Screening artefacts in infrared measurements
Applicability of infrared limb sounding is limited by high-reaching clouds. Since cloud signal in the spectra causes retrieval artefacts, cloud-contaminated spectra are rigorously discarded. Thus, measurements do no longer represent the atmosphere with its entire variability but only the cloud-free atmosphere. This causes a screening bias in averaged data. The figure shows the estimated H2O cloud screening bias for MIPAS.
One could argue that less rigorous cloud screening would reduce the screening bias. The price to pay, however, would be a retrieval bias, which is considered much more harmful and less controllable. Contrary to the screening bias, a retrieval bias would affect also analysis of single (non-averaged) data. The effect of the screening bias in model-observation comparisons can easily be remedied by emulating the cloud screening in the model data as well: Only data of modeled cloud-free scenes should be used for calculation of the mean values.
For futher details of the analysis method, see https://doi.org/10.5194/acp-18-4187-2018.