Cross-scale hotspot transition and scale-dependent drivers of urban bird richness

Biodiversity conservation has become a major challenge in rapidly urbanizing regions, where environmental heterogeneity reshapes ecological patterns across spatial and seasonal dimensions. However, most studies have focused on identifying a single optimal scale while overlooking how biodiversity responses vary across scales. In this study, we developed an integrated framework combining multi-scale analysis, explainable machine learning, and seasonal simulations to investigate urban bird diversity patterns in Shenzhen, China. More than 215,000 citizen-science bird records and environmental variables were analyzed across spatial scales ranging from 300 m to 2000 m. The results revealed strong scale dependency in both model performance and environmental responses. Among all scales, 1000 m achieved the highest predictive accuracy and ecological interpretability, representing the optimal scale for modeling urban bird diversity. Cross-scale hotspot analysis identified stable, attenuated, and emergent hotspot transition patterns, indicating that biodiversity-supporting conditions undergo persistence, attenuation, and spatial reorganization during scaling processes. Environmental drivers exhibited hierarchical shifts across scales, transitioning from fine-scale habitat regulation to broader climatic, topographic, and anthropogenic control, accompanied by widespread nonlinear threshold responses. Seasonal simulations further revealed substantial temporal shifts in hotspot distributions and driver dominance associated with species phenology. Overall, this study provides a scale-adaptive framework for understanding urban biodiversity responses and offers practical implications for ecological monitoring, biodiversity conservation, and spatial planning in rapidly urbanizing landscapes.