Earthquake-triggered landslides frequently occur in active mountain areas, which poses great threats to the safety of human lives and public infrastructures. Fast and accurate mapping of coseismic landslides is important for earthquake disaster emergency rescue and landslide risk analysis. Machine learning methods provide automatic solutions for landslide detection, which are more efficient than manual landslide mapping. Deep learning technologies are attracting increasing interest in automatic landslide detection. CNN is one of the most widely used deep learning frameworks for landslide detection. However, in practice, the performance of the existing CNN-based landslide detection models is still far from practical application. Recently, Transformer has achieved better performance in many computer vision tasks, which provides a great opportunity for improving the accuracy of landslide detection. To fill this gap, we explore whether Transformer can outperform CNNs in the landslide detection task. Specifically, we build a new dataset for identifying coseismic landslides. The Transformer-based semantic segmentation model SegFormer is employed to identify coseismic landslides. SegFormer leverages Transformer to obtain a large receptive field, which is much larger than CNN. SegFormer introduces overlapped patch embedding to capture the interaction of adjacent image patches. SegFormer also introduces a simple MLP decoder and sequence reduction to improve its efficiency. The semantic segmentation results of SegFormer are further improved by leveraging image processing operations to distinguish different landslide instances and remove invalid holes. Extensive experiments have been conducted to compare Transformer-based model SegFormer with other popular CNN-based models, including HRNet, DeepLabV3, Attention-UNet, U${}^2$Net and FastSCNN. SegFormer improves the accuracy, mIoU, IoU and F1 score of landslide detectuin by 2.2%, 5% and 3%, respectively. SegFormer also reduces the pixel-wise classification error rate by 14%. Both quantitative evaluation and visualization results show that Transformer is capable of outperforming CNNs in landslide detection.

Rapid and accurate identification of landslides is an essential part of landslide hazard assessment, and in particular it is useful for land use planning, disaster prevention, and risk control. Recent alternatives to manual landslide mapping are moving in the direction of artificial intelligence—aided recognition of these surface processes. However, so far, the technological advancements have not produced robust automated mapping tools whose domain of validity holds in any area across the globe. For instance, capturing historical landslides in densely vegetated areas is still a challenge. This study proposed a deep learning method based on Light Detection and Ranging (LiDAR) data for automatic identification of historical landslides. Additionally, it tested this method in the Jiuzhaigou earthquake-hit region of Sichuan Province (China). Specifically, we generated a Red Relief Image Map (RRIM), which was obtained via high-precision airborne LiDAR data, and on the basis of this information we trained a Lightweight Attention U-Net (LAU-Net) to map a total of 1949 historical landslides. Overall, our model recognized the aforementioned landslides with high accuracy and relatively low computational costs. We compared multiple performance indexes across several deep learning routines and different data types. The results showed that the Multiple-Class based Semantic Image Segmentation (MIOU) and the F1_score of the LAU-Net and RRIM reached 82.29% and 87.45%, which represented the best performance among the methods we tested.

To study the uncertainties of a collapse susceptibility prediction (CSP) under the coupled conditions of different data-based models and different connection methods between collapses and environmental factors, An’yuan County in China with 108 collapses is used as the study case, and 11 environmental factors are acquired by data analysis of Landsat TM 8 and high-resolution aerial images, using a hydrological and topographical spatial analysis of Digital Elevation Modeling in ArcGIS 10.2 software. Accordingly, 20 coupled conditions are proposed for CSP with five different connection methods (Probability Statistics (PSs), Frequency Ratio (FR), Information Value (IV), Index of Entropy (IOE) and Weight of Evidence (WOE)) and four data-based models (Analytic Hierarchy Process (AHP), Multiple Linear Regression (MLR), C5.0 Decision Tree (C5.0 DT) and Random Forest (RF)). Finally, the CSP uncertainties are assessed using the area under receiver operation curve (AUC), mean value, standard deviation and significance test, respectively. Results show that: (1) the WOE-based models have the highest AUC accuracy, lowest mean values and average rank, and a relatively large standard deviation; the mean values and average rank of all the FR-, IV- and IOE-based models are relatively large with low standard deviations; meanwhile, the AUC accuracies of FR-, IV- and IOE-based models are consistent but higher than those of the PS-based model. Hence, the WOE exhibits a greater spatial correlation performance than the other four methods. (2) Among all the data-based models, the RF model has the highest AUC accuracy, lowest mean value and mean rank, and a relatively large standard deviation. The CSP performance of the RF model is followed by the C5.0 DT, MLR and AHP models, respectively. (3) Under the coupled conditions, the WOE-RF model has the highest AUC accuracy, a relatively low mean value and average rank, and a high standard deviation. The PS-AHP model is opposite to the WOE-RF model. (4) In addition, the coupled models show slightly better CSP performances than those of the single data-based models not considering connect methods. The CSP performance of the other models falls somewhere in between. It is concluded that the WOE-RF is the most appropriate coupled condition for CSP than the other models.

As the Third Pole of the Earth and the Water Tower of Asia, the Tibetan Plateau (TP) nurtures large numbers of glacial lakes, which are sensitive to global climate change. These lakes modulate the freshwater ecosystem in the region but concurrently pose severe threats to the valley population by means of sudden glacial lake outbursts and consequent floods (GLOFs). The lack of high-resolution multi-temporal inventory of glacial lakes in TP hampers a better understanding and prediction of the future trend and risk of glacial lakes. Here, we created a multi-temporal inventory of glacial lakes in TP using a 30-year record of 42,833 satellite images (1990–2019), and we discussed their characteristics and spatio-temporal evolution over the years. Results showed that their number and area had increased by 3285 and 258.82 km² in the last 3 decades, respectively. We noticed that different regions of the TP exhibited varying change rates in glacial lake size; most regions show a trend of expansion and increase in glacial lakes, while some regions show a trend of decreasing such as the western Pamir and the eastern Hindu Kush. The mapping uncertainty is about 17.5%, which is lower than other available datasets, thus making our inventory reliable for the spatio-temporal evolution analysis of glacial lakes in the TP. Our lake inventory data are publicly published, it can help to study climate change–glacier–glacial lake–GLOF interactions in the Third Pole and serve as input to various hydro-climatic studies.

The Ms 7.0 Jiuzhaigou earthquake that occurred on 8 August 2017 triggered hundreds of landslides in the Jiuzhaigou valley scenic and historic-interest area in Sichuan, China, causing heavy casualties and serious property losses. Quick and accurate mapping of post-disaster landslide distribution is of paramount importance for earthquake emergency rescue and the analysis of post-seismic landslides distribution characteristics. The automatic identification of landslides is mostly based on medium- and low-resolution satellite-borne optical remote-sensing imageries, and the high-accuracy interpretation of earthquake-triggered landslides still relies on time-consuming manual interpretation. This paper describes a methodology based on the use of 1 m high-resolution unmanned aerial vehicle (UAV) imagery acquired after the earthquake, and proposes a support vector machine (SVM) classification method combining the roads and villages mask from pre-seismic remote sensing imagery to accurately and automatically map the landslide inventory. Compared with the results of manual visual interpretation, the automatic recognition accuracy could reach 99.89%, and the Kappa coefficient was higher than 0.9, suggesting that the proposed method and 1 m high-resolution UAV imagery greatly improved the mapping accuracy of the landslide area. We also analyzed the spatial-distribution characteristics of earthquake-triggered landslides with the influenced factors of altitude, slope gradient, slope aspect, and the nearest faults, which provided important support for the further study of post-disaster landslide distribution characteristics, susceptibility prediction, and risk assessment.