基于决策树考虑地形特征的场地参数估计方法  

A method of site parameter estimation based on decision tree theory considering terrain features

作　　者：张雨婷 任叶飞[1] 温瑞智[1] 王大任 冀昆[1] ZHANG YuTing;REN YeFei;WEN RuiZhi;WANG DaRen;JI Kun(Key Laboratory of Earthquake Engineering and Engineering Vibration,Institute of Engineering Mechanics,China Earthquake Administration,Harbin 150080,China)

机构地区：[1]中国地震局工程力学研究所,中国地震局地震工程与工程振动重点实验室,哈尔滨150080

出　　处：《地球物理学报》2022年第2期698-710,共13页Chinese Journal of Geophysics

基　　金：大中城市地震灾害情景构建重点专项(2018QJGJ07);国家重点研发计划项目(2019YFE0115700);国家自然科学基金(51878632,U1901602);黑龙江省自然科学基金优秀青年项目(YQ2019E036)资助。

摘　　要：发展基于地形特征的场地参数V_(S30)估计方法因其具有重要应用需求而成为研究热点.以我国新疆维吾尔自治区和河北省的DEM数据和工程钻孔资料验证了基于决策树理论考虑地形特征的V_(S30)估计方法在我国的适用性,检验方法的准确性和对DEM数据精度的敏感性.得到如下结论:(1)基于决策树理论考虑地形坡度、表面纹理和局部凸度划分了两个地区的16类地形类别,建立了考虑这三项地形特征的V_(S30)预测模型;(2)经验证基于决策树理论考虑三项地形特征的V_(S30)估计方法具有普遍适用性,但同时存在区域依赖性,需要分区建立适用的V_(S30)预测模型;(3)发现引入了表面纹理和局部凸度两项地形特征后,较使用单一的地形坡度对V_(S30)预测的准确性有所提升;(4)地形分类对DEM数据精度存在敏感性,高精度数据对陡峭的山脉地区划分可能更为详细,而低精度数据则能够对平坦的平原地区识别可能更为充分,相比较而言900 m精度的DEM数据对于基于地形特征的V_(S30)估计方法相对较为实用.上述模型和方法可为发展我国区域场地分类图提供一种有效的技术途径.The development of site parameter V_(S30) estimation method based on surface terrain features has become a hot topic because of its numerous application demands.The DEM data and engineering boreholes collected from Xinjiang Uygur Autonomous Region and Hebei Province are used to validate whether the V_(S30) estimation method based on decision tree theory considering terrain features is applicable to China,and testify its accuracy in V_(S30) prediction and sensitivity to DEM data resolution.The following conclusions are drawn:(1)Based on the decision tree theory both regions are classified as 16 types of terrain categories by means of three terrain features,i.e.,topographic slope,surface texture and local convexity,and the V_(S30) prediction models are developed considering proxies of these three terrain features;(2)It is validated that the V_(S30) estimation method based on decision tree theory considering terrain features has universal application,but is regionally dependent and calls for developing V_(S30) prediction models separately;(3)It has been observed that the accuracy of V_(S30) prediction is improved after introducing two terrain features(i.e.,surface texture and local convexity)on the basis of topographic slope;(4)Terrain classification is sensitive to DEM data resolution,and high-resolution data is more appropriate for steep mountain areas,while low-resolution data results in more detailed classifications in plain areas.In comparison,the 900 m-resolution DEM data is relatively more practical for terrain-proxy methods in V_(S30) prediction.The proposed models and methods in this study could support for developing an effective technical approach for the development of regional site classification maps in China.

关 键 词：决策树理论 地形坡度 地形特征 地形分类 场地参数 V_(S30)预测模型 

分 类 号：P315[天文地球—地震学]