A.Folch Comments on “Online treatment of eruption dynamics improves the volcanic ash and SO2 dispersion forecast: case of the Raikoke 2019 eruption” by Bruckert et al.

Summary

This paper couples the ICON-ART atmospheric chemistry model with the FPLUME volcanic plume model to furnish source conditions for ash and SO2 emissions from the 21-22 June 2019 Raikoke eruption. Time-averaged column mass loads are compared with AHI data from Himawari-8 satellite and from previous off-line studies (Muser et al. 2020). This is done for 3 configurations: FPLUME-Rad, FPLUME-norad and Mastin-rad. The “Rad” options turn on aerosol-radiation coupling on SO2 plume dispersion. The authors conclude that the coupling with FPLUME improves the ash loading comparison during the first hours and days after the eruption and that, afterwards, the aerosol dynamical processes become more important.

General Comments

• The FPLUME model actually outputs MER and vertical distribution of mass from plume height. However, it seems that, in the online ICON-FPLUME coupling, only MER is considered whereas a Suzuki parameterisation (with fixed parameter values of 4 and 5) is imposed for the vertical distribution of mass. Is there a particular reason for this? To derive both ESPs (i.e. MER and profile) from FPLUME would seem a much more consistent approach. Note that, in addition, FPLUME gives size-resolved vertical profiles, whereas the authors assume the same values for the Suzuki parameters (4 and 5 according to eq.1) for all particle/aerosol bins.
• Vent conditions (Sec.2.3.1). As noted in the text, FPLUME is sensitive to vent conditions, particularly in the case of low plumes. Insufficient momentum (low exit velocity) and/or insufficiency energy at inlet can yield, together with other atmospheric factors, to lack of model convergence. This is not necessarily a model shortcoming, but actually reflects configurations in which a plume could not be sustained and therefore collapses. Please check if this could be the case for columns below 10 km, particularly if exit velocities drop below 80-90m/s (this seems to be the case according to values from Table 1). Rather than using Mastin-derived MERs (see next comment), I suggest re-running FPLUME for the Raikoke phases 01, 02, 04, and 10 with higher values of exit velocity (see if this converges FPLUME). On the other hand, it is also true that FPLUME is a model for sustained plumes. Application to short-lived burst-like transient puffs may certainly fail.
• I do not see the need for the Mastin-derived modelling option in this study. Which is the purpose of it? First of all, with respect to FPLUME-rad only affects the strength of the source (given that same Suzuki is used in both cases). It should, therefore, imply only a scaling of the Amplitude component of the SAL metric. Unfortunately, comparison (e.g. in Figure 5) is not given to check this. If this is modelling option not relevant to the paper it could be removed. Results and conclusions would be unaltered. The paper could then focus on: i) comparison with previous off-line (Muser 2020) and, ii) comparison rad/norad. This would be simpler and easier to follow.
• ICON model configuration (section 2.4). Is there any particular reason for running ICON globally? Can the model be run only over a limited area? If not, could you comment on the grid approach over the area of interest or, if none, is the resolution of the R3B07 configuration uniform across the globe?
• Figure 5. Why it only shows results for FPLUME-Rad? Why the other FPLUME.norad and Mastin-Rad are never shown? On the other hand, can you comment on the poor Structure-component results for ash? 
• Any particular reason for doing 6-hourly model output averages? Considering that Himawari-8 observations can be available 4-times hourly, other approaches having less impact on the results could be considered. Averaging over such long intervals smooths out (e.g. peaks) and has substantial impact on instantaneous model results. How is SAL affected by this?  Can this mask disagreements?

Specific Comments

• L36 (and throughout the text). Should multiple reference citations be ordered chronologically?
• L103. Could you specify the size ranges for each aerosol mode?
• L166. Sc layer?
• Table 1. Please specify that heights are a.s.l. (and not above vent as required by FPLUME) and that emission rate of SO2 is computed using (2). Also, the emission rate of ash is not reported. 
• Figure 2. Why is FPLUME (red dots) given in all phases? 
• Figure 3. Is the color scale adequate? Should the higher value be at around 2 gm-2 or similar?  

This manuscript investigates the improvements obtained in forecasting dispersal of fine ash and SO2 by using the ICON-ART model once it is coupled with a model to reproduce the plume dynamics. FPlume model is used for this intention and the Raikoke eruption is adopted as test case. It also investigates the cloud transport dynamics and the processes which might explain the differentiation between SO2 and ash clouds.

A good description of Eruption Source Parameters is of paramount importance when trying to reproduce the dispersal of volcanic emissions, and it becomes even more essential within operational environment. In this sense, contributions like this manuscript are welcome and needed to keep working and identifying the best technical solutions for optimizing numerical effort and moving forward more reliable predictions.

1. Please in the abstract specificy how long lasted the eruption.
2. Line 7: what do you mean with „the simulated effect... is in the order of 6 km??“
3. Line 103: 7 in word
4. Line 126: ... by Marti et al. (2017)
5. Line 135: plese provide references for this statement.
6. Line 142: what do you mean with: „... to ensure an uninterrupted simulation?“
7. Paragraph 2.3: the Radius of the vent is also part of the equation, so I‘m wondering which value or range of values you adopted here?
8. Line 166: what Sc layer means?
9. Line 176: please refer to Table 1
10. Table 1: I‘d suggest to make this table more complete. I‘d add in the first column the day the phases refer to. I‘d then also add a column specifying the Fine fraction flux (kg/s) or specify in the caption of the table that the flux for fine ash is shown in Figure 2. Please specify in the caption how the SO2 flux is estimated. Please xplain the caption the source of all data showed in the table.
11. Line 186: Figure 2 shows only the MER and not the height
12. Figure 2: Fine ash is <30 or <32 micron? Please correct the title of the plot. What means E on the y-axis? Please clarify. In the caption: ...calculated with Fplume MER times.....and calculated with Mastin MER times...
13. Line 200: H_T is not the plume height averaged over the entire eruption duration? Please clarify.
14. Line 237: Please check the dates, aren‘t they 21 June and 22 June, instead?
15. Line 239: (compare Fig. 3 top and bottom)
16. Line 242: is the temporal evolution of ask loading on a hourly basis?
17. Line 243: plateau? Is not a peak??
18. Line 244: please start a new line when „Muser et al....“
19. Figure 3: Please double check the dates
20. Line 246: please double check the timing, I guess it is 18 UTC of 21 June as the plot shows the hours since the June 21, 12UTC
21. Line 251: in the bracket add: in Figure 4
22. Line 254: here you refer to the effect of meteorological conditions. I guess it would be useful to add come vertical profiles of wind speed in correspondence of the vent.
23. Figure 4: Please be consistent with terminology: is this ash or fine ash? Are these hourly values? I‘d much like to see the occurrence of pulses in these plots, to understand when the eruption dynamics affected the detection.
24. Line 276: here you are suggesting that the fact that you are not describing the gravitational spreading occurring nearby the source is affecting you capability in reconstructing the SO2 forecast in its initial phase. Then please explain why this is not also impacting the quality of the ash forecast?
25. Line 281: so from Figure 4 it turns out that the total amount of ash compares well with the satellite retrieval. However figure A2 shows a wider extent than what is retrieved. In this sense I‘d have expected that this meant forecasted concentration lower than the observed and in this sense an Amplitude < 0 (as at Line 223-225 you say that when the model overestimates the parameter A is positive, and negative otherwise). But Figure 5 shows positive Amplitude for ash at all time intervals. Please explain.
26. Line 285: here you say that S for ash in the beginning is negative. You justify the same results for SO2 saying that this is most likely due to the ash thick presence that obscure the gas component in the retrieval. What might be affecting the negative S for ash?
27. Paragraph 3.3: please double check the dates you refer to...
28. Figure 6: please clarify how the different heights are obtained. Here I see grey heights, are they coming from Fplume or satellite? Please specify. How are you getting the different heights for SO2 and ash? Are you maybe referring to the cloud height?? Please clarify.
29. I guess the caption in Figure 7 and A3 partly clarifies it as it mention this „horizontally averaged vertical distribution of mass“. Please spend more words in explaining the methodology, referring to Muser et al. 2020, is not enough. And use the words „plume“ and „clou“ in a distinct manner.
30. Line 358: I think the statement about the „dense ash plume“ hampering the validity of the comparison needs to be demonstrated. Please rephrase the conclusion.

Review of:

"Online treatment of eruption dynamics improves the volcanic ash

and SO2 dispersion forecast: case of the Raikoke 2019 eruption"

by Bruckert et al.

Overview:

===============    

This is an interesting study providing valuable

insight into the importance of online coupling approach

in volcanic ash and SO2 dispersion modelling.

They show that the online approach leads to significant

differences in the vertical structure of ash and SO2

plumes when radiative effects are taken into account.

This is a novel and important study shedding new light

on the processes involved in complex dynamics of

volcanic plumes and is suitable for publication in 

Atmospheric Chemistry and Physics.

However, I believe that additional work would need to 

be done before publication. I would recommend the 

authors to provide additional evidences to support 

their conclusions. Specifically, they argue that 

"online treatment of eruption dynamics improves the 

volcanic ash and SO2 dispersion forecast". 

Unfortunately, the paper fails to explain why the 

online treatment improves ash/SO2 dispersion forecast.

General comments:

===============

* Throughout the article, the authors highlight the 

  impact of aerosol-radiation interaction on the

  simulations. However, no evidence is provided to

  conclude that the online approach leads to better

  agreement with observations.

  

  One of the arguments used by the authors to claim that the

  online treatment improves model predictions is based on

  the temporal evolution of the total amount of ash (Fig. 4). 

  For example:

  l. 260, p. 10: "Thus, we conclude that the online treatment 

  of plume development improves the ash loading prediction 

  during the first hours and days after the eruption."

  However, this conclusion seems to be incorrect as the

  Fplume-norad simulation shows the best agreement with

  satellite data in Fig. 4. Moreover, differences in the total 

  mass loading could be related to a possible underestimation by  

  satellite retrievals or a poor estimation of the fraction 

  of very fine ash used to compute the mass eruption rate.

  

  The authors had the opportunity to show the improvement

  of the online approach by using the SAL metrics.

  However, only results for the Fplume-rad simulation are

  presented in Section 3.2. 

  

  Suggestion: 

  Improvement of the online approach could be 

  demonstrated by comparing SAL results for Fplume-rad 

  and Fplume-norad. For example, you could compute the 

  SAL metrics as the sum of the absolute values of S, A 

  and L, which results in an index ranging from 0 (optimal) to 6.

* Section 3.3 should be improved. A lot of information

  is missing. Please clarify how spatial averages are 

  computed in Figures 6 and 7.

Specific comments:

===============

* l. 51, p. 2: "They described the gravitational

  spreading of the umbrella cloud by the model of Costa

  et al. (2013). Collini et al. (2013) highlighted a 

  good agreement in ash transport simulations with 

  satellite observations for the Cordon Caulle eruption

  2011 by a combined WRF/ARW-FALL3D forecast system."

  These introductory sentences seems out of context. 

  Please better contextualize this discussion. 

* Eq. (1): The expressions for the vertical profile

  seem to be wrong in Marti et al. (2017) and also here. 

  The Suzuki distribution should be normalized by the

  integral of S, instead of the maximum S. For 

  discrete point sources, you should normalize using

  the sum of S requiring:

  E = \sum S*

* It seems that you distinguish between the terms 

  "eruption phase" and "puff" in some parts of the 

  manuscript, while sometimes are used as synonyms. 

  In order to avoid confusion, it would probably be 

  more convenient to unify the terminology and use only 

  "eruption phase". In my opinion, "puff" is a bit 

  ambiguous for a complex multi-phase eruption like Raikoke.

* What do you mean by "insensitive" here? Variations of 

  10% in column heights didn't affect MER estimations?

  Please clarify

* Figure 2 (title): "MER of very fine ash (<30 um)" ->

  It should be "<32 um", right?

* l. 217, p. 9: "The method assesses predefined objects

  based on a threshold value"

  You don't say how these objects were defined. What

  threshold value have you used?

* l. 230, p. 9: Where do these gaps come from? The mean

  averaging you applied to fill gaps conserves the

  total mass? Or are you adding new mass with this

  approach?

* l. 232, p. 9: Regridding mass loading to a coarser grid 

  by a linear interpolation is not the best approach, as

  mass conservation is not ensured. Do you have an idea

  how much is the total mass difference induced by the

  interpolation method?

* l. 260, p. 10: "Thus, we conclude that the online 

  treatment of plume development improves the ash loading 

  prediction during the first hours and days after the 

  eruption."

  This statement is not correct. It cannot be concluded

  from the results presented in Section 3.1 that the 

  online treatment improves simulations. In fact,

  FPlume-norad outperforms Mastin-rad, and you cannot

  say that FPlume-rad is better than FPlume-norad. I

  think the only valid conclusion here is: "the FPlume

  experiments (ie, FPlume-rad and FPlume-norad) showed

  better agreement with observations"

* l. 264, p. 11: SAL requires defining model and 

  observation objects. For instance, you can identify 

  clusters of grid cells with mass loading exceeding a 

  given threshold. However, you haven't mentioned what 

  threshold was used. This threshold can be defined 

  based on the detection limit of the satellite 

  retrievals. For example, Prata et al. (2021) assumed 

  0.2 g/m2 for volcanic ash. Otherwise, model-observation

  comparisons wouldn't be fair.

  Please clarify what thresholds you have assumed for 

  ash and SO2.

* l. 282, p. 12: Results in Fig. A2 are really good. 

  I think it would be worth including Fig. A2 in the

  main body of the paper, probably replacing Fig. 3.

* l. 295, p. 12: Why are you defining two threshold for

  volcanic ash instead of using a single threshold for

  total ash? Why didn't you define a threshold for the

  giant mode? What threshold have you used for ash in

  Fig. 6?

* l. 296, p. 12: I see no reason to remove those

  'steps'. Why top height is greater than zero before

  the eruption starting time in Fig. 6? This has to do 

  with the smoothing? Since data was smoothed, you

  should at least indicate the vertical resolution of

  the model. Is it comparable to the differences found 

  in top height between rad and no-rad experiments?

* Figure 6: How is the mass averaged height computed? 

  Is a vertical average? In this case, the average is 

  weighted by mass concentration or by mixing ratios? 

  Or is a horizontal average? In this case, the average

  is weighted by mass loading (in g/m2).

  Are vertical profiles also averages? Or are they

  computed at specific locations? How this average is

  performed? Do exclude grid cells without ash/SO2 from

  the average?

* l. 306, p. 13: "in the FPlume-norad case still shows 

  the overlap of the different phase dependent profiles"

  Where is shown this overlap? Please clarify.

* l. 318, p. 14: "the vertical distribution of the

  total ash mass"

  What do you mean by "total mass"? Is the total mass

  within a model layer?

* Figure 7(c): "SO2 mass loading in kg" -> mass loading

  should be in units of g/m2 as in Fig. 3. Why are you

  showing "mass" in (a) and "mass loading" in (c)? What

  is the difference?

* Equation (3): Obtaining the median radius for a 

  multimodal log-normal distribution is not a trivial

  problem. Are you sure the median radius is given by

  such a simple formula? Or this expression only defines

  a "characteristic radius"?

* Figure A1(left) is not relevant.

Technical corrections and minor comments:

===============

* l. 126, p. 5: Remove parentheses from Marti et al. (2017)

* l. 130, p. 5: Correct citing format in Rose and Durant (2009)

* l. 143, p. 5: It is necessary a reference for the 

  Mastin empirical relationship

* l. 155, p. 6: Specify: "ellipsoid" -> "Earth ellipsoid"

* l. 166, p. 6: What do you mean by "marine Sc layer"?

* l. 174, p. 7: "The Raikoke eruption 2019" -> "The 2019 Raikoke eruption"

* l. 204, p. 8: "Himawari-8 Ash and SO2" -> "Himawari-8 Ash and SO2 retrievals"

* l. 212, p. 8: "The ash retrievals were corrected 

  (...) completely cloud covered". Please clarify this

  sentence.

* l. 330, p. 15: "Fig. 7 also shows" -> "Fig. 7d also shows"

* l. 331, p. 15: "the radius is higher on average because" ->

  "the radius is higher on average according to the

  FPlume-rad experiment because"

* Figure 7 (caption): "and SO2 mass (b)" -> "and SO2 mass (c)".