Cordilleran Section - 99th Annual (April 1–3, 2003)

Paper No. 8
Presentation Time: 11:05 AM

REVISED MAPPING OF THE LA REFORMA CALDERA BASED ON REMOTE SENSING WITH MASTER AIRBORNE MULTI-SPECTRAL DATA


DMOCHOWSKI, Jane Ellen1, STOCK, Joann M.1, HOOK, Simon J.2 and HAUSBACK, Brian P.3, (1)Dept. of Geological and Planetary Sciences, California Institute of Technology, 1200 E. California, Pasadena, CA 91125, (2)Jet Propulsion Lab, Mail stop 183-501, Pasadena, CA 91109, (3)Department of Geology, California State Univ of Sacramento, Sacramento, CA 95819, jane@gps.caltech.edu

Multispectral airborne data from the MODIS/ASTER simulator (MASTER) in conjunction with field mapping and spectroscopy were used to produce weight percent silica maps and image classification maps of the La Reforma caldera located east of the Tres Virgenes Volcanoes in Baja California (27.5°N). These data were then used to revise existing maps of the region. MASTER acquires data in 50 spectral channels in the visible near infrared (VNIR), short wave infrared (SWIR) and thermal infrared (TIR) parts of the electromagnetic spectrum. The image classifications and weight percent silica maps were extracted from the VNIR-SWIR and TIR channels respectively. The TIR extraction involved atmospherically correcting the data and performing a temperature-emissivity separation. Atmospheric correction was undertaken using a radiative transfer code (MODTRAN 3.5) driven with a model atmospheric profile obtained from the National Center for Environmental Prediction (NCEP). The emissivity information was extracted from the atmospherically corrected data using the TES algorithm. The emissivity images were then used with chemical analyses from field samples to produce wt. % SiO2 in the regions. The VNIR-SWIR data were atmospherically corrected with the ACORN algorithm to surface reflectance. The reflectance data were used to identify endmember spectra, utilizing the Minimum Noise Transformation, the Pixel Purity Index, and various other classification techniques. Endmembers determined by this process were compared to field spectral data from regional validation sites for petrological identification and radiometric accuracy. Confusion matrices of supervised classifications based on the USGS spectral library yield ~40% overall accuracies, using five defined ground truth classes. Classifications based on the ground truth classes themselves yield higher overall accuracies. The resulting classifications and maps of wt. % SiO2 were used in concert with field investigations to revise mapping of the La Reforma caldera detailing the intracaldera sequence as welded ash-flow tuffs and lavas. Additionally, some units previously mapped as Quaternary mafic products are updated to be more silicic units.