The research, led by Mehnaz Nazir and Tawqir Bashir of Sher-e-Kashmir University of Agricultural Sciences and Technology of Kashmir, together with colleagues at Lovely Professional University and the Zoological Survey of India, combined two years of painstaking fieldwork with sophisticated machine-learning models and landscape connectivity analysis. Between April 2021 and March 2023, the team walked hundreds of kilometres of mountain trails across ten protected areas, recording direct sightings and collecting scats, the latter verified to species level using DNA extraction with a Qiagen Stool Kit. The effort yielded 90 presence records for brown bears, 430 for black bears, and 876 for leopards, a dataset that formed the foundation for one of the most detailed carnivore habitat assessments ever conducted in the region.

What makes the study technically remarkable is its treatment of scale. Habitat selection, the authors emphasise, is not a one-size-fits-all phenomenon; animals respond to environmental features at multiple spatial and temporal scales, from the canopy cover of a single forest stand to the climate of an entire mountain range. To capture this complexity, the researchers calculated the focal mean of 48 ecologically relevant predictor variables, spanning 30 climatic, 9 topographic, 4 land-use, and 5 human-disturbance factors, at eight different spatial scales for brown bears, ranging from 1 to 128 kilometres, a maximum chosen to approximate the species’ average dispersal distance of 108.3 kilometres. For black bears and leopards, they used 17 and 11 radii respectively, reflecting the known dispersal biology of each species.

The modelling framework itself was equally rigorous. The team employed Random Forest, a non-parametric ensemble learning method that builds thousands of decision trees through bootstrap aggregation, an approach particularly well suited to ecological datasets with many predictors and modest sample sizes. Because true absence data are nearly impossible to obtain for cryptic, wide-ranging animals, the researchers generated pseudo-absences, spatially rarefied to a minimum separation of 2 kilometres and excluded from within 2 kilometres of any known presence record, to avoid contaminating the absence class with potentially occupied habitat. Through a two-step optimisation involving t-tests across scales, correlation filtering, and a Model Improvement Ratio based on permuted variable importance, the final models retained 8 variables for leopards, 24 for black bears, and 19 for brown bears. The resulting models performed exceptionally well, with an out-of-bag error rate of just 0.03 and an area under the curve of 0.97.

The ecological findings are as striking as the methodology. For leopards, canopy height emerged as the single most influential variable, with the cats favouring semi-open forests with canopy heights of 10 to 12 metres, south-facing slopes, and moderate precipitation of around 1,000 millimetres annually. Notably, leopards showed a marked tolerance for human presence, remaining suitable even at human population densities of 2,000 to 2,500 people, a finding consistent with the species’ reputation as one of the world’s most adaptable large predators. Black bears, by contrast, keyed in on terrain roughness, selecting rugged mid-elevation slopes above 30 degrees at elevations of 2,000 to 3,000 metres, while avoiding areas with road densities exceeding 0.1 kilometres per square kilometre.

The brown bear told a different story altogether. For this high-altitude specialist, precipitation of the driest month proved to be the dominant driver of habitat selection, a reflection of the critical role moisture plays in sustaining the plant growth and biomass that underpin the bear’s omnivorous alpine diet. Brown bears preferred open meadows with sparse forest cover below 0.2 and low canopy heights of 5 to 8 metres, and they proved acutely sensitive to human disturbance: habitat suitability declined sharply once the human footprint index exceeded 25, and the species tolerated only very low human population densities below 500. In a warming, increasingly crowded Himalaya, that sensitivity is a warning sign, because behavioural or evolutionary adaptations to climate stress tend to unfold over long timescales and may not keep pace with rapid environmental change.

Translating habitat suitability into movement, the researchers converted their models into resistance surfaces using a negative exponential transformation, producing continuous values from 1, representing minimal resistance to movement, to 100, representing near-impermeable terrain. They then deployed the UNICOR connectivity simulator, generating 500 spatially weighted nodes per species and running both cumulative resistance kernel and factorial least-cost path analyses to identify core habitats and dispersal corridors. The results paint a sobering picture of how little of the carnivores’ functional landscape actually falls within formal protection.

For leopards, the total predicted connectivity area spans roughly 12,881 square kilometres, yet only 11.19 percent of it lies inside protected areas. Black bears fare worse: of their 25,678 square kilometres of core connectivity area, just 5.31 percent is formally protected. The brown bear, the most range-restricted and vulnerable of the three, has approximately 25,813 square kilometres of core connectivity area, of which a mere 6.28 percent enjoys protection. Across all species, high-connectivity zones cluster in the Greater Himalayan ranges around Ganderbal, Kupwara, and Bandipora, and in the Pir Panjal range around Shopian, Kulgam, and Baramulla, with the Zabarwan range emerging as a crucial shared corridor linking the two mountain systems through the outskirts of Srinagar.

Perhaps the most consequential finding concerns transboundary connectivity. Least-cost path modelling repeatedly identified corridors extending from Kashmir’s protected areas, including Dachigam National Park, Tral Wildlife Sanctuary, and Kazinag National Park, through largely unprotected landscapes toward Gilgit-Baltistan and northern Pakistan. These routes suggest that the survival of Himalayan carnivore populations may depend on linkages that cross one of the world’s most heavily militarised borders, a reality that demands international cooperation rather than unilateral action. The authors argue that many of the most ecologically significant core areas and corridors lie entirely outside the existing protected area network, in districts such as Kulgam, Budgam, Kupwara, Ganderbal, and Bandipora, and that these spaces warrant urgent inclusion in conservation planning.

The study’s implications extend well beyond Kashmir. The researchers advocate a shift from fortress-style protection toward community-based conservation that aligns corridor protection with local livelihoods, engaging villages, government agencies, and non-governmental organisations in shared stewardship of the landscapes carnivores must traverse. Their integrated habitat-and-connectivity framework, combining multi-scale Random Forest modelling with resistance kernel and least-cost path analysis, offers a scalable template for conserving wide-ranging species in other fragmented mountain systems worldwide. The authors also acknowledge limitations, notably the seasonally restricted brown bear dataset, which reflects only summer habitat use because the species’ high-altitude haunts remain snowbound and inaccessible for much of the year. Even so, the message is unambiguous: protected areas remain vital refugia, but on their own they are not enough. The future of the Himalaya’s great carnivores will be decided in the corridors between them, in the unprotected valleys and ridgelines where bears and leopards still move, and in the political will to keep those pathways open.

Subject of Research: Habitat selection and landscape connectivity of large carnivores in the Kashmir Himalaya and the effectiveness of protected areas

Article Title: Are protected areas enough? Habitat selection and connectivity for large carnivore conservation in the Himalaya

Article References: Are protected areas enough? Habitat selection and connectivity for large carnivore conservation in the Himalaya. (n.d.). https://doi.org/10.1007/s10113-026-02662-3

Image Credits: AI Generated

DOI: 10.1007/s10113-026-02662-3

Keywords: Himalaya, large carnivores, protected areas, habitat connectivity, Himalayan brown bear, Himalayan black bear, leopard, Random Forest, least-cost paths, wildlife corridors, Kashmir, conservation planning