A hybrid machine learning ensemble approach based on a Radial Basis Function neural network and Rotation Forest for landslide susceptibility modeling: A case study in the Himalayan area,India

In this paper, a hybrid machine learning ensemble approach namely the Rotation Forest based Radial Basis Function (RFRBF) neural network is proposed for spatial prediction of landslides in part of the Himalayan area (India). The proposed approach is an integration of the Radial Basis Function (RBF) neural network classifier and Rotation Forest ensemble, which are state-of-the art machine learning algorithms for classification problems. For this purpose, a spatial database of the study area was established that consists of 930 landslide locations and fifteen influencing parameters (slope angle, road density, curvature, land use, distance to road, plan curvature, lineament density, distance to lineaments, rainfall, distance to river, profile curvature, elevation, slope aspect, river density, and soil type). Using the database, training and validation datasets were generated for constructing and validating the model. Performance of the model was assessed using the Receiver Operating Characteristic (ROC) curve, area under the ROC curve (AUC), statistical analysis methods, and the Chi square test. In addition, Logistic Regression (LR), Multi-layer Perceptron Neural Networks (MLP Neural Nets), Naïve Bayes (NB), and the hybrid model of Rotation Forest and Decision Trees (RFDT) were selected for comparison. The results show that the proposed RFRBF model has the highest prediction capability in comparison to the other models (LR, MLP Neural Nets, NB, and RFDT); therefore, the proposed RFRBF model is promising and should be used as an alternative technique for landslide susceptibility modeling.     

How does forest structure affect root reinforcement and susceptibility to shallow landslides?

Forests can decrease the risk of shallow landslides by mechanically reinforcing the soil and positively influencing its water balance. However, little is known about the effect of different forest structures on slope stability. In the study area in St Antönien, Switzerland, we applied statistical prediction models and a physically‐based model for spatial distribution of root reinforcement in order to quantify the influence of forest structure on slope stability. We designed a generalized linear regression model and a random forest model including variables describing forest structure along with terrain parameters for a set of landslide and control points facing similar slope angle and tree coverage. The root distribution measured at regular distances from seven trees in the same study area was used to calibrate a root distribution model. The root reinforcement was calculated as a function of tree dimension and distance from tree with the root bundle model (RBMw). Based on the modelled values of root reinforcement, we introduced a proxy‐variable for root reinforcement of the nearest tree using a gamma distribution. The results of the statistical analysis show that variables related to forest structure significantly influence landslide susceptibility along with terrain parameters. Significant effects were found for gap length, the distance to the nearest trees and the proxy‐variable for root reinforcement of the nearest tree. Gaps longer than 20 m critically increased the susceptibility to landslides. Root reinforcement decreased with increasing distance from trees and is smaller in landslide plots compared to control plots. Furthermore, the influence of forest structure strongly depends on geomorphological and hydrological conditions. Our results enhance the quantitative knowledge about the influence of forest structure on root reinforcement and landslide susceptibility and support existing management recommendations for protection against gravitational natural hazards. Copyright © 2015 John Wiley & Sons, Ltd.     

基于证据权方法的玉树地震滑坡危险性评价

玉树地震诱发了2 036处滑坡。应用地理信息系统与遥感技术,选取与地表破裂距离、峰值加速度(PGA)、高程、坡度、坡向、曲率、坡位、与水系距离、岩性、与断裂距离、与公路距离、归一化植被指数(NDVI)等12个因素作为玉树地震滑坡危险性评价因子,采用加法与减法2种证据权方法,开展玉树地震滑坡危险性评价研究工作。结果表明:基于加法证据权方法得到评价结果的正确率为80.32%,基于减法证据权方法得到结果的正确率为80.19%。将滑坡危险性评价结果图分为极高危险区、高危险区、中危险区、低危险区与极低危险区5类。这一成果可划分出滑坡危险区,为灾后滑坡防治、基础设施重建与自然环境保护提供参考。     

SIMULATION STUDY OF ROAD-CUT EFFECTS ON SLOPE STABILITY

Landslides and rock falls along the highway are common geological hazards in Southwest China. As an influencing factor on potential landslides behavior, roads or distance to roads have been successfully used in landslide susceptibility assessments in mountainous area. However, the relationship between the road-cut and the slope stability is not clear. Therefore, we performed two-dimensional slope stability calculation using the general limit equilibrium (GLE)method incorporated in the software SLOPE/W of GeoStudio for stability analysis of slopes. Our studies show that the man-made roads influence on the slope stability mainly exists in two ways:One is to create a new steep slope, which will result in rock falls and shallow landslides along the roads; the other is to influence the stability of the original slope, which will result in comparatively huge landslides. For the latter, our simulation study reveals that the road location, namely at which part of a natural slope to construct a road is important for the slope stability. For a natural slope with a potential slip surface, if a road is constructed at or near the slope toe where the potential slip surface surpasses, it will greatly degrade the slope's factor of safety (Fs) and make the slope unstable; however, if a rode-cut is near the top of the slope, it will increase the slope's Fs and make the slope more stable. The safety location is different for different slope angle, steeper slope needs a higher location for a safety road-cut in comparison with gentle slopes. Moreover, the slope stability decreases when loading a seismic force and it varies with the slope angle. Firstly, the Fs decreases when the slope angle increasing, and when the slope angle reaches 45°, the Fs then becomes greater with the slope angle increasing.     

Optimal design of structures for earthquake loads by a hybrid RBF-BPSO method

The optimal seismic design of structures requires that time history analyses （THA） be carried out repeatedly. This makes the optimal design process inefficient, in particular, if an evolutionary algorithm is used. To reduce the overall time required for structural optimization, two artificial intelligence strategies are employed. In the first strategy, radial basis function （RBF） neural networks are used to predict the time history responses of structures in the optimization flow. In the second strategy, a binary particle swarm optimization （BPSO） is used to find the optimum design. Combining the RBF and BPSO, a hybrid RBF-BPSO optimization method is proposed in this paper, which achieves fast optimization with high computational performance. Two examples are presented and compared to determine the optimal weight of structures under earthquake loadings using both exact and approximate analyses. The numerical results demonstrate the computational advantages and effectiveness of the proposed hybrid RBF-BPSO optimization method for the seismic design of structures.     

BFA-CM最优化测井解释方法

最优化测井解释方法能充分利用各种测井资料及地质信息,可以有效地评价复杂岩性油气藏.优化算法的选择是最优化测井解释方法的关键,影响着测井解释结果的准确性.细菌觅食算法(BFA)是新兴的一种智能优化算法,具有较强的全局搜索能力,但在寻优后期收敛速度较慢.复合形算法(CM)局部搜索能力极强,将其与BFA算法相结合构成BFA-CM混合算法,既提高了搜索精度又提高了搜索效率.利用BFA-CM最优化测井解释方法对苏里格致密砂岩储层实际资料进行了处理,计算结果与岩心及薄片分析资料吻合得很好.     

GIS‐based regional landslide susceptibility mapping: a case study in southern California

Landslides threaten lives and property throughout the United States, causing in excess of $2 billion in damages and 25–50 deaths annually. In regions subjected to urban expansion caused by population growth and/or increased storm intensities caused by changing climate patterns, the economic and society costs of landslides will continue to rise. Using a geographic information system (GIS), this paper develops and implements a multivariate statistical approach for mapping landslide susceptibility. The presented susceptibility maps are intended to help in the design of hazard mitigation and land development policies at regional scales. The paper presents (a) a GIS‐based multivariate statistical approach for mapping landslide susceptibility, (b) several dimensionless landslide susceptibility indexes developed to quantify and weight the influence of individual categories for given potential risk factors on landslides and (c) a case study in southern California, which uses 11 111 seismic landslide scars collected from previous efforts and 5389 landslide scars newly digitized from local geologic maps. In the case study, seven potential risk factors were selected to map landslide susceptibility. Ground slope and event precipitation were the most important factors, followed by land cover, surface curvature, proximity to fault, elevation and proximity to coastline. The developed landslide susceptibility maps show that areas classified as having high or very high susceptibilities contained 71% of the digitized landslide scars and 90% of the seismic landslide scars while only occupying 26% of the total study area. These areas mostly have ground slopes higher than 46% and 2‐year, 6‐hour precipitation greater than 51 mm. Only 12% of digitized landslides and less than 1% of recorded seismic landslides were located in areas classified as low or very low susceptibility, while occupying 42% of the total study region. These areas mostly have slopes less than 27% and 2‐year, 6‐hour precipitation less than 41 mm. Copyright © 2007 John Wiley & Sons, Ltd.     

Calibration of subsurface batch and reactive-transport models involving complex biogeochemical processes

In this study, the calibration of subsurface batch and reactive-transport models involving complex biogeochemical processes was systematically evaluated. Two hypothetical nitrate biodegradation scenarios were developed and simulated in numerical experiments to evaluate the performance of three calibration search procedures: a multi-start non-linear regression algorithm (i.e. multi-start Levenberg–Marquardt), a global search heuristic (i.e. particle swarm optimization), and a hybrid algorithm that combines the particle swarm procedure with a regression-based “polishing” step. Graphical analysis of the selected calibration problems revealed heterogeneous regions of extreme parameter sensitivity and insensitivity along with abundant numbers of local minima. These characteristics hindered the performance of the multi-start non-linear regression technique, which was generally the least effective of the considered algorithms. In most cases, the global search and hybrid methods were capable of producing improved model fits at comparable computational expense. In other cases, the multi-start and hybrid calibration algorithms yielded comparable fitness values but markedly differing parameter estimates and associated uncertainty measures.     

Landslide intelligent prediction using object-oriented method

Landslides are one of the most serious geological disasters in the world and happen quite frequently in the Three Gorges. Landslide prediction is a very important measure of landslide prevention and cure in the Three Gorges. Traditional methods lack in sufficiently mining the various complex information from a landslide system. They often need much manual intervention and possess poor intelligence and accuracy. An intelligent method proposed in this paper for landslide prediction based on an object-oriented method and knowledge driving is hopeful to solve the above problem. The method adopted Landsat ETM+ images, 1:50,000 geological map and 1:10,000 relief map in the Three Gorges as the data origins. It firstly produced the key factors influencing landslide development and used multi-resolution segmentation algorithm to segment the image objects based on the key landslide factors of engineering rock group, reservoir water fluctuation, slope structure and slope level. Secondly, the method chose some sample objects and adopted the decision tree algorithm C5.0 to mine the landslide forecast criteria according to the factor values of each sample object. Finally, under knowledge driving the method classified the image objects and realized landslide susceptibility analysis and intelligent prediction in the Three Gorges. The method proposed in this paper is object-oriented. Results of a real-world example show that: (1) the object-oriented method possesses much more compact knowledge representation, higher efficiency, more continuous classifying result and higher prediction accuracy compared with the pixel-oriented method; (2) it possesses the overall accuracy of 87.64% and kappa coefficient of 0.8305 and is more accurate than the other seven methods (such as the pixel-oriented methods of Parallelpiped, Minimum Distance, Maximum Likelihood, Mahalanobis Distance, K-means and Isodata and the object-oriented method of Nearest Neighbor); (3) about 46.97% landslides lie in the high susceptibility region, 24.24% landslides lie in the moderate susceptibility region, 27.27% landslides lie in the low susceptibility region and 1.52% landslides lie in the very low susceptibility region. Therefore the method can effectively realize landslide susceptibility analysis and provides a new idea for landslide intelligent and accurate prediction.     

Point process-based modeling of multiple debris flow landslides using INLA: an application to the 2009 Messina disaster

We develop a stochastic modeling approach based on spatial point processes of log-Gaussian Cox type for a collection of around 5000 landslide events provoked by a precipitation trigger in Sicily, Italy. Through the embedding into a hierarchical Bayesian estimation framework, we can use the integrated nested Laplace approximation methodology to make inference and obtain the posterior estimates of spatially distributed covariate and random effects. Several mapping units are useful to partition a given study area in landslide prediction studies. These units hierarchically subdivide the geographic space from the highest grid-based resolution to the stronger morphodynamic-oriented slope units. Here we integrate both mapping units into a single hierarchical model, by treating the landslide triggering locations as a random point pattern. This approach diverges fundamentally from the unanimously used presence–absence structure for areal units since we focus on modeling the expected landslide count jointly within the two mapping units. Predicting this landslide intensity provides more detailed and complete information as compared to the classically used susceptibility mapping approach based on relative probabilities. To illustrate the model’s versatility, we compute absolute probability maps of landslide occurrences and check their predictive power over space. While the landslide community typically produces spatial predictive models for landslides only in the sense that covariates are spatially distributed, no actual spatial dependence has been explicitly integrated so far. Our novel approach features a spatial latent effect defined at the slope unit level, allowing us to assess the spatial influence that remains unexplained by the covariates in the model. For rainfall-induced landslides in regions where the raingauge network is not sufficient to capture the spatial distribution of the triggering precipitation event, this latent effect provides valuable imaging support on the unobserved rainfall pattern.     

基于高分卫星遥感影像的地震应急滑坡编目与分布特征探讨——以2017年8月8日九寨沟7.0级地震为例

地震应急是减轻地震灾害的重要途径之一。地震应急工作具有时间紧迫、事关重大的特点。2017年8月8日四川九寨沟M_S7.0级地震发生后,为快速、准确地提供地震引发的滑坡灾害分布,本研究基于震后第一天获取到的高分辨率遥感影像（高分二号卫星影像、北京二号卫星影像）,通过人工目视解译的方法初步建立了四川九寨沟地震滑坡编目。结果表明,该地震至少触发了622处同震滑坡,分布在沿使用影像边界框定的面积为3919km2的区域内。本研究还利用这个地震滑坡编目,统计了九寨沟地震滑坡数量和滑坡点密度（LND）与地形（坡度、坡向）、地震（地震烈度、震中距）等因素的关系。结果表明九寨沟地震滑坡多发生在坡度为20°—50°的区域内,滑坡的易发性随着坡度的增加而增加。受地震波传播方向的影响,E、SE向是地震滑坡较易发生的坡向。滑坡的易发程度和地震烈度呈正相关,即随着烈度的增大,滑坡易发性增大。滑坡易发性还随着震中距增加而降低,这是由于地震波能量随震中距的增加而衰减导致的。     

用随机启发式优化算法反演2003年云南大姚地震断层面参数的比较研究

基于余震分布确定主震断层面的数学模型,以确定断层面的走向和倾角参数进行计算,研究了遗传算法、模拟退火算法、差分演化算法、粒子群算法等4种最优化反演方法的反演效果和可靠性。结果显示,在涉及到的反演参数较少和非线性不太严重时,4种方法都有较好的表现,差分演化算法、粒子群算法速度快,精度高,遗传算法速度较慢,精度较低,模拟退火由于缺乏并行机制,速度较慢,精度高于遗传算法。余震在求出的断层附近分布图直观地反映出4种方法的效果和可靠性。     

Groundwater-level forecasting under climate change scenarios using an artificial neural network trained with particle swarm optimization

Abstract

Artificial neural networks (ANNs) have recently been used to predict the hydraulic head in well locations. In the present work, the particle swarm optimization (PSO) algorithm was used to train a feed-forward multi-layer ANN for the simulation of hydraulic head change at an observation well in the region of Agia, Chania, Greece. Three variants of the PSO algorithm were considered, the classic one with inertia weight improvement, PSO with time varying acceleration coefficients (PSO-TVAC) and global best PSO (GLBest-PSO). The best performance was achieved by GLBest-PSO when implemented using field data from the region of interest, providing improved training results compared to the back-propagation training algorithm. The trained ANN was subsequently used for mid-term prediction of the hydraulic head, as well as for the study of three climate change scenarios. Data time series were created using a stochastic weather generator, and the scenarios were examined for the period 2010–2020.

Editor Z.W. Kundzewicz; Associate editor L. See

Citation Tapoglou, E., Trichakis, I.C., Dokou, Z., Nikolos, I.K., and Karatzas, G.P., 2014. Groundwater-level forecasting under climate change scenarios using an artificial neural network trained with particle swarm optimization. Hydrological Sciences Journal, 59(6), 1225–1239. http://dx.doi.org/10.1080/02626667.2013.838005     

Comparing the stream re-aeration coefficient estimated from ANN and empirical models / Comparaison d'estimations par un RNA et par des modèles empiriques du coefficient de réaération en cours d'eau

Abstract

Dissolved oxygen (DO) is one of the most useful indices of river's health and the stream re-aeration coefficient is an important input to computations related to DO. Normally, this coefficient is expressed as a function of several variables, such as mean stream velocity, shear stress velocity, bed slope, flow depth, and Froude number. However, in free surface flows, some of these variables are interrelated, and it is possible to obtain simplified stream re-aeration equations. In recent years, different functional forms have been advanced to represent the re-aeration coefficient for different data sets. In the present study, the artificial neural network (ANN) technique has been applied to estimate the re-aeration coefficient (K ₂) using data sets measured at different reaches of the Kali River in India and values obtained from the literature. Observed stream/channel velocity, bed slope, flow depth, cross-sectional area and re-aeration coefficient data were used for the analysis. Different combinations of variables were tested to obtain the re-aeration coefficient using an ANN. The performance of the ANN was compared with other estimation methods. It was found that the re-aeration coefficient estimated by using an ANN was much closer to the observed values as compared with the other techniques.     

Evaluation of global optimization algorithms for parameter calibration of a computationally intensive hydrologic model

With the popularity of complex hydrologic models, the time taken to run these models is increasing substantially. Comparing and evaluating the efficacy of different optimization algorithms for calibrating computationally intensive hydrologic models is becoming a nontrivial issue. In this study, five global optimization algorithms (genetic algorithms, shuffled complex evolution, particle swarm optimization, differential evolution, and artificial immune system) were tested for automatic parameter calibration of a complex hydrologic model, Soil and Water Assessment Tool (SWAT), in four watersheds. The results show that genetic algorithms (GA) outperform the other four algorithms given model evaluation numbers larger than 2000, while particle swarm optimization (PSO) can obtain better parameter solutions than other algorithms given fewer number of model runs (less than 2000). Given limited computational time, the PSO algorithm is preferred, while GA should be chosen given plenty of computational resources. When applying GA and PSO for parameter optimization of SWAT, small population size should be chosen. Copyright © 2008 John Wiley & Sons, Ltd.     

Including root reinforcement variability in a probabilistic 3D stability model

Forests play a significant role in protecting people, settlements in mountainous terrains from hydrogeomorphic hazards, including shallow landslides. Although several studies have investigated the interactions between forests and slope instabilities, a full understanding of them has not yet been obtained. Additionally, models that incorporate forest stand properties into slope failure probability analyses have not been developed. In principle, physical‐based models, which are powerful tools for landslide hazard analyses, represent an appropriate approach to linking stand properties and slope stability. However, the reliability of these models depends on numerous parameters that describe highly complex geotechnical and hydrological processes (e.g. potential failure depth, saturation ratio, root reinforcement, etc.) that are difficult to measure and model. In particular, the spatial heterogeneity of root reinforcement remains a problem, and the use of physically based models from a forest management perspective has been limited. This paper presents a procedure for assessing slope stability in terms of the Factor of Safety that accounts for forest stand characteristics such as tree density, average diameter at breast height and minimum distance between trees. The procedure combines a three‐dimensional (3D) slope stability model with an evaluation of the variability of root reinforcement in terms of a probability distribution, according to forest characteristics. Monte Carlo simulation is used to account for the residual uncertainties in both stand characteristics and 3D stability model parameters. The proposed method was applied in a subalpine catchment in the Italian Alps, mainly covered by coniferous forest and characterized by steep slopes and high landslide risk. The results suggest that the procedure is highly reliable, according to landslide inventory maps [area under the ROC curve (AUC) is 0.82 and modified success rate (MSR) is 0.70]. Thus, it represents a promising tool for studying the role of root reinforcement in landslide hazard mapping and guiding forest management from a slope stability perspective. Copyright © 2017 John Wiley & Sons, Ltd.     

Emergency assessment of seismic landslide susceptibility： a case study of the 2008 Wenchuan earthquake affected area

The 8.0 Mw Wenchuan earthquake triggered widespread and large scale landslides in mountainous regions.An approach was used to map and assess landslide susceptibility in a given area. A numerical rating system was applied to five factors that contribute to slope instability. Factors such as lithology, topography, streams and faults have an important influence as event-controlling factors for landslide susceptibility assessment. A final map is provided to show areas of low,medium, and high landslide susceptibility. Areas identified as having high landslide susceptibility were located in the central,northeastern, and far south regions of the study area. The assessment results will help decision makers to select safe sites for emergency placement of refuges and plan for future reconstruction. The maps may also be used as a basis for landslide risk management in the study area.     

Automated mapping of land components from digital elevation data

An algorithm for automating the mapping of land components from digital elevation data is described. Land components are areas of relatively uniform slope and aspect and often correspond with ridge crests, shoulders, head slopes, back slopes or foot slopes. Aspect regions, which generally span from stream to ridge, are first identified by generalizing an aspect map derived from digital elevation data. The aspect regions are then split successively into land components by grouping pixels above or below an automatically determined contour of elevation or ‘distance from stream’. The contour approximates a slope break. The land components mapped in this way give a complete polygonization of a hilly landscape and are a reasonable approximation of manually mapped land components.     

基于GIS和专家知识的滇西南地区滑坡敏感性模糊逻辑推理方法

为了充分识别和有效减轻滑坡灾害风险,对滇西南南涧(约470 km²)和凤庆—昌宁(约2300 km²)两个研究区开展了基于GIS和专家知识的滑坡敏感性模糊逻辑评价研究。通过检查模型计算得到的历史滑坡点敏感性值与整个研究区域的滑坡敏感性平均值是否不同来评价本方法的性能,用Z值检查来测试差异的统计显著性。计算结果显示,南涧地区的Z值为4.1,相应的P值小于0.001,表明通过模型计算得到的滑坡敏感性值是该区域滑坡事件发生的良好指标;凤庆—昌宁地区的Z值为8.93,相应的P值小于0.001。在此基础上,采用自然断点法对滑坡敏感性值进行分类,根据分类结果将滑坡敏感性水平划分成5个等级:极低(0.0~0.001)、较低(0.001~0.051)、中等(0.051~0.394)、较高(0.394~0.557)和极高(0.557~1.0)。敏感性极低和较低的地区没有发现历史滑坡记录;敏感性极高地区的历史滑坡密度约是敏感性较高地区的4倍,约为敏感性中等地区的10倍。凤庆—昌宁地区的研究结果表明,从区域专家群中提取的滑坡敏感性与环境因子关系的知识可以外延到滇西南其它地区。     

STUDY ON THE DISTRIBUTION PATTERN OF EARTHQUAKE- TRIGGERED LANDSLIDES BASED ON SEISMIC LANDSLIDE SUSCEPTIBILITY ANALYSIS: A CASE STUDY OF LANDSLIDES TRIGGERED BY THE MS6.5 LUDIAN EARTHQUAKE IN 2014

On August 3, 2014, an M_W6.5 earthquake occurred in Ludian County, Yunnan Province, which triggered significant landslides and caused serious ground damages and casualties. Compared with the existing events of earthquake-triggered landslides, the spatial distribution of co-seismic landslides during the Ludian earthquake showed a special pattern. The relationship between the co-seismic landslides and the epicenter or the known faults is not obvious, and the maximum landslide density doesn't appear in the area near the epicenter. Peak ground acceleration (PGA), which usually is used to judge the limit boundary of co-seismic landslide distribution, cannot explain this distribution pattern. Instead of correlating geological and topographic factors with the co-seismic landslide distribution pattern, this study focuses on analyzing the influence of seismic landslide susceptibility on the co-seismic distribution. Seismic landslide susceptibility comes from a calculation of critical acceleration values using a simplified Newmark block model analysis and represents slope stability under seismic loading. Both DEM (SRTM 90m)and geological map (1 ︰ 200^￣￣000)are used as inputs to calculate critical acceleration values. Results show that the most susceptible slopes with the smallest critical accelerations are generally concentrated along the banks of rivers. The stable slopes, which have the larger critical accelerations and are comparably stable, are in the places adjacent to the epicenter. Comparison of the distribution of slope stability and the real landslides triggered by the 2014 M_W6.1 Ludian earthquake shows a good spatial correlation, meaning seismic landslide susceptibility controls the co-seismic landslide distributions to a certain degree. Moreover, our study provides a plausible explanation on the special distribution pattern of Ludian earthquake triggered landslides. Also the paper discusses the advantages of using the seismic landslide susceptibility as a basic map, which will offer an additional tool that can be used to assist in post-disaster response activities as well as seismic landslides hazards zonation.