Harnessing Remote Sensing and AI for Groundwater Resource Mapping: A Study from the Varuna River Region, Uttar Pradesh

Groundwater is an essential natural resource supporting all living beings in ecological and agricultural systems, especially in the Indo-Gangetic Plains (IGP).  The IGP comprises shallow aquifers densely populated with agriculturally productive regions. However, in the past few decades, the area has been under water scarcity for several reasons, including climate change and mismanagement, affecting livelihoods. For all these reasons, an accurate assessment of groundwater availability and identifying groundwater potential zones (GWPZ) are crucial. The modern techniques of GWPZ identification, including AI/ML with remote sensing, play a vital role in determining the potential zones with high accuracy.  The objective of the present study is to determine the GWPZ around the Varuna River region of Uttar Pradesh, India, using remote sensing with machine learning models.

The present study tries to delineate potential areas of groundwater augmentation in the Varuna River area of Uttar Pradesh, India, by using remote sensing techniques supplemented with machine learning algorithms such as Support Vector Machine (SVM), Gradient Boosting Machines (GBM), Random Forest (RF). Satellite imagery, geospatial analysis, and predictive modeling assessed various hydrological, geological, and climatic parameters. With such state-of-the-art tools, this study tries to provide broad coverage of groundwater distribution, locating the region's possible areas, hence contributing toward sustainable water management strategies that will strengthen the communities of people depending on this precious resource.

We have classified the GWPZ into five different zones, ranging from very low, low, moderate, and high to very high zones. Compared to all models, the RF shows the highest predictive accuracy, with Area Under the Curve (AUC) values of 91%, whereas the GBM and SVM AUC curves show 88% and 84%, respectively.