集合住宅淹水深度損失曲線之不確定性分析及其應用

Abstract

根據流行病學災害研究中心之統計資料顯示，全球洪水災害不僅發生頻繁，且受影響的居民相當多，因此災害的風險管理日益受到重視，風險分析為風險管理的首要工作，而淹水災害損失推估為災害風險分析的基礎。本研究乃針對淹水災害損失推估方式進行研討，並考慮建立方式中之不確定性，以建立考慮不確定性之淹水災害損失的推估方法論，最後再進行住宅區之淹水災害風險分析。 目前淹水損失推估的方式有很多種，其中以淹水深度損失曲線法為最常用的方式。淹水深度損失曲線建立的方式大致可分為問卷調查法與合成曲線法。由於問卷調查法除了需要花費大量的人力、物力外，且受限於需有歷史淹水資料，故本研究選擇合成曲線法，並修正傳統上沒有考慮不確定性的問題。 本研究以汐止市為研究區域。整個研究架構大致上可分為資料蒐集、模擬模式、機率分析、危險度分析與結果評估五大階段。首先藉由統計要覽的資料中模擬出五千個家戶，並利用納莉颱風淹水調查資料與國稅局損失申報資料的分析結果，求得各項設備於各淹水深度下的損失機率與各項設備的損失機率密度函數後，再以蒙地卡羅進行模式模擬，以求得各淹水深度區間的淹水損失機率密度函數，並透過機率分析推得不同超越機率下各淹水損失與淹水深度的關係圖。而後將分析的結果進一步應用於住宅區平均每年淹水損失之推估，所推得的年平均淹水損失為一機率密度函數，最後再將此機率密度函數應用於防洪準備金的規劃。相較於傳統淹水深度損失曲線不考慮不確定性的方式，本研究結果應用於防災規劃時具有較大的彈性。

Statistical data from the Center for Research on the Epidemiology of Disasters show that flood disasters occurred frequently and affect a lot of residents. Therefore, risk management plays an important role. In risk management, the flood damage assessment is the basis of risk analysis which is the first work. This research focuses on the ways of flood damage assessment and considers uncertainty to build the methodology of flood damage assessment. Finally, this study estimates the risk of residential area. At present, there are many ways to estimate flood damage, and flood depth damage curve is the most common one. The curve can be constructed in two ways, Questionnaire and Synthesis. Because questionnaire costs a lot of manpower and money and is restricted to having history flood damage data, the method of synthesis is applied in this research. This method considers uncertainty which was not included traditionally. The Shi-Ju City in Taipei Metropolitan was used as a demonstration area for this study. First, this research analyzes the declaration data for individual damage in disaster of National Tax Administration and the Na-Li typhoon damage data. Then the loss density functions of each facility and the damage probability of each depth will be estimated. Secondly, Monte Carlo simulation will be used to determine the probability distribution of damage under a specific flood depth. Thirdly, the probabilistic model will be applied to establish the damage-depth curve under different exceeding probabilities. Finally, the result will be applied to estimate the average annual damage of residential area. The results show that average annual damage is a probability density function. The way of considering uncertainty in this research is more flexible than the traditional way is when being applied to risk management.