| dc.description.abstract | Hydrological hazards often occur in conjunction with extreme precipitation events in Taiwan. The exceptional volume and intensity of the precipitation cause frequent torrential floods, sometimes with devastating effects on life and property. To improve our understanding of extreme events, the study modeled the rainfall-runoff processes using distributed watershed models with high-resolution precipitation input.
Precipitation data is generally collected from rain gauge stations. However, each measurement represents only the amount of rainfall at that particular spot, not precipitation in the surrounding area. Radar approaches are considered to offer a good spatial description of precipitation, but hardly predict precipitation quantities with acceptable accuracy. High resolution radar-rainfall estimates are compared with ground observations for an extreme precipitation event. The Taipei City area and the Shihmen reservoir watershed were chosen as the study sites, and the passage of Typhoon Nari (2001) through these areas was taken as the case study event. It was concluded that radar reflectivity from the Wufenshan radar station can be helpful for identifying precipitation variations during the passage of a land falling tropical cyclone. Spots with extreme rainfall can be identified when radar approaches are performed, but not based on gauge approaches. However, compared to the gauge approaches to the radar-rainfall estimates over the investigated domain tended to be overestimated. The divergence between radar-rainfall and gauge-rainfall can be identified via sub watershed investigations.
The watershed model with high resolution precipitation data was tested on a complex mountainous reservoir region, the Shihmen reservoir watershed. Radar-rainfall estimates were examined on this study. Numerical results generally revealed acceptable agreement between the observed and simulated reservoir stage hydrographs. The model calibration processes verified that the proposed model was effective for flood routing in the Shihmen reservoir watershed. Moreover, simulated results obtained using a grid size equal to 160m by 160m had the strongest agreement between simulated and measured data, and resulted in an execution time reduction of 40% than that of the case with 120m by 120m. Case study showed that inverse-distance weighting method carried the smallest error in estimation compared to all other spatial precipitation interpretations. The ratio approach produced the smallest residual error in simulation results among all other radar approaches. Precipitation is identified to be the main factor forcing model result.
A physical based distributed-parameter model combining surface runoff and groundwater flow is developed for investigating hydrological processes. Surface runoff is composed of both overland flow and river flow components, and the groundwater module considers the unsaturated zone and saturated zone in an unconfined aquifer system. An investigation of hydrological processes, including precipitation, infiltration, evaporation, percolation, surface runoff and groundwater flow are all considered in the proposed simulation model. Comparative analysis shows that the gradient method is superior to the GIS approach for describing the flow above riverbed. This study suggests using the Thiessen polygon method for precipitation interpolation. The best calibrations are obtained at a spatial resolution of 160m by 160m, when the simulated time step is less than five seconds. The proposed model shows good potential for storm based simulations, recession period description and long-term modeling. Therefore, the proposed model is confirmed to be suitable for mountainous watershed, such as Shihmen reservoir watershed. | en_US |