Thermal imaging for input to terrestrial and planetary thermal models (Invited)

Multispectral imaging of emitted thermal radiation is used to estimate surface composition, especially of silicate minerals, but the most common use is to estimate surface temperature T. Thermal modeling requires accurate estimation of T, which for many terrestrial and nearly all planetary studies requires measuring radiance L remotely. The key difficulty is that thermal imaging is underdetermined, with 3 atmospheric parameters and n+1 surface parameters where n is the number of channels (i, j...), even if adjacency effects and anisothermal pixels are overlooked. The n unknowns are emissivities ɛ (i), and if T is to be estimated something about the ɛ spectrum must generally be known or assumed. Many different algorithms have been devised. If n=1, ɛ is assumed from laboratory data; if n=2, T for water is proportional to L(i) and L(j)-L(i) (split-window technique), empirically calibrated. One of the benefits of multispectral imaging is that ɛ(i) can be estimated pixel by pixel, such that T can be recovered with more confidence. Multispectral and hyperspectral data are commonly handled by assuming the maximum value of ɛ. Under some conditions, two-time imaging can be used assuming that ɛ(i) is unchanging. Generally, terrestrial atmospheric corrections, instrumental calibration and ɛ assumptions contribute roughly equally to T inaccuracy of 1-2 K for high spatial resolution data, although for low resolution, restricted atmospheric conditions, and known surface composition this figure can be improved. Peak L occurs at increasing wavenumber as T rises. Thus, T recovery for active volcanoes makes use of midwave (2000-3300 cm^-1) or shortwave (3300-10,000 cm^-1) rather than longwave thermal infrared (700-1250 cm^-1). One complexity is atmospheric variability near active vents; another is the possibility that ɛ may change as lava cools and the crystallinity of the skin changes (Abtahi et al., 2002). At the other extreme Mars and the icy satellites have low T and therefore lower SNR than warmer targets. This especially is true at night, for example on Mars. Thermal inertia, a measure of the resistance of a surface to changing its T as energy is added or subtracted to it, can be estimated from day-night temperature differences. It can in principle be used to learn something about the porosity, rock/soil ratios, the presence of thin veneers of sand, or other non-compositional characteristics of a surface. Quantitative measures of T, ɛ and thermal inertia are needed for thermal modeling. However, calculating thermal inertia requires accounting for topography and albedo and is more challenging than just estimating T, and therefore on both Earth and Mars approximations to it are commonly used photointerpretively, just as images of T images and even derived ɛ are sometimes used photointerpretively also.