都會區淹水模式之比較與應用

Abstract

都會區藉由建置雨水下水道系統進行市區排水以避免產生積淹水之情況，但隨著都會區發展，建築物及不透水鋪面不斷增加，再加上近年來極端氣候發生之頻率有其增加之趨勢，因此時常有短延時暴雨或是颱風等強降雨發生，讓雨水下水道系統超出其可負荷之排水能力，因而使都會區發生淹水情形，造成經濟產業及民眾上的各方面損失與危害。 為了準確模擬淹水範圍以評估淹水影響與損失情況，國內外學者對於都會區淹水發展出許多種之方法與模式，而本研究將藉由四種模式對都會區淹水進行模擬，此四種淹水模式分別為單獨二維漫地流淹水模式(模式A)、雨水下水道系統與二維淹水模式之分離演算模式(模式B)、雨水下水道系統與二維淹水模式之半交互演算模式(模式C)、二維淹水模式結合雨水下水道模式之交互演算模式(模式D)，並藉由準確度、偵測率及預兆得分等方法，評估出何種模式對都會區淹水模擬有較佳之表現。 本研究以臺北市文山區木柵次分區為研究區域，並以2000年象神、2001年納莉颱風以及2012年6月12日及6月16日暴雨四場淹水事件做為淹水模擬事件，並且將二場颱風事件作為長延時降雨淹水事件，另二場6月強降雨作為短延時降雨淹水事件，以分析四種淹水模式在四場淹水事件之模擬情況及其適用之降雨事件，其分析比較結果顯示，在長、短延時降雨淹水事件下，都是以模式D有較佳的模擬結果，其準確度達到91%以上，偵測率達到70%以上，預兆得分達到44%；若要簡化淹水模式，在長延時降雨事件下，可以以雨水下水道系統與二維淹水模式之分離演算模式進行簡化，在短延時降雨事件下，可以以單獨二維漫地流淹水模式進行評估。

In urban areas, storm sewer systems have been established to avoid serious flood inundation induced by flooding events. However, as short duration rain events and typhoon events frequently occur in recent years, and buildings and impervious paving are dramatically increased during urban development, the drainage capacity of storm sewer systems is not enough, resulting in frequent urban flooding. Urban flooding events can bring serious damage to the public and the economy on various aspects. In order to accurately simulate urban flooding events for assessing the impact and loss of urban areas, many scholars have developed several methods and models to cope with these issues. Therefore, in this study, four different urban flooding models are used: the two-dimensional overland flow model (model A), the storm sewer systems model combining with two-dimensional overland flow model by using separated algorithm (model B), the storm sewer systems model combining with two-dimensional overland flow model by using half-interactive algorithm (model C), the storm sewer systems model combining with two-dimensional overland flow model by using interactive algorithm (model D). Then, the performances of these models are evaluated by using the threat score (TS), the probability of detection (POD) and accuracy. In this study, the Muzha area in Taipei city is selected as the study site. Four flooding events are chosen to conduct numerical simulations: Typhoon Xangsane (2000), Typhoon Nari (2001), the heavy rains on June 12, 2012 and on June 16, 2012. The above two typhoon events are regarded as long duration rain flooding events, and the latter two events are viewed as short duration rain flooding events to analyze the correlation between the four urban flooding models and the four flooding events for finding an appropriate selection regarding specific duration of rainfall. For long or short duration rain flooding events, the simulations derived by the model D are the best. The values of accuracy are over 91%, the values of probability of detection (POD) are over 70%, and the values of threat score (TS) are over 44% in all the simulations obtained by the model D. To simplify the urban flooding models, we can use the storm sewer systems model combining with two-dimensional overland flow model by using separated algorithm in long duration rain flooding events. We can use the two-dimensional overland flow model in short duration rain flooding events.