SDM- and Land Cover-Based Resistance Models for Omnidirectional Landscape Connectivity of Korean Water Deer

Background and objective: Habitat loss and fragmentation are primary drivers of biodiversity decline. Integrating species distribution models (SDMs) with landscape connectivity analysis provides a spatially explicit framework for identifying priority conservation areas. This study aimed to (1) develop an ensemble habitat suitability model for the Korean water deer (<i>Hydropotes inermis argyropus</i> Swinhoe, 1870), (2) assess landscape connectivity using circuit theory, and (3) identify priority restoration sites within the Geumgang River watershed, South Korea.Methods: We compiled 2,831 occurrence records and employed ten SDM algorithms with 11 environmental variables at 30 m resolution. Pseudo-absence points were generated using a disk strategy (minimum distance = 3 km). Model performance was evaluated via spatial block cross-validation (2 × 2 checkerboard). An ensemble approach with TSS-based model filtering (≥ 0.4) produced consensus habitat suitability maps. Landscape connectivity was analyzed using Omniscape to derive cumulative current flow, flow potential, and normalized current flow.Results: The ensemble models achieved calibration TSS of 0.607-0.608 and AUC of 0.891-0.906. Elevation and landscape heterogeneity were the most influential predictors. Validation with 18,464 independent roadkill records demonstrated that the SDM-based resistance surface consistently outperformed the expert-assigned LULC-based approach across all five metrics.Conclusion: The integrated framework identified priority restoration sites where targeted interventions could most effectively enhance connectivity and reduce wildlife-vehicle collisions.