雨量站分群與區域淹水指標評估

Rainfall Stations Clustering and Regional Flood Disaster Indicator Assessment

陳智謀(Zhi-Mou Chen)

指導教授 ： 葉一隆 ； 陳庭堅

國立屏東科技大學/工學院/土木工程系所/博士(2015年)

https://doi.org/10.6346/NPUST.2015.00300

摘要

由天然災害研究統計可知，氣旋及暴雨所造成的淹水災害事件威脅人類生命與財產安全。有鑑於此，透過劇烈降雨事件之雨量分析，以及結合颱洪災害事件調查成果，可做為擬定淹水災害防救對策。
本文以屏東縣為研究區域，以淹水災害分析為主軸，探討颱洪災害期間雨量分布特性及建立淹水指標。
對於颱洪災害期間的雨量資料，本文以群集理論進行分析，先利用反距離權重分析法進行空間雨量資料補遺作業，藉此取得各次災害期間之空間降雨分布資料，再利用分析所得之雨量資料執行雨量站群集計算。本文群集分析是採用兩階段之方式計算，先以Ward’s 群集法獲取各雨量站之間的分群距離，再利用均方根標準偏差值決定最佳之分群數，最後再以K-Means群集法及GIS確立各雨量站之分群結果。
對於淹水指標評估，本文利用多屬性決策方法，主要彙整項目包含：歷次淹水災害面積、淹水潛勢分析、淹水致災原因及淹水環境特徵評估等項目。分析程序為：1. 依各項資料的特性進行因子標準化處理；2. 分析各因子之熵值，再以熵值分配其權重；3. 透過熵權重值將各因子整合成淹水指標。
由結果可知，透過雨量站分群計算，相同分群之雨量站具有降雨時空分布較為相近之特性，若雨量站資料因故遺失時，則可以利用相同分群的雨量站進行反距離權重計算，以達到雨量資料補遺的目的。此外，透過淹水指標評估可知，淹水指標與過去颱洪災害淹水範圍及淹水災害潛勢均呈正相關，由此指標之分級則可反應各區域淹水災害危險度之差異。

關鍵字

群集分析 ； 反距離權重分析法 ； 淹水指標 ； 熵值

並列摘要

The natural disaster statistics shows that cyclones and heavy rain caused flooding disaster, hence threaten people’s lives. Therefore, the combination of the analysis of rainfall amount during intense rainfall events and flood disasters survey results to develop flood disaster prevention and relief measures, which is one of the best way to respond the uncertainty of disaster events.
In this study we selected Pingtung County as the study area and focused on flood hazard analysis to investigate rainfall distribution and to develop flood indicator during typhoon and heavy rainfall events.
For the rainfall data during flood disaster event, this study employed cluster analysis method. Firstly, we taken inverse distance weighting method to add rainfall to station having loss rainfall, then rainfall spatial distribution can be obtained during disaster events. The rainfall data was used to conduct clustering of rainfall stations. The clustering analysis taken two-step method that Ward’s clustering method was used to obtain clustering distance of rainfall stations and values of mean square standard deviation were used to decide optimal clustering group. Finally, K-Mean clustering method and GIS were used to establish clustering result.
This study used multi-attributes decision method to assess flooding indicators that assessed items included area of individual flooding disaster, flooding potential analysis, reasons causing flooding, and flooding environmental characteristics. The following steps were taken for the analysis: (1) Normalization of the collected flooding data with the factors obtained in accordance with the characteristics of the various data. (2) The normalization factors were analyzed for their entropy, which determined a weighting value. (3) Various factors were integrated into the flooding disaster indicator based on their weighting values.
From the cluster analysis results showed that the rainfall spatial and temporal distribution was similar in the same clustering group. If rainfall data was lost in a station due to any reason, we can carry inverse distance weighting calculation for addendum rainfall data to the station with the rainfall data in the same clustering group. In addition, the flooding disaster indicator had a positive correlation with the flooding area of past flooding disaster events and flooding disaster potential. The ranking of established indicator can distinguish the extent of flood hazards.

並列關鍵字

Cluster analysis ； Inverse distance weighting method ； Flood disaster indicator ； Entropy

參考文獻

9. 童慶斌，楊奕岑，2004，「氣候變遷對臺灣水文環境之衝擊」，全球變遷通訊雜誌，第44期 第1-8頁。

46. 陳亮全，謝龍生，李欣輯，黃俊宏，2010，「莫拉克颱風勘災機制之規劃」，國研科技，第25期，第34-39頁。

47. 王俞婷，謝龍生，李欣輯，林宣汝，黃俊宏，葉克家，2011，颱洪易致災區域調查技術與應用，國研科技，第30期，第77-89頁。

6. UNISDR, 2013, 2013 Global Assessment Report on Disaster Risk Reduction, Available at: https://www.unisdr.org/we/inform/gar, Accessed on 2016/07/10.

8. Frederick, K. D., and Major, D. C., 1997, “Climate Change and Water Resources,” Climatic Change, Vol. 37, No. 1, pp. 7-23.