本研究選取瓦魯斯溪集水區及 2009 年莫拉克颱風為研究區域與事件,利用羅吉斯迴歸法及頻率比法兩種模式,配合地理資訊系統(GIS),經評估篩選後,選定(高程、坡度、坡向、地質、距水系距離、距斷層距離及累積雨量)等七項崩塌因子建立崩塌潛勢模式,之後加入敏感度分析作驗證。 研究結果顯示:本研究所採之兩種模式,經由誤差矩陣法驗證後發現準確率皆高達7成以上,皆具良好之預測效果,總正確率分別為 73.1 及 75.2 %,對瓦魯斯溪集水區崩塌預警教果有一定程度成效,並帶入地理資訊系統(GIS)繪製崩塌潛感圖;本研究區之崩塌因子敏感度排行依序為:高程>地質>坡度>距斷層>累積降雨>距水系>坡向,冀望本研究之崩塌潛感圖與崩塌因子敏感度排行,可供瓦魯斯溪集水區整體治理規劃經營與崩塌災害預警工作之相關單位作參考依據。
This study uses the Varus River watershed area and the 2009 Typhoon Morakot as a study area and an event, respectively. Through the use of logistic regression analysis and the frequency ratio method models, combined with the geographic information system (GIS) along with evaluation and screening, seven factors were found to be related to landslides (elevation, angle of slope, slope direction, geology, distance from the water system, distance from the fault, and accumulated rainfall). These factors were used to establish the potential landslide model, and then the sensitivity analysis was added for verification. The results of the study demonstrated that the both models adopted in this study were verified by the confusion matrix method and were determined to have an accuracy of more than 70%, and so have good predictive effects. The total correctness rates were 73.1 and 75.2%. A warning of landslides near of the Varus River watershed would have a certain degree of effectiveness, as the Geographic Information System (GIS) could be used to draw a potential landslide area map. The sensitivity of the potential landslide factor ranking in this study is as follows: elevation > geology > angle of slope > distance from the fault > accumulated rainfall > distance from the water system > slope direction. It is hoped that the potential landslide area map and landslide factor sensitivity ranking from this study are make available as a reference to the relevant parties of the Varus River watershed area in terms of overall management planning operations and landslide disasters early warning systems.