Dynamic Evaluation Of Landslide Risk Based On Integrated Monitoring Of Satellite,Airborne And Ground-based Data

Landslide is a common geological disaster.In terms of the intensity of energy conversion,it belongs to the transformation of gravitational potential energy into kinetic energy.From the cause of occurrence,the occurrence of landslide is the result of the combination of control factors and predisposing factors.Because landslide is a complex nonlinear system,landslide risk evaluation is an important means of landslide geological disaster management and monitoring,which can provide decision makers with decision support and provide the public with the distribution of disaster occurrence probability.Because of the complexity of the landslide,a single monitoring method can not meet the requirements of monitoring accuracy and monitoring intensity.The integrated monitoring of satellite,airborne and ground-based data emphasizes the diversity of monitoring methods.The data includes optical remote sensing image data,microwave remote sensing image data,DEM data generated by InSAR technology,UAV remote sensing data,geological survey data and geotechnical test data.The mathematical nature of integration is the fusion of heterogeneous data.The paper uses multi-source data to study the landslide risk assessment model at different scales,including the Xiangjiatun landslide,the Wanzhou section of the Three Gorges reservoir area and the entire Three Gorges reservoir area.The main results of the paper are as follows:(1)A landslide space database was established for the Three Gorges reservoir area,the Wanzhou section of the Three Gorges reservoir area,and the bank section of the Wanzhou district of the Three Gorges reservoir area.The PostgreSQL database is used as the back-end database,and QGIS is used as the desktop platform.The seamless connection between the PostgreSQL database and QGIS is used to realize the convenient and quick interaction between the database data and the operators.(2)For the single landslide,a semi-quantitative evaluation model of the single landslide based on ArcGIS Engine and Visual Studio was developed.Taking the Xiangjiaao landslide as an example,the stability of the landslide under different construction conditions was analyzed.Finally,the evaluation result of the semi-quantitative evaluation model of the landslide was obtained.After calculation,the stability coefficients of Xi-angjiaao landslide under natural conditions,heavy rain conditions and heavy rain and earthquake conditions are 1.26,0.98 and 0.73,respectively.The Monte Carlo simulation calculation shows that the probability of failure of the landslide in the natural state is 0.809.Finally,the risk probability of using the semi-quantitative evaluation model for the single landslide risk is 0.29,and the risk level is medium.(3)For local scales,taking the reservoir bank section of Wanzhou District in the Three Gorges Reservoir Area as the research object,the problem of imbalanced landslide sample in landslide susceptibility evaluation was studied.In the past studies of landslide susceptibility evaluation,the susceptibility calculation was usually regarded as an ordinary binary-classification problem.However,given the size of study areas,the number of landslide and non-landslide samples tends to vary widely.Direct classification calculations result in poor model performance,as the model has a high Accuracy rate but a low Recall rate.Aiming at this problem,the landslide susceptibility calculation process is regarded as an imbalanced binary-classification problem,and Logistic Regression model(LR),GBDT model(Gradient Boosting Decision Tree,GBDT)and weighted GBDT model(Weighted GBDT,Weighted GBDT)are established.And the performance evaluated using the the ROC curve,AUC value,and the Recall value as evaluation indexes.The study found that the LR model has the smallest AUC value and Recall value of 0.845 and 0.004,respectively.The AUC value and Recall value of the GBDT model are 0.976 and 0.426,respectively,while the Weighted GBDT model has the highest AUC value and Recall value,that is 0.977 and 0.845 respectively.The Recall value characterizes the proportion of landslide samples being correctly classified.From the Recall value,it can be seen that the LR model has a very poor discriminating ability for landslide samples,while the Weighted GBDT model has the best discriminating ability.The comparison with the actual landslide position also proves that the Weighted GBDT model has the best performance.(4)For regional scale,taking the Three Gorges reservoir area as the research object,the dynamic evaluation of landslide risk based on big data cloud platform is studied.The geological structure of the Three Gorges reservoir area is complex and diverse.Because of the cyclical fluctuation of the reservoir water level and abundant rainfall,the geological disasters in the Three Gorges reservoir area occur very frequently.The state invests a large amount of manpower and material resources in the prevention and control of geological disasters in the Three Gorges reservoir area every year,all for the purpose of disaster prevention and mitigation,and to safeguard the lives and property of the people in the reservoir area.However,since there are as many as 19 administrative districts involved in the Three Gorges reservoir area,traditional monitoring methods are often limited to one city or region,and it is impossible to form effective monitoring and management of the entire Three Gorges reservoir area.This paper uses Google Earth Engine cloud platform as the carrier of multi-heterogeneous data,and uses the whole Three Gorges reservoir area as the research area to develop a dynamic evaluation model and system for landslide risk based on big data cloud platform,realizing the dynamic risk forecasting of the Three Gorges reservoir area.The system implements landslide hazard prediction based on two methods,the Antecedent Rainfall index(ARI)and the empirical rainfall threshold.At the same time,a rapid assessment of landslide vulnerability based on nighttime remote sensing data was achieved.(5)Finally,taking the “831”rainstorm landslide occurred on August 31,2014 in Chongqing as an example,the spatial and temporal changes of rainfall and landslide risk in the Three Gorges reservoir area from August 28,2014 to September 3,2014 were analyzed.The study found that the maximum daily rainfall on August 30,2014 and August 31,2014 reached 138 mm and 104 mm respectively.The peak rainfall appeared in the northwest of Fengjie County,south of Wuxi County,northeast of Yunyang County and east of Kai County.There is a good agreement with the location of the “831”storm landslide event.At the same time,using the landslide hazard map of the Three Gorges reservoir area,it was found that the landslide risk level in the above areas was higher on August 31,2014,and the landslide risk level remained high from September 1,2014 to September 3,2014,which is because the hysteresis of the impact of cumulative rainfall on landslide risk.