降雨入滲造成地下水位抬升，使邊坡土體重量增加且抗剪強度降低，常為邊坡破壞主因。USGS結合降雨入滲與無限邊坡模式之一維暫態模式TRIGRS假設每一網格的土體於該模式的環境下形同個別自由體，並未考慮土體內的地下水側向流補注，無法真實地反應地下水位。本研究改進李浩瑋(2011)結合TRIGRS與TOPMODEL之研究，將每單位小時由TRIGRS計算得之地下水位使用TOPMODEL修正，並將修正後地下水位當作下一小時的初始地下水位，在TRIGRS中繼續計算，直至降雨事件結束。 本研究以2004年艾利颱風之降雨事件，模擬大漢溪流域的匹亞溪集水區地下水位在淺層土壤中之變化，並且討論土壤飽和度分佈對誘發山崩之影響。結合TRIGRS與TOPMODEL連續推估的解釋山崩總體正確率為89.9%，成功率曲線下面積(AUC)為0.838。若僅使用TRIGRS模擬結果，其總體正確率為76.7%，成功率曲線下面積為0.793，使用TOPMODEL修正TRIGRS模式其總體正確率為85.5%，成功率曲線下面積為0.810。結果顯示結合TRIGRS與TOPMODEL連續推估模式更能有效解釋山崩分布。使用馬莎颱風降雨資料帶入新模式預測山崩之結果，總體正確率為95.1%，預測率曲線下面積為0.816；僅使用TRIGRS模擬結果，總體正確率為96.8%，預測率曲線下面積為0.763，亦為可接受之預估結果。;Rainfall infiltration is the main reason most of the shallow slope failure because the rise of the groundwater level will make the weight of the slope to increase and the shear strength to deteriorate. In the past, the USGS proposed TRIGRS model combining rainfall infiltration and slope stability which is the one-dimensional vertical infiltration transient model to analyze the groundwater level. The model assumed that soil body of each grid as an independent individual and didn’t consider of hydrological influence, so it can not reflect the groundwater level refined. In this study, we modify the method of Hao-Wei Lee (2011) which combined TRIGRS and TOPMODEL models to calculate the groundwater level. At first, using TOPMODEL model to correct the groundwater level for each hour which is obtained by TRIGRS model, and then, this revised groundwater level is treated as the initial groundwater level for the next hour, and the correction will continue to the end of the rainfall event. We use AERE typhoon rainfall event in 2004 to simulate the change of groundwater level of shallow soil layer in Piyaxi watershed of Dahanxi basin and discusses the effect of soil saturation degree distribution on induced landslide. The results show that, if the overall correct rate is 89.9% when the TRIGRS and TOPMODEL continuous estimation model are combined, the area under the success rate curve (AUC) will be 0.838, and if the overall correct rate is 76.7% when only the TRIGRS is used, the AUC is 0.793. Then the overall correct rate is 85.5% when use the TOPMODEL to correct TRIGRS model is used, the AUC is 0.810. It means the combination of TRIGRS and TOPMODEL continuous estimation model can display the more effective interpretation of landslide distribution. Finally, the Matsa typhoon rainfall data is put into the present model to predict the results of landslides, the area under the curve is 0.816, and the overall correct rate is 95.1%. The area under the curve is 0.763, and the overall correct rate was 96.8% when using TRIGRS model only.
