Paper No. 7
Presentation Time: 8:00 AM-6:00 PM
REMOTE DATING OF LANDSLIDES BASED on LIDAR, ANALYSIS OF SOILS, AND STATISTICAL COMPARISON OF SURFACE MORPHOLOGY USING SPECTRAL ANALYSIS
DUPLANTIS, Serin1, HULBE, Christina L.
1, BURNS, Scott F.
1 and MADIN, Ian P.
2, (1)Department of Geology, Portland State University, PO Box 751, Portland, OR 97207, (2)Oregon Department of Geology and Mineral Industries, 800 NE Oregon St # 28, Suite 965, Portland, OR 97232, serin@pdx.edu
The aim of this study is to develop a method for dating landslides remotely in the hope that public safety officials will be able to implement this method to create more useful susceptibility maps. Light Detection and Ranging (LiDAR) and spectral analysis of Digital Elevation Models (DEM) are used to achieve this goal. Relative ages for landslides are based on dated Clackamas River terraces, a fluvial terrace soil chronosequence, and laboratory analysis of collected soil samples. Landslides in two Oregon quadrangles were identified visually using ArcGIS software and LiDAR bare-earth DEMs. A more detailed study of the landslide complexes in a smaller subset area was also conducted. Landslides were field verified and over 70 soil pits were excavated in representative areas within each landslide complex. Soil samples were collected from all soil horizons in each pit and textural descriptions, grain size analyses, and Atterberg Limit tests were completed for many of the soil samples.
Spectral analysis was used to investigate differences in surface morphology among failed surfaces of different ages. Both two-dimensional discrete Fourier transform and wavelet analysis were used to identify dominant wavelengths in selected “sample” areas of the LiDAR DEM over landslides identified in the field. Spectral signatures of the different surfaces were then compared statistically using a Kolomogrov-Smirnov test. The test identifies whether or not pairs of samples are from the same distribution, that is, do the surfaces from which the spectra were calculated have the same (the null hypothesis) or different spectral character? In almost every comparison made here, the statistical tests fail – that is, the samples all appear to be drawn from different frequency distributions – even on surfaces identified to be of similar age. However, it is clear that unfailed and failed surfaces are statistically different and failed surfaces begin to look more similar as they age.
