Record Details
Discriminating between landslide sites and potentially unstable terrain using topographic indices
ScholarsArchive at Oregon State University
| Field | Value |
|---|---|
| Title | Discriminating between landslide sites and potentially unstable terrain using topographic indices |
| Names |
Appt, Jeremy
(creator) Skaugest, Arne (advisor) |
| Date Issued | 2002-07-15 (iso8601) |
| Note | Graduation date: 2003 |
| Abstract | A landslide inventory, statistical analyses and a Geographic Information System (GIS) are used to analyze landslide sites and potentially unstable terrain in the Oregon Coast Range. The objectives are to evaluate the efficacy of locating landslide sites with topographic variables and discriminate the difference between sites where landslides have and have not occurred. The population of known landslides are characterized as up-slope, non-road related, and associated with 1996 storm events. Topographic variables are derived from a Digital Elevation Model (DEM) for index construction forming six groups; i) slopes, ii) contributing areas, iii) ratios of slope and contributing area, iv) curvature v) infinite slope models, and vi) functions of slope and contributing area based on statistical models. Index groups employ different algorithms. Index performance is measured with landslide and aerial densities. Cumulative landslide occurrence is plotted against cumulative area on a continuous domain of the index to locate a maximum landslide density on equal size areas. Indices are used to generate model definitions of potentially unstable terrain based on similarity to the landslide population. Aerial densities of potentially unstable terrain based on index definitions are determined but no common metric is achieved. Statistical analyses on spatially stratified data suggest a significant (α < 0.05) difference between landslides sites and adjoined terrain. The minimum resolution at which a significant difference is achieved based on spatial stratification is a three cell radius surrounding the slide population. Curvature and area discriminate better than simple slope and topographic ratios. The relative performance is mostly a function of DEM error and resolution, and spatial correlation. Hydrologic geomorphic models perform about as well as the topographic ratios, and much less than the simple area index. There is no statistical evidence to suggest that the hydrologic geomorphic models accurately describe a threshold in the Mapleton slide population. The lack of significance is likely due to limitations on the available parameter sets. Logistic regression produced an index with the highest discrimination performance due to a maximum likelihood algorithm. Regression models have a physical basis in and parallel the behavior of linked hydrologic geomorphic and slope stability models. The measured differences in performance are a useful assessment of the DEM – index combination. |
| Genre | Thesis |
| Topic | Landslide hazard analysis -- Oregon |
| Identifier | http://hdl.handle.net/1957/9297 |
