Advanced machine learning for predicting groundwater decline and drought in the Rabat–Salé–Kénitra region, Morocco

Advanced machine learning models, particularly random forest (RF), demonstrate exceptional accuracy in predicting groundwater storage variations in the Rabat–Salé–Kénitra region of Morocco, with a coefficient of determination (R²) of 0.981 and minimal prediction errors. This study highlights RF's superior performance over other models like ANN, GB, SVR, DT, and Prophet, which was integrated for handling seasonality in time-series data. The research emphasizes the urgent need to integrate these predictive models into groundwater management strategies to combat groundwater depletion caused by population growth, agricultural expansion, and climate change-induced droughts. These findings offer policymakers a robust toolkit to regulate water extraction, promote sustainable practices, and ensure the long-term viability of vital groundwater resources in this UNESCO World Heritage Site.

Elmotawakkil, A., & Enneya, N. (2024). Advanced machine learning for predicting groundwater decline and drought in the Rabat–Salé–Kénitra region, Morocco. Journal of Hydroinformatics, 26(11), 2980–3007. https://doi.org/10.2166/hydro.2024.328