Spatiotemporal evolution of carbon balance and imbalance risk in the Yangtze River Delta: A study based on spatial...

Under the dual pressures of climate change and economic development, assessing changes in the carbon balance (CB) is critical for regional sustainable development. Existing studies have focused on the nonlinear effects of driving factors on CB, yet have overlooked the variations caused by spatial heterogeneity, leading to an inaccurate characterization of the relationships between driving factors and CB. This research analyzed the CB dynamics and spatiotemporal evolution of carbon imbalance risk (CR) in the Yangtze River Delta (YRD) from 2001 to 2024, identified the spatially nonlinear threshold effects of driving factors using the GWRF-SHAP-GAM (Geographically Weighted Random Forest, SHapley Additive exPlanations, Generalized Additive Model) framework, and integrated these effects into risk-oriented priority management. The results indicated that the study region remained in a persistent state of carbon imbalance with gradual deterioration, and the proportion of carbon-imbalanced areas rose from 11.3% to 46.6%. Significant "high-risk persistence" and "risk escalation" were observed in CR, and high-risk zones were hard to downgrade, while low-risk zones continuously transitioned toward higher risk levels. Pronounced spatial heterogeneity and nonlinear effects existed between driving factors and CB. Human activity factors showed simple negative associations with distinct single thresholds and served as the dominant contributors in the model, whereas natural environmental factors presented complex bimodal or trimodal patterns with multiple thresholds but relatively weak influences. Priority management strategies were proposed by integrating regional CR with thresholds of key influencing factors. The innovation of this study lies in establishing a comprehensive analytical framework that integrates GWRF, SHAP, and GAM to analyze the spatiotemporal evolution of CB and the spatially heterogeneous nonlinear thresholds of driving factors in the YRD. This framework provides a risk-oriented approach for understanding the statistical associations and supporting precise regulation of CB in urban agglomerations.