Search

Selecting appropriate mapping units is vital for landslide susceptibility mapping as well as pattern investigation, given that various units-based analyses extensively control the prediction performances. This paper investigated the landslide development through the interpretation and field surveys along the Yunling–Yanjing segment of the Lancang River in southeastern Tibet, and fulfil LSM with the consideration of 15 conditioning factors. Two grid unit methods (single-point and multi-point patterns) and two slope unit methods were comparatively analyzed for model training and mapping of landslide susceptibility via machine learning algorithms. The results suggest that the landslides are preferentially distributed in an elevation range of 2000–4000 m, in a slope range of 20–40°, a local relief range of 1000–2500 m, and southwest-oriented slopes. The data extracted by the multi-point method denotes a higher representation of landslide development features. All models possessed positive prediction ability for landslide susceptibility, and the multi-point method based on grid unit performed the best with the AUC exceeding 0.9. The best-performing models indicated that zones of high and very high susceptibility mainly distributed adjacent to the mainstem and some tributaries of the Lancang River. Furthermore, the distribution of "safety islands" (the slopes less prone to landslides) along National Highway G214 was reasonably illustrated as well, which provides a hazard predictability for such an important transportation corridor along the deeply-incised valley in the Lancang River. This study demonstrates a theoretical basis for the regional disaster prevention and mitigation for human activity, and provides methodological references for landslide susceptibility evaluation in similar mountainous areas over the world. [ABSTRACT FROM AUTHOR]

As one of the most destructive geological disasters, a myriad of landslides has revived and developed in the Three Gorges Reservoir area under the combined action of various detrimental factors. Therefore, the pertinently regional landslide susceptibility mapping (LSM) is of great significance for disaster prevention and mitigation. In this study, LSM is prepared by using a boosting-C5.0 decision tree model. Under the landslide verification of on-site investigations, the study area is divided into accumulation and rock areas, and a total of 12 impact factors are selected. TOL and VIF are employed to determine the multicollinearity among the impact factors. The independent training (80%) and validation (20%) datasets are constructed by random sampling for LSM. ANN, C5.0, and SVM are selected for comparative analysis. The results show that there is no rigorous multicollinearity among the impact factors proposed in this paper. The landslide susceptibility in the study area is divided into low, moderate, high, and very high. The highest susceptibility area distributes along the riverside where the landslide ratio is 37.05% in boosting-C5.0 model. Then the ROCs are expropriated to infer the accuracy of each model. The boosting-C5.0 performs the best with the largest area under the curve in both accumulation and rock areas, reaching at 0.991 and 0.990 in the validation sets, respectively. Finally, the composite modification of the 5 validation sets shows that the uncertainty of boosting-C5.0 is concentrated in the intermediate probability areas of susceptibility. This study reveals the feasibility of machine learning in landslide susceptibility assessment, which could provide a basis for the risk management and control of geological disasters. [ABSTRACT FROM AUTHOR]

As a frequent natural disaster, landslides incur significant economic and human losses worldwide. The main idea of this paper is to propose novel integrative models for landslide susceptibility evaluation in a prone area of Chaharmahal va Bakhtiari Province in Iran. To do this, four metaheuristic techniques, namely chimp optimization algorithm (ChOA), crow search algorithm (CSA), satin bowerbird optimization (SBO), and water cycle algorithm (WCA) are used to supervise the training of an artificial neural network (ANN). A spatial database is created by 170 historical landslides and 14 most common conditioning factors. The models are optimized in terms of hyper-parameters and elite ones are then used to produce the susceptibility maps. During the training phase, all models could acquire a reliable understanding of the landslide pattern. The obtained accuracy index of area under the receiving operating characteristic curve (AUROC) showed that the maps produced by the WCA-ANN (AUROC = 0.925) and SBO-ANN (AUROC = 0.900) are more accurate than those of the ChOA-ANN (AUROC = 0.851) and CSA-ANN (AUROC = 0.855). Considering this assessment, the suggested hybrid models are reliable enough for landslide susceptibility modeling. Accordingly, the generated susceptibility map can be used for urban planning and generating alert systems over the studied area. [ABSTRACT FROM AUTHOR]

Multi-platform remote sensing using space-, airborne and ground-based sensors has become essential tools for landslide assessment and disaster-risk prevention. Over the last 30 years, the multiplicity of Earth Observation satellites mission ensures uninterrupted optical and radar imagery archives. With the popularization of Unmanned Aerial Vehicles, free optical and radar imagery with high revisiting time, ground and aerial possibilities to perform high-resolution 3D point clouds and derived digital elevation models, it can make it difficult to choose the appropriate method for risk assessment. The aim of this paper is to review the mainstream remote-sensing methods commonly employed for landslide assessment, as well as processing. The purpose is to understand how remote-sensing techniques can be useful for landslide hazard detection and monitoring taking into consideration several constraints such as field location or costs of surveys. First we focus on the suitability of terrestrial, aerial and spaceborne systems that have been widely used for landslide assessment to underline their benefits and drawbacks for data acquisition, processing and interpretation. Several examples of application are presented such as Interferometry Synthetic Aperture Radar (InSAR), lasergrammetry, Terrestrial Optical Photogrammetry. Some of these techniques are unsuitable for slow moving landslides, others limited to large areas and others to local investigations. It can be complicated to select the most appropriate system. Today, the key for understanding landslides is the complementarity of methods and the automation of the data processing. All the mentioned approaches can be coupled (from field monitoring to satellite images analysis) to improve risk management, and the real challenge is to improve automatic solution for landslide recognition and monitoring for the implementation of near real-time emergency systems. [ABSTRACT FROM AUTHOR]

Landslide hazard is a complex nonlinear dynamical system with uncertainty. The evolution of landslide is influenced by many factors such as tectonic, rainfall and reservoir level fluctuation. Using a time series model, total accumulative displacement of landslide can be divided into the trend component displacement and the periodic component displacement according to the response relation between dynamic changes in landslide displacement and inducing factors. In this paper, a novel neural network technique called ensemble of extreme learning machine (E-ELM) is proposed to investigate the interactions of different inducing factors affecting the evolution of landslide. Grey relational analysis is used to sieve out the more influential inducing factors as the inputs in E-ELM. Trend component displacement and periodic component displacement are forecasted, respectively; then, total predictive displacement is obtained by adding the calculated predictive displacement value of each sub. Performances of our model are evaluated by using real data from Baishuihe landslide in the Three Gorges Reservoir of China, and it provides a good representation of the measured slide displacement behavior. [ABSTRACT FROM AUTHOR]

The aim of this study is to evaluate and compare the performances of 5 machine learning (ML) techniques for predicting the spatial probability of landslide at Atsuma, Japan, and Mt. Umyeon, Korea. 5 ML models used are Naïve Bayes (NB), ensemble learning (random forest (RF) and adaboost (AB)), and deep learning (multilayer perceptron (MLP) and convolutional neural network (CNN)) models. Real landslide events at the study areas are randomly separated to the training set for landslide mapping and the validation set for assessing performance. To assess the performance of the used models, the resulting models are validated using receiver operating characteristic (ROC) curve. The success rate curves show that the areas under the curve (AUC) for the NB, RF, AB, MLP, and CNN are 85.1, 88.8, 88.6, 87.5, and 95.0%, respectively, at Atsuma and 68.7, 85.6, 90.5, 81.6, and 92.0%, respectively, at Mt. Umyeon. Similarly, the validation results show that the areas under the curve for the NB, RF, AB, MLP, and CNN are 84.3, 87.1, 87.1, 86.7, and 89.7%, respectively, at Atsuma and 64.9, 85.5, 83.9, 84.7, and 90.5%, respectively, at Mt. Umyeon. In addition, statistical tests such as Chi-square test and difference of proportions test show that all classified landslide susceptibility maps have statistical significance and the significant difference in classified landslide susceptibility maps from different ML models. The comparison results among 5 ML models show that the CNN model had the best performance and NB model had the worst performance in both study areas. [ABSTRACT FROM AUTHOR]