Research On The Susceptibility And Risk Assessment Of Geological Hazards In Changchun Based On GIS And Stacking Model

Changchun is located in the central and western part of Jilin Province.And it is the political,economic,cultural,and scientific center of Jilin Province.Geological disasters are increasingly restricting the development of Changchun.Therefore,it is important to evaluate the susceptibility and risk of geological hazards to Changchun.The susceptibility assessment of geological hazards is the basis for hazard assessment.And the reliability of the susceptibility evaluation results will affect the prediction of the risk of geological hazards.Therefore,finding a more accurate evaluation result of susceptibility is an important step to complete hazard evaluations.Machine learning models are widely used in the prediction of the susceptibility of geological disasters,but Stacking model,which fuses individual models to improve learning performance,have not been widely used in the prediction of geological disasters.Therefore,this article relies on the 2019 geological disaster hazard assessment project in Changchun and other relevant data to investigate and analyze the distribution of geological hazards in Changchun in detail.In order to analyze the distribution and development of geological disasters in Changchun,a Stacking fusion model was selected to evaluate the susceptibility to geological disasters.The results were compared with three basic models: support vector machine,artificial neural network and random forest.Based on the reliability and applicability of the model,the optimal evaluation model is selected to calculate the danger of geological hazards in Changchun with the disaster probability and the importance of rigon.And reasonable prevention and control suggestions are proposed.The main work has the following results:(1)The physical geography,socioeconomic,meteorological,hydrological and geological conditions of the study area were analyzed.And the distribution of geological hazards in Changchun City was summarized according to the causes and development characteristics of the geological hazards.According to field survey and available data statistics,there are 212 geological disasters in Changchun.Among them,173 collapsed disaster points,accounting for 81.60% of the total number of disaster investigation points;20 ground subsidence,accounting for 9.43%;10 debris flows,4.72%;5 landslides,accounting for 2.36%;4 unstable slopes,accounting for 1.89%.(2)Based on the analysis of the natural,geological and social conditions in the study area,14 factors that influence the hazards are selected by field surveys,zoning data,geographic resources websites and remote sensing images etc.And we analyzed the relationship between factors and the distribution of disasters.The matrix scatter plot and Spearman rank method were used to show the independence of the 14 impact factors.And then the single analysis method was used to show that the significances of the 10 impact factors are less than 0.05,indicating that there is a significant relationship between factors(the elevation,slope,lithology,distance to the river,distance to the fault,distance to the road,coal mining intensity,distance to the coal mine,NDVI and TWI)and the occurrence of geological hazards in study area.These 10 factors could be used as an impact factor to establish the assessment system to evaluate the susceptibility.(3)After the selection of factors,support vector machines,artificial neural networks,and random forests were used as the base models.The base model was independently trained and analyzed first,and then the stacking fusion model is trained.After that,we used the ROC curve and Kappa coefficient to analyze the reliability and applicability of the stacking fusion model.The ROC result shows that the accuracy rate of stacking model is 94.4%,and the prediction rate is 93.4%,which is higher than three basic models.Kappa coefficient of Stacking model is 0.731,which shows that the prediction results are highly consistent with the actual observation results.The analysis results show that the Stacking model can effectively improve the learning efficiency and accuracy.Finally,the prediction results of the susceptibility of geological hazards based on the Stacking fusion model are selected for risk analysis.(4)The past geological hazard activity level and factors are used to indirectly reflect the probability of geological hazards.Combined with the evaluation results of susceptibility and importance of region,the AHP method is used to evaluate the risk of geological hazards in Changchun.Four risk levels are divided: high,medium,low and low level.At last,some methods are put forward to deal with the hazards.