SWAT-aided Surface Runoff And Sediment Yield Modeling And Parameter Uncertainty Analysis

Due to in-depth understanding of hydrological processes and continued research on watershed modeling, distributed hydrological models have gradually begun to replace traditional concept models and to be extensively applied for supporting decision making in many critical areas such as flood control, drought combat, and water resources management. However, because of the complexity and dynamic of hydrological processes, the system and random errors in measurement and data, and faultness of hydrological modeling approachess, uncertainies are introduced to and have influece on model validation and and hydrological forecasting. In-depth and systematical studies on distributed hydrological modeling and parameter uncertainty analysis are ugently needed to enhance the capacity of hydrological forecasting and improve the prediction accuracy in China. This would be practically significant and scientifically valuable to help reduce flood and droughts losses, control soil erosion, and more effectively manage and conserve water resources.To help meet the above needs, this thesis research targeted an advanced and widely used distributed hydrological model-Soil and Water Assessment Tool (SWAT). Through a real-world case study in the upper reach of the Huolin River Basin, the model was established, validated and applied for simulating the hydrogoical processes and sediment transport, leading to reasonable predication accuracy. In order to verify the reliability of the SWAT model, this research chose the SUFI (Sequential Uncertainty Fitting)-2 and GLUE (Generalized likelihood uncertainty estimation) methods to analyze the modeling uncertainties. The results showed that both methods were feasible and able to reflect the uncertainty intervals of model parameters and the ranges of prediction resutls. This research further examined the capibility and effeciency of the two methods in analyzing uncertainties of distributed hydrological models through consideration and comparison of three uncertainty sources in parameters, simulation processes, and computation efficiency. The results indicated that both methods had their own advantages; however, in comparison with the GULE method, the SUFI-2 method was more capable for handling model complexity and meeting high computation requirements. But GLUE methods is more fully reflect the correlation between the parameters relatively for the analysis parameters uncertainty, and the simulation results objective function obtained is ideal relatively. Finally the thesis summerized the findings and discuss possible future work.Through in-depth studies on the above issues, this thesis research established an integrated approach including distributed hydrological modeling, model validation and application, and uncertainty analysis and comparison. The research methology and findings would help build a solid basis for improvement of uncertainty analysis theories and application of distributed hydrological models, as well as provide a valuable reference for other relevant research.