Reducing Local Correlations Among Causal Factor Classifications as a Strategy to Improve Landslide Susceptibility Mapping

• DOI: 10.3389/feart.2021.781674
• Publication Date: 2021-11-22T06:12:50Z
• Authors: Ting Xiao; Lanbing Yu; Weiming Tian; Chang Zhou; Luqi Wang
• Keywords: Spatial correlation; Causal model

A landslide susceptibility map (LSM) is the basis of hazard and risk assessment, guiding land planning utilization, early warning disaster, etc. Researchers are often overly keen on hybridizing state-of-the-art models or exploring new mathematical to improve accuracy in terms a receiver operator characteristic curve. Correlation analysis causal factors necessary routine process before modeling ensure that overall correlation among all low. However, this insufficient detect high local factor classes. The objective study answer three questions: 1) Is there between some parts locally? 2) Does it affect assessment? 3) How can influence be eliminated? To aim, Wanzhou County was taken as test site, where assessment based 12 has been previously performed using frequency ratio (FR) model random forest (RF) model. In work, we conducted spatial “altitude” “rivers” found sizeable overlap altitude-class-1 rivers-class-1. were reclassified, then FR RF used reevaluate analyze loss caused by two factors. results demonstrated LSMs markedly enhanced after reclassification “rivers,” especially for model–based LSM. This research shed light arising from particular geomorphology their impact susceptibility.