Landslide susceptibility mapping based on Support Vector Machine: A case study on natural slopes of Hong Kong, China

The Support Vector Machine (SVM) is an increasingly popular learning procedure based on statistical learning theory, and involves a training phase in which the model is trained by a training dataset of associated input and target output values. The trained model is then used to evaluate a separate set of testing data. There are two main ideas underlying the SVM for discriminant-type problems. The first is an optimum linear separating hyperplane that separates the data patterns. The second is the use of kernel functions to convert the original non-linear data patterns into the format that is linearly separable in a high-dimensional feature space. In this paper, an overview of the SVM, both one-class and two-class SVM methods, is first presented followed by its use in landslide susceptibility mapping. A study area was selected from the natural terrain of Hong Kong, and slope angle, slope aspect, elevation, profile curvature of slope, lithology, vegetation cover and topographic wetness index (TWI) were used as environmental parameters which influence the occurrence of landslides. One-class and two-class SVM models were trained and then used to map landslide susceptibility respectively. The resulting susceptibility maps obtained by the methods were compared to that obtained by the logistic regression (LR) method. It is concluded that two-class SVM possesses better prediction efficiency than logistic regression and one-class SVM. However, one-class SVM, which only requires failed cases, has an advantage over the other two methods as only “failed” case information is usually available in landslide susceptibility mapping.     

基于CART决策树的双尺度流域单元地貌分类研究——以北回归线(云南段)地区为例

针对当前单一地貌划分单元造成的分类结果破碎或漏分问题,该文引入双尺度流域单元划分方法,即采用两种不同大小流域单元的组合作为地貌划分基本单元,以提高地貌划分的细分性和完整性。以30 m ASTER GDEM数据为数据源,基于最佳地形因子组合(高程、地势起伏度、坡度、坡度变率、光照模拟值)、双尺度流域单元、CART决策树算法,实现了北回归线(云南段)地区平原(2类)和山地(7类)共9类地貌的划分,双尺度流域单元划分的最佳流量阈值分别为500、2000。通过平均值、标准差、Moran′s I和人工判读结果对分类结果进行检验,发现基于CART决策树的双尺度流域单元地貌分类方法在北回归线(云南段)地区总体精度可达82.1%,Kappa系数为0.793,总体能够准确识别出研究区的地貌类型空间分布特征,是地貌类型划分的一种可行方法。     

Bridging theory and implementation – Testing an abstract classification system for practical mapping by field survey and 3D aerial photographic interpretation

ABSTRACT

The abstract classification system Nature in Norway (NiN) has detailed ecological definitions of a high number of ecosystem units, but its applicability in practical vegetation mapping is unknown because it was not designed with a specific mapping method in mind. To investigate this further, two methods for mapping – 3D aerial photographic interpretation of colour infrared photos and field survey – were used to map comparable neighbouring sites of 1 km² in Hvaler Municipality, south-eastern Norway. The classification accuracy of each method was evaluated using a consensus classification of 160 randomly distributed plots within the study sites. The results showed an overall classification accuracy of 62.5% for 3D aerial photographic interpretation and 82.5% for field survey. However, the accuracy varied for the ecosystem units mapped. The classification accuracy of ecosystem units in acidic, dry and open terrain was similar for both methods, whereas classification accuracy of calcareous units was highest using field survey. The mapping progress using 3D aerial photographic interpretation was more than two times faster than that of field survey. Based on the results, the authors recommend a method combining 3D aerial photographic interpretation and field survey to achieve effectively accurate mapping in practical applications of the NiN system.     

Loess terrain segmentation from digital elevation models based on the region growth method

Regional automatic segmentation – automatic terrain segmentation according to terrain features – is significant for modern geographical analysis. We propose a new approach of terrain region segmentation based on the region growth method. This method features actual runoff nodes as seed points. The corresponding growth threshold is defined based on statistical analysis of quantitative indexes of topographic features. Terrain segmentation of some regions is completed using the growth threshold. The corresponding edge boundaries of different terrain regions are extracted by image processing. Thus, the automatic segmentation of the terrain region is realized by the edge boundary. The application of the method to a typical Chinese loess landform area and automatic segmentation of three types of terrain regions – gully interfluve land, gully slope land, and gully groove land – are achieved by analyzing characteristics of the curvature structure of surface profiles. Segmentation results, compared with results of visual interpretation from a high-precision digital orthophoto map, show an average accuracy of segmentation of 93.51%. Topographic factor features of segmentation results are statistically analyzed. This study presents an effective and practical approach for segmenting terrain regions. This approach may be incorporated into the theory and method system of digital terrain analysis.     

基于球面DQG的矢量与地形数据无缝集成

现有的全球大规模空间数据可视化系统主要侧重于影像和地形数据的综合表达,针对矢量与地形的集成可视化能力相对较弱。该文以球面退化四叉树格网(Degenerate Quad-tree Grids,DQG)为基础,通过DQG格网的三角化过程构建了地表DEM模型,并提出了从矢量线对象到地形格网表面的映射方法。采用GTOPO30数据集和国界矢量数据进行了相关实验,结果表明:该方法能实现矢量数据与多分辨率DEM的无缝集成,并能有效地避免矢量对象"悬浮"和"入地"等现象。     

Glacier mapping based on Chinese high-resolution remote sensing GF-1 satellite and topographic data

The precise glacier boundary is a fundamental requirement for glacier inventory,the assessment of climate change and water management in remote mountain areas.However,some glaciers in mountain areas are covered by debris.The high spatial resolution images bring opportunities in mapping debris-covered glaciers.To discuss the capability of Chinese GaoFen-1 satellite lacking the short wave infrared band and thermal infrared band in mapping glaciers,this study distinguished supraglacial terrain from surrounding debris by combining GaoFen-1(GF-1)wide-field-view(WFV)images,the ratio of the thermal infrared imagery and morphometric parameters(DEM and slope)with 30 m resolution.The overall accuracy of 90.94%indicated that this method was effective for mapping supraglacial terrain in mountain areas.Comparing this result with the combination of GF-1 WFV and low-resolution morphometric parameters shows that a high-quality DEM and the thermal infrared band enhanced the accuracy of glacier mapping especially debris-covered ice in steep terrain.The user's and producer's accuracies of glacier area were also improved from 89.67%and 85.95%to 92.83%and 90.34%,respectively.GF data is recommended for mapping heavily debris-covered glaciers and will be combined with SAR data for future studies.     

Statistical analysis of drainage density from digital terrain data

Drainage density (D_d), defined as the total length of channels per unit area, is a fundamental property of natural terrain that reflects local climate, relief, geology, and other factors. Accurate measurement of D_d is important for numerous geomorphic and hydrologic applications, yet it is a surprisingly difficult quantity to measure, particularly over large areas. Here, we develop a consistent and efficient method for generating maps of D_d using digital terrain data. The method relies on (i) measuring hillslope flow path distance at every unchanneled site within a basin, and (ii) analyzing this field as a random space function. As a consequence, we measure not only its mean (which is half the inverse of the traditional definition of drainage density) but also its variance, higher moments, and spatial correlation structure. This yields a theoretically sound tool for estimating spatial variability of drainage density. Averaging length-to-channel over an appropriate spatial scale also makes it possible to derive continuous maps of D_d and its spatial variations. We show that the autocorrelation length scale provides a natural and objective choice for spatial averaging. This mapping technique is applied to a region of highly variable D_d in the northern Apennines, Italy. We show that the method is capable of revealing large-scale patterns of variation in D_d that are correlated with lithology and relief. The method provides a new and more general way to quantitatively define and measure D_d, to test geomorphic models, and to incorporate D_d variations into regional-scale hydrologic models.     

多分析尺度下综合判别的地形元素分类方法

地形元素（如山脊、沟谷等）是地表形态类型基本单元,通过地形元素的不同空间组合可形成更高级别的地貌类型。现有的地形元素提取方法大多依靠地形属性计算,难以克服地形元素的空间相关性表达与局部地形属性计算存在不对应的矛盾,Jasiewicz和Stepinski提出的Geomorphons方法——基于高程相对差异信息进行地形元素分类,可避免这一问题,但Geomorphons方法本质上是在单一分析尺度上选择地形特征点用于判别,易受局部地形起伏的影响而造成误分类。针对这一问题,设计出一种多分析尺度下综合判别的地形元素分类方法。应用结果表明：相比Geomorphons方法,利用该方法得到的地形元素的分类结果更为合理。     

基于势能信息熵的黄土小流域地貌演化特征

黄土高原流域地貌系统的地貌演化特征十分复杂,尚有诸多科学问题有待进一步深入研究。以往研究大多集中在流域地貌演化的侵蚀和发育特征等某一方面,缺乏从流域地貌系统及其势能信息熵的视角深入剖析野外多岩土层黄土小流域地貌演化特征的研究。为此,基于系统论的观点和方法,构建多岩土层黄土小流域地貌系统及其势能信息熵的数学模型,并以辛店沟小流域为例,对其地貌演化特征进行研究。结果表明：（1）构建的野外多岩土层黄土小流域地貌系统的概念模型及其势能信息熵的数学模型能够有效对辛店沟小流域进行数值模拟。（2）以黄土侵蚀作用为主的辛店沟小流域从2000—2019年的地貌演化过程是其势能信息熵的熵减过程和黄土地貌不断侵蚀的过程。（3）辛店沟小流域的势能信息熵能较好地反映该小流域的地貌演化阶段和地貌侵蚀过程。     

Mapping Human Terrain in the Joint Army–Navy Intelligence Study of Korea (1945)

The concept of human terrain has become a prominent element of U.S. military strategy. It is a means to capture the cultural–geographical qualities of an enemy or target population. An early effort to map human terrain is found in the Joint Army–Navy Intelligence Study (JANIS) of Korea (1945). We argue that the JANIS report on Korea was paradigmatic for the U.S. military's contemporary geographical work and offers insights into the cultural politics of human terrain mapping. This explains why the JANIS text is cited by the National Geospatial-Intelligence College (NGC) today as an historical model. This article not only offers a window into the history of geography counterinsurgent but also shows that geography has been entwined with empire.     

Landslide Soils and Geomorphology in Bridger-Teton National Forest,Northwest Wyoming

Active landslides are evident throughout Bridger-Teton National Forest (BTNF), and northwestern Wyoming has one of the highest landslide densities in the country. Land use changes and increased demands for infrastructure challenge BTNF personnel to better understand landslide processes in order to make informed land management decisions. Despite recent population growth in the region, research on landslide phenomena is lacking. In this study, soil and geomorphic properties related to landslide occurrence were studied at 18 landslides in the BTNF. Landslides were categorized as active or inactive based on geomorphic features. Landslide soil characteristics including texture, shrink-swell potential, clay mineralogy, and horizonation were compared on active and inactive landslides. The results indicate that soil characteristics related to the degree of soil formation are different on active and inactive landslides. Soil characteristics such as plasticity, shrink-swell potential, and clay mineralogy influence slope stability and were distinctly different on active and inactive landslides, especially in C horizons. This study shows that soil characteristics and slope geomorphic properties may be useful for assessing landslide frequency. Our results support a hypothesis that landslide occurrence in the BTNF is related to weathering of soil and unconsolidated material, which affects clay mineralogy.     

The Relationship between Geomorphic River Structure and Coarse Particulate Organic Matter (CPOM) Storage along the Kangaroo River,New South Wales,Australia

Fluvial landforms provide a physical template upon which to appraise biophysical relationships along river courses. In this study, the spatial pattern of organic matter storage along the Kangaroo River, NSW, is related to geomorphic controls that operate at a range of scales within a nested hierarchy. This snapshot study of CPOM storage found that at the catchment scale the longitudinal pattern of coarse particulate organic matter (CPOM) storage is dependent on the type and downstream pattern of River Styles. At the reach scale, CPOM storage is dependent on the geomorphic unit structure and physical heterogeneity of the river, and associated energy conditions along the reach. At the geomorphic unit scale, CPOM storage capacity is related to the position of geomorphic units relative to the thalweg (i.e. flow characteristics) and associated roughness attributes. At the hydraulic unit scale, CPOM storage capacity is related to local flow velocity and substrate characteristics (clast size and distribution).     

南海珊瑚礁高分辨率遥感地貌分类体系研究

珊瑚礁遥感地貌分类体系在珊瑚礁遥感地貌制图中具有重要的指导作用。目前,珊瑚礁遥感地貌分类体系仍存在构建标准不统一、部分重要地貌类型不突出以及涵盖地貌类型不完备等问题,影响了珊瑚礁遥感地貌制图应用于珊瑚礁科学和管理。本文采用中国南海46个珊瑚礁(环礁、台礁)的高分辨率遥感影像(WorldView-2、Quickbird),并结合西沙群岛15个岛礁的地貌实地调查数据进行南海高分辨率遥感地貌分类体系的构建研究。以各地貌类型所处礁体位置、动力特征、出露程度和沉积类型为划分标准,将相似尺度和重要性相当的地貌类型归为同等级别,共构建了3级19类南海珊瑚礁遥感地貌分类体系。该分类体系划分标准统一、地貌类型数量最多且完备,新命名了内礁坪生物稀疏带、内礁坪生物丛生带,补充了水下礁脊、潮间带浅滩、浅水礁塘等地貌类型,便于直观理解和推断不同尺度的海洋生态相互作用及重要性。同时,其多等级多尺度性适用于分辨率由低到高的南海珊瑚礁遥感地貌制图,有助于珊瑚礁态势演变分析、管理和保护,为维护国家海洋权益作出应有的贡献。     

Automated landslide mapping using spectral analysis and high-resolution topographic data: Puget Sound lowlands, Washington, and Portland Hills, Oregon

Landslide inventory maps are necessary for assessing landslide hazards and addressing the role slope stability plays in landscape evolution over geologic timescales. However, landslide inventory maps produced with traditional methods — aerial photograph interpretation, topographic map analysis, and field inspection — are often subjective and incomplete. The increasing availability of high-resolution topographic data acquired via airborne Light Detection and Ranging (LiDAR) over broad swaths of terrain invites new, automated landslide mapping procedures. We present two methods of spectral analysis that utilize LiDAR-derived digital elevation models of the Puget Sound lowlands, Washington, and the Tualatin Mountains, Oregon, to quantify and automatically map the topographic signatures of deep-seated landslides. Power spectra produced using the two-dimensional discrete Fourier transform and the two-dimensional continuous wavelet transform identify the characteristic spatial frequencies of deep-seated landslide morphologic features such as hummocky topography, scarps, and displaced blocks of material. Spatial patterns in the amount of spectral power concentrated in these characteristic frequency bands highlight past slope instabilities and allow the delineation of landslide terrain. When calibrated by comparison with detailed, independently compiled landslide inventory maps, our algorithms correctly classify an average of 82% of the terrain in our five study areas. Spectral analysis also allows the creation of dominant wavelength maps, which prove useful in analyzing meter-scale topographic expressions of landslide mechanics, past landslide activity, and landslide-modifying geomorphic processes. These results suggest that our automated landslide mapping methods can create accurate landslide maps and serve as effective, objective, and efficient tools for digital terrain analysis.     

DEM 点位地形信息量化模型研究

针对DEM 点位, 首先应用微分几何法对其所负载的语法信息量进行测度, 其次根据地形特征点类型及地形结构特征确定其语义信息量, 然后基于信息学理论构建了DEM 点位地形信息综合量化模型。在此基础上, 以黄土丘陵沟壑区作为实验样区, 对DEM 点位地形信息量提取方法及其在地形简化中的初步实例应用进行了探讨和验证。实验结果显示, 所提出的DEM 点位地形信息量化方案可行;基于DEM 地形信息量指数的多尺度DEM 构建方案, 具有机理明确、易于实现的特点, 并通过优先保留地形骨架特征点, 可以有效减少地形失真, 从而满足不同层次的多尺度数字地形建模和表达要求。对DEM 点位地形信息进行有效量化, 为认识DEM 地形信息特征提供了一个新的切入点, 同时为多尺度数字地形建模提供理论依据与方法支持。     

基于模糊聚类的新疆典型高寒草原土壤pH值空间制图

准确、高效地掌握草原土壤属性的空间分布能够为草地资源境管理提供基础信息和参考依据。相比于传统土壤调查方法,基于模糊逻辑的土壤—环境推理能够提高野外采样效率和预测制图精度,被广泛应用于数字土壤制图。但由于土壤自身的空间变异性及其与环境条件间的非线性,现有推理模型的稳定性较低,尚未在高寒草原区进行应用。选择新疆巴音布鲁克典型亚高山草原地区约4 km^2区域为研究区,以高程、坡度、坡向、沿剖面曲率、沿等高线曲率、地形湿度指数6个地形因子为土壤环境因子,采用模糊C均值聚类(Fuzzy C-means Clustering,FCM)方法对环境因子聚类,得到9个环境因子组合,并在隶属度值高的环境因子组合中心共设置18个典型点。运用土壤—环境推理方法模拟研究区表层土壤pH值空间分布,其变化范围在7.170~8.186之间。选取35个独立样本进行精度检验(均匀采样点16个,横截面采样点9个,垂直带采样点10个),模拟结果与实测值基本吻合,且基于模糊聚类和土壤—环境推理方法的模拟精度高于普通克里格法和反距离权重法。通过基于模糊逻辑和土壤—环境推理的数字土壤制图方法在小尺度区域的运用验证,结果表明基于典型点的采样方案能够快速、有效地对区域土壤属性进行空间模拟,该方法对于类似小尺度的研究区同样有效。     

地形信息对确定DEM适宜分辨率的影响

分辨率会直接影响基于栅格数字高程模型（DEM）的数字地形分析结果,因此在实际应用中,需要选择适宜的DEM分辨率。目前采取的基本方法,基于某种地形信息定量刻画尺度效应曲线,从而确定DEM适宜分辨率,但对于采用不同地形信息时所产生的影响尚缺乏研究。本文针对该方法中通常采用的坡度、剖面曲率、水平曲率等3 种地形信息,每种地形信息提取时,分别使用两种不同的常用算法,在3 个不同地形特征的研究区中,逐一计算其在不同分辨率下的局部方差均值,以刻画尺度效应曲线,确定相应的DEM适宜分辨率,并进行对比分析。结果表明：① 采用剖面曲率或水平曲率所得适宜分辨率结果基本相同,但采用坡度所得出的适宜分辨率结果则有明显差别,后者所得的适宜分辨率更粗;② 采用不同地形信息时,越是在平缓地形为主的研究区,所得的适宜分辨率结果越相近,在复合地形特征的研究区所得到的适宜分辨率区间均明显较宽;③ 地形属性计算时所用的算法对适宜分辨率结果的影响不明显。     

Uncertainty modeling and analysis of surface area calculation based on a regular grid digital elevation model (DEM)

In the field of digital terrain analysis (DTA), the principle and method of uncertainty in surface area calculation (SAC) have not been deeply developed and need to be further studied. This paper considers the uncertainty of data sources from the digital elevation model (DEM) and SAC in DTA to perform the following investigations: (a) truncation error (TE) modeling and analysis, (b) modeling and analysis of SAC propagation error (PE) by using Monte-Carlo simulation techniques and spatial autocorrelation error to simulate DEM uncertainty. The simulation experiments show that (a) without the introduction of the DEM error, higher DEM resolution and lower terrain complexity lead to smaller TE and absolute error (AE); (b) with the introduction of the DEM error, the DEM resolution and terrain complexity influence the AE and standard deviation (SD) of the SAC, but the trends by which the two values change may be not consistent; and (c) the spatial distribution of the introduced random error determines the size and degree of the deviation between the calculated result and the true value of the surface area. This study provides insights regarding the principle and method of uncertainty in SACs in geographic information science (GIScience) and provides guidance to quantify SAC uncertainty.     

新疆地形复杂度的空间格局及地理特征

地形复杂度（TCI）是区域地表外在形态结构的数字化表达,既可表征坡面单元的多样性及其组合形式的复杂性,也可映射地球内外营力在地表留下“烙印”的过程,其客观描述和定量表达可为地表形态定义、刻画及分异等地形地貌理论研究提供重要依据。新疆独特的山盆地貌结构为地貌学研究提供了理想的场所。本文基于“倒金字塔滤网系统”逐步筛选微观和宏观坡面因子并确定权重进而构建地形复杂度模型;运用均值变点法确定新疆地形复杂度的最佳窗口,分析不同地貌单元地形复杂度的空间异质性,进一步探究不同营力作用对地表复杂程度的贡献。结果显示：① 地形复杂度模型综合相关性分析、聚类分析、变异系数法与主成分分析等方法对地形因子进行客观筛选与科学组合,使其具有全面客观性及独立有效性。② 地形复杂度以ASTER GDEM数据（30 m）为基础,得到最佳窗口（14×14）下全疆TCI介于0.13~46.36。山地与盆地相同地貌类型TCI面积占比峰值相近,不同地貌类型可比较其斜率、偏度及峰度得以区分。在经纬向上能更好的刻画出独立山峰及深切峡谷等局部地形分异,TCI≈1可作为平原与山地地貌单元之间的分界值。在高程分区下TCI>2时为各山系（山群）震荡高程区起始处,山麓处TCI曲线差异显著。③ 地形复杂度能较好的体现不同外营力在地表留下的痕迹,一定程度上也可表征外营力对不同成因地貌类型的贡献度。本研究对新疆地形地貌的形态特征和形成原因等方面提供理论依据和科学方法,可为地形地貌研究、生态环境影响以及区域发展评价提供实践指导。