A Landslide Displacement Prediction Research Based on Optimization-parameter ARIMA Model Under the Inducing Factors

Induced by many factors such as rainfall, the landslide displacement prediction is a complex nonlinear dynamic system. By using Baijiabao landslide monitoring data in three-gorge reservoir, we get the characteristics of landslide evolution to extract the related influence factors of landslide deformation. It was found after analysis that the most leading accumulation was rainfall; Based on the mathematical of Arima model, introducing monthly cumulative rainfall to evaluate seasonal trend as model parameter optimization, using as many as 72 monthly relative displacement data to fitting and prediction, finally the average absolute error and the correlation coefficient between predicted results and the measured values are 2.873 and 0.983. Results show that the optimized parameters of the model improve the accuracy compared with the traditional experience method. Under the comprehensive effects of rainfall and other external inducing factors, the accurate forecasting of landslide displacement is really a complicated dynamical system problem. Firstly, we use monitoring data of Baijiabao landslide in the Three Gorges Reservoir Area as an example, form a basically analysis of real-time characteristics of landslide evolution in each stage of its development, then extract the most relevant factors which most affect the deformation of Baijiabao landslide, this process shows that the Baijiabao landslide belongs to a type of rainfall-induced colluvial slope; Secondly, we use the autoregressive integrated moving average model as a method for fitting and prediction for monitoring data, and the seasonal trend parameters of the model are also evaluated by the seasonal factor of monthly cumulative rainfall data at the same time. Finally, we build a study on fitting and prediction of landslide relative displacement data in 72 months, the average absolute error of the model results and the measured values are 2.873, and the modal correlation coefficient reaches 0.983 which is very accurate. This research shows that: when compared with the traditional experience method, the optimal parameter model is more in line with the general law of landslide deformation in each stage of actual development.