Validating the accuracy of SO2 gas retrievals in the thermal infrared (8–14 μm)

• Published in: Bulletin of volcanology Vol. 79; no. 11; pp. 1 - 18
• Main Authors: Gabrieli, Andrea, Porter, John N., Wright, Robert, Lucey, Paul G.
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01-11-2017; Springer Nature B.V
• Subjects: Algorithms; Atmospheric water; Atmospheric water vapor; Brightness temperature; Cells; Coefficients; Earth and Environmental Science; Earth Sciences; Geology; Geophysics/Geodesy; Imaging techniques; Mathematical models; Meteorological satellites; Mineralogy; Modelling; Plumes; Public health; Radiance; Radiative transfer; Remote sensing; Research Article; Sedimentology; Sky radiance; Sulfur; Sulfur dioxide; Sulphur; Surface radiation temperature; Temperature; Volcanic ash; Volcanic plumes; Volcanoes; Volcanology; Water vapor; Water vapour; Ground-based retrieval; SO; Path-concentration; Inversion algorithm; PLSR; Hyperspectral imaging
• ISSN: 0258-8900; 1432-0819
• DOI: 10.1007/s00445-017-1163-3

Quantifying sulfur dioxide (SO
2
) in volcanic plumes is important for eruption predictions and public health. Ground-based remote sensing of spectral radiance of plumes contains information on the path-concentration of SO
2
. However, reliable inversion algorithms are needed to convert plume spectral radiance measurements into SO
2
path-concentrations. Various techniques have been used for this purpose. Recent approaches have employed thermal infrared (TIR) imaging between 8 μm and 14 μm to provide two-dimensional mapping of plume SO
2
path-concentration, using what might be described as “dual-view” techniques. In this case, the radiance (or its surrogate brightness temperature) is computed for portions of the image that correspond to the plume and compared with spectral radiance obtained for adjacent regions of the image that do not (i.e., “clear sky”). In this way, the contribution that the plume makes to the measured radiance can be isolated from the background atmospheric contribution, this residual signal being converted to an estimate of gas path-concentration via radiative transfer modeling. These dual-view approaches suffer from several issues, mainly the assumption of clear sky background conditions. At this time, the various inversion algorithms remain poorly validated. This paper makes two contributions. Firstly, it validates the aforementioned dual-view approaches, using hyperspectral TIR imaging data. Secondly, it introduces a new method to derive SO
2
path-concentrations, which allows for single point SO
2
path-concentration retrievals, suitable for hyperspectral imaging with clear or cloudy background conditions. The SO
2
amenable lookup table algorithm (SO
2
–ALTA) uses the MODTRAN5 radiative transfer model to compute radiance for a variety (millions) of plume and atmospheric conditions. Rather than searching this lookup table to find the best fit for each measured spectrum, the lookup table was used to train a partial least square regression (PLSR) model. The coefficients of this model are used to invert measured radiance spectra to path-concentration on a pixel-by-pixel basis. In order to validate the algorithms, TIR hyperspectral measurements were carried out by measuring sky radiance when looking through gas cells filled with known amounts of SO
2.
SO
2
–ALTA was also tested on retrieving SO
2
path-concentrations from the Kīlauea volcano, Hawai’i. For cloud-free conditions, all three techniques worked well. In cases where background clouds were present, then only SO
2
–ALTA was found to provide good results, but only under low atmospheric water vapor column amounts.