Characterisation of spectroscopic properties of DOAS instruments using high-resolution solar spectra

Abstract. The characterisation of spectroscopic properties of DOAS instruments is important for accurate trace gas retrievals. In this study, we investigate and extend existing methods for the determination of spectroscopic properties using high-resolution solar spectra (also known as Kurucz fit, KF, approaches). We apply these methods to long-term zenith sky DOAS measurements in Kiruna (northern Sweden). This unique data set allows to study the performance and precision of such fitting procedures under different environmental and observational conditions. Also, the effect of the change of the detector from a photodiode array to a modern CCD is investigated. One key finding of our study is that the so-called Ring effect (caused by rotational Raman scattering) leads to a systematic broadening (by typically about 10 %) of the width of the instrument spectral response function (ISRF) derived from a KF compared to the true ISRF derived from atomic line lamp measurements. Especially for measurements of trace gases located close to the ground this broadening can lead to errors of the trace gas results if a KF-derived ISRF is used for the preparation of trace gas reference spectra and Ring spectra. Here it is important that the strength of the Ring effect can strongly change due to clouds, in particular in the presence of optically thick clouds. Measurements in the presence of optically thick clouds should thus not be chosen for the application of the KF. Another specific finding for the Kiruna measurements (also relevant for other high latitude stations) is that the Ring effect changes systematically with season because of the changing surface albedo (caused by snow cover in the winter). From KF, different instrument properties can be obtained. We give specific recommendations for different KF variants for the determination of the ISRF, intensity offsets (e.g. caused by spectrograph straylight), or the wavelength dependence of the light throughput of the instrument. We also show that a strong wavelength dependence of the light throughput (e.g. caused by the Fabry–Pérot etalon effect) can lead to wrong trace gas results. This finding might also be relevant for other instruments affected by strong Fabry–Pérot etalon effects, or containing other optical elements with strong wavelength-dependent light throughputs. Finally, we introduce a method to correct such a wavelength-dependent light throughput using the results of a modified KF.