基于确定性系数和支持向量机的地质灾害易发性评价

doi:10.12082/dqxxkx.2018.180349

摘要/Abstract

摘要：

确定性系数（Certainty Factor,CF）是经典的地质灾害影响因子敏感性分析方法;支持向量机（Support Vector Machine, SVM）作为机器学习的代表方法,能够综合各个影响因子的关系,对地质灾害易发性进行评价。本文以云南省怒江州泸水县为研究区,将高程、坡度、坡向、剖面曲率、距断裂的距离、距河网的距离、距路网的距离、地貌类型、岩土体类型、土地利用类型作为该区域地质灾害影响因子,依据各影响因子灾害面积比和分级面积比曲线对影响因子的状态进行分级。根据381个地质灾害隐患点,采用CF方法计算的各个影响因子的敏感性值,作为SVM的分类数据,建立基于CF-SVM的易发性评估模型,同时与单独SVM模型的评价结果进行对比分析。结果表明,CF-SVM模型得到的极高和高易发区主要分布在怒江两岸河谷地带,涵盖了89.76%的地质灾害隐患点,比单独SVM模型具有更高的成功率;利用ROC曲线和P-R曲线对两个模型进行检验,CF-SVM模型的评价精度分别达到92%和88%,均高于单独的SVM。由此说明,CF-SVM模型对地质灾害易发性评价有较高的预测价值,可以为地质灾害风险评估和管理提供依据。

关键词: 地质灾害, 确定性系数, 支持向量机, 易发性评价, 泸水县

Abstract:

The Certainty Factor (CF) is a classical method of sensitivity analysis for geological hazards, which can be used to quantify complex multi-factors in the same range, and CF value directly represents the contribution of each factor to geological hazards. The Support Vector Machine (SVM) is a representative method for machine learning, which can be used to evaluate the susceptibility of geological hazards on the basis of various impact factors. The main purpose of this paper is to combine CF method with SVM method called CF-SVM model for developing geological hazards susceptibility model which is applied to Lushui County of Yunnan Province. To assess geological hazards susceptibility, the paper selects ten impact factors, including the elevation, slope angle, slope aspect, profile curvature, distance to faults, distance to rivers, distance to roads, rock soil mass types, geomorphologic type, the land use, which were calculated as the most important impact factors. The state of each impact factor was graded based on the geological hazards area ratio curve and the grading area ratio curve. The CF method is applied to calculated the sensitivity values of each factor on the basis of 381 geological hazards, and the under-sampling method was used to select 381 non-geological hazards in Lushui county. The 381 geological hazards and 381 non-geological hazards hazards are the classification data of SVM method and geological hazards susceptibility maps were produced. Research shows that the extremely high and the high susceptibility areas are mainly distributed on the both sides of the Nujiang River. The comparision of geological hazards susceptibility map and actual geological hazards distribution showed that extremely high and the high susceptibility areas can account for about 89.76% of the actual geological hazards for CF-SVM model which is better than the single SVM model. The model performance was evaluated by the receiver operation characteristic (ROC) and precision recall (P-R) in the two models. Through ROC and P-R of the results, the prediction power of the CF-SVM model is superior to the single SVM model, and prediction accuracy of the CF-SVM model reached 92% and 88% respectively. This indicates that the combination of the CF and SVM has a higher predictive value. The research provides guidance and technical reference for the nationwide county geological hazards evaluation, and the proposed model can be used assess and manage the risk of geological hazard.

Key words: geological hazard, certainty factor, support vector machine, susceptibility evaluation, Lushui County

李远远, 梅红波, 任晓杰, 胡旭东, 李梦迪. 基于确定性系数和支持向量机的地质灾害易发性评价[J]. 地球信息科学学报, 2018, 20(12): 1699-1709.DOI:10.12082/dqxxkx.2018.180349

LI Yuanyuan,MEI Hongbo,REN Xiaojie,HU Xudong,LI Mengdi. Geological Disaster Susceptibility Evaluation Based on Certainty Factor and Support Vector Machine[J]. Journal of Geo-information Science, 2018, 20(12): 1699-1709.DOI:10.12082/dqxxkx.2018.180349

图/表 8

图1

表1

图2

图3

表2

影响因子敏感性计算结果"

影响因子	分类	分区栅格/个		隐患点/个	栅格比例/%	隐患点比例/%		CF
高程/m	700~1100		15 482	92	5.01	24.15	0.79
	1100~1700		47 263	115	15.30	30.18	0.49
	1700~2100		49 721	136	16.09	35.70	0.55
	2100~2300		28 357	27	9.18	7.09	-0.23
	>2300		168 130	11	54.42	2.89	-0.95
坡度/°	0~15		31 463	70	10.20	18.37	0.45
	15~20		26 957	52	8.73	13.65	0.36
	20~30		93 922	123	30.40	32.28	0.06
	30~50		150 619	136	48.75	35.70	-0.27
	>50		5992	0	1.94	0	-1
坡向/°	0~45		40 004	44	12.95	11.55	-0.11
	45~135		81 761	128	26.46	33.60	2.13
	135~225		78 137	92	25.29	24.15	-0.05
	225~270		36 140	52	11.70	13.65	0.14
	270~360		72 911	65	23.60	17.06	-0.28
剖面曲率	0~2		126 328	149	40.89	39.10	-0.04
	2~3		66 278	86	21.45	22.57	0.05
	3~6		89 907	103	29.10	27.03	-0.07
	6~8		17 963	23	5.81	6.04	0.04
	8~10		7375	18	2.39	4.72	0.50
	>10		1102	2	0.36	0.52	0.32
距断裂的距离/m	0~1200		153 635	301	49.73	79.00	0.37
	1200~1800		31 168	42	10.09	11.02	0.08
	1800~3000		35 349	22	11.44	5.77	-0.50
	>3000		88 801	16	28.74	4.20	-0.85
距河流的距离/m	0~200		102 325	153	33.12	40.16	0.18
	200~400		76 523	77	24.77	20.21	-0.18
	400~600		51 970	73	16.82	19.16	0.12
	600~1400		72 559	73	23.49	19.16	-0.18
	1400~1600		3145	4	1.01	1.05	0.03
	>1600		2431	1	0.79	0.26	-0.67
距路网的距离/m	0~200		67 711	264	21.91	69.29	0.68
	200~400		41 322	56	13.37	14.70	0.09
	400~800		49 902	38	16.15	9.97	-0.38
	800~1000		16 677	5	5.40	1.31	-0.76
	>1000		133 341	18	43.16	4.72	-0.89
影响因子	分类	分区栅格/个		隐患点/个	栅格比例/%	隐患点比例/%		CF
地貌类型	1		20 422	51	6.61	13.39	0.51
	2		39 326	131	12.72	34.38	0.63
	3		236 763	199	76.63	52.23	-0.32
	4		12 442	0	4.03	0	-1
岩土体类型	1		21 716	58	7.05	15.22	0.54
	2		81 516	23	26.48	8.40	-0.68
	3		109 024	129	35.41	33.86	-0.04
	4		62 997	101	20.46	26.51	0.23
	5		26 740	31	8.69	8.14	-0.03
	6		5875	30	1.91	7.87	0.76
土地利用类型	1		246 535	151	79.80	39.63	-0.50
	2		26 548	19	8.60	4.99	-0.42
	3		1986	35	0.64	9.19	0.93
	4		27 249	127	8.82	33.33	0.74
	5		5168	45	1.67	11.81	0.86
	6		1296	2	0.42	0.52	0.20
	7		170	2	0.06	0.52	0.90

表2

表3

图4

图5

参考文献 35

[1]	牛瑞卿,彭令,叶润青,等.基于粗糙集的支持向量机滑坡易发性评价[J].吉林大学学报(地球科学版),2012,42(2):430-439.
	[ Niu R Q, Peng L,Ye R Q, et al.Landslide susceptibility assessment based on rough sets and support vector machine[J]. Journal of Ji Lin University (Earth Science Edition), 2012,42(2):430-439. ]
[2]	黄发明. 基于3S和人工智能的滑坡位移预测与易发性评价[D].武汉:中国地质大学(武汉),2017.
	[ Huang F M.Landslide displacement prediction and susceptibility assessment based on 3S and artificial intelligence[D]. Wuhan: China University of Geosciences (Wuhan), 2017. ]
[3]	范林峰,胡瑞林,曾逢春,等.加权信息量模型在滑坡易发性评价中的应用——以湖北省恩施市为例[J].工程地质学报,2012,20(4):508-513.
	[ Fan L F, Hu R L, Zeng F C, et al.Applicaton of weighted information value model to landslide susceptibility assessment: A case study of Enshi city[J]. Journal of Engineering Geology, 2012,20(4):508-513. ]
[4]	Poiraud A.Landslide susceptibility-certainty mapping by a multi-method approach: A case study in the Tertiary basin of Puy-en-Velay (Massif central, France)[J]. Geomorphology, 2014,216:208-224. doi: 10.1016/j.geomorph.2014.04.001
[5]	Schicker R, Moon V. Comparison of bivariate and multivariate statistical approaches in landslide susceptibility mapping at a regional scale[J]. Geomorphology, 2012,161-162(7):40-57. doi: 10.1016/j.geomorph.2012.03.036
[6]	李芳,梅红波,王伟森,等.降雨诱发的地质灾害气象风险预警模型:以云南省红河州监测示范区为例[J].地球科学,2017,42(9):1637-1646. doi: 10.3799/dqkx.2017.112
	[ Li F, Mei H B, Wang W S, et al.Rainfall-Induced meteorological early warning of geo-hazards model: application to the monitoring demonstration area in Honghe Prefecture, Yunnan Province[J]. Earth Science, 2016,35(2):284-296. ] doi: 10.3799/dqkx.2017.112
[7]	谭玉敏,郭栋,白冰心,等.基于信息量模型的涪陵区地质灾害易发性评价[J].地球信息科学学报,2015,17(12):1554-1562. doi: 10.3724/SP.J.1047.2015.01554
	[ Tan Y M, Guo D, Bai B X, et al.Geological hazard risk assessment based on information quantity model in fuling district, Chongqing city, China[J]. Journal of Geo-information Science, 2015,17(12):1554-1562. ] doi: 10.3724/SP.J.1047.2015.01554
[8]	苏鹏程,韦方强.澜沧江流域滑坡泥石流空间分布与危险性分区[J].资源科学,2014,36(2):273-281.
	[ Su P C, Wei F Q.Landslides and debris flow hazards and danger zonation along the Lancang River[J]. Resources Science, 2014,36(2):273-281. ]
[9]	陈晓利,周本刚,冉洪流,等.汶川地震中擂鼓镇地区的滑坡崩塌规律及预测[J].吉林大学学报(地球科学版),2010,40(6):1371-1379.
	[ Chen X L, Zhou B G, Ran H L, et al.Analysis and prediction of the spatial distribution of landslides triggered by Wenchuan earthquakes in Leiguzhen Region[J]. Journal of Jilin University( Earth Science Edition), 2010,40(6):1371-1379. ]
[10]	兰恒星,伍法权,周成虎,等.基于GIS的云南小江流域滑坡因子敏感性分析[J].岩石力学与工程学报,2002(10):1500-1506. doi: 10.3321/j.issn:1000-6915.2002.10.014
	[ Lan H X, Wu F Q, Zhou C H, et al.Analysis on susceptibility of GIS based landslide triggering factors in Yunan Xiaojiang watershed[J]. Chinese Journal of Rock Mechanics and Engineering, 2002(10):1500-1506. ] doi: 10.3321/j.issn:1000-6915.2002.10.014
[11]	田春山,刘希林,汪佳.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):154-161,170. doi: 10.16030/j.cnki.issn.1000-3665.2016.06.24
	[ Tian C S, Liu X L, Wang J.Geohazard susceptibility assessment based on CF model and logistic regression models in Guangdong[J]. Hydrogeology & Engineering Geology, 2016,43(6):154-161,170. ] doi: 10.16030/j.cnki.issn.1000-3665.2016.06.24
[12]	Feng Y, Palomar D P.Normalization of linear support vector machines[J]. IEEE Transactions on Signal Processing, 2015,63(17):4673-4688. doi: 10.1109/TSP.2015.2443730
[13]	Daehyon Kim, Seungjae Lee, Seongkil Cho.Input vector normalization methods in support vector machines for automatic incident detection[J]. Transportation Planning & Technology, 2007,30(6):593-608. doi: 10.1080/03081060701698235
[14]	牛全福,冯尊斌,党星海,等.黄土区滑坡研究中地形因子的选取与适宜性分析[J].地球信息科学学报,2017,19(12):1584-1592.
	[ Niu Q F, Feng Z B, Dang X H, et al.Suitability analysis of topographic factors in loess landslide research[J]. Journal of Geo-information Science. 2017,19(12):1584-1592. ]
[15]	Trigila A, Iadanza C, Esposito C, et al.Comparison of logistic regression and random forests techniques for shallow landslide susceptibility assessment in Giampilieri (NE Sicily, Italy)[J]. Geomorphology, 2015,249:119-136. doi: 10.1016/j.geomorph.2015.06.001
[16]	Shortliffe E H.A model of inexact reasoning in medicine[J]. Mathematical Biosciences, 1975,23(3):351-379. doi: 10.1016/0025-5564(75)90047-4
[17]	Heckerman D.Probabilistic interpretations for mycin's certainty factors[J]. Machine Intelligence & Pattern Recognition, 1986,4:167-196.
[18]	Cherkassky V.The nature of statistical learning theory[J]. Technometrics, 1997,8(6):1564. doi: 10.1007/978-1-4757-2440-0
[19]	Tehrany M S, Pradhan B, Mansor S, et al.Flood susceptibility assessment using GIS-based support vector machine model with different kernel types[J]. Catena, 2015,125(125):91-101. doi: 10.1016/j.catena.2014.10.017
[20]	张学工. 关于统计学习理论与支持向量机[J].自动化学报,2000(1):36-46.
	[ Zhang X G.In troduction to statistical learning theory and support vector machines[J]. Acta Automatica Sinica. 2000(1):36-46. ]
[21]	汪海燕,黎建辉,杨风雷.支持向量机理论及算法研究综述[J].计算机应用研究,2014,31(5):1281-1286. doi: 10.3969/j.issn.1001-3695.2014.05.001
	[ Wang H Y, Li J H, Yang F L.Overview of support vector machine analysis and algorithm[J]. Application Research of Computers, 2014,31(5):1281-1286. ] doi: 10.3969/j.issn.1001-3695.2014.05.001
[22]	蒋德明,李益敏,鲍华姝.泸水县滑坡孕灾环境因素敏感性研究[J].自然灾害学报,2016,25(4):109-119.
	[ Jiang D M, Li Y M, Bao H S.Study on sensitivity in disaster-pregnant environmental factors of landslide in Lushui County[J]. Journal of Natural Disasters, 2016,25(4):109-119. ]
[23]	王丽丽,苏程,冯存均,等.数据驱动自适应更新的斜坡地质灾害易发性评价系统[J].岩石力学与工程学报,2016,35(S1):3076-3083.
	[ Wang L L, Su C, Feng C J, et al.A data driven self-adaptive up data landslide susceptibility assessment system[J]. Chinese Journal of Rock Mechanics and Engineering, 2016,35(S1):3076-3083. ]
[24]	张俊,殷坤龙,王佳佳,等.三峡库区万州区滑坡灾害易发性评价研究[J].岩石力学与工程学报,2016,35(2):284-296.
	[ Zhang J, Yin K L, Wang J J, et al.Evaluation of landslide susceptibility for Wanzhou district of Three Gorges reservoir[J]. Chinese Journal of Rock Mechanics and Engineering, 2016,35(2):284-296. ]
[25]	王佳佳,殷坤龙,肖莉丽.基于GIS和信息量的滑坡灾害易发性评价——以三峡库区万州区为例[J].岩石力学与工程学报,2014,33(4):797-808. doi: 10.3969/j.issn.1000-6915.2014.04.018
	[ Wang J J, Yin K L, Xiao L L.Langslide susceptibility assessment based on GIS and weighted information value: A case study of Wanzhou[J]. Chinese Journal of Rock Mechanics and Engineering, 2014,33(4):797-808. ] doi: 10.3969/j.issn.1000-6915.2014.04.018
[26]	Liu X Y, Wu J, Zhou Z H.Exploratory undersampling for class-imbalance learning.[J]. IEEE Transactions on Systems Man & Cybernetics Part B, 2009,39(2):539-550. doi: 10.1109/ICDM.2006.68 pmid: 19095540
[27]	M C.Landslide susceptibility mapping using the matrix assessment approach: A Derbyshire case study[J]. Geological Society, 1998,15(1):247-261.
[28]	焦方谦,赵新生,陈川.证据权模型在泥石流灾害易发性评价中的应用[J].干旱区地理,2013,36(6):1111-1124.
	[ Jiao F Q, Zhao X S, Chen C.Debris flow hazard susceptibility evaluation application with weighted evidences model[J]. Arid Land Geography, 2013,36(6):1111-1124. ]
[29]	武雪玲,任福,牛瑞卿.多源数据支持下的三峡库区滑坡灾害空间智能预测[J].武汉大学学报·信息科学版,2013,38(8):963-968.
	[ Wu X L, Ren F, Niu R Q.Spatial intelligent prediction of landslide hazard based on multi-source data in three gorges reservoir area[J]. Geomatics and Information Science of Wuhan University, 2013,38(8):963-968. ]
[30]	Fawcett T.An introduction to ROC analysis[J]. Pattern Recognition Letters, 2006,27(8):861-874. doi: 10.1016/j.patrec.2005.10.010
[31]	Davis J, Goadrich M.The relationship between Precision-Recall and ROC curves[C]. Proceedings of the International Conference on Machine Learning, New York, 2006:233-240.
[32]	林齐根,刘燕仪,刘连友,等.支持向量机与Newmark模型结合的地震滑坡易发性评估研究[J].地球信息科学学报,2017,19(12):1623-1633. doi: 10.3724/SP.J.1047.2017.01623
	[ Lin Q G, Liu Y Y, Liu L Y, et al.Earthquake-triggered landslide susceptibility assessment based on support vector machine combined with newmark displacement model[J]. Journal of Geo-information Science, 2017,19(12):1623-1633. ] doi: 10.3724/SP.J.1047.2017.01623
[33]	Swets J A.Measuring the Accuracy of Diagnostic Systems[J]. Science, 1988,240(4857):1285-1293. doi: 10.1126/science.3287615 pmid: 3287615
[34]	Fawcett T.An introduction to ROC analysis[J]. Pattern Recognition Letters, 2005,27(8):861-874. doi: 10.1016/j.patrec.2005.10.010
[35]	Davis J, Goadrich M.The relationship between Precision-Recall and ROC curves[C]. Proceedings of the International Conference on Machine Learning, New York, 2006:233-240.

模型	易发性等级	分区栅格数据	栅格比例/%	隐患点个数/个		频率比值
CF-SVM	极高	62 479	20.29	171	44.88	2.21
	高	53 380	17.34	171	44.88	2.59
	中等	44 333	14.40	28	7.35	0.51
	低	49 783	16.17	11	2.89	0.18
	极低	97 896	31.80	0	0	0
SVM	极高	13 951	4.53	56	14.70	3.24
	高	34 679	11.26	85	22.31	1.98
	中等	126 575	41.11	216	56.69	1.38
	低	114 459	37.18	19	4.99	0.13
	极低	18 207	5.91	5	1.31	0.22