鄉鎮市區層級災害耐韌力指標關聯性與權重之DANP分析以及空間聚集性分析之應用

Relationship and Weights of Township Level Disaster Resilience Indicators and the Application of Spatial Clustering Analysis

國內外的相關研究對於面對災害的態度已由抗災的觀點，轉為耐韌力的思維與災害共處的調適作為，而各領域對於耐韌力的議題也漸臻成熟且被廣泛地討論。本研究透過文獻回顧先行釐清關於耐韌力的定義與概念，並廣泛蒐集相關研究中所提出的各類指標作為基礎，以建立適用於本研究分析之災害耐韌力指標系統；之後進行專家問卷以及採用決策實驗室分析基礎之網路層級分析法，以DEMATEL法分析指標的因果關係結構，藉由檢視指標間影響程度，以及利用矩陣運算取得指標間的相互影響關係（interfering relationship）及影響強度，同時產生DEMATEL影響矩陣，再透過納入此一影響矩陣的ANP分析求得指標間的關聯與影響權重（influential weights）。使用DEMATEL影響矩陣的ANP分析之主要優點是可以避免一般計算常態化超矩陣（super matrix）時，必須建立在每一個群組之權重都是相等的假設，而忽略群組間不同效應的問題。針對耐韌力的評估，本研究提出四個評估項目：「社會背景」、「經濟基礎」、「環境條件」、「應變能量」，每個項目之下各列出三個評估指標，共計十二個指標，透過DEMATEL分析中心度與原因度，瞭解各指標與項目間的影響力與與關聯性；之後利用DEMATEL矩陣的轉化進行DANP分析，找出各指標的權重，並將之應用於各鄉鎮市區的耐韌力評估。評估的結果發現，耐韌力較高的地區多位於台灣西部的主要都會區，而耐韌力較低者多位於山區與偏遠地區；本研究並以空間自相關技術應用於災害耐韌力的空間分析，以確認其空間聚集情形。

The recent research of disaster management indicates that the attitude of facing disasters is changing from “resistance” point of view to “resilience” and “adapting” behaviors of living among the possible risks. The issues of “resilience” are also getting matured and being discussed broadly among various disciplines. The definitions and concepts of resilience were clarified first by literature reviews, along with the collections of indicators from relating researches, in order to establish a disaster resilience indicator system that is feasible for this study. By collecting experts' opinions through questionnaires and adopting the analysis procedures of DANP (DEMATEL-based ANP) method, DEMATEL was used to investigate the interfering relationship among indicators, to find the central indicators to represent the effectiveness of factors, and to obtain the influence matrix. The ANP analysis was introduced the DEMATEL influence matrix to identify the connections of indicators and calculate the influential weights. The advantage of introducing the DEMATEL influence matrix is to avoid the equal-weight assumptions of clusters when normalizing the super matrix in ANP, thus it can take the effectiveness of clusters into account during the calculation. This study used 4 assessment items -- “social background”, “economic basis”, “environmental conditions”, and “response capacity”-- and 12 indicators to evaluate the resilience of local townships. By using DEMATEL method, the “Prominence” and “Relation” of items and indicators were identified for the relationship and effectiveness among them. The DEMATEL matrix was transformed for DANP analysis to find the indicator weights. By applying the results from DANP analysis, this study was able to examine the spatial distribution of townships' disaster resilience in Taiwan, and to verify the spatial clustering of disaster resilience through the application of spatial autocorrelation techniques.