loading page

Spectra classification methodology for hyperspectral InfraRed imaging of Mt Etna volcanic plume with a radiative transfer retrieval model
  • Charlotte Segonne,
  • Nathalie Huret,
  • Sébastien Payan
Charlotte Segonne
LaMP / OPGC, Université Clermont Auvergne

Corresponding Author:charlotte.segonne@uca.fr

Author Profile
Nathalie Huret
LaMP / OPGC , Université Clermont Auvergne
Author Profile
Sébastien Payan
LATMOS / IPSL, Sorbonne Université, UVSQ, CNRS
Author Profile

Abstract

Quantification of sulfur dioxide (SO2) emission flux is a fundamental task in volcanology to have insights of the composition and the spatial evolution of volcanic plumes. The ground based InfraRed hyperspectral imager HyperCam, from Telops Company, was deployed during IMAGETNA campaign in 2015 and provided high spatial and spectral resolution images of Mt Etna plume. The spectral range of the hyperspectral imager is [7.7 - 11.8 µm] and the measured images contained 320 x 64 pixels with a spectral resolution of 2 cm-1. To process hyperspectral images in quasi real-time, a fast and reliable radiative transfer retrieval model is required. The LATMOS Atmospheric Retrieval Algorithm (LARA), used to retrieve the slant column densities of SO2, includes an accurate line-by-line radiative transfer model and an efficient minimization algorithm of the Levenberg-Marquardt type. But the calculation time remains too high to infer near real time (NRT) estimation of SO2 fluxes. As first, to reach NRT target, a classification methodology of the brightness temperature spectra was developed and then applied on each measured sequence to significantly decrease the processing time. One image previously took a week of calculation to be retrieved. The classification of the spectra, which takes a couple of hours, allows the retrieval of a complete measurement sequence of the field campaign (~400 images) in only two days. The accuracy of the methodology was confirmed, by comparing the SO2 slant column density images obtained after classification with the one obtained by the accurate and time expensive pixel by pixel retrieval processing. The LARA model, the spectra classification methodology and a comparison of the results will be presented.