Hello. A question arised while discussing with some colleagues about wether or not different normalization schemes for the reference spectrum would affect the analysis.
More specifically, if our reference spectrum is constructed by averaging several zenith spectra together, would it matter for QDOAS if the radiances are merely added together or averaged, or even renormalized by some scale factor?
Also, since our reference file is saved in plain ASCII format, it is not clear how could we let QDOAS know about the exposure time and the number of scans in this case.
Would the answers change if our measured spectra is loaded in mfd-heilderber, mcf-bira or netCDF format?
Any other advice is kindly appreciated.
Best regards,
Nicolás Casaballe
any normalization factor N of the reference spectrum would show up as an additive constant factor log(N) in the fitted optical depth log(spectrum/reference). Such a constant factor would be absorbed by the constant term of the polynomial which is normally included in the fit (qdoas can be configured to fit without any polynomial term, but I don’t think that will work well in practice). A normalization factor would therefore only have an impact on the fitted “slant column” of the constant term of the polynomial, which is rarely used in applications I know of. EDIT: see correction below
I do not believe qdoas accounts for exposure time or number of scans included in a reference in the analysis (because we fit an optical depth polynomial, an absolute radiometric calibration is not really needed), and the chosen input file format does not matter.
Small correction/addition after checking with my colleague: in fact, QDOAS normalizes the reference and the spectrum which is analysed before the fit, so any normalization factor will not even affect the fitted polynomial…