STUDY ON LANDSLIDE EXPLAINABLE DISPLACEMENT EARLY WARNING THRESHOLD MODEL BASED ON TIME SERIES BIG DATA

China has complex geological conditions, strong tectonic and seismic activities, and uneven distribution of rainfall in space and time, resulting in many potential landslides, such as, wide distribution, heavy disaster and great risk. Monitoring and warning is an important means to slow down or reduce the risk of landslide disaster. The study of landslide explainable displacement early warning threshold model based on time-series big data is an important frontier in the field of geological disaster monitoring and early warning. The main scientific problem of the project is how to further improve the scientific setting of landslide early warning threshold model under the condition that detailed survey data and long time-series observation data are generally lacking. To this end, the project takes about 39,000 landslide automatic monitoring points covering 16 key geological disaster prevention and control areas in China as data sources, and takes deformation landslide displacement sample construction, quantitative evaluation of displacement prediction results and the study of landslide explainability early warning threshold model of "geological model + data-driven" as a breakthrough. It is proposed to use machine learning algorithm, correlation analysis, numerical simulation and other methods to establish a sample database of landslide long time series displacement monitoring data, build a landslide explainable displacement prediction model based on machine learning algorithm by region and type, and finally form a landslide explainable early warning threshold model of "geological model + data-driven".