多重尺度格網細化之淹水模擬

Abstract

全球氣候變遷導致臺灣降雨強度逐漸增加，而臺灣又因地形特殊，造成河川坡陡流急，每逢颱洪暴雨，常造成嚴重的災害。因此，若能建立一套有效率且準確的淹水模式，即可提供淹水資訊，強化決策作業、降低洪災損失。 本研究利用數值地形高程(DTM)與地理資訊系統(GIS)處理淹水模式之各項資料，然而近年來遙測技術的迅速發展，地形資料之解析度相對提高，增加了淹水模式的準確性，但在使用這些高解析度的數值地形資料進行大區域的淹水模擬時，則有計算量過大及效率降低之問題。使用多重格網局部細化淹水模式可在保有高準確度的模擬結果下減少模擬時計算資源之浪費。本研究以陳彥宏(2008)局部細化淹水模式為基礎，進一步加入第三層格網，將三種不同解析度（600公尺,200公尺,40公尺）格網相互搭配，於重要區域進行尺度更小之淹水模擬，並搭配平行演算法，使模式能更快速地進行模擬，同時兼顧其準確性。 本研究利用理想案例與現地案例進行模式驗證，理想案例為一個10公里×10公里之正方形封閉模擬區域，搭配理想設計雨型進行模式驗證；現地案例以臺南市曾文溪北岸為模擬區，並以近50年來淹水最嚴重之莫拉克颱風事件（2009年8月）進行模式精確度與效率度之分析探討。經案例模擬證明，在三種不同解析度格網搭配模擬下，其模擬結果能與全區高解析度格網模擬相近；演算效率上，利用多重格網搭配模擬可大幅縮減模擬時間，減少計算資源浪費。快速且準確的模擬結果可提供即時淹水資訊，供相關應變單位與民眾預作警戒及防護工作，達到減災之效果。

Global climate change has been increasing rainfall intensity and variability. The heavy rainfall and the special physiography including steep slope and easy precipitation-driven causes severe damage in Taiwan. Therefore, an efficient and accurate inundation model is developed for generating the inundation information to provide policy decision and reduce flood damage. The digital terrain model (DTM) and geographic information system (GIS) are employed to deal with the data of model. However, the remote sensing technique has obviously improved the spatial resolution and accuracy of digital terrain models in recent years. Despite advances in both computing software and hardware, high resolution flood modelling remains computationally demanding applying to a large study area. The multi-layered grid refinement can effectively reduce the computation time. Based on the thesis of 'Down-scale Inundation Simulation for Local Area of a Basin' by Chen (2008), this study revised adopted three layers grid refinement (600m, 200m and 40m) in flood modelling to provide a much higher resolution grid with less computational time. The three layers grid refinement was applied to two cases. The first was a synthetic one. It was a 10 km×10 km square-shaped enclosed area with a designed rainfall pattern. The second was a real case that the multi-layered model applied to Tseng-Wen river basin with typhoon Morakot event (2009), which triggered the worst flooding that ever happened on this island in half a century. The two cases were compared with accuracy and efficiency. The result of three layers grid refinement is close the level of fine-grid simulation in the global area. Multi-layered approach is not only considering the accuracy, but also raising the efficiency of simulation. The contribution of this study could provide more efficient and more accurate model for the enhance decision-making and reduce flood damage in flood emergency operation.