Integrating driver analysis and predictive modeling of the ecosystem health–human activity relationship for an ecological zoning strategy in the Guangdong–Hong Kong–Macao Greater Bay Area

Balancing the human-nature relationship is a critical challenge for regional management. However, research on the spatial variation mechanisms and development trends of this relationship is limited. Extending beyond traditional single-indicator studies, this study identified the spatial relationship between the Ecosystem Health Index (EHI) and the Human Activity Intensity Index (HAII) in the Guangdong–Hong Kong–Macao Greater Bay Area (GBA) from 2000 to 2020 using the bivariate spatial autocorrelation method. The XGBoost-SHAP model was employed to examine the driving factors of the spatial relationships, and the PLUS model was used to predict trends in spatial relationship changes. The main findings include: (1) Urban core areas in the GBA exhibited characteristics of low EHI and high HAII, with these areas expanding significantly (4.88 %) during the two decades; (2) Anthropogenic factors dominated the variation in EHI–HAII relationships, with the influence of factors on relationship types varying according to urbanization status; (3) areas characterized by low EHI and high HAII are projected to continue expanding outward, and transitional zones between the core and periphery of the GBA may constitute key areas for environmental management. This study utilized prediction results as the basis for ecological zoning and driving factor analysis results as the basis for ecological management, proposing a novel ecological zoning strategy based on relationship analysis to support regional sustainable development. The recommendations primarily include strengthening urban green space construction while reducing human disturbance in high-quality forest areas, controlling population influx around degradation risk areas, and improving transportation accessibility in scattered ecological restoration regions. • A machine learning approach identified driving factors shaping human-nature interactions. • A novel prediction-based zoning method was developed to guide ecological planning. • The proposed framework supports fine-scale sustainable development strategies in rapidly urbanizing regions.