區域淹水災害風險評估及其未確定性分析

Abstract

台灣是一個淹水災害頻傳的國家，淹水災害風險管理日益受到重視，風險分析是風險管理的首要工作，而淹水災害損失推估又是淹水災害風險分析的基礎，本研究乃從分析淹水災害損失資料著手，建立區域淹水災害損失推估機制，以進行區域淹水災害風險分析。 以往一般所謂的風險多著重於事件發生的機率，但是聯合國救災辦公室對淹水災害風險的定義，是以淹水災害損失的期望值當作淹水災害風險的指標。但單一淹水災害事件的損失期望值，並不足以代表區域之淹水災害風險，區域年平均淹水災害損失應是一個比較適當之區域淹水災害風險指標。 區域淹水災害損失推估有許多方法，本研究中選擇並建立淹水深度-損失曲線，配合地理資訊系統與網格式資料建立區域淹水災害損失推估模式。應用網格式資料推估區域淹水災害損失，亦可以保留淹水災害損失之空間分佈特性，研究中亦建立區域淹水災害風險圖，進行區域淹水災害風險空間分析。 由於對淹水水理之瞭解尚未完全，加上淹水災害損失資料本身具有相當之變異性，淹水災害損失調查與淹水潛勢等資料都具有未確定性，研究中乃提出災損評估之機率分佈函數，並利用系統模擬研討區域淹水災害損失與風險之未確定性。 研究中以汐止七堵地區為研討區域，該區域是員山子分洪工程之受益區，本研究利用有/無員山子分洪工程之區域淹水災害潛勢資料，進行區域淹水災害風險推估、區域淹水災害風險空間分析、員山子分洪工程效益評估、員山子分洪工程效益空間分析、區域淹水災害風險之未確定性分析與區域淹水災害風險未確定性空間分析，分析所得之各項資訊，可作為區域淹水災害管理之決策參考。

Flood risk management is an important issue in Taiwan because the island suffers from frequent flood attacks especially for the last decade. Flood risk analysis is the starting point of flood management and flood loss assessment is the basis of flood risk analysis. This paper focuses on the studies of regional flood loss and flood risk assessment as well as the uncertainties associated with them. Risk is usually thought of as the occurrence possibility of an event. Flood risk is defined by the United Nation disaster relief office as the expected value of flood loss. Since multiple events may occur during a year, the expected loss of a single flood event can’t be considered as the regional flood risk. Regional average annual flood loss of events with different occurrence probabilities may be a more proper index to the regional flood risk. There are different approaches to the regional flood loss assessments. The depth-damage curve was used in this study to assess regional flood loss. A grid-based regional flood loss model was proposed which takes into account the variations of socio-economic activities among different grid cells. Through the use of a set of flood hazard potential data with different return periods, regional flood risks were presented as flood loss-exceedance probability (EP) curve. The area under the EP curve represents the expected annual flood loss for the region. A flood risk map was also built with this grid data approach which made the spatial risk presentation n possible. Uncertainty is an inevitable part of the assertion of knowledge. The result of flood risk analysis also had uncertainty. The variations of flood loss data at the same flooding depth were studied. Confidence intervals for regional flood loss and flood risk were constructed through a system simulation approach using the flood loss probability density functions. Townships of Si-Jhih and Ci-Du in northern Taiwan were used as a study area for the demonstration of these proposed algorithms.