HR: 1330h
AN: S12B-0393 [PDF]
TI: Hyperspectral Analysis of Paleoseismic Trench Stratigraphy: Toward Improving the Recognition and
Documentation of Past Earthquakes
AU: * Ragona, D E
EM: dragona@ucsd.edu
AF: IGPP, Scripps Institution of Oceanography, 8765 Biological Grade, La Jolla, CA 92093 United States
AU: MInster, B
EM: jbminster@ucsd.edu
AF: IGPP, Scripps Institution of Oceanography, 8765 Biological Grade, La Jolla, CA 92093 United States
AU: Fialko, Y
EM: yfialko@ucsd.edu
AF: IGPP, Scripps Institution of Oceanography, 8765 Biological Grade, La Jolla, CA 92093 United States
AU: Rockwell, T
EM: trockwel@cox.net
AF: San Diego State University, 5500 Campanile Dr., San Diego, CA 92182
AB:
We are conducting a pilot project to use hyperspectral imagery to assist in the recognition and documentation of paleoseismic
events in trench exposures. Recent advances in hyperspectral imagery suggest that stratigraphy can be analyzed in much the
same way as Aviris imagery of Earth's surface. In principle, hyperspectral images may be able to elucidate and record
otherwise-poor stratigraphy in some exposures, thereby improving the information that can be gleaned from a paleoseismic
site. This technique may also eliminate some problems in interpretation of the earthquake history of a site by illuminating
details of the stratigraphy and structure that are not apparent to the human eye, such as unique unit correlations across
complicated fault ruptures. The trench site chosen for this study is located at Hog Lake in the Anza seismic gap along the
San Jacinto Fault in southern California. The site was selected because of its detailed, well defined stratigraphy. The
method adopted was to obtain a 50 cm side square matrix of samples that could be used to generate a low-resolution image of
the sampled area, in the sense that each sample represents a single pixel. The samples were collected 2.5 cm apart in a
square matrix of 20x20 samples. Each of the 400 samples collected are stored into PVC or metallic cylinders of 3/4" or 1/2"
diameter. All samples were spectrally analyzed at JPL using a FieldSpec Pro instrument that measure radiation in the
350-2,500 nm wavelength window. Five measurements of each sample were performed, along with measurements of the radiation
reflected by a reference surface (Spectralon), under natural light and clear sky conditions. The data obtained was then
processed to obtain reflectance spectra for all samples. Principal Component Analysis was used to create a pixilated image
from the three dominant components. That image shows promising similarity with the standard digital picture of the sampled
trench wall. However, large random measurement errors created problems in some samples when we tried to separate classes of
different materials. We are currently working on the identification of the source of noise to either correct the data or to
improve the experimental procedure. Our current preliminary results show that the hyperspectral data can discriminate between
different lithologies for those samples that provide consistent spectrums with high signal to noise ratio Whole spectrum
comparisons yielded better results than single absorption features identification.
DE: 7221 Paleoseismology
SC: Seismology [S]
MN: 2003 Fall Meeting