不同降雨事件各有不同的天氣特徵，且台灣地區降水現象受地形和综觀條件影響很大，使得氣象模式往往會產生高估或低估降水預報。本論文研究分析2005年3~12月之氣象局自動雨量站資料和模式WRF模擬雨量場資料以探討全台灣季節性和全年10個月的降水分佈特性，希望暸解WRF模式對於季節性降水的預報能力與其對於單一降水事件預報能力間的差異，得知WRF氣象模式對於不同時間尺度降水現象的預報準確度。 颱風伴隨強風豪雨，經常造成台灣地區淹水和人員傷亡，以及土石流、山崩等天然災害。因此為瞭解在颱風強降水事件之下集水區的淹水範圍與淹水過程，於是本論文研究挑選2005年之三個颱風事件，以水文模式來模擬石門水庫流量的多寡，以進行討論研究。 研究結果發現，WRF氣象模式模擬季節性降雨的分佈情形比模擬單一降雨事件(梅雨、颱風)時結果較好，得知WRF模式可以掌握到在台灣地區不同季節的盛行風向所帶來的地形性降雨分佈。本研究針對台灣崎嶇高聳的地形造成迎風面降雨增加和背風面降雨減少情形，有更進一步的分析探討。WRF氣象模式對於極端降水事件之雨量及颱風路徑模擬都有不錯的結果，但WRF模擬雨量大多為高估降雨。若將WRF模擬雨量輸入至水文模式FLO-2D以模擬石門水庫霞雲集水區的流量時，由於受到WRF模擬雨量誤差影響，導致在有些颱風事件模擬流量時結果並不是很理想。最後本論文研究中之水文模式FLO-2D並未設定初始基流和入滲後出流，因此可能導致流量模擬上的誤差。 Different precipitation events occur under different synoptic weather regimes, and they are strongly affected by the Taiwan terrain and large-scale/synoptic conditions, so it is difficult for the meteorological model to have accurate precipitation forecast. This thesis utilized analyzes the automatic rain gauge data of the Central Weather Bureau and WRF-simulated rainfall data provided by Mr. Shi-Chien Wang. 1) to investigate the rainfall characteristics for each weather regime, 2) to examine the predictability of seasonal-to-annual rainfall of the WRF model, and 3) to verify the accuracy of WRF-predicted rainfall at different time scales. The heavy rainfalls associated with typhoons often induce severe flooding, debris flows, landslides, and cause a great lose of human lives in Taiwan. Thus, in order to understand the flooding characteristics under typhoon-induced the heavy-rainfall events, this study will uses the FLO-2D hydrological model to simulate the river discharge over the Shihmen Reservoir watershed for three typhoon events in 2005. Results show that the WRF model simulates better the distribution of seasonal rainfall than that for individual event (Mei-Yu front and typhoon), because the WRF model has a good skill at predicting the orographic rainfall under the prevailing wind in different seasons. This study also discusses the effects for the rainfall increase over the windward slopes and the decrease over the leesides in Taiwan. We find that the WRF model predicts well the rainfall distribution and typhoon tracks for heavy rainfall events, but the WRF model tends to over-predict the precipitation amount. If the WRF-simulated rainfall is taken as input for the FLO-2D hydrological model, the predicted river discharge is highly sensitive to the errors inherent in the WRF-simulated rainfall. Finally, the neglect of initial base flow and infiltration outflow in the FLO-2D model may cause discrepancy in discharge at the later stage of a flooding event.
