Investigating the Impact of Modern Absorption Cross-Section Databases on CO2 Retrievals

Current spectrometers provide high-quality absorption spectra from both ground-based direct sun measurements and spaceborne backscatter measurements. Accurate retrievals of atmospheric CO₂ concentrations from these measured spectra are fundamental for modelling large-scale atmosphere-surface exchange fluxes. When retrieving CO₂ concentrations from measured spectra, high-quality spectroscopic reference data are essential to drive radiative transfer simulations and to enable accurate retrievals. Here, we investigate how various modern molecular absorption cross-section datasets affect CO₂ retrievals in the 1.6 μm and 2 μm wavelength regions. This includes recent parameter sets derived from laboratory measurements at the German Aerospace Center (DLR e.V.) for line-mixing parameterizations (Birk et al., 2024) with separate continuum data. We compare these new data to those from HITRAN 2020 (Gordon et al., 2022) with and without speed-dependent Voigt profile extension as well as to the ABSCO tables (Benner et al., 2016), (Devi et al., 2016).

To evaluate the quality of the spectroscopic databases, we submit high-resolution direct-sun spectra collected by the TCCON (Total Carbon Column Observing Network) spectrometer at Karlsruhe to our RemoTeC retrieval algorithm under variation of the driving spectroscopic parameters. We evaluate systematic spectral residuals as well as spurious dependencies of the retrieved CO₂ columns on slant airmass. We further retrieve CO₂ separately from the P- and R-branches within a spectral window to assess potential mismatches. In addition, we use one year of GOSAT satellite measurements to investigate whether and how differences in CO₂ concentrations retrieved under variation of the spectroscopic parameters show dependencies on geophysical parameters such as latitude, season or surface type. Our analyses show that the DLR cross-sections lead to noticeable improvements in spectral line modelling in the strong 2 µm band, which in turn reduces airmass dependencies. Including the CO₂ continuum from the DLR dataset further reduces airmass-dependent biases, although this improvement is not reflected in the spectral residuals. Using the ABSCO tables results in residuals comparable to those using the HITRAN 2020 cross-sections. However, the airmass bias is low and comparable to the DLR cross-sections. In the weaker 1.6 µm bands, fit quality is comparable across all datasets, with small differences in airmass dependence. For the 2 μm band, the comparison of P- and R-branch retrievals reveals differences of up to 0.2 % and a pronounced airmass-dependent bias in the HITRAN 2020 R-branch. This inter-branch difference vanishes in the 1.6 μm bands when using DLR cross-sections but persists for HITRAN. Furthermore, retrievals with DLR cross-sections show significantly improved agreement between the 2 µm and 1.6 μm bands compared to HITRAN. In the GOSAT analysis, in addition to airmass-dependent effects leading to latitudinal and seasonal biases, we found surface albedo to strongly correlate with differences in retrieved CO₂ concentrations.

Benner et al., 2016: https://doi.org/10.1016/j.jms.2016.02.012

Devi et al., 2016: https://doi.org/10.1016/j.jqsrt.2015.12.020

Birk et al., 2024: https://elib.dlr.de/208834/

Gordon et al., 2022: https://doi.org/10.1016/j.jqsrt.2021.107949