Landslide disasters occur frequently in the mountainous areas in southwest China, which pose serious threats to the local residents. Interferometry Synthetic Aperture Radar (InSAR) provides us the ability to identify active slopes as potential landslides in vast mountainous areas, to help prevent and mitigate the disasters. Quickly and accurately identifying potential landslides based on massive SAR data is of great significance. Taking the national highway near Wenchuan County, China, as study area, this paper used a Stacking-InSAR method to quickly and qualitatively identify potential landslides based on a total of 40 Sentinel SAR images acquired from November 2017 to March 2019. As a result, 72 active slopes were successfully detected as potential landslides. By comparing the results from Stacking-InSAR with the results from the traditional SBAS-InSAR (Small Baselines Subset) time series method, it was found that the two methods had a high consistency, with 81.7% potential landslides identified by both of the two methods. A detailed comparison on the detection differences was performed, revealing that Stacking-InSAR, compared to SBAS-InSAR may miss a few active slopes with small spatial scales, small displacement levels and the ones affected by the atmosphere, while it has good performance on poor-coherence regions, with the advantages of low technical requirements and low computation labor. The Stacking-InSAR method would be a fast and powerful method to qualitatively and effectively identify potential landslides in vast mountainous areas, with a comprehensive understanding of its specialty and limitations.

Landslides, a major natural geohazard, obstruct municipal constructions and may destroy villages and towns, at worst causing significant casualties and economic losses. Interferometric Synthetic Aperture Radar (InSAR) technique offers distinct advantages on landslide detection and monitoring. In this paper, a more systematic workflow is designed for InSAR study of landslides, in terms of three levels: (i) early detection on regional scale, (ii) three-dimensional (3D) surface displacement rates estimation on detailed scale, and (iii) time series analysis on long-term temporal scale. The proposed workflow is applied for landslide research on the Xiaojiang River Basin, China, using ascending and descending Sentinel-1 images acquired from March 2017 to May 2019. First, the landslide inventory has been mapped and updated using InSAR stacking method, supporting geohazard prevention on a regional scale. A total of 22 active landslides are identified, ranging from medium to super large scale. Compared with the existing inventory, three unrecorded landslides are newly detected by our approach, and five recorded landslides are detected significant expansion of their boundaries. Then, specific to a detected landslide, Baobao landslide, a Total Least Squares–Kalman Filter-based approach is presented. Two outcomes are provided for further spatial-temporal pattern analysis: 3D displacement rates, providing an intuitive insight on the spatial characteristics and sliding direction of landslide, which are analyzed to deep the understanding of its kinematic mechanism, and long-term time series, which contribute to deduce the dynamic evolution of landslide, presenting benefits in landslide forecasting.

Landslides are common hazardous geological events, and accurate and efficient landslide identification methods are important for hazard assessment and post-disaster response to geological disasters. Deep learning (DL) methods based on remote sensing data are currently widely used in landslide identification tasks. The recently proposed segment anything model (SAM) has shown strong generalization capabilities in zero-shot semantic segmentation. Nevertheless, SAM heavily relies on user-provided prompts, and performs poorly in identifying landslides on remote sensing images. In this study, we propose a SAM-based cross-feature fusion network (SAM-CFFNet) for the landslide identification task. The model utilizes SAM’s image encoder to extract multi-level features and our proposed cross-feature fusion decoder (CFFD) to generate high-precision segmentation results. The CFFD enhances landslide information through fine-tuning and cross-fusing multi-level features while leveraging a shallow feature extractor (SFE) to supplement texture details and improve recognition performance. SAM-CFFNet achieves high-precision landslide identification without the need for prompts while retaining SAM’s robust feature extraction capabilities. Experimental results on three open-source landslide datasets show that SAM-CFFNet outperformed other comparative models in terms of landslide identification accuracy and achieved an intersection over union (IoU) of 77.13%, 55.26%, and 73.87% on the three datasets, respectively. Our ablation studies confirm the effectiveness of each module designed in our model. Moreover, we validated the justification for our CFFD design through comparative analysis with diverse decoders. SAM-CFFNet achieves precise landslide identification using remote sensing images, demonstrating the potential application of the SAM-based model in geohazard analysis.

Landslides are one of the most destructive natural disasters in the world, threatening human life and safety. With excellent performance as a foundation model for image segmentation, the Segment Anything Model (SAM) has provided a novel paradigm for semantic segmentation research. However, the lack of remote sensing images in the SAM training data limits its ability to recognize landslides. Additionally, despite the transfer learning approach can transfer SAM feature extraction capability to the landslide segmentation task, but it will consume a lot of computational resources and training time. In order to solve these challenges, this study proposes a TransLandSeg model that transfers the segmentation capability of SAM while learning landslide features at a low training cost. To limit model training parameters, the adaptive transfer learning (ATL) module is purposely designed, the image encoder is frozen during model training, only the ATL module and mask decoder are trained, and the knowledge learned from the ATL module is input into the original network. Moreover, to select the best ATL module, we also designed 9 kinds of ATL modules and analyzed the accuracy of the TransLandSeg model with different ATL modules. We selected the Bijie landslide dataset and the Landslide4Sense dataset for model training and testing. The experiment results show that the TransLandSeg model increases the mean intersection over union (MIoU) by 1.48%-13.01% compared to other state-of-the-art semantic segmentation models. In addition, TransLandSeg requires only 1.3% of SAM parameters to enable SAM's powerful capabilities to transfer to landslide segmentation. The source code for this work will be accessible at https://github.com/JunchuanYu/TransLandSeg.

The development characteristics of land subsidence of Xi'an by an advanced InSAR technique are investigated based on Envisat, Radarsat-2 and Sentinel-1 satellites, and it can be found that there still have been five typical land subsidence bowls with maximum subsidence rates beyond 40 mm/year in Xi'an. In 2012, the land subsidence took on slightly slowing down with a ground rebound area of up to 6.3 km². But in 2003–2010, 2016 and 2017, there was more serious land subsidence and especially in 2017 the maximum velocity reached up to −137 mm/year. Yuhuazhai-Nanyaotou subsidence bowl as the greatest subsidence bowl, its area that rate was below 50 mm/year in 2017 was up to 10.64 km². Moreover, these land subsidence bowls mainly concentrate upon tectonic depression area due to huge thickness clay soil and are spatially and temporally coincident with confined aquifer depression cones, suggesting that over-exploitation groundwater plays a decisive role for land subsidence. Ground fissure belts are generally controlled by normal faults in Xi'an, but asymmetry settlement of land surface can also give rise to the abnormal activities of ground fissure belts, which have caused serious threat or damage to building structures. In addition, it could also be found that surface uplift area in 2012 involves in both the tectonic depression area and loess ridge in terms of their same thickness sand layer.

Lakes are an important component of global water resources. In order to achieve accurate lake extractions on a large scale, this study takes the Tibetan Plateau as the study area and proposes an Automated Lake Extraction Workflow (ALEW) based on the Google Earth Engine (GEE) and deep learning in response to the problems of a low lake identification accuracy and low efficiency in complex situations. It involves pre-processing massive images and creating a database of examples of lake extraction on the Tibetan Plateau. A lightweight convolutional neural network named LiteConvNet is constructed that makes it possible to obtain spatial–spectral features for accurate extractions while using less computational resources. We execute model training and online predictions using the Google Cloud platform, which leads to the rapid extraction of lakes over the whole Tibetan Plateau. We assess LiteConvNet, along with thresholding, traditional machine learning, and various open-source classification products, through both visual interpretation and quantitative analysis. The results demonstrate that the LiteConvNet model may greatly enhance the precision of lake extraction in intricate settings, achieving an overall accuracy of 97.44%. The method presented in this paper demonstrates promising capabilities in extracting lake information on a large scale, offering practical benefits for the remote sensing monitoring and management of water resources in cloudy and climate-differentiated regions.