CAJ | 학술논문

以地震滑坡作为研究对象，选取“4?20”芦山地震中芦山县为研究区，结合多源数据，在相关分析后选取10个评价因子，分别是地面高程、坡度、坡向、斜坡形态、地层、斜坡结构、断层平均距离、水系平均距离、植被指数和地震峰值加速度，在数字高程模型基础上采用集水区重叠法划分斜坡单元，再对各评价因子重采样，进而利用基于遗传算法的神经网络算法构建地震滑坡危险性评价模型，完成地震滑坡危险性区划。将基于斜坡单元的危险性区划结果和基于格网单元的区划结果进行比较，结果显示滑坡正确率分别为96．6％和92．6％，斜坡单元的正确率较高；同时通过多组数据的受试者工作特征曲线分析本评价模型的不确定性，每组曲线位置及曲线下面积大小相当。
이지진활파작위연구대상，선취“4?20”호산지진중호산현위연구구，결합다원수거，재상관분석후선취10개평개인자，분별시지면고정、파도、파향、사파형태、지층、사파결구、단층평균거리、수계평균거리、식피지수화지진봉치가속도，재수자고정모형기출상채용집수구중첩법화분사파단원，재대각평개인자중채양，진이이용기우유전산법적신경망락산법구건지진활파위험성평개모형，완성지진활파위험성구화。장기우사파단원적위험성구화결과화기우격망단원적구화결과진행비교，결과현시활파정학솔분별위96．6％화92．6％，사파단원적정학솔교고；동시통과다조수거적수시자공작특정곡선분석본평개모형적불학정성，매조곡선위치급곡선하면적대소상당。
Earthquake‐induced landslide could cause serious damage ,which could be even worse than the quake itself especial in a mountain area . We select Lushan as the research area where Lushan earthquake occurred on April 20 ,2013 . This earthquake caused massive landslides that resulted in a tragic loss of life and economy .Combining multi‐source data , we select 10 evaluation factors after Pearson correlation analysis , including elevation , slope , aspect , shape , curvature classification , vegetation index ,distance from drainages ,slope structure ,lithology ,and distance from faults .Based on the digital elevation model ,we use watershed overlay method to plot slope units ,and resample the <br> evaluation factors .After multi‐group test ,we get the appropriate parameter value of the neural network optimized by Genetic Algorithm model . The risk zoning of earthquake induced landslides is then calculated by using neural network optimized by Genetic Algorithm .The same risk zoning model is used to slope units and grid units .The accuracy rate of slope units is 96 .6% ;while the accuracy rate of grid units is 92 .6% .In comparison with the grid units , the slope units are more accurate . Further ,we analyze the uncertainty in this evaluation model by using multi‐data receiver operating characteristic curve(ROC) .Each curve has the equal position ;and the areas under the line are almost the same .The result indicates that the proposed method has the advantages of high accuracy ,stable performance ,and small amount of data to process .It can be used for early warning and assessment of earthquake .