Browse/search for people

Publication - Dr Matthew Watson

    Thermal deconvolution

    Accurate retrieval of multispectral infrared emissivity from thermally-mixed volcanic surfaces


    Rose, SR, Watson, IM, Ramsey, MS & Hughes, CG, 2014, ‘Thermal deconvolution: Accurate retrieval of multispectral infrared emissivity from thermally-mixed volcanic surfaces’. Remote Sensing of Environment, vol 140., pp. 690-703


    The thermal infrared (TIR) wavelength region has proved highly useful for remotely extracting important parameters of volcanic activity, such as the composition, texture, and temperature of either the surface or gas/aerosol emissions. However, each of these characteristics can vary within the area of one pixel of a remote sensing dataset, which ultimately affects the accuracy of the retrieval of these characteristics. For example, where multiple temperatures occur in a particular pixel, the derived emissivity spectrum and pixel-integrated brightness temperature for that pixel are inaccurate. We present a new approach for deconvolving thermally-mixed pixels in a day/night pair of the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) thermal infrared (TIR) scenes over Kilauea volcano, acquired during an active effusive phase in October 2006. The thermal deconvolution algorithm identifies thermally-mixed pixels and determines the multiple temperature components and their area, using data from the higher spatial resolution short wave infrared (SWIR) channels of ASTER. The effects of thermal mixing on the emissivity retrievals were quantified using a spectral deconvolution approach comparing the original to the thermally deconvolved data. The root mean squared (RMS) error improved slightly from 0.879 to 0.813, whereas the compositional end-members changed more dramatically (e.g., glass decreased from 70.2% to 49.3% and the vesicularity increased from 0.7% to 16.3%). The results provide more accurate temperature and emissivity data derived from ASTER data over thermally-elevated surfaces such as volcanoes and fires. This approach also serves as rapid means for accurately identifying sub-pixel temperatures, commonly obscured in low to medium spatial resolution orbital datasets. Moreover, it minimizes processing time, allowing critical information to be quickly disseminated. © 2013.

    Full details in the University publications repository