降雨誘發山崩是台灣目前主要的山崩災害，因此探討不同集水區之降雨量與誘發山崩之關係有其重要性。近年山崩潛感分析開始加入雨量因子進行統計，為瞭解雨量因子是否有其物理意義，因此挑選若干個設有雨量站的小集水區，觀察降雨量與誘發山崩發生率之關係。本研究在全台灣的沉積岩、板岩、變質岩三個不同的地質區中，選取適當的次集水區作為分析單元，測繪不同降雨事件之誘發山崩，並分析誘發山崩發生率與雨量因子之關係。降雨因子包括：最大時雨量、總雨量、平均降雨量強度、降雨延時以及最大3、6、9、12、24 小時雨量。誘發山崩發生率則以崩壞比表示；某一降雨事件分析單元內誘發山崩面積與該單元面積之比率稱為崩壞比。每一降雨事件之崩壞比與某一雨量因子之關係以最小平方法進行線性一次迴歸，發現各研究區R2 值結果良好，雨量因子與崩壞比具有正相關性，證實雨量因子作為山崩潛感模型促崩因子有其物理意義。本研究採用R2 值達0.5 以上之降雨誘發山崩崩壞比曲線，將曲線於橫軸之截距定義為該研究區降雨門檻值，發現各研究區的確有門檻值現象存在。同時證明，利用線性一次迴歸已足夠表示降雨量雨誘發山崩發生率之關係。本研究發現各地質區雖無統一的崩壞比曲線斜率及降雨門檻值，但是相鄰或者距離較近的地區較有相似的崩壞比曲線斜率及降雨門檻值。Rainfall-induced landslides is a major landslide disaster in Taiwan. Tostudy the relationship between rainfall and induced landslides at differentdrainage basins is important. In recent years, landslide susceptibility analysisbegan to use rainfall as a triggering factor in statistics. For realizing thephysical meaning of rainfall factor, we select a number of small drainage basinswhere a rainfall station in available to observe the relationship between rainfalland the occurrence rate of induced landslides. This study selects 17 study areasto map induce landslides of each rainfall events and to analysis the relationshipbetween rainfall and the occurrence rate of induced landslides from thesedimentary terrain, the slate formation and the metamorphic complex inTaiwan. Rainfall factors include maximum hourly rainfall, total rainfall,average rainfall intensity, rainfall duration, and maximum 3, 6, 9, 12, 24 hourlyrainfalls. We present the occurrence rate of induced landslides by the landslideratio; the ratio of landslide area and total area in each study area.The correlation between the landslide ratio and a rainfall factor is positive.This supports that rainfall as a triggering factor used in a landslidesusceptibility analysis is physically meaningful. This study uses a simple linearregression model to fit the relationship between the landslide ratio and a rainfallfactor in each study area. The goodness of fit for the regression model isgenerally acceptable. We selected those models with R2 value larger than 0.5 tofind out the intercept of the best fit line and the horizontal axis and to define a rainfall threshold. At the same time, we found that the use of a simple linearregression model is good enough to interpret the relationship between rainfalland the occurrence rate of induced landslides. Finally, we observe the landslideratio curve and the rainfall threshold in each geologic province. Although thelandslide ratio curve and the rainfall threshold are not consistent in a wholegeologic province, they are similar in the adjacent or the closer study areas.
