旋转森林模型在滑坡易发性评价中的应用研究

以三峡库区万州段为研究区,从多源空间数据中提取29个致灾因子作为区域滑坡易发性分析的评价指标,在数字高程模型基础上采用集水区重叠法划分斜坡单元,构建旋转森林集成学习模型,定量预测滑坡空间易发性,并生成滑坡易发性分区图。在易发性分区图中,高易发区占11.6%,主要分布在万州主城区和长江及支流两岸;不易发区占45.6%,主要分布在人类工程活动低、植被覆盖度高的区域。采用受访者工作特征曲线和曲线下面积对旋转森林模型的滑坡易发性进行评价,结果显示该模型的预测精度为90.7%,其预测能力优于C4.5决策树。研究表明,应用旋转森林进行滑坡易发性评价具有预测能力强、精度高等优点。     

GIS支持下应用PSO-SVM模型预测滑坡易发性

滑坡灾害易发性预测是滑坡监测、预警与评估的关键技术。如何有效地选取评价因子和构建预测模型是滑坡灾害定量预测研究中的难题。本文以三峡库区长江干流岸坡作为研究区,通过地形、地质和遥感等多源数据融合,提取滑坡孕灾环境和诱发因素的信息作为评价因子。在此基础上,针对滑坡灾害的非线性和不确定性特征,采用粒子群算法对支持向量机模型参数进行全局寻优,构建粒子群算法(particle swarm optimization,PSO)-支持向量机(support vector machine,SVM)模型,定量预测滑坡易发性。最后通过分类精度比较分析基于格网单元和对象单元的滑坡易发性预测精度,结果表明,基于对象单元的PSO-SVM预测精度较高,其曲线下面积为0.841 5,Kappa系数为0.849 0,预测结果与野外实际调查情况较为一致,可为三峡库区滑坡防灾减灾工作提供参考。     

一种结合SMOTE和卷积神经网络的滑坡易发性评价方法

大规模的人类工程活动诱发和加剧了滑坡灾害的致灾情况,严重威胁工程安全和环境安全。滑坡易发性评价是滑坡监测预警的关键技术。针对传统滑坡监测手段数据源有限、缺乏挖掘滑坡灾害空间分布特征及其诱发因素的有效方法等问题,以位于三峡库区的中国重庆市万州区为研究区,基于地形、地质和遥感影像等多源数据,首先提取了22个滑坡易发性评价因子,并对这些因子进行多重共线性检验;然后采用合成少数类过采样技术(synthetic minority oversampling technique, SMOTE)解决滑坡和非滑坡样本比例不平衡问题,建立输入训练集;最后构建卷积神经网络(convolutional neural networks,CNN)模型,定量预测滑坡易发性,生成滑坡易发性分区图。采用受试者工作特征曲线分析评价结果,测试数据集模型精度达89.50%,说明该模型是一种高性能的滑坡易发性评价方法。     

顾及样本优化选择的多核支持向量机滑坡灾害易发性分析评价

滑坡灾害易发性分析评价对地质灾害的防治与管理具有重要意义。针对滑坡灾害样本选择策略,单核支持向量机多特征映射不合理的问题,本文提出顾及样本优化选择的多核支持向量机(multiple kernel support vector machine,MKSVM)滑坡灾害易发性分析评价方法。为了保证样本平衡性并提高负样本的合理性,采用相对频率比(relative frequency,RF)综合评价各状态对于滑坡灾害易发性影响的重要程度,实现各评价因子状态的合理划分;利用确定性系数法(certainty factor,CF)计算各评价因子各状态分级影响滑坡灾害的敏感性,并在此基础上进行加权求和得到各栅格单元的滑坡灾害易发性指数,在滑坡灾害易发性指数极低和低易发区内随机选择与滑坡灾害点数目一致的非滑坡灾害点作为负样本数据。利用MKSVM对各特征空间最优核函数进行线性组合,解决了单一核函数映射不合理的问题,提高了模型的分类准确率和预测精度。以湖南省湘西土家族苗族自治州为研究区,从滑坡灾害易发性分区图、分区统计及评价模型精度3个方面对CF样本策略的MKSVM模型、CF样本策略的单核SVM模型、随机样本策略的MKSVM模型、随机样本策略的单核SVM模型进行了对比分析。结果表明,4种模型的受试者工作特征曲线(receiver operating characteristic,ROC)下的面积(area under curve,AUC)分别为0.859、0.809、0.798、0.766,验证了CF样本策略的合理性、有效性及MKSVM模型的可靠性。     

熵指数融入支持向量机的滑坡灾害易发性评价方法——以陕西省为例

滑坡灾害易发性评价可为滑坡灾害风险管理、国土空间规划及滑坡监测提供科学依据。针对现有滑坡灾害易发性评价模型无法消除易发性评价指标因子在量纲、性质等方面的差异,尚未考虑易发性评价指标因子与滑坡灾害相关性,以及精度较高的经典机器学习模型训练效率较低、参数选取困难等问题,引入熵指数(index of entropy,IOE)和粒子群优化(particle swarm optimization,PSO)算法,提出IOE融入支持向量机(support vector machine, SVM)的滑坡灾害易发性评价方法。首先,基于滑坡灾害易发性评价指标因子,利用IOE模型计算SVM的调节因子;然后,采用PSO算法迭代求解SVM最优解,根据SVM二分类得到的隶属度来区分滑坡灾害易发性;最后,以陕西省作为实验区,从滑坡灾害易发性分区图、分区统计及评价模型精度3个方面将所提方法与SVM方法进行了对比,实验结果表明所提方法的准确性、可靠性优于SVM方法。     

基于不同模型和单元的滑坡易发性评价比较

基于地理信息系统的滑坡灾害空间预测研究发展迅速,出现了多种滑坡空间预测模型。在总结滑坡灾害空间预测研究现状的基础上,简要介绍了决策树和支持向量机2种模型的基本原理。以秭归县一个研究区为例,选取11个滑坡影响因子,采用2种不同的研究单元,分别建立了决策树和支持向量机模型并对滑坡易发性作出了评价。结果表明针对同一模型,面向对象单元的滑坡易发性评价精度优于栅格单元;针对同一数据单元,支持向量机模型的滑坡易发性评价精度优于决策树模型。     

层次分析-熵值定权法应用于山区城镇地质灾害易发性评价

我国西南山区很多城镇面临着崩塌、滑坡等地质灾害的威胁,开展山区城镇地质灾害易发性评价研究具有重要意义。本文对重庆市武隆区羊角场镇进行了地质灾害易发性评价研究,在确定区域易发性评价因子的基础上,综合层次分析法和熵权法两种方法确定评价因子的权重,最后采用易发性指数并基于ArcGIS对研究区进行易发性评价和区划。易发性分区结果与已有的地质灾害分布有较好的对应关系。采用的组合赋权法先用层次分析法确定主观权重,然后又结合熵值法对权重进行修正,在一定程度上减少了主观影响,对地质灾害易发性的量化更加准确与合理。本文的研究思路可为山区城镇地质灾害的风险评估和防治提供参考。     

顾及样本敏感性的滑坡易发性评价

滑坡作为一种危害极大的自然地质现象,严重威胁着人民的生命财产安全。因此,科学、准确地评价滑坡体的易发性至关重要。随着机器学习的发展,基于机器学习的滑坡易发性评价逐渐成为研究热点。而在真实情况中,滑坡区域与非滑坡区域面积占比悬殊,这使得机器学习模型的应用存在较严重的样本不均衡问题。本文采用样本敏感性分析方法,综合多个机器学习模型在不同比例的正负滑坡样本集上的表现,以获取最均衡滑坡样本集;并在此样本集基础上采用深度随机森林模型,在示范研究区开展滑坡易发性评价。最终的评价结果接近真实分布,表明本文方法具有较好的有效性。     

融合自编码与密集残差网络的滑坡易发性评价

针对滑坡易发性评价中编录样本受限、特征信息利用不足，导致预测效果不理想的问题，该文提出融合栈式自编码与密集残差网络的滑坡易发性评价方法。将传统自编码网络进行栈化，重构原始数据以强化特征表达；在卷积神经网络中引入跳层连接，通过密集连接的方式构建一种密集残差网络模型提取数据的深层特征。该文以四川省雅安市为研究区开展滑坡易发性评价，将所提模型与逻辑回归模型、CPCNN-RF模型、U-net模型进行了对比分析。实验结果表明：该方法取得最佳精度，受试者工作特征曲线下的面积为0.883,在一定程度上能够较好地适应小样本预测，易发性制图结果更为准确可靠。     

层次分析法的黄河上游滑坡易发性评价

针对黄河上游龙羊峡至积石峡段滑坡灾害分布易发性评价与区划成图问题,该文以ArcGIS为平台,联系评价区的实际特点,选取地貌类型、地层岩性、降雨、断层、坡度为评价因子,运用层次分析法(AHP)确定各评价因子的权重,建立研究区滑坡易发性评价模型,结合GIS的空间分析功能实现研究区内滑坡灾害的易发性区划。结果表明,滑坡灾害主要集中在龙羊峡库区右岸和群科-尖扎盆地。区划结果与野外实际调查基本吻合,为今后GIS应用于地质灾害区划提供了思路,同时可为区内地质单位进行灾害监测提供基础数据和依据。     

Multi-hazard assessment using machine learning and remote sensing in the North Central region of Vietnam

Natural hazards constitute a diverse category and are unevenly distributed in time and space. This hinders predictive efforts, leading to significant impacts on human life and economies. Multi-hazard prediction is vital for any natural hazard risk management plan. The main objective of this study was the development of a multi-hazard susceptibility mapping framework, by combining two natural hazards—flooding and landslides—in the North Central region of Vietnam. This was accomplished using support vector machines, random forest, and AdaBoost. The input data consisted of 4591 flood points, 1315 landslide points, and 13 conditioning factors, split into training (70%), and testing (30%) datasets. The accuracy of the models' predictions was evaluated using the statistical indices root mean square error, area under curve (AUC), mean absolute error, and coefficient of determination. All proposed models were good at predicting multi-hazard susceptibility, with AUC values over 0.95. Among them, the AUC value for the support vector machine model was 0.98 and 0.99 for landslide and flood, respectively. For the random forest model, these values were 0.98 and 0.98, and for AdaBoost, they were 0.99 and 0.99. The multi-hazard maps were built by combining the landslide and flood susceptibility maps. The results showed that approximately 60% of the study area was affected by landslides, 30% by flood, and 8% by both hazards. These results illustrate how North Central is one of the regions of Vietnam that is most severely affected by natural hazards, particularly flooding, and landslides. The proposed models adapt to evaluate multi-hazard susceptibility at different scales, although expert intervention is also required, to optimize the algorithms. Multi-hazard maps can provide a valuable point of reference for decision makers in sustainable land-use planning and infrastructure development in regions faced with multiple hazards, and to prevent and reduce more effectively the frequency of floods and landslides and their damage to human life and property.     

Landslide inventory validation and susceptibility mapping in the Gerecse Hills,Hungary

Landslides pose a threat to property both in the populated and cultivated areas of the Gerecse Hills (Hungary). The currently available landslide inventory database holds the records from many sites in the area, but the database is out-of-date. Here we address the problem of revising the National Landslides Cadastre landslide inventory database by creating a landslide suscept- ibility map with a multivariate model based on likelihood ratio functions. The model is applied to the TanDEM-X DEM (0.4″ res.), the current landslide inventory of the area, and data acquired from geological maps. By comparing the distributions of four variables in the landslide and non-landslide area with grid computation methods, the model yields landslide susceptibility estimates for the study area. The estimations show to what extent a certain area is similar to the sample areas, therefore, its likelihood to be affected by landslides in the future. The accuracy of the model predictions was checked in the field and compared to the results of our previous study using the SRTM-1 DEM for a similar analysis. The model gave accurate estimates when certain correction measures were applied to the input datasets. The limitations of the model, the input datasets, and the suggested correction measures are also discussed.     

诱发因素影响下的滑坡参数优化预测模型研究

在降雨等外界诱发因素的综合作用下,滑坡位移预测是一个复杂的动力系统问题。利用三峡库区白家包滑坡综合监测数据,分析滑坡演化实时特征,提取影响滑坡变形的最相关因素,研究发现白家包滑坡为降雨主导型堆积层滑坡;采用自回归综合移动模型（ARIMA）模型进行拟合及预测,引入月累积降雨量对模型季节性趋势参数进行评估优化,对白家包滑坡72期月相对位移数据进行拟合及预测研究,最终模型结果和实测值的平均绝对误差和相关系数分别为2.873和0.983。研究结果表明,与传统经验法相比,优化参数模型更符合滑坡变形的一般规律。     

GIS-based landslide susceptibility model considering effective contributing area for drainage time

This study employed GIS modelling to ascertain landslide susceptibility on Mt. Umyeon, south of Seoul, South Korea. In this study, an effective contributing area (ECA) for certain drainage time was purposed as a temporal causative factor and then used for modelling in combination with spatial causative factors such as elevation, slope, plan curvature, drainage proximity, forest type, soil type and geology. Landslide inventory map of 163 landslide locations was prepared using aerial photographic interpretation and field verifications after that digitized using GIS environment in 1:5000 scale. A presence-only-based maximum entropy model was used to establish and analyse the relationship between landslides and causative factors. Before final modelling, a jackknife test was performed to measure the variable contributions, which showed that the slope was the most significant spatial causative factor, and ECA with a drainage time of 12 h was the most significant temporal causative factor. The performances of the final models, with and without significant ECA, were assessed by plotting a receiver operating characteristic curve to be 75.5 and 81.2%, respectively.     

Application of frequency ratio and likelihood ratio model for geo-spatial modelling of landslide hazard vulnerability assessment and zonation: a case study from the Sikkim Himalayas in India

The Likelihood Ratio (LR) Model has been applied as an improvement upon the Frequency Ratio (FR) that computes the ratio of the percentage of the landslide pixels to the percentage of the non-landslide pixels instead of the total number of pixels used in the denominator as in case of the FR. The comparative assessment of the two techniques is made through spatial modelling of GIS vector data using the ArcGIS software. Two different Landslide Information Values were computed for each polygon element of the study area employing the two FR techniques that categorized the study area into five classes of vulnerability using natural breaks (Jenks) technique. Subsequently, vulnerability zonation maps were prepared showing the different levels of landslide vulnerability. The LR technique yielded significantly higher vulnerability assessment accuracy (77%) as compared to the standard FR (71%).     

尺度变化对城市生态环境与人类活动关系的影响研究

针对空间尺度对城市生态环境与人类活动影响的研究，以南昌市为研究区，划分了300×300、500×500、700×700共3种格网尺度。使用遥感生态指数RSEI（目标向量）量化城市生态环境质量，结合POI点、微博签到点与道路网数据（特征向量），利用随机森林回归模型分析不同尺度下两者之间的拟合效果。结果表明：①3种尺度下RSEI与POI点、微博签到点与道路网均呈现较强的负相关性；且负相关性最优的为RSEI和微博签到点数据，最差的为RSEI和道路网数据。②300×300尺度下随机森林回归模型的拟合效果最好。随着尺度的变大，拟合的效果会越来越差。③无论尺度如何变化，利用随机森林回归拟合的标准化残差ε均呈正态分布；且随着尺度的变大，ε值空间分布的随机性也逐渐增大。随机森林回归模型为度量尺度对城市生态环境与人类活动的关系研究提供了有效的途径，也为城市生态文明建设提供了科学的依据和参考。