Forecasting Future Groundwater Recharge from Rainfall Under Different Climate Change Scenarios Using Comparative Analysis of Deep Learning and Ensemble Learning Techniques

Groundwater is the most reliable source of freshwater for household, industrial, and agricultural usage. However, anthropogenic interventions in the water cycle have disrupted sustainable groundwater management. This research aims to comprehend the future of groundwater recharge predominantly due to rainfall under changing climate. In this study, predictors of groundwater recharge such as precipitation, land use land cover (LULC), soil type, land slope, temperature, potential evapotranspiration, and aridity index (ArIn) were used for the Punjab region of India over the duration of 34 years, from 1986 to 2019. To simulate future conditions, various climate change scenarios from the CMIP6 report have been incorporated. Different Artificial Intelligence and Deep Learning models, ranging from the straightforward Linear Regression model to the intricate Extreme Gradient Booting (XGBoost), used these parameters as input. Statistical analysis of the models showed that XGBoost is most effective in predicting the groundwater recharge phenomena. Correlation studies revealed precipitation to be the primary contributor to recharge, followed by the ArIn, while soil type and slope are found to have the strongest inverse correlation. The models’ resilience and performance were investigated by conducting a k-fold cross-validation analysis. The pattern of groundwater recharge is forecasted for the years 2020 to 2035 across Punjab with different climate change scenarios. The study demonstrates how the Punjab area is mirroring its current status around Shared Socioeconomic Pathway (SSP) 370. Groundwater level estimates confirmed its strong correlation with and dependence on groundwater recharge. The analysis is strengthened by comparing the AI-predicted groundwater recharge with the Central Ground Water Board (CGWB) Punjab’s annual estimate.