Occurrence of Bornean Mammals in Two Commercial Forest Reserves and Characteristics

Template B v3.0 (beta): Created by J. Nail 06/2015

Occurrence of Bornean mammals in two commercial forest reserves and characteristics

that influence their detectability

By TITLE PAGE Seth Timothy Wong

A Document Type. Submitted to the Faculty of Mississippi State University in Partial Fulfillment of the Requirements for the Degree of Master of Science in Wildlife, Fisheries and Aquaculture in the Department of Wildlife, Fisheries and Aquaculture

Mississippi State, Mississippi

December 2017

2017

Occurrence of Bornean mammals in two commercial forest reserves and characteristics

that influence their detectability

By APPROVAL PAGE Seth Timothy Wong

Approved:

______Jerrold L. Belant (Major Professor)

______Rahel Sollmann (Committee Member)

______Garrett M. Street (Committee Member)

______Kevin M. Hunt (Graduate Coordinator)

______George M. Hopper Dean College of Forest Resources

Name: Seth Timothy Wong ABSTRACT Date of Degree: December 8, 2017

Institution: Mississippi State University

Major Field: Wildlife, Fisheries and Aquaculture

Major Professor: Jerrold L Belant

Title of Study: Occurrence of Bornean mammals in two commercial forest reserves and characteristics that influence their detectability

Pages in Study: 47

Candidate for Degree of Master of Science

The Southeast Asian island of Borneo boasts an incredible diversity of terrestrial mammals which is threatened by habitat loss. Understanding the abundance and distribution of these species is essential for conservation and management. We assessed the occurrence of terrestrial mammals within two commercial forest reserves in Sabah,

Malaysian Borneo. In particular, we investigated the habitat associations of the Sunda stink-badger, whose patchy distribution is not well understood. To improve detection probability and precision of model parameters, we deployed 2 camera-traps at sample stations. Our results showed that Sunda stink-badgers are likely dietary and habitat generalists, that may benefit from forest disturbance. Additionally, we found that unguligrade species were associated with high detection probability when data from one camera trap was considered and inclusion of a second camera further increased the detectability of ungulates compared to all other species. We suggest that future studies consider physical characteristics of focal species to maximize effectiveness of camera effort and ensure that data collection is efficient and meets project needs.

DEDICATION

I dedicate this thesis to my family, Samson and Frances Wong, for their unconditional love and never-ending support in all my aspirations. To Linda Micke and friends, mentors, and colleagues at the San Francisco Zoo for their guidance, patience, and overwhelming support. Finally, I dedicate this thesis to the Hose’s Civet and Small

Carnivore Project, Borneo team. I would not be where and who I am today without all of you.

ACKNOWLEDGEMENTS

I thank Jerry Belant for giving me this opportunity and guiding me throughout this process. I am also grateful to my committee members, Rahel Sollmann and Garrett

Street, for their guidance and comments in writing this thesis. I thank the Sabah

Biodiversity Center for issuing a research permit and my local collaborator, Johnny

Kissing, along with all the staff from the Sabah Forestry Department for their active involvement and contributions to the field work. Thanks to Andreas Wilting, Azlan

Mohamed, John Mathai, and Jurgen Niedballa for their mentorship and giving me the opportunity to work in the field. My appreciation goes to all local field assistants, Azrie

Petrus, Alex Petrus, Bandy Marinson, Jereo Jaimin, Jeffendy Juari, Jacklan Juari, Roslen

Husin, Reyner Rupin, Alex de Jellson Julis, and Ladwin Rupin, without whom this work would not be possible. This project was supported by the German Federal Ministry of

Education and Research. Fieldwork was supported by the San Francisco Zoo and the

Point Defiance Zoo and Aquarium through the Dr. Holly Reed Conservation Fund.

Underlying RapidEye data was funded by the German Federal Ministry of Economy and

Energy and contributed on behalf of the German Aerospace Center.

iii

TABLE OF CONTENTS

DEDICATION ...... ii

ACKNOWLEDGEMENTS ...... iii

LIST OF TABLES ...... v

LIST OF FIGURES ...... vi

CHAPTER

I. INTRODUCTION ...... 1

References ...... 5

II. HABITAT ASSOCIATIONS OF THE SUNDA STINK-BADGER MYDAUS JAVANENSIS IN THREE FOREST RESERVES IN SABAH, MALAYSIAN BORNEO...... 7

Introduction ...... 7 Methods ...... 8 Results ...... 13 Discussion ...... 14 References ...... 22

III. INFLUENCE OF BODY MASS, SOCIALITY, AND MOVEMENT BEHAVIOR ON IMPROVED DETECTION PROBABILIES WHEN USING A SECOND CAMERA TRAP ...... 25

Introduction ...... 25 Methods ...... 27 Study area ...... 27 Camera survey ...... 28 Data preparation ...... 28 Occupancy modeling ...... 29 Influence of species traits ...... 30 Results ...... 31 Discussion ...... 32 References ...... 42

LIST OF TABLES

2.1 Mean, standard deviation (SD), and range of values for habitat covariates collected during 2008–2010 and 2014–2015 from three study designs conducted in three forest reserves: Deramakot, Tangkulap-Pinangah, and Segaliud Lokan; Sabah, Malaysian Borneo...... 18

2.2 Parameter estimates from candidate models for detection and occupancy of Sunda stink-badgers from surveys in Deramakot Forest Reserve (DFR), Tangkulap-Pinangah Forest Reserve (TPFR), and Segaliud Lokan Forest Reserve (SLFR), Sabah, Malaysia during years 2008–2010 and 2014–2015. Bold models only represent best supported models based on difference in Akaike’s Information Criterion score (ΔAIC). SE = standard error; CI = 95% confidence interval...... 19

3.1 Characteristics for 20 mammalian species with > 10 detections during a camera-trap survey conducted during October–December 2014 in Deramakot Forest Reserve, Sabah, Malaysian Borneo...... 37

3.2 Average number of records (± standard deviation [SD]) and sites (± SD) detected for 20 mammal species across 63 stations during a camera-trap survey conducted during October–December 2014 in Deramakot Forest Reserve, Sabah, Malaysian Borneo...... 38

3.3 Null occupancy and detection probability estimates and corresponding coefficient of variation (CV) values for 20 mammalian species detected across 63 stations during a camera-trap survey conducted during October–December 2014 in Deramakot Forest Reserve, Sabah, Malaysian Borneo...... 39

LIST OF FIGURES

1.1 Location of study areas: Deramakot Forest Reserve and Tangkulap Forest Reserve, Sabah, Malaysian Borneo...... 4

2.1 Location of three forest reserves on Borneo (black square; inset) and the areas sampled to estimate Sunda stink-badger occupancy during the coarse-grid surveys (October–December 2014 in Deramakot Forest Reserve; May–August 2015 in Tangkulap- Pinangah Forest Reserve), fine-grid survey (January–March 2015), and an earlier survey during 2008–2010...... 21

3.1 Location of Deramakot Forest Reserve, central Sabah, Malaysian Borneo (black square: inset)...... 41

CHAPTER I

INTRODUCTION

Understanding habitat associations and distributions of species is integral for effective conservation and management. This is particularly important in regions like

Southeast Asia, where biodiversity is under threat from large-scale habitat loss and conversion (Sodhi et al. 2004, 2010; Duckworth et al. 2012). Since the early 1900s, mammalian carnivore populations have declined as a result of habitat loss and fragmentation and direct exploitation through hunting (Karanth and Chellam 2009;

Lynam 2010; Ngoprasert et al. 2012). This is of concern due to the role carnivores play as top predators and keystone species that are essential in maintaining fully-functioning ecosystems (Treves and Karanth 2003).

On the Southeast Asian island of Borneo, there exists a suite of carnivore families, which includes Viverridae, Herpestidae, Mustelidae, and Mephitidae (Schreiber et al. 1989). This diverse mammalian community (hereafter, small carnivores) fills an array of roles, including predators and seed dispersers (Schreiber et al. 1989; Mudappa et al. 2010). However, due to fragmentation, loss of natural habitat, and hunting pressure, species within this assemblage are likely declining and threatened with extinction, according to the International Union for Conservation of Nature (Schipper et al. 2008;

Shepherd et al. 2011). Conserving small carnivores is important for the persistence of functional forest ecosystems. In turn, determining habitat factors that influence the

presence and distribution of small carnivores is important for identifying potential conservation areas and developing strategies for their persistence (Belant and Wilting

2013).

Studying small carnivores in tropical forests is challenging due to their elusive behavior and the logistical difficulties posed by the environment they live in (Mathai et al. 2010; Belant and Wilting 2013; Sunarto et al. 2013). Therefore, little is known about the ecological requirements and habitat associations of these species. Our understanding of secretive and elusive species within tropical forests can be improved by the use of remote cameras (Sunarto et al. 2013). Remote cameras require relatively little effort to survey large areas over extended periods of time (McCallum 2013; Sunarto et al. 2013).

However, the equipment is expensive, especially cameras that perform reliably under extreme tropical conditions (Sunarto et al. 2013). Therefore, survey designs need to consider financial and data requirements to answer proposed research questions.

In this study, we investigated the habitat associations of the Sunda stink-badger,

Mydaus javanensis, a little-studied mephitid restricted to Southeast Asia. We collected data during two 6-month surveys in two commercial forest reserves in Sabah, Malaysian

Borneo (Figure 1.1), each managed under a different level of timber exploitation. To determine whether we could improve the detection probability and occupancy estimates for Sunda stink-badger and other forest mammals we also assessed the influence of species physical characteristics and behaviors on detection probability and precision of parameter estimates in occupancy modeling with a second camera trap. Though employing additional effort at a survey station can benefit occupancy studies (Pease et al.

2016, O’Conner et al. 2017), it may not be necessary for all species. We provide

guidance on maximizing effectiveness of camera effort for occupancy studies to ensure efficient data collection based on project focus and needs.

Figure 1.1 Location of study areas: Deramakot Forest Reserve and Tangkulap Forest Reserve, Sabah, Malaysian Borneo.

White line represents the boundaries of each commercial forest reserve.

References

Belant, J. L. and Wilting A. (2013). Forward: Methods for detecting and surveying tropical carnivores. Raffles Bulletin of Zoology Supplement No. 28: i – iii.

Duckworth, J. W., Batters, G., Belant, J. L., Bennett, E. L., Brunner, J., Burton, J., Challender, D. W. S., Cowling, V., Duplaix, N., Harris, J. D., Hedges, S., Long, B., Mahood, S. P., McGowan, P. J. K., McShea, W. J., Oliver, W. L. R., Perkin, S., Rawson, B. M., Shepherd, C. R., Stuart, S. N., Talukdar, B. K., Van Dijk, P. P., Vie, J-C., Walston, J. L., Whitten, T., and Wirth, R. (2012). Why South-East Asia should be the World’s Priority for Averting Imminent Species Extinctions, and a Call to Join a Developing Cross-Institutional Programme to Tackle this Urgent Issue. S.A.P.I.E.N.S, 5(2): 77–92.

Karanth, U. K. and Chellam, R. (2009) Carnivore conservation at the crossroads. Oryx 43: 1–2.

Lynam, A. J. (2010). Securing a future for wild Indochinese tigers: Transforming tiger vacuums into tiger source sites. Integrative Zoology 5: 324–334.

Mathai, J., Hon, J., Juat, N., Peter, A., and Gumal, M. (2010). Small carnivores in a logging concession in the Upper Baram, Sarawak, Borneo. Small Carnivore Conservation 42:1–9

McCallum, J. (2013). Changing use of camera traps in mammalian field research: haitats, taxa and study types. Mammal Review 43:196–206.

Mudappa, D., Kumar, A., and Chellam, R. (2010). Diet and fruit choice of the brown palm civet Paradoxurus jerdoni, a viverrid endemic to the Western Ghats rainforest, India. Tropical Conservation Science 3(3):282–300.

Ngoprasert, D., Lynam, A. J., Sukmasuang, R., Tantipisanuh, N., Chutipong, W., Steinmetz, R., Jenks, K. E., Gale, G. A., Grassman, L. I., Kitamura, S., Howard, J., Cutter, P., Cutter, P., Leimgruber, P., Songsasen, N., and Reed, D. H. (2012). Occurrence of three felids across A network of protected areas in Thailand: prey, intraguild, and habitat associations. Biotropica 44:810–817.

Schipper, J., Hoffmann, M., Duckworth, J. W., and Conroy, J. (2008). The 2008 IUCN red listings of the world’s small carnivores. Small Carnivore Conservation 39:29–34.

Schreiber, A., Wirth, R., Riffel, M., and Van Rompaey, H. (1989). Weasels, Civets, Mongooses, and their Relatives: An action plan for the conservation of mustelids and viverrids. Gland, Switzerland: International Union for Conservation of Nature and Natural Resources.

Shepherd, C. R., Belant, J. L., Breitenmoser-Würsten, C., Duplaix, N., Ambu, L. N., and Wilting, A. (2011). Conservation challenges and opportunities for Borneo’s carnivores. TRAFFIC Bulletin 23:89–91.

Sodhi, N. S., Koh, L. P., Brook, B. W., and Ng, P. K. L. (2004). Southeast Asian biodiversity: an impending disaster. TRENDS in Ecology and Evolution. 19:654–660.

Sodhi, N. S., Posa, M. R. C., Lee, T. M., Bickford, D., Koh, L. P., and Brook, B. (2010). The state and conservation of Southeast Asian biodiversity. Biodiversity and Conservation 19:317–328.

Sunarto, Sollmann, R., Mohamed, A., and Kelly, M. J. (2013). Camera trapping for the study and conservation of tropical carnivores. Raffles Bulletin of Zoology Supplement No. 28:21–42.

Treves, A. and Karanth, U. K. (2003). Human-carnivore conflict and perspectives on carnivore management worldwide. Conservation Biology 17(6):1491–499.

CHAPTER II

HABITAT ASSOCIATIONS OF THE SUNDA STINK-BADGER MYDAUS

JAVANENSIS IN THREE FOREST RESERVES IN

SABAH, MALAYSIAN BORNEO

Introduction

The Sunda stink-badger, Mydaus javanensis, is one of only two species from the family Mephitidae occurring outside of the Americas. The species occurs on the islands of Sumatra, Borneo, Java, and the Natuna Islands, and inhabits diverse habitats including primary, secondary, and disturbed forests; open areas adjacent to forests; and oil palm plantations (Payne and Francis 1985). On Borneo, the species appears largely restricted to the Malaysian state of Sabah where it is often among the most frequently recorded carnivore species in camera-trap studies (Samejima et al. 2016). There are occasional sightings of Sunda stink-badger in the Malaysian state of Sarawak and the Indonesian state of Kalimantan where the species was last recorded in the early 20th century (Wilting et al. 2015).

It is unknown why Sunda stink-badger distribution on Borneo is patchy (Payne et al. 1985; Wilting et al. 2015), especially in Kalimantan. The apparent recent absence in

Kalimantan may be due to the species being formerly hunted (Bock 1882; Puri 1997;

Samejima et al. 2016). Soil type could be an important habitat factor as Sunda stink- badger occurrence may be linked to earthworm density and soil characteristics (Wilting et

al. 2015), though the species’ diet also includes bird eggs, carrion, insects, and plants

(Long and Killingley 1983; Neal and Cheeseman 1996; Payne and Francis 1985). Sunda stink-badger occurrence could also be influenced by forest type; badgers may select for disturbed or open forests as suggested by detection rates in oil palm plantations in Sabah

(Samejima et al. 2016).

We investigated environmental and anthropogenic factors that might influence occurrence of Sunda stink-badgers using camera-trapping data and occupancy modeling.

In particular, we assessed the hypothesis that Sunda stink-badger occurrence is positively associated with earthworm abundance. Additionally, we assessed the influence of forest disturbance on Sunda stink-badger occurrence, as the species is currently considered disturbance-tolerant (Samejima et al. 2016), suggesting at least limited use of disturbed forest.

Methods

We used camera-trap survey data from three forest sites: Deramakot Forest

Reserve (DFR; 5°14–28´N, 117°19–36´E), Tangkulap-Pinangah Forest Reserve (TPFR);

5°17–30´N, 117°11–21´E; Figure 1) and Segaliud Lokan Forest Reserve (SLFR; 5°20–

27´N, 117°23 –39´E). Currently, DFR (550 km2) and SLFR (572 km2) are designated as

Class II Production Forest (primarily intended for commercial timber production) under the Sabah Forest Enactment (Forest Enactment 1968; Kitayama 2013) while TPFR (501 km2) was reclassified in 2015 from Class II to Class I Protection Forest (area managed for biodiversity conservation and enhancement of ecosystem functions). Both forest reserves are certified by the Forest Stewardship Council (FSC; Forest Stewardship

Council 1996) and are currently managed by the Sabah Forestry Department. Since 1995, 8

DFR has experienced reduced-impact logging (RIL) strategies whereby placement of logging roads and skid trails and harvesting methods are conducted to reduce forest disturbance. Before 2001, TPFR was harvested using conventional selective logging techniques, but logging has ceased to allow forest regeneration. Similarly, SLFR, which has been privately managed by KTS Plantation Sdn Bhd. since 1994, was logged through conventional selective methods, with portions previously clear-cut. Since 1998, SLFR has implemented RIL practices. Overall, forest disturbance is greater in TPFR and SLFR than in DFR due to more intensive logging. Hunting is prohibited in all forest reserves

(Mohamed et al. 2013).

We surveyed within DFR during October–December 2014, establishing 64 stations at 2.5-km intervals (coarse-grid survey; Figure 1). At each station, we set two infrared motion detection cameras (Reconyx PC850 Hyperfire Whiteflash LED,

Reconyx, Wisconsin, USA), each on a logging road or animal trail about 5 m apart. We programed cameras to take 3 consecutive images during each detection, with camera options set to “no delay” between detections. We cleared vegetation to reduce false triggering of cameras. We checked cameras after 30 days to download images and replace batteries, then retrieved cameras after >60 days of operation. During May–August

2015, we replicated this coarse grid survey in TPFR. We sampled 49 km2 of DFR intensively during January–March 2015 (fine-grid survey). We established 64 stations in

16 clusters with each cluster containing 4 stations set in a 2 x 2 array with 500 m between stations. The 16 clusters were set in a 4 x 4 array with 1.5 km between edges of clusters.

Similar to the course-grid study, at each station, we placed 2 cameras <5 m apart and at

opposite sides of a feature (i.e. logging road or animal trail). Cameras were checked after

30 days and retrieved after >60 days of operation.

To increase records of Sunda stink-badger for occupancy modeling, we incorporated data collected during 2008–2010 (see Mohamed et al. 2013, Sollmann et al.

2017). The study comprised 47, 64, and 55 camera stations across 123 km², 122 km², and

114 km² of DFR, TPFR, and SLFR, respectively (Figure 1). Camera stations were set at

1.4-km intervals with 2 cameras at each station facing each other along logging roads or animal trails. Cameras were operational for 42 (DFR and TPFR) or 47 (SLFR) consecutive days before collection.

At each station during our coarse and fine-grid surveys, we estimated canopy cover and lateral obstruction from understory vegetation (Table 1). We established a 20 x

20 m plot using two 20-m ropes oriented along the cardinal directions and centered at the midpoint between the two cameras. Canopy cover data was obtained from images taken using a hand-held GPS (Garmin GPSmap 62sc, Garmin ltd, Canton of Schaffhausen,

Switzerland) at the center of the plot and 10 m in each cardinal direction. We converted images to black and white, where vegetation was represented with black pixels and sky was represented as white pixels. We used the GNU Image Manipulation Program (GIMP,

Gimp Team 2014) to estimate percentage canopy cover based on the ratio of black to total pixels and estimated mean percentage canopy cover of the 5 images taken at each station. To quantify lateral obstruction, we took images of understory vegetation from each plot center against an orange 1.5 x 1 m tarpaulin held at ground level and 10 m from the plot center in each cardinal direction. We edited images using GIMP software to crop images so only the area of the tarpaulin was represented. We converted images to black

and white, where vegetation was represented as black pixels and the tarpaulin as white pixels. We divided the number of black pixels by total pixels to estimate the percentage lateral obstruction in each image, and averaged values from the 4 images at each station to estimate the mean percentage lateral obstruction from understory vegetation.

For the fine-grid survey, we also estimated leaf litter cover and earthworm abundance at each station. We measured leaf litter cover in nine 1 x 1m subplots located at the center, 10 m in the cardinal directions, and 14.14 m in the ordinal directions of the plot. We assigned each subplot a value from 0 to 4, with 0 representing no leaf litter, 1 representing 0–25% cover, 2 representing 26–50% cover, 3 representing 51–75% cover, and 4 representing 76–100% cover. We estimated relative leaf litter cover at each station by taking an average of the 9 subplot values. We estimated worm abundance by counting worms in three 25 x 25 x 10 cm excavated samples of soil (Fragoso and Lavelle 1992;

Rickart et al. 1991). Samples were located 5 m from plot center at 0, 120, and 240 degrees.

Using a landcover map derived from RapidEye 5 m resolution satellite imagery

(RapidEye 2011) of the three forest reserves acquired during 2011 and 2012 (see

Niedballa et al. 2015), we calculated 5 covariates for the coarse-grid surveys and 2008–

2010 surveys: distance from each station to the nearest oil palm plantation, water, bare ground, and shrub land cover and forest score, a weighted mean of land cover percentages within a 50-m radius of each station. Distance to water provided an indication of accessibility to a water source, a basic resource for animals. Distance to oil palm, bare ground, and shrub were used as indices of potential edge effects and served as proxies for coarse-scale forest disturbance. Forest score was calculated to quantify the

degree of forest disturbance at each station (Niedballa et al. 2015). We assigned an integer value to each land cover pixel to rank forest disturbance: bare areas, grassland, oil palm plantations and water were assigned 0; shrub was assigned 1; secondary and degraded forest 2; and dense and primary forest 3. An average of pixel values produced a forest score from 0 to 3, with lower numbers indicating higher forest disturbance. Since stations were closer together in the fine-grid survey, there was little variation in distance to nearest land-cover type (i.e., land cover is locally autocorrelated). Therefore, only forest score was calculated for stations in the fine-grid survey and distance measures were excluded.

To investigate Sunda stink-badger habitat associations, we used likelihood-based occupancy modeling (MacKenzie et al. 2006). We used package camtrapR version 0.99.5

(Niedballa et al. 2016) in Program R version 3.2.2 (R Core Team 2015) to organize and extract metadata from camera images and create species detection/non-detection matrices for Sunda stink-badger records across all stations. We considered 5 days as a sample occasion and used only full 5-day occasions in detection matrices.

We implemented single season occupancy models in program R using the package unmarked version 0.11-0 (Fiske and Chandler 2011). Due to differences in study design and timing of data collection, we analyzed the fine-grid, coarse-grid, and 2008–

2010 surveys separately. To avoid over-parameterizing models and increasing uncertainty in estimates, we conducted a 2-stage modeling process. We first determined the most suitable model for detection probability using single covariate models for differences in camera placement (i.e. logging road or animal trail) and differences in site surveyed while holding occupancy probability constant across camera-trap stations. Covariates

were included in the final detection model only if they were the best supported model based on model ranking using Akaike’s Information Criterion (ΔAIC = 0, where all models with ΔAIC < 2 are considered equally supported; Burnham and Anderson 2002) and 95% confidence intervals (CI) of coefficient estimates did not include 0. Conditional on the selected detection model, we tested covariates for occupancy probability in the same manner. We estimated average detection and occupancy probability for each survey based on the best supported model. Particularly, we used the fine-grid survey to investigate worm abundance effects on occupancy, while we used distance to nearest land-cover type only in the coarse-grid and 2008–2010 survey models for occupancy probability.

Results

During the coarse-grid surveys, we obtained 36 detections (29 in DFR and 7 in

TPFR) of Sunda stink-badger over 8,415 trap-nights. Sunda stink-badger detection was lower in TPFR than in DFR; using this detection model, distance to water was the best supported model for probability of occupancy though the 95% CI of the coefficient included 0 (Table 2). Detection probability (mean + standard deviation [SD]) for the coarse-grid survey was 0.110 + 0.025 in DFR and 0.020 + 0.011 in TPFR with an average occupancy probability of 0.317 + 0.071.

During the fine-grid survey (DFR only), we obtained 71 Sunda stink-badger detections across 37 sites over 3,925 trap-nights. Sunda stink-badger detection was not affected by camera placement as 95% CI of the coefficient included 0 and the best supported model was the null model, therefore, detection was held constant. For occupancy in the fine-grid, neither earthworm abundance nor other habitat covariates had 13

significant effects on Sunda stink-badger occurrence as 95% CI for all coefficients included 0. The null model along with canopy cover and leaf litter cover models were equally supported (i.e., ΔAIC<2). Based on the null model, detection probability was

0.125 + 0.018 with an average occupancy probability of 0.717 + 0.088 for the fine-grid survey.

From the 2008–2010 survey, we obtained 216 detections (49 in DFR, 78 in TPFR, and 89 in SLFR) over 7,535 trap-nights. Neither camera placement nor site significantly influenced detection probability (i.e. 95% CI included 0 and null model was best supported model). With detection probability held constant, forest score was the best supported model (95 % CI: -2.471, -0.135), and Sunda stink-badger occupancy was greater at sites with lower forest scores. Average detection probability during the 2008–

2010 survey was 0.212 + 0.015 with average occupancy probability of 0.686 + 0.050.

Discussion

Our data from the fine-grid survey did not support the hypothesis that Sunda stink-badger occurrence is associated with earthworm abundance (Samejima et al. 2016), at least not at a local scale. We also found no association with leaf litter cover, which can affect the density of endogeic earthworms in other forest types (Jordan et al. 2009;

Nachtergale et al. 2002). Leaf litter in tropical forests harbors diverse and abundant invertebrates (Gonzalez and Seastedt 2000, 2001). Sunda stink-badgers reportedly feed on numerous foods including insects, bird’s eggs, carrion, and plants (Long and

Killingley 1983; Neal and Cheeseman 1996; Payne and Francis 1985). Our results suggest that Sunda stink-badgers do not rely on earthworms at a fine-scale and are more likely dietary generalists similar to other mephitids (Cantu-Salazar et al. 2005; Donadio 14

et al. 2004; Zapata et al. 2001). It is also possible that the variation in earthworm abundance captured in the fine-grid survey (Table 1) was not sufficient to detect any association with badger occurrence. There may also be a larger-scale influence of earthworm distribution and densities on badger occurrence: soils in DFR may be more fertile compared to other regions in Sabah, which in turn may be more fertile compared to other parts of Borneo where poorer soils may support fewer earthworms overall. The lack of a significant association with leaf litter cover may also be due to limited variation in leaf litter cover or the potential overall high abundance of food in our study sites.

Our results provided mixed support for the suggestion that Sunda stink-badgers are disturbance-tolerant (Samejima et al. 2016). In the fine-grid and coarse-grid surveys, we found no significant association between Sunda stink-badger occupancy and forest disturbance. However, for the 2008–2010 survey in which more Sunda stink-badgers were recorded, we found a higher probability of occupancy in more disturbed forest.

Forest scores for the coarse and fine grid surveys were consistently high (DFR: 2.40 ±

0.34 in coarse grid; 2.30 ± 0.35 in fine grid; TPFR: 2.13 ± 0.49 in coarse grid). In contrast, there was more variation in forest score across the three study sites in the 2008–

2010 survey (DFR: 2.02 ± 0.05, TPFR: 1.77 ± 0.44; SLFR: 1.67 ± 0.37), which facilitated our ability to detect an effect of forest score on badger occupancy. The association with forest disturbance found in the 2008–2010 survey could be due to the

Sunda stink-badger’s generalist diet. Generalist species are often adaptable to disturbed and degraded habitats (Devictor et al. 2008), in part due to the abundance of food such as insects (Lambert et al. 2006). Therefore, Sunda stink-badgers may benefit from, rather than tolerate, disturbed areas.

We found no significant relationship between Sunda stink-badger occupancy and canopy cover or lateral obstruction from understory vegetation. Other skunk species use open canopy and dense understory habitats for denning and predator avoidance (Doty and

Dowler 2006; Lesmeister et al. 2009). In forests, lower percentage canopy cover and dense understory vegetation are indicative of disturbed habitats (Lambert et al. 2006;

Mohamed et al. 2013). Our percentage canopy cover and lateral obstruction from understory vegetation represent a limited gradient for forest disturbance (Table 1), which may explain the lack of observed response within our study sites. In areas with greater disturbance, Sunda stink-badgers may use sites with lower canopy cover and higher understory density.

We found no significant association between occupancy probability and distance to oil palm plantation, bare earth, or shrub landcovers. However, Sunda stink-badgers were less likely to occur at larger distances to water, though that effect was not strong

(the CI of the coefficient included 0). It is likely that average distance to water in this study may have been too great (> 500 m) to reveal a stronger effect or association, though the ranging behavior of the species is not known. Similarly, mean distances to oil palm plantation were > 5.5 km. Sunda stink-badgers have been recorded in oil palm plantations and forest patches within plantations (Bernard et al. 2014; Yue et al. 2015). However, these records occurred in close proximity to forest, and presence of forest may be important for their occurrence in oil palm plantations. Though mean distances to bare earth and shrub landcovers were < 250 m, we did not find any strong associations with occupancy which supports why the species has been recorded in open gardens adjacent to forests and near villages (Payne et al. 1985; Samejima et al. 2016). As our results did not

suggest any strong relationships with particular land-covers, Sunda stink-badger are likely habitat generalists.

If we consider Sunda stink-badgers as habitat and dietary generalists, it remains unknown why the current distribution on Borneo is patchy. Our results represent conditions within three forest reserves sustainably managed or recovering from overharvest in central Sabah, Malaysia. In contrast, areas of Kalimantan where Sunda stink-badgers were previously considered common have experienced more intensive human disturbance (i.e. large-scale conversion to oil palm plantations; Samejima et al.

2016). There may be a disturbance threshold, where higher degrees of disturbance (e.g. large-scale conversion to oil palm and intensively logged areas) may result in lower probabilities of occupancy. Additional surveys across a larger gradient of habitat disturbance are needed to better assess the range of conditions that can support Sunda stink-badgers. Additionally, studies assessing the possible impact of hunting on Sunda stink-badger populations and distribution are warranted, as over-hunting remains a possible reason for the species’ patchy distribution on Borneo (Samejima et al. 2016).

Understanding the effects of habitat disturbance and hunting pressure (e.g. through hunter interviews) will be particularly useful for future management of the Sunda stink-badger as forests on Borneo continue to be degraded by human activities.

Table 2.1 Mean, standard deviation (SD), and range of values for habitat covariates collected during 2008–2010 and 2014–2015 from three study designs conducted in three forest reserves: Deramakot, Tangkulap-Pinangah, and Segaliud Lokan; Sabah, Malaysian Borneo.

Covariatea Mean SD Minimum Maximum Coarse-grid survey (2014–2015) Canopy cover 0.788 0.186 0.159 0.950 Lateral obstruction 0.525 0.159 0.180 0.878 Forest score 2.3 0.4 0.9 3.0 Distance to plantation 6.35 3.98 0.01 15.79 Distance to water 0.77 0.69 0.02 3.11 Distance to bare earth 0.23 0.24 0.00 1.19 Distance to shrub 0.03 0.04 0.00 0.20

Fine-grid survey (2015) Average earthworm abundance 2.4 1.6 0.0 8.7 Leaf litter cover 2.4 0.8 0.9 4.0 Canopy cover 0.820 0.137 0.378 0.924 Lateral obstruction 0.493 0.163 0.120 0.865 Forest score 2.3 0.3 1.4 3.0

2008–2010 Survey Forest score 1.8 0.5 0.4 2.8 Distance to plantation 5.65 3.24 0.06 13.53 Distance to water 0.47 0.44 0.00 2.21 Distance to bare earth 0.08 0.14 0.00 0.69 Distance to shrub 0.01 0.01 0.00 0.05 a Canopy cover and lateral obstruction are represented as proportion of vegetation coverage at each station. Distances are in km. Leaf litter cover is average of 9 categorical values from 0 to 4, with 0 representing no leaf litter, 1 representing 0–25% cover, 2 representing 26–50% cover, 3 representing 51–75% cover, and 4 representing 76–100% cover.

Table 2.2 Parameter estimates from candidate models for detection and occupancy of Sunda stink-badgers from surveys in Deramakot Forest Reserve (DFR), Tangkulap-Pinangah Forest Reserve (TPFR), and Segaliud Lokan Forest Reserve (SLFR), Sabah, Malaysia during years 2008–2010 and 2014–2015. Bold models only represent best supported models based on difference in Akaike’s Information Criterion score (ΔAIC). SE = standard error; CI = 95% confidence interval.

Survey Model Coefficient Estimate (SE) CI ΔAIC Coarse-grid survey (2014-2015) Detectiona Site b1(TFR) -1.540 (0.493) -2.507, -0.574 0 Null - - - 8.33 Camera placement b1(road) -0.419 (0.452) -1.305, 0.467 10.94 b2(mixed) 0.317 (0.780) -1.211, 1.845 Occupancyb

19 Distance to water b1(distance to water) -0.720 (0.382) -1.467, 0.027 0 Distance to bare earth b1(distance to bare) -0.445 (0.316) -1.063, 0.174 2.47 Null - - - 2.76 Canopy cover b1(canopy cover) 3.030 (2.390) -1.648, 7.709 2.82 Distance to shrub b1(distance to shrub) -0.265 (0.287) -0.827, 0.296 3.83 Distance to plantation b1(distance to plantation) -0.199 (0.322) -0.830, 0.431 4.37 Forest score b1(forest score) -0.369 (0.681) -1.702, 0.965 4.47 Lateral obstruction b1(lateral obstruction) 0.377 (1.950) -3.445, 4.200 4.73 Fine-grid survey (2015) Detectionc Null - - - 0.00 Camera placement b1(road) 0.045 (0.293) -0.528, 0.618 1.98 Occupancyd Canopy cover b1(canopy cover) -5.270 (4.210) -13.522, 2.982 0.00 Leaf litter cover b1(leaf litter cover) -0.775 (0.482) -1.720, 0.169 0.06 Null - - - 0.82

Table 2.2 (continued)

Forest score b1(forest score) -0.802 (1.030) -2.828, 1.224 2.18 Lateral obstruction b1(lateral obstruction) -0.812 (2.160) -5.043, 3.420 2.68 Earthworm abundance b1(earthworm abundacne) 0.097 (0.289) -0.469, 0.664 2.69 2008-2010 Survey Detectione Null - - - 0.00 Camera placement b1(road) -0.001 (0.173) -0.341, 0.338 2.00 Site b1(SLFR) -0.214 (0.247) -0.697, 0.269 3.23 b2(TFR) -0.125 (0.254) -0.622, 0.372 Occupancyd Forest score b1(forest score) -1.300 (0.596) -2.471, -0.135 0.00 Null - - - 4.42 Distance to plantation b1(distance to plantation) -0.274 (0.208) -0.680, 0.133 4.66

20 Distance to water b1(distance to water) -0.120 (0.199) -0.508, 0.269 6.07

Distance to bare earth b1(distance to bare) -0.103 (0.180) -0.456, 0.250 6.10 Distance to shrub b1(distance to shrub) 0.008 (0.203) -0.389, 0.407 6.41 a Occupancy was constant, and b1(TPFR) expresses the difference in detection relative to DFR; b1(road) and b2(mixed) express the difference in detection relative to animal trails. b Conditional on best detection model including different detection in survey site. c Occupancy is constant, and b1(road) expresses the difference in detection relative to animal trails. d Null model was best detection model; detection was held constant. e Occupancy was constant, and b1(road) expresses the difference in detection relative to animal trails; b1(SLFR) and b2(TPFR) express difference in detection relative to DFR.

Figure 2.1 Location of three forest reserves on Borneo (black square; inset) and the areas sampled to estimate Sunda stink-badger occupancy during the coarse- grid surveys (October–December 2014 in Deramakot Forest Reserve; May–August 2015 in Tangkulap-Pinangah Forest Reserve), fine-grid survey (January–March 2015), and an earlier survey during 2008–2010.

References

Bernard, H., Baking, E.L., Giordano, A.J., Wearn, O.R., Ahmad, A.H., 2014. Terrestrial mammal species richness and composition in three small forest patches within an oil palm landscape in Sabah, Malaysian Borneo. Mamm. Study 39, 141–154.

Bock, C.,1882. The Head-hunters of Borneo: a Narrative of Travel Up the Mahakkam and Down the Barito; also, Journeyings in Sumatra, second ed. Sampson Low, Marston, Searle, & Rivington, London.

Burnham, K.P., Anderson, D.R., 2002. Model Selection and Multi–Model Inference. Springer Link, Berlin.

Cantu-Salazar, L., Hidalgo-Mihart, M.G., Lopez-Gonzalez, C.A., Gonzalez-Romero, A., 2005. Diet and food resource use by the pygmy skunk (Spilogale pygmaea) in the tropical dry forest of Chamela, Mexico. J. Zool. Lond. 267, 283–289.

Doty, J.B., Dowler, R.C., 2006. Denning ecology in sympatric populations of skunks (Spilogale gracilis and Mephitis mephitis) in west-central Texas. J. Mammal. 87, 131–138.

Devictor, V., Julliard, R., Jiguet, F., 2008. Distribution of specialist and generalist species along spatial gradients of habitat disturbance and fragmentation. Oikos 117, 507– 514.

Donadio, E., Di Martino, S., Aubone, M., Novaro, A.J., 2004. Feeding ecology of the Andean hog-nosed skunk (Conepatus chinga) in areas under different land use in north-western Patagonia. J. Arid Environ. 56, 709–718.

Fiske, I.J., Chandler, R.B., 2011. Unmarked: an R package for fitting hierarchial models of wildlife occurrence and abundance. J. Stat. Softw. 43, 1–23.

Sabah Forest Enactment, 1968. State of Sabah Forest Enactment, 1968 (Sabah No. 2 of 1986).http://www.lawnet.sabah.gov.my/lawnet/sabahlaws/StateLaws/ForestEnact ment1968.pdf (accessed 10 January 2017).

Forest Stewardship Council, 1996. FSC principles and criteria for forest stewardship. https://ic.fsc.org/en/certification/principles-and-criteria (accessed 10 January 2017).

Fragoso, C., Lavelle, P., 1992. Earthworm communities of tropical rainforests. Soil Biol. Biochem. 24, 1397–1408.

Gonzalez, G., Seastedt, T.R., 2000. Comparison of the abundance and composition of litter fauna in tropical and subalpine forests. Pedobiologia 44, 545–555.

Gonzalez, G., Seastedt, T.R., 2001. Soil fauna and plant litter decomposition in tropical and subalpine forests. Ecology 82, 955–964.

Jordan, D., Hubbard, V.C., Ponder Jr, F., Berry, E.C., 2000. The influence of soil compaction and the removal of organic matter on two native earthworms and soil properties in an oak-hickory forest. Biol. Fertil. Soils 31, 323–328.

Kitayama, K., 2013. Co-benefits of Sustainable Forestry: Ecological Studies on a Certified Bornean Rain Forest. Springer Japan, Tokyo.

Lagan, P., Mannan, S., Matsubayashi, H., 2007. Sustainable use of tropical forests by reduced-impact logging in Deramakot Forest Reserve, Sabah, Malaysia. Ecol. Res. 22, 414–421.

Lambert, T.D., Malcolm, J.R., Zimmerman, B.L., 2006. Amazonian small mammal abundances in relation to habitat structure and resource abundance. J. Mammal. 87, 766–776.

Lesmeister, D.B., Gompper, M.E., Millspaugh, J.J., 2009. Habitat selection and home range dynamics of eastern spotted skunks in the Ouachita Mountains, Arkansas, USA. J. Wildl. Manage. 73, 18–25.

Long, C.A., Killingley, C.A., 1983. The badgers of the world. Charles C. Thomas, Springfield.

MacKenzie, D.I., Nichols, J.D., Royle, J.A., Pollock, K.H., Bailey, L.L., Hines, J.E., 2006. Occupancy estimation and modeling: inferring patterns and dynamics of species occurrence. Academic Press, Burlington.

Mohamed, A., Sollmann, R., Bernard, H., Ambu, L.N., Lagan, P., Mannan, S., Hofer, H., Wilting, A., 2013. Density and habitat use of the leopard cat (Prionailurus bengalensis) in three commercial forest reserves in Sabah, Malaysian Borneo. J. Mammal. 94, 82–89.

Nachtergale, L., Ghekiere, K., Schrijver, A.D., Muys, B., Luyssaert, S., Lust, N., 2002. Pedobiologia 46, 440–451.

Neal, E., Cheeseman, C., 1996. Badgers. T & AD Poyser, London.

Niedballa, J., Sollmann, R., Mohamed, A., Bender, J., Wilting, A., 2015. Defining habitat covariates in camera-trap based occupancy studies. Sci. Rep. 5, doi:10.1038/srep17041

Niedballa, J., Sollmann, R., Courtiol, A., Wilting, A., 2016. camtrapR: an R package for efficient camera trap data management. Methods Ecol. Evol. 7, 1457–1462.

Payne, J., Francis, C.M., Phillipps, K., 1985. A field guide to the mammals of Borneo. The Sabah Society and WWF Malaysia, Kota Kinabalu and Kuala Lumpur.

Puri R., 1997. Hunting Knowledge of the Penan Benalui of East Kalimantan, Indonesia. (Unpublished PhD Thesis). University of Hawaii, Hawaii.

R Core Team, 2015. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna. https://www.R-project.org.

RapidEye, 2011. Satellite Imagery Product Specifications Version 3.2.

Rickart, E.A., Heaney, L.R., Utzurrum, R.C.B., 1991. Distribution and ecology of small mammals along an elevational transect in Southeastern Luzon, Philippines. J. Mammal.72, 558–469.

Samejima, H., Meijaard, E., Duckworth, J.W., Yatsuma, S., Hearn, A.J., Ross, J., Mohamed, A., Alfred, R., Bernard, H., Boonratana, R., Pilgrim, J.D., Eaton, J., Belant, J.L., Kramer-Schadt, S., and Wilting, A., 2016. Predicted distribution of the Sunda Stink-badger Mydaus javanensis (Mammalia: Carnivora: Mephitidae) on Borneo. Raff. Bull. Zool. 33, 61–70.

Sollmann, R., Mohamed, A., Niedballa, J., Bender, J., Ambu, L., Lagan, P., Mannan, S., Ong, R.C., Langner, A., Gardner, B., and Wilting, A., 2017. Quantifying mammal biodiversity co-benefits in certified tropical forests. Diversity Distrib. 23, 317– 328.

The GIMP team, 2014. GNU Image Manipulation Program. www.gimp.org (accessed 31 July 2014).

Wilting, A., Duckworth, J.W., Meijaard, E., Ross, J., Hearn, A., Ario, A., 2015. Mydaus javanensis. The IUCN Red List of Threatened Species 2015: e.T41628A45209955.http://dx.doi.org/10.2305/IUCN.UK.20154.RLTS.T41628A 45209955.en.

Yue, S., Brodie, J.F., Zipkin, E.F., Bernard, H., 2015. Oil palm plantations fail to support mammal diversity. Ecol. Appl. 25, 2285–2292.

Zapata, S.C., Travaini, A., Martinez-Peck, R., 2001. Seasonal feeding habits of the Patagonian hog-nosed skunk Concepatus humboldtii in southern Patagonia. Acta Theriol. 46, 97–102.

CHAPTER III

INFLUENCE OF BODY MASS, SOCIALITY, AND MOVEMENT BEHAVIOR ON

IMPROVED DETECTION PROBABILIES WHEN USING

A SECOND CAMERA TRAP

Introduction

Understanding species distributions and habitat use are necessary for effective conservation and management (Kanagaraj et al. 2013). Modeling species distributions can predict areas where species may occur and help identify areas with potential for human-wildlife conflict (e.g. Kanagaraj et al 2011; Ngoprasert et al. 2011). Distribution models are based on occurrence data which are obtained from direct observation or indirect records such as scat and tracks (Kanagaraj et al. 2013; Sunarto et al 2013), but many wildlife species are rare and elusive, making studies employing direct observation difficult (Mathai et al. 2010; Sunarto et al. 2013). Consequently, many studies have based their findings on indirect surveys for scat and tracks (Smith et al. 2005; Mortelliti and

Boitani 2008). Specifically, remote cameras have emerged as a common tool for collecting reliable data on elusive species without the need for direct observation

(Sunarto et al 2013).

Remote cameras are useful for collecting species records across large areas with relatively little effort (O’Connell et al. 2011). One issue with cameras is that they represent a highly directional and point-based sampling method; even species in the

immediate vicinity of cameras will not be detected if they do not pass the camera’s sensor field. This causes false-negative errors which can result in incorrect conclusions about species occurrence and habitat associations (Tyre et al. 2003). Deploying additional cameras at a sampling station can increase the number of species detected, number of sites where a species is detected, improve detection probability, and reveal habitat associations that would otherwise be missed with fewer detectors (Pease et al. 2016;

O’Conner et al. 2017). Adding a second camera, however, increases equipment cost, and the effectiveness of an additional detector may vary depending on species physical characteristics and behaviors.

A positive association has been observed between species detection and body size

(Lyra-Jorge et al. 2008; Tobler et al. 2008). Also, sociality may influence species detectability, with social species that travel or forage in groups more likely to be detected

(Treves et al. 2010). Movement speed can affect detectability (Rowcliffe et al. 2011), where animals that move quickly past a camera’s detection zone are less likely to be detected. Deploying an additional camera at sample sites may therefore be particularly useful for smaller-bodied, solitary, and faster-moving species.

We evaluated species characteristics that may influence detection probability estimates derived from one- or two-camera based occupancy models, using empirical data from a survey in Sabah, Malaysian Borneo. Given our working hypothesis that animal detections are influenced by physical characteristics and behavior, we assessed the effect of species body mass, social behavior, dietary niche, and foot posture (the latter 2 characteristics as indices of movement behavior and potential speed). We predicted that

1) larger bodied, social, and slower moving species would have high detection probability

with one camera, 2) detection probability would be higher for species exhibiting multiple of these traits, and 3) Species with high one-camera detection rates have a lower difference in detection probability between 2- and 1-camera designs. Understanding the effects of species physical characteristics and behavior associated with detection probability can aid in planning future camera surveys.

Methods

Study area

We conducted our survey in Deramakot Forest Reserve (DFR) located in the lowlands of central Sabah, Malaysian Borneo. Since 1995, DFR (550 km2) has experienced reduced-impact logging strategies whereby placement of logging roads and skid trails and harvesting methods are designed to reduce forest disturbance. In 1997, it received certification from the Forest Stewardship Council (Lagan et al. 2007). Currently,

DFR is managed by the Sabah Forestry Department and is designated as Class II

Production Forests (primarily intended for commercial timber production) under the

Sabah Forest Enactment (Forest Enactment 1968; Kitayama 2013). There is little seasonal variation in rainfall and temperature within DFR (Kitayama 2013). Mean annual rainfall during 2008–2010 was about 309 cm and was evenly distributed throughout the year. The mean annual temperature of DFR is 27.0°C, with a mean daily temperature of

25.2°C (mean daily minimum of 19.8°C and mean daily maximum of 35.5°C). Hunting is illegal within the forest reserve (Lagan et al. 2007; Mohamed et al. 2013).

Camera survey

We surveyed the terrestrial mammal community within DFR during October–

December 2014, establishing 63 stations in a grid with 2.2-km spacing (Figure 1). At each station, we set two infrared motion detection cameras about 5 m apart (Reconyx

PC850 Hyperfire Whiteflash LED, Reconyx, Wisconsin, USA) and 30–45 cm above ground, each oriented toward a logging road or animal trail. We programed cameras to take 3 consecutive images during each detection, with camera options set to “no delay” between detections. We cleared vegetation to reduce false triggering of cameras. We checked cameras after 30 days to download images and replace batteries, then retrieved cameras after >60 days of operation.

Data preparation

We identified mammals in images to species, with muntjacs (Bornean yellow muntjac (Muntiacus antherodes) or red muntjac (M. muntjac)), mongooses (collared mongoose (Herpestes semitorquatus) or short-tailed mongoose (H. brachyurus)), and mousedeer (greater mousedeer (Tragulus napu) or lesser mousedeer (T. kanchil)) identified only by genus due to similarities in body mass, sociality, dietary niche, and foot posture between sister species. We used the package “camtrapR” version 0.99.5

(Niedballa et al. 2016) in program R version 3.2.2 (R Core Team 2015) to organize and build a record database from all cameras. We recorded the average percent of independent detections (< 30 minutes apart) that were recorded on both cameras at a station (+ standard deviation, SD). For each mammalian species or pair with >10 detections at multiple stations, we conducted a literature search to determine average body mass, sociality (social or solitary), dietary niche (herbivore, omnivore, or 28

carnivore), and foot posture (plantigrade, digitigrade, or unguligrade; Table 1). We considered dietary niche as an indicator for how a species moves while foraging and passing camera detection zones. Particularly, all species which primarily consume motile prey were considered carnivorous. We used foot posture classification as an alternate indicator for the manner and potential speed at which a species moves.

Occupancy modeling

Since the number of detections varied among cameras within a station, we bootstrapped the random selection of a single camera from each station for 150 iterations to generate datasets based on a one-camera design. For each iteration, we used the package “camptrapR” (Niedballa et al. 2016) to generate detection/non-detection matrices for each mammal species with more than 10 detections across all stations. We considered 5 days as a sample occasion and used only full 5-day occasions in detection matrices. We implemented single season occupancy models (MacKenzie et al. 2006) in program R using the package “unmarked” version 0.11-0 (Fiske and Chandler 2011). We estimated the null (i.e., no covariates) occupancy and detection probabilities for each species in each iteration. From these 150 iterations, we took a subset of the first 50 models which converged and calculated standard errors and confidence intervals for both occupancy and detection probability estimates. For each parameter estimate, we also calculated the coefficient of variation (CV) to represent precision of estimates. We then calculated the average occupancy and detection probability (± SD) and average CV (±

SD) for estimates across all species from the subset of 50 model iterations. Using the full record database, which included all records taken by both cameras at each station, we

estimated null occupancy and detection probability for each mammalian species with >

10 detections and calculated a CV for these parameters.

To evaluate the effect of using a single-camera design, we compared parameter

CVs between one-camera and two-camera detection estimates for each species. We also calculated the confidence interval (CI) coverage, which is the proportion of 95% CIs for estimates of detection probability from the one-camera design which included the corresponding two-camera estimate for each species.

Influence of species traits

We fitted linear models describing the influence of body mass, social behavior, dietary niche, foot posture, and two-way interactions between each of these traits on detection probability for each species. To avoid model overfitting and improve interpretability, we used the least absolute shrinkage and selection operator (LASSO), a regularization technique which penalizes the magnitude of coefficients through a tuning parameter λ (Tibshirani 1996). We performed logit transformation of detection probability response variable before implementing LASSO regression in program R using the package “glmnet” version 2.0-10 (Friedman, Hastie, and Tibshirani 2010). We performed 5-fold cross validation (due to sample size, default 10-fold cross validation was not appropriate for our data; Kohavi 1995) to determine the optimal value of λ and assessed the resulting model coefficient values. We replicated this process with the difference between two-camera and one-camera detection estimates for each species as the response variable. Difference in detection values were cube root transformed

(transformation method accounting for both positive and negative values) to fit normal

distribution for linear modeling. We assessed resulting model coefficients for traits associated with an increase in detection probability.

Results

We collected 4198 records of 35 mammal species over 4390 trap nights. An average of 18.82% of detections (± 10.34%) occurred on both cameras across all stations.

We collected >10 detections for 20 species for which we compiled information on average body mass, sociality, dietary niche, and foot posture (Table 1). Of these 20 species, sambar deer (Rusa unicolor), muntjac, mousedeer, bearded pig (Sus barbatus), pig-tailed macaque (Macaca nemestrina), and Malay civet (Viverra tangalunga) were detected frequently with at least an average of 110 detections from a single camera trap.

Total number of detections for each species (with >10 detections) in the two-camera design was about twice the average number of records from individual cameras (Table 2).

Species were recorded on average at 7.0 (± 3.6) more stations in the two-camera design.

Null estimates of occupancy probability increased from a one-camera to two- camera design by 1.3–28.8% (with an average increase of 11.5% + 8.1%) across all species excluding Asian elephant (Elephas maximus) and Sunda pangolin (Manis javanica) where the occupancy probability estimate decreased by 0.2% and 18.3% respectively. Coefficients of variation for occupancy probability estimates decreased from a one-camera to two-camera design by 1.12– 21.88 (average decrease of 12.47 +

9.22) for all species excluding Sunda pangolin where the CV increased by 1.55.

Confidence interval coverage for occupancy probability estimates was <25% for three species: 0% for sambar deer and pig-tailed macaque, with both species having a

probability of occupancy estimate of 100%, and 10% for Malay civet. For all other species, CI coverage ranged from 46%–100% (mean = 81%).

Null estimates of detection probability increased from the one-camera to the two- camera design by 0.1–19.2% across all species (Table 3), with an average increase of

5.2% (+ 5.1%). The CVs of two-camera detection probability estimates decreased by 1.3–

29.7 for all species compared to one-camera estimates, with an average decrease of 23.1

(+ 8.3). Confidence interval coverage from one-camera detection probability estimates of the corresponding two-camera detection estimate was >25% for five species: 0% coverage for muntjac, mousedeer, and bearded pig; 16% coverage for sambar deer; and

8% coverage for Malay civet. For all other species, CI coverage ranged from 42 to 100%

(mean = 81%).

Only foot posture was influential on the single-camera detection probability. An unguligrade foot posture had a positive influence on detection probability (ß = 0.902), while digitigrade and plantigrade foot postures had coefficients = 0. The cross-validated

LASSO procedure suggested no other variables were associated with detection probability as all coefficient values were 0. Foot posture was also the single influential predictor for difference between 2-camera and 1-camera detection probabilities. An unguligrade foot posture had a positive influence on the difference in detection probability with ß = 0.128. All other coefficient values were zero.

Discussion

Our results of null occupancy and detection estimates for 20 mammal species from one- and two-camera stations support previous suggestions that multiple cameras improve species detection probabilities (Pease et al. 2016, O’Conner et al. 2017). 32

Addition of a second camera trap increased the number of detections and sites detected for all species as animals present at a sample site were likely missed less often. Increasing the number of cameras can therefore benefit camera studies where detection rates for species are low (e.g. Ahumada et al. 2017). Occupancy estimates generally increased along with precision of estimates, where occupancy estimates for both sambar deer and pig-tailed macaque increased to 100% with two cameras likely due to the high proportion of presence in detection histories. For Sunda pangolin, the nature of the detection history

(spatial and temporal distribution of additional detections and sites detected) using two cameras likely led to lower estimates of occupancy probability compared with one camera estimates, though coefficient of variation for estimates were similar between one- and two- camera designs.

Precision of parameter estimates is important in avoiding misguided habitat associations which can be used to determine further research and management decisions

(Tyre et al. 2003). As investigating habitat associations is a primary question addressed by occupancy studies, additional sampling can reveal relationships otherwise missed with less effort (Pease et al. 2016). Lower coefficients of variation for parameter estimates in two-camera models demonstrated that deploying a second camera at a station can improve precision of parameter estimates (i.e. MacKenzie and Royle 2005). However, our null-model parameter estimates showed relatively high (>25%) confidence interval coverage for many species, where the 95% confidence intervals of one-camera model estimates included the two-camera estimates. Occupancy probability estimates for sambar deer, pig-tailed macaque, and Malay civet had relatively low (<25%) confidence interval coverage, and detection probability estimates for muntjac, mousedeer, bearded pig, and

Malay civet also had relatively low to no confidence interval coverage. This suggests that a second camera in occupancy studies would be beneficial for occupancy estimates and detectability of species with similar body mass or social and movement behaviors.

We found that foot posture influenced one-camera detection probability and precision of parameter estimates. Unguligrade species had higher detection probabilities compared to digitigrade and plantigrade species. Though unguligrade species are capable of high speeds (Halfpenny and Biesiot 1986) and faster moving species are less likely to be detected (Rowcliffe et al. 2011), unguligrade species are potentially slower moving due to foraging and vigilance behaviors (e.g. Little et al. 2014; 2016). Contrasting with initial predictions, when data from a second camera was incorporated, unguligrade species were also associated with higher differences in detection probability between one- and two-camera designs compared to all other species. Our results suggest that though one camera is likely sufficient for detecting ungulates, an additional camera at sites would further increase their detectability. Improving detectability and precision of parameter estimates for ungulates are important because they are often the primary target for hunters and therefore susceptible to over-exploitation (Brodie et al. 2014; Hoffmann et al. 2015).

Though sociality was suggested to increase detectability of species (Treves et al.

2010), we did not find associations between social behavior and single-camera detection probability or differences in detection probability estimates. Our results suggest that detection probability for social and solitary species are equally variable, and the improvement in detectability for social species was lower than improvements associated with foot posture. Also, we did not find any effect of species’ dietary niche on detection

probability estimates. Previous studies have highlighted that carnivores have higher detectability due to a tendency toward traveling and foraging along well-defined trails and roads in contrast to the movement and foraging strategies of omnivores and browsing ungulates (Rowcliffe et al. 2011, Cusack et al. 2015). Though we did not investigate the specific influence of camera placement along animal trails and logging roads on detection probability, our results showed that both detection probability estimates with one camera and improvements in detection between one and two cameras were equally variable across carnivores, omnivores, and herbivores. Social behavior and dietary niche did not have strong associations with detection probability when one camera was used and likely are not associated with improvements in detection probability with a second camera trap.

Beyond our analysis of ecological traits, we observed higher improvements in detection probability with a second camera for species which were detected frequently

(>50 detections from one camera), though changes in precision of parameter estimates based on CVs were low. Conversely, species that were detected less frequently (<50 detection using one camera) had greater improvements in precision of parameter estimates with the two-camera design but smaller changes in detection probability. An increase in precision of parameter estimates for less frequently detected species is likely due to an increase in the number of sites detected while detection probability is less improved with the few additional detections from two cameras. The increase in detection probability for frequently detected species is due to the larger number of additional detections from a second camera, while the number of sites detected from one camera is already sufficient for precise parameter estimates. Therefore, our results provide further support for suggestions that increasing the number of cameras at a station to improve

detection and precision of parameter estimates is situational and species-dependent

(Pease et al. 2016; O’Conner et al. 2017). For occupancy studies where the modeling process accounts for imperfect detection, a second camera may only benefit species that are detected less frequently by improving precision of parameter estimates.

Investing in a second camera at a sample station may sometimes be beneficial, but incurs additional costs. For equipment-limited projects, this can also result in reduced survey areas which can limit the detection of rare species (MacKenzie and Royle 2005).

Though our study and previous work (Mackenzie and Royle 2005, Pease et al. 2016;

O’Conner et al. 2017) have suggested that additional camera effort produces additional detections, increases the number of sites where species are detected, and improves detection probability, we stress that the magnitude of these effects is influenced by species behavior. A two-camera study design benefits the detectability of unguligrade species, whereas one camera may be sufficient for the detection of all other non-ungulate species. Additionally, a second camera may only benefit less frequently detected species by improving precision of model estimates. We advise that occupancy studies first conduct a pilot study to gather preliminary information on species detection rates, to improve study efficiency and help ensure that reliable information is obtained to meet project objectives.

Table 3.1 Characteristics for 20 mammalian species with > 10 detections during a camera-trap survey conducted during October–December 2014 in Deramakot Forest Reserve, Sabah, Malaysian Borneo.

Foot posture Species common name Species scientific name Body mass (kg) Social behavior Dietary niche Digitigrade Asian elephant Elephus maximus 3000.0 Social Herbivore Sunda clouded leopard Neofelis diardi 20.0 Solitary Carnivore Sunda pangolin Manis javanica 7.0 Solitary Carnivore Malay civet Viverra tangalunga 5.0 Solitary Omnivore Mongoose Herpestes sp. 2.6 Solitary Carnivore Leopard cat Prionailurus bengalensis 2.1 Solitary Carnivore Plantigrade Bornean orangutan Pongo pygmaeus 67.5 Solitary Omnivore Sun bear Helarctos malayanus 55.5 Solitary Omnivore Malayan porcupine Hystrix brachyura 8.5 Solitary Herbivore

Pig-tailed macaque Macaca nemestrina 6.5 Social Omnivore Thick-spined porcupine Hystrix crassispinis 4.6 Solitary Herbivore Banded civet Hemigalus derbyanus 2.3 Solitary Omnivore Sunda stink-badger Mydaus javanensis 1.7 Solitary Omnivore Long-tailed porcupine Trichys fasciculata 1.7 Solitary Herbivore Moonrat Echinosorex gymmura 1.0 Solitary Carnivore Unguligrade Banteng Bos javanicus 500.0 Social Herbivore Sambar deer Rusa unicolor 115.0 Solitary Herbivore Bearded pig Sus barbatus 70.0 Social Omnivore Muntjac Muntiacus sp. 15.6 Solitary Herbivore Mousedeer Tragulus sp. 3.2 Solitary Herbivore

Table 3.2 Average number of records (± standard deviation [SD]) and sites (± SD) detected for 20 mammal species across 63 stations during a camera-trap survey conducted during October–December 2014 in Deramakot Forest Reserve, Sabah, Malaysian Borneo.

Species One-camera Two-cameras Common name Scientific name Records Sites Records Sites Asian elephant Elephus maximus 16.8 ± 4.7 6.8 ± 1.4 37 11 Sunda clouded leopard Neofelis diardi 17.7 ± 4.0 10.3 ± 2.0 36 19 Sunda pangolin Manis javanica 5.9 ± 1.5 5.3 ± 1.3 12 9 Malay civet Viverra tangalunga 139.2 ± 14.2 34.9 ± 2.2 277 46 Mongoose Herpestes sp. 10.4 ± 3.5 6.5 ± 1.8 21 13 Leopard cat Prionailurus bengalensis 15.2 ± 8.2 3.4 ± 1.2 26 6 Bornean orangutan Pongo pygmaeus 17.5 ± 3.6 14.5 ± 2.7 35 27 Sun bear Helarctos malayanus 58.5 ± 9.9 22.3 ± 2.2 116 33 Malayan porcupine Hystrix brachyura 49.2 ± 10.6 16.7 ± 2.1 101 25 Pig-tailed macaque Macaca nemestrina 110.0 ± 11.2 40.2 ± 2.9 223 54 Thick-spined porcupine Hystrix crassispinis 20.3 ± 4.4 10.9 ± 1.6 41 16 Banded civet Hemigalus derbyanus 17.1 ± 4.9 10.3 ± 2.2 35 20 Sunda stink-badger Mydaus javanensis 16.9 ± 4.4 8.0 ± 1.9 34 15 Long-tailed porcupine Trichys fasciculata 8.4 ± 1.5 6.4 ± 1.4 17 11 Moonrat Echinosorex gymmura 31.7 ± 7.1 11.5 ± 1.6 65 17 Banteng Bos javanicus 15.2 ± 4.0 4.9 ± 1.0 31 7 Sambar deer Rusa unicolor 215.7 ± 40.5 42.8 ± 2.5 434 55 Bearded pig Sus barbatus 247.1 ± 21.1 48.8 ± 1.8 496 55 Muntjac Muntiacus sp. 256.4 ± 28.7 52.1 ± 1.5 519 56 Mousedeer Tragulus sp. 691.0 ± 42.6 54.0 ± 1.1 1403 57

Number of records and sites for one-camera were derived from 150 simulated record databases where a single random camera was selected from each survey station. Two- cameras included all records from both cameras at each station.

Table 3.3 Null occupancy and detection probability estimates and corresponding coefficient of variation (CV) values for 20 mammalian species detected across 63 stations during a camera-trap survey conducted during October–December 2014 in Deramakot Forest Reserve, Sabah, Malaysian Borneo.

Species One-Camera Two-Cameras Common name Scientific name Occu Det Occu Det dDet dCV Asian elephant Elephus maximus 0.229 (43.0) 0.076 (44.8) 0.227 (30.5) 0.101 (28.8) 0.026 -16.0 Sunda clouded leopard Neofelis diardi 0.269 (39.3) 0.070 (40.4) 0.418 (22.7) 0.090 (23.5) 0.019 -16.9 Sunda pangolin Manis javanica 0.571 (60.3) 0.019 (77.1) 0.388 (61.8) 0.033 (66.9) 0.014 -10.2 Malay civet Viverra tangalunga 0.568 (11.9) 0.193 (10.2) 0.744 (7.7) 0.256 (7.0) 0.064 -3.3 Mongoose Herpestes sp. 0.274 (48.6) 0.066 (51.9) 0.331 (32.3) 0.069 (34.1) 0.003 -17.8 Leopard cat Prionailurus bengalensis 0.087 (57.9) 0.171 (41.2) 0.099 (39.0) 0.207 (22.1) 0.037 -19.1 Bornean orangutan Pongo pygmaeus 0.787 (37.7) 0.027 (46.8) 1.000 (0.4) 0.038 (17.1) 0.011 -29.7 Sun bear Helarctos malayanus 0.412 (18.0) 0.143 (16.2) 0.551 (12.1) 0.196 (10.2) 0.053 -6.0

39 Malayan porcupine Hystrix brachyura 0.309 (23.0) 0.128 (20.3) 0.411 (16.3) 0.175 (13.0) 0.047 -7.3

Pig-tailed macaque Macaca nemestrina 0.711 (10.5) 0.144 (11.9) 1.000 (NA) 0.183 (7.2) 0.040 -4.8 Thick-spined porcupine Hystrix crassispinis 0.281 (37.4) 0.073 (39.0) 0.294 (22.6) 0.134 (19.7) 0.061 -19.3 Banded civet Hemigalus derbyanus 0.394 (47.3) 0.052 (51.8) 0.512(25.4) 0.068 (28.5) 0.016 -23.4 Sunda stink-badger Mydaus javanensis 0.164 (39.3) 0.102 (36.3) 0.295 (24.4) 0.113 (22.8) 0.012 -13.5 Long-tailed porcupine Trichys fasciculata 0.283 (72.7) 0.038 (77.6) 0.412 (49.6) 0.039 (54.1) 0.001 -23.5 Moonrat Echinosorex gymmura 0.202 (28.9) 0.148 (22.8) 0.286 (20.9) 0.191 (14.6) 0.043 -8.2 Banteng Bos javanicus 0.091 (46.0) 0.141 (34.5) 0.117 (35.8) 0.201 (21.3) 0.060 -13.2 Sambar deer Rusa unicolor 0.734 (9.0) 0.178 (9.7) 1.000 (0.0) 0.231 (6.2) 0.053 -3.5 Bearded pig Sus barbatus 0.786 (7.0) 0.246 (7.1) 0.858 (5.1) 0.389 (4.6) 0.143 -2.5 Muntjac Muntiacus sp. 0.838 (6.0) 0.242 (6.9) 0.890 (4.5) 0.389 (4.5) 0.147 -2.4 Mousedeer Tragulus sp. 0.854 (5.2) 0.477 (3.9) 0.905 (4.1) 0.668 (2.5) 0.192 -1.3 Two-camera estimates were derived from a record database with all species detections during the survey, while one-camera estimates represent mean values derived from a subset of 50 simulated database including only records from a single, randomly selected camera at each survey station. The difference in detection probability (ΔDet) and CV (ΔCV) were calculated by

subtracting one-camera estimates from two-camera estimates. Species which were detected frequently (>50 detections from one camera) are in bold.

Figure 3.1 Location of Deramakot Forest Reserve, central Sabah, Malaysian Borneo (black square: inset).

Black dots represent 63 camera-stations surveyed during October–December 2014.

References

Ahumada JA, Silva CEF, Gajapersad K, Hallam C, Hurtado J, et al. Community structure and diversity of tropical forest mammals: data from a global camera trap network. Philos Trans R Soc Lond B Biol Sci; 2011. 366: 2703–2711.

Brodie JF, Giordano AJ, Zipkin EF, Bernard H, Mohd-Azlan J, Ambu L. Correlation and persistence of hunting and logging impacts on tropical rainforest mammals. Conserv Biol. 2015; 29: 110–121.

Cusack JJ, Dickman AJ, Rowcliffe JM, Carbone C, Macdonald DW, Coulson T. Random versus Game Trail-Based Camera Trap Placement Strategy for Monitoring Terrestrial Mammal Communities. PLoS ONE. 2015; 10: e0126373. https://doi.org/10.1371/journal.pone.0126373

Dunn RH, Rasmussen DT. Skeletal morphology of a new genus of Eocene insectivore (Mammalia, Erinaceomorpha) from Utah. J Mammal. 2009; 90: 321–331.

Forest Stewardship Council. FSC principles and criteria for forest stewardship. 1996; Available from: https://ic.fsc.org/en/certification/principles-and-criteria

Fiske IJ, Chandler RB. Unmarked: An R Package for Fitting Hierarchical Models of Wildlife Occurrence and Abundance. J Stat Softw. 2011; 43:1–23.

Friedman J, Hastie T, Tibshirani R. Regularization Paths for Generalized Linear Models via Coordinate Descent. J Stat Softw. 2010; 33: 1–22.

Gordon CH, Stewart EAM. The Use of Logging Roads by Clouded Leopards. Cat News. 2007; 47: 12–13.

Halfpenny JC, Biesiot, EA. A Field Guide to Mammal Tracking in North America. 2nd ed. Colorado; Johnson Printing Company; 1986.

Hearn AJ, Ross J, Macdonald DW, Bolongon G, Cheyne SM, Mohamed A, et al. Predicted distribution of the Sunda clouded leopard Neofelis diardi (Mammalia: Carnivora: Felidae) on Borneo. Raffles Bull Zool. 2016; Suppl 33: 149–156.

Hedges S, Meijaard E. Reconnaissance survey for banteng (Bos javanicus) and banteng survey methods traing project, Kayan-Mentarang National Park, East Kalimantan, Indonesia. Indonesia: World Wildlife Fund – Indonesia and Care for International Forestry Research; 1999.

Heydon MJ. The ecology and management of rain forest ungulates in Sabah, Malaysia: Implications of forest disturbance. Final Report. Arberdeen (U.K.). Institute of Tropical Biology, Department of Zoology, University of Aberdeen; 1994.

Hoffman M, Duckworth JW, Holmes K, Mallon D, Rodrigues ASL, Stuart SN. The difference conservation makes to extinction risk of the world's ungulates. Conserv Biol. 2015; 29: 1303–1313.

Hwang YT, Lariviere S. Mydaus javanensis. Mammalian Species. 2003; 723:1–3. doi: http://dx.doi.org/10.1644/723

Jennings AP, Zubaid A, Veron G. Home ranges, movements and activity of the short- tailed mongoose (Herpestes brachyurus) on Peninsular Malaysia. Mammalia. 2010; 74:43–50.

Jennings AP, Mathai J, Brodie J, Giordano AJ, Veron G. Predicted distributions and conservation status of two threatened Southeast Asian small carnivores: the banded civet and Hose’s civet. Mammalia. 2013; 77: 261–271.

Kanagaraj R, Wiegand T, Kramer-Schadt S, Anwar M, Goyal SP. Assessing habitat suitability for tiger in the fragmented Terai Arc landscape of India and Nepal. Ecography. 2011; 34: 970–981.

Kanagaraj R, Wiegand T, Mohamed A. Modelling species distributions to map the road towards carnivore conservation in the tropics. Raffles Bull Zool. 2013; Suppl 28: 85–107.

Kimura T. Correlation between the Morphology of the Feet and Muscle Fiber Composition in the Anterior Tibial Muscle. Anthropol Sci. 1996; 104: 1–14.

Kohavi R. A study of cross-validation and bootstrap for accuracy estimation and model selection. In: Proceedings of the 14th International Joint Conference on Artificial Intelligence Volume 2. San Francisco: Morgan Kaufmann Publishing; 1995. pp. 1137–1143.

Kitayama K. Co-benefits of Sustainable Forestry: Ecological Studies on a Certified Bornean Rain Forest. Tokyo: Springer Japan; 2013.

Lagan P, Mannan S, Matsubayashi H. Sustainable use of tropical forests by reduced- impact logging in Deramakot Forest Reserve, Sabah, Malaysia. Ecol Res. 2007; 22(3): 414–421.

Little AR, Demarais S, Gee KL, Webb SL, Riffell SK, Gaskamp JA, et al. Does Human Predation Risk Affect Harvest Susceptibility of White-Tailed Deer During Hunting Season? Wildl Soc Bull. 2014; 38: 797–805.

Little AR, Webb SL, Demarais S, Gee KL, Riffell SK, Gaskamp JA. Hunting intensity alters movement behaviour of white-tailed deer. Basic Appl Ecol. 2016; 17: 360– 369.

Lyra-Jorge MC, Ciocheti G, Pivello VR, Meirelles ST. Comparing methods for sampling large- and medium-sized mammals: camera traps and track plots. Eur J Wildl Res. 2008; 54:739–744.

Mathai J, Hon J, Juat N, Peter A, Gumal M. Small carnivores in a logging concession in the Upper Baram, Sarawak, Borneo. Small Carniv Conserv. 2010; 42:1–9.

Mackenzie DI, Royle JA. Designing occupancy studies: General advice and allocating survey effort. J Appl Ecol. 2005; 42:1105–1114.

MacKenzie DI, Nichols JD, Royle JA, Pollock KH, Bailey LL, Hines JE. Occupancy estimation and modeling: inferring patterns and dynamics of species occurrence. Burlington: Academic Press; 2006.

Medway L. Mammals of Borneo: Field keys and an annotated checklist. Kuala Lumpur, Malaysia: Monographs of the Malaysian Branch of the Royal Asiatic Society; 1983.

Mohamed A, Sollmann R, Bernard H, Ambu LN, Lagan P, Mannan S, et al. Density and habitat use of the leopard cat (Prionailurus bengalensis) in three commercial forest reserves in Sabah, Malaysian Borneo. J Mammal. 2013; 94: 82–89.

Mohamed A, Ross J, Hearn AJ, Cheyne SM, Alfred R, Bernard H, et al. Predicted distribution of the leopard cat Prionailurus bengalensis (Mammalia: Carnivora: Felidae) on Borneo. Raffles Bull Zool. 2016; Suppl 33: 180–185.

Mortelliti A, Boitani L. Evaluation of scent-station surveys to monitor the distribution of three European carnivore species (Martes foina, Meles meles, Vulpes vulpes) in a fragmented landscape. Mamm Biol. 2008; 73: 287–292.

Ngoprasert D, Steinmetz R, Reed DH, Savini T, Gale GA. Influence of fruit on habitat selection of Asian bears in a tropical forest. J Wildl Manage. 2011; 75: 588– 595.

Niedballa J, Sollmann R., Courtiol A, Wilting A. camtrapR: an R package for efficient camera trap data management. Methods Ecol Evol. 2016; 7: 1457–1462.

Normua F, Higashi S, Ambu L, Mohamed M. Notes on oil palm plantation use and seasonal spatial relationships of sun bears in Sabah, Malaysia. Ursus. 2004; 15(2): 227–231. doi: http://dx.doi.org/10.2192/15376176(2004)015<0227:NOOPPU>2.0.CO;2

Nowak RW. Walker’s Mammals of the World. Baltimore, Maryland: The Johns Hopkins University Press; 1999.

O’Connell AF, Nichols JD, Karanth KU, editors. Camera Traps in Animal Ecology: Methods and Analysis. Tokyo, Dordrecht etc.: Springer; 2011.

O’Connor KM, Nathan LR, Liberati MR, Tingley MW, Vokoun JC, Rittenhouse TAG. Camera trap arrays improve detection probability of wildlife: Investigating study design considerations using an empirical dataset. PLoS ONE. 2017; 12: e0175684. https://doi.org/10.1371/journal.pone.0175684

Oliver LR. Pigs, Peccaries, and Hippos. Status Survey and Conservation Action Plan. Gland, Switzerland: International Union for Conservation of Nature and Natural Resources; 1993.

Payne J, Francis CM. A field guide to the mammals of Borneo. Kota Kinabalu and Kuala Lumpur, Malaysia: The Sabah Society and WWF Malaysia; 1985.

Pease BS, Nielsen CK, Holzmueller EJ. Single-Camera Trap Survey Designs Miss Detections: Impacts on Estimates of Occupancy and Community Metrics. PLoS ONE. 2016; 11(11): e0166689. https://doi.org/10.1371/journal.pone.0166689

R Core Team. R: A Language and Environment for Statistical Computing. In: R Foundation for Statistical Computing, editor. Vienna, Austria; 2015.

Rabinowitz A. Notes on the behavior and movements of leopard cats, Felis bengalensis, in a dry tropical forest mosaic in Thailand. Biotropica. 1990; 22: 397–403.

Ross J, Hearn AJ, Macdonald DW, Alfred R, Cheyne SM, Mohamed A, et al. Predicted distribution of the Malay civet Viverra tangalunga (Mammalia: Carnivora: Viverridae) on Borneo. Raffles Bull Zool. 2016; Suppl 33: 78–83.

Ross J, Hearn A, Macdonald DW, Semiadi G, Alfred R, Mohamed A, et al. Predicted distribution of the banded civet Hemigalus derbyanus (Mammalia: Carnivora: Viverridae) on Borneo. Raffles Bull Zool. 2016; Suppl 33: 111–117.

Rovero F, Marshall AR. Camera trapping photographic rate as an index of density in forest ungulates. J Appl Ecol. 2009; 46: 1011–1017.

Rovero F, Martin E, Rosa M, Ahumada JA, Spitale D. Estimating Species Richness and Modelling Habitat Preferences of Tropical Forest Mammals from Camera Trap Data. PLoS ONE. 2014; 9: e103300. https://doi.org/10.1371/journal.pone.0103300

Rowcliffe MJ, Carbone C, Jansen PA, Kays R, Kranstauber B. Quantifying the sensitivity of camera traps: an adapted distance sampling approach. Method Ecol Evol. 2011; 2: 464–476.

Sabah Forest Enactment. State of Sabah Forest Enactment, 1968 (Sabah No. 2 of 1986). 1968; Available from: http://www.lawnet.sabah.gov.my/lawnet/sabahlaws/StateLaws/ForestEnactment1 968.pdf

Samejima H, Meijaard E, Duckworth JW, Yasuma S, Hearn AJ, Ross J, et al. Predicted distribution of the Sunda stink-badger Mydaus javanensis (Mammalia: Carnivora: Mephitidae) on Borneo. Raffles Bull Zool. 2016; Suppl 33: 61–70.

Servheen C, Herrero S, Peyton B. Bears. Status Survey and Consrevation Action Plan. Gland, Switzerland: International Union for Conservation of Nature and Natural Resources; 1998.

Shil SK, Quasem MA, Rahman ML, Kibria ASMG, Uddin M, Ahasan ASML. Macroanatomy of the Bones of Pelvis and Hind Limb of an Asian Elephant (Elaphas maximus). Int J Morphol. 2013; 31: 1473–1478.

Smith DA, Ralls K, Cypher BL, Maldonado JE. Assessment of scat-detection dog surveys to determine kit fox distribution. Wildl Soc Bull. 2005; 33: 897– 904.

Sollmann R, Mohamed A, Samejima H, Wilting A. Risky business or simple solution – Relative abundance indices from camera-trapping. Biol Conserv. 2013; 159: 405– 412.

Sunarto, Sollmann R, Mohamed A., Kelly MJ. Camrea trapping for the study and conservation of tropical carnivores. Raffles Bull Zool. 2013; Suppl 28: 21–42.

Taylor ME. Locomotion in some East African viverrids. J Mammal. 1970; 51: 42–51.

Taylor ME. Foot structure and phylogeny in the Viverridae (Carnivora). J Zool Lond. 1988; 216: 131–139.

Tibshirani R. Regression shrinkage and selection via the lasso. J Royal Statist Soc B. 1996; 58: 267–288.

Tobler MW, Carillo-Percastegui SE, Leite Pitman R, Mares R, Powell G. An evaluation of camera traps for inventorying large- and medium-sized terrestrial rainforest mammals. Anim Conserv. 2008; 11: 169–178.

Treves A, Mwima P, Plumptre AJ, Isoke S. Camera-trapping forest–woodland wildlife of western Uganda reveals how gregariousness biases estimates of relative abundance and distribution. Biol Cons. 2010; 143: 521–528.

Tyre AJ, Tenhumberg B, Field SA, Niejalke D, Paris K, Possingham HP. Improving precision and reducing bias in biological surveys: estimating false-negative error rates. Ecol Appl. 2003; 13: 1790–1801.

Van Weers DJ. Notes on Southeast Asian porcupines (Hystricidae, Rodentia) IV. On the taxonomy of the subgenus Acanthion F. Cuvier, 1823 with notes on the other taxa of the family. Beaufortia. 1979; 29: 215–272.