TUNER-compliant error estimation for MIPAS: methodology

Abstract

This paper describes the error estimation for temperature and trace gas mixing ratios retrieved from the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS) limb emission spectra. The following error sources are taken into account: measurement noise, propagated temperature and pointing noise, uncertainties in the abundances of spectrally interfering species, instrument line shape errors, and spectroscopic data uncertainties in terms of line intensities and broadening coefficients. Furthermore, both the direct impact of volatile and persistent gain calibration uncertainties, offset calibration, and spectral calibration uncertainties, as well as their impact through propagated calibration-related temperature and pointing uncertainties, are considered. An error source specific to the MIPAS upper atmospheric observation mode is the propagation of the smoothing error crosstalk of the combined NO and temperature retrieval. Whenever non-local thermodynamic equilibrium modelling is used in the retrieval, related kinetic constants and mixing ratios of species involved in the modelling of populations of excitational states also contribute to the error budget. Both generalized Gaussian error propagation and perturbation studies are used to estimate the error components. Error correlations are taken into account. Estimated uncertainties are provided for a multitude of atmospheric conditions. Some error sources were found to contribute both to the random and the systematic component of the total estimated error. The sequential nature of the MIPAS retrievals gives rise to entangled errors. These are caused by error sources that affect the uncertainty in the final data product via multiple pathways, i.e., on the one hand, directly, and, on the other hand, via errors caused in a preceding retrieval step. These errors tend to partly compensate for each other. The hard-to-quantify effect of the horizontally non-homogeneous atmosphere and unknown error correlations of spectroscopic data are considered to be the major limitations of the MIPAS error estimation. © Author(s) 2022.