多元时间序列分析的滑坡演化阶段划分

滑坡的时间-位移曲线一般具有3个阶段特征，即初始变形阶段、等速变形阶段和加速变形阶段，不同演化阶段加速度具有不同的变化特点.目前往往是依据对加速度曲线特征的分析来人为划分演化阶段，缺少相应的理论支持和定量计算.针对上述问题，选取月降雨量、月库水位高程变化量对滑坡的累计位移建立多因素的时间序列预测模型.然后利用Chow分割点检验理论，以所建模型中F和LR统计量最大值点作为分割点对滑坡演化阶段进行划分.以新滩滑坡和三峡库区白水河滑坡为例，利用累计位移、降雨及库水位变化数据进行计算验证.结果表明，对多元时间序列模型进行Chow分割点检验可对滑坡的演化阶段进行准确划分，为滑坡的临滑预警预报提供重要判据. 

Evolution Stage Division of Landslide Based on Analysis of Multivariate Time Series

Curves of cumulative displacement vs. time of the landslide generally include the stages of the initial deformation, constant speed deformation and accelerated deformation. Different deformation stages have their own different acceleration characteristics. Now the evolution of landslide can be divided by analyzing the characteristics of acceleration curve. But it lacks theoretical basis and quantitative calculation. To the above problems, the rainfall of current month and reservoir level fluctuation of current month are selected as influencing factors, and the multivariate time series model is adopted to predict the cumulative displacement. Taking the maximum points of F statistic and log-likelihood ratio statistic of the multivariate time series model as segmentation points divides the evolution stages of landslide based on the theory of Chow breakpoint test. This model is used to deal with the data of cumulative displacement, rainfall and reservoir level fluctuation of Xintan landslide and Baishuihe landslide in the Three Gorges reservoir area. The results indicate that the Chow breakpoint test on the multivariate time series prediction model based on the inducing factors and the landslide displacement has certain validity and application value. This division of the landslide evolution stage can provide an important criterion for the early-warning prediction of landslide.