Spatiotemporal evolutionary patterns and influencing factors of urban carbon emissions in China

• Used the GBRT model for constructing urban carbon emissions dataset (1992–2021). • Proposed the GASE model for ranking node importance in carbon emissions networks. • Explored the spatial heterogeneity and influence factors of urban carbon emissions. Exploring the spatial and temporal evolution of urban carbon emissions in China and revealing their influencing factors is significant for formulating differentiated carbon emission reduction strategies and further promoting urban sustainability. This research initially used the Gradient Boost Regression Tree (GBRT) model to construct the urban carbon emission dataset from 1992 to 2021. On this basis, local Moran’s index, Social Network Analysis (SNA), and the Multiscale Geographically Weighted Regression (MGWR) model were employed to analyze spatial–temporal evolution patterns and influencing factors of urban carbon emissions. Additionally, a novel method ranking the importance of nodes based on the Global Average Structural Entropy (GASE) model was proposed to evaluate the influence of nodes in the urban carbon emissions network. The results show that the GBRT model exhibited stronger generalization ability and higher accuracy in estimating urban carbon emissions compared to other regression models. Urban carbon emissions had a significant positive spatial correlation, and the level of spatial aggregation underwent an initial increase, followed by a decrease, and finally, it reached a stable state. Furthermore, the proportion of the ring length (the count of edges within the ring structure) of 2 increased as the density and stability of urban carbon emissions network structures increased. Moreover, the newly proposed GASE model demonstrates superior discrimination and stability in the ranking of node importance. The nodes with high influence in the carbon emissions network were primarily distributed in the Yangtze River Delta (YRD), Pearl River Delta (PRD), Beijing-Tianjin-Hebei (BTH), and some provincial capitals. Lastly, Normalized Difference Vegetation Index (NDVI), ratio of Ecological Functional Area (EF), ratio of Urban and Artificial Facilities Functional Area (UAF), and NightTime Light (NTL) progressively enhanced the promotion of urban carbon emissions, while Relief Degree of Land Surface (RDLS) contributed to a gradual reduction in the inhibition of urban carbon emissions. In contrast, the Industrial Structure (IS) variable and Temperature (Temp) exerted an increasing inhibitory effect on urban carbon emissions, and there was significant spatial heterogeneity among all urban carbon emissions influencing factors.