基于加权证据权模型的青藏高原东部巴塘断裂带滑坡易发性评价

巴塘断裂带位于青藏高原东部,呈北东—南西向展布,全新世活动强烈,沿断裂带崩塌、滑坡、泥石流等地质灾害极为发育。基于遥感解译和野外地质调查,在巴塘断裂带两侧10 km范围内识别出滑坡93处;在分析滑坡空间发育特征的基础上,选取地形地貌(地面高程、地形坡度和地形坡向)、地形湿度指数、地层岩性、活动断裂、降雨量、水系、人类工程活动和植被覆盖等10个因素作为滑坡易发程度的主控因素,采用加权证据权法建立滑坡易发性评价模型,开展巴塘断裂带滑坡易发性评价;成功率(ROC)曲线检验结果表明此次滑坡易发性评价的准确率为82.3%。采用基于自然断点法将滑坡易发程度划分为极高易发、高易发、中等易发和低易发4个级别,结果表明滑坡易发性受巴塘断裂带和河流控制显著,极高易发区和高易发区主要分布在巴塘断裂带、金沙江和巴曲河谷及一级支流两侧,中等易发区主要分布在巴曲各支流中上游,低易发区主要分布在人类工程活动弱的高山地带以及地形相对平缓的区域。滑坡易发性评价结果很好地反映了巴塘断裂带现今滑坡发育分布特征,对该区重大工程规划建设和防灾减灾具有科学指导意义。     

基于频率比与集成学习的滑坡易发性评价——以金沙江上游巴塘—德格河段为例

金沙江上游巴塘—德格河段地处青藏高原东部,该区地质、地形、地貌极其复杂,滑坡灾害最为发育,开展区域滑坡易发性评价对防灾减灾工作有着重要的意义。本文以金沙江上游巴塘—德格河段为研究区,在滑坡编录与野外实际调查的基础上,通过对滑坡分布规律和影响因素分析,选取高程、坡度、坡向、曲率、地形起伏度、地表切割度、地表粗糙度、地层岩性、断层、水系和道路等11个影响因子,构建了滑坡易发性评价指标体系。利用皮尔森系数去除高相关性影响因子,运用频率比方法定量分析各个因子与滑坡发育的关系。通过频率比模型选取非滑坡样本,采用集成学习算法模型进行滑坡易发性评价,根据易发性指数将研究区划分为极高易发区、高易发区、中易发区、低易发区及极低易发区5个等级。由滑坡易发性分区图和ROC曲线表明,高和极高易发区主要沿金沙江沿岸和沟谷分布,随机森林模型的成功率曲线下面积AUC=0.84,历史滑坡灾害位于高-极高易发区的灾害数占总滑坡数的84.8%,梯度提升树模型的成功率曲线下面积AUC=0.79,历史滑坡灾害位于高-极高易发区灾害数占总滑坡数的79.3%。由AUC值和历史灾害的分布可知,随机森林模型比梯度提升树模型在本研究区滑坡易发性评价中有着更好的评价精度和更高的预测能力。     

基于深度学习的CZ铁路康定—理塘段滑坡易发性评价

CZ铁路康定至理塘段地处青藏高原东部边缘,区域内地形地貌多变、地质构造复杂,滑坡灾害极其发育,严重威胁着CZ铁路康定至理塘段的规划建设和未来安全运行。因此,选取高程、坡向、平面曲率、剖面曲率、地形起伏度、地表切割度、地形湿度指数、归一化植被指数、岩性、距断层距离、距河流距离、距道路距离共计12个影响因子构建滑坡空间数据库,采用深度学习的卷积神经网络(convolutional neural network,CNN)模型进行滑坡易发性评价,根据易发性指数将研究区划分为极高易发区(13.76%)、高易发区(14.00%)、中易发区(15.86%)、低易发区(18.17%)、极低易发区(38.21%)5个等级,并与人工神经网络(artificial neural network,ANN)模型进行对比。结果表明,CNN模型的评价精度AUC(0.87)大于ANN(0.84)模型,且极高易发区的频率比值高于ANN模型,CNN模型在本研究区有着更高的预测能力;极高和高易发区主要分布在水系较为发育的地区,沿着雅砻江和其他河流两侧2 km范围内呈带状分布。滑坡易发性评价结果较好地反映了研究区滑坡灾害发育的分布现状,能够为该区的CZ铁路建设和未来安全运行过程中的防灾减灾工作提供科学的依据。     

基于GIS的白龙江流域甘肃段滑坡易发性评价

在甘肃省白龙江流域地质灾害资料收集及现场调查的基础上, 统计分析了该区滑坡发育与地层岩性、坡度、坡向、高程、断裂、植被等因素之间的关系, 建立了白龙江流域滑坡易发性评价指标体系。采用基于GIS的层次分析法评价模型, 完成了滑坡易发性分区评价, 将研究区滑坡按易发程度划分为高易发区、中易发区、低易发区和极低易发区, 其中, 高易发区占研究区总面积的13.59%, 主要分布在断裂带、白龙江两侧以及软弱岩土体分布的区域; 中易发区占27.85%;主要分布在白龙江支流以及主要道路两侧的一定范围内; 低易发区占33.09%, 主要分布在海拔相对较高、植被覆盖度较高、基本上无断裂带通过的区域; 其余区域为极低易发区, 占25.46%。对比分析显示评价结果与实际滑坡发育情况吻合, 可以较好地反映区内滑坡灾害发育的总体特征。      

基于深度神经网络模型的雅安市滑坡易发性评价CSCD

准确的滑坡易发性评价结果是滑坡风险评估的基础,对防灾减灾工作有着重要的意义。文章以雅安市为研究区,在野外地质调查的基础上,选取高程、坡度、坡向、平面曲率、剖面曲率、地形湿度指数、泥沙输运指数、径流强度指数、归一化植被指数、年均降雨量、地震动峰值加速度、地形起伏度、距断层距离、地层岩性、距河流距离、距道路距离等16个因子,构建研究区滑坡易发性评价指标体系,采用度神经网深络(DNN)模型进行滑坡易发性评价,根据易发性指数将研究区划分为极高易发区(12.2%)、高易发区(7.0%)、中易发区(9.8%)、低易发区(17.0%)、极低易发区(54.1%)五个等级,并与人工神经网络(ANN)模型进行对比,用ROC曲线的AUC值进行精度检验。结果表明,DNN模型的评价精度AUC(0.99)大于ANN(0.96)模型。因此,相比ANN模型,DNN模型在该研究区有着更好的拟合能力和预测能力,滑坡极高和高易发区主要分布于雅安市人类工程活动强烈的低海拔地区,沿着道路和水系分布,距道路距离、高程、年均降雨量是影响雅安滑坡发育的主要影响因子。     

基于逻辑回归–信息量的川藏交通廊道滑坡易发性评价

川藏交通廊道位于青藏高原中东部,是世界上隆升和地貌演化最快的区域之一。在内外动力耦合作用下,区内滑坡灾害极其发育,严重制约着公路、铁路和水电工程的规划建设。在区域地质资料收集和整理的基础上,选取岩性、坡度、坡向、坡形、地形起伏度、地形粗糙度、断裂密度和河流距离8个因素为评价因子,结合传统信息量和逻辑回归模型的优势,采用逻辑回归–信息量模型对研究区滑坡进行易发性评价。通过对评价因子的多重共线性和显著性检验,得到评价因子不存在多重共线性且均对滑坡发生具有显著影响。采用ROC曲线对评价结果进行检验,其AUC值为0.81,表明评价模型能很好地预测滑坡的发生。易发性评价结果表明:研究区高易发区主要集中龙门山断裂带、金沙江断裂带、澜沧江断裂带、怒江断裂带、边坝–洛隆断裂带等大型活动断裂带控制区,以及区内坡度陡峭、地形起伏度大的大型河流深切河谷的两岸;中易发区在区内分布广泛,主要分布在岸坡较陡、地形起伏度中等的大型河流支流的两岸。研究结果有利于加深对川藏交通廊道滑坡发育分布的认识,也可为研究区的工程规划建设和防灾减灾提供科学依据。     

GIS支持下CF与Logistic回归模型耦合的九寨沟景区滑坡易发性评价

2017年8月8日九寨沟M_S7.0地震诱发了数以千计的崩滑体,产生的大量松散固体碎屑在降雨作用下极易启动转化为新的滑坡或泥石流形成次生灾害,因此对九寨沟景区进行滑坡易发性评价尤为必要。基于震前、震后高精度遥感影像对比分析结合现场调查,共获取1047处滑坡,总面积为3.88 km2。在分析滑坡发育分布与影响因素关系的基础上,本文选取了构造因子、地形因子、地质因子及其他因子等9个指标,采用确定性系数(CF)模型、逻辑回归(Logistic)模型以及两种模型耦合分析进行滑坡易发性评价。研究结果表明,坡度、坡向、高程和地层岩性是影响滑坡分布的主要因子;研究区被划分为低易发区(60.72%)、中度易发区(24.18%)、高易发区(9.89%)和极高易发区(5.21%),高-极高易发区基本沿沟谷分布,面积为99 km2,其中熊猫海、老虎海周边均为滑坡极高易发区;采用耦合模型比单一模型评价结果更加合理,其结果可作为景区滑坡防治和分段分时开放的参考依据。     

基于斜坡单元的山区城镇滑坡灾害易发性评价：以康定为例

山区地质灾害易发性评价对城镇地质灾害风险管理具有重要意义。本文以康定市为例,以斜坡单元为最小评价单元,选取高程、坡度、坡向、曲率、工程地质岩组、距道路距离、距断裂距离、距水系距离和斜坡结构等9个滑坡影响因子,根据各因子滑坡面积比曲线与证据权值曲线的突变点,划分滑坡影响因子二级状态,并对各影响因子进行相关性分析,剔除相关性较高的距道路距离因子,在此基础上,采用证据权模型进行滑坡易发性评价。对已有治理工程的斜坡单元,本文尝试利用折减系数法对其易发性进行进一步评价。结合现场调查,将研究区滑坡易发性程度划分为：极高易发、高易发、中等易发、低易发。评价结果表明,自然工况下极高易发区主要位于康定市炉城镇以及研究区北侧二道桥村一带,高易发区主要位于雅拉河、折多河与瓦斯沟河谷两侧,对治理工程所在的斜坡单元进行折减后,极高易发区面积由11.21%降至8.42%,滑坡比率由4.03降低至2.3,研究结果符合实际情况,模型精度达77.8%。评价结果较好地反映了康定市区的滑坡易发性分布情况,可为城镇精细化评价提供一定的参考依据。     

基于RBF神经网络?信息量耦合模型的滑坡易发性评价

滑坡易发性评价是滑坡灾害管理的基础工作,也是制定各项防灾减灾措施的重要依据。针对传统的信息量模型在评价过程中确定权重值存在准确性不高的缺点,文章提出RBF神经网络和信息量耦合模型。以甘肃省岷县为研究区,筛选坡度等9个指标因子构建了滑坡灾害易发性评价指标体系,应用RBF神经网络-信息量耦合模型（RBFNN-I）进行滑坡灾害易发性评价,利用合理性检验和ROC曲线对模型的评价结果进行精度检验。结果表明:（1）RBFNN-I模型的AUC值为0.853,相比单一的RBFNN和I模型分别提高了6.3%和9.7%,说明RBFNN-I模型具有更好的评价精度;（2）岷县滑坡灾害的极高易发区和高易发区主要分布在临潭—宕昌断裂带、洮河及其支流、闾井河和蒲麻河两侧河谷地带,距断层距离、降雨量、距道路距离和NDVI是影响岷县滑坡灾害分布的主控因子。     

川藏铁路加查—朗县段地质灾害发育特征研究

川藏铁路加查至朗县段位于青藏高原东南部雅鲁藏布江中游,地形地貌复杂、构造活动强烈,是我国地质灾害高易发区,崩塌、滑坡和泥石流等地质灾害发育密度大、危害严重。在资料收集和遥感解译的基础上,对川藏铁路加查—朗县段的地质灾害进行野外调查,在铁路线两侧各5 km约780 km2范围内发现了崩塌、滑坡和泥石流共139处,沿雅鲁藏布江断裂带新发现拉岗村高速远程滑坡和日阿莫大型滑坡,并研究了该区地质灾害的发育特征和形成机理。调查结果表明：在雅鲁藏布江断裂对区域地貌和岩体结构控制作用下,崩塌、滑坡等地质灾害沿断裂带呈带状密集分布是该区地质灾害发育分布的主要特征之一,约有53%的崩塌滑坡滑动方向垂直于断裂走向,30%的崩塌滑坡与断裂带走向近于平行;在地壳强烈隆升和河流侵蚀作用下,雅鲁藏布江宽谷段和峡谷段的地质灾害发育特征具有明显差异;断裂活动特别是断裂剧烈活动诱发地震导致该区具有高速远程滑坡发生的背景,如拉岗村高速远程滑坡;在断裂活动、降雨、人类工程活动等内外动力耦合作用下该区地质灾害形成机理更加复杂,部分滑坡稳定性差且多次发生活动,给该区重大工程规划建设和防灾减灾造成重要影响。     

基于SBAS InSAR的鲜水河断裂带蠕滑型滑坡特征研究

鲜水河断裂带是青藏高原东部川滇地块的一条重要边界断裂,全新世以来活动强烈,断裂带沿线岩土体结构破碎强烈,在断裂活动诱发地震、断裂蠕滑和强降雨等因素作用下,断裂带沿线滑坡、泥石流等地质灾害发育密度大,危害严重。在前人研究的基础上,采用短基线集(SBAS InSAR)的方法,基于日本对地观测卫星(ALOS 1)所获得的2007—2011年期间15景PALSAR数据,对鲜水河断裂带道孚至炉霍段的活动速率进行分析计算,获取了该段断裂带内蠕滑型滑坡5年间的时间序列形变特征。研究结果表明：鲜水河断裂带道孚至炉霍段近年来以蠕滑滑动为主,蠕滑速率为(94±078) mm/a,断裂的蠕滑作用对区域构造应力场和断裂带内滑坡具有重要的控制作用,表现为距离鲜水河断裂带越近,影像间相干性越强,稳定的相干点越多,干涉效果越好,滑坡滑动累计位移越大。沿鲜水河断裂道孚至炉霍段,共识别出98个蠕滑型滑坡,沿鲜水河断裂带两侧呈线性展布,并分析了典型蠕滑型滑坡的地表形变特征。基于SBAS InSAR的雷达数据处理方法,可以有效地分析地表的缓慢变形以及区域性蠕滑型滑坡的发育发展变化规律,研究结果对于鲜水河断裂带沿线防灾减灾及类似构造活动地区的地质灾害研究具有一定的指导作用。     

基于确定性系数的降雨型滑坡影响因子敏感性分析

本文选择东南沿海地区具有典型降雨型滑坡的淳安县作为研究区,在完成全县地质灾害详细调查的基础上,选取高程、坡度、坡向、曲率、工程地质岩组、距断层距离、距道路距离、土地利用和植被等9个滑坡影响因子,利用GIS技术与确定性系数分析方法,对这9个影响因子开展敏感性分析。研究结果表明:(1) 寒武、震旦、石炭和白垩系是滑坡易发地层,侵入岩组、紫红色砂岩、碳酸盐岩夹碎屑岩、碳酸盐岩为主的岩组是滑坡高敏感性岩组;滑坡受断层影响总体上随着距离断层由近及远逐渐降低;(2) 坡度范围10°~35°是滑坡的易发坡度,30°~35°滑坡数量达到峰值;SE和S等朝南坡向是滑坡最易发坡向;高程范围为100~200m是滑坡最易发区间;凹坡最易发生滑坡,而凸坡则滑坡敏感性最差;非林地、茶叶、竹林和经济林等是滑坡高敏感植被类型;(3) 住宅用地、耕地、园地等与人类活动密切相关的用地类型是滑坡易发地类;距道路距离因子对滑坡敏感性低,相关性不明显。上述各滑坡影响因子最利于滑坡发生的数值区间确定,将为研究区进一步开展降雨型滑坡区域易发性评价及预测奠定基础。     

Structural and geomorphological aspects of the Kat landslides (Tokat—Turkey) and susceptibility mapping by means of GIS

Kat County, which is located in a slope of hilly region and constructed in the side of a mountain along the North Anatolian Fault Zone, is frequently subject to landslides. The slides occur during periods of heavy rainfall, and these events cause destruction to property, roads, agricultural lands and buildings. In the last few decades, a lot of houses and buildings have been damaged and destroyed. Settlement areas have remained evacuated for a long time. The slope instabilities in the study area are a complex landslide extending from north to south containing a lot of landslides. Field investigations, interpretation of aerial photography, analyses of geological data and laboratory tests suggest that some factors have acted together on the slopes to cause the sliding. In the wet season, the slopes became saturated. As the saturation of the earth material on the slope causesa rise in water pressure, the shear strength (resisting forces) decreases and the weight (driving forces) increases; thus, the net effect was to lower the safety factor. Previous failures have affected the rock mass, leading to the presence of already sheared surfaces at residual strengths. The relation between the joint planes and the instability of the slope in the study area was discussed and it was found that the potential slope instabilities are mainly in the directions of NW–SE, NE–SW and N–S. The landslide susceptibility map obtained by using the geographical information system showed that a large area is susceptible and prone to landslides in the northern part of the study area.An erratum to this article can be found at     

基于GIS与WOE-BP模型的滑坡易发性评价

区域滑坡易发性研究对地质灾害风险管理具有重要意义.以往研究中,将多元统计模型与机器学习方法相结合用于滑坡易发性评价的研究较少.以三峡库区万州区为例,首先选取9种指标因子（坡度、坡向、剖面曲率、地表纹理、地层岩性、斜坡结构、地质构造、水系分布及土地利用类型）作为滑坡易发性评价指标.基于证据权模型（weights of evidence,WOE）计算得到的对比度和滑坡面积比与分级面积比的相对大小,对各指标因子进行状态分级;再利用粒子群法优化的BP神经网络模型（PSO-BP）得到各指标因子权重.综合两种模型确定的状态分级权重和指标因子权重（WOE-BP）计算滑坡易发性指数（landslide susceptibility index,LSI）,基于GIS平台得到全区滑坡易发性分区图.结果表明:水系、地层岩性和地质构造是影响万州区滑坡发育的主要指标因子;WOE-BP模型的预测精度为80.8%,优于WOE模型的73.1%和BP神经网络模型的71.6%,可为定量计算指标因子权重和优化滑坡易发性评价提供有效途径.      

A comparison of landslide susceptibility mapping of the eastern part of the North Anatolian Fault Zone (Turkey) by likelihood-frequency ratio and analytic hierarchy process methods

The North Anatolian Fault is known as one of the most active and destructive fault zones which produced many earthquakes with high magnitudes both in historical and instrumental periods. Along this fault zone, the morphology and the lithological features are prone to landslides. Kuzulu landslide, which is located near the North Anatolian Fault Zone, was triggered by snow melting without any precursor, occurred on March 17, 2005. The landslide resulted in 15 deaths and the destruction of about 30 houses at Kuzulu village. There is still a great danger of further landslides in the region. Therefore, it is vitally important to present its environmental impacts and prepare a landslide susceptibility map of the region. In this study, we used likelihood-frequency ratio model and analytical hierarchy process (AHP) to produce landslide susceptibility maps. For this purpose, a detailed landslide inventory map was prepared and the factors chosen that influence landslide occurrence were: lithology, slope gradient, slope aspect, topographical elevation, distance to stream, distance to roads, distance to faults, drainage density and fault density. The ArcGIS package was used to evaluate and analyze all the collected data. At the end of the susceptibility assessment, the area was divided into five susceptibility regions, such as very low, low, moderate, high and very high. The results of the analyses were then verified using the landslide location data and compared with the probability model. For this purpose, an area under curvature (AUC) and the seed cell area index assessments were applied. An AUC value for the likelihood-frequency ratio-based model 0.78 was obtained, whereas the AUC value for the AHP-based model was 0.64. The landslide susceptibility map will help decision makers in site selection and the site-planning process. The map may also be accepted as a basis for landslide risk-management studies to be applied in the study area.     

Landslide susceptibility mapping by frequency ratio and logistic regression methods: an example from Niksar–Resadiye (Tokat,Turkey)

Turkey confronts loss of life and large economic losses due to natural disasters caused by its morphologic structure, geographical placement, and climate characteristics. The Kuzulu (Koyulhisar) landslide, which caused loss of life and property on 17th March 2005, occurred in an area near the country’s most important active fault, the North Anatolian Fault Zone. To mitigate and prevent landslide damages, prediction of landslide susceptibility areas based on probabilistic methods has a great importance. The purpose of this study was to produce a landslide susceptibility map by the logistic regression and frequency ratio methodologies for a 733-km² area near the North Anatolian Fault Zone from the southeast of Niksar to Resadiye in Tokat province. Conditioning parameters, such as elevation, slope gradient, slope aspect, distance to streams, roads, and faults, drainage density, and fault density, were used in the analysis. Before susceptibility analysis, the landslides observed in the area were separated into two groups for use in analysis and verification, respectively. The susceptibility maps produced had five different susceptibility classes such as very low, low, moderate, high, and very high. To test the performance of the susceptibility maps, area under curve (AUC) approach was used. For the logistic regression method, the AUC value was 0.708; while for the frequency rate method, this value was 0.744. According to these AUC values, it could be concluded that the two landslide susceptibility maps obtained were successful.     

National-scale assessment of landslide susceptibility to rank the vulnerability to failure of rock-cut slopes along expressways in Korea

The objective of this study is to perform a preliminary national-scale assessment of the landslide susceptibility of rock-cut slopes along expressways in Korea. A geographic information system (GIS) database was compiled based on data from topographical and geological maps, and rock-cut slope data, including the locations of past landslides. Seven factors (i.e., slope height, slope length, slope gradient, upper slope gradient, lithology, distance from nearest fault, and dip direction of slope) were extracted from the GIS database to assess the relationship between each factor and landslide events. Weight of evidence (WOE), analytic hierarchy process (AHP), and fuzzy logic methods, as well as hybrid methods, were used to establish the rating of classes for each factor, weightings for the factors, and to combine multiple factor layers into landslide-susceptibility maps. A comparison of the results obtained using several different methods, based on the area under curve technique, revealed that the WOE method showed the highest accuracy of 74%. The annual cost of traffic congestion resulting from slope failures was evaluated to identify those rock-cut slopes where detailed investigations and landslide warning systems are required.