Calculating radiant flux from thermally mixed pixels using a spectral library

Hot surfaces associated with volcanoes, wild fires and geothermal areas are often thermally heterogeneous with respect to the spatial resolution of satellite sensors. A single pixel temperature derived from a satellite image can therefore represent a continuum of surface temperatures that may vary by hundreds of degrees Celsius. For thermally mixed pixels it is therefore more meaningful to estimate radiant flux [Watts] and/or radiant exitance [Watts per metre squared]. Here we introduce a new method for calculating radiant flux from thermally heterogeneous surfaces with temperatures in the 100 to 1100 °C range. It involves modelling radiance spectra using a spectral library. Two spectral libraries were created to represent two different sensor configurations i) a VNIR-SWIR imaging spectrometer and ii) a two channel SWIR imager, both characterized by a 30 m spatial resolution. We compare our approach against that of the “dual-band method”. The spectral library approach was able to calculate radiant flux to within 30% of the actual value for targets radiating at or above 0.7 MW (i.e. when using an imaging spectrometer) or 7.1 MW (i.e. when using just two SWIR wavebands). The dual-band approach, on the other hand, required targets to be radiating at least 12 MW before a 30% accuracy level could be obtained. All of the approaches could accurately fit the spectral radiance values that they modelled. However, they could not reliably determine subpixel temperature distributions. This indicates that it might never be possible to retrieve subpixel temperature distributions reliably using short-wave infrared spectra alone. This finding has significant implications for the remote sensing of hot targets.