General comments

This study assesses the 11-year solar cycle signals in the middle atmosphere in multiple-model ensemble simulations. This study starts with initial solar cycle signals in short wave heating rate, ozone, and temperature anomalies and continues with an analysis of whether the top-down mechanism explaining the downward propagation of these initial solar signals can be found in the presented models. I find the study highly relevant for long-term discussion of indirect solar effects and recommend it for publication with minor comments listed below.

Specific comments

I am not convinced that three sets of historical-like simulations with 9, 6 and 10 ensemble members, respectively, can be called “large ensembles” (l2+l330) since we do not know how large the large ensemble needs to be (Milinski et al, 2020).

Can you specify the threshold in the abstract (l8)?

What does “partly confirmed” (l25) mean? 

The authors should elaborate more on the fact that the solar signal may not be stationary (Thejll et al, 2003) related to modulation by QBO and PDO (l30).

I would omit the “controversial” label (l39) even though these studies may reduce the confidence level of the solar-NAO connection as you state.

As shown in previous studies (e.g. Mitchell et al, 2014; Kuchar et al, 2015) the upper stratospheric equatorial temperature anomaly related to the solar cycle has been detected showed a statistically significant signal with structure and amplitude of 1–1.25 K. Temperature response in Fig. 2A maximizes at 0.6 K. I would say that models a bit underestimate the response even with comparison (l167) with Kunze et al (2020). These facts should be discussed and even analyzed more thoroughly in your models. 

Based on Fig. 1.c (l159), the authors suggest that a nonlinear response can occur when the solar forcing is strong enough but I would soften these statements given the large spread and not enough samples for high sfu values.

I would omit the publications of Gray et al (2010) which provides a review of the Kodera and Kuroda mechanism and Mitchell et al (2015; CMIP5) which does not show any BDC response (l219) and only highlight the link between weaker BDC and lower-stratospheric temperature induced by the 11-year solar cycle. 

How different (l240)?

Would you find relevant to reproduce composite differences between Smax and Smin as in e.g. A8 for ERA5 and assess whether the response of temperature and zonal wind in a reanalyzed dataset also reveals a sensitivity to weak and strong solar epochs? 

Using vector figures instead of raster ones may help to improve the quality of your publication.

Due to the extensiveness and unique methodology of the study, I think the whole community would appreciate an adoption of Open Science approaches to allow reproduce the extensive analysis in this study (e.g. Laken, 2016). In particular, I would recommend any kind of willingness of the authors to publish the code allowing to reproduce the figures in the paper. There are multiple ways how to proceed, either to allow the access upon request or via portals allowing to assign Digital Object Identifier (DOI) to the research outputs, e.g. ZENODO. I think it could enhance the quality and reliability of this publication.

I really appreciate the authors’s willingness to use the robust bootstrap method to but why do you use only 1000 samples? Furthermore, this should be used to assess the significance level of the correlation coefficient to secure methodological consistency. Or was the temporal autocorrelation taken into account in your composites? Can you discuss how the inclusion of the effective sample size (see Section 5 in Bretherton et al, 1999) would influence the t-test results? Do your composite samples comply with the t-test assumptions?

Technical comments

Please specify what CCR in your figures stands for

l288 replace EAR with ERA5

l290 replace EAR with ERA5

References

Bretherton, C. S., Widmann, M., Dymnikov, V. P., Wallace, J. M., & Bladé, I. (1999). The Effective Number of Spatial Degrees of Freedom of a Time-Varying Field, Journal of Climate, 12(7), 1990-2009. Retrieved Jan 19, 2022, from https://journals.ametsoc.org/view/journals/clim/12/7/1520-0442_1999_012_1990_tenosd_2.0.co_2.xml

Kuchar, A., Sacha, P., Miksovsky, J., & Pišoft, P. (2015). The 11-year solar cycle in current reanalyses: a (non)linear attribution study of the middle atmosphere. Atmospheric Chemistry and Physics, 15(12), 6879–6895. https://doi.org/10.5194/acp-15-6879-2015

Kunze, M., Kruschke, T., Langematz, U., Sinnhuber, M., Reddmann, T., and Matthes, K.: Quantifying uncertainties of climate signals in chemistry climate models related to the 11-year solar cycle – Part 1: Annual mean response in heating rates, temperature, and ozone, Atmos. Chem. Phys., 20, 6991–7019, https://doi.org/https://doi.org/10.5194/acp-20-6991-2020, 2020.

Laken, B. A. (2016). Can Open Science save us from a solar-driven monsoon? Journal of Space Weather and Space Climate, 6, A11. http://doi.org/10.1051/swsc/2016005020.

Milinski, S., Maher, N., and Olonscheck, D.: How large does a large ensemble need to be?, Earth Syst. Dynam., 11, 885–901, https://doi.org/10.5194/esd-11-885-2020, 2020.

Mitchell, D.M., Gray, L.J., Fujiwara, M., Hibino, T., Anstey, J.A., Ebisuzaki, W., Harada, Y., Long, C., Misios, S., Stott, P.A. and Tan, D. (2015), Signatures of naturally induced variability in the atmosphere using multiple reanalysis datasets. Q.J.R. Meteorol. Soc., 141: 2011-2031. https://doi.org/10.1002/qj.2492

Thejll, P., Christiansen, B., and Gleisner, H.: On correlations between the North Atlantic Oscillation, geopotential heights, and geomagnetic activity, Geophys. Res. Lett., 30, https://doi.org/10.1029/2002GL016598, 2003

Huo et al. explores a topic of 11-year solar cycle influence on the atmosphere through the top-down mechanism. Surface effects of such mechanism are very uncertain and have been largely disputed in a recent literature, and therefore the authors focus mostly on the middle atmospheric part of the story, where the forcing is created for further potential downward propagation. Authors aim to explore the potential reasons for multi-model uncertainties by looking at the solar signals in the middle atmospheric shortwave heating rates, temperature, ozone, and zonal winds, and by contrasting those to climatological biases of models. This is an interesting topic and the paper also uses a unique set of data (long-term large-ensemble three-model simulations), however it requires some major changes before it can be published.

Here are some major comments (some of them also appear later in the list of specific ones):

1. The paper states itself as an 11-year cycle study, however authors do not focus that much on the max-min differences but mostly on the 45-year means, which is rather representative of the grand minima type of variability.
2. The paper mostly relies on the correlation analysis and seems to not exploit the full potential of the available data. Two out of three experiments are used only for a little bit, even though they are mentioned multiple times throughout the text in descriptions. In fact, I struggled to see where the LOWFREQ one is used at all.
3. The main findings of the paper are straightforward (higher signals in the lower latitudes and higher altitudes, stronger signals during active phases, model dependent top-down mechanism, and potential causes for it in the model climatological biases), however the way the analysis is created to show this is overcomplicated. There are 12 figures in the main text and 9 in the appendix, while most of them (and related text) show statistically insignificant results or pure empty space, and the text discussing them is way too technical. It looks more like a report rather than a paper aimed at conveying a clear message and findings to the reader. All this can be substantially shortened.
4. In the same direction but specifically about the figures: the amount and use of figures needs to be critically reviewed by the authors in a way that figures serve the story and not that the text just describes many similar plots. Also, the choice of Figures in the main text vs the appendix is sometime confusing. For example, you have almost half of a big paragraph discussing Figure A2, while Figure 7 shows the same as Figure 6 but at a different level and is mentioned in just one sentence. Figures 9 or 11 are just full of empty space.

Specific comments:

Title: you analyze only the NH extended wintertime, consider reflecting it in the title somehow

l25: "upward" -> "upward-propagating"

l35: comma is missing after "The "top-down" mechanism"

l50: "much larger than the" -> "much larger than of the"

l53: "the dynamics and the uncertainty of the model" -  sounds too vague, please rephrase

l63 and l65: "upper middle atmosphere" and "lower middle atmosphere" sound odd, as you haven't defined the middle atmosphere boundaries and its lower and upper parts. Consider using "upper stratosphere/lower mesosphere" and "lower stratosphere" instead.

Section 2.1: It is worth noting that all three models used are of the ECHAM family. There also have been papers intercomparing ECHAM5 and ECHAM6 GCMs, as well as the performances of their (original and modified) radiative transfer schemes, which would be useful in the interpretation of the SWHR and temperature signals.

l93: Explain what is SOLCHEK, it is not described anywhere.

l104: Describe what ozone is used in MPI-ESM-HR and how it is treated in the three experiments

Section 2.2: You list three experiments together, and it is expected therefore that all of them will be heavily used during the analysis (e.g., as differences between them). However, the FIX and the LOWFREQ ones are used only for a tiny bit. Please highlight that the paper mostly relies on the FULL one and the others are used only for small specific purposes. It is also unclear why you don't look at the classical differences between the experiments and just rely on the correlations of filtered data instead, given also how much statistics you would have with all the ensemble members. FULL-FIX would give you long-term trend + 11-year signal, while FULL-LOWFREQ would give you the 11-year signal, i.e. you would be able to extract both the long-term variability and the 11-year signal without a need for correlations or multi-linear regressions. You are free, of course, to choose what to analyze and what methodology to try, but to me it looks a bit like a missed potential!

l125: how do you justify a 45-year running window? It basically gives you an average over 4 adjacent solar cycles, but it also greatly decreases the overall signal, given that you average solar min and max years together. This also contradicts the title and the main motivation of the paper, which were stated for the 11-year cyclicity. In your case, for most of the paper you rather explore the long-term variability, i.e. your results are much closer to the grand minima impacts set-up than to the question of how the 11-year cycle modulates the atmosphere, even though the mechanisms are similar.

l138: why do you use 90% here and not 95% as for the correlation coefficient?

l147: "in the tropical region" - it not only the tropics, but all sunlit regions with the strongest effects in the tropics. Please rephrase, otherwise it reads like the effects are present only in the tropics

Figure 1: Please simplify the x-axis title (e.g., why do you need "10.7cm" there?). "F10.7 index (sfu)" or "Solar radio flux (sfu)" would suffice.

L151-152: Why do you use annual mean for SWHR and T but DJF-mean for F10.7?

l159: why do you use 10hPa for ozone? According to all previous studies it maximizes rather around 5 hPa, i.e., between 1 and 10 hPa or 35-40 km depending on a study (Maycock et al., 2018 https://acp.copernicus.org/articles/18/11323/2018/; Ball et al., 2019 https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2018GL081501; Dhomse et al., 2022 https://acp.copernicus.org/articles/22/903/2022/)

l160-165: these periods of non-linearity at f10.7>200 sfu look very interesting. Given that these are not so many data points, why don't you provide the periods specifically? This will simplify potential explanations with other forcings.

Figure A2: Please add uncertainty estimates to your lines (either over the ensemble or over the period)

l169: Again, here it could be useful to discuss the related performances of the radiative transfer codes (e.g., Sukhodolov et al., 2014 https://gmd.copernicus.org/articles/7/2859/2014/; Nissen et al., 2007 https://acp.copernicus.org/articles/7/5391/2007/ and others)

l178-180: It looks like the T-correlation is lower than the SWHR one everywhere and not only during 1850-1920

l192: How can you achieve more than one correlation coefficient between two time-series? What do you mean here?

Figure 2 and further: not clear what does CCR mean.

Figure 2 and further: I understand that the relation gets unlinear under high SDs, but how or why do you get two lines over the same SD periods there? Why the shading so weird there and often completely de-attached from the ensemble mean line, while it is stated as ensemble spread, which suggests that the ensemble mean should be somewhere in the middle… Sorry it is a very confusing plot and you need to carefully introduce it to the reader, given that similar plots are used for the rest of the paper

l197-200: Why do you say that the FOCI correlations are robust if all of them are below your significance threshold? Also, for the other two models most of them are also positive and non-significant, therefore I don’t see how you can contrast FOCI to the other two.

l215-216: I don’t see much of a stat-significant warming at tropical 70 hPa.  Also how would you explain stat-significant areas in the troposphere? Is it an artifact of 90% instead of 95%?

Figure 4: Why the countours get interrupted in the EMAC and MPI-ESM cases?

l217: Note that the 11-year cycle-related lower-strat warming has been heavily disputed (Chiodo et al., 2014 https://acp.copernicus.org/articles/14/5251/2014/; Kuchar et al., 2017 https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2017JD026948)

l238: “statistical top-down propagation” – did you mean statistically significant?

l250-252: do you associate all stat-significant areas in the troposphere as resulting from the top-down propagation? Or only those around the midlatitude jet?

l251-253: you don’t show this, so please mark as “(not shown)”

Figure 7: This figure shows the same as Fig 6, but for a different level, almost nothing in it is statistically significant, and you mention it in one sentence only. Given that the figure occupies more than half of the page, it doesn’t look worthy to have it in the main text.

Figures 8-12: these are big figures with lots of empty space and marginal stat significance. Please find a way to replot what you want to show (signal dependencies on model biases) in a more condensed way. Otherwise it is a waste of space and complication for a reader.

l274: please specify the period used for ERA5. Also, it doesn’t look correct that you compare the ERA5 climatology, shown as one line, to the spread of transient model points and treat it as biases. If you want to show model climatological biases, it will be much clearer to just add a lat-pressure figure to the appendix with models minus ERA5 for T and U. Also, there have been many papers validating these models, so it is necessary to verify if your biases are consistent with those reported in the literature.

l288 and l290: “EAR5” -> “ERA5”

Table 1: “EAR5” -> “ERA5”. Also, “December dT” is rather a title for the whole table, while the first-row first-column place should for “level”

l290-291: how are you sure that the interactive chemistry is responsible here? Either provide some evidence or rephrase as a potential cause, but better still with references (e.g, Chiodo and Polvani, 2016 https://journals.ametsoc.org/view/journals/clim/29/12/jcli-d-15-0721.1.xml)

l295: you say that you use anomalies, however the x-axis units look like it is the absolute values

l306-307: Doesn’t EMAC have the cold bias instead of the warm that you mention?

Figure 11: Again, this is just a lot of empty space, that you want to put to the main text

l328: “relativley" ->  “relatively”

l352: Please explain what do you mean by a composite here (i.e., max-min differences)

General Comments

This study attempts to ascertain reasons for the diversity in modelling results regarding the 11-year solar cycle signal in the middle atmosphere. It does this by examining the solar signal in three different models, starting with the shortwave heating, ozone, and temperature anomalies at 1hPa (the direct solar signals), and then looking at the temperature and zonal wind anomalies lower down in the stratosphere, to see whether any evidence of a 'top-down' transmission of these initial solar signals can be found. When the models show variation in these indirect solar signals, possible reasons for this are examined. 

I think the paper is well and logically structured, the overall scientific treatment is quite rigorous, and the ideas presented relevant to discussions about the atmospheric solar cycle imprint in models. Having said that, I think there are some issues with the paper in its current format which would need to be addressed before publication, outlined below. I would therefore recommend publication with minor revisions.

Specific comments

Please define ECHAM/MESSy (L45)

Section 2.1 (climate models) needs significant reworking to improve readability.  Please define all abbreviations as they appear in the text (e.g. NEMO3.6 L83, JSBACH L83, SOLCHEK L92, T42L47MA L93, MECCA L95, JVAL L94, RAD-FUBRAD L95, UBCNOx L95, GR15L40 L96).

I'm not sure that figure 1 exactly supports what you say in L151-153. I think you need to add linear trendlines to the figures for values below about 150/180sfu. Also, I think the upper limit of this linear trend is possibly lower than 180sfu, maybe 150sfu. Again, a trendline would help to clarify this. 

Figure 1 in general is quite difficult to decipher, given the multiple models plotted and many data points. I would suggest replacing it with line plots, with the lines indicating ensemble-mean values, and a shadow region indicating the ensemble spread, similar to your other figures.

Please explain why the spread in ensemble results above a certain value becomes bifurcated in figures 3, 6,7,  11, as indicated by the shadow regions.

L192 delete ) after ozone

L196 inset ) after Fig 3

Figures 4 & 5: Upon examination, I am concerned that we are seeing some aliasing with the QBO. Most of the subplots in figure 5 do show definite QBO-like equatorial zonal wind anomalies. FOCI is probably the greatest concern because at least some of these QBO-like zonal wind anomalies appear significant (e.g. Feb-Mar at ~30hPa). This could be significant because the QBO state (easterly vs westerly) has been shown to have an influence on the polar night jet strength, i.e. the Holton-Tan effect (see Holton & Tan 1980). Authors should address this concern, ideally by filtering their results for QBO phase (westerly vs. easterly, or neutral).

L249 insert such before both: '...such that both the positive...'

Figure 11: please adjust the x-axis scale range for each model as the scatter plots are coming out too squished horizontally. Consider doing to same for figures 8 & 9; you can always just highlight the different scale ranges in the figure captions/in-text discussion.

I think you should consider redoing all your line-style plots that show ensemble members and ensemble means (figures 1-3, 6-12) so they just show the ensemble mean and spread in the ensemble members, like in figure A3. 

L264 remove quotation marks around "opposing"

L274 add ) after Fig. 8

L287 and L363 change pole to polar

L288 change EAR5 to ERA5 (also in L290, caption and column heading for table 1)

References:

Holton, J. R., & Tan, H. C. (1982). The quasi-biennial oscillation in the Northern Hemisphere lower stratosphere. Journal of the Meteorological Society of Japan. Ser. II, 60(1), 140-148.