Study of early identification method for large landslides in high vegetation coverage areas of Southwest China

Abstract

Landslide disasters with dense vegetation and steep terrain, and high concealment frequently occur in Southwest China. Current field surveys, unmanned aerial vehicle (UAV) photogrammetry, and Interferometric Synthetic Aperture Radar (InSAR) technologies all have limitations in complex environments with high vegetation coverage. In this study, the landslide in Xinmo Village, Mao County, Aba Prefecture, Sichuan Province, was used as the research object. The slope types were divided according to the regional stratum occurrence and slope direction, and the dip slope was identified as the pre-selected area for the landslide. Nine vegetation indexes were constructed based on Landsat 8 Operational Land Imager (OLI) data, and Modified Soil Adjusted Vegetation Index (MSAVI) with high correlation was selected as the indicator of landslide change to estimate the vegetation coverage. The relationship between vegetation anomalies and landslide creep was analyzed by superimposing slope structure and vegetation spatial variation characteristics. The results showed that from May 2015 to May 2017, the vegetation coverage in the landslide main source area, above the deformation body, local collapse area, and around the washouts showed a significant decrease; i.e., as the time of landslide was approaching, some vegetation in the study area was affected by the landslide deformation and the growth condition became worse. Between April and May 2017, the vegetation coverage in the area not affected by the landslide was less than 0.6 (i.e., bare ground area) decreased abruptly, with change rates of 78.4, 87.7, and 89.7%, respectively, which is consistent with the development pattern of vegetation in the growing period; while the reduction rate of image elements in the vegetation abnormal area was only 20.5%, which judged that the vegetation might be affected by landslide creep and the growth and development were hindered. The study shows that there is an obvious spatial–temporal correlation between vegetation anomalies and landslide deformation during the landslide creep phase, which indirectly reflects the evolution process of landslide gradual destabilization and provides a theoretical basis for the early identification of landslides in high vegetation coverage areas.