Correcting Mars Global Surveyor Thermal Emission Spectrometer (MGS-TES) High Altitude (40 - 65 km) Temperature Retrievals for Instrumental Correlated Noise and Biases

Mars Global Surveyor Thermal Emission Spectrometer (MGS-TES) nadir-soundings have been used to derive atmospheric temperatures up to roughly 40 km [Conrath et al., JGR 105 2000, Smith et al., JGR 106, 2001], and MGS-TES limb soundings have been used to extend the atmospheric temperature data set to > 60 km in altitude [Smith et al., JGR 106, 2001]. The ~40 - ~65 km altitude range probed by the MGS-TES limb sounding is particularly important for capturing key dynamical features such as the warm winter polar mesosphere [e.g., Smith et al., JGR 106, 2001; McCleese et al., Nature Geoscience 1, 2008], and the response of thermal tides to dust opacity [e.g. Wilson and Hamilton, J. Atmos. Sci. 53, 1996]. Thus accurate and precise temperature profiles at these altitudes are particularly important for constraining global circulation models. They are also critical for interpreting observations of mesospheric condensate aerosols [e.g., Määttänen et al., Icarus 209, 2010; McConnochie et al., Icarus 210, 2010)]. We have indentified correlated noise components in the MGS-TES limb sounding radiances that propagate into very large uncertainties in the retrieved temperatures. We have also identified a slowly varying radiance bias in the limb sounding radiances. Note that the nadir-sounding-based MGS-TES atmospheric temperatures currently available from the Planetary Data System are not affected by either of these issues. These two issues affect the existing MGS-TES limb sounding temperature data set are as follows: Considering, for example, the 1.5 Pascal pressure level (which typically falls between 50 and 60 km altitude), correlated-noise induced standard errors for individual limb-sounding temperature retrievals were 3 - 5 K in Mars Year 24, rising to 5 - 15 K in Mars Year 25 and 10 - 15 K in Mars Year 26 and 27. The radiance bias, although consistent on ~10-sol time scales, is highly variable over the course of the MGS-TES mission. It results in temperatures (at the 1.5 Pascal level) biased low by as much as 7 K in some periods and biased high by as much as 4 K in other periods, although typical temperature biases are less than 2 K in magnitude and the average bias over the course of the mission is near zero. We have developed techniques that essentially eliminate the variable radiance bias and the correlated noise problems, resulting in a dramatically more precise and accurate high altitude temperature data set. We expect, for example, to reduce temperature uncertainties at the 1.5 Pascal level to ~1 - ~2 K. The radiance bias is addressed simply by subtracting a running average space-pointed spectra. The correlated-noise is addressed by identifying the background noise covariance matrix from the space-pointed spectra, and then performing the retrievals in a basis that diagonalizes this matrix. We will present comparisons of this improved data set with the previous version, and with the Mars Climate Sounder data set [e.g., Kleinböhl et al., JGR 114, 2009].