In recent decades, carbon sequestration (CS) capacity in Northern China has changed significantly, but the main factors leading to the spatial heterogeneity of CS change were still unclear. In this study, we analyzed the spatio-temporal characteristics of CS change from 2000 to 2020 in Northern China, and further used correlation analysis and multiple linear regression model (MLR) to identify the main factors leading to the spatial differentiation of CS change. Furthermore, we adopted random forest model (RF) to compare the importance of these factors and geographically weighted regression (GWR) to spatially illustrate the heterogeneity of their influence on CS change. Our results showed that the most obvious increases of CS were concentrated in the Loess Plateau, Yanshan Mountains and Taihang Mountains. After eliminating redundant and low-impact variables, we screened six factors that can well predict the spatial differentiation of CS change in Northern China. Based on the selected predictors, the MLR could explain 62.9% of the spatial variation of CS change, while the RF and GWR could explain 82.2% and 65.7% under the same predictors, respectively. Meanwhile, the spatial feature of each predictor's influence on CS change showed obvious differences. Among all predictors, afforestation was the most important factor leading to the spatial variation of CS change, and aridity index had the largest contribution among the climatic factors. In addition, we found that aridity index and potential evapotranspiration could explain better than commonly used precipitation and temperature. This study deepens the understanding of the spatial heterogeneity of CS change in Northern China and provides further suggestions for improving regional CS capacity.