Deep and Machine Learning for monitoring groundwater levels and hydrological changes using GRACE and SENTINEL-1 for the Ganga River basin

Accurate estimating groundwater levels in river basins is paramount for hydro-geological research and sustainable water resource management. This research presents a deep-learning approach for accurate groundwater level estimates in the Ganga River Basin. Our method leverages the transformative potential of Vision Transformers (ViT) and their variants, with an emphasis on the Swin-Transformer variant enhanced with Normalization Attention Modules (NAMs), by utilizing the combined band information of Sentinel-1 synthetic aperture radar (SAR) and GRACE satellite data. We have assembled an extensive dataset and trained with computer vision to tackle the Ganga River basin’s particular difficulties. Cutting-edge Vision Transformers showed great promise in groundwater level estimation, with the model based on Swin-Transformer NAM achieving an exceptional Mean Absolute Error (MAE) of 1.2. The basin and beyond will benefit greatly from this study’s significant advancement of sustainable groundwater level and utilization practices.