Spatial and Temporal Rainfall-runoff Simulation Integration with Uncertainty

• Main Authors: Jr-Chuan Huang, 黃誌川
• Other Authors: Mei-Ling Hsu
• Format: Others
• Language: zh-TW
• Published: 2002
• Online Access: http://ndltd.ncl.edu.tw/handle/94492156402972311639

博士 === 國立臺灣大學 === 地理環境資源學研究所 === 91 === Because of Taiwan’s special environmental characteristics including highly fractured geology, precipitous landform, and frequent heavy rainfall, many hydrological model developed abroad cannot be effectively applied. Traditional hydrological models treat specific watershed as a single homogenous unit, only simulate and calibrate runoff hydrographs in its outlet, regardless of the spatial pattern within the watershed. With the progress of remote sensing techniques, like LIDAR, NEXRAD, collection and updating of spatially and temporally variable hydrology-related characteristics of the watershed have become more and more efficient. If hydrological models could effectively incorporate the spatially varied information within watersheds, then they could be applied to many more related research fields, and greatly enhance their value. This research develops a hydrological model (SEDIM) which incorporates spatial information derived from DEMs and GIS data base to simulate surface and subsurface runoff within given watersheds for given storm events. The model adopts continuity equation, diffuse wave model, and the Manning’s equation for surface runoff routing, while uses the Green and Ampt infiltration theory and Darcy’s Law for subsurface runoff treatment. Uncertainty analysis is used to retrieve optimal parameter values and to estimate the confidence interval for the simulation results. The environmental data of the Heng_Chi catchment, and 6 rainfall events were used first for parameter calibration, and then 4 other storms were used to verify the results. Simulation efficiency coefficients for almost all the calibration and the verification runs reach 80%. Generally speaking, the model has very good performance, there is only one hour difference between the estimated and the actual lag time of the flood peak, and the estimated error for the peak flows are mostly under 15%.