Groundwater Storage Dynamics in Purba Bardhaman District, West Bengal, India: Insights from GLDAS Using Multi-temporal Scale Statistical Analysis and Multiple Linear Regression

Groundwater is a vital yet increasingly stressed resource for rice-based agriculture, domestic supply, and local livelihoods in the alluvial plains of Purba Bardhaman District, West Bengal, India. This study examines groundwater storage (GWS) dynamics from 2003–2023 using harmonized 0.125° × 0.125° gridded products from the Global Land Data Assimilation System (GLDAS) and rainfall from the India Meteorological Department (IMD) across 32 coordinate locations. A multi-scale statistical assessment, complemented by mRMR-assisted data-driven regression modelling, was applied to quantify trends, identify regime shifts, and isolate dominant hydro-meteorological controls on GWS variability. Significant negative trends (p < 0.05) were detected in all monthly temporal scale with steepest decline in Winter, Monsoon, Post-Monsoon and Annual scales. Abrupt regime shift in GWS was detected in 2008/2009. Profile/root-zone soil moisture exhibited near-perfect correlations (r = 0.867 to 0.998), while cumulative rainfall (CR3) and baseflow showed moderate-to-strong influences (r = 0.550 to 0.845). mRMR-optimized MLR models achieved high fidelity (R² = 0.670 – 0.898), underscoring soil moisture and lagged rainfall as key drivers. The central novelty of this work is the integration of regime-shift detection with mRMR-based predictor screening to link abrupt groundwater transitions to physically interpretable hydro-meteorological drivers at sub-district scale. The results indicate that groundwater recovery windows are shrinking even during monsoon months, consistent with an extraction–recharge imbalance in an intensively irrigated landscape. These findings would be conducive in developing the targeted management for rice-based systems by strengthening recharge enhancement, improving irrigation efficiency, and prioritizing high-decline zones.