摘要
以深圳市历史崩塌灾害数据为基础,选取高程、坡度、坡向、水系密度、道路密度、距断层距离和岩性作为评价指标,随机抽取70%的数据作为训练集,剩余30%的数据作为验证集,分别基于频率比模型(FR模型)、支持向量机模型(SVM模型)和二者的组合模型(FR-SVM组合模型)对深圳市崩塌灾害易发性进行评价,同时分析了核函数选择和数据预处理方法对SVM模型和FR-SVM组合模型性能的影响。结果表明:1)深圳市崩塌灾害高度易发区和较高度易发区集中分布在梧桐山山脉周围及横岗—罗湖、温塘—观澜一带;2)ROC曲线和灾害分布验证表明,FR-SVM组合模型相对于FR和SVM单模型而言精度更高,具有较好的实用性;3)SVM模型和FR-SVM组合模型中,采用Rbf核函数及归一化数据预处理方法的模型准确率更高;4)深圳市的不稳定斜坡中65%分布在崩塌灾害的中度易发区及以上分区,福田区及龙岗区危险性较高。易发性分区结果可为深圳市的地质灾害防治和土地规划提供参考。
Based on the historical collapse disaster records in Shenzhen,the elevation,slope,aspect,drainage density,road density,distance from the fault,and stratum lithology were selected as evaluation indexes.70%of the data were randomly selected as the training set and the remaining 30%as the validation set.The collapse disasters susceptibility in Shenzhen was evaluated based on the frequency ratio model(FR model),the support vector machine model(SVM model)and the combined model(FR-SVM combined model),respectively.The influence of kernel function selection and data preprocessing method on the performance of the SVM model and FR-SVM combined model were analyzed.The results indicate that:1)The moderate-high collapse disasters susceptibility zones mainly distribute around the Wutong Mountains,Henggang-Luohu,and Wentang-Guanlan area;2)ROC curve and disaster distribution verify that the FR-SVM combined model has better practicability and accuracy than the FR and SVM models;3)The standalone SVM model and the FR-SVM combined model,the Rbf kernel function and normalized data initialization preprocessing method yields has higher model accuracy;4)65%of unstable slopes in Shenzhen are distributed in the moderate susceptibility and above areas;Futian and Longgang are more dangerous than other districts.The outcomes of susceptibility mapping can offer useful insights for disaster prevention and land-use planning in Shenzhen.
作者
王大兵
张清山
冉斌
周苏华
WANG Dabing;ZHANG Qingshan;RAN Bin;ZHOU Suhua(Guizhou Provincial Transportation Construction Project Quality Supervision and Law Enforcement Detachment,Guiyang 550008,China;College of Civil Engineering,Hunan University,Changsha 410082,China;Fourth Engineering Co.,Ltd.,of China Railway 20th Bureau Group Co.,Ltd.,Qingdao 266075,China)
出处
《市政技术》
2025年第8期119-128,151,共11页
Journal of Municipal Technology
基金
贵州省交通运输厅科技计划项目(2023-312-030,2025-112-018)。
关键词
崩塌灾害
易发性
频率比
支持向量机
collapse disasters
susceptibility
frequency ratio
support vector machine
