基于Random Forest和AHP的贵德县北部山区滑坡危险性评价

Landslide hazard evaluation in the northern mountainous area of Guide County based on Random Forest and AHP

滑坡危险性评价是滑坡灾害防治和管理的重要依据。文章基于层次分析法和随机森林模型,结合距离函数法,探索性地提出了一种新的组合赋权法(RF-AHP)。采用RF-AHP对青海省贵德县北部山区滑坡进行了危险性评价,对比探讨了AHP、RF和RF-AHP三种模型评价结果与实际滑坡灾害的吻合性,结果表明:(1)RF-AHP在高危险区和极高危险区面积占比38.38%的情况下,包括了60.13%的滑坡灾害,结果准确性相比AHP和RF两种模型有较大提升;(2)随着危险性等级的逐步提高,RF-AHP区划结果中相应分区的灾害实际发生的比率也随之增高,并对三种方法出现结果差异的客观原因进行了分析讨论,证明RF-AHP适用于滑坡危险性评价工作。

Evaluation of landslide susceptibility provide an importance evidence of landslide disaster prevention and landslide warning.A new combined weight process(RF-AHP)based on random forest(RF),analytic hierarchy process(AHP)and distance functions is proposed.The model is applied to the northern mountainous area of Guide County.The compatibility and applicability of three models are discussed.The results show that(1)the RF-AHP model accounts for 38.38%of the high risk area and the extremely high risk area,it includes 60.13%of the landslide disasters,is greatly improved compared to the two models of AHP and RF.(2)With the gradual increase in risk levels,the actual rate of the corresponding disasters in the results of RF-AHP also gradually increases.The objective causes of the difference in results between the three methods were discussed.It can be concluded from the case study that combined weight process is suitable for evalution of landslide susceptibility.