Enhanced characterization of SO2 plume height and column density using the second UV spectral band of TROPOMI
Abstract. Volcanic emissions of sulfur dioxide (SO2) affect the environment, climate, and society. Their detection and quantification rely extensively on remote sensing techniques, which are used to track SO2 and monitor volcanic activity worldwide. In particular, nadir-viewing satellites measuring total SO2 vertical column densities (VCDs) have provided valuable insights into volcanic emissions for decades. However, the determination of the SO2 layer height (LH) is more challenging. In this study, we present an improved SO2 LH (and VCD) retrieval algorithm, applicable to the second UV spectral band (BD2) of the TROPOspheric Monitoring Instrument (TROPOMI). This band exhibits a stronger SO2 absorption than the third band (BD3) that is traditionally used for SO2 retrievals. To assess the impact of various spectral, atmospheric, and observation conditions, we conducted sensitivity analyses from a set of synthetic spectra representative of TROPOMI measurements using the Look-Up Table COvariance-Based Retrieval Algorithm (LUT-COBRA). Our results demonstrate that BD2 retrievals result in more accurate estimates of the SO2 heights and columns, particularly in the upper troposphere and lower stratosphere (UTLS), with LH errors reduced by at least a factor of 2. The algorithm was applied to real TROPOMI observations from volcanic eruptions and degassing episodes, and compared to BD3 retrievals. BD2 shows an improved sensitivity, with less noise, and a detection limit as low as 2 DU, surpassing the current operational TROPOMI SO2 product by an order of magnitude. Furthermore, our plume height estimates align closely with independent measurements from the Infrared Atmospheric Sounding Interferometer (IASI) and Microwave Limb Sounder (MLS), confirming the reliability of the approach.
Competing interests: The authors declare that they have no conflict of interest. At least one of the (co-)authors is a member of the editorial board of Atmospheric Measurement Techniques.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.
This work builds on the previously reported LUT-COBRA algorithm (Theys et al., 2022) for volcanic SO2 plume height retrievals, by applying it to the UV2 measurements from TROPOMI. Through theoretical calculations and sensitivity tests, the authors demonstrated that the TROPOMI UV2 measurements at shorter wavelengths (~305 nm) can provide better accuracy for SO2 layer height than UV3 measurements. The authors also described the implementation of the LUT-COBRA algorithm with the TROPOMI UV2 band, and compared the retrieved plume heights with those from the operational TROPOMI SO2 height algorithm (based on a machine learning technique) as well as those from infrared (IASI) and microwave limb (MLS) sensors. Overall, this is a well-written paper, and the topic should be of interest to the atmospheric science and remote sensing communities. While the technique has been previously described, the application to TROPOMI UV2 clearly shows improvement in the retrieved SO2 heights. The paper can be further improved by addressing some technical points (see specific, mostly minor comments below) and I’d recommend minor revisions before the paper can be accepted for publication in AMT.
Specific comments:
Lines 140-142: In Jacobian calculations, the authors used LIDORT (instead of VLIDORT) and did not explicitly consider aerosols or rotational Raman scattering. Can the authors comment on the uncertainties in the retrieved SO2 plume heights associated with these factors?
Lines 149-150 and Table 3: the interpolation error for Jacobians at relatively large SO2 VCDs can be quite substantial especially at short wavelengths – can the authors comment on the selected SO2 VCD nodes in the LUTs?
Figure 5: is the apparent increase in error for SO2 LH > 25 km due to the coarse resolution of LH nodes in the LUT?
Figure 6 and lines 239-240: some of the error sources are not completely independent (e.g., T prof. and Air prof., O3 prof. and O3 VCD). Not sure if quadrature summation is justified here.
Section 4.1: The construction of the weighted O3 profiles was done to reduce the size of the LUTs, right? How does this affect O3 profile related errors?
Lines 315-318: what is the typical number of eigen vectors used in equation 9? Is this done for all spectra (or just those with ill-conditioned covariance matrix)?
Table 4 and section 4.2: all examples given here appear to be under relatively small or moderate SZAs. Can the authors present some results for higher SZAs, where the O3 profile effect could be more significant?
Figure 8: there appears to be a general tendency of pixels with lower VCDs to have lower plume heights. Is there an explanation for this?
Figure 10c: the BD2 retrievals appear to be quite noisy for the eastern branch of the plume – how confident are the authors in the retrieved heights here?
Figures 11 and 14: it is a bit difficult to directly compare IASI with TROPOMI using these maps. The profile plots (Figures 12 and 15) are more informative for these two cases.