Enhancing Groundwater Level Predictions by Integrating Precise Agricultural Water Consumption with Machine Learning Algorithms

In intensely irrigated regions, groundwater levels are strongly influenced by crop water consumption and human activities. Although machine learning methods have proven effective for groundwater prediction, few studies have explicitly incorporated irrigation water use as a key predictor. In this study, we developed a groundwater prediction framework that integrates precise agricultural water consumption using machine learning techniques, with Shijiazhuang in China as the study area. Crop water consumption was simulated by coupling multi-source data with the AquaCrop model. A Random Forest-based groundwater prediction framework was then constructed to explore the effects of irrigation water consumption on groundwater level changes from 2018 to 2024. Sentinel-2A remote sensing imagery was employed to extract regional crop cultivation patterns, achieving an Overall Accuracy (OA) greater than 0.94 and a Kappa coefficient exceeding 0.89. Subsequently, AquaCrop model was applied to simulate crop yields and water consumption, with simulated yields showing strong agreement with official statistics, thus enabling the generation of a long-term time series of crop water consumption. Building on these results, the simulated irrigation water consumption was incorporated into the groundwater level prediction model. The performance of models using different predictor combinations was compared. Results show that including water consumption significantly enhanced the predictive accuracy, reducing root mean square error (RMSE) by 6.34%, 4.17%, and 3.97% in the three groups, and reducing the average error by 9.09%, 19.33% and 21.37% respectively. Spatially, model errors were also notably reduced. Overall, this study demonstrates that integrating irrigation water consumption into a groundwater model can effectively quantify the response of groundwater levels to climatic and anthropogenic factors, providing a scientific basis for groundwater resource management and ecological restoration.