Aim Deforestation is rapidly altering Southeast Asian landscapes, resulting in some of the highest rates of habitat loss worldwide. Among the many species facing declines in this region, clouded leopards rank notably for their ambassadorial potential and capacity to act as powerful levers for broader forest conservation programmes. Thus identifying core habitat and conservation opportunities are critical for curbing further Neofelis declines and extending umbrella protection for diverse forest biota similarly threatened by widespread habitat loss. Furthermore, a recent comprehensive habitat assessment of Sunda clouded leopards (N. diardi) highlights the lack of such information for the mainland species (N. nebulosa), and facilitates a comparative assessment. Location Southeast Asia. Methods Species-habitat relationships are scale-dependent, yet <5% of all recent habitat modeling papers apply robust approaches to optimize multivariate scale relationships. Using one of the largest camera trap datasets ever collected, we developed scale-optimized species distribution models for two con-generic carnivores, and quantitatively compared their habitat niches. Results We identified core habitat, connectivity corridors, and ranked remaining habitat patches for conservation prioritization. Closed canopy forest was the strongest predictor, with ~25% lower Neofelis detections when forest cover declined from 100 to 65%. A strong, positive association with increasing precipitation suggests ongoing climate change as a growing threat along drier edges of the species' range. While deforestation and land use conversion were deleterious for both species, N. nebulosa was uniquely associated with shrublands and grasslands. We identified 800km2 as a minimum patch size for supporting clouded leopard conservation. Main Conclusions We illustrate the utility of multi-scale modeling for identifying key habitat requirements, optimal scales of use, and critical targets for guiding conservation prioritization. Curbing deforestation and development within remaining core habitat and dispersal corridors, particularly in Myanmar, Laos, and Malaysia, is critical for supporting evolutionary potential of clouded leopards and conservation of associated forest biodiversity.
% Closed Forest 16km-radius Window
Using Hansen et al.’s (2013) original percent forest cover layer (http://earthenginepartners.appspot.com/science-2013-global-forest/download_v1.2.html), we reclassified >40% forest cover as ‘closed forest’. We then used FRAGSTATS (McGarigal, Cushman, Neel, & Ene, 2012) to calculate the percentage of the landscape occupied by closed forest habitat within a 16km-radius moving window. The data is in raster .tif format for use in GIS. We resampled the source data from its native 30m resolution to 250m resolution. Coordinate system: Asia South Albers Equal Area Conic.
CF_PLAND_16000.tif
Shrubland/Grassland Correlation Length 16km-radius Window
Original data for this GIS raster layer was obtained from the ESA CCI Land Cover Layer 300m 2015 (www.esa-landcover-cci.org). We first reclassified the original data layer, denoting shrubland/grassland as present (1) or absent (0). We then used FRAGSTATS (McGarigal, Cushman, Neel, & Ene, 2012) to calculate the correlation length or area-weighted mean of gyration within a 16km-radius moving window. Correlation length measures the average distance an individual can travel within a habitat patch, or in this case the average extensiveness of shrubland/grassland habitat within a 16km-radius window. We re-sampled the source data from its native 300m resolution to 250m. Units are in meters. Coordinate system: Asia South Albers Equal Area Conic.
SHG_GYRATE_16000.tif
Compound Topographic Index Focal Mean 500m-radius Window
This GIS raster was derived from Jarvis, Reuter, Nelson, & Guevara's (2008) digital elevation model (http://srtm.csi.cgiar.org) using the Geomorphometry & Gradient Metrics Toolbox (Evans, Oakleaf, Cushman, & Theobald, 2014) in ArcGIS 10.2.2 (ESRI, Redlands, CA, USA, 2011). Compound topographic index (CTI) characterizes flow accumulation. Low elevation drainages incur high CTI, whereas mountaintops and ridgelines exhibit low CTI. We then used neighborhood statistics in ArcGIS to calculate the focal mean within a 500m-radius moving window. We re-sampled the native 90m resolution DEM to 250m. Coordinate system: Asia South Albers Equal Area Conic.
CTI_FM_500.tif
% Mosaic Cropland & Native Vegetation 1km-radius Window
Original data for this GIS raster was obtained from the ESA CCI Land Cover Layer 300m 2015 (www.esa-landcover-cci.org). We first reclassified the original data layer, denoting mosaic cropland and native vegetation habitat as present (1) or absent (0). We then used FRAGSTATS (McGarigal, Cushman, Neel, & Ene, 2012) to calculate the percent of the landscape occupied by mosaic habitat within a 1km-radius moving window. We re-sampled the source data from its native 300m resolution to 250m. Coordinate system: Asia South Albers Equal Area Conic.
Mosaic_PLAND_1000.tif
Protected Area Correlation Length 8km-radius Window
Data for this GIS raster layer was obtained from the IUCN & UNEP-WCMC 2017 World Database on Protected Areas (http://www.protectedplanet.net/). We first reclassified the original data layer, denoting protected status as present (1) or absent (0). We then used FRAGSTATS (McGarigal, Cushman, Neel, & Ene, 2012) to calculate the correlation length or area-weighted mean of gyration within an 8km-radius moving window. Here correlation length measures the average distance an animal could travel within protected areas within an 8km-radius window. Data are at 250m resolution; units are in meters. Coordinate system: Asia South Albers Equal Area Conic.
PA_GYRATE_8000.tif
Mean Annual Precipitation Focal Mean 32km-radius Window
Data for this GIS raster layer was obtained from the WorldClim Global Climate Data Portal (http://worldclim.org/version2). We then used neighborhood statistics in ArcGIS to calculate the focal mean of MAP within a 32km-radius moving window. We re-sampled the source data from its native 30-arcsec resolution (~1km) to 250m. Units are in millimeters of precipitation. Coordinate system: Asia South Albers Equal Area Conic.
MAP_FM_32000.tif
Mainland Clouded Leopard Habitat Suitability Model - 4 Thresholds
The zip file contains four GIS raster layers representing a range of thresholds of the mainland clouded leopard habitat suitability model. Layers include 50th, 75th, 90th, and 97.5th percentiles of the multi-scale GLMM. Suitable habitat is classified as 1; all habitat below a specified threshold is classified as 0. Raster resolution: 250m; coordinate system: Asia South Albers Equal Area Conic.
CL_GLMM_thresholds.zip
Mainland Clouded Leopard Habitat Suitability Model - Continuous
This GIS raster layer is the projected multi-scale habitat suitability model for the mainland clouded leopard, based on beta coefficients of the GLMM and a mean of 79.47 camera trap nights averaged across all camera stations. Model projection units are number of clouded leopard observations predicted across multivariate niche space, given 79.47 active camera trap nights. Thus the specific output value is of little importance; the model output should rather be interpreted on a relative scale. Habitat >4000m elevation was reclassified as non-suitable and assigned the lowest model output value. Resolution: 250m; coordinate system: Asia South Albers Equal Area Conic.
CL_GLMM_continuous.tif
Top 28 High Quality Habitat Patches for the Mainland Clouded Leopard
This GIS raster layer contains the top 28 patches of high quality habitat for the mainland clouded leopard, based on the 90th percentile threshold of the habitat suitability model. We identified and ranked all habitat patches >1000km2, based on descending habitat area. Table S8 in the Supplementary Information file associated with Macdonald et al. (2019) provides additional information for each habitat patch, including associated protected areas, mean correlation length within each patch, and percent of patch area designated as protected. Resolution: 500m; coordinate system: Asia South Albers Equal Area Conic.
CL_HS_3classes_Top28_PATCHES.tif