Using convolutional neural network (CNN) methods and satellite images for landslide identification and classification is a very efficient and popular task in geological hazard investigations. However, traditional CNNs have two disadvantages: (1) insufficient training images from the study area and (2) uneven distribution of the training set and validation set. In this paper, we introduced distant domain transfer learning (DDTL) methods for landslide detection and classification. We first introduce scene classification satellite imagery into the landslide detection task. In addition, in order to more effectively extract information from satellite images, we innovatively add an attention mechanism to DDTL (AM-DDTL). In this paper, the Longgang study area, a district in Shenzhen City, Guangdong Province, has only 177 samples as the landslide target domain. We examine the effect of DDTL by comparing three methods: the convolutional CNN, pretrained model and DDTL. We compare different attention mechanisms based on the DDTL. The experimental results show that the DDTL method has better detection performance than the normal CNN, and the AM-DDTL models achieve 94% classification accuracy, which is 7% higher than the conventional DDTL method. The requirements for the detection and classification of potential landslides at different disaster zones can be met by applying the AM-DDTL algorithm, which outperforms traditional CNN methods.

Accurate and timely landslide susceptibility mapping (LSM) is essential to effectively reduce the risk of landslide. In recent years, deep learning has been successfully applied to landslide susceptibility assessment due to the strong ability of fitting. However, in actual applications, the number of labeled samples is usually not sufficient for the training component. In this paper, a deep neural network model based on semi-supervised learning (SSL-DNN) for landslide susceptibility is proposed, which makes full use of a large number of spatial information (unlabeled data) with limited labeled data in the region to train the mode. Taking Jiaohe County in Jilin Province, China as an example, the landslide inventory from 2000 to 2017 was collected and 12 metrological, geographical, and human explanatory factors were compiled. Meanwhile, supervised models such as deep neural network (DNN), support vector machine (SVM), and logistic regression (LR) were implemented for comparison. Then, the landslide susceptibility was plotted and a series of evaluation tools such as class accuracy, predictive rate curves (AUC), and information gain ratio (IGR) were calculated to compare the prediction of models and factors. Experimental results indicate that the proposed SSL-DNN model (AUC = 0.898) outperformed all the comparison models. Therefore, semi-supervised deep learning could be considered as a potential approach for LSM.

Debris flow susceptibility analysis is a prerequisite of risk assessment. The main objective of this study was to explore the accuracy and practicability of mapping units for evaluation of debris flow susceptibility. These units include grid cell units (GCUs), and watershed units (WUs) with the flow thresholds 10 000 (WU 10 000) and 5000 (WU 5000). The frequency ratio (FR) model was selected as the statistical method. Yongji County (YJC) of Jilin Province, China was selected as the research site, and a total of 123 debris flow disasters were surveyed. Eight influencing factors were considered and a total of three models were constructed. The predictive capabilities of the models were verified using an ROC curve and AUC. The results showed the three models to be accurate and the evaluation results of the GCU were found to be more accurate than others. However, when considering the effects of geology and geomorphology on the occurrence of debris flows, the WU was more feasible than the GCU. Therefore, the results indicate that the evaluation of debris flow susceptibility should be carried out based on the WU of the appropriate flow threshold in combination with the actual prevention and control of debris flow disasters.

Landslides are one of the most frequent geomorphic hazards, and they often result in the loss of property and human life in the Changbai Mountain area (CMA), Northeast China. The objective of this study was to produce and compare landslide susceptibility maps for the CMA using an information content model (ICM) with three knowledge-driven methods (the artificial hierarchy process with the ICM (AHP-ICM), the entropy weight method with the ICM (EWM-ICM), and the rough set with the ICM (RS-ICM)) and to explore the influence of different knowledge-driven methods for a series of parameters on the accuracy of landslide susceptibility mapping (LSM). In this research, the landslide inventory data (145 landslides) were randomly divided into a training dataset: 70% (81 landslides) were used for training the models and 30% (35 landslides) were used for validation. In addition, 13 layers of landslide conditioning factors, namely, altitude, slope gradient, slope aspect, lithology, distance to faults, distance to roads, distance to rivers, annual precipitation, land type, normalized difference vegetation index (NDVI), topographic wetness index (TWI), plan curvature, and profile curvature, were taken as independent, causal predictors. Landslide susceptibility maps were developed using the ICM, RS-ICM, AHP-ICM, and EWM-ICM, in which weights were assigned to every conditioning factor. The resultant susceptibility was validated using the area under the ROC curve (AUC) method. The success accuracies of the landslide susceptibility maps produced by the ICM, RS-ICM, AHP-ICM, and EWM-ICM methods were 0.931, 0.939, 0.912, and 0.883, respectively, with prediction accuracy rates of 0.926, 0.927, 0.917, and 0.878 for the ICM, RS-ICM, AHP-ICM, and EWM-ICM, respectively. Hence, it can be concluded that the four models used in this study gave close results, with the RS-ICM exhibiting the best performance in landslide susceptibility mapping.Entropy 2019, 21, 372 2 of 24 involving land use management as an efficient approach to reduce property damage and economic loss in landslide-prone areas [1,[6][7][8][9]. The outcome maps would be useful for general planned development activities and disaster management in the future, such as choosing new urban areas and infrastructural activities, as well as for environmental protection.Landslide susceptibility maps can be obtained using both qualitative (inventory-based and knowledge-driven methods) or quantitative approaches (data-driven methods and physically based models) [4,[10][11][12][13][14][15][16][17].Landslide inventory-based techniques, as a prelude to all other methods, include the collection of past landslide data, construction of databases, and production of susceptibility maps based on those data [18]. Landslide inventory mapping can be carried out using a variety of methods that were updated and summarized by Corominas et al. [17].Knowledge-driven methods that estimate landslide potential from the practical experience and expertise of the researcher are used by geomorphologists to analyze...

No abstract