Machine Learning Models Evaluation for Predicting Dam Filling Rates: A Case Study of AL Massira Dam, Morocco

Author : Imane Zine Elabidine, Anas Bahi, Ahmed Akhssas, Rachid Sebbari, Driss Bari, Tarik Chafiq

Abstract :Dams are key infrastructures in regulating water resources through their flood control, water supply, and sustainable management solutions. The study proposes a structured approach for selecting the best machine learning models that can be used in the prediction of filling rates in dams using daily hydro-meteorological data. It targets the AL Massira Dam in Morocco’s Oum Er-Rbia basin and uses various machine learning techniques such as Auto-Regressive Integrated Moving Average (ARIMA), Support Vector Machine (SVM), eXtreme Gradient Boosting (XGBoost), and Multilayer Perceptron (MLP). The methodology involves two distinct modeling approaches: (1) time series forecasting using ARIMA for one-day-ahead predictions with lagged data and (2) multivariate modeling integrating variables such as rainfall, temperature, soil moisture, and potential evapotranspiration for seven-day-ahead projections. Using a meticulously preprocessed dataset from 2018 to 2022, preliminary findings revealed that Support Vector Regression (SVR) and Multilayer Perceptron (MLP) models delivered the most robust multi-day predictions. The SVR model showed the best performance for AL Massira, with a root mean square error of 0.90, mean absolute error of 0.51, and a correlation coefficient of 0.84. The MLP model also realized relatively robust predictions, with an RMSE of 0.95, MAE of 0.71, and a correlation coefficient of 0.83. These will be useful for dynamic water resource management and adaptive dam operations, especially in view of changing climatic conditions

Keywords :Dam filling rates, Machine learning, Time series forecasting, Hydrological modeling, Weather data analysis, Adaptive water management.

Conference Name :International Conference on Geographic Information System Techniques and Modeling (ICGISTM -25)

Conference Place Tokyo, Japan

Conference Date 4th Jul 2025