Groundwater potential mapping using conventional and machine learning approaches with multi-indicator validation of Aflaj systems in Oman
Year: 2026
Type: article
Abstract: Study region The Alaqida Basin in northern Oman is an arid mountainous watershed where groundwater is a critical water resource. Groundwater occurrence varies spatially with rainfall, geology, topography, drainage, structural features, soil properties, and land use, highlighting the need for reliable groundwater-potential assessment. Study focus This study develops a hybrid groundwater-potential zonation framework integrating the Analytic Hierarchy Process (AHP) and Extreme Gradient Boosting (XGBoost). Seven groundwater-conditioning factors—rainfall, geology, soil, slope, drainage density, lineament density, and land use/land cover—were integrated in GIS to generate an AHP-based groundwater-potential map. The resulting classes were used as reference labels to train XGBoost using 40,083 pixel-based samples. Model performance was evaluated using 5-fold spatial block cross-validation and conventional and ordinal classification metrics. New hydrological insights for the region High and moderate groundwater-potential zones occupy 50.7% of the basin, while low and very low zones account for 45.2% and 4.1%, respectively. XGBoost reproduced the AHP-derived classes with cross-validated accuracy of 0.929, macro F1-score of 0.878, Cohen’s kappa of 0.888, quadratic weighted kappa of 0.941, and Matthews correlation coefficient of 0.889. Low-confidence predictions occupy 5.5% of the basin. Favorable zones coincide with major wadis, coarse alluvial deposits, and structurally enhanced areas, while correspondence with wells, recharge dams, and traditional Aflaj systems provides independent support for the hydrogeological plausibility of the mapped groundwater-potential pattern.
Language: English
Subfield: Environmental Engineering
Field: Environmental Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,600
OpenAlex ID: W7223214586