Understanding soil moisture dynamics and key drivers for single and mixed cover cropping systems using machine learning

Abstract

Soil moisture plays a vital role in crop productivity, water resource management, and long-term agricultural sustainability. Cover crops are promoted to improve soil health and agricultural sustainability, but their effects on water use and availability in the semi-arid western US remain elusive. This study aimed to understand soil moisture dynamics and the factors influencing them in single-species and mixtures of cover crops integrated into a winter wheat production system. Volumetric water content (VWC), weather variables, and soil temperature were monitored for two years under pea ( Pisum sativum L.), oat ( Avena sativa L.), and pea and oat (PO) mixtures as cover crops and a no-cover crop (NCC) control. An extreme gradient boosting machine learning model with SHapley Additive exPlanations (SHAP) identified key environmental and management drivers regulating soil moisture at the surface (5 cm) and subsurface (30 cm) soil layers. All cover crops had greater VWC at 5 cm but lower VWC at 30 cm than NCC, while the PO mixture increased precipitation use efficiency. Crop water productivity was highest in PO, which was significantly greater than in NCC and other treatments. The XGBoost model achieved greater accuracy (R 2 up to 0.73, RMSE < 0.01 m 3 m −3 , MAE < 0.01 m 3 m −3 ) at 30 cm across diverse cropping systems. At 5 cm, the prediction accuracy of the model was moderate, with RMSE and MAE close to 0.05 and R 2 ≈ 0.45, reflecting inherent noise in surface soil moisture due to the stochastic occurrence of rainfall and irrigation events. The SHAP analysis identified weekly total water input at 5 cm and soil temperature at 30 cm as key moisture drivers, with effects varying by cover crop species. These findings highlight the potential of cover crop mixtures to improve water capture and utilization and of agronomy-guided machine learning to capture complex soil-plant-atmospheric interactions and support agricultural water management in arid and semi-arid regions.

The authors' abstract, as published at the source. Agricultural Water Management, 2026 · DOI ↗