Summary: The manuscript focuses on the climate response to the strong volcanic eruptions that occurred in the 6^th century. The authors analyse an ensemble of climate simulations with the model MPI-ESM, compare their results with other previous simulations and with temperature reconstructions based on dendroclimatological data. The analysis of model data include the atmospheric response (temperature, precipitation, wind) and the response of the ocean circulation in the North Atlantic. One important issue is the duration of the response of these variables, as previous simulations and decollimated reconstructions seem to disagree - at lest in some regions.

Recommendation: In my opinion the scope and focus of the article are valuable. There are some knowledge gaps about the response of large volcanic eruptions and the origins of Little Ice Ages in the last millennia, that need to be filled.

However, the manuscript leaves clear room for improvement, and I believe it requires considerable revisions. The structure is sometimes confusing, the language is also not always clear, and the data analysis is not deep enough. This is reflected in the unclear take-home-messages of the manuscript. Disagreements between simulations are explained by general model differences. Disagreements between reconstructions and simulations are explained by possible errors in the volcanic forcing or deficiencies of the proxy data, mainly in the tree-ring data, but the manuscript does not include more specific and solid explanations. Sometimes, the discussion is inconsistent, and some example of this are given in the list of particular points below.

1. ‘ sea level pressure and a decrease in hydrological variables occur’

The text could be here clearer. Does this sentence mean that both precipitation and

evaporation becomes smaller?

2. ‘However, most reconstruction data sets go back to about 1200

CE, and the further back in time, the fewer proxy records remain, and the more uncertainties they contain (Masson-Delmotte,2013; Neukom et al., 2019).

3. The sentence is not clearly formulated (perhaps it is not grammatically correct). Consider an alternative formulation like ‘ Further back in time the network of proxy data becomes sparser and uncertainties in each individual temperature record also grow.

4. ‘The aim of this study is to investigate whether a multidecadal to centennial cooling may have occurred in the mid 6th to 7^tth century.’

This sentence states the main objective of the study. However, the reader will not be wiser after reading the manuscript. The disagreement between the proxy record that do indicate a centennial cooling do not agree with the model results, but the study does not include a solid explanation for this disagreement (apart from speculating that perhaps the high altitude of the Alp records may be responsible for the long temperature recovery).

This is perhaps my main concern of the manuscript. It is in general too descriptive and does not go deep enough into explaining those disagreements.

5. ‘ the short term (years), as well as the long term (decadal to centennial)

the short term (annual)...

6. ‘sea ice impacts, we also study atmospheric and ocean circulation, hydrology and the ocean-sea ice feedbacks in maintaining the climate signal’

the ‘climate signal’ is too unspecific. Please, help the reader by being more specific, for instance ‘in maintaining the volcanic induced cooling’

7. ‘For this study, we ran ten ensemble members for 160 years from 520-680 CE.’

for 160 years, starting in 520 CE. Or alternatively, covering the period 520-680 CE

8. For each ensemble member the atmospheric diffusivity was changed by 1 · 10−5 to simulate slightly different climate states by the year 536 CE, the year of the first large volcanic eruption.

physical units are missing

9. Historical Land Use Data Set for the Holocene (HYDE3.2, Klein Goldewijk, 2016). Considering several options (e.g. linear ramp-up) we decided to simply let the land-cover data be constant for the first 850 years of the past2k runs.

We prescribed a constant land-cover for the first 850 years of...

10. 'The tree-ring sites are displayed in Fig. A1. For the model-tree-ring comparison a land mask was applied to the model 2m air temperature analyzing the NH extratropics between 40◦ and 75◦ N.'

Here and in other instances in the manuscript, it is not clear whether the reconstructed NH temperature was just calculated as the to be the simple average of the local temperature reconstructions at the tree-ring sites or whether there was a more sophisticated reconstructions method, for instance by calibrating a statistical model to replicate the NH mean temperature ( as in Stoffel et al. 2015). The present manuscript lists in Table 2 just 6 records. Is the NH temperature the average of only these 6 records or is it the temperature reconstructed by Stoffel et al. ? I think that the simple average of these 6 records cannot meaningfully be considered a Northern Hemisphere average.

11. 'Towards the end of the simulation period the ensemble shows a larger spread than at the beginning of the simulations, which corresponds to the ocean heat content state being more different between members in the end than at the beginning of the simulations.'

Could this be an indication that the ensemble set-up is not adequate to investigate the main objective of the study, and that the spread of ocean initial conditions is too narrow ?

12. The results section includes several paragraphs that actually would belong to a (missing) Discussion section. An example is this paragraph:

'Zhong et al. (2011), and Miller et al. (2012) argued that the ocean - sea-ice feedback could play a major role in sustaining a century long cooling after a cluster of four volcanic eruptions in the mid 13th century. In contrast to these studies, we simulate a multi-decadal sea ice response in the mid 6th to 7th century. '

Consider also this reformulation: In contrast to these studies, our simulated cooling is shorter and lasts only for a few decades.

13. After the 536/540 CE double event, the ensemble mean of the model simulations does not return to zero sea-ice cover anomalies before 560 CE.

the ice-cover in the ensemble mean takes longer to recover and only reaches the climatological mean value by year xxxx.

14. 'Fig 3 (TRW) and triangles (MXD) in the 2 m air temperature maps. The 2σ (1σ) standard deviations for 2m air temperature SLP, evaporation (and precipitation) are stippled (hatched). '

The ensemble standard deviation sigma (2xsigma) ...are stippled (hatched)

15. ‘there is a land-sea contrast present for evaporation in summer, where the signal is opposite over the ocean.’

In summer, the sign of the evaporation anomalies over ocean and land is opposite.

16. ‘the north side of the climatological high pressure systems reflecting an atmospheric circulation separation at around 45◦ N.’

this sentence is unclear. Does it mean that the sign of the anomalies is apposite north and south of 45 N?

17. ‘The long term response is shown in the right side of Figure 3.’

on the right side or in the right half of the picture. Better still is to label all the panels and refer to them accordingly.

18. 'The increase in precipitation over the Mediterranean in boreal summer in the model simulations in this study are related to the shifting of the inter tropical convergence zone (ITCZ) into the Southern Hemisphere (SH) after the eruptions (not shown here), as well as a weakening of the high and low SLP over the North Atlantic (Figure 3b). After a large volcanic eruption, the ITCZ shifts away from the cooler hemisphere, in this case the NH (Schneider et al., 2009). '

I guess that during the boreal summer summer , the Northern Hemisphere is the warmer hemisphere, not the cooler hemisphere. Perhaps the authors mean that the negative temperature anomalies are stronger in the NH?.

19. ‘appeared to be opposite, with a drying over Southern Europe and a wettening over Northern Europe. They accounted this to the models not capturing the winter NAO well and therefore simulating a different response.’

The reference to Iles and Hegerl to discuss precipitation response in boreal summer over Europe is misplaced. Iles and Hegerl refer to the winter precipitation response, so it is not correct to state that Iles and Hegerl agree with the results obtained here for summer precipitation. It would also strange to claim that a wrong simulation of the winter NAO response can explain the wrong sign of summer precipitation anomalies in Europe.

20. 'The significance was calculated from the 1200 year control run by

taking the 2 to 20 year means of the 1200 years, and then taking 4 random time steps from that time series for the 4 large eruptions. This was done 1000 times for each variable, and the standard deviation was then calculated from those new random time series. 1 time or 2 times the standard deviation (1 and 2σ) were then used to calculate the significance.'

I cannot understand how the standard deviations were calculated. This paragraph seems to me rather unclear

21. 'In boreal winter, there is a see-saw pattern visible in the 2 year SLP response with an increased low pressure over Greenland and a decreased low pressure over Northern Europe, corresponding to the seesaw winter temperature pattern between Greenland and Scandinavia, as described by Van Loon and Rogers (1978). The changes in boreal winter reflects a positive Arctic....'

This whole paragraph is rather unclear. I think there is an error in the first sentence (increased low pressure in both Greenland and Northern Europe?) , but in general it is difficult to follow. It contains a mixture of own results and previous results, and it is difficult to disentangle which is which. After reading the paragraph, it is unclear whether the model does produce a NAO response or not.

I would recommend to first describe the new results, and then briefly compare them with previous results. In a discussion section this comparison can be then deeper and more detailed.

22. ‘The summer cooling over the continents can have a serious effect on the vegetation and society summer can lead to crop failure and famine in areas that are close to the temperature limit for grow....

This paragraph on the impacts of low temperatures is misplaced here. It is not related with the previous or the following paragraphs. This could go to a Discussion section or be deleted without any loss.

23. From Fig. 4b and c can be seen that the subpolar gyre (south of’it can be seen...

24. I found the following two paragraphs to be inconsistent with the ensemble set-up:

‘The studies from Zhong et al. (2011) about the onset of the LIA also concluded the response to be depended on the initial state of the North Atlantic, as only 2 out of 4 simulations (one warm and one normal NA state) lead to a cooling long enough to resemble the LIA. Compared to their study, our NA state is relatively warm, but it is hard to compare as a different model and set up were used.’

‘be the initial state of the ocean when the volcanic eruptions occur. This is less likely, as 10 ensemble members were run, which showed a range of variability in the same range as the 0-1850 CE variability, where the response to the volcanic eruptions’

The first paragraph states that the initial ocean state is important for a successful simulation of the LIA. The second paragraph indicates that the ocean initial conditions in the ensemble are wide enough separated. However, the initial conditions were prepared by just perturbing one single parameter in the atmosphere submodel, and else where the manuscript states that the spread of the ensemble at the end of the simulations is clearly larger than at the beginning, indicating that the initial ocean states are not that much separated. This requires a deeper discussion.

25. ‘For the model - tree-ring comparison, the model temperature anomalies were taken for grid cells corresponding to the latitude/longitude range for the tree-ring locations. In Fig. 5 the comparison for the NH, the Alps, Altai and Northern Scandinavia are shown.'

Again, it is unclear how the reconstructed NH temperature has been calculated.

26. ‘The temperature anomalies from the model simulations and the 2 sigma variability range fall within the 2 sigma variability of the NH of the model simulations and the timing of the peak cooling after the four large volcanic eruptions agree very well. ‘

I guess the authors mean that the reconstructed NH temperature fall within the model ensemble spread ?

27. ‘Figure 5b shows the model - tree-ring comparison for Northern Scandinavia (NScan). Just as for the NH, the variability of the model simulations fall within the variability of the tree-ring temperatures. ‘

see previous comment

28. ‘agree very well both in timing and in signal. This could be because the tree-ring data for Fennoscandinavia consists of MXD data, so there is less time lag and smoothing in the signal (Esper et al., 2015). More deviation is visible for the ensemble mean peak cooling for the 574 and 626 CE eruptions.’

This is an example of, in my opinion, cherry-picking results. It seems that for some of the eruptions the MXD data agree better with the model because the wood density proxies better represent the true temperature. However, this is not the case for the other two eruptions. Why?

29. ‘because the volcanic forcing in the model is overestimating the cooling in the mid-latitudes. ‘

because the prescribed volcanic forcing is too strong

30. ‘The concept of a LALIA period was raised by Büntgen et al. (2016), based on tree-ring data. There is a good agreement between the tree-ring temperatures and the model temperatures after normalization. This was done with regard to the time'

In my understanding, there is no normalization involved here. The records have just been re-aligned to a common mean, but they have not been re-scaled to a common standard deviation.

31. ‘Perhaps the century long lasting cooling may be only apparent in the Alps and Altai tree-ring records, as the cooling is a local feature occurring at high altitude of the mid-latitudes. Our model resolution is too coarse to fully capture the topography’

This is speculation. It needs to be more strongly supported.

32. This comment may be more a matter of taste but I find that the Summary and Conclusion section is rather repetitive of what has been just exposed in the Results sections. The summary, if the authors wish to keep one, can be considerably shortened - there is no need to repeat all results in detail again. On the other hand, the real conclusions, starting in line 443, could be more sharply written e.g. that none of the simulations of the ensemble reproduce a century-long cooling, but that this could be due to a too narrow choice of ocean initial conditions, that the most important feedbacks mechanisms for multidecadal cooling involve sea-ice cover. and that the findings here do not agree with some of the palaeoclimatological data, but do agree with other dendro data .

This is an interesting paper showing new model based information on seasonal climate response following major volcanic events during the 6th and 7th century. In addition, the paper shows also proxy based reconstructions for different regions of the NH and a few simple model/reconstruction comparisons.

There are a couple of shortcomings, gaps and weak parts of the manuscript that need to be revised, I thus recommend major revisions

Comments (in no particular order):

Abstract

The abstract needs a major revision, currently it is a compilation of individual results without a clear statement. There are statements about proxy reconstructions, model output and model-data comparison, but there is no connection between the sections and also no dynamic explanations of what was found and new conclusions about climate during the 6/7 century. The first sentence is confusing as there is no actual detection/attribution study (yet) to support the finding that volcanoes are the main cause of cooling and not natural variability or a combination thereof. Further, several palaeoproxies are mentioned, but it is not clear what they refer to, whether they are used to date volcanic eruptions or to derive past climate.

General manuscript

It would be interesting and valuable to have a paragraph on the socio-economic impact of the cooler and more variable climate during this period. This would be a nice link to the other sections of the paper.

The paper is not up to date with the latest publications and the introduction needs to be revised accordingly. The first paragraph of the introduction is incomplete, confusing and needs better focus and the inclusion of more appropriate and new references to the state of the art in palaeoreconstructions for the study period. In general, the paper needs to be updated with recent findings related to reconstructions, post-volcanic responses in the palaeocontext, regional interpretations, and data/model comparison studies. Furthermore, the paper also needs considerations on uncertainties. In addition, two studies from 10 years ago on the onset of the LIA are cited (lines 39 ff). They are superseded by new findings.

It is not clear why Stoffel et al., Büntgen et al. and Esper et al. are used for comparisons with the model (Fig. 5). New reconstructions are available (see Büntgen et al. 2021 Nat. Comm for a review) and possibly an ensemble series could have been used instead of single reconstructions, which do not reflect the true NH conditions but are locally biased. It is also not clear why the authors compare the grid-based model output with local tree-ring-based reconstructions for three regions. In lines 390ff. they mention that such a comparison can be misleading, so they might reconsider this section. A more appropriate comparison could be made with continental reconstructions for Europe (Luterbacher et al. 2016, Env. Res. Lett) and for Asia (Zhang et al. 2018 Nat. Sci. Reports, 8, 7702).

The methods and associated measures to compare reconstructions and model output need to be explained in more detail.

The authors report on the summer precipitation behaviour in the Mediterranean region in the model world. This part needs to be revised, as in reality there is hardly any precipitation in the warm season and if there is, it is mostly on the northern rim. Even in post-volcanic summers there is no clear signal in observations and reconstructions of the last centuries (Wegmann et al. 2014; Fischer et al. 2007, CRL). Please note that there are hydroclimate reconstructions from different areas of the Mediterranean with which the model output can be compared.

In general, the work has a bias towards summer, which is not surprising as the tree-ring reconstructions resolve summer conditions. However, it would be important to provide some more insight into the conditions during the cold season and how the volcanic influence could change the annual cycle after the short and decadal volcanic influence.

Section 3.1. volcanic response:

It seems that the Figures 2,a, c and d are not commented and interpreted in the main text. Are the time series in Figure 2 referring to summer? The plots are small and details cannot be seen. Please could you increase the readability of the figures in general, thank you.

It is not entirely clear what Figure 3 shows. Are all post-volcanic years during the study period averaged and shown in relation to pre-volcanic non-volcanic conditions? Please state this more clearly in the caption and main text. Also, please provide more explanation of the multi-decadal analysis and how it is carried out.

In Figure 3b I see a strongly negative NAO with higher absolute pressure in the subpolar regions and lower pressure in the subtropics. This does not seem correct given the independent evidence of a strongly positive NAO following strong tropical volcanoes.

It might be good to show the statistically significant areas and instead use a field sign test to show which areas are different compared to the reference period.

Please consider the interesting paper by Moreno-Chamarro, et al. 2017 re: Winter amplification of the European Little Ice Age cooling by the subpolar gyre. Nature Sci. Reports, 7, 9981. It is not about the 6/7th century, but about an active period during the Maunder Minimum and the role of internal variability versus forced influence on European seasonal climate. It might be worthwhile to review this publication as well to see whether similar processes might have been in effect and to include it in the interpretation.

Summary and conclusions

This part needs to be shortened and cleared of duplications. It is more a listing of some results without clear connection and explanation and what the main conclusions are from this study.

Fig. A2: Please provide the units of a) and also how the NAO indices have been calculated

This paper develops and analyzes an ensemble of climate model simulations covering the period of 4 large eruptions in the 5th and 6th century as well as the decades following.  These modeling results are really very exciting because they provide brand new and much needed insight into the potential behavior of the climate system following large and important eruptions (including two closely-spaced eruptions in the 6th century) in the first millennium and the potential for new paleoclimate proxy/model comparisons of this important but still sparsely known period of the Common Era.  These results will therefore be of great interest in understanding the range of responses to volcanic eruptions and relevant for both modelers and proxy paleoclimatologists. 

My primary critical observation is that the manuscript is excessively descriptive.  I think a stronger manuscript would result from placing the model (and the proxy-model) comparison directly in the larger context of uncertainties about first millennium climate and response to large eruptions, isolating the model behavior that is most interest to understanding this unique time period, a more complete treatment of the proxy data, and more accounting for uncertainty (the real strength of the model ensemble is the range of possibly behaviors of the ocean-atmosphere system that can be observed and used to quantify variability in the response).  My major comments below are mostly about these issues, and some minor comments and suggestions follow.

Major comments: 

Abstract and Introduction:  I think the paper (and readers) would benefit from a re-writing of the abstract and a reframing of the importance of this work.  As it currently stands, one of the particularly interesting aspects of the 6th/7th century eruptions (that persistence or not of the cooling and therefore the duration of the 'Late Antique Little Ice Age') is mentioned only rather vaguely, and much of the abstract is given over to a simple description of the model phenomena.  A stronger abstract would set the stage for the proxy disagreements (and note the sparseness of the proxy data as exacerbating the difficulty in understanding these uncertainties) and then frame the results in terms of this as well as the importance of understanding large events like the closely spaced eruptions of the 6th century.  Likewise, the Introduction would benefit from using more recent publications about the Little Ice Age as well as a more structured framing of the uncertainty and motivation for the study.  

Initial discussion of proxy data (Lines 53 to 59).  This section needs to be enhanced, as an appreciation for the uncertainties and causes of uncertainties for first millennium climate reconstructions, particularly with the resolution needed to resolve volcanic eruptions precisely, is important for the comparisons that come later in the paper.  While Ahmed et al. 2013 (which is properly PAGES 2k Network 2013) was a distillation of temperature proxy data for the last IPCC, it is superseded by a number of papers, including PAGES2k Consortium 2017 (Emile-Geay et al. 2017).  The authors might also want to cite Esper et al. 2018 (doi is 10.1016/j.dendro.2018.06.001) which analyzes tree-ring proxy uncertainties in the early part of the last millennium (and therefore these uncertainties will be even greater in the first millennium of the Common Era).  Particularly here: there is a substantial body of literature now (some of it discussed later) about the ability of different tree-ring proxy measurements to resolve or 'smear' volcanic cooling - MXD vs tree-ring width.  Similarly, multiproxy approaches that mix seasons, hemispheres, or are low resolution might not resolve the volcanic signal or may require additional post-processing of model data to make an appropriate comparison.  Since the source of possible uncertainties in proxy reconstruction of 6th/7th century climate is important for the comparisons that come later, I think a more thorough and up-to-date discussion is warranted here in the introduction.

Proxy data (Section 2.2): As above, I think a more complete and clear description of the proxy data here would be useful for later in the paper when comparisons become important.  Table 2 lists the individual proxy data that are available (this is good to have this, since the representation of tree-ring width and MXD can sometimes be subsumed when using a reconstruction), but the wording in Section 2.2 is confusing - for instance, what does 'The first four sites combined are the "NH land" compilation by Stoffel et al. (2015)' mean?  Does this mean that the 4 sites listed first in the Table were also used by Stoffel?  This isn't clear. By the time one arrives at Figure 5 and the associated text, it isn't clear what/which of each of these proxies is going into the comparison, so a more thorough discussion of the proxy data used and what each reconstruction in Figure 5 contains is necessary.  The following line say 'The data sets 135 contain a mix of tree ring width (TRW) and maximum latewood density (MXD).' and this is true, but only the NSCAN MXD data are available for the 6th and 7th century - the rest are tree-ring width and subject to the potential problems described in the following lines.  Again, this section seems rather sparse and is not clearly organized, and yet limitation of the proxy data (or their particular time series properties) will become important later in the paper.  Since the authors prepared this manuscript, there as been a new ensemble reconstruction of Common Era temperature (Buntgen et al. 2021, doi is 10.1038/s41467-021-23627-6) - while I realize this paper actually came out after this manuscript was submitted and the authors cannot have been expected to use these reconstructions (of course!) I would encourage them to at least consider them (formally analytically or informally as comparison, since the early part of the LALIA is examined in the paper), at the discretion of the editor.  Finally, it would be worth I think mentioning why multiproxy reconstruction like PAGES2k, LMR etc. are likely to be unsuitable for this comparison and why the authors rely (and rightly I think) on tree-ring data. 

Proxy-model comparison: This section is  unclear in places and speculative without support in others; for instance, (Line 355) I'm not sure what 'The temperature anomalies from the model simulations and the 2 sigma variability range fall within the 2 sigma variability of the NH of the model simulations' means?  I also find it to be too qualitative - what does 'good agreement' mean and how to measure it?  In Line 362, 'More deviation is visible' is also vague.  I think this section would benefit from a more straightforward and quantitative exploration of the proxy-model comparisons.  In Line 365, I'm not sure how something could be both 'less good ... but still remarkable'?  However, also the full range of the model ensemble should really be considered in the comparison - the 'real world' is just simply one iteration of what could have happened under different initial conditions, forcing uncertainty, feedbacks, and interactions and stochastic variability.  So the comparison is not simply to the multimodel ensemble mean or even peak cooling, but taking into account the full range of ensemble variability and seeing the tree-ring data as one 'ensemble member' of possible actual and physically plausible atmosphere-ocean states.  

Later in Line 376, the authors write that 'There is a good agreement between the tree-ring temperatures and the model temperatures after normalization' - but again this lack of precision doesn't do justice to the comparison - indeed there appears to be reasonable association fr the major eruptions for NH temperature from Stoffel (including some MXD) and the NSCAN MXD, but for Alps and Altai the lag recovery is longer.  So simply saying there is a good agreement masks interesting differences. In Line 381, this seems very highly speculative: 'could be due to the prescribed volcanic forcing in the model, and that the 547 eruption might have had a stronger impact on NH land than the model simulates.' - why wouldn't the same apply to Stoffel or NSCAN then?  There would need to be some support for this to claim it as a source of the discrepancy.  On Line 390, again this seems highly speculative: 'Perhaps the century long lasting cooling may be only apparent in the Alps and Altai tree-ring records, as the cooling is a local feature occurring at high altitude of the mid-latitudes.'  Again, in Line 395 the authors speculate that 'The change in hydro-climate corresponds to the soil moisture availability for the trees and thus could have impacted tree ring growth', but again this is just speculation, and indeed for the Alps, which have the longest lag at odds with the model, the 20 years summer precipitation anomaly (Figure 3c) is positive and the winter signal is mixed. 

Potentially the most parsimonious answer is that tree-ring width has a tendency to increase the 'tail' of the post-volcanic cooling and change the timing of recovery to baseline.  But the authors give this only a brief mention in Line 371.  

Summary and Conclusion: This section is largely a restatement of the paper, but would be stronger with a distillation of the main points of the article and major conclusions. 

Additional Comments:

Line 12, Line 13 'land–sea'

Line 50: perhaps 'multidecadal cooling, as has been reconstructed by Buntgen et al. (2016).'

Line 55: 'Common Era'

Line 101-102: this sentence seems out of place?  'A common paleo-model set-up is to use 1850 pre-industrial conditions, due to model simulation limitations'

Line 103: This is confusing as written - but why use standard deviations instead of the temperature deviations in the bootstrap? 

Figure 1: are these data from Jungclaus et al. and Toohey and Sigl?  Best to include a citation with the figure caption as well

Table 1: I assume the months (January) are assigned because the actual month of the eruption is not known for these?  It would be worth mentioning this (and some of the potential consequences, e.g. Stevenson et al. 2017) in the methods section as well

Line 152: '2 K' - relative to which baseline?  I presume the 521-680 CE mean mentioned in Line 142, but please clarify

Line 157: 'decreased for ∼20 years' - it is difficult to see this in Figure 2 because of the scale - can you provide a range of the actual return to baseline periods? Particularly since the closely-spaced 536/540 eruptions would be expected (I think) to collectively show a longer recovery time than the individual eruptions in the 570s and 626 event

Line 169-175: Some other more recent papers to consult (and cite as appropriate) might include Lehner et al. 2013 and Slawinska and Robock 2018

Line 270: Perhaps useful to consult Fischer et al. 2007 (doi is 10.1029/2006GL027992, 2007) and Rao et al. 2017 (doi is 10.1002/2017GL073057) with respect to Mediterranean (and Europe) response to eruptions in historical and paleoclimate data

Figure 5 caption: 'Fennoscandia' ?

Line 372: as well as estimate of the forcing, spatial distribution of AOD anomalies, feedbacks, uncertain in timing of the eruption (Stevenson et al.) - so, lots of potential uncertainties.  

Line 377-: I'm not sure this requires any further explanation - the models and data have different references periods and likely different means, but what is of interest is the volcanic signal, so a renormalization isn't that remarkable 

My review of submission cp-2021-49 by van Dijk and co-authors focuses on the paleoclimate part of the paper and not on the models, as the latter are not my field of expertise. In their contribution, the authors seek to answer the question whether long-lasting cooling occurred over the Northern Hemisphere following a cluster of large 6^th and 7^th century eruptions which occurred in 536, 540, 574 and 626 CE according to sulfur deposits in bipolar ice cores. The authors do so by comparing proxies (mostly tree-ring reconstructions) with model output.

The idea of the paper is nice, but I have a few major comments that shall be addressed by the authors in a revised version:

• The title is misleading, no proxy record has hitherto posited that the 6^th century eruptions would have been at the origin of a long-lasting hemispheric cooling. Instead, a study based on data from the Alps and Altai (by Buntgen and colleagues, 2016) has just pointed to marked cooling at these two sites. Other tree-ring records do not suggest a comparable cooling. Speaking of hemispheric cooling is thus an overstatement and should be changed.
• Along the same line, starting from line 32ff the authors state that cooling might have exceeded that of the LIA and focus on two site chronologies that were presented in 2016. While the authors rightly present the results of this study, and add the reply provided by Helama and colleagues from 2017, they ignore a vast body of proxy studies that have been published on the topic and where the chronologies cover many sites of the NH. By focusing only on the LALIA study, they ignore a large body of spatial and temporal reconstructions covering the period of interest. The authors should therefore present a more balanced assessment of the existing data by including e.g., Schneider et al. (2015, ERL), NTREND (spatial and temporal; Wilson et al., 2016 QSR ; Anchukaitis et al., 2017 QSR), Guillet et al. (2017 NGEO) or the most recent TRW-based paper from the tree-ring community published lately by Buntgen et al (2021) in NCOMM.
• Chapter 2.2: It is not clear to the reader how the authors did the tree-ring analysis. They provide a long discussion on advantages of MXD over TRW data, but it is very unclear how the authors did the reconstructions and what they did with the data. How were the sites/data chosen? More details need to be provided here as it remains very unclear to the referee how the proxy series were developed.
• The same holds true for the NH approach: why did they not use the spatial reconstruction data from Anchukaitis et al. (2017) or Guillet et al. (2017)?
• Another major drawback is the restriction of the comparison of model with tree-ring data (lines 349ff) just between the Alps, Altai (both known for excessive cooling in tree-ring records) and Scandinavia. Why dd the authors not rely on the full set of tree-ring reconstructions and include a comparison for Siberia, Central Asia and North America?

For the paper to become acceptable, the breadth of the proxy records needs to widened and the methods need to be described in much more detail.

Minor points:

Line 2-4 (Abstract): to which “multiple paleo proxies” are the authors referring to? This is a misleading statement as the proxy records pointing to massive and long lasting cooling are few. This needs remediation

Line 12/13: “see” should be changed to “sea”

Lines 19-21: This needs some rephrasing, stating that the cooling was 20 years in the proxy records is somewhat an overstatement. The initial cooling was in fact there, but temperature recovered rather quickly to more normal conditions and reached “fully normal” after two decades. Some clarification would be good here.

Line 30: what lines of evidence do you have to compare the 6^th century cooling to the conditions that led to the LIA? I suggest that either references are provided or that this statement is removed.

Line 152: How does peak cooling in models compare with proxies? How does the amplitude of cooling compare between the two datasets?

Line 155: what do you mean with background level for AOD? <0.1?

Line 376: the LALIA concept is based on records from two sites. The authors should go beyond these sites and analyze all data that exists in the NH. It is unclear why the study is limited to Alps, Altai and Fennoscandia

Line 386: use the data presented in the Buntgen et al. (2021) ensemble study instead