Spatiotemporal relationships, influencing factors and policy implications of coastal man–land system spatial resilience based on interpretable machine learning models: A case study of China’s southeastern coastal region

Huan Song, Zeyu Wang

Spatial resilience, as a projection of system resilience at the landscape scale, offers a novel spatial interpretation for analyzing man–land interactions in coastal zones. This study builds an evaluation system from “element-landscape-system” levels, based on the conceptual framework of spatial resilience in coastal man–land systems. It examines the spatiotemporal evolutionary features of spatial resilience at various scales using multisource spatiotemporal big data. Interpretable machine learning techniques were used to construct SHapley Additive exPlanations (SHAP) models and investigate influencing factors. Results yielded three main findings: First, spatial resilience of man–land systems in the southeastern coastal region exhibits significant spatiotemporal heterogeneity with a slow optimization trend, forming a multi-scale spatial pattern characterized by “higher resilience in the element layer in the north than in the south, higher resilience in the landscape layer along the coast than in bays, and stronger resilience in the system layer in the north than in the south.” Second, habitat quality, landscape connectivity index, and marine aquaculture area were the core factors influencing spatial resilience at the element, landscape, and system levels, contributing 35.17%, 29.24%, and 28.87%, respectively. Third, the variables with higher contribution rankings exhibited threshold effects on spatial resilience at different scales. To promote integrated land–sea management and optimize territorial spatial layout, this study, based on the advantages of regional resource endowments, formulates differentiated zonal management strategies and puts forward targeted policy recommendations.