Often when performing DOAS retrievals, there may be a slight wavelength misalignment between the limb or nadir spectra, the reference, and the various cross-sections. This will lead to errors in the fitted column abundances. To correct for this, shift and stretch algorithms are employed. It is sometimes difficult to determine when and to what extent these corrections are necessary and so often automated shift and stretch algorithms are employed. These routines are designed to return values of zero shift and stretch if there is no misalignment. Wavelength shifts occur when one spectral quantity (spectra or cross-section) is shifted by a constant amount with respect to another. Wavelength stretches are similar to wavelength shifts but the amount of the shift increases or decreases linearly with respect to some reference wavelength. Higher order stretches are of lesser importance as misalignments are usually quite small (<1 nm) and so any variation with wavelength is well approximated as linear.
The CPFM wavelength assignments are estimated to be accurate to 0.1 nm. As shifting of the nadir and limb relative to the horizontal flux by amounts in excess of 0.1 nm acted only to decrease the quality of the fits, this seems to be an accurate figure. Shifting by amounts less than 0.1 nm had nearly negligible impact on both the fit quality and the retrieved ACDs. This is due in large part to the fact that the CPFM instrument has a resolution of about 1 nm and all cross-sections are degraded to this using the instrument function. In general, shifting of the cross-sections did not improve the spectral fitting. The sole exception to this were the H2O cross-sections which contributed to the measured spectra only for wavelengths longward of 500 nm. Optimum fits were obtained by shifting the cross-sections 0.4 nm towards the blue. In practice, these shifts were performed by adjusting the wavelength grid and then interpolating the spectra or cross-sections back onto the original wavelength grid using a cubic spline. Wavelength stretches were not considered but it appears that they are not necessary (C.T. McElroy, personal communication, 1998).