High-Resolution Mapping of Cropland Soil Organic Carbon in Northern China

Mapping the high-precision spatiotemporal dynamics of soil organic carbon (SOC) in croplands is crucial for enhancing soil fertility and carbon sequestration and ensuring food security. We conducted field surveys and collected 1121 soil samples from cropland in Changzhi, northern China, in 2010 and 2020. Random Forest (RF) models combined with 19 environmental covariates were used to map the topsoil (0–20 cm) SOC in 2010 and 2020, and uncertainty maps were used to calculate the dynamic changes in cropland SOC between 2010 and 2020. Finally, RF and Structural Equation Modeling (SEM) were employed to explore the effects of climate, vegetation, topography, soil properties, and agricultural management on SOC variation in croplands. Compared to the prediction model using only natural variables (RF_C), the model incorporating agricultural management (RF_A) significantly improved the simulation accuracy of SOC. The coefficient of determination (R2) increased from 0.77 to 0.85, while the Root Mean Square Error (RMSE) decreased from 1.74 to 1.53 g kg−1, and the Mean Absolute Error (MAE) was reduced from 1.10 to 0.94 g kg−1. The uncertainty in our predictions was low, with an average value of only 0.39–0.66 g kg−1. From 2010 to 2020, SOC in the Changzhi croplands exhibited an overall increasing trend, with an average increase of 1.57 g kg−1. Climate change, agricultural management, and soil properties strongly influence SOC variation. Mean annual precipitation (MAP), drainage condition (DC), and net primary productivity (NPP) were the primary drivers of SOC variability. Our findings highlight the effectiveness of agricultural management for predicting SOC in croplands. Overall, the study confirms that improved agricultural management has great potential to increase soil carbon stocks, which may contribute to sustainable agricultural development.