摘要:經濟部水利署為落實防洪減災的任務,近年來致力於精進洪水預報技術,發展出「水利署洪水預報系統」,期望能在水患發生前,透過降雨推估和逕流、洪水演算,有效操作水庫閘門蓄減洪峰,並對洪氾淹水潛勢地區發佈警報,緊急撤離民眾,降低洪水災害的生命損失。唯台灣山脈阻擋、谷地槽導等複雜地形效應,使得現有雨量觀測系統仍難以準確估計颱風時山區的降雨歷線和分佈;湍急逕流經常改變河川斷面,不易建立水位-流量率定曲線,流量估計誤差可能很顯著。降雨、逕流觀測誤差,和基於線性假設的逕流模式應用於高時變性、高動態範圍的山區颱風降雨,導致逕流模式參數錯誤;再加上不準確的定量降雨預報等,共同構成逕流歷線的偏估,為河川水位預報最主要的誤差來源。「系集降雨預報應用於洪水預報之研究」二年期計畫,是在2005-2006「洪水預報與減災應變格網建置計畫」、2007-2009年「水文模式與分散式洪水預報系統整合應用計畫」、2010-2011年「機率式洪水預報系統之研發」建置的水利署洪水預報系統既有基礎上,針對三大工作目標進行技術研發與系統擴充:(1)利用系集降雨預報,建置河川洪水位機率預報系統:引進數值天氣預報模式為基礎的系集定量降雨預報,發展河川洪水位機率預報產品;並針對洪水系集預報高計算量需求,研發計算效率高、可長期模擬之降雨-逕流演算模組。(2)協助水利署彙整各河川局不同洪水預報系統的預報資訊,以供颱洪時期間防災預警研判。(3)擴充水利署洪水預報系統之預報流域:建置花蓮溪、卑南溪及後龍溪洪水預報系統。第一年度計畫已完成引進系集降雨預報產品,納入平行演算機制,開發計算效率高、可長期模擬之降雨-逕流演算模組(BKW與GSMA模式),同時協助防災中心訂定水利署洪水預報產品資料標準格式和抬頭編碼,以利彙整、展示DRAINS系統和各河川局洪水預報系統輸出的洪水位、流量資訊等技術服務工作,以及以既有之DRAINS架構建置花蓮溪與卑南溪流域洪水預報系統。第二(本)年計畫持續進行高效能逕流模式機制與效能改進,並建置蘭陽溪、淡水河、濁水溪、曾文溪、林邊溪與太麻里溪流域之高效能逕流模式;建置河川洪水位機率預報成果展示網頁,並以淡水河、濁水溪、曾文溪、林邊溪與太麻里溪流域為例,進行事後預報,評估河川洪水位機率預報成效;建置「河川洪水預警綜合展示平台」以便於颱洪時期能更快速的整合暨共同展示各河川局回報之河川洪水預報資訊;以及以既有之DRAINS架構建置後龍溪流域洪水預報系統。
Abstract: To implement flood control and disaster reduction, the Water Resources Agency (WRA) of Ministry of Economic Affairs devoted to improve flood forecasting technology in recent years, by calculating precipitation, runoff and water level before flood, it is expected to decrease flood with reservoir operations, to give early warning and evacuate people in the potential flooding area, and to decrease losses. Due to the complicated terrain effect, it is difficult to estimate the hydrograph and distribution of precipitation in the mountain area during typhoon seasons with the existing observation system; also it is not easy to build rating curves because the channel cross sections are easily changed, and might cause apparent error of discharge. The observation error of precipitation and runoff, and using the runoff model based on linear hypothesis to the mountain area during typhoon periods, both cause the false parameter in the runoff model; and with the inaccurate QPESUMS forecasting data, all the above result in the runoff hydrograph bias, which is the main reason of water level forecasting error.Based on the established results in 2005-2006 “Development of the Flood Forecasting and Mitigation Grid” project and 2007-2009 “Integration and application of hydrologic models and the distributed flood forecasting system” project, this two-year project “A study to apply ensemble QPF with operational flood forecasting” is to do some research and development and further develop the system by completing the 3 main items, which are: (1) applying the ensemble forecasting data to build flood probability forecasting system; (2) giving assistance to WRA in integrating forecasting data from different flood forecasting system in each River Management Office; (3) expanding DRAINS to other river basin.The main accomplishments for the first year of this project are:For the 1st subject, we design and develop a precipitation-runoff model which has high computation performance and can run long term simulation. For the need of max computation of ensemble flood forecasting, Block Kinematic Wave model (BKW) which has high performance and can run long term simulation is developed, and another model called Geomorphic Soil Moisture Accounting model (GSMA) is developed to work with BKW. These two models for Lanyang river basin are also completed.The 2nd subject is to apply ensemble forecasting data to build flood probability forecasting system. We use Lanyang river basin for example, choose two serious events from past few years and do 事後預報 to assess the outcomes of flood probability forecasting.In the 3rd subject, we integrate the forecasting data from different flood forecasting system in each River Management Office, and help WRA to understand the most possible situations in river administered by central government. And the 4th subject is to provide water level forecasted by DRAINS as the demand of WRA 4 times (07, 11, 16, 19LST) a day, based on the required format. The last subject is to develop Hualien river and Beinan river flood forecasting system based on DRAINS.