Evaluation of landslide susceptibility mapping techniques using lidar-derived conditioning factors (Oregon case study). Issue 6 (1st November 2016)

• Record Type: Journal Article
• Authors: Mahalingam, Rubini; Olsen, Michael J.; O'Banion, Matt S.
• Is Part Of: Geomatics, natural hazards & risk. Volume 7:Issue 6(2016)
• Page Start: 1884
• Page End: 1907
• Publication Date: 2016-11-01
• Subjects: Landslides -- LiDAR -- Oregon -- statistical techniques -- landslide inventory; Geomatics -- Periodicals; Geomatics; Periodicals; 526.905
• DOI: 10.1080/19475705.2016.1172520
• Languages: English
• ISSNs: 1947-5705

Abstract

ABSTRACT: Landslides are a significant geohazard, which frequently result in significant human, infrastructure, and economic losses. Landslide susceptibility mapping using GIS and remote sensing can help communities prepare for these damaging events. Current mapping efforts utilize a wide variety of techniques and consider multiple factors. Unfortunately, each study is relatively independent of others in the applied technique and factors considered, resulting in inconsistencies. Further, input data quality often varies in terms of source, data collection, and generation, leading to uncertainty. This paper investigates if lidar-derived data-sets (slope, slope roughness, terrain roughness, stream power index, and compound topographic index) can be used for predictive mapping without other landslide conditioning factors. This paper also assesses the differences in landslide susceptibility mapping using several, widely used statistical techniques. Landslide susceptibility maps were produced from the aforementioned lidar-derived data-sets for a small study area in Oregon using six representative statistical techniques. Most notably, results show that only a few factors were necessary to produce satisfactory maps with high predictive capability (area under the curve >0.7). The sole use of lidar digital elevation models and their derivatives can be used for landslide mapping using most statistical techniques without requiring additional detailed data-sets that are often difficult to obtain or of lower quality.