Objective The aim was to explore the application of the improved vertical drought index (MPDI) in extracting time series features to enhance the accuracy of digital mapping of soil texture in plain areas, in order to provide new insights for the digital mapping of soil texture in plain regions.

Methods Taking Fengqiu County in Henan Province as the research area, the combining MPDI time series feature variables were constructed based on the Land Surface Dynamic Feedback (LSDF) method with conventional variables such as terrain factors, climate factors, and vegetation factors as covariates. The ordinary kriging, support vector machine, random forest, and convolutional neural network models were used to perform digital mapping of soil texture (percentages of clay, silt, and sand), and the SHAP method was applied to interpret variable importance.

Results The research results showed that the prediction accuracy from support vector machine, random forest, and convolutional neural network models were all improved after introducing MPDI time series features. Among them, the convolutional neural network model achieved the best prediction results, with the lowest mean absolute error (MAE), root mean square error (RMSE), and relative root mean square error (RRMSE) for clay, silt, and sand content, which were 4.937, 6.344, 0.296 in MAE; 5.363, 6.677, 0.266 In RMSE; and 10.007, 12.215, 0.228 in RRMSE, respectively. The concordance correlation coefficient (CCC) and coefficient of determination (R²) were also the highest, showing 0.343, 0.227 in clay; 0.268, 0.143 in silt; and 0.312, 0.189, in sand, respectively. At the same time, the SHAP importance ranking results indicated that MPDI time series feature variables played an important role in model prediction and were key variables for soil texture prediction.

Conclusion The introduction of MPDI time series features reduces prediction errors to varying degrees across different models, improves prediction accuracy, which will propose a new environmental variable, and provide new insights for digital mapping of soil texture in plain areas.