Exploring the differential responses of ecosystem services to urbanization: A perspective from the metropolitan area as a whole and internal service bundles

• Explores differential ecosystem services (ES) responses to urbanization across multiple scales. • Uses FCM-SOM to analyze ES clusters and their responses at the metropolitan scale. • Applies SHAP for attribution analysis of urbanization’s impact on ES. • Visualizes non-linear urbanization-ES relationships through partial dependence plots. Rapid economic growth and urban expansion have significantly disrupted the original structure and functions of ecosystems. Achieving a balanced development between the economy and ecology has become a critical challenge. Current research on urbanization and ecosystem services often fails to differentiate between urban agglomerations and rarely examines the nonlinear interactions between them. This study employs ecological models, machine learning techniques, and post-hoc interpretive models to comprehensively analyze the spatiotemporal responses of ecosystem services to urbanization. First, the carbon storage, water yield, habitat quality, soil conservation, food production, cultural landscape, and urbanization levels of typical urban clusters from 2000 to 2020 were assessed using the InVEST model, the RUSLE model, and the Entropy Weight Method, revealing their spatiotemporal evolution characteristics. Second, the XGBoost-SHAP model is used to quantitatively evaluate the effects of urbanization on ecosystem services, highlighting the nonlinear relationships through partial dependence plots. Finally, the Self-Organizing Map – Fuzzy C-Means (SOM-FCM) model is employed to identify ecosystem service bundles and explore the similarities and differences among clusters within urban agglomerations. The findings indicate that: (1) From 2000 to 2020, habitat quality, carbon storage, and cultural landscape in typical urban clusters exhibited an overall declining trend, while water yield showed fluctuations. Soil conservation increased by 5%-14% from 2000 to 2010, but decreased by 1%-3% from 2010 to 2020. Meanwhile, the comprehensive level of urbanization gradually rose. (2) The impact of CUL on HQ is significant across all three major urban clusters, while the impact on SC is the smallest. Among them, the impact of CUL in the Yangtze River Delta on various ES is the lowest. (3) There are variations in the types and quantities of ecosystem service clusters across different urban agglomerations, with the overall effect of urbanization on ecosystem services shifting from negative to positive. The urbanization impact on each cluster exhibits distinct hierarchical features, with relatively minor effects on soil retention. This study provides a new theoretical foundation and decision-making support for promoting coordinated development between urbanization and ecosystem services, offering significant policy implications and practical value.