Forecasting of groundwater level variation under changing climate in Chhattisgarh state using deep learning technique

ABSTRACT Groundwater is the most precious natural resource in modern days. India is the largest consumer of groundwater globally, with over 25% of the world's groundwater extraction. Climate change affects the groundwater level (GWL) both in direct and indirect ways. Recently developed deep learning (DL) models are considered only the direct drivers of the groundwater dynamic. Including indirect key drivers such as anthropogenic activities and lithology to forecast GWLs using machine learning techniques is poorly understood. This paper aims to consider both the direct and indirect key drivers for forecasting seasonal GWLs. A modified approach based on a DL model has been formulated in this context that considers land cover dynamics, lithological properties, and climatic variables such as temperature and precipitation. The model was calibrated and validated to forecast seasonal GWLs for four Shared Socioeconomic Pathway (SSPs) scenarios. The results show that the median of R2 and Nash–Sutcliffe efficiency in calibration is 0.83 and 0.81, respectively, and in validation, 0.84 and 0.82, respectively, which is acceptable. Overall, the results obtained broadly correspond to an acceptable degree of accuracy. The proposed methodology is applicable for seasonal GWL forecasting and can be useful to farmers and key stakeholders.