This manuscript presents a comprehensive evaluation of the 11-year solar cycle imprints from the middle atmosphere to the surface, especially the so-called “top-down” mechanism, based on 10-member CMIP5 historical simulations from the MiKlip (MPI-ESM-HR). Considering the diverse and debated conclusion on this topic in the existing studies, I think this paper is worth to be published after a minor review. The following are my comments on this manuscript:

Lines 17-18: “……on the North Atlantic Oscillation might lead to ……”, I would suggest writing in a way like: “……on the sea level pressure in the North Atlantic might lead to a synchronization of the North Atlantic Oscillation via the 11-year solar cycle”

Lines 32-33: “……with opposite sign in response……” This also suggests the large internal variability could veil the possible solar imprints.

Lines 93-104: all these studies are based on the same climate model, i.e. CESM-WACCM, of course, different versions, and different forcings. So even using the same climate model, the detected solar imprints or the interpreted results also could be controversial.

Lines 134-135: I guess the sea ice module is included in the MPIOM, right? Please add a few words here.

Line 142: “We focus on the period 1880-1999” Why did not use all model years (1850-2005)?

Line 160: How did you perform the EOF analysis on the ensemble members and the ensemble mean? Did EOF on stitching data (i.e. 1200 model years) or did EOF on each member separately, but keeping the phases (sign in EOF1) the same as the observation? I ask this question because the EOFs could be opposite signs, if performing the EOF separately, need to double-check to make sure the patterns of all the members are in the same phase.

Section 2：Significance level is shown in the lead-lag correlation and MLR analysis results, but the method description is missing. Considering after a band-pass filter, the effective degree of freedom of the time series will reduce a lot, so it’s necessary carefully calculate the EDF and describe the method in this section.

By the way, how did you assess the robustness of the solar imprints?

Line 174: “180 SSN”, here hard to understand this term for a reader not familiar with the MLR, please explain more.

Line 176: “the solar forcing lags the model data”, lags? for investigating the lagged response in climate, should solar forcing lead the model data, right? Did I misunderstand something?

Lines 175-176: Are the other predictors keeping the same as the lag 0 when shifting the solar time series?

Line 207: “period from 1850-1999” the period from 1850-2005?

Line 209: “pretty high minimum” do you mean compared to the SC19? I think the response of SWHR to the minimum of SC20 fits quite well, but a higher response to the maximum of SC20. So, please clarify your statement in this sentence. )

Figure 1: a) and b) are missing in the figure. And I would expect a similar figure as Figure 1, but for temperature in the tropical mean (supplementary figure or panel into Figure 1), this figure will nicely show the sensitivity of the temperature response to the solar cycle amplitude. 

I also suggest plotting the original SSN time series and the scaled SSN time series used in MLR (180 SSN?), as a supplementary figure.

Line 219: “a robust” How did you define this?

Line 223: “a relative short time series of satellite…” maybe provide a specific time period? I think this period probably includes different solar cycles from the simulated time series. Repeating the model analysis in the same data period as the satellite would help to compare

Figure 2:  A very brief caption. Please add more details. Like, is it a result of MLR? or a composite of annual mean zonal mean temperature in the solar maximum years? If it’s the latter, how did you define the solar maximum years (maximum of SSN of each solar cycle I guess)? What does it look like in the solar minimum?

It seems the second warming is absent in your simulations.

Section 4: a very long paragraph, that needs a break somewhere.

Line 256: What does it imply, if the SSN is below the SC14 maximum? How strong the solar cycle could be if it’s above the SC maximum? I think a figure of the original SSN time series will help to give the reader directly the impression.

Line 266: please add a reference for this statement.

Line 270: As the “dynamical response of the PNJ” is directly related to the meridional temperature gradient, I’m curious, can the solar-induced TST warming really increase the poleward temperature gradient? I think it’s not the case for some ensemble members, at least for the EM4.

Figure 4: same as Figure 2, the caption is too brief. Is it a result of MLR? MLR performed on ensemble mean or averaged regression coefficients of all the members?

Horizontal components of the E-P flux are hard to see in the bottom row of Figure 4, is it due to a very less reflection? Maybe scaling it to show more clearly.

Line 339: “insulation” Do you mean “insolation”?

Lines 370-371: I think the temperature response in EM4 is very different from EM1, it’s almost the opposite state in the Pole region.

About Figure 5, no significant temperature response in the tropical stratopause and no response of the lower mesospheric subtropical jet in EM4. Is the warming migration to the high latitude in EM1 due to solar insolation? If yes, why it disappears in EM4?

Line 384: “Drews et al., (2022)” => “Drews et al. (2022)”, they used a different climate model (CESM-WACCM)

Lines 413-415 and Figure 6: did a “top-down” mechanism show in observational data? An anomalous zonal mean zonal wind in the troposphere and surface in Feb?

Line 444-446: I guess that’s why we need some positive (slow) feedback from the ocean. But I’m not sure your model can simulate this. By the way, if the “top-down” signals are not robust, how can we expect the surface response?

Figure 7: I suggest adding the surface wind or surface zonal wind in Figure 7. Same comments on the caption as above.

The spatial pattern of solar imprints in SLP may have different active centers from the EOF1 (NAO). Could you please compare them? Is the EOF1 (NAO) in your model the same as the observation?

Lines 458-463: this should be in the method section.

---best,

Wenjuan

Spiegel et al. examine the impact of the 11-year solar cycle on decadal predictions, by using a large data-set from the MPI decadal prediction system. The length and size of their ensemble experiments offer an unprecedented opportunity to critically evaluate the possibility (suggested in many other studies) concerning a possible role of the solar cycle as a source of predictability (via the "top-down" mechanism). This is a very controversial and yet relevant topic, given the potential implicatuions for decadal prediction of boreal wintertime climate in the North Atlantic. This paper nicely shows that the top-down signal in the MPI model is small and there is hardly any synchronization of the NAO. In particular, the paper also debunks the possibility raised in another recent paper (Drews et al., 2022) that the time-dependence of the solar/NAO signal highlighted in Chiodo et al. (2019) might only arise in decades with a stronger solar cycle. The paper nicely shows that indeed, the modeled response is weak and mostly not significant and indeed, the main conclusion against a solar modulation of the NAO and synchronization for their model is absolutely warranted. I think that the paper will contribute to the scientific debate on this matter and deserves to be published, after some corrections as detailed below.

GENERAL COMMENTS

- The main conclusions of the paper are supported by the evidence provided in the paper. However, it would be good to at least acnowledge and/or discuss the possibility that the MPI-ESM-HR model system, just like other model systems, might suffer from the "signal-to-noise" paradox issue, according to which, the models might underestimate any externally forced signals in dynamical aspects of the climate system (see e.g., Scaife and Smith, https://www.nature.com/articles/s41612-018-0038-4 ), such as the PNJ and stratosphere-troposphere coupling.

- It might be also good to at least discuss the implications of the part of the solar forcing which is missing in these runs; the particles forcing (MEE/SEP). I would argue that particles are unlikely to change the picture and the EPP signal is perhaps even more complicated than the irradiance/UV signal, and the models are at their infancy in simulating this forcing... but it needs to be at least discussed, given the number of related publications arguing that EPP play an important role in the solar influences on the polar vortex/ NAM and NAO. The other general point concerns the lengthy discussions of the individual months. I think many of the parts of sections 5-6 could be shortened, especially if there's no robust signal to document.

-Lastly, the paper examines the top-down mechanism in detail and rules out a robust effect in their model... but what about any bottom-up impacts? They should be at least mentioned somewhere in the discussion, to put this paper in the broader context. The existing literature is cited but there could be more discussion of the comparison with Chiodo et al., 2019 and the aspects that are new here (i.e. testing the link between the time-dependence of the signal and the magnitude of the solar cycle amplitude).

SPECIFIC COMMENTS

L17-19: could add that previous work didn't really do a decadal prediction set-up --> rather, continuous long-term climate runs!

L20 - it's more than just "confirmation" from other modeling groups - a cleaner model study is really missing...!

L22 we aim for an unbiased evaluation --> "unbiased" seems a bit too harsh and indirect criticism towards previous studies --> replace with "objective and improved"...?

L30 is rather weak -> suggest removing "rather"

L31 remove "basically"

L36-37 unclear what the driver of the "anomalies" is - a bit more clarity would be needed. Would these be ensemble mean anomalies (or anomalies in the individual ensembles) that are correlated with the solar cycle, or what is the "driver" of these anomalies..?

L36 "most likely independent from the seasonal march in the middle atmosphere" -> would this mean that the timing of the tropospheric anomalies does not match the downward propagation of the (apparent) solar signals? If so, I'd reword this to use the word "timing" so that it's clearer to the reader what we are talking about.

L38 "rather sporadically than in a systematic way" -> sporadic in that they depend on the ensemble member, or sporadic in that they depend on the time window being analysed? again, more specificity might be appreciated.

L40 "might rather be interpreted" - I'd repharase to "might be". Just remove "interpreted" to simplify the wording of this part of the abstract. 

L41 "between the solar forcing" - it might be good to again be more specific and say that the paper handles the UV/irradiance forcing specifically... as there might be also other pathways for a solar influence on climate, namely via MEE/particle forcing. Just add "irradiance" or "UV" to "solar forcing"

L40 "as a statistical artefact" - it may well be that the synchronization is simply due to internal dynamics of the atmosphere...? if that's the case, then it may not be a "statistical artifact" in the strict sense, but rather, that it's not solar driven...

L73 "...convergence in the Eliassen-Palm flux (EPF)...positive wind anomaly" a convergence in EPF would result in a deceleration of U. I guess what is meant here is thus the "divergence" of the EPF?

L77 - Discussion of Matthes 2004 - I'd be more specific here and say that this study specifically studied the evolution of the signal on sub-monthly time-scales

L81 "from February on" -> from February onwards

L83 "very individual temporal progressions" - again, I'd use the word "timing" to clarify what is meant here. E.g. you could reword this to "the exact timing of the downward propagation depends on the individual study"

Side note: I'm not sure Marsh '07 really show a downward solar impact - rather, they show, by means of a time-slice experiment, that the signal is robust in the upper stratosphere but in the polar regions, their signal is rather weak and non-significant...

around L90: it might be good to add somewhere here that there are also other potential pathways for a solar influence on climate but the irradiance-UV component is the most studied one and the one that models have an easier time simulating...?

L110 I think you could also add here that the availability of the large ensembles with the observed time-varying solar cycles allows you to explicitly test the dependence of the signal on the solar cycle amplitude (hypothesis put forward by Drews et al 2022).

L209 I'd suggest removing unscientific terms such as "pretty". 

L219 replace "receive" with "obtain"

Figure 2 seems a bit blurry and the individual components of the figure (e.g. colorbar and right panel) seem to have been manually assembled together. I'd suggest including a high resolution version of this figure and use another software to produce the figure without having to manually assemble the individual panels. This also concerns the other Figs. 3-8.

Figure 3 nice figure, but I have the impression that the data is not uniformly scattered in the x-dimension (Sunspot number) - the dots are all structured in "vertical bands". Why is that happening? Is the SSN sampled only every 5-10 units...? Also, do dots correspond to each individual member? 

L410-440 (and intro) I think it would also be good to add the most recent paper by Gray 2016 to this discussion... and in particular, Ma et al., 2021 - who argued that the early-winter signal is the most robust component of the signal. 

 

L441 I think you can be more confident here and replace "assume" with "conclude"

L444-447 What about the "bottom up" signal? I agree that much of this discussions disqualifies the top-down... but can we also rule out a role of any direct surface (TSI-driven) signals?

L544-546 Could the authors provide some evidence to support the interpretation that the background state determines if the signal is transferred or not?

L550 remove "time series"

L551 not sure what is meant with "manifold phase relations". can the authors clarify?

L563 "...made in model development.." agreed that having larger ensembles helps diagnosing robust signals and probably that's partly the reason why signals in older studies were deemed as "solar"... on the other hand, I do not think that model physics being simpler in those early studies is another reason for the apparently stronger signal. Those models had all the necessary physics to capture the mechanism, in principle. 

L583 I'd add somewhere here that these results agree with the conclusions of Chiodo et al., 2019 - and here, they even strengthen those conclusions concerning the little role of solar forcing in modularing the NAO, as the runs of this paper can even more directly be compared with observations and the solar signal can be tests in an even more realistic set-up (with transient forcings, transient solar cycle, etc.)