Evaluating Mars Atmosphere Nadir-Sounding Temperature Retrieval Methods

Abstract:

Motivated by the prospect of future TIR instruments in Mars orbit, including the recently launched Emirates Mars Infrared Spectrometer (EMIRS) on the Emirates Mars Mission (EMM) [Altunaiji et al. 2017, 6th International Workshop on the Mars Atmosphere, # 4413], we have experimented with using general circulation model (GCM) output as an artificial “ground-truth” to evaluate retrieval algorithms. This has led us to evaluate the extent to which the existing Mars Global Surveyor’s Thermal Emission Spectrometer (TES) data set is capable of capturing expected-by-GCM temperature structures, and to assess whether a different approach might have better accuracy and/or precision.

Our artificial ground truth comes from the Mars Climate Database (MCD) version 5.3 [Forget et al. 1999, JGR, 104, 24155-24176; Millour et al. 2018, Scientific Workshop: “From Mars Express to ExoMars”]. The MCD GCM output is turned into radiance using a multiple-scattering discrete ordinates radiative transfer model. We add simulated noise based on TES views of empty space, with the noise amplitude and the relative contribution of spectrally-correlated noise components both adjustable to allow consideration of other instruments. The retrieval algorithms that we have considered to-date include: the original TES temperature retrieval, a “modified original TES” with a much weaker second-order regularization constraint, and a completely new retrieval based on a discrete ordinates multiple scattering forward model and Marquardt-Levenberg optimization with a smoothness constraint.

We have found that the original TES temperature retrieval works well except for: (1) in vicinity of the northern winter polar jet, because the regularization constraint prevents the retrieved profile from accommodating the strong curvature (i.e. strongly varying lapse rate) in the artificial ground truth of the GCM; (2) at high latitudes during dust storm conditions, apparently because of the same curvature problem rather than due to aerosol opacity effects. Both the “modified original TES” and new Marquardt-Levenberg based retrieval give improved results for both problem areas, and we are currently working to evaluate which approach gives the best combination of accuracy and precision in the presence of realistic instrumental noise.