A Strategy for Atmospheric Correction

Introduction: The Thermal Emission Imaging System (THEMIS) instrument on the Mars Odyssey spacecraft [1] is currently acquiring infrared (IR) multispectral images of Mars. The instrument has a spatial resolution of 100 meters/pixel in the IR, and acquires data in 9 different wavelength bands, 8 of which are sensitive to radiance emitted by the surface. A complication in the analysis of such data is the separation of the total radiance reaching the detector into surface, atmospheric, and instrument background components, which is necessary for an accurate interpretation. A method for determining the multiplicative component of the atmospheric abosrption is presented: this method depends on the accurate determination of the additive component via the radiance offset method [2]. The multiplicative correction is applied to a number of THEMIS images in a systematic survey of how environmental variables (e.g. surface temperature, elevation, relief) affect the quality of the method. Emissivities derived from the corrected images are compared to Thermal Emission Spectrometer (TES) emissivity values as an additional constraint on the quality of the correction. The final result will be a database of corrected image parameters, which can then be used to identify the limitations of this particular method. Background: The radiance values recorded by THEMIS can be considered as a sum of radiances from various sources: surface emission that passes through the atmosphere, atmospheric radiation (" upwelling "), atmospheric radiation reflected off of the surface (" downwelling "), atmospheric scattering, and instrument effects. In order to separate the surface from the atmospheric and instrument components, it is necessary to make some simplifications. First, it is assumed that a majority of the instrument effects have been removed from processed THEMIS cubes. Second, the atmosphere is assumed to be laterally and vertically homogeneous over the image to be corrected: this assumption is more likely to be valid over a small image than a large one, so subsets of full THEMIS strips will produce more reasonable corrections than the full strips will. The validity of this assumption is also affected by cloud cover within the image. The last assumption is that the downwelling radiance dominates the additive atmospheric component. These simplifications can be represented in terms of blackbody radiation (BB), surface reflectivity (R), the surface emissivity (!), and the atmospheric absorption (!atm) as: Radiance = Surface radiance + downwelling + b Rad = !atm·!·BBsurf + !atm·R·BBatm+b