Comparison of Landslide Susceptibility Evaluation by Deep Random Forest and Random Forest Model: A Case Study of Lueyang County, Hanzhong City
-
摘要:
针对浅层的机器学习模型泛化能力低而导致其滑坡易发性评价模型预测精度不高的问题,笔者围绕陕西省汉中市略阳县城中心为研究区,采用深度随机森林构建区域地灾易发性评价模型来提升预测精度。依据略阳县滑坡成灾机理研究成果,选取坡度、相对高差、坡向、坡型、工程地质岩组、断裂距离、水系距离、公路铁路距离、植被覆盖等9个因子作为易发性评价指标;将研究区栅格单元按5 m × 5 m进行划分并提取评价因子值,输入深度随机森林评价模型,从而获得研究区易发性评价图。依据评价结果略阳县地质灾害可划分为极高易发区、高易发区、中易发区、低易发区4个等级,面积所占比例分别为5.31%、22.97%、42.11%、29.61%,其划分结果与研究区内地质灾害实际发育情况吻合,合理反映研究区地灾分布的总体特征。深度随机森林的地质灾害易发性预测模型在ROC曲线下面积值(AUC)为91.2%,高于随机森林预测模型的86.3%,表明该模型具有一定的合理性与可行性,可为区域滑坡易发性评价进一步提供新方法。
Abstract:To address the problem of low prediction accuracy of landslide susceptibility evaluation model due to the difficulty of knowledge reuse and generalization of shallow machine learning model, this paper takes Lueyang County, Hanzhong City, Shaanxi Province as the study area and uses deep random forest to build a regional geological disaster susceptibility evaluation model to improve the prediction accuracy. Firstly, based on the research results of landslide mechanism in Lueyang County, nine factors such as slope, relative height difference, slope direction, slope type, engineering geological rock group, fault distance, river system distance, road and railroad distance, and vegetation cover are selected as susceptibility evaluation indexes; secondly, the study area is divided into 5 m × 5 m raster cells and the values of evaluation factors are extracted and input into the depth random forest evaluation model; finally, the susceptibility evaluation map of the study area is obtained. Based on the evaluation results, geological hazards in Lueyang County can be classified into four levels: very high susceptibility, high susceptibility, medium susceptibility, and low susceptibility, with the proportion of area being 5.31%, 22.97%, 42.11%, and 29.61%. The classification results are consistent with the actual development of geological hazards and reasonably reflect the overall characteristics of geological hazard distribution in the study area. In addition, the area under the ROC curve of the geological hazard susceptibility prediction model of deep random forest is 91.2%, which is higher than 86.3% of the random forest prediction model, indicating that the model is reasonable and feasible, and can provide new ideas for the evaluation of regional landslide susceptibility.
-
Key words:
- landslide /
- Lueyang County /
- susceptibility evaluation /
- Deep Random Forest Model
-
-
图 3 深度随机森林模型流程结构(据Zhou, 2017)
Figure 3.
表 1 各评价因子信息量表
Table 1. Weighted information values of individual evaluation factors
因子 分级 Si(km2) Ni(个) Ii 坡度 <10° 10.535625 8 −0.085244 10°~20° 13.3961 21 0.639636 20°~30° 35.77955 46 0.441342 30°~40° 54.226925 38 −0.165514 40°~50° 38.6219 12 −0.978836 >50° 13.119575 12 0.100879 相对高差 16~175 m 24.930925 53 0.944260 175~238 m 55.38815 46 0.004353 238~300 m 49.224425 29 −0.339017 300~379 m 26.43665 8 −1.005232 379~605 m 9.6986 1 −2.081904 坡向 0~45° 23.746850 21 0.067150 45°~90° 21.137625 23 0.274517 90°~135° 17.924900 10 −0.393528 135°~180° 18.869250 14 −0.108399 180°~225° 20.641625 26 0.420864 225°~275° 22.018400 18 −0.011429 275°~315° 20.152475 13 −0.248300 315°~360° 21.185550 12 −0.378335 曲率 <−0.5(凹形坡) 11.1099 12 0.267147 −0.5~0.5
(直线形坡)143.597 119 0.002190 >0.5(凸形坡) 10.969775 6 −0.413307 工程地质岩组 坚硬岩组 65.882425 40 −0.308915 半坚硬岩组 29.980725 13 −0.645528 软硬相间岩组 58.766425 46 −0.054852 松散岩组 11.050425 39 1.451170 距断裂距离 <100 m 31.79 26 −0.010915 100~200 m 25.87 26 0.195167 200~500 m 49.79 53 0.252595 500~1000 m 34.60 23 −0.218224 >1000 m 23.64 9 −0.775554 距河流水系
距离<200 m 69.69 86 0.400425 200~400 m 52.38 31 −0.334384 400~600 m 27.98 17 −0.308056 600~800 m 11.37 1 −2.240549 >800 m 4.28 2 −0.569794 距公路、铁路
距离<100 m 18.99 58 1.306584 100~500 m 46.12 41 0.072488 500~1000 m 38.06 14 −0.809915 1000~1500 m 26.29 11 −0.681226 >1500 m 36.23 13 −0.834794 NDVI −0.41~0.07 1.88 0 0.07~0.32 7.64 26 1.415100 0.32~0.52 26.25 56 0.947905 0.52~0.68 91.74 50 −0.416840 0.68~0.84 38.18 5 −1.842838 表 2 易发性等级划分与滑坡实际发生比率对比表
Table 2. Comparison of susceptibility classification and actual landslide occurrence rate
评价
方法易发性
等级a
易发分区面积占比(%)b
分区滑坡数量c
滑坡百分比(%)c/a
滑坡发生比率随机森林 低 16.34 1 0.73 0.04 中 69.21 74 54.01 0.78 高 11.24 40 29.20 2.60 极高 3.21 22 16.06 5.00 深度
随机森林低 29.61 7 5.11 0.17 中 42.11 41 29.93 0.71 高 22.97 62 45.26 1.97 极高 5.31 27 19.71 3.71 -
[1] 黄发明, 殷坤龙, 蒋水华, 等. 基于聚类分析和支持向量机的滑坡易发性评价[J]. 岩石力学与工程地质学报, 2018, 37(1): 156-167
HUANG Faming, YIN Kunlong, JIANG Shuihua, et al. Landslide susceptibility assessment based on clustering analysis and support vector machine[J]. Chinese Journal of Rock Mechanics and Engineering, 2018, 37(1): 156-167.
[2] 李亭, 田原, 邬伦, 等. 基于随机森林方法的滑坡灾害危险性区划[J]. 地理与地理信息科学, 2014, 30(6): 25-30 doi: 10.3969/j.issn.1672-0504.2014.06.006
LI Ting, TIAN Yuan, WU Lun, et al. Landslide susceptibility Mapping Using Random Forest[J]. Geography and Geo-Information Science, 2014, 30(6): 25-30. doi: 10.3969/j.issn.1672-0504.2014.06.006
[3] 刘坚, 李树林, 陈涛, 等. 基于优化随机森林模型的滑坡易发性评价[J]. 武汉大学学报·信息科学版, 2018, 43(7): 1085-1091
LIU Jian, LI Shulin, CHEN Tao, et al. Landslide Susceptibility Assesment Based on Optimized Random Forest Model[J]. Geomatics and Information Science of Wuhan University, 2018, 43(7): 1085-1091.
[4] 孟庆华. 秦岭山区地质灾害风险评估方法研究—以陕西凤县为例[D]. 北京: 中国地质科学院, 2011
MENG Qinghua. Study on the Methods of Geo-hazards Risk Assessment in Qinling Mountain: A case studyof Feng County, Baoji City, Shaanxi Province[D]. Beijing: Chinese Academy of Geological Sciences, 2011.
[5] 孟晓捷, 张新社, 曾庆铭, 等. 基于加权信息量法的黄土滑坡易发性评价——以1: 5万天水市麦积幅为例[J]. 西北地质, 2022, 55(2): 249-259
MENG Xiaojie, ZHANG Xinshe, ZENG Qingming, et al. The Susceptibility Evaluation of Loess Landslide Based on Weighted Information Value Method—Taking 1: 50000 Map of Maiji District of Tianshui City As an Example[J]. Northwestern Geology, 2022, 55(2): 249-259.
[6] 邱海军. 区域滑坡崩塌地质灾害特征分析及其易发性和危险性评价研究——以宁强县为例[D]. 西安: 西北大学, 2012: 83
QIU Haijun. Study on the Regional Landslide Characteristic Analysis and Hazard Assessment: A case study of Ningqiang County[D]. Xi’an: Northwest University, 2012: 83.
[7] 邱维蓉, 吴帮玉, 潘学树, 等. 几种聚类优化的机器学习方法在灵台县滑坡易发性评价中的应用[J]. 西北地质, 2020, 53(1): 222-233
QIU Weirong, WU Bangyu, PAN Xueshu, et al. Application of Several Cluster-optimization-based Machine Learning Methods in Evaluation of Landslide Susceptibility in Lingtai County[J]. Northwestern Geology, 2020, 53(1): 222-233.
[8] 田春山, 刘希林, 汪佳, 等. 基于CF和Logistic回归模型的广东省地质灾害易发性评价[J]. 水文地质工程地质, 2016, 43(6): 154-170
TIAN Chunshan, LIU Xilin, WANG Jia. et al. Geohazard susceptibility assessmentbased on CF model and Logistic Regression models in Guangdong[J]. Hydrogeology & Engineering Geology, 2016, 43(6): 154-170.
[9] 王佳运, 毕俊擘, 杨旭东, 等. 山西吕梁山区城镇边坡风险分级与优化[J]. 地质通报, 2020, 39(12): 2004-2012 doi: 10.12097/j.issn.1671-2552.2020.12.014
WANG Jiayun, BI Junbo, YANG Xudong, et al. Risk classification and optimization of to-wn side slope in Lüliang Mountain[J]. Geological Bulletin of China, 2020, 39(12): 2004-2012. doi: 10.12097/j.issn.1671-2552.2020.12.014
[10] 王杰, 马凤山, 郭捷, 等. 一种改进的区域滑坡危险性评价模型及其应用[J]. 中国地质灾害与防治学报, 2011, 22(2): 14-19 doi: 10.3969/j.issn.1003-8035.2011.02.003
WANG Jie, MA Fengshan, GUO Jie, et al. An improved regional landslide hazard assessment, model and its application[J]. The Chinese Journal of Geological Hazard and Control, 2011, 22(2): 14-19. doi: 10.3969/j.issn.1003-8035.2011.02.003
[11] 王涛, 吴树仁, 石菊松, 等. 秦岭中部太白县地质灾害发育特征及危险性评估[J]. 地质通报, 2013, 32(12): 1976-1983
WANG Tao, WU Shuren, SHI Jusong, et al. Case Study Of Landslide Characteristics And Hazard Assessment In Taibai County, Central Qinling Mountains[J]. Geological Bulletinof China, 2013, 32(12): 1976-1983.
[12] 吴常润, 赵冬梅, 刘澄静, 等. 基于GIS和信息量模型的陇川县滑坡易发性评价[J]. 西北地质, 2020, 53(2): 308-320
WU Changrun, ZHAO Dongmei, LIU Chengjing, et al. Landslide Susceptibility Assessment of Longchuan County Based on GIS and Information Value Model[J]. NorthwesternGeology, 2020, 53(2): 308-320.
[13] 吴润泽, 胡旭东, 梅红波, 等. 基于随机森林的滑坡空间易发性评价: 以三峡库区湖北段为例[J]. 地球科学, 2021, 46(1): 321-330
WU Runze, HU Xudong, MEI Hongbo, et. al. Spatial Susceptibility Assessment of Landslides Based on Random Forest: A Case Study from Hubei Section in the Three Gorges Reservoir Area[J]. Earth Science, 2021, 46(1): 321-330.
[14] 吴孝情, 赖成光, 陈晓宏, 等. 基于随机森林权重的滑坡危险性评价: 以东江流域为例[J]. 自然灾害学报, 2017, 26(5): 119-129
WU Xiaoqing, LAI Chengguang, CHEN Xiaohong, et al. A landslide hazard assessment based on random forestweight: a case study in the Dongjiang River Basin[J]. Journal of Natural Disarsters, 2017, 26(5): 119-129.
[15] 向喜琼. 区域滑坡地质灾害危险性评价与风险管理[D]. 成都: 成都理工大学, 2005: 22
XIANG Xiqiong. Regional Landslide Hazard Assessment and Risk Management[D]. Chengdu: Chengdu University of Technology, 2005: 22.
[16] 薛强, 张茂省, 李林等. 基于斜坡单元与信息量法结合的宝塔区黄土滑坡易发性评价[J]. 地质通报, 2015, 34(11): 2108-2115
XUE Qiang, ZHANG Maosheng, LI Lin. Loess landslide susceptibility evaluation based on slope unit and information value method in Baota District, Yan’an. Geological Bulletin of China, 2015, 34(11): 2108-2115.
[17] 张春山, 韩金良, 孙炜锋, 等. 陕西陇县地质灾害危险性分区评价[J]. 地质通报, 2008, 27(11): 1795-1801 doi: 10.3969/j.issn.1671-2552.2008.11.006
ZHANG Chunshan, HAN Jinliang, SUN Weifeng, et al. Assessments of geohazard danger zoning in Longxian County, Shaanxi, China[J]. Geological Bulletin of China, 2008, 27(11): 1795-1801. doi: 10.3969/j.issn.1671-2552.2008.11.006
[18] 张茂省, 贾俊, 王毅, 等. 基于人工智能(AI)的地质灾害防控体系建设[J]. 西北地质, 2019a, 52(2): 103-116
ZHANG Maosheng, JIA Jun, WANG Yi, et. al. Construction of Geological Disarster Prevention and Control System Based on AI[J]. Northwestern Geology, 2019a, 52(2): 103-116.
[19] 张茂省, 薛强, 贾俊, 等. 山区城镇地质灾害调查与风险评价方法及实践[J]. 西北地质, 2019b, 52(2): 125-135 doi: 10.19751/j.cnki.61-1149/p.2019.02.013
ZHANG Maosheng, XUE Qiang, JIA Jun, et. al. Methods and Practice for the Investigationand Risk Assessment of Geo-hazards in Mountains Towns[J]. Northwestern Geology, 2019b, 52(2): 125-135. doi: 10.19751/j.cnki.61-1149/p.2019.02.013
[20] 张向营, 张春山, 孟华君, 等. 基于Random Forest和AHP的贵德县北部山区滑坡危险性评价[J]. 水文地质工程地质, 2018, 45(4): 142-149
ZHANG Xiangying, ZHANG Chunshan, MENG Huajun, et al. Landslide hazard evaluation in the northern mountainous area of Guide County based on Random Forest and AHP[J]. Hydrogeology & Engineering Geology, 2018, 45(4): 142-149.
[21] 张永双, 苏生瑞, 吴树仁, 等. 强震区断裂活动与大型滑坡关系研究[J]. 岩石力学与工程学报, 2011, 30(增刊2): 3503-3513
ZHANG Yongshuang, SU Shengrui, WU Shuren, et al. Research On Relationship Between Fault Movement And Large-Scale Landslide In Intensive Earthquake Region[J]. Chinese Journalof Rock Mechanics and Engineering, 2011, 30(Sup. 2): 3503-3513.
[22] 赵建华, 陈汉林, 杨树峰, 等. 基于决策树算法的滑坡危险性区划评价[J]. 浙江大学学报(理学版), 2004, 31(4): 465-470
ZHAO Jianhua, CHEN Hanlin, YANG Shufeng, et al. Landslide risk assessment based on decision tree arithmetic[J]. Journal of Zhejiang University (Science Edition), 2004, 31(4): 465-470.
[23] 周静静, 赵法锁, 李辉, 等. 陕西省地质灾害与影响因素相关性研究[J]. 灾害学, 2019, 34(2): 228-234 doi: 10.3969/j.issn.1000-811X.2019.02.041
ZHOU Jingjing, ZHAO Fasuo, LI Hui, et al. Correlational Research Between Geological Hazards and impact Factors in Shaanxi Province[J]. Journal of Catastrophology, 2019, 34(2): 228-234. doi: 10.3969/j.issn.1000-811X.2019.02.041
[24] 周样样. 陕南地区强降雨条件下突发型地质灾害成因机制研究[D]. 西安: 长安大学, 2013
ZHOU Yangyang. Study on Formation Mechanism of Abrupt Geological Hazards of Southern Shaanxi Region in Condition of Strong Rainstorm[D]. Xi’an: Chang’an University, 2013.
[25] Chung C J F, Fabbri A G. Validation of spatial prediction modelsfor landslidehazard mapping[J]. Natural Hazards, 2003, 30(3): 451-472. doi: 10.1023/B:NHAZ.0000007172.62651.2b
[26] Dou J , Yunus A P , Merghadi A , et al. Different sampling strategies for predicting landslide susceptibilities are deemed less consequential with deep learning[J]. Science of the Total Environment, 2020, 720: 137320. https://doi.org/10.1016/j.scitotenv.2020.137320
[27] Maher Ibrahim Samee, Biswajeet Pradhana, Saro Lee, et al. Application of convolutional neural networks featuring Bayesian optimization for landslide susceptibility assessment[J]. Catena, 2020, 186: 10424. https://doi.org/10.1016/j.catena.2019.104249
[28] Saro, Lee, Joo Hyung Ryu, Joong Sun Won, et al. Determination and application of the weights for landslide susceptibility mapping using an artificial neural network[J]. Engineering Geology, 2004, 71(3-4): 289-302. doi: 10.1016/S0013-7952(03)00142-X
[29] Snoek J, Larochelle H, Adams R P. et al. Practical Bayesian Optimization of Machine Learning Algorithms[J]. Advances in Neural Information Processing Systems, 2012, 4: 1–12.
[30] Fang Zhice, Wang Yi , Ling Peng, et al. Integration of convolutional neural network and conventional machine learning classifiers for landslide susceptibility mapping[J]. Computer & Geosciencees, 2020, 139: 104470. https://doi.org/10.1016/j.cageo.2020.104470
[31] Zhou Zhihua, Ji Feng. Deep Forest: Towards An Alternative to Deep Neural Networks[J]. Statistics, 2017, 71(3–4): 289–302.
-