Spatiotemporal evolution and influencing factors of residential building carbon emissions in Hot Summer and Cold Winter Climate Zone based on XGBoost-SHAP algorithm

This study systematically investigated the spatiotemporal evolution and driving factors of residential building carbon emissions in the Hot Summer and Cold Winter Climate Zone. We first utilized panel data from nine provinces and municipalities from 2007 to 2021 to examine the spatiotemporal evolution patterns of carbon emissions through methods including the Slope value, Theil index, and Standard Deviational Ellipse. Subsequently, we conducted an in-depth analysis of the differential driving mechanisms of carbon emissions using the XGBoost-SHAP framework optimized by Bayesian hyperparameters. The results indicated that total carbon emissions increased from 185.64 million tons in 2007 to 334.61 million tons in 2021, with growth rates exhibiting significant phase-specific characteristics. Spatially, the distribution exhibited notable upstream-downstream gradient, characterized by north-south agglomeration and east-west dispersion, with the contribution rate of intra-regional disparities consistently exceeding 90%. The relative influence of the dimensions ranked as follows: population scale > technological level > economic development > building attributes > policy support > climatic characteristics. The upstream, midstream, and downstream regions demonstrated differentiated development patterns of "energy-dominated," "population-dominated," and "energy-economy dual-driven" modes, with all influencing factors generally showing significant nonlinear effects and threshold characteristics. In response to regional differences, a "three-layer linkage" regional coordination mechanism was proposed, along with differentiated emission reduction strategies including "dual-track progression", "refined management", and "system integration".