Evaluation and Analysis of Urban Ecosystem Health with an Optimized Machine-Learning Model to Promote Ecological Differentiated Management in Metropolitan Areas

Ecosystem health assessment is a prerequisite for understanding the dynamics of urban ecosystems. However, there is a lack of response analysis of urban ecosystem health before and after considering the material flows of key ecosystem services. Therefore, this study developed an urban ecosystem health assessment framework (VORMES) that integrates ecosystem services related to material flows and proposed the support vector regression redefined by northern goshawk optimization (NGO-SVR) to improve the evaluation accuracy. Then, taking the Nanjing metropolitan area as an example, the spatiotemporal evolution of the ecosystem health was assessed. The comparative framework variation index (CFVI) quantifies the differences between the VORMES and traditional frameworks, with spatial agglomeration and hotspot distribution used to explore their underlying causes. The results showed that the ecosystem health exhibited a multilayered composite core-edge bipolar diffusion pattern spatially. Compared to traditional frameworks, the assessment result of the VORMES framework was lower in general, and the CFVI had a significant spatial aggregation pattern. The hotspot analysis found that cold spot areas were concentrated in north-central areas of low total ecosystem service provision, while hotspot areas were concentrated on east-central areas of balanced ecosystem service. The external supply and internal equilibrium of ecosystem services are key factors driving this difference. Moreover, the SVR and the support vector regression with Bayesian optimization (BO-SVR) were chosen for the algorithm comparison test. The performance of NGO-SVR has an obvious advancement relative to SVR. The fitting accuracy was ranked as NGO-SVR > BO-SVR > SVR. Finally, based on the assessment results, differentiated management strategies for the one belt and three zones are proposed to safeguard the sustainable development of metropolitan ecosystems. This study provides a powerful tool for guiding metropolitan ecosystem restoration, offering valuable insights for metropolitan areas that face similar ecological pressures within and beyond China.