The revised version of the manuscript is clearly improved and most of the referee comments were considered in the revision. Some changes were not applied to the whole manuscript (for example the words “optimal” or “best” were mostly removed, but remain in the abstract, in the conclusions and in some other places; also the update to the use of average RMS instead of sum of RMS is not done all along the manuscript) and some specific comments remain.
A major concern remains about the goal of this work and how exactly future scientists can rely on the conclusions done in the manuscript. The precise goal of this work is not clearly stated, which makes it difficult to make the link with the conclusions. If I understand it correctly, the goal of this work is to allow selecting a refractive index for an eruption, using IASI (or other TIR) observations. From that selection, one can then get information on the particle composition. However, I think that too many uncertainties remain on the selection of the “best”, “optimal” refractive index (and indeed those words were mostly removed from the revised version) to allow that goal to be reached. Therefore, I see the manuscript more as a succession of case studies for a number of eruptions, with a nice analysis of a selection of IASI spectra and a vague conclusion about which refractive index can be selected (vague because the differences in the results for different RI are too small to be really conclusive, especially considering all uncertainties and biases that can play a role in the calculations). At the end the method is thus not strong enough to be considered a general method for obtaining ash composition from IASI. At least I am not convinced. But maybe this is not precisely the goal of the analysis? Some more thoughts are also in comment 5.
More specifically:
1. The abstract is a bit too “enthusiastic” about the conclusion that can be reached
2. The comparisons of MBCRM with lbl and RTTOV are not clearly detailed (more below).
3. Comparison with lbl: the authors should provide the reference to the selected line parameters (for example HITRAN version …, or any other of course); also the information is not given about the simulations themselves: which spectral resolution was used, was an instrument line shape applied and which one, … Those all can have a decisive impact on the comparisons. About the results, I find the RMS to be quite significant (especially for nadir only, usually differences increase with the viewing angle) although mostly lower than the IASI NEdT. Between 1200 and 1600 cm-1 the RMS of the differences is even higher than the IASI noise for a number of channels, so I don't agree with the sentence in lines 93-94, especially since that part of the spectrum is used in this work. If these RMS of differences are compared to, for example, a similar analysis with RTTOV, one can see that the RTTOV bias with respect to LBL is much lower: https://nwp-saf.eumetsat.int/site/software/rttov/download/coefficients/comparison-with-lbl-simulations (and then in the list below you can select the IASI instrument). I do not mean that this RT code is not good, just that this bias / uncertainty (we don't know if the sign is constant, right?) is not negligible, and that combined to the IASI noise it leads to a significant uncertainty that reflects on the choice of RI. And this is not at all discussed.
4. Comparison with RTTOV: we can't see the difference good enough on the plot to assess it. I have a feeling that differences are up to 1K. This has to be taken in the discussion also; not that RTTOV is perfectly simulating aerosols, but that there's in addition to the IASI noise significant uncertainty in the aerosols radiative transfer that can amount up to 1K (if I see well, and again this is nadir only and will be larger for large viewing angles). In addition, again, the details of the RTTOV calculations are omitted. Which way are the aerosols simulated: Chou scaling (the default, very fast approximation with no explicit treatment of scattering and, to my personal knowledge, not appropriate for ash aerosols if looking for high accuracy simulations) or Discrete Ordinate method? This has a non-negligible impact on the simulations! How many streams if DOM is used? ...
5. I really miss a discussion about all the uncertainties, biases and errors (IASI noise, RT bias/uncertainty, additional uncertainty linked to uncertainty on surface and atmospheric temperatures, …) compared to the small differences between the simulations for different RI. Figure 9 (very useful) shows very clearly that many RI are within standard deviation of the “best”, for many of the analysed eruptions, and this is without considering the uncertainties I just mentioned. This is the part that fails to convince me that the method allows to really gain useful information on the particle composition. At best with the current information in the manuscript, I would say that the method allows selecting a decent refractive index to use in simulations when trying to obtain other ash parameters such as optical depth of the plume, or particle size. A decent conclusion could also be that, after careful discussion and analysis of all uncertainties, it is actually very challenging to gain information on particle composition by selecting a refractive index based on comparisons between simulations and observations, and that all those uncertainties have to be reduced before one can actually get particle composition information from IASI.
Specific / technical:
- Carefully check for the use of “optimal” or “best”, and also “sum of RMS” or “total RMS” all along the paper, for consistency with changes done in this version
- Lines 64-66: I think it needs rephrasing: now it looks like the particle size causes the large negative bias, but it is the ash plumes. I recommend splitting in 2 sentences.
- Lines 70-71: needs rephrasing. The IASI observations are not used to simulate BTS, but the simulated BTS are compared with the IASI observations if I understand the work correctly.
- Line 82: It is still written “we developed an original radiative transfer code”, while the code comes from a previous publication by different authors. I strongly suggest to rephrase.
- Lines 310-314: I am uncomfortable with the fact that the authors simply copied my explanation from the referee comment. It does not really belong here in this form and reads very weird in the context.
- Line 385: the SO2 layer height should also be given
- Lines 388-390: I am still a bit puzzled with this. I understand the authors explanation (in the answer to referee comments) why to remove this spectral range from the final RMS calculation, but I still do not understand why at step (a) this spectral part is not removed. It feels inconsistent. Also I think that at least a short explanation should be included in the manuscript, not just in referee answer
- Figure 9: the text says it shows the mean RMS, the caption says the minimum RMS. I guess it is indeed the mean.
- Line 601-602: I don’t agree with this sentence, see the major concerns above: “Furthermore, this result supports the feasibility of estimating ash RI from satellite infrared sounder measurements. »
- Line 662: Even with the correction, I still do not understand what would be biased (bias is a difference with respect to a reference data set and I see no comparison to any reference here). Or do you simply mean that no change of htop and tau was observed when changing the refractive index?
- Line 733: Why here is the list of suitable RI limited to 3, while just 2 lines above there were 5? If there is a link with SO2, I do not understand it and maybe it could be explained? And then in the following lines only 2 RI are mentioned, this is not very consistent.
- Line 884: the report referenced is not in English, I am unsure how to deal with this
- Lines 1079-1081: this is a much too strong conclusion…
- The conclusion section in its whole is not very clear to me, but I think this is in line with the fact that the goal of the work is not clearly stated in the introduction. |