| Abstract -- 
              There are many sediments will transport from the slopeland area 
              to the downslope area of watershed in Taiwan natural environment 
              each year, especially this phenomenon increased rapidly such as 
              landslide cause many sediment disasters after the Gi-Gi earthquake. 
              Landslide analysis include: the disaster influence area designation, 
              the residents evacuation route, beginning factors analysis, control 
              works planning, land use management, disaster evaluation and predication, 
              etc., but there are many influence factors will still cause decision 
              the time landslide difficulty and it need detail research for predication 
              landslide for protect human life in recently. This project used the Da-Hu landslide at Maioli county, Da-Pu landslide 
              at Chiayi county, Chiu-Fen-Er-Shan landslide at Nantou county as 
              the experimental sites to investigation the local rainfall intensity, 
              the magnitude of groundwater table raise and time of rising after 
              rainfall, rising segment and falling segment, etc., to simulate 
              this process with triangle groundwater table predication model. 
              We hope this model can be used for the time of landslide predication 
              and combine warning system equipment to avoid direct disaster impact.
 According to the investigation data, the rainfall intensity and 
              the magnitude of groundwater table raise relationship are closely, 
              but the investigation data are so few and each rainfalls are continuously, 
              thus we can not divide each rainfall influence, moreover the groundwater 
              table change and slope stability predication. It still need continuous 
              investigation and join another groundwater and geotechniqual tests 
              for examine more clear mechanism and carry out the practical concept 
              model.
 |