Explainable Machine Learning Framework for Soil Moisture Stress Assessment and Irrigation Decision Support

Water management is one of the vital activities in agriculture, and soil moisture is one of the significant parameters in irrigation planning. The research presents an Explainable Machine Learning Framework for Soil Moisture Stress Assessment and Irrigation Decision Support, using data associated with Pune, Maharashtra. Historical data of soil-moisture, rainfall, and seasonal factors were used in the experiment to predict the soil moisture of the next day. Comparative analysis using Linear Regression, Decision Tree, and Random Forest was performed using chronological splitting, and all three models were assessed using MAE, RMSE, and R². Linear Regression algorithm applied on combined historical soil-moisture, rainfall-history, and seasonal factors recorded the best result with RMSE equals to 1.7977 and R² of 0.7819. The analysis using SHAP highlighted the importance of current soil moisture, whereas RMSI categorized the prediction results of the soil moisture for relative irrigation planning and decision-making support.