Climate-resilient strategies for sustainable groundwater management in Mahanadi River basin of Eastern India

Abstract

A comprehensive groundwater (GW) monitoring approach is necessary for the long-term sustainability of regional economies and livelihoods, especially with the threats of population explosion, rapid urbanization, and climate change. By using modern technologies like integrating of machine learning, geographic information systems (GIS), and remote sensing (RS) data, climate-resilient monitoring strategies can be developed. This study aims to estimate groundwater levels using records from the TerraClimate dataset (1958–2020) and predict future GW patterns up to 2050 using climate model data. The focus is on the Mahanadi River basin in India, utilizing GIS, RS, and Google Earth Engine cloud. Future climate trend analysis (2021–2050) was conducted using the mean ensemble CMIP6 models (i.e., EC-Earth3 and MIROC6) historical, SSP2-4.5, and SSP5-8.5 datasets. Additionally, spatiotemporal vegetation indices were analyzed using MODIS data (2010–2017). This research employs an innovative ensemble boosting of six machine learning algorithms to predict groundwater levels and develop climate-resilient agricultural strategies in the river basin. The approach uses six nature-inspired wrapper algorithms to identify the best features contributing to groundwater level predictions for pre-monsoon and post-monsoon seasons. Validation was done using regional groundwater level data and various machine-learning classification matrices. The Boruta algorithm and SHapley Additive exPlanations methods were applied to select features for delineating hydrometeorological conditions in the study area. The slope of linear regression results showed a negative trend in precipitation in the lower part of the basin (around − 0.371 mm/year) and a positive trend in the upper part (around 0.238 mm/year). In the pre-monsoon period, the Extreme Gradient Boosting and Adaptive Boosting models achieved the best accuracy of 92% based on the area under the receiver operating characteristics curve. Based on the ensemble of two CMIP6 GCMs, under SSP5 8.5, 33.25% of the area is classified as having very high GWL exposure during the pre-monsoon period, compared to 30.58% in historical data. Additionally, under SSP5 8.5, 23.80% of the area is classified as having very high GWL exposure during the post-monsoon period, compared to 18.73% in historical data. The heavy reliance on groundwater for irrigation is a major cause of groundwater depletion in the catchment area. These results can inform sustainable agriculture planning, policymaking, and management in the future.