Integrating ecological resilience assessment and machine learning for ecological security patterns in Guangdong Province, China

Ecological security patterns (ESPs) provide an important framework for coordinating ecological conservation and urban development in rapidly urbanizing regions. However, conventional approaches to ecological source identification often emphasize current ecosystem functions while paying limited attention to ecological resilience under ongoing urbanization and environmental uncertainty. Meanwhile, the application of machine learning to resistance-surface construction in such contexts remains limited. To address these gaps, this study develops an integrated framework that combines ecological resilience assessment, machine learning, and circuit theory to construct an ESP for Guangdong Province, China. Ecological resilience was first assessed to identify potential ecological sources, while Light Gradient Boosting Machine (LightGBM) was then used to construct a resistance surface by capturing nonlinear relationships between human activities and ecological resistance. Circuit theory was subsequently applied to identify ecological corridors and key ecological nodes. The results show that ecological sources are unevenly distributed across Guangdong, with higher resilience concentrated in mountainous and forested areas. The LightGBM model achieved the highest test-set AUC among six candidate algorithms. The resulting resistance surface further reveals substantial spatial variation in ecological pressure, with intensive human activities concentrated around major urban agglomerations. Ecological corridors and pinch points identified through circuit theory highlight critical areas for maintaining regional connectivity. The findings demonstrate that integrating ecological resilience with machine learning can provide a more adaptive basis for ESP construction in rapidly urbanizing regions and offer practical support for targeted ecological conservation and connectivity management.