Machine learning-based modeling of groundwater recharge under three climate change scenarios in the Densu Basin of Ghana, West Africa

This study assesses the impacts of climate change on groundwater recharge in the Densu Basin using machine learning models. Increasing climate variability threatens groundwater sustainability, requiring reliable predictive tools. The Standardized Precipitation Index (SPI) and climate variables were used as inputs in Multi-Linear Regression (MLR), Random Forest (RF), and Artificial Neural Network (ANN) models. Due to their similar performance, RF and ANN predictions were combined to improve accuracy. Bias-corrected projections from multiple General Circulation Models (GCMs) under different emission scenarios indicate a future marked by higher temperatures and lower rainfall, leading to substantial reductions in groundwater recharge, particularly in the northern basin. The ensemble mean of ANN and RF models provided consistent projections across all climate scenarios. Results highlight that short-term precipitation deficits (SPI-3 and SPI-6) strongly influence recharge variability, underscoring the basin's sensitivity to near-term droughts. These findings emphasize the need for adaptive water management strategies to safeguard groundwater resources under a changing climate.