Integrating the XGBoost Model and GeoShapley Method to Reveal the Impacts of Carbon Balance Spatial Linkages on Urban Resilience

Although many studies have examined the relationship between carbon emissions and Urban Resilience (UR), regional Carbon Balance (CB) network interactions remain underexplored. Existing studies mainly apply interpretable machine learning methods, especially the SHAP framework. However, they often ignore the spatial dependence of regional data. To address this gap, this study adopts a spatial network perspective and a spatial machine learning approach to examine the relationship between CB spatial linkages and UR. By integrating carbon emission data, Net Primary Productivity data, and multi‐source urban indicators, an XGBoost GeoShapley framework is constructed to analyze the Yangtze River Delta (YRD) and assess how CB spatial connections influence UR. The main results are as follows. (1) The Carbon Balance Network (CBN) exhibits progressively strengthening connectivity over time, with intercity linkages increasing across the region. Shanghai, Suzhou, Wuxi, and Nanjing remain core nodes, while Hefei and Yangzhou gain influence, indicating a more diversified network structure. (2) Among CBN structural features, betweenness shows the strongest marginal contribution and jointly influences UR with socioeconomic factors. (3) CBN variables exert significant nonlinear effects on UR, and clear threshold patterns are observed. (4) Interaction results indicate that urbanization level and industrial structure upgrading significantly moderate CBN effects on UR. These findings provide implications for promoting coordinated high resilience and low carbon development in the YRD urban agglomeration.