Comparison of Machine Learning Methods for Potential Active Landslide Hazards Identification with Multi-Source Data

DRs-UNet: A Deep Semantic Segmentation Network for the Recognition of Active Landslides from InSAR Imagery in the Three Rivers Region of the Qinghai–Tibet Plateau

At present, Synthetic Aperture Radar Interferometry (InSAR) has been an important technique for active landslides recognition in the geological survey field. However, the traditional interpretation method through human–computer interaction highly relies on expert experience, which is time-consuming and subjective. To solve the problem, this study designed an end-to-end semantic segmentation network, called deep residual shrinkage U-Net (DRs-UNet), to automatically extract potential active landslides in InSAR imagery. The proposed model was inspired by the structure of U-Net and adopted a residual shrinkage building unit (RSBU) as the feature extraction block in its encoder part. The method of this study has three main advantages: (1) The RSBU in the encoder part incorporated with soft thresholding can reduce the influence of noise from InSAR images. (2) The residual connection of the RSBU makes the training of the network easier and accelerates the convergency process. (3) The feature fusion of the corresponding layers between the encoder and decoder effectively improves the classification accuracy. Two widely used networks, U-Net and SegNet, were trained under the same experiment environment to compare with the proposed method. The experiment results in the test set show that our method achieved the best performance; specifically, the F1 score is 1.48% and 4.1% higher than U-Net and SegNet, which indicates a better balance between precision and recall. Additionally, our method has the best IoU score of over 90%. Furthermore, we applied our network to a test area located in Zhongxinrong County along Jinsha River where landslides are highly evolved. The quantitative evaluation results prove that our method is effective for the automatic recognition of potential active landslide hazards from InSAR imagery.

Susceptibility Analysis of Geohazards in the Longmen Mountain Region after the Wenchuan Earthquake

Multitemporal geohazard susceptibility analysis can not only provide reliable results but can also help identify the differences in the mechanisms of different elements under different temporal and spatial backgrounds, so as to better accurately prevent and control geohazards. Here, we studied the 12 counties (cities) that were severely affected by the Wenchuan earthquake of 12 May 2008. Our study was divided into four time periods: 2008, 2009–2012, 2013, and 2014–2017. Common geohazards in the study area, such as landslides, collapses and debris flows, were taken into account. We constructed a geohazard susceptibility index evaluation system that included topography, geology, land cover, meteorology, hydrology, and human activities. Then we used a random forest model to study the changes in geohazard susceptibility during the Wenchuan earthquake, the following ten years, and its driving mechanisms. We had four main findings. (1) The susceptibility of geohazards from 2008 to 2017 gradually increased and their spatial distribution was significantly correlated with the main faults and rivers. (2) The Yingxiu-Beichuan Fault, the western section of the Jiangyou-Dujiangyan Fault, and the Minjiang and Fujiang rivers were highly susceptible to geohazards, and changes in geohazard susceptibility mainly occurred along the Pingwu-Qingchuan Fault, the eastern section of the Jiangyou-Dujiangyan Fault, and the riparian areas of the Mianyuan River, Zagunao River, Tongkou River, Baicao River, and other secondary rivers. (3) The relative contribution of topographic factors to geohazards in the four different periods was stable, geological factors slowly decreased, and meteorological and hydrological factors increased. In addition, the impact of land cover in 2008 was more significant than during other periods, and the impact of human activities had an upward trend from 2008 to 2017. (4) Elevation and slope had significant topographical effects, coupled with the geological environmental effects of engineering rock groups and faults, and river-derived effects, which resulted in a spatial aggregation of geohazard susceptibility. We attributed the dynamic changes in the areas that were highly susceptible to geohazards around the faults and rivers to the changes in the intensity of earthquakes and precipitation in different periods.

Statistical Time-Series Analysis of Interferometric Coherence from Sentinel-1 Sensors for Landslide Detection and Early Warning

Landslides are one of the most destructive natural hazards worldwide, affecting greatly built-up areas and critical infrastructure, causing loss of human lives, injuries, destruction of properties, and disturbance in everyday commute. Traditionally, landslides are monitored through time consuming and costly in situ geotechnical investigations and a wide range of conventional means, such as inclinometers and boreholes. Earth Observation and the exploitation of the freely available Copernicus datasets, and especially Sentinel-1 Synthetic Aperture Radar (SAR) images, can assist in the systematic monitoring of landslides, irrespective of weather conditions and time of day, overcoming the restrictions arising from in situ measurements. In the present study, a comprehensive statistical analysis of coherence obtained through processing of a time-series of Sentinel-1 SAR imagery was carried out to investigate and detect early indications of a landslide that took place in Cyprus on 15 February 2019. The application of the proposed methodology led to the detection of a sudden coherence loss prior to the landslide occurrence that can be used as input to Early Warning Systems, giving valuable on-time information about an upcoming landslide to emergency response authorities and the public, saving numerous lives. The statistical significance of the results was tested using Analysis of Variance (ANOVA) tests and two-tailed t-tests.