Abstract: In order to more realistically describe the explicit relation between the displacement of the reservoir bank accumulation layer landslide and its influencing factors, and overcome the shortcomings of the existing superimposed component response model which is difficult to construct the displacement regression function, a new decomposition mode—fluctuation ratio is defined, while the displacement increment is expressed as the product of the fluctuation ratio and the trend item increment. Wangyemiao landslide, a typical accumulation landslide in the Three Gorges Reservoir area, is investigated as an example. The lowpass FFT filtering method is used to obtain the trend term displacement component. The monthly rainfall, the monthly reservoir water level and the previous month's fluctuation ratio are determined as the core factor combination. The K-means algorithm is used to establish the correlation function between each single factor and the fluctuation ratio. Then a multi-factor generalized model composed of the evoked factor term and the control term is proposed. Compared with the actual monitoring displacement and the values predicted by several other methods, our model has better prediction effect and simple operation; furthermore, the displacement components can be clearly explained. Four other typical landslides in the Three Gorges Reservoir area are selected for comprehensive comparative study, the results further verify its rationality. This method can accurately reflect the comprehensive response of landslide displacement with time lapse, multi-factor influence and state constraint, which provides a new idea for studying the evolution mechanism and time-space prediction of this type of landslide.

Key words: fluctuation ratio, multi-factor, reservoir bank landslide, displacement analysis, evolutionary process