Morphometric assessment and soil erosion susceptibility maping using ensemble extreme gradient boosting (XGBoost) algorithm: a study for Hunza-Nagar catchment, Northern Pakistan

Abstract

Soil erosion and groundwater resources are two fundamental global concerns intricately linked through various hydrological and morphometric processes. Morphometrics with soil erosion assessment is crucial for managing hydrological processes and implementing preventative strategies. Utilizing Geographical Information system and Remote Sensing techniques, morphometric, morphotectonic, and soil erosion susceptibility in the tectonically active Hunza-Nagar catchment were explored, spanning 1455.05 km² with elevations from 1763–7697 m above sea level. With this motive, linear, areal, and relief morphometric variables were investigated. Analysis of the linear aspects indicated the sub-dendritic drainage pattern with streams ordered from 1 to 4th order. The calculated parameters recorded huge variations, including stream length of 384.92 km, bifurcation ratio of 1.65, drainage density of 2.65 km/km², drainage intensity of 0.25 km⁻¹, drainage texture of 0.49, stream frequency of 0.07 km⁻² and form factor of 0.41, respectively. The circulatory ratio of 0.46 indicates structural influence, elongation ratio of 0.72 reflects moderate to steep slopes with low flood regimes, length of overland flow of 1.33 km shows high infiltration and shape index of 2.47 underscores a higher risk of soil erosion in the catchment. Soil erosion susceptibility analysis was conducted using the XGBoost model, renowned for its proficiency in predictive modeling and classification tasks. The model was trained and tested on a dataset comprising factors pertinent to soil erosion dynamics. Subsequently, the trained model was applied to assess soil erosion susceptibility across the study area. The final Susceptibility map was classified from low to very high susceptible zones. Confusion matrix and Receiving operative characteristic curve (ROC) were used to validate the model. These results offer crucial insights into geohydrological characteristics, supporting global conservation efforts in conservation of natural resources and soil practices.