Dr Suraj Kumar Bhagat, Assistant Professor, Centre for Interdisciplinary Research has published a paper titled “HydroPredictor A Hybrid Machine Learning Model for Addressing Data Scarcity in Groundwater Prediction” in the Q1 journal having an impact factor 3.8. In collaborations with researchers in Morocco, Africa, their research has created HydroPredictor, a smart AI tool that mixes two powerful techniques: one that excels at handling categories (like soil types), and the other that uncovers hidden patterns in complex data, for finding hidden water reserves, helping communities manage precious resources wisely amid climate change.
Abstract
Predicting groundwater in data-scarce, environmentally sensitive regions is challenging due to sparse data, spatial variability, and nonlinear hydrogeological processes. We introduce HydroPredictor, a hybrid ML framework combining CatBoost’s categorical handling with regularized MLP’s nonlinear feature learning. Trained on 315 geo-referenced samples from Morocco’s Feija Basin using 10 environmental predictors (elevation, rainfall, soil permeability, NDVI, TWI), it employs Optuna hyperparameter optimization and 5-fold cross-validation. HydroPredictor delivers 89.23% test accuracy, F1-score of 0.8937, and AUC > 0.90 across classes. Statistical tests (Friedman, Wilcoxon; p < 0.05) confirm superiority over RF, SVM, and MLP, offering robust, scalable groundwater mapping for sustainable management.
Practical Implementation/ Social Implications of the Research
Accurate maps prevent over-pumping, boost farming/animal husbandry yields, and cut migration by ensuring water security—39% more viable land in some arid cases. They support SDG 6 by educating communities, avoiding contamination crises (e.g., arsenic), and fostering cooperative management in water-stressed regions like Rajasthan or Morocco.
Collaborations – Morocco, Africa
Future Research Plans
Incorporate live satellite/sensor data (e.g., GRACE-FO for water storage, Sentinel for NDVI) and IoT wells for dynamic predictions under changing rainfall or drought.
