Precise roadkill environmental factors identification for sustainable planning and design through spatial and temporal modelling

Roads are major drivers of habitat fragmentation and amphibian mortality. Effective mitigation requires understanding species-specific, temporal, and multiscale environmental drivers, yet such integrated analyses are uncommon. Additionally, amphibian species in our region are understudied, and fine-scale roadkill drivers remain globally underexplored. This study aims to identify precise roadkill predictors for these species by integrating these approaches and to develop a framework for translating the findings into context-specific ecological planning and design. We combined standardized surveys and citizen science roadkill data for five anuran species in a Taiwanese biodiversity hotspot and analyzed them at three levels: multispecies, single-species, and single-species temporal subsets. We employed a dual Random Forest and Logistic Regression consensus approach, assessing a comprehensive set of predictors, including fine-scale roadside features, across multiple spatial scales (25–250 m) with an innovative uphill–downhill buffer design. We identified 24 robust consensus predictors, with habitat-related (e.g., water proximity, forest cover) and road-related fine-scale factors (e.g., streetlights, retaining walls, drainage channels) being equally important. Species-specific and temporal models (AUC up to 0.851) significantly outperformed the multispecies model, showing that predictor relevance and the scale of effect shift seasonally at inter- and intra-specific levels, particularly during breeding migrations. Our findings underscore that generalized approaches are insufficient for effective roadkill mitigation and highlight the need for context-specific, evidence-based ecological planning and design. The study provides a transferable framework that links ecological traits and predictor relevance to actionable, multi-stage planning, design, and operational management guidelines. This approach shows how integrating ecological precision with planning relevance can enhance biodiversity conservation for sustainable development.