Please note that all figures and tables link to the full power point presentation!!
AGU Conference, San Fransisco, 2002
Kuhn and Michael S. Ramsey
The Advanced Spaceborne Thermal Emission Reflectance Radiometer (ASTER) is the only high resolution multispectral thermal infrared (TIR) imager currently in Earth- orbit. The data being returned are ideal for discerning physical variations on the surface of active lava domes such as Soufriére Hills, Montserrat. The 90 meter spatial resolution of the TIR data provides accurate measurements of the surface temperature and emissivity. This information can be used to map the glass, vesicle and petrological distribution on the dome’s surface, and therefore provide fundamental insights into lava emplacement processes. For this study, six nighttime ASTER scenes of the dome and surrounding region captured over the past two years were chosen. These images show the entire dome, are relatively cloud-free and have significant thermal anomalies present, including summit and pyroclastic flow deposits. In order to validate the image data, detailed field-based information will be collected including temperature, GPS and spectroscopic data (similar to the ASTER spectral band passes). In addition, a searchable database of activity based on Montserrat Volcano Observatory (MVO) reports has been created and used as a framework for the image data.
Field measurements will be taken in conjunction with a nighttime overpass of ASTER. Active areas were determined both visually and with the aid of high resolution radiometers. The heightened activity continues to preclude close field measurements; therefore, long range laser profiling and GPS will be used to locate anomalies on the dome surface. A CIMEL spectro-radiometer with identical TIR wavelengths to ASTER and a FRIR will be implemented to map surface temperature and spectral variations. Numerous target areas are chosen including spines, lobes, incandescence, and fresh pyroclastic deposits. Samples of the most recent pyroclastic deposits will be collected and analyzed for mineral, phenocryst, and vesicle content using both petrographic and infrared analyses. This will be the most detailed suite of thermal infrared data collected at an active silicic dome and will provide a more complete understanding of the capabilities of remotely acquired TIR data to accurately describe active dome processes. This approach can easily be applied to other active areas that have the potential of transitioning from effusive to explosive dome growth, subsequent collapse and ensuing pyroclastic activity.
An understanding of extrusion behavior, particularly an adequate degassing of volatiles, can provide information on the triggers of explosivity. Variations in surface texture between two end-members of obsidian and vesicles, specifically provides insight of emplacement time or extrusion rate, volatile content, and internal structure of the dome. The preliminary work for this study was done by Ramsey and Fink (1999) on Little Glass Mountain, a rhyolitic flow near the summit caldera of Mount Shasta, in northern California, that formed 1.1 ka (Heiken 1978; Donnelly-Nolan et al. 1990). Vesicle percentage can be estimated using a sub-pixel analysis with two end-members, obsidian and blackbody. Highly ordered silicates exhibit spectral features that are muted or lost with amorphous glasses. In addition, the absorption feature from 7-11 µm is widened in obsidian. Vesicles, on the other hand, are concave voids, much greater in size than the emitted wavelength. This allows for a photon to interact with the surface multiple times, with absorption occurring each time in concordance to reflection, and an increase the emissivity at the absorption band. Presuming that emission spectra combine linearly in the thermal infrared wavelengths, aerial percentages of obsidian and vesicles can be estimated using a linear spectral unmixing technique. Results are favorable when the end-members are known. The active state of the Soufriére Hills dome, however, provides more complicating factors.
Remote sensing has fast become an important tool used for many applications from urban planning and development to sea surface temperature modeling. One application that is not as geographically pervasive, but ever increasingly important, is monitoring of active volcanoes. Location priority is not just given to volcanoes with a surrounding high population density, but also to those that are explosively eruptive and are currently in an active phase. Although Soufriére Hills, located in the Lesser Antilles on the island of Montserrat, is monitored heavily by the Montserrat Volcano Observatory, or MVO, remote sensing is still a critical method. Information using the ASTER satellite can be acquired by using bands from visible near infra-red to thermal wavelengths and beyond, but this paper will focus on one thermal capability only.
Soufriére Hills, an active andesitic dome, has undergone three different eruptive phases, since reactivation in 1992. The first began in 1995 and continued through 1998, with cycles of growth and collapse. Early 1998 to late 1999, the duration of the second phase, is characterized by no extrusion, but dome collapse and small explosions. The third phase, from late 1999 to present, shows renewed extrusion, with two major collapses (Montserrat Volcano Observatory, 2002). Eventually, the knowledge gained from remotely monitoring Soufriére Hills will be applied to other locations as a hazard assessment and mitigation tool.
IV. Data Sets and Methods
Six nighttime thermal infrared ASTER images spanning 2000 to 2002 where used in this portion of the study. The ASTER science team applies atmospheric correction algorithms to L1B datasets, which are available online as the same identification number preceded by a L2_09 tag. Band scale factors, from the metadata file, were applied to three of the images that capture a thermal anomaly within English’s Crater: #’s 1, 3, and 4. (Data was not available for geo-referencing purposes; therefore a topographic map was merged onto each image, with the crater and coastline outlined.) A temperature/emissivity separation using the emissivity normalization technique with an assumed value of 0.985, was applied to each image subset (1a, 3a, 4a), and warped onto the DEM. A density slice was applied to the temperature image (1b, 3b, 4b), to show the temperature distribution on the active dome. A cluster of four pixels were chosen based on the hottest temperatures. Pixel 1 always represents the hottest temperature and pixel 2 always represents the next hottest. The cluster is outlined in each image (1a, b, and c, 3a, b, and c, and 4a, b, and c). A z-profile, or spectra, was plotted using the emissivity image (1c, 3c, and 4c), which is enhanced with a gaussian stretch for visualization.
Lastly, each emissivity image was spectrally subsetted to include bands 11, 12, 13, and 14. Band 10 was removed to reduce water absorption features due to humid environments. A linear spectral unmixing technique was applied to each of the emissivity images (1c, 3c, and 4c), using two input spectra (figure 4) as end-members. Table 1 illustrates the aerial percentage results for each of the four pixels in the cluster.
All data processing was performed using the Environment for Visualizing Images (ENVI) software package by Research Systems Inc. (RSI).
Results for the linear spectral unmixing test are presented in Table 1. For each pixel in each cluster, aerial percentage of obsidian and blackbody are reported along with the RMS error. The specific locations in bold show promising results: the totals near 100 percent. The remaining location results show negative percent cover of obsidian and blackbody aerial percentage totaling greater than 100 percent. These results, however, are the best mathematical using the two specified end-members. Results were validated using a MATLAB program developed for an automated spectral deconvolution algorithm using blind end-members (Zorn and Ramsey, 2002), but with the known library. The active state of the Soufriére Hills dome perhaps presents complicating factors to the linear spectral unmixing technique. A third end-member may exist, or the pixel spectra are conceivably mutated by gasses or other activity. The results may also be attributed tosmall scale surface roughness. Future work is required to field-validate the deconvolution results.
Donnelly-Nolan JM, Champion DE, Miller CD, Grove TL, Trimble DA (1990) Post-11,000-year volcanism at Medicine Lake Volcano, Cascade Range, California. J Geophys Res 95:19693- 19704
Heiken G (1978) Plinean-type eruption at the Medicine Lake Highland, California, and the nature of the underlying magma. J Volcanol Geotherm Res 4:375-402
Keary, P and Vine, FJ (1996) Global techtonics. Blackwell Science, Cambridge, 333 p.
Ramsey, MS and Fink, JH (1999) Estimating silicic lava vesicularity with thermal remote sensing: a new technique for volcanic mapping and monitoring. Bull Volcanol., 61:32-39
Zorn, NV and Ramsey, MS (2002) An
automated spectral deconvolution algorithm: Application to thermal infrared
studies of Earth and Mars. Solar System Remote Sensing, LPI Contribution
No. 1129, p. 93-94