圖書典藏及影音數位平台

經濟部水利署為水災應變需要，於106-107年執行「淹水預警功能之測試與運作」研究計畫，完成臺中、臺南、高雄及屏東之淹水預報系統，可於每時間提供未來3小時淹水模擬，經測試結果顯示可反應現地淹水趨勢。為使預報成果與現地情況更加相似，本所針對系統中運算結構及資料進行盤點，藉由資料分析調整部分系統結構，以提升即時淹水模擬效能。 本計畫將蒐集國內外有關系統測試、水理模式資料及運算處理等文獻，以瞭解運算結構及改善方法，且針對系統中需提升的部分提出相對應的精進建議。此外對水理模式進行全面性檢視，以氣象資料處理、地文資料建置及模式演算方法等問題剖析對模擬結果的影響程度。依據已探討的問題進行最佳化研究，挑選出重點結構改善系統誤差並利用後處理方式改善隨機誤差，進而提升系統運算效率。另外將以數場颱洪事件進行測試，分析預報結果並檢討改善方法。

In recent years, with the improvement of hydrological observation technology, many hydrological observation techniques have been gradually applied to drainage planning. WRPI has verified the hydrological model applicable to Taiwan through the verification of the local observation data. Also, through the establishment of the SOP of building the hydrological model, the accuracy and computational efficiency of the hydrological model was able to improve. To implement regional flood prevention, since 2015 Water Resources Planning Institute (WRPI) has used the results of the third-generation Flood Inundation Map to build the integrated platform for the application of high-performance 2D inundation simulation called "Flood Early Warning System." It has successfully integrated the FEWS_Taiwan to the Flood Inundation Map. In the system, the SOBEK model for Taichung City and Tainan City in Flood Inundation Map uses the forecast data of the CWB to predict the possible flooding range for the next three hours. After integrating SOBEK 2D flooding mode, FEWS_Taiwan has improved the flooding mode from static situational simulation to dynamic real-time data simulation through immediate and forecasted rainfall information, which is a crucial milestone for early warning of flooding areas. As the computer computing speed increases, the use of FEWS_Taiwan has stabilized the scheduling capability, coupling with a fast parallel computing architecture and the stable update of the basic data that improves the correctness and calculation speed of the model, the goal of instant flood warning can be achieved. In order to make the forecasting results faster and simulation similar to the local situation, this year WRPI evaluation and analysis the operational system structure and data of the system and adjusts part of the system structure by data analysis to improve the efficiency of the real-time flooding forecasting. The porject will continue to collect domestic and foreign literature to understand the computing structure and improvement methods, through the previous years’ project achievements, found the main factors affecting real-time flood simulation, including the basic data source, model structure and operational platform. For the computing structure part, there are two suggested directions at this stage: one is the flexible expansion and reduction of computing resources; the other is to remove computing resources for other requirements during non- typhoon flooding. In the part of the hydraulic model calculation mechanism, the factors affecting the accuracy of real-time flooding simulation can be roughly divided into three aspects: the model calculation method, the geographic data and the rainfall-runoff module. At the same time using new hydraulic model D-Flow FM (base on the opensource model) to testing the possibility of 2D flooding simulation model transfer in the future. Finally, the real-time 2D flooding simulation in Tainan area will be used as a target to test several flood events in the area in recent years, analyze the forecast results and review the improvement methods to find out the effect of improving the immediate flooding simulation.