Landslides pose a greater potential risk to the Sichuan-Tibet Transportation Project, and extensive landslide inventory and mapping are essential to prevent and control geological hazards along the Sichuan-Tibet Transportation Corridor (STTC). Recently proposed landslide detection methods mainly focused on new landslides with high vegetation. In addition, there are still challenges in automatic detection of old landslides using optical images. In this paper, two methods, namely mask region-based convolutional neural networks (Mask R-CNN) and transfer learning Mask R-CNN (TL-Mask R-CNN), are presented for detecting and segmenting new and old landslides, respectively. An optical remote sensing dataset for landslide recognition along the Sichuan-Tibet Transportation Corridor (LRSTTC) is constructed as an evaluation benchmark. Our experimental results show that the recall rate and F1-score of the proposed method for new landslide detection can reach 78.47% and 79.80%, respectively. Transfer learning is adopted to detect old landslides, and our experimental results show that evaluation indices can be further improved by about 10%. Furthermore, TL-Mask R-CNN has been applied to identify ice avalanches based on the characteristics of landslides. It appears that our proposed methods can detect and segment landslides effectively along the STTC with the constructed LRSTTC dataset, which is essential for studying and preventing landslide hazards in mountainous areas.

Landslides have attracted extensive attention and research worldwide because of their tremendous destructiveness, and many catastrophic landslides have occurred along river corridors and reservoir bank slopes. The detection and detailed monitoring of potential landslides are prerequisites for landslide disaster prevention, with recent advance in spaceborne Interferometric Synthetic Aperture Radar (InSAR) having effectively addressed this challenge. In this study, a wide-area landslide detection and monitoring framework, combining multiple InSAR techniques, is proposed to investigate active landslides along the upper reaches of the Yellow River from Longyang Gorge to Liujia Gorge (UYRLL), north-west China. A total of 597 active landslides have been mapped in this region for the first time. Our analyses suggest that a large percentage of the detected landslides are distributed at an elevation of 2,000-3,000 m with slope angles of 5-30°and their spatial distribution density is correlated with regional tectonic activity. Multi-Temporal InSAR (MTI) techniques have also been adopted to analyze the surface motion characteristics of two typical landslides (i.e., the Lijia landslide and the Xijitan landslide), showing that these two landslides are dominated by linear motion. The landslide movements also contain slight non-linear oscillations, inferred from wavelet analysis to be associated with reservoir water area changes and seasonal rainfall. The proposed framework herein can be extended to the detection and monitoring of active landslides in other areas with similar geomorphic features, and the above findings on landslide characteristics can also contribute to landslide prevention and mitigation in the upper Yellow River area.

Emergency response after earthquakes, especially rapid access to building damage information, is of great significance to ensure timely rescue and reduce casualties. However, manual field surveys of building damages are inefficient and dangerous, and optical satellite data are more susceptible to cloud interference after earthquakes. Synthetic Aperture Radar (SAR) is now widely used in disaster response efforts due to its full-time and all-weather capability. According to the change in coherence between SAR images before and after earthquakes, it is possible to identify damaged buildings that cause coherence loss. However, the accuracy of traditional coherence-based damage detection methods is relatively low due to biases in coherence estimation and inconsistency in spatio-temporal baselines. In this study, we propose a new method to produce a post-earthquake Building Damage Proxy Map (BDPM) based on multi-temporal Sentinel-1 coherence, which incorporates homogeneous SAR pixel coherence estimation and histogram matching techniques. The former is used to reduce estimation biases and the latter to reduce the effect of baseline inconsistencies in adjacent coherence maps. We successfully applied this method to the 2022 Mw 6.2 Afghanistan earthquake, the 2023 strong earthquake sequence in Turkey, and the 2023 M s 6.2 Jishishan, China earthquake. We also verified its accuracy (over 80%) by comparing the BDPM with results from the United Nations Institute for Training and Research (UNITAR) and analyzed various factors affecting the distribution of damaged buildings. These analyses confirm the effectiveness of our method for generating BDPM using free medium-resolution Sentinel-1 data, which will greatly assist in earthquake emergency response.

Wide area landslide detection is a major international research hotspot in the field of geological hazards, and the integration of multi-temporal optical satellite images and spaceborne interferometric synthetic aperture radar (InSAR) appears to be an effective way to realize this. In this paper, a technical framework is presented for wide area landslide detection: (i) multi-temporal satellite optical images are used to detect landslides with distinguishable geomorphological features; (ii) Generic Atmospheric Correction Online Service (GACOS) assisted InSAR stacking is employed to generate annual surface displacement rate maps in radar line of sight using satellite SAR images from both ascending and descending tracks, which are in turn utilized to automatically detect active landslides from ground motion using hotspot analysis, and (iii) the distribution characteristics of the detected landslides are investigated by examining their relationships with topographic and hydrological factors. Three expressways in Sichuan Province, China—namely the Yakang (Ya’an-Kangding), Yaxi (Ya’an-Xichang), and Lushi (Luding-Shimian) expressways—and their surrounding regions (a total area of approximately 20,000 square kilometers) were chosen as the study area. A total of 413 landslides were detected, among which 320 were detected using multi-temporal satellite optical images, and 109 were detected using GACOS-assisted InSAR stacking. It should be noted that only 16 landslides were detected by both approaches; these landslides all exhibited not only obvious geomorphological features but also ground motion. A statistical analysis of the topographic and hydrological factors shows that of the detected landslides: 81% are distributed at elevations of 1000–2500 m, over 60% lie within the elevation range of 100~400 m, and 90% present with medium and steep slopes (20°~45°), and 80% are located within areas seeing an annual rainfall of 950~1050 mm. Nine landslides were found to pose potential safety hazards to the expressways. The research findings in this paper have directly benefitted the Sichuan expressways; equally important, it is believed that the technical framework presented in this paper will provide guidance for hazard mitigation and the prevention of transportation hazards in the future.