Exploring the morpho-tectonic nature, hydrological and physical characteristics of a watershed and prioritizing sub-watersheds surface runoff potentialities by integrating MCDM and ensemble machine learning models

Abstract

Much effective rainfall often leads to natural and human-induced hazards when unused. Therefore, monitoring and managing water resources by assessing comprehensive surface runoff (SR) potential is crucial instead of relying on broad sub-watershed (SW) generalizations. This study aims to identify the characteristics of landscape evolution and hydro-physical factors that influence SR. It prioritizes the SR potentials of SWs by examining multi-dimensional aspects of hybrid MCDM models integrated with ensemble machine learning techniques. Findings show SW shapes vary from elongated (0.57) to oval (0.87). Stream length ratios ranging from SW4 (0.61) to SW5 (9.99) indicate diverse slope magnitudes and geomorphic development stages. Moderate drainage densities (0.65–0.89 km/km²) indicate underlying litho-structural influences on drainage networks. Low to moderate drainage textures (1.33–2.33) suggest potential SR zones, mainly due to the Proterozoic gneissic complex dominance. Landscape evolution analysis shows SWs are at an early mature stage (average 0.49). The prioritization results from the hydro-morphotectonic (MPT) and hydro-physical (HYP) TOPSIS and VIKOR integrated extra trees regression (ETR) models achieve the best accuracy, followed by XGBoost, Bagging, and Voting models. Among the 24 models analyzed, TOPCOM-ETR (RMSE = 0.106), VIKHYP-ETR (RMSE = 0.093), and TOPMPT-ETR (RMSE = 0.129) are found most effective for prioritizing SWs. The study identifies that sloping SWs 1–2, 9–11, besides concave SWs 5 and 8, are optimal for maximizing SR. The TOPCOM-ETR model highlights the foothill SWs with the highest runoff potential based on relief, slope, and hydrological features. The findings of this study can help guide government agencies in making informed decisions to tackle water-related challenges and uncertainties.