With increasing loss and fragmentation of habitats driving the emerging global extinction crisis, paired with limited resources for conservation, there is an immense need to identify and prioritize the most important areas for conservation actions. The goal of this study was to measure, map and rank core areas and corridors for mainland clouded leopard (a forest indicator species) across its entire range in Southeast Asia. We used an empirically based landscape resistance model developed from range‐wide camera survey data, cumulative resistant kernel analysis to define core areas and least‐cost network analysis to identify corridors for long‐distance dispersal. We then ranked core areas based on their strength and size, and corridors based on their strength and the strength of core areas they connect. We found that the most important core areas and corridors are concentrated in Southeast Asia, largely in Myanmar, Laos and Malaysia. Myanmar contains nearly the entirety of the first and third highest ranked core areas, as well as the most important network of corridors in SE Asia. Almost the entire territory of Laos constitutes one large potential core area, ranked as the second most important across the clouded leopard’s range. A large number (22) of very small (<8000 km²) and fairly isolated core areas are in China. Only 24% of clouded leopard core areas and 17% of corridors are protected. This is the first example of using empirical models to prioritize conservation actions across the full range of a large carnivore. Our analysis identifies the location, size and connectivity of the most important remaining habitats of the clouded leopard across its range, which could provide quantitative guidance in the efforts to maximize the efficacy of regional conservation initiatives to conserve this species and the ecosystems it inhabits.

Context: Understanding the environmental and anthropogenic factors influencing habitat selection of multiple species is a foundation for quantifying human impacts on biodiversity and developing effective conservation measures.Objectives: To determine the effect of multiple scales of environmental/topographic and anthropogenic variables and landscape patterns on habitat suitability of terrestrial mammals in Bhutan, assess the effectiveness of the current protected area network, identify areas of high species richness outside of the existing protected area, and evaluate the potential effectiveness of indicator and umbrella species for conservation planning.Methods: We modelled multi-scale habitat selection of sixteen species of terrestrial mammals across Bhutan using data from a nation-wide camera trap survey. We used the predicted species distribution maps to assess the multi-species conservation effectiveness of the existing protected area network. We performed simulations to identify high priority areas for multiple species based on their habitat suitability, proximity to existing protected areas and overall connectivity within the predicted distribution of species. We used correlation analysis among predicted occurrence maps and multivariate cluster analysis to identify potential indicator species. We evaluated the potential utility of each species as umbrella species by assessing how well optimal protected areas for that species would protect suitable habitat for all 16 species simultaneously.Results: Protected areas and forest cover were strongly associated with habitat use of most modelled species. Additionally, topographical features, like terrain roughness and slope position, contributed to habitat selection of multiple species, but often in different ways. Environmental and topographical variables were mostly selected at medium to broad scales. Anthropogenic variables (agriculture and built-up areas) were negatively associated with habitat suitability of most species at both fine and broad scales. Conservation effectiveness assessment of existing protected areas found protected areas in south-central Bhutan have high effectiveness in terms of both mean and total richness protected. Similarly, biological corridors in the south-central region offered high mean richness protection. Our simulation of optimal areas for additional protection found areas abutting protected areas in southern Bhutan offered high relative species richness protection. Our umbrella species analysis found muntjac, wild pig, serow, sambar and Asian golden cat are the most effective umbrella species for broader biodiversity protection. Our indicator species analysis found tiger, gaur, dhole, clouded leopard, Asian black bear and common leopard as effective indicator species.Conclusions: This study highlights the need to protect optimally located species-rich areas outside the current protected areas. This kind of multi-species habitat assessment provides important information to optimize future conservation and development plans at national and regional scales.

Context: After decades of political and economic isolation, Myanmar is now the focus of large international investments, particularly from China, which raises questions of how to balance national development with safeguarding the Indo-Burma biodiversity hotspot. Objective: To evaluate the impact of five major developments in Myanmar on forest ecosystems, using clouded leopard as a focal and umbrella species for wider biodiversity conservation. Methods: Based on an empirical habitat relationships model, we identified core areas and corridors in Myanmar, and compared them across the development scenarios. We simulated population dynamics and genetic diversity in each scenario using an individual-based, spatially explicit cost-distance population genetics model. Results: The predicted current clouded leopard population may be larger than the current carrying capacity of the landscape, raising the possibility that the species’ population has not yet equilibrated with recent habitat loss and degradation. All the developments combined resulted in 36% decrease in landscape connectivity and 29% decrease in simulated clouded leopard population size, including substantial reduction in genetic diversity. Each development was predicted to have a negative effect; however, emerging economic zones had disproportionally large impacts (− 24% in connectivity and − 25% in population size), resulting in fragmentation of the largest core areas. Similarly, the Indian Highway and Silk Road caused fragmentation of the largest core habitats, and the Pipeline Railroad significantly decreased connectivity in the main stronghold for clouded leopards. Conclusions: Spatially-explicit assessments like the one presented here provide quantitative evaluation on development impacts and help optimize the trade-offs between development and conservation. The rapid and increasing development of Myanmar and surrounding Southeast Asian nations pose an enormous threat to the biodiversity of the region. Optimizing the trade-off between development goals and conservation is essential to minimize the effects of rapid land use change on biodiversity.

Separation of savanna land cover components is challenging due to the high heterogeneity of this landscape and spectral similarity of compositionally different vegetation types. In this study, we tested the usability of very high spatial and spectral resolution WorldView-2 (WV-2) imagery to classify land cover components of African savanna in wet and dry season. We compared the performance of Object-Based Image Analysis (OBIA) and pixel-based approach with several algorithms: k-nearest neighbor (k-NN), maximum likelihood (ML), random forests (RF), classification and regression trees (CART) and support vector machines (SVM). Results showed that classifications of WV-2 imagery produce high accuracy results (>77%) regardless of the applied classification approach. However, OBIA had a significantly higher accuracy for almost every classifier with the highest overall accuracy score of 93%. Amongst tested classifiers, SVM and RF provided highest accuracies. Overall classifications of the wet season image provided better results with 93% for RF. However, considering woody leaf-off conditions, the dry season classification also performed well with overall accuracy of 83% (SVM) and high producer accuracy for the tree cover (91%). Our findings demonstrate the potential of imagery like WorldView-2 with OBIA and advanced supervised machine-learning algorithms in seasonal fine-scale land cover classification of African savanna.