Predictive Hydrogeophysical Modelling of Subsurface Conditions Using Geo-Electrical Data Along a Water Channel in Akwa Ibom State, Southern Nigeria

Groundwater exploration in alluvial and river-influenced terrains of Southern Nigeria is challenged by subsurface heterogeneity and the high cost of invasive drilling. This study presents a predictive hydrogeophysical modelling framework that integrates Vertical Electrical Sounding (VES), Electrical Resistivity Tomography (ERT), and machine learning to estimate deeper aquifer properties from shallow resistivity data along the Kwa Iboe River channel. A total of 27 VES and 9 ERT profiles were analyzed to characterize geo-electrical parameters. Machine learning models, including Random Forest, Neural Networks, and Ridge Regression, were evaluated for predicting deeper resistivity, porosity, hydraulic conductivity, transmissivity, and aquifer thickness. Random Forest achieved superior accuracy (R² = 0.92–0.96) in resistivity prediction, while Ridge Regression performed best for aquifer thickness (R² = 0.75). The results delineated high-yield zones with low resistivity (<20 Ωm) and high transmissivity (>1000 m²/day), mainly near the river. Strong correlations were observed between resistivity and hydrokinetic parameters (e.g., r = 0.99 between ρb and ρw). Spatial maps illustrated aquifer behavior influenced by lithology, topography, and river recharge. This integrated, data-driven approach offers a cost-effective alternative to traditional drilling, aiding sustainable groundwater development and environmental planning in tropical, riverine settings.