This figure shows the results of the jackknife procedure on the full Maxent model. doi:10.1371/journal.pone.0049142.g001 

Context in source publication

Context 1

... from the World Database of Protected Areas [30]. concessions. Concessions for industrial oil palm plantations (IOPP) are granted by governments to allow the establishment and management of industrial, monoculture oil palm estates. In Indonesia, since decentralisation in the year 2000 [31], the majority of new IOPP permits have been issued by district level governments on lands categorised as ‘conversion forests’ by national land use plans. ‘Conversion forests’ (HPK Hutan Produksi Konversi , and APL Areal Penggunaan Lain ) include regions allocated explicitly for non-forest purposes. A smaller fraction of the permits have been issued on lands classified as ‘production forests’, in areas where there is a deficit of ‘conversion forests’, for example in the province of Central Kalimantan. Similar procedures are followed by the state governments of Sabah and Sarawak, although parts of commercial forest reserves (up to ten percent) can be de-gazetted and reclassified as agricultural lands and converted to industrial plantations. For Indonesian Borneo (Kalimantan), maps of oil palm concession boundaries were obtained from various provincial governments at a 1:250,000 scale. For South and East Kalimantan provinces, the maps are current as of 2005 [32,33,34,35]. For Central and West Kalimantan provinces, the maps are current as of 2007 and 2008, respectively [36,37]. If the data were in hard- copy format, they were scanned and digitized in ArcGIS by the Indonesian NGO Living Landscapes Indonesia. For the Malaysian state of Sarawak, maps of oil palm concessions were obtained from AidEnvironment and the Sarawak Dayak Iban Association. Because no official government data of land use are publicly available for Sarawak, these maps were compiled from a range of different sources across different years. In the absence of official up-to-date government maps for Sarawak, we cannot verify the accuracy of the concession data we used in our analysis. For Sabah, there was a dearth of data available regarding spatially explicit concessions. Therefore, to include plantation areas within the analyses we took the ‘‘current’’ plantation extent as a surrogate. To derive this information, we digitised (in ArcGIS 9.3) the extent of plantations throughout Sabah from SPOT 5 2.5 m (2011–2010), SPOT 5 10 m (2009–2007) and Landsat 30 m (2005) satellite images. Industrial tree plantation (ITP) concessions. An ITP concession is a right granted by a government to develop an area of land into an industrial monoculture timber plantation (e.g. Acacia mangium, Hevea or Eucalyptus spp.). In Indonesia, ITP permits are issued by the Indonesian Ministry of Forestry on lands classified as ‘production forests’. Production forests’ (HPH Hutan Produksi , 300 m asl . and HPT Hutan Produksi Terbatas . 300–500 m asl ) comprise areas allocated for commercial logging, where conversion to another land-use is prohibited. However, the conversion of natural forests to timber plantations is not recognized as deforestation by the United Nations Framework Convention on Climate Change because tree plantations are legally defined as ‘forest’ [38]. In Sarawak and Sabah, ITP permits are issued by the Forestry Department on lands classified as Commercial Forest Reserves (equivalent to Indonesia’s production forests). In Sarawak, maps of industrial timber plantation concessions (called Reforestation Licenses, LPF) current as of 2008 were obtained from the Bruno Manser fund report in pdf format [39], scanned and digitized in ArcGIS. For Sabah, maps of Industrial tree plantations (called ITPs) were obtained from digitised Landsat 2000 data (derived from WWF Germany). Logging concessions in natural forests. Companies pos- sessing logging concession licenses have the right to extract natural timber from natural forests. Deforestation (or open clearing) is prohibited in logging concessions, as timber should be extracted in a sustainable manner. For Kalimantan, maps of logging concessions were obtained from the national spatial planning agency (BAPLAN) of the Ministry of Forestry, in 1:250,000 scale and in digital format. The maps are current as of 2009–2010. For Sabah, a map of ‘production forest’ areas partitioned into forest management units was obtained from the Sabah Forestry Department. For Sarawak, maps of logging concessions for year 1996 were obtained from the Bruno Manser Fund report in pdf format [39], scanned and digitized in ArcGIS. In the absence of official up-to-date government maps for Sarawak, we cannot verify the accuracy of the logging concession map we used in our analysis. The model fit as measured by the mean area under the curve (AUC) of the receiver operating characteristics (ROC) was 0.810 which is considered to be ‘excellent’ [40]. The contextual layer with the highest percentage contribution to the model was annual rainfall (39%), followed by land cover (19%), soil type (15%) and temperature range (14%) (Table 2). Similarly, the jackknife procedure (Figure 1) also indicated annual rainfall by itself contributed more than any other individual variable and that excluding annual rainfall from the model reduced the gain of the model more than the exclusion of any other variable. The latter indicates that annual rainfall contributes to a gain within the model that is not present in any of the other variables. The modelled orang-utan distribution map covers 155,106 km 2 (21% of Borneo’s landmass) (Table 3). It shows four main distinct distribution areas (Figure 2c): a) Sabah and the north-eastern region of East Kalimantan where P. p. morio occurs; b) the southern and central East Kalimantan area where P. p. morio also occurs; c) the Central Kalimantan and south-western part of West Kalimantan area where P. p. wurmbii occurs; and d) the northern part of West Kalimantan and southern part of Sarawak where P. p. pygmaeus is found. The largest area of orang-utan distribution is found in Central Kalimantan, followed by West Kalimantan, East Kalimantan, Sabah, Sarawak, and South Kalimantan (Table 3). The area in which orang-utans might occur in South Kalimantan only covers 13 km 2 . An estimated 29% of the current orang-utan distribution in Borneo is found in natural forests exploited for timber, where logging is allowed but forest conversion is prohibited. A smaller proportion (22%) of orang-utan distribution lies within protected areas where logging and conversion are prohibited (Table 3, Figure 2d). An almost equal percentage (19%) overlaps with largely undeveloped (i.e. still forested) industrial oil palm concessions, and 6% overlaps with largely undeveloped Industrial Tree Plantations. These concessions are still forested but are expected to become converted to plantations in the near future. Finally, an estimated 24% of the orang-utan distribution range occurs outside of protected areas and outside of concessions, with 11% and 13% in ‘‘conversion’’ forests, and in ‘‘production’’ forests, respectively. This paper provides the latest estimate of the Bornean orangutan distribution based on the largest geo-referenced survey dataset available. This map makes great improvements upon previous estimates [5,25]. Nevertheless, within all modelling approaches, commission errors (when a species is mistakenly thought to be present) and omission errors (when a species is mistakenly thought to be absent) are inevitable [41]. We have tried to minimize both kinds of inaccuracies by: (1) allowing the knowledge of orang-utan specialists to facilitate the post- processing of these maps; (2) by running fine-tune iterations of the model to enhance model performance; and, (3) by eliminating areas known to be unsuitable for orang-utans, i.e. areas in which there is no forest anymore (4) as well as areas where orang-utans are naturally found to be absent or only used by transient individuals. This current distribution map has therefore, made significant advancements in understanding the spatial extent of orang-utan distribution and we hope will be pertinent for informing land-use decision making, conservation actions, and associated land use policies. The most important predictor variables for orang-utan distribution in Borneo identified during our analysis were annual rainfall, land cover, soil types and mean daily temperature range. Orang-utans are not predicted to occur in areas of high annual rainfall (eastern part of Sarawak and the northern part of Central Kalimantan for example). High rainfall can influence orang-utan distribution through several indirect processes such as leaching of soils which could lead to less productive forests [42,43,44] and/or creating high-cloud cover, which has been shown to lead to a reduction in solar radiation and thereby to lower primary productivity in Borneo [45]. Among the remaining variables, land cover is obviously important for orang-utan distribution since orang-utans in general occur only in areas with some natural tree cover left [17]. The influence of mean daily temperature on orangutan distribution is also likely to be indirect, although we are not sure about the underlying mechanisms such as potential influences of temperature on thermoregulation that could influence activity patterns, density and distribution [46]. Areas with a very low mean daily temperature do not contain orang-utans and those occur in Brunei (note that we do not suggest that other factors such a rivers, other geographical boundaries or past hunting are not relevant for this gap in the orang-utan distribution) and the far eastern part of Sarawak and the southern part of East Kalimantan. Also, one area of extremely high daily mean temperature variation in Sarawak does not contain orang-utans. Protected areas. No specific study has addressed the relative survival of orang-utans in protected areas compared to other types of land-use type, yet some general patterns can be inferred. Protected areas are the only land-use type where deforestation, logging, and hunting ...