基于XGBoost模型的三峡库区燕山乡滑坡易发性评价与区划

滑坡易发性评价是精细化滑坡灾害风险评价的基础。为了提升滑坡易发性评价模型的精度和稳健性,以三峡库区万州区燕山乡为例,选取工程地质岩组、堆积层厚度等九个影响因子构建滑坡易发性评价指标体系,应用信息量模型定量分析滑坡发育与指标之间的关系。在此基础上,随机选取70%/30%的滑坡样本作为训练/验证数据集,应用极致梯度提升模型(extreme gradient boosting, XGBoost)开展易发性评价。随后从模型预测精度和模型稳定性两方面将其与决策树模型(decision tree, DT)和梯度提升树模型(gradient boosting decision tree, GBDT)进行对比。结果表明：研究区堆积层滑坡主要受长江水系、堆积层厚度和工程地质岩组影响。XGBoost模型具有最高的准确率(94.3%)和预测精度(97.3%)。在模型稳定性验证中,平均预测精度最高(97.3%),优于DT(91.3%)和GBDT(95.7%),模型标准差和变异系数均为0.01,低于其余两种模型。XGBoost在区域滑坡易发性评价与制图中得到了可靠的结果,为滑坡灾害空间预测提供了新的技术支撑。     

采用多模型融合方法评价滑坡灾害易发性:以湖北省五峰县为例

不同的易发性评价模型可以得到有差异的滑坡空间预测结果,选取最优模型甚至综合各模型的优势是提高易发性评价精度的有效方法。为检验模型融合思路的有效性,以鄂西地区五峰县渔洋关镇为研究区,提取坡度、地层、断层、河流、公路等7个滑坡成因条件,分别采用信息量模型、证据权模型和频率比模型进行滑坡易发性评价;并将3种模型分别进行归一化、主成分分析(PCA,Principal component analysis)和优势融合,得到了6幅易发性分区图。结果表明:优势耦合模型精度最高(90.3%),频率比模型次之(89.7%),归一化融合模型和PCA融合模型分别为89.3%和89.1%,以上4种结果的精度均高于证据权模型(87.7%)和信息量模型(87.6%);6幅预测图对应的评价结论与历史滑坡空间分布的实际情况相符。空间一致性对比结论表明,主成分融合模型与优势耦合模型的同格率高达68%,其预测结果避免了单个模型预测结论带来的偶然性和片面性,说明多模型融合方法与优势耦合模型在提高滑坡易发性预测精度上是可行性的,该思路对其他地区滑坡灾害易发性评价具有借鉴意义。     

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.      

基于信息量与Logistic模型的矿区滑坡灾害易发性评价

针对矿区长期煤矿开采引起的滑坡灾害频发问题,快速高效地模拟和评价矿致滑坡灾害易发性是实现采矿地区科学防灾减灾的关键。基于此,本文应用信息量与Logistic回归模型结合多源高分辨率光学遥感数据等,选取相对高差、坡度、坡向、距断层距离、NDVI、距采空区距离6个滑坡影响因子来评价采煤矿区滑坡灾害易发性。结果表明:(1)信息量与Logistic回归模型耦合的综合预测准确率为96%,信息量模型滑坡预测准确率为95%,实验结果表明耦合模型的预测精度优于单一信息量评价模型,评价模型的合理性和预测精度皆符合检验要求;(2)研究结果也表明了采用信息量+Logistic回归模型耦合能较为客观准确、快速高效地评价地下采矿引起的滑坡灾害易发范围,评价结果可为类似地区高效快速划定滑坡灾害易发区间提供技术支撑。     

基于SOM-I-SVM耦合模型的滑坡易发性评价

在使用机器学习模型对滑坡进行易发性评价时,通常会在滑坡影响范围之外随机选取非滑坡样本点,具有一定的误差。为了提高滑坡易发性评价的精度,将自组织映射（self-organizing map,SOM）神经网络、信息量模型（information,I）以及支持向量机模型（support vector machine,SVM）进行耦合,提出一种基于SOM-I-SVM模型的滑坡易发性评价方法,并将SOM神经网络与K均值聚类算法进行对比,验证模型的可靠性。以十堰市茅箭区为例,首先通过对环境因子的相关性及重要性分析,筛选出距水系距离、坡度、降雨量、距构造距离、相对高差、距道路距离、地层岩性等7个因子,建立滑坡易发性评价指标体系,在此基础上计算出各因子的分级信息量值,并作为模型的输入变量进行滑坡易发性评价。分别采用SOM神经网络和K均值聚类算法选取非滑坡样本,然后将样本数据集代入I-SVM模型预测滑坡易发性。将SVM、I-SVM、KMeans-I-SVM、SOM-I-SVM等4种模型预测精度进行对比,其ROC曲线下面积（AUC）分别为0.82,0.88,0.90,0.91,说明SOM-I-SVM模型能...     

基于信息量和卷积神经网络的黄土高原滑坡易发性评价

黄土高原在地质环境与人类活动的复杂互馈作用下易导致黄土崩滑灾害频发,亟需选择适用性的影响因子和训练模型开展滑坡易发性评价研究.本研究以黄土高原为研究区,基于野外滑坡调查和资料收集,构建涵盖地形地貌、基础地质环境、气象水文、人类活动、土壤物理化学性质以及植被覆盖的评价体系,采用信息量模型（ Ⅳ）分别联接到随机森林模型（RF）和卷积神经网络模型（CNN）构建耦合模型 Ⅳ-RF和 Ⅳ-CNN,开展滑坡易发性评价研究.结果表明,耦合模型（ Ⅳ-RF、 Ⅳ-CNN）的精度均高于独立模型（RF、CNN）,4种模型的AUC值分别为0.916、0.938、0.878、0.853, Ⅳ-CNN具有更强的预测能力和精度. Ⅳ-CNN模型的极高、高、中、低、极低易发性区域面积占比分别为8.78%、7.47%、15.34%、19.82%、47.87%,主要分布在黄土高原南部和东部地质环境复杂和人类活动强烈的山地、黄土梁峁地区.坡度、侵蚀类型、地貌类型、粘粒含量、距道路距离在贡献率分析中排在前5位,是影响滑坡发育的主控因子.本研究旨在为黄土高原滑坡灾害的预测和防治工作提供可靠的科学依据,为滑坡易发性评价研究深化...     

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。     

基于粗糙集和BP神经网络的滑坡易发性评价

区域滑坡易发性评价是国土规划和滑坡中长期防治的重要依据。为进一步提高滑坡易发性评价的准确性,以恩施市龙凤镇为研究区,运用地理信息系统GIS技术,获取了包括工程岩组、坡度、地质构造等在内的13个初始评价因子,利用基于遗传约简算法的粗糙集理论对初始评价因子进行属性约简,去掉冗余属性后获得最小约简,即8个核评价因子:工程岩组、高程、地形曲率、道路、水系、坡度、坡向、径流强度指数,并以此作为BP神经网络的输入层,构建RS-BPNN预测模型,获得滑坡易发性指数LSI及滑坡易发性等级分区图。其中高易发区面积占总面积的12.82%,该区包含的滑坡面积占总滑坡面积的78.11%,通过ROC曲线测试,模型预测精度为90.9%。结果表明,RS-BPNN模型预测性能良好,进一步提高了滑坡易发性评价的精度和准确性,有较高的工程实用价值。      

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。     

基于不同耦合模型的县域滑坡易发性评价对比分析

为有效预测县域滑坡发生的空间概率,探索不同统计学耦合模型滑坡易发性定量评价结果的合理性和精度,以四川省普格县为研究对象。选取坡度、坡向、高程、工程地质岩组、断层和斜坡结构等6项孕灾因子作为评价指标体系,基于信息量模型（I）、确定性系数模型（CF）、证据权模型（WF）、频率比模型（FR）分别与逻辑回归模型（LR）耦合开展滑坡易发性评价。结果表明:各耦合模型评价结果和易发程度区划均是合理的,极高易发区主要分布于则木河、黑水河河谷两侧斜坡带,面积介于129.04～183.43 km2（占比6.77%～9.62%）,各模型评价精度依次为WF-LR模型（AUC=0.869）>I-LR模型（AUC=0.868）>CF-LR模型（AUC=0.866）>NFR-LR模型（AUC=0.858）。研究成果可为川西南山区县域滑坡易发性定量评估提供重要参考。     

Ensemble-based landslide susceptibility maps in Jinbu area, Korea

Ensemble techniques were developed, applied and validated for the analysis of landslide susceptibility in Jinbu area, Korea using the geographic information system (GIS). Landslide-occurrence areas were detected in the study by interpreting aerial photographs and field survey data. Landslide locations were randomly selected in a 70/30 ratio for training and validation of the models, respectively. Topography, geology, soil and forest databases were also constructed. Maps relevant to landslide occurrence were assembled in a spatial database. Using the constructed spatial database, 17 landslide-related factors were extracted. The relationships between the detected landslide locations and the factors were identified and quantified by frequency ratio, weight of evidence, logistic regression and artificial neural network models and their ensemble models. The relationships were used as factor ratings in the overlay analysis to create landslide susceptibility indexes and maps. Then, the four landslide susceptibility maps were used as new input factors and integrated using the frequency ratio, weight of evidence, logistic regression and artificial neural network models as ensemble methods to make better susceptibility maps. All of the susceptibility maps were validated by comparison with known landslide locations that were not used directly in the analysis. As the result, the ensemble-based landslide susceptibility map that used the new landslide-related input factor maps showed better accuracy (87.11% in frequency ratio, 83.14% in weight of evidence, 87.79% in logistic regression and 84.54% in artificial neural network) than the individual landslide susceptibility maps (84.94% in frequency ratio, 82.82% in weight of evidence, 87.72% in logistic regression and 81.44% in artificial neural network). All accuracy assessments showed overall satisfactory agreement of more than 80%. The ensemble model was found to be more effective in terms of prediction accuracy than the individual model.     

Quantitative landslide susceptibility mapping at Pemalang area,Indonesia

For quantitative landslide susceptibility mapping, this study applied and verified a frequency ratio, logistic regression, and artificial neural network models to Pemalang area, Indonesia, using a Geographic Information System (GIS). Landslide locations were identified in the study area from interpretation of aerial photographs, satellite imagery, and field surveys; a spatial database was constructed from topographic and geological maps. The factors that influence landslide occurrence, such as slope gradient, slope aspect, curvature of topography, and distance from stream, were calculated from the topographic database. Lithology was extracted and calculated from geologic database. Using these factors, landslide susceptibility indexes were calculated by frequency ratio, logistic regression, and artificial neural network models. Then the landslide susceptibility maps were verified and compared with known landslide locations. The logistic regression model (accuracy 87.36%) had higher prediction accuracy than the frequency ratio (85.60%) and artificial neural network (81.70%) models. The models can be used to reduce hazards associated with landslides and to land-use planning.     

Landslide susceptibility mapping using certainty factor, index of entropy and logistic regression models in GIS and their comparison at Mugling�CNarayanghat road section in Nepal Himalaya

Landslide susceptibility maps are vital for disaster management and for planning development activities in the mountainous country like Nepal. In the present study, landslide susceptibility assessment of Mugling?CNarayanghat road and its surrounding area is made using bivariate (certainty factor and index of entropy) and multivariate (logistic regression) models. At first, a landslide inventory map was prepared using earlier reports and aerial photographs as well as by carrying out field survey. As a result, 321 landslides were mapped and out of which 241 (75?%) were randomly selected for building landslide susceptibility models, while the remaining 80 (25?%) were used for validating the models. The effectiveness of landslide susceptibility assessment using GIS and statistics is based on appropriate selection of the factors which play a dominant role in slope stability. In this case study, the following landslide conditioning factors were evaluated: slope gradient; slope aspect; altitude; plan curvature; lithology; land use; distance from faults, rivers and roads; topographic wetness index; stream power index; and sediment transport index. These factors were prepared from topographic map, drainage map, road map, and the geological map. Finally, the validation of landslide susceptibility map was carried out using receiver operating characteristic (ROC) curves. The ROC plot estimation results showed that the susceptibility map using index of entropy model with AUC value of 0.9016 has highest prediction accuracy of 90.16?%. Similarly, the susceptibility maps produced using logistic regression model and certainty factor model showed 86.29 and 83.57?% of prediction accuracy, respectively. Furthermore, the ROC plot showed that the success rate of all the three models performed more than 80?% accuracy (i.e. 89.15?% for IOE model, 89.10?% for LR model and 87.21?% for CF model). Hence, it is concluded that all the models employed in this study showed reasonably good accuracy in predicting the landslide susceptibility of Mugling?CNarayanghat road section. These landslide susceptibility maps can be used for preliminary land use planning and hazard mitigation purpose.     

Logistic regression versus artificial neural networks: landslide susceptibility evaluation in a sample area of the Serchio River valley,Italy

This article presents a multidisciplinary approach to landslide susceptibility mapping by means of logistic regression, artificial neural network, and geographic information system (GIS) techniques. The methodology applied in ranking slope instability developed through statistical models (conditional analysis and logistic regression), and neural network application, in order to better understand the relationship between the geological/geomorphological landforms and processes and landslide occurrence, and to increase the performance of landslide susceptibility models. The proposed experimental study concerns with a wide research project, promoted by the Tuscany Region Administration and APAT-Italian Geological Survey, aimed at defining the landslide hazard in the area of the Sheet 250 “Castelnuovo di Garfagnana” (1:50,000 scale). The study area is located in the middle part of the Serchio River basin and is characterized by high landslide susceptibility due to its geological, geomorphological, and climatic features, among the most severe in Italy. Terrain susceptibility to slope failure has been approached by means of indirect-quantitative statistical methods and neural network software application. Experimental results from different methods and the potentials and pitfalls of this methodological approach have been presented and discussed. Applying multivariate statistical analyses made it possible a better understanding of the phenomena and quantification of the relationship between the instability factors and landslide occurrence. In particular, the application of a multilayer neural network, equipped for supervised learning and error control, has improved the performance of the model. Finally, a first attempt to evaluate the classification efficiency of the multivariate models has been performed by means of the receiver operating characteristic (ROC) curves analysis approach.     

Comparing data-driven landslide susceptibility models based on participatory landslide inventory mapping in Purwosari area,Yogyakarta, Java

There are different approaches and techniques for landslide susceptibility mapping. However, no agreement has been reached in both the procedure and the use of specific controlling factors employed in the landslide susceptibility mapping. Each model has its own assumption, and the result may differ from place to place. Different landslide controlling factors and the completeness of landslide inventory may also affect the different result. Incomplete landslide inventory may produce significance error in the interpretation of the relationship between landslide and controlling factor. Comparing landslide susceptibility models using complete inventory is essential in order to identify the most realistic landslide susceptibility approach applied typically in the tropical region Indonesia. Purwosari area, Java, which has total 182 landslides occurred from 1979 to 2011, was selected as study area to evaluate three data-driven landslide susceptibility models, i.e., weight of evidence, logistic regression, and artificial neural network. Landslide in the study area is usually affected by rainfall and anthropogenic activities. The landslide typology consists of shallow translational and rotational slide. The elevation, slope, aspect, plan curvature, profile curvature, stream power index, topographic wetness index, distance to river, land use, and distance to road were selected as landslide controlling factors for the analysis. Considering the accuracy and the precision evaluations, the weight of evidence represents considerably the most realistic prediction capacities (79%) when comparing with the logistic regression (72%) and artificial neural network (71%). The linear model shows more powerful result than the nonlinear models because it fits to the area where complete landslide inventory is available, the landscape is not varied, and the occurence of landslide is evenly distributed to the class of controlling factor.     

A comparative study of sequential minimal optimization-based support vector machines,vote feature intervals,and logistic regression in landslide susceptibility assessment using GIS

Landslide susceptibility assessment using GIS has been done for part of Uttarakhand region of Himalaya (India) with the objective of comparing the predictive capability of three different machine learning methods, namely sequential minimal optimization-based support vector machines (SMOSVM), vote feature intervals (VFI), and logistic regression (LR) for spatial prediction of landslide occurrence. Out of these three methods, the SMOSVM and VFI are state-of-the-art methods for binary classification problems but have not been applied for landslide prediction, whereas the LR is known as a popular method for landslide susceptibility assessment. In the study, a total of 430 historical landslide polygons and 11 landslide affecting factors such as slope angle, slope aspect, elevation, curvature, lithology, soil, land cover, distance to roads, distance to rivers, distance to lineaments, and rainfall were selected for landslide analysis. For validation and comparison, statistical index-based methods and the receiver operating characteristic curve have been used. Analysis results show that all these models have good performance for landslide spatial prediction but the SMOSVM model has the highest predictive capability, followed by the VFI model, and the LR model, respectively. Thus, SMOSVM is a better model for landslide prediction and can be used for landslide susceptibility mapping of landslide-prone areas.     

Improvement of statistical landslide susceptibility mapping by using spatial and global regression methods in the case of More and Romsdal (Norway)

Statistical models are one of the most preferred methods among many landslide susceptibility assessment methods. As landslide occurrences and influencing factors have spatial variations, global models like neural network or logistic regression (LR) ignore spatial dependence or autocorrelation characteristics of data between the observations in susceptibility assessment. However, to assess the probability of landslide within a specified period of time and within a given area, it is important to understand the spatial correlation between landslide occurrences and influencing factors. By including these relations, the predictive ability of the developed model increases. In this respect, spatial regression (SR) and geographically weighted regression (GWR) techniques, which consider spatial variability in the parameters, are proposed in this study for landslide hazard assessment to provide better realistic representations of landslide susceptibility. The proposed model was implemented to a case study area from More and Romsdal region of Norway. Topographic (morphometric) parameters (slope angle, slope aspect, curvature, plan, and profile curvatures), geological parameters (geological formations, tectonic uplift, and lineaments), land cover parameter (vegetation coverage), and triggering factor (precipitation) were considered as landslide influencing factors. These influencing factors together with past rock avalanche inventory in the study region were considered to obtain landslide susceptibility maps by using SR and LR models. The comparisons of susceptibility maps obtained from SR and LR show that SR models have higher predictive performance. In addition, the performances of SR and LR models at the local scale were investigated by finding the differences between GWR and SR and GWR and LR maps. These maps which can be named as comparison maps help to understand how the models estimate the coefficients at local scale. In this way, the regions where SR and LR models over or under estimate the landslide hazard potential were identified.     

Predictive landslide susceptibility mapping using spatial information in the Pechabun area of Thailand

For predictive landslide susceptibility mapping, this study applied and verified probability model, the frequency ratio and statistical model, logistic regression at Pechabun, Thailand, using a geographic information system (GIS) and remote sensing. Landslide locations were identified in the study area from interpretation of aerial photographs and field surveys, and maps of the topography, geology and land cover were constructed to spatial database. The factors that influence landslide occurrence, such as slope gradient, slope aspect and curvature of topography and distance from drainage were calculated from the topographic database. Lithology and distance from fault were extracted and calculated from the geology database. Land cover was classified from Landsat TM satellite image. The frequency ratio and logistic regression coefficient were overlaid for landslide susceptibility mapping as each factor’s ratings. Then the landslide susceptibility map was verified and compared using the existing landslide location. As the verification results, the frequency ratio model showed 76.39% and logistic regression model showed 70.42% in prediction accuracy. The method can be used to reduce hazards associated with landslides and to plan land cover.     

Spatial prediction models for shallow landslide hazards: a comparative assessment of the efficacy of support vector machines,artificial neural networks,kernel logistic regression,and logistic model tree

Preparation of landslide susceptibility maps is considered as the first important step in landslide risk assessments, but these maps are accepted as an end product that can be used for land use planning. The main objective of this study is to explore some new state-of-the-art sophisticated machine learning techniques and introduce a framework for training and validation of shallow landslide susceptibility models by using the latest statistical methods. The Son La hydropower basin (Vietnam) was selected as a case study. First, a landslide inventory map was constructed using the historical landslide locations from two national projects in Vietnam. A total of 12 landslide conditioning factors were then constructed from various data sources. Landslide locations were randomly split into a ratio of 70:30 for training and validating the models. To choose the best subset of conditioning factors, predictive ability of the factors were assessed using the Information Gain Ratio with 10-fold cross-validation technique. Factors with null predictive ability were removed to optimize the models. Subsequently, five landslide models were built using support vector machines (SVM), multi-layer perceptron neural networks (MLP Neural Nets), radial basis function neural networks (RBF Neural Nets), kernel logistic regression (KLR), and logistic model trees (LMT). The resulting models were validated and compared using the receive operating characteristic (ROC), Kappa index, and several statistical evaluation measures. Additionally, Friedman and Wilcoxon signed-rank tests were applied to confirm significant statistical differences among the five machine learning models employed in this study. Overall, the MLP Neural Nets model has the highest prediction capability (90.2 %), followed by the SVM model (88.7 %) and the KLR model (87.9 %), the RBF Neural Nets model (87.1 %), and the LMT model (86.1 %). Results revealed that both the KLR and the LMT models showed promising methods for shallow landslide susceptibility mapping. The result from this study demonstrates the benefit of selecting the optimal machine learning techniques with proper conditioning selection method in shallow landslide susceptibility mapping.