Predicting water scarcity in northern Bangladesh using deep learning and climate data

Water scarcity, exacerbated by climatic variability and human activities, poses a significant challenge in northern Bangladesh. This study presents a comprehensive water scarcity map by integrating drought and groundwater potential maps using advanced deep learning techniques. A deep learning model with optimizer is employed to predict current and future water scarcity under shared socioeconomic pathways (SSP1-2.6, SSP2-4.5, and SSP5-8.5). The primary focus is on how integrating these datasets with deep learning and climate projections enhances the prediction and management of water scarcity, enabling innovative resource planning. Findings reveal that SSP1-2.6 significantly reduces water scarcity and drought risks, particularly during Kharif-1 and Rabi seasons, while SSP5-8.5 intensifies water scarcity, especially in Rabi. Model validation using total operating characteristic and area under the curve metrics confirms strong predictive performance. This study advances water scarcity assessment, offering a detailed and actionable framework for sustainable water resource management and climate adaptation strategies.