Research On Spatial Distribution Characteristics Of Geologic Disaster And Landslide Susceptibility Assessment In Huoxi Coalfield

Landslide,one of the most influential geological disasters,not only brings great damage to land cover and ecological environment,but also seriously threatens people’s lives and property.As the most important coal-producing region,Shanxi Province distributes extensive geologic disasters such as landslide due to mass mining of coal resources.Taking one of the six major coalfields —Huoxi Coalfield in Shanxi Province as the research area,this paper conducted numerical modelling and quantitative evaluation to landslide susceptibility,using remote sensing(RS)and geographic information system(GIS)method and based on the analysis of the spatial distribution characteristics of geologic disaster.First,spatial distribution characteristics of geologic disaster were studied based on the spatial statistics.Then,the correlation of spatial distribution among the landslides and evaluation factors was analyzed.Further,the evaluation models of landslide susceptibility were built,whose accuracy was compared accordingly.Finally,the distribution of landslide susceptibility was mapped using the best model in Huoxi Coalfield,and the quality of the map was evaluated.Through this study,the following conclusions can be obtained:(1)The spatial distribution characteristics of geological disasters in Huoxi Coalfield were analyzed using the nearest neighbor indexes(NNI),aggregation index(z-score value),more distance spatial classification analysis function value(Ripley K),the box-counting dimension and information dimension respectively.NNI value reflects random pattern distribution of geological disasters is;z-score value shows obvious aggregation characteristics of geological disasters distribution;Ripley’s K value proves the above randomness and clustering features of the spatial distribution of geological disasters,so it has spatial scale dependency characteristics;the methods of box-counting dimension and information dimension show that the spatial distribution shape of geological hazards in the Huoxi Coalfield have obvious fractal characteristics,which is affected by the number of geologic hazard.(2)The correlation characteristics between evaluation factor(landform,geology,hydrology,ground cover,and human activity)and the spatial distribution of landslides has been carried by using RS data source and GIS spatial analysis method.For the landform factor,landslides are mainly distributed in the low mountains,hills,the low and mediun mountains,whose elevation between 700-1500 m;the sunny slope of steep slope,steep hill and slope of study area;and the area which surface curvature and the section curvature value since-1 to 1.For geological factors,landslide occurs mainly in the shale sand,coal,limestone stratum and area that close to fault.For the surface cover factor,landslides are mainly distributed in industrial and mining land,cultivated land,grass areas and some area where normalized vegetation index value is less than 0.3.For human activity factor,landslides occur mainly within 500 m from the road and mining disturbance area.(3)Based on the method of cross check constructed three kinds of landslide susceptibility assessment model(Logistic regression model,artificial neural network and support vector machine(SVM)model),and based on the accuracy of fitting;the receiver-operating characteristic(ROC)curve;area under curve(AUC)value and sensitive index;the precision of three models got quantitative assessment.The results of fitting accuracy,the logistic regression model,artificial neural network and support vector machine(SVM)model,respectively,in the modeling phase were 74.26,73.52,87.22,in the validation phase were 69.68,68.2,70.12;using the predicted value from the specificity of the point value and model in modeling and validation phase,drawing ROC curve about three models,and calculating the corresponding AUC value based on the curves.Logistic regression model,artificial neural network and support vector machine(SVM)model in the modeling phase,respectively,were 0.807,0.798,0.914,validation phase were 0.804,0.797,0.917;sensitivity index of three models are: 70.17,66.05,74.10.Through the above comparison,support vector machine(SVM)model for Huoxi Coalfield is the optimization model for landslide susceptibility assessment.(4)Quantitative analysis for evaluation factors about the manifestation in landslide susceptibility assessment model is carried on.Based on the optimal model--support vector machine(SVM)model,first,evaluating the precision about the simulation results of the evaluation factors under different spatial resolution,by selecting different spatial resolution from 30 m to 500 m shows that when the spatial resolution is 80 m the highest accuracy could be got,then the importance of each evaluation factors in the model were compared,the results show that three factors,the most relevant for landslide susceptibility assessment,were followed by lithology,distance from the road and slope.(5)Based on the evaluation factors which spatial resolution is 80 m,obtained landslide susceptibility map about Huoxi Coalfield through support vector machine(SVM)model,and the landslide susceptibility of HuoXi Coalfield can be divided into four scales: very high,high,medium and low by the method of quantile law.The results show that the distribution about the number of landslides points in the very high,high,medium and low were 235,87,16,and 0,accounting for the total number of landslides points,respectively,were 69.53%,25.74%,4.73% and 0.Therefore,this study obtained the result accuracy of landslide susceptibility assessment in Huoxi Coalfield is above 69.53%.This study obtained the spatial distribution characteristics of the Huoxi Coalfield geological disasters and the quantitative evaluation of landslide susceptibility constructed based on the optimization model.It can not only provides reference for the investigation about artificial slope in research area,land development and consolidation,regional optimization layout geographical conditions monitoring and the rational mining coal resources,but also provides reference value for the related research in other similar coal region and management work.