ObjectiveAs the largest Carbon pool in the terrestrial ecosystem, Soil Organic Carbon(SOC) plays an important role in Soil quality and crop yield. Accurate prediction of the spatial distribution of SOC on cropland is essential for the development of agricultural management measures. In the framework of Digital Soil Mapping(DSM), an important method to improve the precision of SOC spatial prediction is to select an effective environmental covariate. In previous studies, the mean values of remote sensing indices and climate variables for a certain period or point in time were usually used as input variables, while temporal characteristics and events were rarely used for SOC prediction. Therefore, in order to reduce the impact of the lack of part of physical information and climate characteristics, phenological variables and extreme climate variables were added in this study. The response characteristics to the spatial variability of SOC of cultivated land and the feasibility of predicting the spatial distribution of SOC were discussed.MethodThe research area of this paper is Shanggao County, Jiangxi province. A random forest model was used, in which remote sensing data, DEM-derived variables, phenological parameters and climatic characteristics were selected as environmental covariates, and the model results were corrected for residuals using Ordinary Kriging(OK). The prediction effect and prediction accuracy of the model under different types of variable combinations were compared.ResultThe results show that chronological variables, phenological variables, and extreme climate variables can improve the prediction performance of the model, and the residual error as an error item can further improve the accuracy of the model. The combination of chronological variables, phenological variables, extreme climate variables, topographic variables, and residuals had the highest prediction accuracy, improving R², MAE, and RMSE by 90.00%, 58.95%, and 57.14% compared to the combination of topographic variables, remote sensing variables, and climate variables. The analysis of variable contribution rates shows that SU, a3 and TXx were important variables affecting the distribution of cultivated land SOC in the study area.ConclusionPhenological variables and extreme climate variables have good application prospects. In the future, it is necessary to verify the validity of extreme climate variables as environmental variables in predicting soil properties under different land use and large-scale study areas.