SCIAMACHY: Spectral Calibration in the SWIR Channels

Abstract

SCIAMACHY (SCanning Imaging Absorption spectroMeter for Atmospheric CHartographY) was a scanning nadir and limb spectrometer covering the wavelength range from 212 nm to 2386 nm in 8 channels. It doubled its originally specified in-orbit lifetime of five years before the com- munication to the ENVISAT platform failed in April 2012. We are now in the postprocessing phase F. The instrument was designed to measure column densities and vertical profiles of trace gas species in the mesosphere, in the stratosphere and in the troposphere. It can detect a large amount of atmospheric gases (e.g. O3, H2CO, CHOCHO, SO2, BrO, OClO, NO2, H2O, CO, CH4 , among others ) and can provide information about aerosols and clouds. The standard spectral calibration of the SCIAMACHY detectors is done with an on-board Spectral Line Source (SLS) by comparing theoretical line positions with line positions retrieved during calibration measurements. However, the used PtCrNe lamp does not have enough strong lines in the SWIR range (channels 6 -8 ) upwards of 1000 nm. The retrieval of in-flight line positions is additionally hampered by many damaged detector pixels of the SWIR detectors. An analysis of the mission data showed that for channels 6-8 no in-flight correction of the on-ground spectral calibration can be done with the standard approach, even at the beginning of the mission with a significant lower number of damaged pixels. We therefore started a study to investigate an alternative spectral calibration approach. In this approach a highly resolved reference spectrum is fitted with a DOAS like algorithm to retrieve the wavelength for each detector pixel. This approach was already tested by us for the Sentinel-4 UVN and the former Earth Explorer 8 candidate mission CarbonSat. For the method to work it is essential to have a good knowledge of the in-flight spectral response function and its change over time. Preliminary results show that the shape of the response function deviates from the current assumption of a Gaussian. For a successful fit it is also necessary to handle the many outliers caused by the bad pixels in an appropriate way. In this paper we present the first results of our investigations, focussing on channel 6 and the methane retrieval window.