MIPAS OFL REC Analyses
These are the Residual and Error Correlation (REC) Analyses for
monthly mean MIPAS Recalculated/Offline Level 2 Data distributed by ESA.
Residual spectra are the difference between spectra measured by the instrument and spectra generated by the retrieval forward model at the final iteration.
Ideally, these should contain only random measurement noise but in practice a number of features are present indicating systematic errors either in the forward model or the instrument characterisation.
Residual and Error Correlation (REC) analysis is a statistical technique for analysing such data. The priniciple is to identify correlations between persistent features in the residual spectra and the signatures expected from different atmospheric species, other potential sources of forward model error and calibration errors represented by various derivatives of the spectrum with respect to wavenumber. This is now performed routinely as part of the monitoring of MIPAS data quality.
See CalVal report
PO-TN-OXF-GS-0020.pdf
(REC Analysis of MIPAS Data Jul02-Feb03) for further explanation.
- Explanation of Plots
- Target Species
Spectral signatures of a large number of error sources are fitted simultaneously for each altitude and latitude band, but only those associated with pressure and the 6 retrieved species are plotted. Different colours/symbols indicate different latitude ranges (set by nominal OMs 001-006). x-axis is an approximate conversion of residual signature to VMR or % pressure error based on a mid-latitude day-time profile.
-
Black dashed line indicates +/- climatological 1sigma variability, dotted lines represent +/-10% and 100% of the profile value. Positive values indicate larger signature in atmospheric spectrum than forward model, indicating an underestimate of the `true' profile.
-
Apart from the basic plot for each species/month, there are also plots comparing current and previous month, residuals averaged separately for day/night conditions, and residuals averaged separately for `target' microwindows (eg fitting H2O signature in H2O microwindows only) and `other' microwindows (eg fitting H2O signature in all the non-H2O microwindows).
-
Spectral Derivatives 0th, 1st and 2nd derivative signatures are fitted to each microwindow/altitude independently. Colours indicate microwindow target species, symbols indicate altitude range. Microwindow labels are listed in order of increasing wavenumber along the top but the set of small arrows indicates actual position along the x-axis.
-
The y-axis represents a scaling of the signature in terms of some instrumental error. Following this analogy, positive y values indicate an underestimate of either the gain (0th derivative), wavenumber of atmospheric lines (1st derivative), or AILS width (2nd derivative), in the sense of regarding the atmospheric spectra as `true' values compared to the forward model. Only points representing large numbers of residuals are plotted.