-trap survey for Bawean deer in (Rahman et al., a) and a survey for Introduction Bawean warty pigs in (Rademaker et al., ), there has been no long-term study of the mammal biodiversity he Sunda Islands of the Indonesian archipelago are of Bawean. As a result, no data are available about the ecol- Tamongst the world’s most biologically diverse regions ogy and behaviour of wild populations of the island’s threa- (Myers et al., ). Island biodiversity is particularly threa- tened endemic species (but see Rode-Margono et al., a), tened by habitat loss and conversion, overexploitation and in- which hinders effective conservation planning. vasive species, and more recently by effects of anthropogenic The aims of our -month camera-trap study on Bawean climate change (Fordham & Brook, ). Conservation of Island were to () provide relative encounter rates for all recorded species, () produce absolute population estimates for the Bawean warty pig, () investigate which habitat EVA JOHANNA RODE-MARGONO (Corresponding author) and HANNAH KHWAJA The North of England Zoological Society/Chester Zoo, Upton-by-Chester, Chester, and environmental factors influence relative encounter CH2 1LH, UK. E-mail [email protected] rates, () examine activity patterns, social structure and MARK RADEMAKER* Forest and Nature Conservation Program, Wageningen reproductive patterns, and () compare our results for the University, Wageningen, The Netherlands Bawean deer to those of Rahman et al. (a). GONO SEMIADI ( https://orcid.org/0000-0002-9351-9746) Research Centre for Biology, Indonesian Institute of Sciences, Cibinong, Bogor, Indonesia *Previously at: Department of Animal Management, Van Hall Larenstein Study area University of Applied Sciences, Leeuwarden, The Netherlands Received October . Revision requested November . The km island of Bawean, in the Java Sea, is dominated Accepted July . First published online March . by semi-evergreen forest, with a dry season during June– Oryx, 2020, 54(6), 892–900 © 2019 Fauna & Flora International doi:10.1017/S0030605318000996 Downloaded from https://www.cambridge.org/core. IP address: 170.106.35.234, on 30 Sep 2021 at 10:44:18, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0030605318000996 Bawean warty pig and Bawean deer 893 November and a wet season during December–May Data analysis (Hamada et al., ). Central to the island is an extinct vol- cano, which reaches m altitude. All remaining forests on We report relative abundance for all species encountered, Bawean are encompassed within the . km Bawean with the relative abundance index (RAI) defined as all inde- Island Nature Reserve and Wildlife Sanctuary and are pendent detections of a given species summed for all camera under varying levels of protection (Semiadi & Meijaard, traps over all days, multiplied by , and divided by the ’ ; Fig. a). total number of camera-trap nights (O Brien et al., ). We employed a -hour interval to define independent events (Rovero et al., ; Rademaker et al., ). As data were not Methods normally distributed and could not be transformed, we used non-parametric Kruskal–Wallis tests and Mann–Whitney Data collection U post-hoc tests to check for differences in relative abun- dance index between seasons and locations. For the influ- Camera trapping took place during November – ence of habitat variables on encounter rates, we report December . We divided the forested area into , single-season occupancy modelling outcomes using the grid cells of × m, and selected cells randomly Rv... (R Core Team, ) package unmarked (Fiske & as locations for camera traps (Long Range IR/E, Chandler, ). We removed records that were incomplete Cuddeback, De Pere, USA), which were moved periodically because of missing covariate values. Models assessed the between locations. During August–October, because of in- effects of all previously described site-level covariates on creasing concern about the low numbers of deer recorded probability of occupancy, and the effects of the observation- up to then, we set five camera traps preferentially for deer, level covariates moonlight, temperature and rainfall on based on local advice. We assumed these locations were still probability of detection. Sample sizes for Bawean deer random for warty pigs. To prevent spatial autocorrelation, were too small to run these analyses. camera traps were spaced at least m apart. In small forest To estimate population density using random encounter fragments, because of difficult terrain, we selected the first modelling we defined the camera-trap rate as the total num- random point and then placed subsequent camera traps ber of independent captures divided by the total number of every m in a randomly generated direction. We set camera-trap days. We calculated absolute population num- camera traps to record -s videos, with a -minute interval bers for four trapping periods of – days. The random until the next trigger. encounter model assumes that the population is closed; At each camera trap location we recorded the latitude i.e. there is a fixed number of individuals in the area and longitude (with a global positioning system), altitude throughout the estimation period (Rowcliffe et al., ). and habitat variables. These included the major habitat As there is no immigration or emigration on Bawean, poten- type (rice cultivation: n = ;gardencultivation:n=; tial bias comes only from births and deaths, the effects of shrubland and degraded forest: n = ;teakplantation: which we assume to be negligible for the chosen trapping n=;tallforest:n=; community forest: n = ;fordefi- periods. We could not standardize period length because nitions see Rademaker et al., ), mean tree diameter of logistical challenges. We also report the combination of at breast height, mean tree height in a × mplot all four periods because single periods violate the assump- around the camera-trap location, and tree density using tion of at least camera-trap locations (Rovero et al., the T-square method (Rode et al., )withtwosample ). The random encounter model used here is described points (for details see Rademaker et al., ). We calcu- in detail in Rademaker et al. () and follows Rowcliffe lated the mean litter depth in four × msubplotsin et al. (, , ). Day range was defined as the the corners of the plot. Distances to the nearest village mean speed of movement of the animals in front of the and protected area border (approximately coinciding camera in m/s, extrapolated to km/day and multiplied by with the forest border) were calculated in ArcGIS v. the proportion of time spent active. The result of the ran- (ESRI, Redlands, USA). Minimum and maximum daily dom encounter modelling was multiplied by the mean temperatures and precipitation data were obtained from group size of the relevant species, as a group is the entity Sangkapura meteorological station. Lunar illumination of detection. However, the result must be considered to be was retrieved from MOONDV v. (Thomas, ). For a minimum estimate, as a video may miss some individuals the random encounter model, we also recorded the angle in a social group. We extrapolated the estimated densities of detection, radial distance at detection and distance trav- per km to the total protected area on Bawean to provide elled for the first three video records of Bawean warty pigs absolute population estimates.