Search

Machine learning models have been widely used in landslide susceptibility prediction. However, landslide multidimensional feature extraction, model generalization ability, and prediction quantification of deep learning are still challenging. This paper proposes a new approach, the landslide density-based bidirectional long short-term memory (LD-BiLSTM) model with multichannel input and an optimized sampling strategy to predict and map landslide susceptibility in active seismic mountainous areas of Sichuan Province, China. First, to ensure the generalization ability of the LD-BiLSTM model, other regions in Sichuan were selected as the model training area independent of the prediction area (Luding County). Multichannel landslide datasets were constructed to extract high-dimensional geospatial features of landslides. Subsequently, the landslide density of each grid cell was utilized as the label for the corresponding input sample. The LD-BiLSTM model was improved by using transfer learning and sampling optimization strategies, which makes our method attenuate the impact of historical landslide inventory deviation on the spatial susceptibility mode compared with the existing DL model, which usually uses landslide objects (LO) as input sample labels. Model performance evaluation results show that the LD-BiLSTM model (precision = 0.903, recall = 0.899, F1-score = 0.901, Area under receiver operating characteristic curve (AUC) = 0.940) outperformed the LO-BiLSTM model (precision = 0.812, recall = 0.815. F1-score = 0.813, AUC = 0.910) in the case areas. Meanwhile, the performance of the LD-BiLSTM model (AUC = 0.9407) significantly outperformed both the information value (IV) (AUC = 0.7207) model and the random forest (RF) (AUC = 0.8116) models in the landslide prediction area (Luding), which confirms that the proposed LD-based method is superior to traditional LO-based methods. Significantly, our approach can effectively extract the spatial distribution of landslides and predict potential landslides in complex high-mountain environments. [ABSTRACT FROM AUTHOR]

The accuracy of landslide displacement prediction can effectively prevent casualties and economic losses. To achieve accurate prediction of the Majiagou landslide displacement in the Three Gorges Reservoir (TGR), China, a hybrid machine learning prediction model considering the deformation hysteresis effect is proposed. The real-time deep displacement measurements were captured by using in-place inclinometers with Fiber Bragg grating (FBG) sensors. The time series method was adopted to divide the total displacement into a trend term and periodic term. Trend displacement was determined by the geological condition and predicted by the fitting method. Periodic displacement was controlled by external factors such as rainfall and fluctuation of reservoir water level. Before making the prediction, the grey correlation analysis was adopted to confirm that the fluctuation of the reservoir water level was the main influence factor. In view of the deficiency that current prediction methods could not quantitatively determine the lag time of landslide deformation and thus select the influencing factors empirically, the dynamic analysis of the correlation between periodic influence factors and periodic displacement was carried out in this paper, and the deformation lag time was identified to be 18 days by using set pair analysis (SPA) method. Finally, the optimal influence factors were selected and the prediction model of Majiagou landslide based on support vector machine optimized by particle swarm optimization (SPA-PSO-SVM) was established. Results showed that the root mean square error (RMSE) and the mean absolute percentage error (MAPE) of the proposed SPA-PSO-SVM prediction model are 0.28 and 12.8, respectively. Compared with the PSO-SVM model, the prediction accuracy of the proposed model had been improved significantly. The reliability and effectiveness of the SPA-PSO-SVM prediction model is verified and it has apparent advantages while predicting landslide displacement with deformation hysteresis effect involved. [ABSTRACT FROM AUTHOR]

The environmental factors of landslide susceptibility are generally uncorrelated or non-linearly correlated, resulting in the limited prediction performances of conventional machine learning methods for landslide susceptibility prediction (LSP). Deep learning methods can exploit low-level features and high-level representations of information from environmental factors. In this paper, a novel deep learning–based algorithm, the fully connected spare autoencoder (FC-SAE), is proposed for LSP. The FC-SAE consists of four steps: raw feature dropout in input layers, a sparse feature encoder in hidden layers, sparse feature extraction in output layers, and classification and prediction. The Sinan County of Guizhou Province in China, with a total of 23,195 landslide grid cells (306 recorded landslides) and 23,195 randomly selected non-landslide grid cells, was used as study case. The frequency ratio values of 27 environmental factors were taken as the input variables of FC-SAE. All 46,390 landslide and non-landslide grid cells were randomly divided into a training dataset (70%) and a test dataset (30%). By analyzing real landslide/non-landslide data, the performances of the FC-SAE and two other conventional machine learning methods, support vector machine (SVM) and back-propagation neural network (BPNN), were compared. The results show that the prediction rate and total accuracies of the FC-SAE are 0.854 and 85.2% which are higher than those of the SVM-only (0.827 and 81.56%) and BPNN (0.819 and 80.86%), respectively. In conclusion, the asymmetric and unsupervised FC-SAE can extract optimal non-linear features from environmental factors successfully, outperforms some conventional machine learning methods, and is promising for LSP. [ABSTRACT FROM AUTHOR]

In this paper, we propose a methodology for landslide susceptibility assessment at a regional scale in Yunnan, southwestern province of China. A landslide inventory map including 3,242 landslide points was prepared for the study area. Five factors recognized as correlated to landslide (namely, lithology, relative relief, tectonic fault density, rainfall, and road density) were analyzed and mapped in geographic information system. An index expressing the correlation between each factor and landslides [called class landslide susceptibility index (CLSI)] was proposed in the study. While analyzing landslide distribution in a large area, point aggregation might be expected. To quantify the uncertainty caused by aggregation, class landslide aggregation index was proposed. To account for the importance of each of the factors in the landslide susceptibility assessment, some weights were calculated by means of analytic hierarchy process. We propose a weighted class landslide susceptibility model (WCLSM), obtained by the combination of CLSI values of each factor with the correspondent weight. WCLSM performance in the study area was evaluated comparing the results obtained by first modeling all landslides and then by performing a time partition. The model was run including only landslides that occurred before 2009 and then validated with respect to landslides that occurred after 2009. The prediction-rate curve shows that the WCLSM model provides a good prediction for the study area. Of the study area, 21.4 % shows very high and high susceptibility and includes the 87.7 % of the number of landslides that occurred after 2009. [ABSTRACT FROM AUTHOR]

Sliding mass of landslides highly endangered the area along travel path, especially landslides with long travel distance. It is necessary to develop an effective prediction model for preliminarily evaluating landslide travel distance so as to improve disaster prevention and relocation. This paper collected 54 landslides with 347-4,170 m travel distance triggered by the 2008 Wenchuan earthquake to discuss the effectiveness of various influential factors on landslide travel distance and obtained an empirical model for its prediction. The results revealed that rock type, sliding source volume, and slope transition angle were the predominant factors on landslide travel distance. The validity of proposed model was verified by the satisfactory agreement between observations and predictions. Therefore, this model might be practically applicable in Wenchuan earthquake area and other similar geomorphological and geological regions. [ABSTRACT FROM AUTHOR]

As one of the major problems of geo-engineering, landslides often influence the safety of linear engineering projects that cross mountainous areas. Therefore, when selecting suitable routes for such projects, it is important to assess their susceptibility to landslides. In this paper, we used a natural gas pipeline in the northeast of the Yunnan-Guizhou Plateau of China as a case study to analyze landslide susceptibility. Based on engineering geological analogy, the analytical hierarchy process, and the least-squares method, a regional landslide susceptibility assessment model was developed and was programmed using GIS ArcEngine components under the Visual Studio.NET environment. The landslide susceptibility along the Zhong-Wu natural gas pipeline from Zhongxian County to Wuhan was assessed based on this model and classified into five levels: very safe, safe, moderate, susceptible, and very susceptible. The high accuracy and prediction capability of the model were confirmed by comparing the model results with past landslide data and performing a prediction test. The results indicated that the assessment model used in this study is reliable and can be used for landslide susceptibility assessment and route selection in other areas. [ABSTRACT FROM AUTHOR]