Spatiotemporal Reconstruction of Groundwater Depletion in India’s Oldest Coalfield Using Machine Learning and GIS, With InSAR and Fractal Analysis Insights

Abstract

Rapid mining expansion, urban growth, and climate variability have intensified groundwater (GW) stress in several industrial regions of India. This is particularly true in the country’s oldest coalfield, the Raniganj area. However, long-term high-resolution monitoring of groundwater storage (GWS) and groundwater levels (GWLs) in such heavily modified terrains remains constrained by sparse observations and data gaps. This study develops a data-driven machine-learning framework to reconstruct spatio-temporal GWL dynamics from 2002 to 2024 by integrating GRACE/GRACE-FO–derived GWS, four static topographic–geomorphic factors, thirteen satellite-based hydro-environmental variables, and ground-based observations, with field verification. Among 39 model realizations using four algorithms, random forest consistently outperformed support vector regression, artificial neural network, and extreme gradient boosting, achieving robust percentage-based performance metrics that are operationally acceptable (NRMSE = 15.01%, NMAE = 11.5%, NMSE = 0.02), and strong agreement with observed time series at most wells (r> 0.6). A two-step gap-filling strategy was applied to generate continuous pre- and postmonsoon GWL simulations, indicating that premonsoon conditions best represent long-term variability and depletion. The results identify critical depletion hotspots in Durgapur, the Kataberia coalbed methane fields, and north-eastern mining areas, with a maximum long-term decline rate of 13.4 cm yr−1. Reconstructed depletion patterns exhibit a strong spatial correspondence with (small baseline subset interferometry) SBAS–InSAR–derived land subsidence and persistent behavior, as indicated by Hurst exponents of fractal analysis, greater than 0.67. The proposed framework provides reliable quantitative information to support GW assessment, urban planning, hazard mitigation, and sustainable management in intensively mined regions.