Extraction and application analysis of landslide influential factors based on LiDAR DEM: a case study in the Three Gorges area, China

The aim of this study was to identify some new factors that may impact the occurrence and distribution of landslides based on light detection and ranging digital elevation model (LiDAR DEM), and to examine whether these factors can apply to distinguish between landslide and non-landslide pixels. Twenty-one landslide influential factors were identified. Thereinto, there were ten novel factors, namely the texture factors of slope and surface roughness, including the contrast (Con), correlation (Cor), angular second moment, entropy, and homogeneity (Hom) textures. Qualitative and quantitative analysis and feature selection method were applied to examine the application of these factors. The analysis results indicate that these factors have certain abilities to distinguish between landslide and non-landslide objects. And the selected optimal factors combination that derived from feature selection method was DEM, slope, Hom_d, Con_s, Cor_s, Hom_s, Con_r, Cor_r, and Hom_r (_d, _s, and _r represent DEM, slope, and surface roughness textures, respectively). In conclusion, the identified landslide influential factors can provide effective information for landslide identification. And the new texture factors of slope and surface roughness could act as important measurements that can improve the precision of landslide inventory mapping, susceptibility mapping, and risk assessment.