Integrated assessment and prediction of irrigation water quality in the Kuttiyadi River Basin using Fuzzy-AHP and XGBoost

Water pollution threatens river ecosystems and agriculture, as poor-quality irrigation water drives salinity, sodicity, alkalinity, and toxicity. These impacts reduce soil fertility, impair infiltration, and lower agricultural productivity. This study assessed irrigation water quality in the Kuttiyadi River basin, Kerala, India, using an integrated framework combining hydrochemical characterization, fuzzy logic-based multi-criteria decision-making, and machine learning prediction. Seasonal parameters were benchmarked against IS 11624:2019 standards. A Fuzzy Irrigation Water Quality Index (FIWQI) was built using Fuzzy-AHP, while an Extreme Gradient Boosting (XGBoost) regression model was trained on eight indicators to predict irrigation suitability. This combination of fuzzy logic with advanced ML is a significant contribution, enabling both interpretability and high predictive power. Findings showed generally favourable irrigation conditions, though Piper diagrams revealed seasonal hydrochemical shifts. The XGBoost model achieved strong accuracy (R2 = 0.98 for Training and 0.84 for Testing). The hybrid fuzzy-ML approach is scalable, adaptive, and supports real-time irrigation management, offering policymakers a reliable tool for sustainable river basin agriculture.