Spatial prediction of landslide susceptibility using data mining-based kernel logistic regression,naive Bayes and RBFNetwork models for the Long County area (China)

The main goal of this study is to assess and compare three advanced machine learning techniques, namely, kernel logistic regression (KLR), naïve Bayes (NB), and radial basis function network (RBFNetwork) models for landslide susceptibility modeling in Long County, China. First, a total of 171 landslide locations were identified within the study area using historical reports, aerial photographs, and extensive field surveys. All the landslides were randomly separated into two parts with a ratio of 70/30 for training and validation purposes. Second, 12 landslide conditioning factors were prepared for landslide susceptibility modeling, including slope aspect, slope angle, plan curvature, profile curvature, elevation, distance to faults, distance to rivers, distance to roads, lithology, NDVI (normalized difference vegetation index), land use, and rainfall. Third, the correlations between the conditioning factors and the occurrence of landslides were analyzed using normalized frequency ratios. A multicollinearity analysis of the landslide conditioning factors was carried out using tolerances and variance inflation factor (VIF) methods. Feature selection was performed using the chi-squared statistic with a 10-fold cross-validation technique to assess the predictive capabilities of the landslide conditioning factors. Then, the landslide conditioning factors with null predictive ability were excluded in order to optimize the landslide models. Finally, the trained KLR, NB, and RBFNetwork models were used to construct landslide susceptibility maps. The receiver operating characteristics (ROC) curve, the area under the curve (AUC), and several statistical measures, such as accuracy (ACC), F-measure, mean absolute error (MAE), and root mean squared error (RMSE), were used for the assessment, validation, and comparison of the resulting models in order to choose the best model in this study. The validation results show that all three models exhibit reasonably good performance, and the KLR model exhibits the most stable and best performance. The KLR model, which has a success rate of 0.847 and a prediction rate of 0.749, is a promising technique for landslide susceptibility mapping. Given the outcomes of the study, all three models could be used efficiently for landslide susceptibility analysis.

     

Application of GIS-based statistical modeling for landslide susceptibility mapping in the city of Azazga,Northern Algeria

Landslide susceptibility mapping is a necessary tool in order to manage the landslides hazard and improve the risk mitigation. In this research, we validate and compare the landslide susceptibility maps (LSMs) produced by applying four geographic information system (GIS)-based statistical approaches including frequency ratio (FR), statistical index (SI), weights of evidence (WoE), and logistic regression (LR) for the urban area of Azazga. For this purpose, firstly, a landslide inventory map was prepared from aerial photographs and high-resolution satellite imagery interpretation, and detailed fieldwork. Seventy percent of the mapped landslides were selected for landslide susceptibility modeling, and the remaining (30%) were used for model validation. Secondly, ten landslide factors including the slope, aspect, altitude, land use, lithology, precipitation, distance to drainage, distance to faults, distance to lineaments, and distance to roads have been derived from high-resolution Alsat 2A satellite images, aerial photographs, geological map, DEM, and rainfall database. Thirdly, we established LSMs by evaluating the relationships between the detected landslide locations and the ten landslides factors using FR, SI, LR, and WoE models in GIS. Finally, the obtained LSMs of the four models have been validated using the receiver operating characteristics curves (ROCs). The validation process indicated that the FR method provided more accurate prediction (78.4%) in generating LSMs than the SI (78.1%),WoE (73.5%), and LR (72.1%) models. The results revealed also that all the used statistical models provided good accuracy in landslide susceptibility mapping.

     

Assessing landslide susceptibility using Bayesian probability-based weight of evidence model

This study aims to demonstrate the application of a Bayesian probability-based weight of evidence model to map landslide susceptibility in the Tevankarai stream watershed, Kodaikkanal, India. Slope gradient, relief, aspect, curvature, land use, soil, lineament density, flow accumulation and proximity to roads were the landslide conditioning factors we considered in order to assess susceptibility. The weight of evidence model uses the prior probability of occurrence of a landslide event to identify areas prone to landslides based on the relative contributions of landslide conditioning factors. A pair-wise test of conditional independence was performed for the above factors, allowing the combination of conditioning factors to be analyzed. The contrast (difference of W ⁺ and W ^?) was used as weight for each factor’s type. The best observed combination consisted of the relief, slope, curvature, land use and distance to road factors, showing an accuracy of 86.1 %, while the accuracy of the map with all factors was 83.9 %.     

Identification of landslide hazard and risk ‘hotspots’ in Europe

Landslides are a serious problem for humans and infrastructure in many parts of Europe. Experts know to a certain degree which parts of the continent are most exposed to landslide hazard. Nevertheless, neither the geographical location of previous landslide events nor knowledge of locations with high landslide hazard necessarily point out the areas with highest landslide risk. In addition, landslides often occur unexpectedly and the decisions on where investments should be made to manage and mitigate future events are based on the need to demonstrate action and political will. The goal of this study was to undertake a uniform and objective analysis of landslide hazard and risk for Europe. Two independent models, an expert-based or heuristic and a statistical model (logistic regression), were developed to assess the landslide hazard. Both models are based on applying an appropriate combination of the parameters representing susceptibility factors (slope, lithology, soil moisture, vegetation cover and other- factors if available) and triggering factors (extreme precipitation and seismicity). The weights of different susceptibility and triggering factors are calibrated to the information available in landslide inventories and physical processes. The analysis is based on uniform gridded data for Europe with a pixel resolution of roughly 30 m × 30 m. A validation of the two hazard models by organizations in Scotland, Italy, and Romania showed good agreement for shallow landslides and rockfalls, but the hazard models fail to cover areas with slow moving landslides. In general, the results from the two models agree well pointing out the same countries with the highest total and relative area exposed to landslides. Landslide risk was quantified by counting the number of exposed people and exposed kilometers of roads and railways in each country. This process was repeated for both models. The results show the highest relative exposure to landslides in small alpine countries such as Lichtenstein. In terms of total values on a national level, Italy scores highest in both the extent of exposed area and the number for exposed population. Again, results agree between the two models, but differences between the models are higher for the risk than for the hazard results. The analysis gives a good overview of the landslide hazard and risk hotspots in Europe and allows a simple ranking of areas where mitigation measures might be most effective.     

A method for quick assessment of earthquake-triggered landslide hazards: a case study of the Mw6.1 2014 Ludian,China earthquake

Rapid assessment of the distribution of earthquake-triggered landslides is an important component of effective disaster mitigation. The effort should be based on both seismic landslide susceptibility and the ground shaking intensity, which is usually measured by peak ground acceleration (PGA). In this paper, we address this issue by analyzing data from the Mw6.1 2014 Ludian, China earthquake. The Newmark method of rigid-block modeling was applied to calculate the critical acceleration of slopes in the study area, which serve as measurement of slope stability under seismic load. The assessment of earthquake-triggered landslide hazard was conducted by comparing these critical accelerations with the distribution of known PGA values. The study area was classified into zones of five levels of landslide hazard: high, moderate high, moderate, light, and very light. Comparison shows that the resulting landslide hazard zones agree with the actual distribution of earthquake-triggered landslides. Nearly 70% of landslides are located in areas of high and moderately high hazard, which occupy only 17% of the study region. This paper demonstrates that using PGA, combined with the analysis of seismic landslide susceptibility, allows a reliable assessment of earthquake-triggered landslides hazards. This easy-operation mapping method is expected to be helpful in emergency preparedness planning, as well as in seismic landslide hazard zoning.

     

三峡库区万州区滑坡灾害易发性评价研究

 滑坡灾害易发性研究在滑坡灾害风险管理与城市规划等方面具有非常重要的现实意义。以往的研究中,鲜有对指标因子状态划分作有关深入分析和讨论的。鉴于此,以滑坡灾害频发的三峡库区万州区为研究对象：首先,选取影响滑坡发生的7个致灾因子(地层岩性、地质构造、水系分布、坡度、坡向、坡体结构及土地利用)作为滑坡易发性的评价指标,依据各指标条件下滑坡累计发生频率曲线斜率的变化,并结合滑坡面积比和分级面积比曲线对指标因子的状态进行分级;其次,根据全区655个历史滑坡数据,分别运用信息量模型和逻辑回归模型建立各自的滑坡易发性评价体系;再则,采用快速聚类法(K-means cluster)对以上2种方法所得到的易发性结果进行分级,并基于GIS平台,得到全区滑坡易发性区划图;最后,从模型结果、精度、适用条件等方面对2个模型进行讨论和比较,研究结果表明：信息量模型和逻辑回归模型的预测精度分别为73.0%和54.9%,前者预测能力要优于后者。     

A novel hybrid intelligent model of support vector machines and the MultiBoost ensemble for landslide susceptibility modeling

The main aim of this study is to propose a novel hybrid intelligent model named MBSVM which is an integration of the MultiBoost ensemble and a support vector machine (SVM) for modeling of susceptibility of landslides in the Uttarakhand State, Northern India. Firstly, a geospatial database for the study area was prepared, which includes 391 historical landslides and 16 landslide-affecting factors. Then, the sensitivity of different combinations of these factors for modeling was validated using the forward elimination technique. The MBSVM landslide model was built using the datasets generated from the best selected factors and validated utilizing the area under the receiver operating characteristic (ROC) curve (AUC), statistical indexes, and the Wilcoxon signed-rank test. Results show that this novel hybrid model has good performance both in terms of goodness of fit with the training dataset (AUC = 0.972) and the capability to predict landslides with the testing dataset (AUC = 0.966). The efficiency of the proposed model was then validated by comparison with logistic regression (LR), a single SVM, and another hybrid model of the AdaBoost ensemble and an SVM (ABSVM). Comparison results show that the MBSVM outperforms the LR, single SVM, and hybrid ABSVM models. Thus, the proposed model is a promising and good alternative tool for landslide hazard assessment in landslide-prone areas.

     

An ensemble landslide hazard model incorporating rainfall threshold for Mt. Umyeon,South Korea

In this study, a new ensemble method was developed to assess landslide hazard models in Mt. Umyeon, South Korea, using the results of a physically based model as a conditioning factor (CF). Hydrological conditions were obtained from the national-scale rainfall threshold. To incorporate rainfall threshold in landslide initiation, national landslide inventory data were used to prepare I-D and C-D thresholds. A series of factor of safety (FS) distribution maps were prepared using a physically based model with a 12-h cumulative rainfall threshold. We created an ensemble model to overcome limitations in the physically based model, which could not incorporate important environmental variables such as hydrology, forest, soil, and geology. To determine the effect of CFs on landslide distribution, spatial data layers of elevation, drainage proximity, soil drainage characters, stream power index, sediment transport index, topographic wetness index, forest type, forest density, tree diameter, soil type geology, and the FS distribution map were analyzed in a maximum entropy-based machine learning algorithm. Validation was performed with a receiver operating characteristic curve (ROC). The ROC showed 65.9% accuracy in the physically based model, whereas the ensemble model had higher accuracy (79.6%) and a prediction rate of 89.7%. The ensemble landslide hazard model is a new approach, incorporating the FS distribution map into the available independent environmental variables.

     

地理信息系统/遥感技术支持下三峡库区青干河流域滑坡危险性评价

以统计模型为基础、地理信息系统作为工具的滑坡灾害评价模式已经得到普遍认可和使用,数字高程模型(DEM)、遥感影像、区域地质调查资料已经成为区域滑坡评价研究的因子数据源。选择三峡库区青干河流域顺向坡滑坡多发地段为研究区,在滑坡编目数据库基础上,通过:(1)数字高程模型获取高程、坡度、地形聚水能力因子;(2)遥感影像获取植被指数;(3)区域地质调查资料、数字高程模型计算斜坡类型定量因子TOBIA指数及获取岩石地层单元因子。采用二分类变量逻辑回归评价方法对上述6种因子建立滑坡危险性评价模型,开展地理信息系统/遥感技术支持下顺向坡滑坡危险性评价研究。研究结果表明,根据模型概率值分布和已知滑坡发育关系,可以将研究区划分为高危险区、中等危险区、低危险区3个等级,高危险区包含70%已知滑坡,中等危险区包含14%已知滑坡,评价结果和实际滑坡发育情况吻合,合理地反映区内滑坡灾害发育的总体特征。     

Enhancing the accuracy of rainfall-induced landslide prediction along mountain roads with a GIS-based random forest classifier

Along mountain roads, rainfall-triggered landslides are typical disasters that cause significant human casualties. Thus, to establish effective mitigation measures, it would be very useful were government agencies and practicing land-use planners to have the capability to make an accurate landslide evaluation. Here, we propose a machine learning methodology for the spatial prediction of rainfall-induced landslides along mountain roads which is based on a random forest classifier (RFC) and a GIS-based dataset. The RFC is used as a supervised learning technique to generalize the classification boundary that separates the input information of ten landslide conditioning factors (slope, aspect, relief amplitude, toposhape, topographic wetness index, distance to roads, distance to rivers, lithology, distance to faults, and rainfall) into two distinctive class labels: ‘landslide’ and ‘non-landslide’. Experimental results with a cross validation process and sensitivity analysis on the RFC model parameters reveal that the proposed model achieves a superior prediction accuracy with an area under the curve  of 0.92. The RFC significantly outperforms other benchmarking methods, including discriminant analysis, logistic regression, artificial neural networks, relevance vector machines, and support vector machines. Based on our experimental outcome and comparative analysis, we strongly recommend the RFC as a very capable tool for spatial modeling of rainfall-induced landslides.

     

Assessment and validation of GIS-based landslide susceptibility maps: a case study from Feltrino stream basin (Central Italy)

Landslide susceptibility studies focus on producing susceptibility maps starting from landslide inventories and considering the main conditioning factors. The validity of susceptibility maps must be verified in terms of model accuracy and prediction skills. This paper deals with a GIS-based landslide susceptibility analysis and relative validation in a hilly-coastal test-area in Adriatic Central Italy. The susceptibility analysis was performed via bivariate statistics using the Landslide-Index method and a detailed (field-based) landslide inventory. Selection and mapping of conditioning factors and landslide inventories was derived from detail geomorphological analyses of the study area. The susceptibility map was validated using recent (shallow) landslides in terms of both model accuracy and prediction skills, via Success rate and Prediction rate curves, respectively. In addition, a pre-existing official landslide inventory was applied to the model to test whether it can be used when a detailed (field-based) inventory is not available, thereby extending its usability in similar physiographic regions. The outcome of this study reveals that slope and lithology are the main conditioning factor of landslides, but also highlights the key role of surficial deposits in susceptibility assessment, for both their type and thickness. The validation results show the effectiveness of the susceptibility model in both model accuracy and prediction skills given the good percentage of correctly classified landslides. Moreover, comparison of the susceptibility map with the official Regional landslides inventory proves the possibility of using the developed susceptibility model also in the absence of detailed landslide mapping, by considering inventories that are already available.     

Probabilistic stability analysis of Bazimen landslide with monitored rainfall data and water level fluctuations in Three Gorges Reservoir,China

Landslide is a common geological hazard in reservoir areas and may cause great damage to local residents’ life and property. It is widely accepted that rainfall and periodic variation of water level are the two main factors triggering reservoir landslides. In this study, the Bazimen landslide located in the Three Gorges Reservoir (TGR) was back-analyzed as a case study. Based on the statistical features of the last 3-year monitored data and field instrumentations, the landslide susceptibility in an annual cycle and four representative periods was investigated via the deterministic and probabilistic analysis, respectively. The results indicate that the fluctuation of the reservoir water level plays a pivotal role in inducing slope failures, for the minimum stability coefficient occurs at the rapid decline period of water level. The probabilistic analysis results reveal that the initial sliding surface is the most important area influencing the occurrence of landslide, compared with other parts in the landslide. The seepage calculations from probabilistic analysis imply that rainfall is a relatively inferior factor affecting slope stability. This study aims to provide preliminary guidance on risk management and early warning in the TGR area.     

Landslide hazard assessment and judgment of reliability: a geomechanical approach

Landslide hazard maps are often defined as reliable a posteriori, in accordance with the real landslides occurring from the time of the map production. However, to be useful for planning, a reliability judgment concerning the hazard mapping should be a priori, based on data uncertainty characterization, and must be driven by the knowledge of the slope instability mechanisms. The landslide hazard assessment, when based on the deterministic diagnosis of the processes, may really lead to really providing clues about how and why the slope could fail (landslide susceptibility) and, possibly, when (landslide hazard). Such deterministic assessment can be pursued only through the interpretation and the geo-hydro-mechanical modelling of the slope equilibrium. In practice, though, the landslide hazard assessment is still seldom dealt with slope modelling, in particular when it addresses intermediate to regional zoning. The paper firstly offers an overview of the key steps of a methodology called the multiscalar method for landslide mitigation, MMLM, which that is a methodological approach for the intermediate to regional landslide hazard assessment using the hydro-mechanical diagnoses of landsliding. The validation of the MMLM to the geologically complex outer sectors of the Southern Apennines (Daunia-Lucanian mountains; Italy) is also delineated, together with a practical approach to incorporate a reliability judgment in the landslide susceptibility/hazard mapping.     

Spatial data analysis and application of evidential belief functions to shallow landslide susceptibility mapping at Mt. Umyeon,Seoul, Korea

Predictive mapping of landslide occurrence at the regional scale was performed at Mt. Umyeon, in the southern part of Seoul, Korea, using an evidential belief function (EBF) model. To generate the landslide susceptibility map, approximately 90 % of 163 actual landslide locations were randomly selected as a training set, and about 10 % of them were used as a validation set. Spatial data sets relevant to landslide occurrence (topographic factors, hydrologic factors, forest factors, soil factors, and geologic factors) were analyzed in a geographic information system environment. In this study, landslide susceptibility was assessed on the basis of mass function assignment (belief, disbelief, uncertainty, and plausibility) and integration within a data-driven approach. The most representative of the resulting integrated susceptibility maps (the belief map) was validated using the receiver operating characteristic (ROC) method. The verification result showed that the model had an accuracy of 74.3 % and a predictive accuracy of 88.1 %. The frequency ratio (FR) model was also used for comparison with the EBF model. Prediction and success rates of 72.1 and 85.9 % were achieved using the FR model. The validation results showed satisfactory agreement between the susceptibility map and the existing landslide location data. The EBF model was more accurate than the FR model for landslide prediction in the study area. The results of this study can be used to mitigate landslide-induced hazards and for land-use planning.     

滑坡的易滑度分区及其概率预报模式

滑坡的发生主要由地形、地质和降雨三方面因素决定。基于区域地质-气象耦合分析的思路,有可能提高区域性滑坡的预报精度。首先通过将降雨条件和地质环境条件相结合的方法,提出一种新的滑坡概率预报模式,利用数理统计分析,提出利用最大24h雨强和前15d实效降雨量作为滑坡灾害发生的短期预报判据;然后以重庆市区作为研究对象,选择岩性组合、地形坡度、边坡形态、岩体结构和水文地质五大因素及其21种状态为预测变量,利用信息量法进行了易滑度的分区;最后,对概率预报滑坡的可行性进行了实例分析和探讨。     

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

 以滑坡灾害发育较多的三峡库区万州区为研究区,基于指标因素状态分级和因素相关性分析结果,选取坡度、坡向、坡体结构、地层岩性、地质构造、水的作用以及土地利用7项影响因素,以全区700多个滑坡灾害点为样本数据,依据各因素状态下发生的滑坡频率曲线和信息量曲线的突变点为等级划分的临界值来确定因素状态,并在此基础上建立易发性评价指标体系。基于GIS的栅格数据模型,应用信息量理论开展研究区易发性评价,研究结果表明：易发性高和较高的区域主要分布在土地利用总体规划中的建设用地、侏罗系中统上沙溪庙组第二、三段(J2s2,J2s3)、库水变动带和河网影响带以及万州城区。统计结果表明,处在高易发和较高易发区面积为1 210 km2,其中高易发区和较高易发区分别占研究区总面积的9.71%和25.9%,研究区易发性评价精度高达87%。本文完整的论述了县域滑坡灾害易发性评价的理论方法和技术路线,并以三峡库区万州区为例开展滑坡灾害易发性评价、结果分析以及预测精度评价等,为该区域滑坡灾害防治规划与预测预报提供技术支持,为全国范围内县域滑坡灾害易发性评价提供理论指导和技术参考。     

基于地理信息系统和信息量模型的滑坡危险性评价——以重庆万州为例

针对滑坡危险性评价，以重庆万州滑坡地质灾害为例，采用地理信息系统技术，选取高程、地层岩性、坡向、坡度、距离河流的远近、距离道路的远近、建筑物分布和到遂宁组和沙溪庙组地层的距离等8个指标作为评价因子，利用信息量模型对万州研究区的滑坡地质灾害进行危险性评价。评价结果表明，地理信息系统和信息量模型能够很好地为滑坡灾害的危险性研究服务，可用来解决过去地质灾害危险性评价中效率低、精度筹、费时、费力等问题，从而实现滑坡地质灾害的信息化和科学化。     

基于GIS的汶川地震滑坡灾害影响因子确定性系数分析

 2008年5月12日14时28分,四川省汶川发生了8.0级大地震,地震诱发了数以万计的滑坡灾害。在大约48 678 km2的区域内,采用震后航空像片与多源卫星影像解译并结合野外调查验证的方法,共圈定出48 007个地震滑坡灾害。在此基础上,选取地层、岩性、断裂、地震烈度、宏观震中、地表破裂调查点、地形坡度、坡向、顺坡向曲率、高程、水系与公路共12个影响因子作为汶川地震诱发滑坡影响因子,利用GIS强大的空间分析能力与确定性系数方法,对这12个影响因子进行敏感性研究。研究结果表明：(1) 寒武与震旦系是地震滑坡易发地层,侵入岩组、灰岩为主的岩组是地震滑坡发育的高敏感性岩组;(2) 地震滑坡受中央断裂影响最大,同时还受控于前山断裂,受后山断裂的影响较小;(3) 地震滑坡易发性分别随着地震烈度、与震中的距离、与地表破裂点距离的增加而减少;(4) 坡度大于40°是地震滑坡的易发坡度,E,ES方向为地震滑坡的易发坡向,高程范围为1 000～2 000 m,尤其是高程1 000～1 500 m范围为地震滑坡易发区;(5) 400 m水系缓冲区和2 000 m公路缓冲区范围内滑坡易发性较高。确定研究区内各地震滑坡影响因子最利于滑坡发生的数值区间,为进一步地震滑坡区域评价及预测奠定基础。     

FRPFP模型滑坡监测预警关联规则挖掘分析——以三峡库区江津到奉节段为例

采用传统的关联规则用于岩土工程监测预警领域的知识发现,在数据庞大情形下单机机器学习实时性差,无法获得多因素综合作用的规则。由于未对前后部项进行约束,得到的关联规则冗余度高,含有大量不符因果逻辑的规则。基于此,提出一种前后部项约束关联规则并行化FRPFP (fore-part and rear-part parallel FP-growth)算法,并在大数据分布式处理平台Spark下进行实现。通过对三峡库区奉节至江津库段滑坡的孕灾因子统计分类,采用7个滑坡发育基础因子和4个滑坡诱导因子作为前部集合,滑坡前缘、中部、后缘监测点位移参数为后部集合,采集研究区25个滑坡11年监测数据。以FRPFP算法为模型架构基于关联规则的滑坡监测预警大数据系统,设计区域滑坡危险性规则挖掘、典型滑坡危险性规则挖掘、滑坡发生原因分析挖掘3个功能,用于库岸滑坡稳定性预测和分析,为认清库岸滑坡的破坏机制和提升其预报水平提供新的思路。     

降雨影响下的区域滑坡危险性动态评价研究——以三峡库区万州主城区为例

 降雨引发的滑坡具有区域性的群发效应,能够在短时间内造成大量的灾难性损失。基于此,提出一种可考虑不同降雨期影响的区域滑坡危险性评价方法。该方法以瞬态降雨入渗的区域斜坡稳定性计算模型为基础,将滑坡危险性定义为在一定持续降雨期内各栅格单元体失稳的概率。通过岩土体物理力学参数的不确定性进行各栅格单元体失稳概率的求解,继而获得区域内滑坡的危险性分布。基于ArcGIS软件开发出区域滑坡危险性动态评价工具。以三峡库区万州主城区为例,详细介绍危险性评价工具的数据处理过程以及参数选取方法,并以2种不同的降雨工况进行比较计算。现场斜坡稳定性的调查与计算结果的对比及统计分析表明：滑坡的危险性分布图与真实滑坡的稳定性情况基本一致,并在一定程度上反应了该地区斜坡稳定性的时空分布特征,测试并验证了评价工具的正确性。