Groundwater Potential Mapping in Northwestern Nigeria: A Comparison of the Analytic Hierarchy Process and Galileo-Enhanced CAT-XGB
Abstract
Global water scarcity intensifies the need for high-resolution groundwater assessment tools, yet traditional approaches struggle to reconcile qualitative expert weights with the extensive spatial coverage offered by remote sensing data. This study presents a comparative framework evaluating expert-driven Multi-Criteria Decision Analysis (a ten-layer Analytic Hierarchy Process (AHP)) against data-driven ensemble machine learning for groundwater potential mapping in the geologically complex terrain of Northwestern Nigeria. Standard thematic layers were integrated with embeddings from a pretrained remote sensing foundation model to train ensemble classifiers, with groundwater potential validated against Vertical Electrical Sounding (VES) data. The foundation-model-enhanced CAT-XGB ensemble achieved notable predictive performance, consistently outperforming standalone tree-based classifiers and traditional AHP. Spatial difference mapping revealed that the enhanced ensemble captured critical, localized hydrogeological anomalies smoothed over by expert-driven weighting. Bridging heuristic decision-making with foundation-model-enhanced machine learning significantly improves spatial precision and reliability, potentially providing a field-validated tool for optimizing borehole siting and sustainable aquifer management in data-scarce, geologically complex regions.
Related
- From water deficit to growth suppression: meta-analytic evidence for a functional hierarchy of drought adaptation in Hippophae rhamnoides.
- Projected increase in stress on African potential groundwater resources due to climate change and increased human water use
- Groundwater potential mapping using conventional and machine learning approaches with multi-indicator validation of Aflaj systems in Oman
- The commons at the intersection of forest fires and land tenure: insights from Galicia (Northwestern Iberia)
- Astronomical forcing and coal accumulation model of the Middle Jurassic Xishanyao Formation in the Turpan–Hami Basin, Northwestern China
BibTeX
@misc{osumeje2026groundwater,
title = {{Groundwater Potential Mapping in Northwestern Nigeria: A Comparison of the Analytic Hierarchy Process and Galileo-Enhanced CAT-XGB}},
author = {Joseph Orojah Osumeje and Sachet Sandeep Utekar and Morli Li and Dani Jones},
howpublished = {EarthArXiv},
year = {2026},
month = oct,
doi = {10.31223/x5mv5h},
url = {https://eartharxiv.org/repository/view/15434/},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Groundwater Potential Mapping in Northwestern Nigeria: A Comparison of the Analytic Hierarchy Process and Galileo-Enhanced CAT-XGB
AU - Joseph Orojah Osumeje
AU - Sachet Sandeep Utekar
AU - Morli Li
AU - Dani Jones
PY - 2026
DA - 2026/10/09
PB - EarthArXiv
DO - 10.31223/x5mv5h
UR - https://eartharxiv.org/repository/view/15434/
AB - Global water scarcity intensifies the need for high-resolution groundwater assessment tools, yet traditional approaches struggle to reconcile qualitative expert weights with the extensive spatial coverage offered by remote sensing data. This study presents a comparative framework evaluating expert-driven Multi-Criteria Decision Analysis (a ten-layer Analytic Hierarchy Process (AHP)) against data-driven ensemble machine learning for groundwater potential mapping in the geologically complex terrain of Northwestern Nigeria. Standard thematic layers were integrated with embeddings from a pretrained remote sensing foundation model to train ensemble classifiers, with groundwater potential validated against Vertical Electrical Sounding (VES) data. The foundation-model-enhanced CAT-XGB ensemble achieved notable predictive performance, consistently outperforming standalone tree-based classifiers and traditional AHP. Spatial difference mapping revealed that the enhanced ensemble captured critical, localized hydrogeological anomalies smoothed over by expert-driven weighting. Bridging heuristic decision-making with foundation-model-enhanced machine learning significantly improves spatial precision and reliability, potentially providing a field-validated tool for optimizing borehole siting and sustainable aquifer management in data-scarce, geologically complex regions.
ER -