A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering, School of Engineering and Technology

Thesis (M.Eng.) - Asian Institute of Technology, 2005

Soil erosion has been recognized as a global threat against the sustainability of natural ecosystem. But, modeling of soil erosion and sediment yield has been regarded as the one of difficult task as it is a highly dynamic process in spatial scale. The objective of this study was to compute and verify the erosion, deposition and sediment yield in distributed manner in catchment scale. Both empirical and process-based models are applied in different physiographic conditions in this study. Universal Soil Loss Equation (USLE) is an empirical model developed by US Department of Agriculture (USDA). Two forms of USLE, known as Revised Universal Soil Loss Equation (RUSLE) and Modified Universal Soil Loss Equation (MUSLE) along with a process-based soil erosion and sediment transport model have been applied in this study. RUSLE, which lumps enter-rill and rill erosion together, is one of the popular soil erosion models because of its simplicity. Rainfall erosivity, soil erodibility, slope length and steepness, management practices and supporting conservation practices are the fundamental parameters of RUSLE. RUSLE method has been applied to three sub-basins of Chi-Mun basin and Bagmati basin in GIS environment. Digital elevation model (DEM) of two resolutions: 90 m and 1000 m are used in the study in order to analyze the effect of DEM resolution on soil erosion and sediment transport process. RUSLE results show that the model performance is better for finer resolution than the courser one. The closest agreement was observed for the resolution of 30 m in most of the cases. Similarly, modified form of USLE, known as MUSLE and a process based soil erosion and sediment model developed by University of Tokyo are also applied to one sub-basin of Chi-Mun basin. Process-based model was first calibrated and validated for the water discharge before applying to sediment discharge computation. Performance of empirical models and process-based model was compared in the sub-basin. Process-based model performance was better for the time series computation of sediment yield. From June 1990 to October 1990, the % eror of the process-based model is within 70% with observations. It should be noted that RUSLE is intended primarily concerned on predicting long-term annual average (sheet and rill erosion) erosion by water on distributed slopes, not for sediment yield computation. So, experimental results should be carried out in the plot/field scale so that erosion estimation from RUSLE can be verified before the verification of sediment yield at the catchment outlet. And, saturated and unsaturated zone part should be added to the surface overland flow model used in the study in order to make reliable estimate of water discharge and subsequently sediment discharge as well.