Groundwater potentiality
mapping using ensemble machine learning algorithms for sustainable groundwater;
case of Ouled Bousbaa area (Morocco)
Hanane Toudamrini, Ahmed Algouti, Abdellah Algouti and Akram Elghouat
Disaster Advances; Vol. 18(12); 80-90;
doi: https://doi.org/10.25303/1812da080090; (2025)
Abstract
Groundwater recharge is crucial for managing freshwater resources and has become
a global issue due to climatic changes, particularly in arid and semiarid areas.
This study uses machine-learning algorithms (MLA) to facilitate groundwater potentiality
mapping (GWPM) via spatial modeling. For high precision, Extreme Gradient Boosting
(XGB) and Random Forest (RF) have been tested for GWPM. For this reason, a database
of springs and well inventories have been prepared and randomly divided into 75%
for training and 25% for model validation. GWPM is also statistically linked to
various relevant factors conditioning groundwater recharge including LS factor,
elevation, MRVBF, curvature, NDVI, NDWI, TWI, drainage density, distance to the
river, rainfall, permeability and fault density.
Validating GWP models uses the receiver operating characteristic curve (ROC-AUC).
The results show that RF (AUC=0,995) and XGB (AUC=0,990) are included in excellent
class based on the ROC curve method. Furthermore, GWPMs are efficient techniques
for sustainable groundwater resource management.
