The soil in China's subtropical forest ecosystems possesses a tremendous capacity carbon sequestration potential,making a substantial contribution to mitigating climate change. Consequently,precise assessment and management of the spatial distribution of forest soil organic carbon (SOC) are essential in combating global climate change. In this study,to explore the spatial distribution of SOC in subtropical forest land and its influencing factors,data were drawn from 186 permanent monitoring plots surveyed in 2017 at the Experimental Center of Tropical Forestry under the Chinese Academy of Forestry,which served as the primary research sites. Using a combination of the random forest model,the random forest with residual kriging model,and the SHapley Additive exPlanations (SHAP) interpretative method,the spatial distribution patterns of SOC content were analyzed. This analysis integrated measured vegetation data,terrain features,remote sensing indices,and climate variables as covariates,allowing identification of the primary factors influencing SOC variation. The results revealed that each factor impacted SOC content in distinct ways,with their modes of action and degrees of influence demonstrating certain regular patterns. The SOC content in the study area ranged from 4.13 to 34.80g/kg and showed significant correlations with climate variables,altitude,biomass,and other factors (P < 0.05). Notably,topography and climate variables were the main factors influencing the spatial distribution of SOC,collectively explaining 74.23% of the prediction results. Measured vegetation data,including aboveground biomass,belowground biomass,and herb cover,also had significant effects on SOC content,accounting for 25.77% of the explained variance. Simultaneously,SOC content exhibited a nonlinear increasing trend with the elevation,mean annual precipitation,and biomass,while showing a nonlinear decreasing trend with the temperature difference between mean warmest month temperature and mean coldest month temperature,hargreaves reference evaporation,and the Simpson index of shrubs. The overall spatial distribution trends of SOC predicted by the two methods were largely consistent,with higher SOC accumulation observed in areas with higher elevations,greater forest cover,and more precipitation. Compared to the random forest model,the random forest with residual kriging model showed better alignment with observed values in terms of standard deviation and coefficient of variation. This model offers higher predictive accuracy and interpretability by accounting for spatial autocorrelation and environmental correlations. In conclusion,this study provides a theoretical basis for understanding the spatial distribution of SOC content in forest land and its influencing factors. It offers valuable insights for formulating forest management strategies aimed at increasing soil carbon sequestration and reducing carbon emissions.