This study investigates the secondary gravity wave (GW) response to Hunga Tonga-Hunga Ha’apai (HTHH) volcanic eruption, which took place on the 15th of January 2022. The work combines an extensive network of meteor radar wind measurements with simulations performed by the MESORAC/HIAMCM modeling system, to identify the global characteristics of the secondary GW response. The results add to the rapidly expanding body of literature surrounding the whole-atmosphere impacts of the HTHH event, for which secondary GWs play an important role in a range of atmospheric and ionospheric processes.

While the paper presents interesting and relevant results, the manuscript can be improved by including a more detailed description of a number of analysis and discussion points. Since many figures can also be improved, a considerable revision is in place. Some general and specific comments are given below. 

General comments

Experimental setup

A more detailed (but concise) description of the modeling setup (as described in Vadas et al. 2023a) should be included, which includes a discussion of the impact of dissipation and resolution on the analysis presented in the current work. 

By the nature of the experimental design, only the effects of secondary gravity waves can be investigated using the HIAMCM. In the discussion (lines 277-278), it is mentioned that the HIAMCM zonal and meridional winds exhibit variable phase relations between both wind components and a much less coherent phase front than expected for a Lamb wave. However, there is no mention of Lamb waves being present in the analysis or HIAMCM results. Can the authors elaborate on how the results support this discussion point?

A discussion should be included on the fact that the HIAMCM model generally underestimates GW amplitudes by ~50% and up to 150% (Vadas et al., 2023a), which would then include a comparison between the fitted observed and simulated GW packet amplitudes. Currently, it is difficult to make out whether or not the HIAMCM model adequately captures the wave forms needed to explain the observed HTHH wave-response. The concluding statement (lines 312-315) that the secondary GWs can explain the majority of the analyzed radar wind GW perturbations therefore does not carry much weight, given also that other possible sources of wind perturbations are not discussed in detail. 

Polar vortex and Figure 2

The statement that HIACMC shows that the polar vortex disturbs the westward traveling GWs (lines 120-124) should be discussed in more detail. At first glance, a discussion of the impact of the polar vortex on the HTHH GWs is not given in the cited literature (e.g. Vadas et al., 2023b). For example, what are the mechanisms behind the polar vortex altering GW propagation, and what was the state of the polar vortex during the HTHH event? Fig. 2 does not seem to show any clearly discernible polar vortex structure, as alluded to in the text. 

On lines 175-176 the impact of the polar vortex on the observed wind variations is further noted as being caused by GWs generated by the vortex. As a reader, it is impossible to verify such a statement, given also the other possible sources of wind variability mentioned in the text. While it is well known that the polar vortex can generate GWs, how and to what extent that applies to the current analysis should be discussed in more detail (including a discussion on the current state of the vortex). It is further noted that orographic GWs add noise to the European sector winds. But can orographic GWs not just as well add noise to the South American sector winds? 

Could the authors discuss the source of wind perturbations above Europe and Scandinavia in Figure 2? These are already present at the 12:00 snapshot (Fig. 2a), and grow rapidly before the GWs reach the sites (16:00 snapshot, Fig. 2b), with the ‘noise’ quickly becoming comparable in magnitude (or even greater than) the GW signal itself. It seems hard to argue that this is due to the polar vortex, since both the ‘Tonga run’ and ‘base run’ simulations specify the same vortex structure, and the noise is limited to the European sector. 

Figures 3, 4 & 5

These figures should have labels added to the axis, and the font size should be enlarged to make the legend and labels more readable. Due to the small panel size, variations in the observed winds are difficult to distinguish, giving the figures an overall noisy impression. The y-axis scale on the observation can probably be changed to limits [-40, 40] without loss of information.

Throughout the text it is argued that increases in wind variability can be identified by looking at these figures (e.g., lines 176-178 and line 239), but this is not so straightforward. Perhaps increases or decreases in wind variability at certain time periods can instead be quantified, for example using a sliding window approach of the calculated variance. Currently, it is difficult to judge if a supposed increase in variability (e.g. before the arrival of the westward HTHH GWs at the MEN radar) is just that, or if the increased in variability is due to the presence of some (possibly unaccounted for) waveform. 

As the left-hand and right-hand panels of Fig.3,4,5 show fundamentally different things, it would be beneficial if the fitted waveforms and amplitudes are overlain (and possibly zoomed in on) for both the observatory and simulated winds. In addition, showing the observed and modeled wind fluctuations above and below each other (rather than side-by-side), would make the difference in arrival time easier to distinguish.

As mentioned earlier, HIAMCM simulated GWs are smaller by ~50% than observed due to the choice of turbulent diffusion coefficient D0 (Vadas et al. (2023a)). The impact of D0 on simulated GW amplitudes should therefore also be discussed within the context of Fig.3,4,5, including how it impacts the choice of y-axis scaling for the right-hand plots, and the amplitude of the fitted waveforms. 

Q2DW

The impact of the Q2DW on the GW propagation lacks detail. It is not clear from the text or earlier cited work what the amplitude of the Q2DW is and how it compares to the concurrent tidal and low-frequency GW amplitudes. Discussing the mechanisms of the Q2DW impacts is important, as one of the concluding statements of this work is that the observations reveal that the Q2DW played an important role in the GW propagation. 

L263 - Stober et al. (2023) are cited in support of the statement that the GW packet is doppler shifted due to a strong Q2DW, thus explaining the difference in arrival time between the observations and HIAMCM for McM and DAV. It is unclear how the cited work supports the current statement; could this be elaborated further?

L205-208 That the GW packet faced a strong headwind towards the south due to the presence of the Q2DW would surely depend on the phase of the Q2DW in the region of propagation? While on lines 204-205 it is clarified that the Q2DW meridional wind component showed clear northward winds over the South American continent, the Antarctic stations (and the associated propagation paths) are at considerably different longitudes (as shown in Fig. 1). Given the zonal wavenumber 3 structure of the Q2DW, it is not obvious what the phase of the Q2DW was (i.e., headwind or tailwind) facing the Antarctic stations, and it may well have been a tailwind.

While the Q2DW was said to be present, its magnitude and relative importance versus the semidiurnal and diurnal tidal mode amplitudes is not discussed. That the superposition of these wave components leads to high wind speeds, does not necessarily imply that Q2DW winds are especially large (even though this may very well be the case), or that the semidiurnal and diurnal tides do not have a similarly large impact on GW propagation. The presence of other (large amplitude) tidal waves along the different great circle propagation paths (and their representation in the HIAMCM model), should therefore be discussed in more detail. 

Other comments

L17 - The wording would be more scientific if ‘huge’ is removed.

L22 – Total Electron Content (TEC) abbreviation should be defined.

L43 – The way the sentence starting with “In this study,...” is worded, suggests that TEC perturbations are also discussed in the current work, which they are not. 

L47 - "strongest gravity waves launched by the HTHH". Maybe this is a matter of semantics, but can secondary GWs really be considered as being launched by the HTHH? Perhaps ‘resulting from’ would be more appropriate. 

L62 – If the 2 km resolution with a 5-kilometer vertical averaging window centered around the respective altitude is identical to the technique described in Stober et al. 2023, this article should be cited here. If the methodology has been substantially modified since Stober (2023), more information about the analysis technique should be given instead. 

Fig. 2 - The quality would be improved if the axis labels would not overlap with the plot. 

L77 - “Could” to “can”.

L78 – The dashed red line not only connects (or lies close to the) ALO and HTHH, but also KUN, MEN, and CAR, as noted in the caption of Fig. 1. The caption and text statements should be the same. 

L92 – Sentence can be clarified by rewording to  "...specify realistic large-scale meteorological fields through which the resolved GWs propagate (Becker et al., 2022)".

L100 - “cover” to “covering”

L124 - ‘This is consistent…’ should be reworded to clarify what exactly ‘this’ refers to. Currently it can be read as if the cited work also supports the result that the polar vortex disturbs GW propagation, which does not appear to be the case.

L144 - “theoretical arrival times” suggests the arrival times are based on theory, while they are derived from best-fit phase speed using all stations that detected this wave. Perhaps ‘derived’ would be more appropriate?

L145-150 In the discussion (L243 - 244), it is mentioned that the distance between the solid left and right vertical lines is 3 hours. Please add this information here already. 

L152-154 Mentioning the power of the CMOR meteor radar relative to the other radars, suggests that the radar power may represent a considerable source of variability for the other radars. Considering also the impact of diurnal meteor count variations on wind variability alluded to on lines 163-165, the impact of the system power as well as diurnal meteor count variations on wind variability should be discussed in more detail.

L165 Perhaps ‘This provides…’ can be reworded to ‘Our results therefore support…’, as it currently reads as if the meteor count rate itself somehow impacts asymmetric azimuthal GW propagation.

L217 ‘The reasons for..’ could be changed to ‘Possible reasons for…’, since the discussion of the Q2DW impact on the DAV and McM stations is largely speculative. Furthermore, the discussion on lines 254-266 mention the impact of Southern Hemisphere MLT temperatures, which does not make sense chronologically. Merging the Q2DW and MLT temperature impacts on the DAV and McM stations into a single section would make sense. 

L223 - MR should be defined.

L252-253 In Vadas and Azeem, the mesopause sound speed (at 98 km) was estimated to be 270 m/s and not 280 m/s. It would also be nice to see in the text for which conditions this number was obtained (March 25 at 100°W and 34°N). 

L270 - It is unclear where the wave speed of 202.5 m/s comes from. This number has not been mentioned earlier in the text. 

L279 - It is unclear what the visibility of a Lamb wave in the meteor radar wind data entails (Lamb waves are noted as being "hardly visible" - can evidence of this wave somehow be seen in Fig 3,4,5?). The sentence starting on line 279 also seems to imply that Lamb waves are effectively filtered out by the spatial and temporal averaging kernel employed in this work. If so, the statement that the observed wave period and horizontal wavelength observed in this work (lines 275-277) cannot be reconciled with that of a Lamb wave does not make sense, since by design, Lamb waves cannot be measured.

Fig. 7 - It would be good to label the ALO stations for the eastward and westward wave packets separately, and to add gridlines to the figure. 

Many of the DOI’s in the reference list are formatted wrongly, having "https://doi.org" twice.

References

Vadas, S. L., & Azeem, I. (2021). Concentric secondary gravity waves in the thermosphere and ionosphere over the continental United States on March 25–26, 2015 from deep Convection. Journal of Geophysical Research: Space Physics, 126(2), e2020JA028275.

Stober, G., Liu, A., Kozlovsky, A., Qiao, Z., Krochin, W., Shi, G., ... & Mitchell, N. (2023, April). Identifying gravity waves launched by the Hunga Tonga–Hunga Ha′ apai volcanic eruption in mesosphere/lower-thermosphere winds derived from CONDOR and the Nordic Meteor Radar Cluster. In Annales geophysicae (Vol. 41, No. 1, pp. 197-208). Göttingen, Germany: Copernicus Publications.

Vadas, S. L., Becker, E., Figueiredo, C., Bossert, K., Harding, B. J., & Gasque, L. C. (2023a). Primary and secondary gravity waves and large‐scale wind changes generated by the Tonga volcanic eruption on 15 January 2022: Modeling and comparison with ICON‐MIGHTI winds. Journal of Geophysical Research: Space Physics, 128(2), e2022JA031138.

Vadas, S. L., Figueiredo, C., Becker, E., Huba, J. D., Themens, D. R., Hindley, N. P., ... & Bossert, K. (2023b). Traveling ionospheric disturbances induced by the secondary gravity waves from the Tonga eruption on 15 January 2022: Modeling with MESORAC-HIAMCM-SAMI3 and comparison with GPS/TEC and ionosonde data. Journal of Geophysical Research: Space Physics, 128(6), e2023JA031408.

This paper investigates the secondary waves originating from the 15th January 2022 Hunga eruption and observed in the MLT region by 13 meteor radars around the world. The observations are compared with numerical simulations. A central claim of the paper is that the observed MLT waves are secondary gravity waves.

This is an interesting study. The new observational data set, and its analysis, are very valuable. The modeling and comparison to the observations appear sound. The paper is overall clearly written. There are nevertheless a few issues which should be addressed before publication. In particular, it seems to me that the statement about the observed waves being secondary GWs is not fully supported by the current analysis. The paper would also benefit from some clarifications (see main comments below).

Main comments:

1) The reasons for the two-step modeling are not explained in the paper. Perhaps they can be found elsewhere, but they should be repeated here. In particular, why not include the primary waves directly in HIACCM, and let them dissipate and directly trigger secondary gravity waves ?

2) It is argued that the waves observed by the meteor radars in the MLT are secondary gravity waves, and not primary waves (in particular not the mode referred to as L1 by Liu et al., 2023 or Pekeris wave by Watanabe et al., 2022). The line of argument does not sound compelling to me, since the propagation speed of the waves examined is similar to that of the L1 mode, as mentioned by the authors.

One argument I see in favor of the authors' interpretation is that the amplitude of the L1 mode is expected to be larger West of Tonga (e.g., Fig S1 of Liu et al., 2023), while the observations (in agreement with HIAMCM, apparently) suggest the opposite. Another possibility is that L1 does not extend to the MLT. Perhaps the authors could elaborate on this in the discussion.

3) Related to the previous comment, the information about GW amplitudes at various distances and in various propagation directions is missing from the discussion and could be useful in the argumentation. Even if there is a discrepancy with the observations regarding the absolute amplitude, there is valuable information in the simulation about the relative amplitude of the fluctuations between the stations.

4) Figures:

Figure 2: At early times, the wave fronts in HIAMCM are quite concentric. Hence, a keogram at 96 km in the eastward and westward directions would help identify the wave propagation speeds in HIAMCM.

In Figures 3, 4 and 5, the labels should be enlarged. The information on the distance is already provided in Figure 1 and does not need to be repeated, however the azimuth is missing.

Specific comments

l 34: ExB-coupling: please write the formula in latex with $$, or expand.

l49: 'at various distances from the eruption site': you might add 'at various distances and azimuths'

l 52: 'characteristic amplitude variations at certain stations' do you mean 'characteristic amplitude variations between certain stations' or 'characteristic waveform' ?

l 61: I suggest rephrasing the statement regarding vertical resolution, which at present is confusing. A 2 km vertical resolution suggests that you can detect perturbation with a 4-km vertical wavelength. This is definitely not the case using a 5-km moving average.

around l 60: it would be useful for the reader to know which components of the wind vector are measured by meteor radars

l65: 'collecting' -> within ?

l 65-71: For legibility, I would recommend removing the list of radar sites from the main text, and refer to Table 1. If you adopt this suggestion, you should add the references for the instruments to Table 1.

l 81: Is it necessary to include the 90 km altitude ? The difference is 1.5 %

Table 1: it would be good to recall the location of the volcano in the caption

l 89: Some information is missing from the HIACCM model description: grid spacing, both vertical and horizontal, upper boundary condition, ... They can probably be found elsewhere, but should be repeated here.

l 89-101: a few details on the workflow of MESORAC and its initialization would be welcome here. From the acronym, I understand that the local GW forcing related to the eruption (which is inferred from geostationary satellite images) is propagated/dissipated using ray tracing equations, but this needs to be explicitly stated. Furthermore, some additional information on the model would be welcome (even if it can be found in the Vadas et al. 2023 paper), e.g. what is used as background flow for the ray tracing ?

l 115-118: I am missing something here, do you mean that wind filtering occurred in the stratosphere West of the volcano, and this did not trigger 2ndary GWs ? The main difference I see in Fig. 2 is between the eastward and westward sectors.

l 122: 'Results from the HIAMCM also show that the HTHH GWs traveling westward along the GC towards Europe are disturbed by the polar vortex in the Northern Hemisphere and, thus, almost no coherent wavefront arrives over Scandinavia, although this would have been the shortest distance to HTHH.' : the second part of the sentence is confusing. Do you mean that ' almost no coherent wavefront arrives over Scandinavia from the North-West (or another direction)'.

There are some differences between the simulations in the European sector which are not related to HTHH (they emerge before the waves have reached that region) and might be hiding a HTHH signature there. Are you confident that this is not happening ?

l 123-124: 'gained strength over the Atlantic Ocean': for waves emanating from a point source on the sphere, we expect a geometric strengthening effect with refocusing of the rays after the mid-point to the antipodes (see, for instance, Matoza et al., 2022; Amores et al., 2022). Are you referring to that effect ?

l 125: add 'comparison' to the section title ?

l 144: 'all meteor radar': only a subset is shown in Fig. 3

l 150: I would end the sentence after 'in the model'

l 165: 'This provides... ' would be better placed right after 'on the same GC' on line 163

l 201: I would rephrase 'The Q2DW may play an important role'-> 'The Q2DW may explain this discrepancy'.

l 200-205: A figure comparing the low-frequency wind field and the Q2DW between HIACCM (or MERRA-2) and meteor radar observations, for instance in an appendix or supplement, would strengthen this point.

l 256: 'noctilucent clouds in Tierra del Fuego' a reference would be welcome here

l 276: Apart from the introduction, it is the first time in this paper that mention is made of the horizontal wavelength. How is it derived ? The argument about the period being inconsistent with L1 is not really convincing: as far as I see from Fig. 3 (for instance at CONDOR or Cariri), HIAMCM sometimes also underestimates the wave period, and it does strongly underestimate the wave amplitude.

A data availability statement is missing. See https://www.atmospheric-chemistry-and-physics.net/policies/data_policy.html

Typos:

l 105: The sentence 'The HTHH GWs are extracted by subtracting a reference run from the disturbance simulation.' seems redundant with the previous one and could be deleted.

l 261: missing comma after previously

In the reference section, please check the url, some have twice the doi.org prefix.

References:

Watanabe, S., K. Hamilton, T. Sakazaki, and M. Nakano, 2022: First Detection of the Pekeris Internal Global Atmospheric Resonance: Evidence from the 2022 Tonga Eruption and from Global Reanalysis Data. J. Atmos. Sci., 79, 3027–3043, https://doi.org/10.1175/JAS-D-22-0078.1.