Wavelength calibration of spectra measured by
the Global Ozone Monitoring Experiment by use
of a high-resolution reference spectrum

Jos H.G.M. van Geffen and Roeland F. van Oss
Applied Optics  42, 2739-2753 (2003).

5. Concluding remarks

The Fast Delivery Service at KNMI provides ozone and cloud information based on measurements of the GOME instrument (aboard ERS-2) on a near-real time basis, i.e. within three hours after observation. The Fast Delivery Service includes a level 0-to-1 processor to derive spectral information (level-1 data) from the raw (level-0) data received at KNMI for this purpose. One of the steps in this processor is a wavelength calibration, which must be accurate enough to facilitate the retrieval of reliable ozone columns and profiles. To this end, a wavelength calibration method has been developed that uses a high-resolution solar spectrum as reference spectrum and performs a combination of shift and squeeze on the spectra to derive the best wavelength grid. The accuracy of the method is about 0.002 nm for wavelengths below about 290 nm, and 0.001 nm for higher wavelengths.

The result of the calibration, that is the change of the wavelength of a point in a given window with respect to an initial guess, appears to vary along an orbit and from orbit to orbit. This obviously limits the number of spectra that can be averaged before the wavelength calibration to improve the signal-to-noise ratio of the spectrum to be calibrated. This orbital variation is discussed in a forthcoming paper,[23] which also investigates the change in the calibration results of solar spectra over a period of six years.

Once daily GOME measures a solar spectrum, when it flies over the North Pole into the sunlight along one of its orbits. GOME then flies towards the Sun, hence there is a Doppler-shift in the wavelength of lambda v/c, where v is the satellite's velocity (about 8.3 km/s) and c the speed of light. This shift is between about 0.008 nm at lambda=300 nm and 0.022 nm at lambda=800 nm. The calibration method's fitting of a shift and a squeeze against the solar reference spectrum, automatically corrects for this Doppler-shift.

There are a few improvements possible on the calibration method. The accuracy of the reference spectrum below 300 nm, for example, could be improved upon. The reference spectrum is based on ground-based and balloon-based measurements of the solar spectrum, and it thus contains some atmospheric absorption features, over and above the distinct solar Fraunhofer lines. It is very desirable to have a pure high-resolution solar spectrum, e.g. measured from space, as reference for the calibration of measured solar spectra. A high-resolution earthshine spectrum as reference for measured earthshine spectra could also be useful. The use of earthshine reference spectrum, however, would introduce problems in determining how good the method works, because of the large variation in atmospheric absorption features in measured spectra. Alternatively, one could use a true solar reference spectrum and also use ozone cross sections in the fit when optimising for ozone retrieval, though the subsequent ozone retrieval itself may then no longer be independent of the wavelength calibration.

The accuracy of the method may further be improved by extending the calibration to also fit the resolution of the slit function used to convolve the high-resolution reference spectrum. This will make the method computationally slower, which is not acceptable within the constraint of near-real time delivery of the data of the Fast Delivery Service and was therefore not done. Furthermore, fitting the resolution is perhaps only useful if the slit function itself is known well, which is not the case for GOME.

The Fast Delivery Service is based on a set of nine wavelength windows plus the instrument health parameters, together known as the EGOI-data. The wavelength calibration method has been designed for these nine EGOI-windows. The method is, however, not restricted to these windows: it can be used with any set of wavelength windows within the available reference spectrum. The method is thus suitable for use with any high-resolution (ir)radiance spectrometer, such as the satellite instruments SCIAMACHY (aboard ENVISAT; launched in 2002), OMI (aboard EOS-Aura; to be launched in 2004), and GOME-2 (aboard METOP; to be launched in 2005).

The GOME level-1 spectra produced with the GOME Data Processor (GDP) have wavelength calibrations based on the spectrum of an onboard calibration lamp and a shift is applied to the spectra (no squeeze). The GDP-provided wavelength grid of the level-1 spectra appears to be insufficiently accurate for some level 1-to-2 retrievals, such as for ozone profiles. The accuracy of the DOAS-retrieval of e.g. ozone, BrO and NO2 will also benefit from a more accurate wavelength calibration. It is therefore useful to have the possibility to improve the wavelength calibration of GDP level-1 spectra within user specified wavelength windows. For the purpose of such a re-calibration, the GomeCal software package has been written and is made available on-line via http://www.knmi.nl/gome_fd/gomecal/. The GomeCal package also provides an improved polarisation correction as well as an additional correction for the degradation of the GOME instrument and a radiometric calibration.

<=== contents of this paper
===> references

<=== Publications page
<=== Post-doc. research at KNMI page

Jos van Geffen -- Home  |  Site Map  |  Contact Me

created: 11 February 2003
last modified: 3 February 2008