Towards Retrieval of Sulphur Dioxide from TROPOMI Radiance Measurements
van Gent, Jeroen1; Theys, Nicolas1; Danckaert, Thomas1; De Smedt, Isabelle1; Van Roozendael, Michel1; Lerot, Christophe1; Brenot, Hugues1; Spurr, Robert2

The TROPOMI instrument, to be launched on-board the Sentinel 5-precursor platform in 2015, is expected to deliver data at an unprecedented spatial resolution of 7x7 km with a revisit time of one day. For the retrieval of atmospheric sulphur dioxide (SO2) concentrations, this resolution will allow us to reveal fine details in both anthropogenic emissions and SO2 clouds of volcanic origin. The much higher data rate of TROPOMI, as compared to current instruments such as GOME 2 (METOP A/METOP B) or OMI (EOS Aura), poses constraints regarding computational speed on the L1b-L2 retrieval algorithms. In addition, the fine spatial resolution may produce non-linear effects, as strong SO2 sources are better resolved and may lead to saturation of the SO2 signal. In this paper, we present the technique used in our prototype algorithm for the future near real-time retrieval of SO2 total column values from TROPOMI L1b measurements. For computational efficiency, the method is based on the well-known DOAS technique, where a slant column density (SCD) is obtained by fitting cross-section spectra of the chemical species of interest to the observed spectrum. Subsequently, the SCD is converted into a vertical column (VCD) by means of an air mass factor (AMF), derived with a radiative transfer model. The DOAS-technique intrinsically relies on the assumption of an optically thin atmosphere. Traditionally, SO2 is retrieved by applying a suitable fitting window in the 310-325 nm range. In order to avoid non-linear effects, thereby maintaining the validity of the DOAS method, we propose to use three different fitting windows: 310-325 nm; 323-335 nm; and 360-390 nm. The first window serves as a baseline. When saturation of the SO2 signal is detected, the algorithm selects an alternative fitting window. For typical S5P observation geometries, we find that the three fitting windows are best suited for SO2 total columns with VCD < 20 DU; 20 DU < VCD < 100 DU; and VCD > 100 DU respectively. In addition, we present some results for the three-window approach with our direct-fitting algorithm for the experimental retrieval of SO2 effective plume altitude. Although the direct-fitting algorithm is too computationally expensive for near real-time processing of TROPOMI L1b data, we envisage its application in the automated processing of SO2 clouds from volcanic eruptions. In particular we show that the use of multiple fitting windows enhances the accuracy of simultaneous retrievals of SO2 VCD and plume altitude for the strongest eruptions. In the current development phase, both the prototype DOAS algorithm and the experimental direct fitting code are tested on the scanning instrument GOME-2 and (like TROPOMI) non-scanning imaging absorption spectrometer OMI.