Research Projects

Effective groundwater management in India faces challenges due to limited high-resolution, real-time data. Groundwater is a critical resource for agriculture, drinking water, and industry, yet its sustainable management requires detailed spatiotemporal data. Current groundwater level (GWL) monitoring in India is often done on a quarterly basis, hindering timely decision-making. While real-time data acquisition systems like the National Hydrology Project (NHP) provide some improvement, there is still a need for high-resolution data to better understand groundwater dynamics and inform management strategies. This project aims to integrate advanced machine learning (ML) and numerical models to enhance both the temporal and spatial resolution of groundwater level data, addressing these data gaps and improving predictions of groundwater availability. The project focuses on three main scientific objectives: 1. Data Collection and Preprocessing: In the initial phase, the project will integrate multiple data sources—including remote sensing (e.g., GRACE), meteorological (rainfall, evapotranspiration), and irrigation data—to refine existing coarse GWL data. This will involve cleaning, preprocessing, and feature engineering to create a robust dataset that captures key variables influencing GWL. 2. Machine Learning Model Development: Machine learning models will be developed and tested to predict GWL. The most effective model will be selected based on its ability to improve both the temporal and spatial resolution of GWL data, enabling more accurate predictions during periods with missing data or coarse resolutions. 3. Integration with Numerical Models: The refined GWL data, including ML-predicted and observed values, will be integrated into numerical groundwater flow models to simulate spatial variability across aquifers. Coupled groundwater-surface water models will be used to calibrate these predictions, incorporating both observed runoff data and refined GWL data. This integration will help predict groundwater levels at finer resolutions and improve the understanding of groundwater flow dynamics. The central hypothesis is that by combining machine learning techniques with numerical groundwater flow models, it will be possible to achieve high-resolution temporal and spatial predictions of GWL, even in data-scarce regions. The project will significantly advance the understanding of groundwater systems by providing better tools for real-time monitoring and long-term predictions. If successful, the research will improve groundwater management strategies, contributing to more sustainable water resource use, particularly in water-scarce regions of India. The integration of ML with numerical models also represents a novel approach to hydrological modeling, with broad implications for the study of aquifers globally.