Abstract: Landslide displacement in Three Gorges Reservoir area is of periodicity due to water level change, rainfall and so on. Based on the time series analysis, landslide displacement can be divided into the trend displacement reflecting the long-term trend of landslide, which is the response of geologic structure; and the periodic displacement reflecting the volatility of landslide, which is mainly affected by external factors such as rainfall. Taking Taping landslide in Three Gorges Reservoir area for example and considering the influences of water level change and rainfall, the trend displacement and periodic displacement are evaluated by Hodrick-Prescott (H-P) filter forecasting method. Difference auto-regressive integrated moving average (DARIMA) model is utilized to smooth the curve of trend displacement, and then compute the predicted value of trend displacement. Vector auto-regressive (VAR) model is used to predict the periodic displacement. The overall predicted displacement is obtained by adding the predicted values of trend displacement and periodic displacement, which is compared with the monitoring displacement and one predicted by other forecasting methods. The results show that the predicted displacements by this proposed method are in better agreement with the monitoring data; the proposed comprehensive model can better reflect the trend and volatility of landslide displacement.

Key words: landslide, displacement prediction, time series, H-P filter method, auto-regressive integrated moving average (ARIMA) model, vector auto-regressive (VAR) model