Implications of landscape genetics and connectivity of snow leopard in the Nepalese Himalayas for its conservation

Snow leopards (Panthera uncia) inhabit alpine and subalpine zones of Central Asia (ca 2500–5800 m), relying on steep, rugged terrain and wild ungulate prey. Despite a global estimate of 2710–3386 mature individuals, populations are declining and occupy only about 27% of potential range, with threats including habitat loss and fragmentation, depletion of wild prey, livestock depredation prompting retaliatory killings, disease, and climate-driven treeline shifts reducing alpine habitat. Reduced connectivity among suitable habitat patches risks inbreeding, genetic drift, bottlenecks, and metapopulation collapse. Although genetic tools (mtDNA and microsatellites) have enabled population monitoring, Nepal lacks analyses of spatial population genetic structure using Bayesian clustering, and connectivity between suitable habitats has not been evaluated. This study aims to: (i) map suitable habitats for snow leopards in Nepal using habitat suitability modelling, (ii) identify realistic dispersal corridors linking these areas via connectivity analyses, and (iii) describe population genetic structure within and among study areas to inform conservation of a viable metapopulation. Nepal is an apt model due to strong community involvement in snow leopard conservation, potentially facilitating protection of key corridors and habitats.

Prior work developed snow leopard-specific primers enabling noninvasive genetic studies for species, sex, individual identification, and demographic inference; studies assessed genetic diversity at microsatellites, and range-wide phylogeography supporting three subspecies. Abundance and density estimates using fecal DNA and camera trapping exist for parts of Nepal, yet only one Nepal-focused study reported descriptive genetic diversity without spatial structure via Bayesian clustering. Habitat suitability modelling has been conducted in Nepal and Tibet, but inter-area connectivity has not been analyzed. Other Maxent-based studies identified climatic variables—particularly annual mean temperature and altitude, along with ruggedness—as key determinants of snow leopard distribution (e.g., Qomolangma and Sanjiangyuan reserves). Debates over the species’ IUCN downlisting cite limited sampling potentially overestimating abundance. Connectivity is recognized as critical for maintaining gene flow and metapopulation viability, but empirical integration of habitat suitability, connectivity, and genetics in Nepal remained a gap addressed by this work.

Study areas: Four sites in Nepal’s northern Himalayas were surveyed: Lower Mustang (LM; ca 100 km²), Upper Manang (UM; ca 105 km²) within Annapurna Conservation Area (ACA), and North Sagarmatha (N-S; Gokyo and Phortse; ca 50 km²) and South-West Sagarmatha (SW-S; Namche and Thame; ca 50 km²) within Sagarmatha National Park (SNP). In Thame valley, no snow leopards were detected. Field sampling and occurrence data: From 2014–2016, 268 putative snow leopard samples (261 feces, 6 hair, 1 urine) were collected using SLIMS protocols. Camera traps (Bushnell HD Trophy Cam) were deployed across 2014–2016 with 32–48 cameras, totaling 4567 trap-nights across wet and dry seasons. Opportunistic scrape records were collected (2004–2016). In total, 628 occurrence records at 482 locations were compiled, primarily from SNP and ACA (UM and LM). Species and sex identification: DNA was extracted and a ~148 bp mtDNA cytochrome-b fragment was amplified using CYTB-SCT-PUN-F’/R’ for snow leopard species confirmation. Sex identification targeted the Y-chromosome AMELY intron (~200 bp) by PCR with triplicate runs per sample, including positive/negative controls; results were confirmed via replicate criteria as specified. Individual identification and genotyping: Six polymorphic microsatellite loci on six chromosomes (e.g., PUN1157, PUN229, PUN124, PUN935, PUN894, PUN132) with fluorescently labeled primers were amplified in two multiplex combinations. PCRs (7 µl) used Qiagen 2x Master Mix, 5X Q solution, and 2 µl DNA; products were diluted 1:50 and sized on an ABI 31 Analyzer. A multiple-tube approach with three replicates per sample was used. Allele calling was done in GeneMarker v1.85; consensus genotypes and binning with Autobin; genotyping error checks (null alleles, large allele dropout, stuttering) with Micro-Checker. Population genetics and structure: Identical genotypes were assessed in CERVUS 3.0.7; basic diversity metrics computed in GenAlEx 6.5. Population structure was inferred with STRUCTURE 2.3.4 under correlated allele frequencies and ancestry models. K was set from 1–10, with 5 runs per K; results visualized in Structure Selector; the most probable K determined with ΔK (AK) and runs combined in CLUMPP 1.1.2. (Five iterations for each K with MCMC steps and burn-in as reported in the text.) Habitat suitability modelling: Environmental predictors encompassed topography (altitude, ruggedness), climate (annual mean temperature, mean diurnal range, annual precipitation), habitat (land cover), and human disturbance (distance to nearest roads). Data sources: SRTM (altitude, ruggedness, solar radiation), WorldClim 2.0 (temperature, precipitation), FAO Global Land Cover, and OpenStreetMap (roads). All layers were processed in ArcMap 10.6.1, standardized to 100 m resolution ASCII rasters. Maxent 3.4.1 was used with 25% data for model testing; model performance evaluated by ROC AUC (>0.75 considered useful). Variable selection was based on pilot analyses; variable importance was assessed by Maxent jackknife. Connectivity analysis: The inverse of habitat suitability yielded a resistance surface for connectivity modelling. Euclidean geographic distances were computed with Geographic Distance Matrix Generator 1.2.3. Resistance distances and voltage maps were generated using Circuitscape 4.0 (circuit theory) to identify likely movement corridors and gene flow directions and to compute pairwise resistance distances among genetic sampling localities. Landscape genetics integration: Genetic distance (FST; Genepop 4.2) was compared to geographic and resistance distance matrices using Mantel tests in PASSAGE 2.0 to assess isolation by distance (IBD) and isolation by resistance (IBR). Ethics: Research was permitted by the Department of National Parks and Wildlife Conservation and the National Trust for Nature Conservation of Nepal.

- Sampling and genotyping: Of 268 collected samples, 128 were genetically confirmed as snow leopard (124 scat, 4 hair). Microsatellite loci could not be identified in 20 snow leopard samples (excluded). A total of 108 microsatellite samples remained; after quality filtering, 63 genotypes were obtained, corresponding to 22 individuals. All loci were free of large allele dropout and stuttering; two loci showed evidence of null alleles (estimated 20.27% and 14.45%) and elevated homozygosity. - Population structure: STRUCTURE supported K=3 clusters (by ΔK). In Sagarmatha, samples segregated into a northern (N-S) and south-western (SW-S) cluster; LM and UM were more admixed. - Genetic diversity (Table 1): Observed heterozygosity moderate (Ho ~0.54–0.61). Fixation indices indicated less differentiation between LM and UM, moderate differentiation in SW-S, and none in N-S; overall significant genetic differentiation across areas. - Habitat suitability model: Excellent predictive performance (AUC training = 0.974). Key predictors: altitude, annual mean temperature, annual precipitation, and distance to roads. Secondary contributors: land cover, solar radiation, mean diurnal temperature range. Excluding distance to roads produced the largest loss in explanatory power. Ruggedness alone was not useful. - Spatial habitat pattern: The most suitable habitats form a relatively narrow belt between ~3500–4500 m elevation (threshold suitability index > 0.5), widest in central-western Nepal (e.g., Mustang) and narrow in the east, indicating higher fragmentation vulnerability there. Optimal patches are typically south-facing slopes near 4000 m, in cold, dry climates with shrubs, rocks, and open grasslands. - Connectivity and corridors: Voltage maps indicate narrow, bottleneck-prone corridors in Central and Eastern Nepal, versus multiple alternative routes in Western Nepal (LM, UM). Three potential corridors link LM and UM: C1 (Nar/Phu valleys north-west to Mustang), C2 (direct link crossing high ridge between Khatung Kang and Tilicho Lake), and C3 (southern link between Lamjung and Mustang). Two corridors (C4, C5) connect western (LM, UM) to eastern Sagarmatha, merging westward and traversing Gaurishankar Conservation Area, Langtang National Park, and Manaslu Conservation Area, with segments through Tibet’s Qomolangma National Nature Reserve due to high mountain barriers. - Landscape genetics: Mantel tests showed weak but significant IBD (r=0.22155; Z=264,714,021.85418; p<0.01) and weak significant IBR (r=0.15785; Z=455,729.95660; p<0.001). - Movement ecology context: Some LM samples were genetically close to SW-S (~300 km apart), consistent with dispersal via stepping-stone habitats (GCA, LNP, MCA) and long-distance movements documented by recent GPS data (up to 27–40 km/day; home ranges ~200–500 km²).

The study integrates habitat suitability, connectivity, and population genetics to address how landscape features and human disturbance shape snow leopard distribution and gene flow in Nepal. Climatic variables (annual mean temperature, altitude, precipitation) and proximity to roads emerged as primary determinants of suitable habitat, aligning with prior studies emphasizing climate. Contrary to expectations, ruggedness was not a strong predictor in this dataset, likely reflecting scale limitations of the ruggedness layer rather than biological irrelevance. The preferred elevational belt (3500–4500 m) corresponds to alpine zones above treeline where primary prey occur. Genetically, UM and LM show sufficient connectivity to mitigate drift, supported by multiple potential corridors (C1–C3). Corridor C2 is likely less used due to extreme elevation barriers, consistent with camera-trap evidence lacking shared individuals between UM and LM despite proximity. Bayesian clustering suggests some LM–SW-S affinity over ~300 km, implying occasional long-distance dispersal mediated by stepping-stone habitats across protected areas (GCA, LNP, MCA). Circuitscape also indicated southern and western access to Sagarmatha, and cluster patterns imply a possible northern route via the Tibetan plateau (Nangpa Pass), which was not explicitly modeled. Within Sagarmatha, N-S and SW-S appear genetically isolated, potentially reflecting recent recolonization from different sources and barrier effects of major rivers (Dudh Khosi, Imja) and heavy tourist traffic along the only feasible corridor. Overall, human presence, roads, and trekking routes likely constrain movements, with narrow, bottlenecked corridors particularly in the east increasing fragmentation risk. Protecting and functionally maintaining identified corridors is therefore critical to sustaining metapopulation dynamics and genetic diversity.

This work delineates Nepal’s suitable snow leopard habitats and identifies the main dispersal corridors (C1–C5) connecting them. Landscape features and human disturbance contribute to isolation, with narrow, vulnerable corridors especially in eastern regions. Protecting these natural corridors is essential to facilitate migration, maintain gene flow among declining populations, and preserve a viable metapopulation. Conservation of these linkages will also benefit other large carnivores (wolves, lynx, golden jackal, red fox) documented in the study areas. The findings provide a spatial blueprint for prioritizing corridor protection and management to ensure long-term persistence.

- Spatial data granularity: The ruggedness layer resolution may have been too coarse to capture micro-topographic features relevant to snow leopard habitat selection, potentially underestimating ruggedness effects. - Sampling and genotyping: Limited number of unique individuals (22) and exclusion of 20 samples lacking microsatellite profiles could constrain power to detect fine-scale structure. Two loci exhibited evidence of null alleles (up to ~20%), which can bias estimates. - Geographic scope: Analyses focused on Nepal; potential northern connectivity via the Tibetan plateau (e.g., Nangpa Pass) was inferred from clustering but not explicitly modeled. Stepping-stone populations in Qomolangma Nature Reserve were limited and isolated by high mountains, adding uncertainty to cross-border connectivity. - Corridor functionality: Identified corridors include narrow bottlenecks and areas of high human activity (roads, trekking routes). Their functional use by snow leopards requires empirical validation; camera-trap data did not document shared individuals between nearby sites (UM, LM). - Human disturbance data: Distance to roads was included, but other disturbance proxies (e.g., trail use intensity, settlements, tourism seasonality) were not explicitly modeled and may affect movement. - Data access: Occurrence locations cannot be publicly shared due to poaching risk, limiting external validation.