Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Managing sun bears in a changing tropical landscape

Managing sun bears in a changing tropical landscape Introduction Deforestation and unsustainable hunting in the tropics have degraded many landscapes to the point that they now no longer support viable populations of threatened species (Sodhi et al ., ; Bennett, ). Traditional approaches to conserving these landscapes and their wildlife populations have tended to focus on protected area management (IUCN, ; DeFries et al ., ). However, it is questionable whether the existing protected areas provide adequate habitat cover, quality or indeed protection for threatened species (Catullo et al ., ), as designation is often based on socio‐economic and political priorities rather than assessments of wildlife habitat requirements (Margules & Pressey, ). This has, for example, led to significant sized patches of lowland tropical forests being excised from protected areas for commercial logging (Daily et al ., ). The threat posed by deforestation tends to disproportionately affect large‐bodied mammals because of their large range requirements (Kinnaird et al ., ). Thus, to halt the ongoing loss of large‐bodied mammal populations, identification of high conservation value areas is required (Hoffmann et al ., ), but the detailed information on the distribution of species required for this purpose is often lacking (Rondinini et al ., ). The requisite distributional data for developing conservation priorities, whilst available for many species in the form of regional range data (Rodrigues et al ., ), are less useful for conservation agencies that require finer‐scale information, such as spatial distribution and use of habitat under varying levels of protection. The sun bear is a large‐bodied mammal species that historically ranged from India to Indonesia and China. Today, its range has contracted considerably. Sun bears are now patchily spread throughout much of this former range and have disappeared entirely from three of its 12 former range countries (Servheen, ). Agricultural expansion into forest habitat, especially on the islands of Indonesia and Borneo, and poaching for trade, especially in mainland Southeast Asia, are the primary drivers of this dramatic range and population reduction (Meijaard, ). The previously data deficient sun bear was recently reclassified by the International Union for Conservation of Nature (IUCN) as Vulnerable (IUCN, ). Yet, the conservation management of this species is hampered by a lack of basic, yet fundamentally important, ecological knowledge of habitat preferences and geographic distributions. Sun bears have a varied diet (Wong et al ., ; Fredriksson et al ., ), which enables the species to live in a range of habitats (Linkie et al ., ). Previous research in the Kerinci Seblat (KS) region, which estimated sun bear occupancy from four different habitat types, showed that occupancy was highest in a primary–secondary hill forest and lowest in a primary submontane forest (Linkie et al ., ). However, given the rapid rates of forest conversion across the KS region (Linkie et al ., ; Uryu et al ., ), sun bear distribution and occupancy in these habitats are likely to change. Sun bear research has predominantly investigated habitat use within lowland forest reserves and protected areas (Servheen, ; Wong et al ., , ; Fredriksson et al ., ). The value of higher elevation and unprotected forests for sun bears has therefore received little attention (Linkie et al ., ). This is a critical information gap because the disproportionately high and unyielding rates of deforestation in lower elevation forest are undoubtedly shaping the spatial distribution and habitat use patterns of sun bears, as well as other large‐bodied mammal species, in the tropics (Kinnaird et al ., ; Wibisono et al ., ). An important limitation for sun bear habitat management is poor land‐use planning, with large forest areas of high biological value still being cleared for plantations or commercial logging, which uses unsustainable practices (Fitzherbert et al ., ). Comparative studies on the influence of different land‐use types on the diversity and distribution of biodiversity, or components of it, are essential for policy makers and conservation managers (Gillison et al ., ). Here, we provide the first assessment of sun bear habitat use across a large tropical forest landscape in Sumatra that covers all of the main land‐use types, ranging from strictly protected areas and conservation areas, production forests and forests that are classified as legally convertible to non‐forest by Indonesian law, defined here as convertible forests. We investigated (1) sun bear habitat use at a broad‐scale looking at protection status, (2) habitat use at a finer‐scale within these study areas and (3) use the main findings to make broad recommendations for the conservation of sun bears, which should also apply to other large‐bodied mammals, living in human‐altered tropical landscapes. Methods Study area The west‐central Sumatra region consists of several land‐use types that include a protected area, Kerinci Seblat National Park (KSNP; 13,300 km 2 ); conservation area, Batang Hari Protection Forest (BHPF; 1,700 km 2 ); production forests (4,888 km 2 ) and convertible forests (1,110 km 2 ; Fig. a). This vast protected area network and diverse range of forest habitats should represent a stronghold for sun bears and other threatened species of large‐bodied mammals (Holden et al ., ; Linkie et al ., ). However, with a mean deforestation rate across the region of 1.10%yr −1 (2004–2008/9; Uryu et al ., ), forest conversion into agricultural farmlands represents a significant threat. The west‐central Sumatra project landscape covering Kerinci Seblat National Park (KSNP), Batang Hari Protection Forest (BHPF) and neighbouring forests, showing (a) the different land‐use types, and (b) the location of camera trap study areas. This study focused on six camera trapping areas: two inside KSNP; two that straddled the KSNP border with placement also in convertible and/or production forests; one inside BHPF; and one inside an active logging concession bordering BHPF, referred to hereafter as PT‐AMT (Fig. b; Table ). This covered three of the eight mainland Sumatran provinces (Bengkulu, Jambi and West Sumatra) and all of the main tropical forest types in which sun bears are known to live, encompassing selectively logged and primary lowland/hill forest to pristine montane forests (Servheen, ). These six areas have different levels of forest degradation and protection status (Table ). Camera trap study areas used for estimating sun bear habitat use in the west‐central Sumatra project landscape Study area Number of camera traps Camera trap area (km 2 ) Altitudinal range (m) Main habitat type and protection status Camera trapping dates PT‐AMT 18 124 365–1022 (mean: 653) Active logging concession bordering primary lowland/hill forest Feb–May 2008 Batang Hari Protection Forest (BHPF) 21 165 340–1461 (mean: 742) Primary hill/submontane forest bordering logging concession June–Sept 2008 Kerinci Seblat National Park (KSNP) Sipurak 21 88 694–1254 (mean: 901) Primary hill/submontane forest inside KSNP adjacent to ex‐logging concession Dec 2009–Mar 2010 Bungo 21 90 363–1630 (mean: 753) Primary–secondary hill forest predominantly inside an ex‐logging concession, with a few sites inside KSNP Apr–July 2010 RKE 23 104 947–1941 (mean: 1194) Primary submontane forest inside KSNP Aug–Nov 2010 Ipuh 21 118 145–1032 (mean: 511) Primary–secondary lowland forest predominantly inside an ex‐logging concession, with a few sites inside KSNP Nov 2010–Feb 2011 Data collection From February 2008 to April 2011, a detection/non‐detection sampling technique using camera trap data was used to estimate sun bear habitat use in the KS region and its neighbouring forests. Between 18 and 23 camera traps were deployed in each area, resulting in a total of 125 camera trap placements across the west‐central Sumatra project landscape (Table ). A combination of film and digital infrared camera models was used including PhotoScout (PTC Technologies Inc, Boston, MA, USA), Moultrie (Moultrie ™ , Alabaster, AL, USA) and Bushnell (Bushnell Corporation, Overland Park, KS, USA). Cameras were placed along ridge and animal trails at a height of approximately 0.5 m above the ground. To obtain a sufficient number of sampling sites at each study area, camera traps were spaced with a minimum distance of 2 km based on sun bear movement and home range data (Wong et al ., ). This camera spacing was to ensure that no sun bear within the sampling areas had a zero probability of being detected, thereby adhering to a critical assumption of the occupancy‐based method used in this study (MacKenzie et al ., ). Cameras were active 24 h per day, set with a 1‐min delay between exposures, programmed to record the time and date of each event and were active for approximately 90 days to avoid possible violation of the population closure assumption (Karanth & Nichols, ). Cameras were visited every 2 weeks to replace their film, memory cards and batteries, as well as to check their condition. Quantification of landscape covariates A dataset of 10 potential landscape variables that were thought to influence sun bear habitat use was constructed within ArcGIS v9.3 software (ESRI, Redlands, CA). The raw data were obtained from several sources: elevation and slope (Rabus et al ., ); the position of public roads, logging roads, rivers, villages and forest edge (Indonesian National Coordination Agency for Surveys and Mapping); land‐use type (protected areas, forest reserves, production forests and convertible forests; WWF‐Indonesia forest data); forest type (primary and degraded forest); and study area (PT‐AMT, BHPF, Sipurak, Bungo, RKE and Ipuh). Elevation and slope covariates were extracted at a 30 × 30 m resolution, and a single value per site was obtained by averaging all the pixel values within each sampling site (camera trap placement). The spatial covariates of distance to public roads, logging roads, rivers, villages and forest edge (from within the forest) were calculated within ArcGIS v9.3 using the Spatial Analyst extension from each sampling site coordinate. Statistical analysis and predictive model construction The 10 landscape variables were categorized as being either broad‐scale or fine‐scale. Broad‐scale variables included categorical covariates of land‐use type, forest type and study area. Fine‐scale variables included elevation, slope and distance to public roads, logging roads, rivers, villages and forest edge. The continuous landscape variable data extracted from each sampling site were imported into spss v.18.0 software ( spss Inc., Chicago, IL, USA) and logarithmically transformed to reduce the likelihood of extreme values having a disproportionate influence on the overall dataset (Royston & Sauerbrei, ). To test for non‐independence between covariates, a Spearman's rank correlation coefficient ( r s ) was calculated between pairwise variables. Correlated covariates were combined using principal component analysis (Abdi & Williams, ), using the data reduction technique in spss v.18.0 software, to produce a single covariate that was then used in the final data analysis. The covariates of distance to public roads and villages ( r s = 0.70), distance to forest edge and villages ( r s = 0.53) and distance to logging roads and elevation ( r s = 0.55) were correlated ( P < 0.05) and entered within unique combined covariates (i.e. Dist. public roads/villages; Dist. Forest edge/village; Dist. Logging/elevation). As different covariates can have different ecological effects, each correlated covariate was also modelled separately, but not in the same model. Due to the possibility of non‐independence between camera trap sampling sites, the Spatial Statistics Toolset extension in ArcGIS v9.3 was used to test for the presence of spatial autocorrelation across the camera trap sampling unit coordinates by calculating Moran's I statistic. This study applied a sampling design in which the number of camera trap sample sites ( n ) was visited multiple times on K sampling occasions. In turn, this allowed for inferences over sun bear habitat use, whilst accounting for imperfect detection, to be made. Detection/non‐detection data were collected from camera trap surveys to obtain encounter histories for sun bears. The detection (1) or non‐detection (0) sequence of sun bears over a 3‐month trapping duration per study area was then recorded as a detection history. Within this detection history, the 3‐month sampling period was divided into six consecutive 2‐week sampling occasions. Detection histories for each of the six study areas were combined to produce one dataset containing a total of 125 sampling sites. The resulting detection history framework ( n = 125, K = 6) was entered into PRESENCE v2.3 software (Hines, ). Within the PRESENCE software, a two‐stage logistic regression analysis was performed to determine which broad‐scale factors (i.e. land‐use type, study area and forest type) influenced sun bear habitat use (ψ) and then how several finer‐scale factors explained habitat use within these study areas. Models included detection probability as a constant, p (.), or as a function of site‐specific covariates, p (covariate). Candidate models were ranked by their second‐order information criterion (AIC c ) values, corrected for small sample sizes, and the Akaike weights ( w i ; Burnham & Anderson, ). The Akaike weight represents the ratio of ΔAIC c values for the whole set of candidate models, providing a strength of evidence for each model. The top‐ranking candidate model was used to determine the sun bear habitat use for the west‐central Sumatra project landscape. The corresponding beta (β) coefficients from the covariates in the top model, derived from PRESENCE software, were used to construct a predictive habitat suitability model ( P ) in ArcGIS v9.3 using the following equation: P = 1 1 + e − ( β 0 + ∑ β i X i ) where β 0 is the constant coefficient (intercept) and β 1 , β 2 , … β i represent the regression coefficients of the associated independent variables X 1 , X 2 … X i . The predictive power of the habitat suitability model was validated using sun bear habitat use data that were derived from a separate camera trapping survey collected from February 2011 to May 2011. Subsequently, an additional 40 camera trap sampling sites were selected within KSNP. This was achieved by randomly selecting four forest areas along the KSNP border that varied in habitat type. A total of 10 camera traps were deployed within each area along animal and ridge trails with a spacing greater than 2 km apart to minimize spatial autocorrelation. Camera traps were left for a period of 1 month at each sampling site. Detection/non‐detection data for each site were collected and entered into PRESENCE software to generate site‐specific habitat use estimates. A Spearman's rank correlation coefficient was used to test for a correlation between model estimated values of habitat use and those generated by the independent dataset for each corresponding 90 m 2 suitability model pixel (Linkie et al ., ). Results From February 2008 to April 2011, sun bears were camera trapped for a combined sampling effort of 10,935 trap nights comprising 1620 trap nights in PT‐AMT, 1737 in BHPF, 2042 in Sipurak, 1762 in Bungo, 1884 in RKE and 1890 in Ipuh. Sun bears were detected in 47 of the 125 units, corresponding to a naïve occupancy estimate of 0.376. Individual sun bear records ranged in altitude from 365 m to 1791 m above sea level (a.s.l.). Broad‐scale patterns At a broad‐scale, sun bear habitat use was found to be strongly influenced by land‐use type (Model 1.1; Table ). Modelling land‐use types as a function of sun bear habitat use produced the following beta coefficient values ( β i ± SE): β KSNP = 0.800 ± 0.193, β BHPF = −0.311 ± 0.242, β production forest = −1.12 ± 1.03 and β conertible forest = −1.70 ± 1.33. The positive beta regression coefficient for KSNP (protected area) indicates this land‐use type increases the probability of habitat use by sun bears, whilst the negative beta regression coefficients for production and convertible forests imply the opposite. Consequently, sun bear habitat use ( ψ ^ ± SE) was highest in the protected forests of KSNP (0.688 ± 0.092), followed by BHPF (0.621 ± 0.110), production forests (0.418 ± 0.121), and lowest in convertible forest (0.286 ± 0.122). However, modelling land‐use types as a function of sun bear detection probability, p (land use), found no significant differences between production forests (0.391 ± 0.059; ±SE), BHPF (0.371 ± 0.117), convertible forests (0.322 ± 0.113) and KSNP (0.305 ± 0.050; Model 1.3, Table ). Summary of model selection procedure for sun bear habitat use within 125 camera trap sites across the west‐central Sumatra project landscape Model no. Model K ΔAIC c w i X 2 P ‐value ĉ Broad‐scale analysis 1.1 ψ(Land use) p (.) 3 0.00 0.472 66.97 0.726 0.81 1.2 ψ(Land use + Study area) p (.) 4 1.87 0.185 55.76 0.523 0.91 1.3 ψ(Land use + Study area) p (Land use) 5 2.14 0.162 56.98 0.519 0.85 1.4 ψ(Study area) p (.) 3 2.15 0.161 53.98 0.672 0.82 1.5 ψ(.) p (.) 2 6.40 0.019 59.46 0.648 0.83 Fine‐scale analysis 2.1 ψ(Dist. public roads/villages + elevation) p (.) 4 0.00 0.321 5.39 0.631 0.82 2.2 ψ(Dist. public roads/villages) p (.) 3 0.42 0.261 5.40 0.592 0.85 2.3 ψ(.) p (.) 2 1.80 0.131 5.38 0.620 0.84 2.4 ψ(Dist. public roads/villages + elevation) p (Dist. public roads/villages) 5 2.14 0.110 5.48 0.570 0.86 2.5 ψ(Dist. public roads/villages) p (Dist. public roads/villages) 4 2.52 0.091 5.39 0.621 0.82 2.6 ψ(Elevation) p (.) 3 2.64 0.086 5.56 0.547 0.86 ψ is the probability a site is occupied by a sun bear and p is the sun bear detection probability where ψ(.) p (.) assumes sun bear presence and detection probability are constant across sites; K is the number of parameters in the model; ΔAIC c is the difference in AIC c values between each model; w i is the AIC c model weight; X 2 is the test statistic for model fit; P ‐value is the probability of observing a test statistic ≥ X 2 based on 999 parametric bootstraps; and ĉ is the estimated overdispersion parameter. Fine‐scale patterns At a fine‐scale, sun bear habitat use was best explained by distance to public roads/villages and elevation, and this model provided an adequate description of the data (Model 2.1; Table ). The summed model weights for each factor with respect to habitat use were as follows: distance to roads/villages (78.3%) and elevation (51.7%). Sun bears showed a preference for forest that was located further away from public roads and villages and at lower elevations (Table ). The corresponding estimated odd ratios for a one unit increase in each of the covariates, OR ^ j = e β ^ j , were OR ^ public roads/villages = 11.2 ± 0.05 and OR ^ elevation = 0.16 ± 0.12. This final model was not affected by spatial autocorrelation (Moran's I = 0.03, P > 0.05) and thus used to construct a predictive habitat suitability model. Logit models estimating sun bear habitat use, with beta (β) coefficient and standard error (SE) estimates for the top‐ranked models (from Table ) Model no. Model β 0 (constant) ± SE β 1 (Roads/villages) ± SE β 2 (elevation) ± SE 2.1 ψ(Dist. public roads/villages + elevation) p (.) 14.732 ± 2.48 2.414 ± 0.55 −1.844 ± 0.78 2.2 ψ(Dist. public roads/villages) p (.) 7.530 ± 2.52 1.96 ± 0.63 – 2.3 ψ(.) p (.) −0.332 ± 0.20 – – 2.4 ψ(Dist. public roads/villages + elevation) p (Dist. public roads/villages) 14.834 ± 2.47 2.443 ± 0.55 −1.844 ± 0.78 2.5 ψ(Dist. public roads/villages) p (Dist. public roads/villages) 7.631 ± 2.52 1.988 ± 0.63 – 2.6 ψ(Elevation) p (.) – – −1.226 ± 0.97 The predictive habitat suitability model showed that sun bears predominantly used habitats within the KSNP border (Fig. ). The proximity of public roads and villages to the forest areas outside KSNP reduced sun bear habitat suitability. Habitat quality in production and convertible forest areas was significantly lower than protected areas. The predicted values of sun bear habitat use were positively and significantly correlated with sun bear habitat use estimates that were derived from independent camera trap data ( n = 40, r s = 0.557, P < 0.001). Sun bear habitat suitability in west‐central Sumatra covering Kerinci Seblat National Park (KSNP), Batang Hari Protection Forest (BHPF) and neighbouring forests. Discussion This study is the first to assess sun bear habitat use across a large tropical landscape. The predictive habitat suitability model revealed how increased human disturbances, indicated by the presence of public roads, villages and elevation, negatively influenced sun bear habitat use across the west‐central Sumatra project landscape. Habitat use was predicted to be higher in protected areas than in forest reserves, production and convertible forests. Areas within the BHPF border ranked high in sun bear habitat use; however, the presence of public roads and villages surrounding the BHPF reduced its suitability. Whilst the World Heritage Site of KSNP was shown to be the most important area, its continuing good habitat status for sun bears is now uncertain since its recent placement on the UNESCO Danger List due to the high levels of illegal logging and agricultural encroachment (UNESCO, ). Modelling sun bear habitat use This study demonstrates the use of a suitable and accurate method to model habitat suitability for sun bears and indeed other threatened large‐bodied mammals that have proved a challenge to study in the wild. The detection/non‐detection sampling framework accounts for imperfect detection (MacKenzie et al ., ) and allows for a number of predictor variables to be included, providing inferences about the factors influencing sun bear habitat use across multiple scales (Gray et al ., ). Consequently, this method modelled with environmental covariates resulted in a habitat suitability model with a strong predictive power, as tested by the separate camera trap survey. However, as the occupancy‐based sampling method relies on temporal closure of species populations during the survey period, this modelling assumption can put severe restrictions on the practicality of using occupancy to model species habitat suitability. Sun bears, and large‐bodied mammal species in general, often focus habitat use in areas where food resources are abundant (Schoen, ; Nomura & Higashi, ; Steinmetz et al ., ). This foraging behaviour is often referred to as the Ideal Free Distribution (IFD) theory, where species distribution is proportional to resource availability (Fretwell & Lucas, ). A key insight into the IFD theory is that as the density of individuals on a habitat patch increases, the suitability of that patch then decreases. Consequently, sun bear habitat use across the west‐central Sumatra region may be dependent on the species occupancy state in certain habitat types. However, this theory has often been disproved in experimental studies. For example, black bears, Ursus americanus , have been shown not to conform to the IFD theory but rather to the Ideal Despotic Distribution theory where subordinate individuals are constrained in their choice of area by dominant individuals (Beckmann & Berger, ). Furthermore, the IFD theory relies on a set of assumptions, one of which is that resource quality of each habitat patch does not change over time (Fretwell & Lucas, ), which is an unlikely explanation for sun bear habitat use in the Sumatra landscape. Food availability is likely to be higher in lower elevation forests. However, logging activities and agricultural expansion in these habitats can affect sun bear habitat use by altering foraging patterns. These human disturbances can prevent access to productive forage areas, reduce the abundance of key resources and alter fruit productivity, availability and distribution. Consequently, the predictive sun bear habitat suitability model derived from detection/non‐detection data does not describe core sun bear ecological niche characteristics but rather represents an adaptation to local conditions. This result is likely to be influenced by habitats that are highly fragmented, such as those found across the KS region and neighbouring forests. Sun bear habitat use derived from the detection/non‐detection data revealed that sun bears used a wide range of habitat types, but more fully use submontane and montane regions (ψ ≥ 0.9). As a result, these habitat types can be interpreted as suitable for sun bears. Consequently, management recommendations based on these findings should be treated with some caution. These habitat types still need to be considered against their relative importance because sun bear abundance will vary in these forest types. For example, previous studies have shown the importance of lowland forests < 500 m a.s.l. (Servheen, ; Wong et al ., , ; Steinmetz et al ., ), where food productivity is higher (Steinmetz et al ., ). It is likely that some of the factors, for example, accessibility, that are driving forest loss across west‐central Sumatra also influence sun bear habitat use in a similar way (Linkie et al ., ). Thus, the fine‐scale model may have been indirectly testing for forest loss, by incorporating factors that are related to accessibility, within the habitat use model. Additionally, this study only obtained sun bear data of altitudes up to 1791 m a.s.l. but the predictive habitat suitability model extrapolated high‐quality habitat into higher elevation forests (montane forests). Despite these areas having low human disturbances, vegetation is scarce. Subsequently, predictions of sun bear habitat use in montane forests over this elevation cannot be reliably made. The habitat suitability model developed in this study was based on 47 sites with detections and 78 sites without detections. A greater number of sampling sites would have been preferable; however, with a significant sampling effort of 10,935 trap nights over a wide geographic coverage, the detection/non‐detection data may give an adequate representation of the sun bear distribution across the entire study area. Implications for sun bear conservation management Most of the lowland forests in the KS region are being cleared by illegal logging and agricultural expansion (Jepson et al ., ; Linkie et al ., ), which remove large fruiting trees and other food sources that are predicted to influence sun bear detection (Steinmetz et al ., ). However, as there was no significant difference between sun bear detection probability in different land‐use types, it is likely that sun bears still use landscapes deemed unsuitable by the predictive habitat suitability model, that is, production and convertible forests. Yet, their predicted habitat use in these landscapes is a reflection of these areas having been degraded. The higher probabilities of sun bear habitat use in protected forest areas, particularly at higher elevations, may be an artefact of forest clearance (Rood et al ., ). Deforestation patterns have altered the suitability of lowland forests through their rapid and widespread conversion to poorer quality habitat types, that is, production and convertible forests. Sun bears are more likely to use undisturbed lowland forests that have higher fruit productivity than higher elevation forests (Steinmetz et al ., ). Previous sun bear studies have also shown that sun bear occupancy was significantly higher in degraded lowland forests than in primary montane forests (Linkie et al ., ), indicating that these degraded lowland habitats potentially offer better quality habitat than higher elevation forests. Subsequently, conservation strategies should be aimed at protecting the remaining lowland forests thus preserving suitable habitat for sun bears. Sun bears are predicted to use most of the forest inside KSNP, except in those areas where public roads enter the National Park, which reduces habitat quality. However, a recent study assessed the trends of four sun bear subpopulations in and around KSNP (Wong et al ., in press). Over the seven‐year monitoring period, the two subpopulations inside KSNP remained stable, whilst another showed signs of an increase. In contrast, the subpopulation located outside of KSNP, and in the area with the highest levels of deforestation, underwent a significant decline. In the past, there has been a lack of biological knowledge in land‐use planning in Sumatra, which has hindered the conservation of large‐bodied mammals through rapid and unmitigated agricultural expansion. However, through the recognition of the current state of Indonesia's forests, district and provincial governments are starting to integrate ecosystem services and biodiversity in their respective land‐use plans (Barano et al ., ). One approach that promotes the sustainable development of agriculture whilst accounting for the conservation of biodiversity and habitat is the High Conservation Value Forest (HCVF) approach (Jennings, ). The HCVF approach is now being used as a criterion for sustainable production in some of the fastest expanding agricultural plantation crops (Dennis et al ., ). In particular, the Roundtable of Sustainability Palm Oil has adopted the HCVF approach, making it a cornerstone of their sustainability standard. However, there are a number of critical weaknesses to the HCVF approach (Yaap et al ., ). One such weakness of the approach is that the quality and accuracy of assessments are dependent on the practitioner, so greater accuracy in methodologies and reporting is needed. This study identified several high‐quality patches outside the borders of KSNP for immediate protection. The predictive habitat suitability model developed in this study can therefore provide an accurate and cost‐effective framework for HCVF assessments that will promote the conservation of sun bears and other large‐bodied mammals across a large landscape. Consequently, for effective conservation of large‐bodied mammal species, priorities should be focused in areas of high habitat suitability through better protection and management (Sunarto et al ., ). Determining the effect of land‐use types on the habitat use of large‐bodied mammals is an important indicator that can be used to address the reassessment of land‐use plans and should be used in compliance with future spatial plans (Rood et al ., ). To conserve sun bears and other threatened large‐bodied mammals, conservation management strategies should include better definition and management for existing production and convertible forest areas, such as increasing productivity within these areas instead of expanding, establishment of protective forest buffer zones to maintain crucial elements of the landscapes and reintegration of degraded landscapes back into protected areas, whilst emphasizing the importance of HCVF and prohibiting further conversion of threatened forests. Acknowledgements We would like to thank the Indonesian Ministry of Forestry and Ir Luhut Sihombing (Director of KSNP) for granting permission to conduct this research and to Doddy Yu Saputra and Yoan Dinata for assisting our field research. We are grateful to WWF‐Indonesia for allowing us to use their deforestation data and Fauna & Flora International–Indonesia Programme for supplying field equipment. This study was possible due to the generous and much appreciated support from Rufford Small Grants, Chester Zoo, People's Trust for Endangered Species, Free the Bears Fund, Rare Species Conservation Trust, Cleveland Metroparks Zoo, Zoo Atlanta and 21st Century Tiger. We would like to thank Jose Lahoz‐Monfort and Lisanne Petracca for GIS guidance and Matt Struebig, Tatyana Humle, Katherine Zeller and two anonymous referees for useful comments that improved an earlier manuscript. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Diversity and Distributions Wiley

Managing sun bears in a changing tropical landscape

Download PDF
10 pages

Loading next page...
 
/lp/wiley/managing-sun-bears-in-a-changing-tropical-landscape-1rDlyv09sf

References (58)

Publisher
Wiley
Copyright
Copyright © 2013 John Wiley & Sons Ltd
ISSN
1366-9516
eISSN
1472-4642
DOI
10.1111/ddi.12020
Publisher site
See Article on Publisher Site

Abstract

Introduction Deforestation and unsustainable hunting in the tropics have degraded many landscapes to the point that they now no longer support viable populations of threatened species (Sodhi et al ., ; Bennett, ). Traditional approaches to conserving these landscapes and their wildlife populations have tended to focus on protected area management (IUCN, ; DeFries et al ., ). However, it is questionable whether the existing protected areas provide adequate habitat cover, quality or indeed protection for threatened species (Catullo et al ., ), as designation is often based on socio‐economic and political priorities rather than assessments of wildlife habitat requirements (Margules & Pressey, ). This has, for example, led to significant sized patches of lowland tropical forests being excised from protected areas for commercial logging (Daily et al ., ). The threat posed by deforestation tends to disproportionately affect large‐bodied mammals because of their large range requirements (Kinnaird et al ., ). Thus, to halt the ongoing loss of large‐bodied mammal populations, identification of high conservation value areas is required (Hoffmann et al ., ), but the detailed information on the distribution of species required for this purpose is often lacking (Rondinini et al ., ). The requisite distributional data for developing conservation priorities, whilst available for many species in the form of regional range data (Rodrigues et al ., ), are less useful for conservation agencies that require finer‐scale information, such as spatial distribution and use of habitat under varying levels of protection. The sun bear is a large‐bodied mammal species that historically ranged from India to Indonesia and China. Today, its range has contracted considerably. Sun bears are now patchily spread throughout much of this former range and have disappeared entirely from three of its 12 former range countries (Servheen, ). Agricultural expansion into forest habitat, especially on the islands of Indonesia and Borneo, and poaching for trade, especially in mainland Southeast Asia, are the primary drivers of this dramatic range and population reduction (Meijaard, ). The previously data deficient sun bear was recently reclassified by the International Union for Conservation of Nature (IUCN) as Vulnerable (IUCN, ). Yet, the conservation management of this species is hampered by a lack of basic, yet fundamentally important, ecological knowledge of habitat preferences and geographic distributions. Sun bears have a varied diet (Wong et al ., ; Fredriksson et al ., ), which enables the species to live in a range of habitats (Linkie et al ., ). Previous research in the Kerinci Seblat (KS) region, which estimated sun bear occupancy from four different habitat types, showed that occupancy was highest in a primary–secondary hill forest and lowest in a primary submontane forest (Linkie et al ., ). However, given the rapid rates of forest conversion across the KS region (Linkie et al ., ; Uryu et al ., ), sun bear distribution and occupancy in these habitats are likely to change. Sun bear research has predominantly investigated habitat use within lowland forest reserves and protected areas (Servheen, ; Wong et al ., , ; Fredriksson et al ., ). The value of higher elevation and unprotected forests for sun bears has therefore received little attention (Linkie et al ., ). This is a critical information gap because the disproportionately high and unyielding rates of deforestation in lower elevation forest are undoubtedly shaping the spatial distribution and habitat use patterns of sun bears, as well as other large‐bodied mammal species, in the tropics (Kinnaird et al ., ; Wibisono et al ., ). An important limitation for sun bear habitat management is poor land‐use planning, with large forest areas of high biological value still being cleared for plantations or commercial logging, which uses unsustainable practices (Fitzherbert et al ., ). Comparative studies on the influence of different land‐use types on the diversity and distribution of biodiversity, or components of it, are essential for policy makers and conservation managers (Gillison et al ., ). Here, we provide the first assessment of sun bear habitat use across a large tropical forest landscape in Sumatra that covers all of the main land‐use types, ranging from strictly protected areas and conservation areas, production forests and forests that are classified as legally convertible to non‐forest by Indonesian law, defined here as convertible forests. We investigated (1) sun bear habitat use at a broad‐scale looking at protection status, (2) habitat use at a finer‐scale within these study areas and (3) use the main findings to make broad recommendations for the conservation of sun bears, which should also apply to other large‐bodied mammals, living in human‐altered tropical landscapes. Methods Study area The west‐central Sumatra region consists of several land‐use types that include a protected area, Kerinci Seblat National Park (KSNP; 13,300 km 2 ); conservation area, Batang Hari Protection Forest (BHPF; 1,700 km 2 ); production forests (4,888 km 2 ) and convertible forests (1,110 km 2 ; Fig. a). This vast protected area network and diverse range of forest habitats should represent a stronghold for sun bears and other threatened species of large‐bodied mammals (Holden et al ., ; Linkie et al ., ). However, with a mean deforestation rate across the region of 1.10%yr −1 (2004–2008/9; Uryu et al ., ), forest conversion into agricultural farmlands represents a significant threat. The west‐central Sumatra project landscape covering Kerinci Seblat National Park (KSNP), Batang Hari Protection Forest (BHPF) and neighbouring forests, showing (a) the different land‐use types, and (b) the location of camera trap study areas. This study focused on six camera trapping areas: two inside KSNP; two that straddled the KSNP border with placement also in convertible and/or production forests; one inside BHPF; and one inside an active logging concession bordering BHPF, referred to hereafter as PT‐AMT (Fig. b; Table ). This covered three of the eight mainland Sumatran provinces (Bengkulu, Jambi and West Sumatra) and all of the main tropical forest types in which sun bears are known to live, encompassing selectively logged and primary lowland/hill forest to pristine montane forests (Servheen, ). These six areas have different levels of forest degradation and protection status (Table ). Camera trap study areas used for estimating sun bear habitat use in the west‐central Sumatra project landscape Study area Number of camera traps Camera trap area (km 2 ) Altitudinal range (m) Main habitat type and protection status Camera trapping dates PT‐AMT 18 124 365–1022 (mean: 653) Active logging concession bordering primary lowland/hill forest Feb–May 2008 Batang Hari Protection Forest (BHPF) 21 165 340–1461 (mean: 742) Primary hill/submontane forest bordering logging concession June–Sept 2008 Kerinci Seblat National Park (KSNP) Sipurak 21 88 694–1254 (mean: 901) Primary hill/submontane forest inside KSNP adjacent to ex‐logging concession Dec 2009–Mar 2010 Bungo 21 90 363–1630 (mean: 753) Primary–secondary hill forest predominantly inside an ex‐logging concession, with a few sites inside KSNP Apr–July 2010 RKE 23 104 947–1941 (mean: 1194) Primary submontane forest inside KSNP Aug–Nov 2010 Ipuh 21 118 145–1032 (mean: 511) Primary–secondary lowland forest predominantly inside an ex‐logging concession, with a few sites inside KSNP Nov 2010–Feb 2011 Data collection From February 2008 to April 2011, a detection/non‐detection sampling technique using camera trap data was used to estimate sun bear habitat use in the KS region and its neighbouring forests. Between 18 and 23 camera traps were deployed in each area, resulting in a total of 125 camera trap placements across the west‐central Sumatra project landscape (Table ). A combination of film and digital infrared camera models was used including PhotoScout (PTC Technologies Inc, Boston, MA, USA), Moultrie (Moultrie ™ , Alabaster, AL, USA) and Bushnell (Bushnell Corporation, Overland Park, KS, USA). Cameras were placed along ridge and animal trails at a height of approximately 0.5 m above the ground. To obtain a sufficient number of sampling sites at each study area, camera traps were spaced with a minimum distance of 2 km based on sun bear movement and home range data (Wong et al ., ). This camera spacing was to ensure that no sun bear within the sampling areas had a zero probability of being detected, thereby adhering to a critical assumption of the occupancy‐based method used in this study (MacKenzie et al ., ). Cameras were active 24 h per day, set with a 1‐min delay between exposures, programmed to record the time and date of each event and were active for approximately 90 days to avoid possible violation of the population closure assumption (Karanth & Nichols, ). Cameras were visited every 2 weeks to replace their film, memory cards and batteries, as well as to check their condition. Quantification of landscape covariates A dataset of 10 potential landscape variables that were thought to influence sun bear habitat use was constructed within ArcGIS v9.3 software (ESRI, Redlands, CA). The raw data were obtained from several sources: elevation and slope (Rabus et al ., ); the position of public roads, logging roads, rivers, villages and forest edge (Indonesian National Coordination Agency for Surveys and Mapping); land‐use type (protected areas, forest reserves, production forests and convertible forests; WWF‐Indonesia forest data); forest type (primary and degraded forest); and study area (PT‐AMT, BHPF, Sipurak, Bungo, RKE and Ipuh). Elevation and slope covariates were extracted at a 30 × 30 m resolution, and a single value per site was obtained by averaging all the pixel values within each sampling site (camera trap placement). The spatial covariates of distance to public roads, logging roads, rivers, villages and forest edge (from within the forest) were calculated within ArcGIS v9.3 using the Spatial Analyst extension from each sampling site coordinate. Statistical analysis and predictive model construction The 10 landscape variables were categorized as being either broad‐scale or fine‐scale. Broad‐scale variables included categorical covariates of land‐use type, forest type and study area. Fine‐scale variables included elevation, slope and distance to public roads, logging roads, rivers, villages and forest edge. The continuous landscape variable data extracted from each sampling site were imported into spss v.18.0 software ( spss Inc., Chicago, IL, USA) and logarithmically transformed to reduce the likelihood of extreme values having a disproportionate influence on the overall dataset (Royston & Sauerbrei, ). To test for non‐independence between covariates, a Spearman's rank correlation coefficient ( r s ) was calculated between pairwise variables. Correlated covariates were combined using principal component analysis (Abdi & Williams, ), using the data reduction technique in spss v.18.0 software, to produce a single covariate that was then used in the final data analysis. The covariates of distance to public roads and villages ( r s = 0.70), distance to forest edge and villages ( r s = 0.53) and distance to logging roads and elevation ( r s = 0.55) were correlated ( P < 0.05) and entered within unique combined covariates (i.e. Dist. public roads/villages; Dist. Forest edge/village; Dist. Logging/elevation). As different covariates can have different ecological effects, each correlated covariate was also modelled separately, but not in the same model. Due to the possibility of non‐independence between camera trap sampling sites, the Spatial Statistics Toolset extension in ArcGIS v9.3 was used to test for the presence of spatial autocorrelation across the camera trap sampling unit coordinates by calculating Moran's I statistic. This study applied a sampling design in which the number of camera trap sample sites ( n ) was visited multiple times on K sampling occasions. In turn, this allowed for inferences over sun bear habitat use, whilst accounting for imperfect detection, to be made. Detection/non‐detection data were collected from camera trap surveys to obtain encounter histories for sun bears. The detection (1) or non‐detection (0) sequence of sun bears over a 3‐month trapping duration per study area was then recorded as a detection history. Within this detection history, the 3‐month sampling period was divided into six consecutive 2‐week sampling occasions. Detection histories for each of the six study areas were combined to produce one dataset containing a total of 125 sampling sites. The resulting detection history framework ( n = 125, K = 6) was entered into PRESENCE v2.3 software (Hines, ). Within the PRESENCE software, a two‐stage logistic regression analysis was performed to determine which broad‐scale factors (i.e. land‐use type, study area and forest type) influenced sun bear habitat use (ψ) and then how several finer‐scale factors explained habitat use within these study areas. Models included detection probability as a constant, p (.), or as a function of site‐specific covariates, p (covariate). Candidate models were ranked by their second‐order information criterion (AIC c ) values, corrected for small sample sizes, and the Akaike weights ( w i ; Burnham & Anderson, ). The Akaike weight represents the ratio of ΔAIC c values for the whole set of candidate models, providing a strength of evidence for each model. The top‐ranking candidate model was used to determine the sun bear habitat use for the west‐central Sumatra project landscape. The corresponding beta (β) coefficients from the covariates in the top model, derived from PRESENCE software, were used to construct a predictive habitat suitability model ( P ) in ArcGIS v9.3 using the following equation: P = 1 1 + e − ( β 0 + ∑ β i X i ) where β 0 is the constant coefficient (intercept) and β 1 , β 2 , … β i represent the regression coefficients of the associated independent variables X 1 , X 2 … X i . The predictive power of the habitat suitability model was validated using sun bear habitat use data that were derived from a separate camera trapping survey collected from February 2011 to May 2011. Subsequently, an additional 40 camera trap sampling sites were selected within KSNP. This was achieved by randomly selecting four forest areas along the KSNP border that varied in habitat type. A total of 10 camera traps were deployed within each area along animal and ridge trails with a spacing greater than 2 km apart to minimize spatial autocorrelation. Camera traps were left for a period of 1 month at each sampling site. Detection/non‐detection data for each site were collected and entered into PRESENCE software to generate site‐specific habitat use estimates. A Spearman's rank correlation coefficient was used to test for a correlation between model estimated values of habitat use and those generated by the independent dataset for each corresponding 90 m 2 suitability model pixel (Linkie et al ., ). Results From February 2008 to April 2011, sun bears were camera trapped for a combined sampling effort of 10,935 trap nights comprising 1620 trap nights in PT‐AMT, 1737 in BHPF, 2042 in Sipurak, 1762 in Bungo, 1884 in RKE and 1890 in Ipuh. Sun bears were detected in 47 of the 125 units, corresponding to a naïve occupancy estimate of 0.376. Individual sun bear records ranged in altitude from 365 m to 1791 m above sea level (a.s.l.). Broad‐scale patterns At a broad‐scale, sun bear habitat use was found to be strongly influenced by land‐use type (Model 1.1; Table ). Modelling land‐use types as a function of sun bear habitat use produced the following beta coefficient values ( β i ± SE): β KSNP = 0.800 ± 0.193, β BHPF = −0.311 ± 0.242, β production forest = −1.12 ± 1.03 and β conertible forest = −1.70 ± 1.33. The positive beta regression coefficient for KSNP (protected area) indicates this land‐use type increases the probability of habitat use by sun bears, whilst the negative beta regression coefficients for production and convertible forests imply the opposite. Consequently, sun bear habitat use ( ψ ^ ± SE) was highest in the protected forests of KSNP (0.688 ± 0.092), followed by BHPF (0.621 ± 0.110), production forests (0.418 ± 0.121), and lowest in convertible forest (0.286 ± 0.122). However, modelling land‐use types as a function of sun bear detection probability, p (land use), found no significant differences between production forests (0.391 ± 0.059; ±SE), BHPF (0.371 ± 0.117), convertible forests (0.322 ± 0.113) and KSNP (0.305 ± 0.050; Model 1.3, Table ). Summary of model selection procedure for sun bear habitat use within 125 camera trap sites across the west‐central Sumatra project landscape Model no. Model K ΔAIC c w i X 2 P ‐value ĉ Broad‐scale analysis 1.1 ψ(Land use) p (.) 3 0.00 0.472 66.97 0.726 0.81 1.2 ψ(Land use + Study area) p (.) 4 1.87 0.185 55.76 0.523 0.91 1.3 ψ(Land use + Study area) p (Land use) 5 2.14 0.162 56.98 0.519 0.85 1.4 ψ(Study area) p (.) 3 2.15 0.161 53.98 0.672 0.82 1.5 ψ(.) p (.) 2 6.40 0.019 59.46 0.648 0.83 Fine‐scale analysis 2.1 ψ(Dist. public roads/villages + elevation) p (.) 4 0.00 0.321 5.39 0.631 0.82 2.2 ψ(Dist. public roads/villages) p (.) 3 0.42 0.261 5.40 0.592 0.85 2.3 ψ(.) p (.) 2 1.80 0.131 5.38 0.620 0.84 2.4 ψ(Dist. public roads/villages + elevation) p (Dist. public roads/villages) 5 2.14 0.110 5.48 0.570 0.86 2.5 ψ(Dist. public roads/villages) p (Dist. public roads/villages) 4 2.52 0.091 5.39 0.621 0.82 2.6 ψ(Elevation) p (.) 3 2.64 0.086 5.56 0.547 0.86 ψ is the probability a site is occupied by a sun bear and p is the sun bear detection probability where ψ(.) p (.) assumes sun bear presence and detection probability are constant across sites; K is the number of parameters in the model; ΔAIC c is the difference in AIC c values between each model; w i is the AIC c model weight; X 2 is the test statistic for model fit; P ‐value is the probability of observing a test statistic ≥ X 2 based on 999 parametric bootstraps; and ĉ is the estimated overdispersion parameter. Fine‐scale patterns At a fine‐scale, sun bear habitat use was best explained by distance to public roads/villages and elevation, and this model provided an adequate description of the data (Model 2.1; Table ). The summed model weights for each factor with respect to habitat use were as follows: distance to roads/villages (78.3%) and elevation (51.7%). Sun bears showed a preference for forest that was located further away from public roads and villages and at lower elevations (Table ). The corresponding estimated odd ratios for a one unit increase in each of the covariates, OR ^ j = e β ^ j , were OR ^ public roads/villages = 11.2 ± 0.05 and OR ^ elevation = 0.16 ± 0.12. This final model was not affected by spatial autocorrelation (Moran's I = 0.03, P > 0.05) and thus used to construct a predictive habitat suitability model. Logit models estimating sun bear habitat use, with beta (β) coefficient and standard error (SE) estimates for the top‐ranked models (from Table ) Model no. Model β 0 (constant) ± SE β 1 (Roads/villages) ± SE β 2 (elevation) ± SE 2.1 ψ(Dist. public roads/villages + elevation) p (.) 14.732 ± 2.48 2.414 ± 0.55 −1.844 ± 0.78 2.2 ψ(Dist. public roads/villages) p (.) 7.530 ± 2.52 1.96 ± 0.63 – 2.3 ψ(.) p (.) −0.332 ± 0.20 – – 2.4 ψ(Dist. public roads/villages + elevation) p (Dist. public roads/villages) 14.834 ± 2.47 2.443 ± 0.55 −1.844 ± 0.78 2.5 ψ(Dist. public roads/villages) p (Dist. public roads/villages) 7.631 ± 2.52 1.988 ± 0.63 – 2.6 ψ(Elevation) p (.) – – −1.226 ± 0.97 The predictive habitat suitability model showed that sun bears predominantly used habitats within the KSNP border (Fig. ). The proximity of public roads and villages to the forest areas outside KSNP reduced sun bear habitat suitability. Habitat quality in production and convertible forest areas was significantly lower than protected areas. The predicted values of sun bear habitat use were positively and significantly correlated with sun bear habitat use estimates that were derived from independent camera trap data ( n = 40, r s = 0.557, P < 0.001). Sun bear habitat suitability in west‐central Sumatra covering Kerinci Seblat National Park (KSNP), Batang Hari Protection Forest (BHPF) and neighbouring forests. Discussion This study is the first to assess sun bear habitat use across a large tropical landscape. The predictive habitat suitability model revealed how increased human disturbances, indicated by the presence of public roads, villages and elevation, negatively influenced sun bear habitat use across the west‐central Sumatra project landscape. Habitat use was predicted to be higher in protected areas than in forest reserves, production and convertible forests. Areas within the BHPF border ranked high in sun bear habitat use; however, the presence of public roads and villages surrounding the BHPF reduced its suitability. Whilst the World Heritage Site of KSNP was shown to be the most important area, its continuing good habitat status for sun bears is now uncertain since its recent placement on the UNESCO Danger List due to the high levels of illegal logging and agricultural encroachment (UNESCO, ). Modelling sun bear habitat use This study demonstrates the use of a suitable and accurate method to model habitat suitability for sun bears and indeed other threatened large‐bodied mammals that have proved a challenge to study in the wild. The detection/non‐detection sampling framework accounts for imperfect detection (MacKenzie et al ., ) and allows for a number of predictor variables to be included, providing inferences about the factors influencing sun bear habitat use across multiple scales (Gray et al ., ). Consequently, this method modelled with environmental covariates resulted in a habitat suitability model with a strong predictive power, as tested by the separate camera trap survey. However, as the occupancy‐based sampling method relies on temporal closure of species populations during the survey period, this modelling assumption can put severe restrictions on the practicality of using occupancy to model species habitat suitability. Sun bears, and large‐bodied mammal species in general, often focus habitat use in areas where food resources are abundant (Schoen, ; Nomura & Higashi, ; Steinmetz et al ., ). This foraging behaviour is often referred to as the Ideal Free Distribution (IFD) theory, where species distribution is proportional to resource availability (Fretwell & Lucas, ). A key insight into the IFD theory is that as the density of individuals on a habitat patch increases, the suitability of that patch then decreases. Consequently, sun bear habitat use across the west‐central Sumatra region may be dependent on the species occupancy state in certain habitat types. However, this theory has often been disproved in experimental studies. For example, black bears, Ursus americanus , have been shown not to conform to the IFD theory but rather to the Ideal Despotic Distribution theory where subordinate individuals are constrained in their choice of area by dominant individuals (Beckmann & Berger, ). Furthermore, the IFD theory relies on a set of assumptions, one of which is that resource quality of each habitat patch does not change over time (Fretwell & Lucas, ), which is an unlikely explanation for sun bear habitat use in the Sumatra landscape. Food availability is likely to be higher in lower elevation forests. However, logging activities and agricultural expansion in these habitats can affect sun bear habitat use by altering foraging patterns. These human disturbances can prevent access to productive forage areas, reduce the abundance of key resources and alter fruit productivity, availability and distribution. Consequently, the predictive sun bear habitat suitability model derived from detection/non‐detection data does not describe core sun bear ecological niche characteristics but rather represents an adaptation to local conditions. This result is likely to be influenced by habitats that are highly fragmented, such as those found across the KS region and neighbouring forests. Sun bear habitat use derived from the detection/non‐detection data revealed that sun bears used a wide range of habitat types, but more fully use submontane and montane regions (ψ ≥ 0.9). As a result, these habitat types can be interpreted as suitable for sun bears. Consequently, management recommendations based on these findings should be treated with some caution. These habitat types still need to be considered against their relative importance because sun bear abundance will vary in these forest types. For example, previous studies have shown the importance of lowland forests < 500 m a.s.l. (Servheen, ; Wong et al ., , ; Steinmetz et al ., ), where food productivity is higher (Steinmetz et al ., ). It is likely that some of the factors, for example, accessibility, that are driving forest loss across west‐central Sumatra also influence sun bear habitat use in a similar way (Linkie et al ., ). Thus, the fine‐scale model may have been indirectly testing for forest loss, by incorporating factors that are related to accessibility, within the habitat use model. Additionally, this study only obtained sun bear data of altitudes up to 1791 m a.s.l. but the predictive habitat suitability model extrapolated high‐quality habitat into higher elevation forests (montane forests). Despite these areas having low human disturbances, vegetation is scarce. Subsequently, predictions of sun bear habitat use in montane forests over this elevation cannot be reliably made. The habitat suitability model developed in this study was based on 47 sites with detections and 78 sites without detections. A greater number of sampling sites would have been preferable; however, with a significant sampling effort of 10,935 trap nights over a wide geographic coverage, the detection/non‐detection data may give an adequate representation of the sun bear distribution across the entire study area. Implications for sun bear conservation management Most of the lowland forests in the KS region are being cleared by illegal logging and agricultural expansion (Jepson et al ., ; Linkie et al ., ), which remove large fruiting trees and other food sources that are predicted to influence sun bear detection (Steinmetz et al ., ). However, as there was no significant difference between sun bear detection probability in different land‐use types, it is likely that sun bears still use landscapes deemed unsuitable by the predictive habitat suitability model, that is, production and convertible forests. Yet, their predicted habitat use in these landscapes is a reflection of these areas having been degraded. The higher probabilities of sun bear habitat use in protected forest areas, particularly at higher elevations, may be an artefact of forest clearance (Rood et al ., ). Deforestation patterns have altered the suitability of lowland forests through their rapid and widespread conversion to poorer quality habitat types, that is, production and convertible forests. Sun bears are more likely to use undisturbed lowland forests that have higher fruit productivity than higher elevation forests (Steinmetz et al ., ). Previous sun bear studies have also shown that sun bear occupancy was significantly higher in degraded lowland forests than in primary montane forests (Linkie et al ., ), indicating that these degraded lowland habitats potentially offer better quality habitat than higher elevation forests. Subsequently, conservation strategies should be aimed at protecting the remaining lowland forests thus preserving suitable habitat for sun bears. Sun bears are predicted to use most of the forest inside KSNP, except in those areas where public roads enter the National Park, which reduces habitat quality. However, a recent study assessed the trends of four sun bear subpopulations in and around KSNP (Wong et al ., in press). Over the seven‐year monitoring period, the two subpopulations inside KSNP remained stable, whilst another showed signs of an increase. In contrast, the subpopulation located outside of KSNP, and in the area with the highest levels of deforestation, underwent a significant decline. In the past, there has been a lack of biological knowledge in land‐use planning in Sumatra, which has hindered the conservation of large‐bodied mammals through rapid and unmitigated agricultural expansion. However, through the recognition of the current state of Indonesia's forests, district and provincial governments are starting to integrate ecosystem services and biodiversity in their respective land‐use plans (Barano et al ., ). One approach that promotes the sustainable development of agriculture whilst accounting for the conservation of biodiversity and habitat is the High Conservation Value Forest (HCVF) approach (Jennings, ). The HCVF approach is now being used as a criterion for sustainable production in some of the fastest expanding agricultural plantation crops (Dennis et al ., ). In particular, the Roundtable of Sustainability Palm Oil has adopted the HCVF approach, making it a cornerstone of their sustainability standard. However, there are a number of critical weaknesses to the HCVF approach (Yaap et al ., ). One such weakness of the approach is that the quality and accuracy of assessments are dependent on the practitioner, so greater accuracy in methodologies and reporting is needed. This study identified several high‐quality patches outside the borders of KSNP for immediate protection. The predictive habitat suitability model developed in this study can therefore provide an accurate and cost‐effective framework for HCVF assessments that will promote the conservation of sun bears and other large‐bodied mammals across a large landscape. Consequently, for effective conservation of large‐bodied mammal species, priorities should be focused in areas of high habitat suitability through better protection and management (Sunarto et al ., ). Determining the effect of land‐use types on the habitat use of large‐bodied mammals is an important indicator that can be used to address the reassessment of land‐use plans and should be used in compliance with future spatial plans (Rood et al ., ). To conserve sun bears and other threatened large‐bodied mammals, conservation management strategies should include better definition and management for existing production and convertible forest areas, such as increasing productivity within these areas instead of expanding, establishment of protective forest buffer zones to maintain crucial elements of the landscapes and reintegration of degraded landscapes back into protected areas, whilst emphasizing the importance of HCVF and prohibiting further conversion of threatened forests. Acknowledgements We would like to thank the Indonesian Ministry of Forestry and Ir Luhut Sihombing (Director of KSNP) for granting permission to conduct this research and to Doddy Yu Saputra and Yoan Dinata for assisting our field research. We are grateful to WWF‐Indonesia for allowing us to use their deforestation data and Fauna & Flora International–Indonesia Programme for supplying field equipment. This study was possible due to the generous and much appreciated support from Rufford Small Grants, Chester Zoo, People's Trust for Endangered Species, Free the Bears Fund, Rare Species Conservation Trust, Cleveland Metroparks Zoo, Zoo Atlanta and 21st Century Tiger. We would like to thank Jose Lahoz‐Monfort and Lisanne Petracca for GIS guidance and Matt Struebig, Tatyana Humle, Katherine Zeller and two anonymous referees for useful comments that improved an earlier manuscript.

Journal

Diversity and Distributions – Wiley

Published: Jul 1, 2013

There are no references for this article.