Spatiotemporal evolution of ecosystem health and its nonlinear responses to driving factors in the Hangzhou Bay Area

Abstract

Under the combined effects of global climate change and rapid urbanization, the ecosystem health (EH) of coastal urban agglomerations faces significant pressure. Focusing on the Hangzhou Bay Area, this research integrated multi-source remote sensing data to construct an EH evaluation system based on the Vigor-Organization-Resilience-Services (VORS) framework. Using a 1-km grid as the basic analytical unit, this study analyzed the spatiotemporal evolution characteristics of EH from 2000 to 2020 and utilized the XGBoost-SHAP model to identify key driving factors and nonlinear response mechanisms. The results indicate the following: (1) the EH of the study area exhibited a spatial differentiation pattern characterized by being superior in the southwest and inferior in the northeast and coastal regions. Despite an overall improving trend, clusters of low health levels still persist in Jiaxing and certain coastal areas along Hangzhou Bay. (2) DEM and NDVI were the dominant factors influencing EH. The impacts of socioeconomic and climatic variables (population density, precipitation, nighttime lights, and temperature) were relatively weaker, yet all demonstrated distinct nonlinear responses and threshold characteristics. These findings validate the efficacy of explainable machine learning in spatial ecological modeling, offering a scientific foundation for ecological conservation, spatial management, and differentiated governance in the Hangzhou Bay Area.