初探風險指標與淹水事件之關聯性──以新北市為例

Exploring the Relationship between Risk Indicators and Flooding Events – A case study in New Taipei City, Taiwan

在全球氣候變遷的影響下，淹水災害越發頻繁，增加洪患管理的難度。為了系統性地解構災害事件，災害風險評估是一個重要的分析手段，其分析成果可以做為洪患管理的參考，爰此，淹水災害的風險評估是一個重要的課題。本研究欲探討風險指標對於淹水事件的發生機率是否會產生影響，並且探討風險指標對於災害事件的影響是否存在空間的異質性。本研究首先對風險指標進行空間群聚的分析，描繪風險指標在研究區的空間群聚狀況；接著透過羅吉斯迴歸研析風險指標對淹水事件的關聯性；最後透過地理加權羅吉斯迴歸探討影響能力的空間異質性。研究成果發現羅吉斯迴歸模型的預測有約60.9% 的正確性，其中淹水潛勢面積比例對於淹水災害的發生有最高且正向的影響力；弱勢族群對於水災的影響屬於負相關。而地理加權羅吉斯迴歸的模式表現比起羅吉斯迴歸還好，可以顯著提升模式的配適性，並對影響力空間異質性的空間樣態進行初探。本研究之成果可做為未來淹水災害的調適策略之參考。

Global climate change has affected our environment and increased the frequency of flooding disasters. To strengthen disaster management, flood risk assessment provides a systematical analysis to assess the elements of flooding disaster. The risk assessment for flooding disaster is an essential issue for disaster reduction. This study intends to explore whether risk indicators can influence the probability of occurrence of floods. This work first applies hotspot analysis to reveal the spatial distribution and cluster of each risk indicator. Logistic regression is then adopted to build a model and explore the relationships between risk indicators and flooding events. Finally, geographically weighted logistic regression is used to analyze the spatial heterogeneity phenomenon in the last part. Analytical results indicate that the logistic model has approximately 60.9% accuracy in the study area. A positive correlation is found between flooding events and inundation potential area percentage, while minority groups have a negative correlation with flooding events. Moreover, the geographically weighted logistic regression approach has better goodness-of-fit than the original one. This investigation also reveals that the spatial heterogeneity phenomenon still exists in the relationships between risk indicators and flooding events. The results of this work can be utilized as a reference for further flooding adaptation strategies.