Comparative Evaluation of Machine Learning Models for Hydrological Variable Prediction in Groundwater Management: Erfoud Radier Station in Morocco

¹ Laboratory of Engineering Sciences, National School of Applied Sciences, Ibn Tofaïl University, Kenitra, Morocco
² Moulay Ismail University of Meknes, Department of Computer Science, Meknes, Morocco
³ National School of Mines Rabat B.P 753, 14000 Agdal-Rabat, Morocco

^* Corresponding author: This email address is being protected from spambots. You need JavaScript enabled to view it.

Abstract

We consider it crucial to recover missing climate data for water resource management in semi-arid regions. In this work, we investigated five machine learning models (ANN, SVM, RF, DT, and KNN) to estimate missing mean temperature values from the station Erfoud Radier in the Guir-Ziz-Rheris (GZR) basin. Our models were evaluated using the MSE, RMSE, MAE, and R2 metrics. The most effective model is the SVM model, with the highest R2 =0.912 and the least error (MSE = 7.142; RMSE = 2.673; MAE = 1.842). The second best model is the ANN model, with the highest R2 (0.885) and a slightly lower error. The RF and KNN models performed poorly, and the DT model performed poorly. Our results confirm the effectiveness of machine learning in the reconstruction of climatic time series, and suggest interesting directions for modeling groundwater levels and water resources management.