摘要
为解决传统评价模型中精度不高、容易出现过拟合问题,提出基于信息量-随机森林(IV-RF)耦合模型对山地丘陵县滑坡灾害进行易发性评价。在关联性分析基础上,构建含有高程、坡向、坡度、工程地质岩组、距道路距离、平面曲率6个指标在内的评价体系,借助ArcGIS的空间分析功能,建立基于信息量-随机森林耦合模型进行滑坡灾害易发性分区评价,并通过信息量模型进行对比分析与验证,最后以新干县为研究区进行实例分析。结果表明:影响滑坡的6个评价指标中,距道路距离这一指标对其影响最大;滑坡高易发区和较高易发区呈线性分布在道路附近;信息量-随机森林耦合模型较信息量模型的预测精度高了20.7%,说明信息量-随机森林耦合模型更适用于山地丘陵区的滑坡易发性评价。
In order to solve the problem of low accuracy and over-fitting in the traditional evaluation model,this paper proposes to evaluate the susceptibility of landslide disaster in mountainous and hilly counties based on the coupling model of Information Value-Random Forest(IV-RF).On the basis of correlation analysis,building contains elevation,slope direction,slope,engineering geological petrofabric,distance from the road,surface curvature,six indexes evaluation system,using spatial analysis function of ArcGIS,based on the amount of Information Value-Random Forest coupling model of landslide hazards assessment,and through comparative analysis and validation Information Value model,Finally,a case study is carried out in Xinjan county.The results show that the distance from road has the greatest influence on landslide among the six evaluation indexes.The high and higher landslide prone areas are linearly distributed near roads.The prediction accuracy of the Information Value-Random Forest coupling model is 20.7%higher than that of the Information Value model,indicating that the Information Value-Random Forest coupling model is more suitable for landslide susceptibility assessment in mountainous and hilly areas.
作者
陈博
王洋
黄信
黄精涛
CHEN Bo;WANG Yang;HUANG Xin;HUANG Jingtao(School of Water Resources&Environmental Engineering,East China University of Technology,330013,Nanchang,PRC)
出处
《江西科学》
2022年第5期914-919,964,共7页
Jiangxi Science
