Understanding the organic carbon content of forest soil will aid in studying the spatial distribution pattern of regional soil organic carbon (SOC) storage. Monitoring and researching forest SOC content is a crucial task that usually involves outdoor sampling and indoor experiments, which takes up much time. To improving its work efficiency, estimation models for topsoil organic carbon content are established. Correlation analysis was employed to evaluate the impact of factors (including elevation, slope, slope orientation, curvature, topographic wetness index, normalized difference vegetation index, enhanced vegetation index, and total nitrogen) on SOC content. Models for forest SOC content were constructed by machine learning algorithms using the above factors to enhance the efficiency of carbon storage estimation. Ultimately, the best model was used to generate a map of the SOC content. Research shows that: The Pearson correlation coefficient (r) between soil total nitrogen and SOC content is highest in both 0-5cm and 5-10cm soil layers (r=0.71, r=0.87). Optimal models for SOC content in the 0-5cm and 5-10cm soil layers are the random forest regression model and the boosted regression tree model, respectively. The coefficient of determination (R2) of the models are above 0.9. In the both soil layers, the performance of models constructed using regression tree algorithms is better than those constructed using linear regression, with the former having a greater R2 than the latter. Specifically, the R2 of the 0-5cm soil layer are 0.998 and 0.789, and the R2 of the 5-10cm soil layer are 0.997 and 0.996.