Using ARIMA model to establish landslide of rainfall alarm criteria in Lishan area

• Main Authors: Kai-Yuan Cheng, 鄭開元
• Other Authors: 蘇苗彬
• Format: Others
• Language: zh-TW
• Published: 2015
• Online Access: http://ndltd.ncl.edu.tw/handle/40780783790850557843

碩士 === 國立中興大學 === 土木工程學系所 === 103 === Lishan area is a typical landslide region, the groundwater level rises during heavy rain, resulting in a lower stability of slope, and thus increasing the possibility of collapse. This study uses the information of rainfall and ground water level change from the existing fully automatic monitoring station in Lishan region. Establishment of time series for the relationship between rainfall and groundwater level change are performed. The research method used is Autoregressive Integrated Moving Average model (referred to as “the ARIMA model”) proposed by Box and Jenkins in 1970. Univariate time series model of groundwater level is achieved by means of the change in groundwater level in Lishan during Typhoon Morakot period. Establishment of the relationship between the rainfall packed by Morakot and the change in groundwater level using transfer function for further forecasting. The torrential rain event of June 10, 2012 is applied to validate whether the model is in line with the trend or not. The best prediction model is tested with three different rainfall distribution tests to assess whether the criteria value shall be amended. The research results showed that not all the selected monitoring station records are suitable for forecasting the groundwater level change. The influence of hydrogeology, construction renovation and status of selected event can all affect the result. However, the feasible monitoring station with good relationship between rainfall and groundwater level change can be chosen to calculate endangerous rainfall value. It can accordingly apply to the slope stability analysis during the heavy rainfall and groundwater level rise, in order to modify the reference value of early warning systems on landslide region.