The flat-headed cat (Prionailurus planiceps) is one of the world's least known, highly threatened felids with a distribution restricted to tropical lowland rainforests in Peninsular Thailand/Malaysia, Borneo and Sumatra. Throughout its geographic range large-scale anthropogenic transformation processes, including the pollution of fresh-water river systems and landscape fragmentation, raise concerns regarding its conservation status. Despite an increasing number of camera-trapping field surveys for carnivores in South-East Asia during the past two decades, few of these studies recorded the flat-headed cat.
In this study, we designed a predictive species distribution model using the Maximum Entropy (MaxEnt) algorithm to reassess the potential current distribution and conservation status of the flat-headed cat. Eighty-eight independent species occurrence records were gathered from field surveys, literature records, and museum collections. These current and historical records were analysed in relation to bioclimatic variables (WorldClim), altitude (SRTM) and minimum distance to larger water resources (Digital Chart of the World). Distance to water was identified as the key predictor for the occurrence of flat-headed cats (>50% explanation). In addition, we used different land cover maps (GLC2000, GlobCover and SarVision LLC for Borneo), information on protected areas and regional human population density data to extract suitable habitats from the potential distribution predicted by the MaxEnt model. Between 54% and 68% of suitable habitat has already been converted to unsuitable land cover types (e.g. croplands, plantations), and only between 10% and 20% of suitable land cover is categorised as fully protected according to the IUCN criteria. The remaining habitats are highly fragmented and only a few larger forest patches remain.
Based on our findings, we recommend that future conservation efforts for the flat-headed cat should focus on the identified remaining key localities and be implemented through a continuous dialogue between local stakeholders, conservationists and scientists to ensure its long-term survival. The flat-headed cat can serve as a flagship species for the protection of several other endangered species associated with the threatened tropical lowland forests and surface fresh-water sources in this region.
Citation: Wilting A, Cord A, Hearn AJ, Hesse D, Mohamed A, Traeholdt C, et al. (2010) Modelling the Species Distribution of Flat-Headed Cats (Prionailurus planiceps), an Endangered South-East Asian Small Felid. PLoS ONE 5(3): e9612. https://doi.org/10.1371/journal.pone.0009612
Editor: Michael Somers, University of Pretoria, South Africa
Received: September 24, 2009; Accepted: February 4, 2010; Published: March 17, 2010
Copyright: © 2010 Wilting et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: The different field work projects of the authors were supported by the Clouded Leopard Project and Point Defiance Zoo and Aquarium, the Panthera Foundation, WWF-Germany, WWF-Malaysia, Minnesota Zoo, Cleveland MetroParks, Houston Zoo, Association for Wild Cat Conservation, the Society for the Conservation of Species and Populations, Deutsche Forschungsgemeinschaft (grant Fi-698/5-1), the Darwin Initiative, Felidae Conservation Fund, International Trust for Nature Conservation, Wild About Cats, HDH Wills 1965 Charitable Trust and the Cat Action Treasury. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
The diminutive flat-headed cat Prionailurus planiceps (Vigors and Horsfield, 1827), with adults weighing as little as 1.59 kg  (Figure 1), occurs in southern Thailand, Peninsular Malaysia, Sumatra and Borneo - (Figure 2). Some authors also suggest that its range extends into the southern part of Myanmar . This species has never been studied intensively in the wild , –. In 2008, a revision of the IUCN Red List suggested an increasing risk of extinction for the flat-headed cat and its status was changed from “Vulnerable” to “Endangered” .
Most information on the distribution and natural history of the flat-headed cat comes from opportunistic incidental sightings, specimen collectors , –, and a few animals kept in captivity –. Morphological adaptations such as its dental structure – and the slight web between the toes as well as incidental observations suggest that this cat is well adapted to hunt small prey in shallow water and on muddy shores . For instance, Leyhausen reported that a captive flat-headed cat displayed more interest in a mouse in a pond than on land , and the stomach of an adult shot at a riverbank contained only fish . Individuals have been observed or collected in undisturbed primary and secondary forests mostly along rivers, streams and in flooded areas , , . Only Khan [in 5] has reported sightings in oil palm plantations and even speculated that this species benefits from the expansion of these plantations in Peninsular Malaysia . However, as we are aware of no other similar records from anthropogenically transformed habitats, it is unlikely that flat-headed cats survive and reproduce in palm oil or rubber plantations. Nothing is known regarding the home-range sizes and population densities of flat-headed cats and their presumably nocturnal behaviour (considering the records obtained in this study) makes surveying and monitoring difficult. Although camera-trapping efforts have increased in magnitude and extent during the past few years , only a few studies have recorded this species, with the number of photographs being so low that abundance estimates could not be calculated.
Several developments in ecological niche modelling (ENM) have provided new tools to estimate species ranges and identify suitable habitats –. For many threatened species, only a few historical (museum) and recent records are available which may result in a low accuracy and a poor fit of most ecological niche models , . However, the maximum entropy (MaxEnt) framework  appears to be robust even if only few occurrence records are available –. Although ENM has been used to predict the distribution of little-known carnivores , , , we are not aware of any application of this framework to assess the potential distribution of any Asian felid species.
The aim of this study is to describe the historical and current distribution of the flat-headed cat, predict its potential habitat occupancy by applying a MaxEnt model throughout its range, identify key localities suitable to contribute to the long-term survival of the species, and ultimately contribute to an update of the conservation status of this endangered species.
Materials and Methods
Species Occurrence Samples
In January 2009, during the Clouded Leopard and Small Cat Summit organised by the IUCN Cat Specialist Group and the IUCN Conservation Breeding Specialist Group, only 14 records of the flat-headed cat were compiled by the attending scientists and conservationists . We collected additional records from the existing scientific literature and researchers who were unable to join the summit and conducted a survey in several natural history museums. We only included reports of direct sightings, camera-trapping pictures or dead specimens. We considered all reported locations to be accurate for species identification since the flat-headed cat has a very distinctive appearance and is not easily mistaken for any other felid or other small carnivore. Owing to their unknown origin, wild caught zoo animals kept in Thailand and Malaysia were not included in the analysis. We also included the results of a questionnaire conducted by MAJ in Sarawak (northwest Borneo) as the interviewees were familiar with the flat-headed cat.
A total of 107 records (47 historical and 60 recent) was obtained (Table S1). We checked for and excluded all multiple occurrences of sites (particularly within museum records), resulting in a final number of 88 independent records (29 historical and 59 recent) with a minimum distance of one km between the records. Records were defined as recent if they were collected within the last 25 years (since 1984); the majority was obtained during the last 10 years, although the precise year was not available in all cases. The oldest historical record from the Museum für Naturkunde in Berlin dated back to 1864. For 21 recent records precise geographic coordinates were available, for the remainder a general description of the locality. For these records we assigned coordinates using published resources such as Google Earth 5.0.1 or the database of BirdLife International (www.birdlife.org). We assume that the positional accuracy of the geographic coordinates assigned to the records was between one and ten kilometres.
A series of bioclimatic variables including eleven temperature and eight precipitation metrics was obtained from the WorldClim database (Version 1.4, http://www.worldclim.org/bioclim.htm) . WorldClim data are derived from monthly temperature and precipitation values using long-term time series from 1950 to 2000 from a global network of 4,000 climate stations with a spatial resolution of one km2. The WorldClim parameters express spatial variation in annual means, seasonality and extreme or limiting climatic factors and represent biologically meaningful variables for characterising species distributions . Topographic data acquired during the SRTM mission and re-sampled from a 90×90 m2 to a 1×1 km2 spatial resolution were downloaded from the WorldClim data base. Surface water bodies (rivers and lakes) were extracted from the country inland water data of the Digital Chart of the World (DCW; http://biogeo.berkeley.edu/bgm/gdata.php) and the minimum distance to water in km was calculated per pixel using ArcGIS 9.3 software and used as an input variable for the model. These abiotic variables were then analysed together with the species occurrence data using the Maximum Entropy algorithm as implemented in MaxEnt software Version 3.2.1. Ten individual MaxEnt models were run in batch mode with the following settings: Auto features (feature types are automatically selected depending on the training sample size), perform jackknife tests, logistic output format, random test percentage = 25, regularisation multiplier = 1, maximum iterations = 1000, convergence threshold = 0.0001 and maximum number of background points = 10,000. We used the mean probabilities predicted by the ten independent models as estimates for subsequent analyses. The MaxEnt model prediction was regarded as the potential former distribution, since historical and recent occurrences were treated with equal weight in the models.
As the modelling algorithm is a major source of uncertainty in the prediction of species distribution – we assessed the robustness of the results of the MaxEnt algorithm by comparing them with the results of two other algorithms, Environmental Distance and Support Vector Machines (SVM) as implemented in the openModeller software (http://openmodeller.sourceforge.net). We computed the area under the receiver operating characteristic curve (AUC) to assess model discriminatory power for the different algorithms. We further compared the predicted probabilities of occurrence for 5,000 points randomly distributed throughout the land surface within the study region between the MaxEnt, SVM and Environmental Distance models using Pearson's correlation coefficient r .
To extract suitable habitats from the predicted continuous surfaces of mean probabilities of occurrence we created five different scenarios using 0%, 10%, 20%, 30% and 40% omission thresholds (including both training and test samples). Subsequent analyses were done for all scenarios, but the maps shown were based on a conservative approach (10% omission threshold).
Land Cover within Predicted Distribution
As we included both historical and recent records in the analysis, the inclusion of land cover information in the model was not feasible owing to the time elapsed between historical records and the collection of land cover data, and the extent of habitat modification during this period. We therefore used land cover information to refine the predicted former distribution of the flat-headed cat. Three land cover maps were chosen for this purpose: the Global Land Cover 2000 (GLC 2000), the GlobCover Land Cover (GlobCover) Version 2.1 and–available only for Borneo - a classification developed by SarVision LLC (Table 1).
Based on our field records and the limited existing knowledge about appropriate habitats for the flat-headed cat, we reclassified all land cover maps into five categories (Table 2): 0 = unsuitable habitat (croplands, bare or burnt areas, artificial areas, upper montane forest), 1 = very poor habitat (mosaic cropland/vegetation, lower montane forest, closed to open evergreen shrubland), 2 = poor habitat, maybe suitable as a corridor (mosaic vegetation, upland forests), 3 = good habitat (lowland forest) and 4 = very good habitat (regularly or permanently flooded forest, peat swamp forest, mangroves).
We then analysed the current land cover status at both historically and recently recorded localities. For this purpose, we created a buffer zone around each locality representative for the ecological requirements of the flat-headed cat. Because actual home range sizes of flat-headed cats are unknown, we used data on home-range sizes of the leopard cat Prionailurus bengalensis as an approximation, a closely related, similar sized, sympatric felid. Known home-range sizes for female leopard cats vary between 1.75 km2 on Iriomote Island  and 2.1 km2 on Borneo , and for males between 5.8 km2  and 7.5 km2  in Thailand. We took the rounded mean value of these extremes of 4 km2 as an estimate of the home-range size of the flat-headed cat. To accommodate uncertainties associated with the precise position of point localities we used a buffer area of three times the extrapolated home-range (12 km2). This analysis was conducted using the Spatial Analyst extension of ArcGIS 9.3 for all three land cover classifications. If reclassification of land cover habitats was appropriate, then the reclassified habitat types in the buffer areas surrounding recent records should be mainly in categories 3 and 4. We also assessed the habitats in the buffer areas around historical records to detect whether they have already been transformed into unsuitable environmental conditions.
Human Population Density
To refine initial model predictions, human population density data supplied by the LandScan 2007™ High Resolution Global Population Data Set (UT-Battelle, LLC) compiled on a 30″×30″ latitude/longitude grid were used. We reclassified human population densities into 5 categories: class 0 = more than 25 inhabitants km−2 (unsuitable for the occurrence of flat-headed cats), class 1 = 10–25 inhabitants km−2, class 2 = 5–10 inhabitants km−2, class 3 = 1–5 inhabitants km−2, class 4 = 0 inhabitants km−2 (no human disturbance, presumably optimal for flat-headed cats). As with the land cover data, we extracted human population densities in buffer areas of 12 km2 around historical and recent records.
Estimating the Loss of Suitable Habitats
The major threat for flat-headed cats is presumably the transformation of their habitats to arable land or plantations [e. g. 3]. We therefore assessed the loss of suitable habitats owing to land cover conversion for the five model scenarios. The proportions of the reclassified land cover classes were extracted and classes 3 and 4 defined as appropriate habitats for the flat-headed cat. The relative proportion of these two classes in relation to the area with non-suitable cover types (classes 0–2) was taken as a measure of the loss of suitable habitat caused by anthropogenic factors.
Key Localities and Protected Areas
To identify habitats with a high probability of current occurrence of flat-headed cats we calculated a habitat suitability index (HSI) . For this purpose, we multiplied the averaged prediction of the ten MaxEnt model runs with the categorised land cover classes and with the reclassified human population density classes. As we assume that the land cover and the predictions by the MaxEnt model are more important than the human population density for the occurrence of flat-headed cats we doubled the weight of these variables. To calculate the HSI of flat-headed cats in each pixel (resolution 1 km2) we modified the equation of Allen et al.  to read HSI = (M2×L2×H)1/5, where M is the mean probability predicted by the ten MaxEnt models, L is the land cover class and H is the human population density category. In order to scale the HSI values between 0 and 1, we rescaled the land cover classes and human population density classes as follows: class 4 was rescaled to 1, class 3 to 0.75, class 2 to 0.5, class 1 to 0.25 and class 0 kept its value. Thus, if any component is unsuitable ( = 0) the HSI will be 0 as well. The application of this equation ensures that even an area with a relatively high human population density (class 1) would retain a high probability to harbour flat-headed cats if both the mean probability predicted by the MaxEnt model and its land cover class are favourable (class 3 or 4). Based on the HSI maps we could identify large contiguous forest blocks with potentially suitable habitat for the flat-headed cat.
The 2009 World Database of Protected Areas (www.wdpa.org) was used to determine the level of legal protection throughout the study area and the proportion of potential habitat within totally protected areas. For this aim, potential habitats were defined as areas covered with vegetation and human population classes 3 or 4 within the predicted potential former distribution for all five threshold scenarios. For comparison we analysed the proportion of all forested areas classified as totally protected under the IUCN criteria in the study region.
Where appropriate, results are presented as means ± standard deviation. We used chi-square distributed log-likelihood G tests as implemented in Systat (version 12, Systat Inc., San Jose, CA, USA) to test for differences between buffer areas of historical and recent records in the distribution of forest classes and human population densities. If the values for land cover or human population density categories for each pixel within a buffer zone are spatially autocorrelated, then neighbouring pixel values would carry less information than truly independent values and the calculated p-value for the log-likelihood G will underestimate the true type I error probability . The appropriate strategy is to estimate the “effective sample size” by accounting for possible spatial autocorrelation and adjusting the log-likelihood G value accordingly . Because we were not aware of a published approach to correct for spatial autocorrelation in 2 x k contingency tables, we proceeded as follows: We followed Cerioli  who, for 2×2 contingency tables, divided the log-likelihood G by a correction factor (1+λ) where λ is a measure of the degree of spatial autocorrelation for the relevant lattice (e.g., an appropriate summation of Moran's spatial autocorrelation coefficient computed over an entire contiguous lattice). The buffer zones for our records are disjointed rather than contiguous, and therefore, spatial autocorrelation between buffer zones is unlikely and Cerioli's λ, computed over the entire distribution map, would be unduly conservative. We therefore assumed that there was no spatial autocorrelation between buffer zones. Instead, to protect type I error probabilities we assumed that significant spatial autocorrelation did occur across the entire buffer zone around each flat-headed cat record and calculated an indicator λ for it as (i) the maximum order number of neighbouring pixels in relation to the central pixel of each buffer zone = ((√number of pixels per buffer zone)-1)/2; (ii) the maximum number of distance classes between pixels of a buffer zone = (√number of pixels per buffer zone)-1. The log-likelihood G would then, in analogy to , be adjusted by dividing the original G by the correction factor = (1+λ) and calculating the type I error probability for the adjusted G using the chi-square distribution and the same degrees of freedom as for the original G. These adjustments reduced the log-likelihood G but did not substantially affect the p-values (Table 3). With the most extreme approach, i.e., treating all buffer zone pixels as if they were identical and using the number of pixels per buffer zone as correction factor, the p–values for the comparison of historical and recent records were still significant for the GLC2000 (p = 0.044) and SarVision LLC maps (p = 0.00027) in terms of habitat suitability and for human population density (p = 0.0023), only the GlobCover map produces then a p of 0.16.
Flat-Headed Cat Records
Our literature survey and contact with scientists currently working in the field yielded several new records of the flat-headed cat from Thailand, Malaysia, Brunei Darussalam and Indonesia. Table S1 contains a complete list of all recent and historical records we had access to. Most recent records (43) came from Borneo: three were recorded in Brunei Darussalam, 12 in Kalimantan, Indonesian Borneo and 28 in Malaysian Borneo (14 in Sabah and 14 in Sarawak). We also obtained information on the origin of 27 museum specimens (17 geographically independent) collected in Borneo in historical times. From Sumatra we obtained 13 independent recent and only four independent historical records, since most of the museum specimens did not have a precise collection location. From the southern part of Thailand and from Peninsular Malaysia we received recent records from four forest patches (Toh Daeng Peat Swamp forest in Thailand, Selangor Peat Swamp Forest, Krau Game Reserve and Pahang Peat Swamp Forest in Malaysia) and 14 historical records (8 were geographically independent). None of the scientists working in the southern part of Myanmar could confirm the presence of the flat-headed cat there (Su Su in litt., Anthony Lynam in litt., Nay Myo Shwe in litt.). Figure 3 depicts localities of recent and historical records.
Circles indicate the location of historical, triangles the location of recent records.
For 88 out of a total of 107 records, geographic coordinates had to be estimated from descriptions of localities. Photographs from camera-traps or direct sightings of flat-headed cats with precise location information were available for 19 locations (15 on Borneo and four on Sumatra). During one of these sightings along a small tributary of the Kinabatangan River in Sabah (Malaysia) two authors (AW & AM) filmed a wild flat-headed cat for several minutes (Video S1).
MaxEnt Models and Their Robustness
For all ten models, the fit as measured by the mean area under the curve (AUC) of the receiver operating characteristic (ROC) was high, with mean values of 0.982 (0.980 to 0.983) for the training data and 0.968 (0.966 to 0.977) for the test data (Table 4). Similarly, the model's discriminatory power (measured by AUC) of the two other modelling algorithms yielded values of 0.89 for Support Vector Machines (SVM) and 0.99 for Environmental distance. The comparison of the models demonstrated strong similarity of the predicted distributional areas between MaxEnt and the other modelling algorithms (correlation between MaxEnt and Environmental Distance r = 0.752, N = 5,000, p<0.01; between MaxEnt and SVM r = 0.831, N = 5,000, p<0.01). Based on this spatial agreements between the different model algorithms we consider the results and conclusions from the MaxEnt model to be robust. Therefore, we present and discuss only the results of the MaxEnt model.
The most important abiotic variable was minimum distance to water, explaining 53.6±2.9% of variation in the distribution of the flat-headed cat. The omission of minimum distance to water as the main explanatory variable significantly decreased the gain of the model, indicating its important information content not present in the other 20 input variables. The second most important variable was precipitation of the driest month, explaining 13.6±8.7% of the variance. The third most important variable was altitude explaining 8.0±2.2% of the model variance. Figure 3 compares localities of historical and recent records with the distribution of the flat-headed cat as predicted by the mean MaxEnt model. Areas with high probabilities of predicted occurrence were mainly confined to lowland areas.
Environmental Conditions at Sites of Records
Our results are consistent with the assumption that the occurrence of flat-headed cats primarily depends on the availability of freshwater habitats. Over 70% of records were collected less than three kilometres away and only four records (<5%) were located more than eight kilometres away from the nearest major river or lake (Figure 4). As improved access to remote areas today may have extended sampling to areas historically little sampled, we tested for possible sampling biases in both historical and recent records. There was no difference between the distributions of distances to freshwater habitats of historical and recent records (Mann-Whitney U-test, U = 744, N = 88, p = 0.31). In contrast, recent records were significantly more likely to be located at higher altitudes than historical records (U = 402, N = 88, p<0.001), although almost 80% of recent and more than 90% of historical records are located below 100 m asl (Figure 5).
Land Cover within the Predicted Distribution and Loss of Suitable Habitat
Figure 6a shows the reclassified GlobCover land cover map for areas predicted to be suitable habitats for the flat-headed cat when applying a 10% omission threshold. Figure 6b illustrates the same results projected onto the SarVision data set of Borneo, the most precise vegetation classification and the only one which distinguishes between montane, upland and lowland forest. Many of the predicted potentially suitable historical habitats were assigned to the unsuitable or poor classes 0 to 2 in all three land cover maps. The relative proportion of already degraded habitat (e.g. arable land, plantations, mosaic vegetation) was similar for all omission threshold scenarios (0%, 10%, 20%, 30% and 40% thresholds) between the three land cover maps (Table 5). The total estimate of loss of suitable habitat varied between 54% (GlobCover 0% omission threshold) and 68% (SarVision 0% omission threshold) (Figure 7). As the SarVision land cover classification only covered Borneo, comparisons of relative losses between different land cover maps may be of limited use.
Results are shown for a strict scenario allowing only 10% omission of all samples.
The landscape around historical records is significantly more degraded than around recent records (Table 3). In the buffer zones around historical records most suitable habitat has already been modified into unsuitable or poor land cover (classes 0–2) (54% GlobCover, 72% GLC 2000 and 80% SarVision). Between 31% (GlobCover) and 42% (GLC 2000 and SarVision) of buffer areas around recent records were categorised in land cover classes 0–2 (Figure 8).
Human Population Density
Human population density was significantly higher in buffer areas around historical than around recent records (Table 3): 66% of buffer areas around historical records were allocated to the high human population density classes 0–2, whereas only 30% of buffer areas of recent records belonged to these classes.
Protected Areas and Prediction of Key Localities
Figure 9 shows the probability of occurrence of flat-headed cats based on the habitat suitability index HSI together with the location of protected areas. The proportion of suitable habitat under total protection in the three land cover maps and five omission scenarios indicated that currently between 10% and 20% of suitable habitat is under total protection as defined by the IUCN . The proportion under total protection decreased with the higher omission thresholds in all three land cover maps (Figure 10). The proportion of protection of all forested areas was 22% for GlobCover, 23% for SarVision LLC and 24% for GLC 2000 classification. Based on Figure 9 we identified 19 forest areas throughout the entire range of the flat-headed cat which are likely to make significant contributions to the long-term survival of the flat-headed cat (Table 6).
For comparison, currently protected areas and predicted key localities 1–19 (Table 6) identified in this study are illustrated.
Location of Historical and Recent Records
Our database of all recent and historical records of the flat-headed cat (Table S1) is provided to assist further conservation plans and will serve as a useful starting point for the collation of future sightings of this endangered species. Identifying the precise location for most records was not easy but as long as errors in the assignment of coordinates are random with respect to the habitat composition of the calculated buffer areas, this will increase the overall error but not bias the results and is unlikely to affect the principal conclusions.
The number of sightings and camera-trapping pictures with precise location information was low, even though the number of camera-trapping studies in South-East Asia has increased during the past 20 years and we had first-hand information from many of these studies. Information from 17 camera-trapping photographs was available for use in this study. This is the lowest number of photographs from camera-traps of any South-East Asian felid and is a testimony to the extremely limited knowledge regarding this species and possibly the limits of its range and distribution. For instance, in Sumatra the presence of tigers (Panthera tigris), Asian golden cat (Pardofelis temminckii), marbled cat (Pardofelis marmorata), Sunda clouded leopard (Neofelis diardi) and leopard cat is regularly confirmed by camera-trap photographs. On Borneo, an increasing number of camera-trapping studies documented the regular occurrence of the leopard cat, Sunda clouded leopard, marbled cat and the Borneo endemic bay cat (Pardofelis badia). Even though records of the latter species were very rare , its geographic range and distribution might now be better understood than that of the flat-headed cat , .
As far as we are aware there are no camera-trap records of the flat-headed cat from Thailand and Peninsular Malaysia. However, most camera-trapping initiatives are designed to record large cat species, typically tigers in Peninsular Thailand/Malaysia and on Sumatra. For this purpose, cameras are placed along roads or ridges, habitats unlikely to be used by flat-headed cats and therefore inappropriately to record this species. Almost none of the camera-trapping studies placed their cameras along the edges of lakes, ponds or rivers. This could easily lead to a sampling bias and thus may affect capture probabilities of flat-headed cats.
The GIS analysis is a first step towards identifying habitat preferences of the flat-headed cat. Our results indicate that the most relevant factor explaining its distribution is the distance to freshwater sources such as major rivers and large lakes. Although with the currently available DCW map it was not possible to assess distances to the nearest minor freshwater source, our results suggest that larger watercourses and water bodies are needed and that small ones alone are unlikely to be sufficient, maybe because they face a higher risk to dry up.
This assumption is supported by the most important bioclimatic variable, precipitation during the driest month, which suggests that rainfall during the dry season is a critical factor for the presence of flat-headed cats, presumably because the rainfall supports the persistence of small-sized water bodies. This might also explain why the range of the flat-headed cat does not expand further north, as the climate is more seasonal there . The model also predicted that South Kalimantan (Figure 3) provides no suitable habitat for the flat-headed cat. This may also be explained by climatic conditions, as this region has typically a more severe dry season .
Most flat-headed cat records were located in extreme lowland areas below 100 m asl. The difference in altitude between historical and recent records might be a consequence of sampling bias in historical times. For logistical reasons, collectors of museum specimens many decades ago spent disproportionately more time in lowland areas along coastlines or rivers. Even though roads or helicopters provide access to more remote areas at higher altitudes today, most recent records are still from the lowlands at less than 100 m asl.
Range and Distribution
For species with limited information about their historical and current range, predictive models are a useful tool to assist conservation planning , , –. The flat-headed cat is one of the least known felids worldwide and its special habitat requirements make it an ideal candidate for the application of predictive modelling. The predicted potential former distribution (Figure 3) and the map of current habitat suitability (Figure 9) differ substantially from the recently published distribution map in the most recent status assessment for the IUCN Red List (Figure 2) . Our maps suggest that coastal and lowland areas are key habitats for the flat-headed cat, but these are mostly missing from the IUCN distribution map. Although the Red List was reassessed in 2008 and the increasing threat to the flat-headed cat was recognised by the species assessors (all authors of this paper: AH, JS, JR, AW & SS), the distribution map of the 1996 Cat Action Plan map  was only slightly modified by the IUCN with the help of the GLC 2000 land cover classification. This must have led to an overemphasis on large forest tracts at higher altitudes and a neglect of smaller, already more fragmented and anthropogenically modified coastal forests.
Loss of Suitable Habitat and Human Population Growth
This study provides a first indication of how much of flat-headed cat habitat has already been converted into unsuitable habitat. South-East Asia has one of the world's highest deforestation rates with an annual forest reduction of 1.3%  and by 2100 three quarters of its original forests are expected to be lost ; on Borneo alone 1.3 million ha of lowland forest are lost annually . The high levels of habitat reduction (between 54% and 68%) as estimated by our models indicate that the actual loss of habitat is immense. This may even be a conservative estimate of loss, since our model accepted remaining forest patches of any size as suitable habitat even though it is likely that small and fragmented forests will not support viable populations of flat-headed cats.
A second estimate of loss of suitable habitat was obtained from the land cover in buffer areas around historical locations. Between 54% (GlobCover) and 80% (SarVision) of previously suitable habitat has already been transformed to unsuitable or poor habitats in the land cover categories 0 to 2 (Figure 8). That between 31% and 42% of the habitat in buffer areas around recent records belongs to categories of unsuitable habitat reflects that some of our “recent” records were collected 25 years ago. Given current deforestation rates much suitable habitat has vanished in the interim. Two records from Sumatra provide good examples for this land cover and land use change. One record from the Riau Province involved a flat-headed cat trapped by hunters in a small forest patch between oil palm plantations, and a second record from Tapan came from an area which today is already transformed into oil palm plantations.
Similarly, the data on human population density showed that 66% of historical records are at present surrounded by high human population densities (Figure 8), resulting from the rapid population growth rates in South-East Asia .
Other Threats to the Flat-Headed Cat
Our analysis considered habitat loss and human population density, but not proximate threats to flat-headed cats such as hunting, over-fishing or fresh-water pollution. Presumably environmental pollution, due to gold mining and agricultural practices is largely reducing distribution and long-term survival prospects of this species. Malaysia and Indonesia are already the largest producers of palm oil – and the globally increasing demand for biofuel and other products derived from palm oil further threatens the remaining forested areas . Expansion of plantations often involves wetland drainage and loss of upstream forests. This results in reduced water runoff during the drier months and may therefore, given the relationship of flat-headed cat distribution to dry season rainfall levels, contribute to the loss of habitat. As we could not incorporate these threats into our estimates, the observed loss of suitable habitat is likely to be conservative. These threats require urgent and careful further elucidation, since the long-term consequences of large-scale agrochemical use, especially in oil palm plantations, are uncertain.
Our results show that currently only between 10% and 20% (where 20% refers to the 0% omission scenario) of potentially suitable habitat is classified as completely protected by IUCN criteria. For the other scenarios the area under full protection was calculated in the range between 10% and 16%. This is a disproportionally low percentage since–throughout the study region - between 22% and 24% (depending on the land cover classification) of the forested area is fully protected. The discrepancy is explained by large national parks, such as Taman Negara, Kerinci, Gunung Leuser or Sungai Kayan and Sungai Mentarang, which are located in more mountainous regions and do not contain large areas of suitable habitats for the flat-headed cat. Owing to the low proportion of totally protected flat-headed cat habitats it can be assumed that currently suitable areas will be converted into plantations in the near future. Higher omission rates in the model reduce the fraction of potential habitats giving a focus on core distribution areas. If protected areas were located within core zones of the distribution of the flat-headed cat, the proportion of protected areas should increase with higher omission rates. Our data show the opposite effect (Figure 10), indicating that protected areas are mainly located within marginal habitats and areas with a lower probability of flat-headed cat occurrence. This highly alarming result is not surprising since well-watered lowlands are among the areas most desired for expansion of agriculture and industry, leading to a general disinclination to declare large protected areas within them. The World Database of Protected Areas only includes those that meet IUCN criteria. Although some of the protected areas not included in this database do contain the flat-headed cat (e.g., the commercially used forests in and around Deramakot Forest Reserve in Sabah), without any doubt the proportion of the flat-headed cat distribution range under any level of official protection is very low.
Key Localities in Peninsular Thailand and Malaysia
Peninsular Thailand/Malaysia has the worst perspective for the long-term survival of the flat-headed cat. No single continuous forest block comprises potential flat-headed cat habitat: all the larger forest reserves and national parks in Peninsular Thailand/Malaysia are located in the interior of the country, away from the species' predicted distribution range. In southern Thailand, Toh Daeng Peat Swamp forest and a few very small forest patches might be the only areas where flat-headed cats still occur in Thailand. In Peninsular Malaysia, we identified only two suitable forest areas, the Selangor peat swamp forest (west coast) and the Pahang Peat Swamp Forest (east coast). Recent sightings from Pahang Peat Swamp Forest raise hope that these forests still support a population of flat-headed cats. However, even this forest patch is already threatened and degraded into several smaller forest blocks.
Key Localities in Sumatra
Sumatra holds five key areas for the flat-headed cat. Two of these, Way Kambas National Park and Hutan Lunang Nature Reserve, are already small and isolated. Camera-trapping pictures from Way Kambas suggest that this wildlife reserve might have the potential to protect the flat-headed cat at the most southern part of its distribution range . In northern Sumatra, south-western Gunung Leuser National Park, close to the coastline, with the adjacent Singkil Barat Nature Reserve comprises an area of great potential for the flat-headed cat. A recent camera-trapping picture from Suak (G. M. Fredriksson in litt.) supports this conclusion from the model. Recent sightings along Merang river in South Sumatra and from Berbak National Park ,  endorse our prediction that this area is important for the flat-headed cat. The largest suitable Sumatran forest for the flat-headed cat is in Riau Province. This large forest block ranges from Senepis Buluhala in the north, where the Sumatran Tiger Conservation Project obtained a camera-trap photograph, to the forests of Kerumutan Wildlife Reserve in the south. It is already fragmented and Riau has the highest deforestation rate in Indonesia, having lost 65% of its forest in the last 25 years . Within this area there is a strong move by several NGOs to protect the remaining forest specifically for Sumatran tigers, but these efforts also need to include the lowland swamp forest in order to protect species depending on wetlands such as the flat-headed cat.
Key Localities in Borneo
Borneo covers the largest part of the range of the flat-headed cat and holds most of the potentially suitable habitats. Based on our model predictions using topo-climatic conditions, the Malaysian state of Sarawak historically had the largest suitable flat-headed cat areas. However, the lowland coastal areas have borne the brunt of recent land development and agricultural expansion, so that today most natural forests are found in the interior, the least favourable habitats for the flat-headed cat as predicted by our model. We identified two areas in Sarawak which most likely hold viable flat-headed cat populations: the comparatively smaller Samusam National Park and the larger Maludam National Park. Both are isolated areas located along the north coast of Borneo. In addition to these two key localities recent sightings from Loagan Bunut National Park reveal the importance of this small national park for the flat-headed cat in Sarawak.
Sabah holds three lowland forest complexes. The largest is Tabin Wildlife Reserve (where Yasuda et al. obtained camera-trap photographs ) together with the northern coastal areas and Kulamba Wildlife Reserve. Currently, peat swamp and mangrove forests in the north are not completely protected, but if conserved they could potentially link this forest block with the Kinabatangan Wildlife Sanctuary (WS), a series of small forest patches along the lower Kinabatangan River. Although the Kinabatangan WS is highly fragmented, the high number of flat-headed cat sightings (mainly by tourists on night river cruises) along the tributaries of the Kinabatangan shows that this wildlife sanctuary is still a suitable area for the flat-headed cat (Video S1). The third forest complex identified in Sabah is an area south (Northern Malua) and north (Deramakot, Tangkulap & Segaliud Lokan) of the upper Kinabatangan. All forest reserves located in this forest complex are for commercial timber production under low-impact selective logging strategies, including a long felling cycle and extended regeneration times, and the pursuit to be certified by the Forest Stewardship Council (Deramakot and Tangkulap) or the Malaysian Timber Certification Council (Segaliud Lokan). The high number (15) of recent camera-trap photographs from this forest block  (AW & AM unpublished data) indicates that low-impact commercial use may not be entirely incompatible with the habitat requirements of flat-headed cats. In addition to these three key habitats in Sabah, flat-headed cats were recently reported in the Ulu Segama FR (AJH & JR) and in Maliau Basin (CT) but as these areas are mainly surrounded by upland forests we did not include them in the list of key localities.
In Brunei Darussalam, the large undisturbed forests of East Brunei with the completely protected areas of Tasek Merimbun (where Yasuda et al. obtained camera-trap photographs ), Ulu Mendaram and Belait peat swamps are the main shelter for flat-headed cats. This area is one of the largest undisturbed forest complexes in Borneo.
As in the other parts of the study region, most undisturbed and protected areas in Kalimantan are in the inland uplands. However, several larger forest blocks remain along the coastline and in lowland areas. In West Kalimantan the Danau Sentarum forest complex, one of the few large inland forest blocks, was predicted by our model to contain appropriate habitats for the flat-headed cat. A second key locality in West Kalimantan, which is currently mainly unprotected, ranges from south of Pontianak along the coastline to Gunung Palung National Park which was predicted by our model to be a good flat-headed cat habitat. In the southern part of Central Kalimantan we identified two interesting forest blocks. Tanjung Puting along the south coast is a large, already relatively isolated area. With an area of 5,300 km2, the Sabangau Peat Swamp Forest is the largest remaining peat swamp forest in Indonesia . In the vicinity of this protected area, large forests suitable for the flat-headed cat as predicted by our model remain. Its enormous size and the recent evidence of flat-headed cats from this area  identify this forest as one of the most important sites for the flat-headed cat. However, this area is threatened by fires and drainage, as are many of the peat swamp forests in Kalimantan. In East Kalimantan, the Muara Sebuku Nature Reserve and the larger forests to its south might have the potential to be a home for flat-headed cats. In Kalimantan, recent records also come from more upland forests  (Wulfraat in litt.). Although these sightings may suggest that flat-headed cats follow rivers upstream, we did not include these areas in our selection of key localities as there are only very few records obtained so far.
This overview of the potential former and recent distribution of the flat-headed cat is a first step to improve conservation efforts for this threatened felid. Verification of model predictions by field surveys is urgently needed, with camera-trapping efforts directed towards: (1) areas around lowland freshwater sources within forests where flat-headed cats are more likely to be detected; (2) plantation-dominated landscapes to corroborate–or cast doubt on–Khan's records of them living in plantations  for which we could not find any support and where further research needs to show whether flat-headed cats occur at forest edges and enter plantations; (3) small forest fragments that have been isolated for decades to see what sort of minimum patch size is realistic; (4) upland forests along major rivers and on well-watered upland plateaux with good forest cover to see how much of a role, if any, these areas could play in the species conservation; (5) investigations of large landscapes of logging estates to see if the findings from Deramakot  are generally representative. Postulations that flat-headed cats and other swamp forest specialists such as the otter civet (Cynogale bennettii) or the hairy-nosed otter (Lutra sumatrana) live in low-impact logging landscapes would give a high support to argue for their maintenance as low-impact logging estates rather than converting them into plantations.
Besides the intensification of large-scale survey efforts in lowland, peat swamp and mangrove forests, detailed site-specific multiple year studies using different methods simultaneously (e.g. radio-tracking, camera-trapping) would be needed to refine current knowledge on habitat and food preferences and requirements. Suitable areas for this purpose are the Pahang Peat Swamp Forest, Senepis Buluhala, Sabangau Peat Swamp Forest, Tasek Merimbun National Park, Kinabatangan Wildlife Sanctuary or Deramakot Forest Reserve where flat-headed cats have been recorded recently.
Our study showed that large areas of the predicted former distribution range have already been transformed into croplands or plantations, and based on the records from this study and our experiences we cast doubt on the statement that flat-headed cats live and reproduce in plantations. Furthermore, the low proportion of key forest areas under complete protection emphasises the urgency of further conservation action. We consider the flat-headed cat suitable to serve as a flagship species for the protection of several other endangered species associated with the threatened tropical lowland and peat swamp forests.
Summary of recent and historical records of the flat-headed cat Prionailurus planiceps.
(0.04 MB XLS)
We thank Marc Ancrenaz, Ang BeeBiaw, Henry Bernard, Mark Bezuijen, Namfon Cutter, Mark Rayan Darmaraj, Sahat July Tirta Dewi, Veerle Dossche, Gabriella M. Fredriksson, Magnus Gagnelid, Lon I. Grassman, Budsabong Kanchanasaka, Kae Kawanishi, Andrew Kitchener, Sanath Kumaran, Tim Laman, Barney Long, Anthony Lynam, Debbie Martyr, Jennifer McCarthy, Erik Meijaard, Wilson Novarino, Samhan Nyawa, Rajanathan Rajaratnam, Chris Shepherd, Nay Myo Shwe, Robert Stuebing, Su Su, Stephan Wulffraat, Dennis Yong and M. Yunus for discussion and assistance. We thank SarVision LLC and WWF for providing the Borneo land cover classification and the following museum institutions for providing a list of their flat-headed cat specimen: AMNH American Museum of Natural History, RMNH Naturalis, Leiden National Museum of Natural History, SMNS Staatliches Museum für Naturkunde Stuttgart, USNM Smithsonian Institution of National Museum of Natural History, ZMB Museum für Naturkunde Berlin, and ZRC Zoological Reference Collection University of Singapore. We thank the Department of Wildlife and National Parks, Malaysia, the Malaysian Nature Society, the Zoological Park Organization of Thailand and the Directorate General of Forest Protection and Nature Conservation - Department of Forestry, Indonesia. Furthermore AW, AJH, AM and JR are grateful to the Economic Planning Unit of Malaysia for issuing research permits for Sabah and the Sabah Wildlife Department, Sabah Forestry Department and the University Malaysia Sabah for their support. AJH & JR thank Yayasan Sabah, Danum Valley Management Committee and the FACE Foundation for issuing permission to conduct their research and Glen Reynolds and the Royal Society SEARRP for logistical support. SMC undertook this work in collaboration with David W. Macdonald. and their work on Bornean felids is part of the WildCRU/Panthera collaboration, and was carried out within the OuTrop-CIMTROP multi-disciplinary research project in the northern Sabangau forest, Central Kalimantan, Indonesia. SMC gratefully thanks Dr Suwido Limin and the Centre for the International Cooperation in Management of Tropical Peatlands (CIMTROP) for supporting her research and providing invaluable logistical support. MAJ thanks the Sabah Forestry Cooperation, Sarawak Forestry Department and the Universiti Malaysia Sarawak.
Conceived and designed the experiments: AW AC. Performed the experiments: AW AC DH. Analyzed the data: AW AC HH. Wrote the paper: AW AC. Contributed to the data collection: AW AJH AM CT SMC SS MAJ JR AS CB JS JWD. Edited the manuscript: AJH DH AM CT SMC SS MAJ JR ACS AS SD CB JS JWD HH.
- 1. Muul I, Lim BL (1970) Ecological and morphological observations of Felis planiceps. J Mammal 51: 806–808.I. MuulBL Lim1970Ecological and morphological observations of Felis planiceps.J Mammal51806808
- 2. Lydekker R (1895) Handbook of the Carnivora. R. Lydekker1895Handbook of the Carnivora.Part I. Cats, civets, and mongooses. Allen & Co, London, UK. Part I. Cats, civets, and mongooses. Allen & Co, London, UK.
- 3. Hearn A, Sanderson J, Ross J, Wilting A, Sunarto S (2008) Prionailurus planiceps. A. HearnJ. SandersonJ. RossA. WiltingS. Sunarto2008Prionailurus planiceps.In: IUCN 2009. IUCN Red List of Threatened Species. Version 2009.1.<www.iucnredlist.org>.
- 4. Sunquist M, Sunquist F (2009) Prionailurus planiceps. In: Wilson DE, Mittermeier R, editors. M. SunquistF. Sunquist2009Prionailurus planiceps.DE WilsonR. MittermeierA Handbook of the Mammals of the World, Volume 1: Carnivores, Lynx Edicions, Barcelona, Spain. A Handbook of the Mammals of the World, Volume 1: Carnivores, Lynx Edicions, Barcelona, Spain.
- 5. Nowell K, Jackson P (1996) Wild Cats: Status survey and conservation action plan. K. NowellP. Jackson1996Wild Cats: Status survey and conservation action plan.IUCN, Gland. IUCN, Gland.
- 6. Bezuijen MR (2000) The occurrence of the flat-headed cat Prionailurus planiceps in south-east Sumatra. Oryx 34: 222–226.MR Bezuijen2000The occurrence of the flat-headed cat Prionailurus planiceps in south-east Sumatra.Oryx34222226
- 7. Leyhausen P (1979) Cat behavior: The predatory and social behavior of domestic and wild cats. P. Leyhausen1979Cat behavior: The predatory and social behavior of domestic and wild cats.Garland STPM Press, New York. Garland STPM Press, New York.
- 8. Sunquist M, Sunquist F (2002) Wild Cats of the World. M. SunquistF. Sunquist2002Wild Cats of the World.University of Chicago Press, Chicago. University of Chicago Press, Chicago.
- 9. Mivart G (1892) The Cat. G. Mivart1892The Cat.Charles Scribner's Sons. New York. Charles Scribner's Sons. New York.
- 10. Lekagul B, McNeely JA (1977) Mammals of Thailand. B. LekagulJA McNeely1977Mammals of Thailand.Association for the Conservation of Wildlife, Bangkok. Association for the Conservation of Wildlife, Bangkok.
- 11. Payne J, Francis CM, Phillipps K (1985) A field guide to the mammals of Borneo. J. PayneCM FrancisK. Phillipps1985A field guide to the mammals of Borneo.Kota Kinabalu. The Sabah Society. Kota Kinabalu. The Sabah Society.
- 12. Khan M (1986) Some notes for the records: Flat-headed cat. Cat News 5: 11.M. Khan1986Some notes for the records: Flat-headed cat.Cat News511
- 13. Rowcliffe JM, Carbone C (2008) Surveys using camera traps: are we looking to a brighter future? Anim Conserv 11: 185–186.JM RowcliffeC. Carbone2008Surveys using camera traps: are we looking to a brighter future?Anim Conserv11185186
- 14. Chefaoui RM, Hortal J, Lobo JM (2005) Potential distribution modeling, niche characterization and conservation status assessment using GIS tools: a case study of Iberian Copris species. Biol Conserv 122: 327–338.RM ChefaouiJ. HortalJM Lobo2005Potential distribution modeling, niche characterization and conservation status assessment using GIS tools: a case study of Iberian Copris species.Biol Conserv122327338
- 15. Gaubert P, Papes M, Peterson AT (2006) Natural history collections and the conservation of poorly known taxa: ecological niche modeling in central African rainforest genets (Genetta spp.). Biol Conserv 130: 106–117.P. GaubertM. PapesAT Peterson2006Natural history collections and the conservation of poorly known taxa: ecological niche modeling in central African rainforest genets (Genetta spp.).Biol Conserv130106117
- 16. Guisan A, Broennimann O, Engler R, Vust M, Yoccoz NG, et al. (2006) Using niche-based models to improve the sampling of rare species. Conserv Biol 20: 501–511.A. GuisanO. BroennimannR. EnglerM. VustNG Yoccoz2006Using niche-based models to improve the sampling of rare species.Conserv Biol20501511
- 17. Papes M, Gaubert P (2007) Modelling ecological niches from low numbers of occurrences: assessment of the conservation status of poorly known viverrids (Mammalia, Carnivora) across two continents. Divers Distrib 13: 890–902.M. PapesP. Gaubert2007Modelling ecological niches from low numbers of occurrences: assessment of the conservation status of poorly known viverrids (Mammalia, Carnivora) across two continents.Divers Distrib13890902
- 18. Buermann W, Saatchi S, Smith TB, Zutta BR, Chavesi JA, et al. (2008) Predicting species distributions across the Amazonian and Andean regions using remote sensing data. J Biogeogr 35: 1160–1176.W. BuermannS. SaatchiTB SmithBR ZuttaJA Chavesi2008Predicting species distributions across the Amazonian and Andean regions using remote sensing data.J Biogeogr3511601176
- 19. DeMatteo KE, Loiselle BA (2008) New data on the status and distribution of the bush dog (Speothos venaticus): Evaluating its quality of protection and directing research efforts. Biol Conserv 141: 2494–2505.KE DeMatteoBA Loiselle2008New data on the status and distribution of the bush dog (Speothos venaticus): Evaluating its quality of protection and directing research efforts.Biol Conserv14124942505
- 20. Sánchez-Cordero V, Stockwell D, Sarkar S, Liu H, Stephens CR, et al. (2008) Competitive interactions between felid species may limit the southern distribution of bobcats Lynx rufus. Ecography 31: 757–764.V. Sánchez-CorderoD. StockwellS. SarkarH. LiuCR Stephens2008Competitive interactions between felid species may limit the southern distribution of bobcats Lynx rufus.Ecography31757764
- 21. Thorn JS, Nijman V, Smith D, Nekaris KAI (2009) Ecological niche modelling as a technique for assessing threats and setting conservation priorities for Asian slow lorises (Primates: Nycticebus). Divers Distrib 15: 289–298.JS ThornV. NijmanD. SmithKAI Nekaris2009Ecological niche modelling as a technique for assessing threats and setting conservation priorities for Asian slow lorises (Primates: Nycticebus).Divers Distrib15289298
- 22. Lim BK, Peterson AT, Engstrom MD (2002) Robustness of ecological niche modeling algorithms in Guyana. Biodiv Conserv 11: 1237–1246.BK LimAT PetersonMD Engstrom2002Robustness of ecological niche modeling algorithms in Guyana.Biodiv Conserv1112371246
- 23. Phillips SJ, Anderson RP, Schapire RE (2006) Maximum entropy modeling of species geographic distributions. Ecol Model 190: 231–259.SJ PhillipsRP AndersonRE Schapire2006Maximum entropy modeling of species geographic distributions.Ecol Model190231259
- 24. Hernandez PA, Graham CH, Master LL, Albert DL (2006) The effect of sample size and species characteristics on performance of different species distribution modeling methods. Ecography 29: 773–785.PA HernandezCH GrahamLL MasterDL Albert2006The effect of sample size and species characteristics on performance of different species distribution modeling methods.Ecography29773785
- 25. Pearson RG, Raxworthy CJ, Nakamura M, Townsend Peterson A (2007) Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagaskar. J Biogeogr 34: 102–117.RG PearsonCJ RaxworthyM. NakamuraA. Townsend Peterson2007Predicting species distributions from small numbers of occurrence records: a test case using cryptic geckos in Madagaskar.J Biogeogr34102117
- 26. Hernandez PA, Franke I, Herzog SK, Pacheco V, Paniagua L, et al. (2008) Predicting species distributions in poorly-studied landscapes. Conserv Biol 17: 1353–1366.PA HernandezI. FrankeSK HerzogV. PachecoL. Paniagua2008Predicting species distributions in poorly-studied landscapes.Conserv Biol1713531366
- 27. Povey K, Howard JG, Sunarto S, Priatna D, Ngoprasert D, et al. (2009) Clouded Leopard and Small Felid Conservation Summit Final Report. K. PoveyJG HowardS. SunartoD. PriatnaD. Ngoprasert2009Clouded Leopard and Small Felid Conservation Summit Final Report.IUCN/SSC Conservation Breeding Specialist Group: Apple Valley, MN. IUCN/SSC Conservation Breeding Specialist Group: Apple Valley, MN.
- 28. Hijmans RJ, Cameron SE, Parra JL, Jones P, Jarvis A (2005) Very high resolution interpolated climate surfaces for global land areas. Int J Climatol 25: 1965–1978.RJ HijmansSE CameronJL ParraP. JonesA. Jarvis2005Very high resolution interpolated climate surfaces for global land areas.Int J Climatol2519651978
- 29. Saatchi S, Buermann W, ter Steege H, Mori S, Smith TB (2008) Modeling distribution of Amazonian tree species and diversity using remote sensing measurements. Remote Sensing of Environment 112: 2000–2017.S. SaatchiW. BuermannH. ter SteegeS. MoriTB Smith2008Modeling distribution of Amazonian tree species and diversity using remote sensing measurements.Remote Sensing of Environment11220002017
- 30. Elith J, Graham CH, Anderson RP, Dudík M, Ferrier S, Guisan A, et al. (2006) Novel methods improve prediction of species' distributions from occurrence data. Ecography 29: 129–151.J. ElithCH GrahamRP AndersonM. DudíkS. FerrierA. Guisan2006Novel methods improve prediction of species' distributions from occurrence data.Ecography29129151
- 31. Dormann CF, Purschke O, García Márquez JA, Lautenbach S, Schröder B (2008) Components of uncertainty in species distribution analysis; a case study of the grey shrike. Ecology 89: 3371–3386.CF DormannO. PurschkeJA García MárquezS. LautenbachB. Schröder2008Components of uncertainty in species distribution analysis; a case study of the grey shrike.Ecology8933713386
- 32. Wisz MS, Hijmans RJ, Li J, Peterson AT, Graham CH, et al. (2008) Effects of sample size on the performance of species distribution models. Divers Distrib 14: 763–773.MS WiszRJ HijmansJ. LiAT PetersonCH Graham2008Effects of sample size on the performance of species distribution models.Divers Distrib14763773
- 33. Zheng B, Agresti A (2000) Summarizing the predictive power of a generalized linear model. Statistics in Medicine 19: 1771–1781.B. ZhengA. Agresti2000Summarizing the predictive power of a generalized linear model.Statistics in Medicine1917711781
- 34. Izawa M, Doi T, Ono J (1990) Ecological study on the two species of Felidae in Japan. Wildl Conserv 140–143.M. IzawaT. DoiJ. Ono1990Ecological study on the two species of Felidae in Japan.Wildl Conserv140143
- 35. Rajaratnam R, Sunquist M, Rajaratnam L, Ambu L (2007) Diet and habitat selection of the leopard cat (Prionailurus bengalensis borneoensis) in an agricultural landscape in Sabah, Malaysian Borneo. J Trop Ecol 23: 209–217.R. RajaratnamM. SunquistL. RajaratnamL. Ambu2007Diet and habitat selection of the leopard cat (Prionailurus bengalensis borneoensis) in an agricultural landscape in Sabah, Malaysian Borneo.J Trop Ecol23209217
- 36. Grassman LI (2000) Movements and diet of the leopard cat Prionailurus bengalensis in a seasonal evergreen forest in south-central Thailand. Acta Theriol 45: 421–426.LI Grassman2000Movements and diet of the leopard cat Prionailurus bengalensis in a seasonal evergreen forest in south-central Thailand.Acta Theriol45421426
- 37. Rabinowitz A (1990) Notes on the behaviour and movements of leopard cats, Felis bengalensis, in a dry tropical forest mosaic in Thailand. Biotropica 22: 397–403.A. Rabinowitz1990Notes on the behaviour and movements of leopard cats, Felis bengalensis, in a dry tropical forest mosaic in Thailand.Biotropica22397403
- 38. Allen AW, Jordan PA, Terrell JW (1987) Habitat suitability index models: moose, Lake Superior region. U.S. 47 p. AW AllenPA JordanJW Terrell1987Habitat suitability index models: moose, Lake Superior region. U.S.47Fish Wildlife Service Biological Report 82 (10.155). Fish Wildlife Service Biological Report 82 (10.155).
- 39. Cliff AD, Ord JK (1981) Spatial Processes. Models and applications. London: Pion. AD CliffJK Ord1981Spatial Processes.Models and applicationsLondonPion
- 40. Haining R (2003) Spatial Data Analysis. Theory and Practice. Cambridge, UK: Cambridge University Press. R. Haining2003Spatial Data Analysis.Theory and PracticeCambridge, UKCambridge University Press
- 41. Cerioli A (1997) Modified tests of independence in 2×2 tables with spatial data. Biometrics 53: 619–628.A. Cerioli1997Modified tests of independence in 2×2 tables with spatial data.Biometrics53619628
- 42. Dudley N (2008) Guidelines for Applying Protected Area Management Categories. Gland, Switzerland: IUCN. N. Dudley2008Guidelines for Applying Protected Area Management Categories.Gland, SwitzerlandIUCN
- 43. Sunquist M, Leh C, Sunquist F, Hills DM, Rajaratnam R (1994) Rediscovery of the Bornean bay cat. Oryx 28: 67–70.M. SunquistC. LehF. SunquistDM HillsR. Rajaratnam1994Rediscovery of the Bornean bay cat.Oryx286770
- 44. Meijaard E (1997) The Bay Cat in Borneo. Cat News 27: E. Meijaard1997The Bay Cat in Borneo.Cat News27
- 45. Azlan MJ, Sanderson J (2007) Geographic distribution and conservation status of the bay cat Catopuma badia, a Bornean endemic. Oryx 41: 394–397.MJ AzlanJ. Sanderson2007Geographic distribution and conservation status of the bay cat Catopuma badia, a Bornean endemic.Oryx41394397
- 46. Hughes JB, Round PD, Woodruff DS (2003) The Indochinese-Sundaic Faunal Transition at the Isthmus of Kra: An Analysis of Resident Forest Bird Species Distributions. J Biogeogr 30: 569–580.JB HughesPD RoundDS Woodruff2003The Indochinese-Sundaic Faunal Transition at the Isthmus of Kra: An Analysis of Resident Forest Bird Species Distributions.J Biogeogr30569580
- 47. Holmes D (1999) Climate. Kota Kinabalu: Natural History Publications. D. Holmes1999Climate.Kota KinabaluNatural History PublicationsIn: Smythies BE, Davison GWH. 4th ed. The birds of Borneo eds.
- 48. Araújo MB, Williams PH (2000) Selecting areas for species persistence using occurrence data. Biol Conserv 96: 331–45.MB AraújoPH Williams2000Selecting areas for species persistence using occurrence data.Biol Conserv9633145
- 49. Ferrier S, Watson G, Pearce J, Drielsma M (2002) Extended statistical approaches to modelling spatial pattern in biodiversity in northeast New South Wales. I. Species-level modelling. Biodiv Conserv 11: 2275–2307.S. FerrierG. WatsonJ. PearceM. Drielsma2002Extended statistical approaches to modelling spatial pattern in biodiversity in northeast New South Wales. I. Species-level modelling.Biodiv Conserv1122752307
- 50. FAO (2007) State of the World's Forests. FAO2007State of the World's Forests.Food and Agriculture Organisation of the United Nations, Rome. Food and Agriculture Organisation of the United Nations, Rome.
- 51. Sodhi NS, Koh LP, Brook BW, Ng PKL (2004) Southeast Asian biodiversity: an impending disaster. Trends Ecol Evol 19: 654–660.NS SodhiLP KohBW BrookPKL Ng2004Southeast Asian biodiversity: an impending disaster.Trends Ecol Evol19654660
- 52. Rautner M, Hardiono M, Alfred RJ (2005) Borneo: Treasure Island at Risk. M. RautnerM. HardionoRJ Alfred2005Borneo: Treasure Island at Risk.Report for WWF Germany. Report for WWF Germany.
- 53. PRB (2007) 2007 World population data sheet. PRB20072007 World population data sheet.Population Reference Bureau. Washington DC. Population Reference Bureau. Washington DC.
- 54. Koh LP, Wilcove DS (2007) Cashing in palm oil for conservation. Nature 448: 993–994.LP KohDS Wilcove2007Cashing in palm oil for conservation.Nature448993994
- 55. Butler RA, Koh LP, Ghazoul J (2009) REDD in the red: palm oil could undermine carbon payment schemes. Conserv Letters 2: 67–73.RA ButlerLP KohJ. Ghazoul2009REDD in the red: palm oil could undermine carbon payment schemes.Conserv Letters26773
- 56. Venter O, Meijaard E, Possingham H, Dennis R, Sheil D, et al. (2009) Carbon payments as a safeguard for protecting threatened tropical mammals. Conserv Letters 2: 123–129.O. VenterE. MeijaardH. PossinghamR. DennisD. Sheil2009Carbon payments as a safeguard for protecting threatened tropical mammals.Conserv Letters2123129
- 57. Soemarsono (1996) First photograph of rare Flat-headed cat reported from Way Kambas National Park, Sumatra. Soemarsono1996First photograph of rare Flat-headed cat reported from Way Kambas National Park, Sumatra.Press release Sumatran Tiger Project, Lampung Tengah, Indonesia. Press release Sumatran Tiger Project, Lampung Tengah, Indonesia.
- 58. Bezuijen MR (2003) The flat-headed cat in the Merang river region of south Sumatra. Cat News 38: 26–27.MR Bezuijen2003The flat-headed cat in the Merang river region of south Sumatra.Cat News382627
- 59. Uryu Y, Mott C, Foead N, Yulianto K, Budiman A, et al. (2008) Deforestation, Forest Degradation, Biodiversity Loss and CO2 Emissions in Riau, Sumatra, Indonesia, WWF-Indonesia Technical Report, Jakarta, Indonesia. Y. UryuC. MottN. FoeadK. YuliantoA. Budiman2008Deforestation, Forest Degradation, Biodiversity Loss and CO2 Emissions in Riau, Sumatra, Indonesia, WWF-Indonesia Technical Report, Jakarta, Indonesia.
- 60. Yasuda M, Matsubayashi H, Rustam , Numata S, Rafiah JAS, et al. (2007) Recent Cat Records by Camera Traps in Peninsular Malaysia and Borneo. Cat News 47: 12–14.M. YasudaH. MatsubayashiRustamS. NumataJAS Rafiah2007Recent Cat Records by Camera Traps in Peninsular Malaysia and Borneo.Cat News471214
- 61. Mohamed A, Samejima H, Wilting A (2009) Records of five Bornean cat species from Deramakot Forest Reserve in Sabah, Malaysia. Cat News 51: 12–15.A. MohamedH. SamejimaA. Wilting2009Records of five Bornean cat species from Deramakot Forest Reserve in Sabah, Malaysia.Cat News511215
- 62. Shepherd PA, Rieley JO, Page SE (1997) The relationship between forest structure and peat characteristics in the upper catchment of the Sungai Sebangau, Central Kalimantan. In: Rieley JO, Page SE, editors. Biodiversity and Sustainability of Tropical Peatlands. pp. 191–210.PA ShepherdJO RieleySE Page1997The relationship between forest structure and peat characteristics in the upper catchment of the Sungai Sebangau, Central Kalimantan.JO RieleySE PageBiodiversity and Sustainability of Tropical Peatlands191210Samara Publishing, Cardigan, UK. Samara Publishing, Cardigan, UK.
- 63. Cheyne SM, Morrogh-Bernard H, Macdonald DW (2009) First flat-headed cat photographed in Sabangau Peat-swamp Forest, Indonesian Borneo. Cat News 51: 16.SM CheyneH. Morrogh-BernardDW Macdonald2009First flat-headed cat photographed in Sabangau Peat-swamp Forest, Indonesian Borneo.Cat News5116
- 64. Meijaard E, Sheil D, Daryono (2005) Flat-headed cat record in east Kalimantan. Cat News 43: 24.E. MeijaardD. SheilDaryono2005Flat-headed cat record in east Kalimantan.Cat News4324
- 65. Yasuma S (1994) An invitation to the mammals of East Kalimantan. pp. 176–180.S. Yasuma1994An invitation to the mammals of East Kalimantan.176180PUSREHUT Special Publication No 3. PUSREHUT Special Publication No 3.
- 66. UNESCO/IUCN (2004) Tropical rainforest heritage of Sumatra Indonesia. Paris.UNESCO/IUCN2004Tropical rainforest heritage of Sumatra Indonesia.Paris
- 67. Lim BL, Abdul Rahman O (1961) Observations on the habits in captivity of two species of wild cats, the leopard cat and the flat-headed cat. Malays Nat Hist J 15: 8–51.BL LimO. Abdul Rahman1961Observations on the habits in captivity of two species of wild cats, the leopard cat and the flat-headed cat.Malays Nat Hist J15851
- 68. Medway L (1965) Mammals of Borneo. L. Medway1965Mammals of Borneo.151 pField keys and an annotated checklist. Malaysian Branch of the Royal Asiatic Society, Kuala Lumpur, Malaysia. Field keys and an annotated checklist. Malaysian Branch of the Royal Asiatic Society, Kuala Lumpur, Malaysia.
- 69. Azlan MJ, Noorafizah D, Sanderson J (2007) Historical records of felid collections in the Sarawak Museum. Cat News 47: MJ AzlanD. NoorafizahJ. Sanderson2007Historical records of felid collections in the Sarawak Museum.Cat News47