The lead author is one of the world’s key scholars  in Historical Climatology and Climate History cooperating with natural scientists and historians, mainly from Europe. The Huaynaputina is a stratovolcano in southern Peru. Its eruption in February 1600 CE triggered a persistent summer and winter cooling in the North Atlantic region during the early 17. It-was the largest eruption in the Andes in historical times. The paper explores the eruption-induced cooling mechanism in some detail. The study compares simulations and a North Atlantic subpolar gyre shift in annual proxies in archives of nature to more detailed with documentary evidence.

The Huyaputina injected fewer sulfates into the stratosphere than other LIA eruptions with a compatible cooling effect. In particular the eruption generated much more and longer lasting winter cooling than summer cooling in Central and Northern Europe, which is unusual. It is hypothesized that reduced heat transport by the North Atlantic Subpolar Gyre SPG in terms of a cooling of the North Atlantic may be one of the reasons, though many aspects remain unclear. Historical written records as well as contemporary historical observations of relevant climate and environmental conditions demonstrate patterns of cooling and sea ice expansion consistent with, but not necessarily indicative of an eruption trigger for the proposed SPG slowdown mechanism.

                  The arguments of scientists and modellers should still be improved in view of the limited understanding of people from the historical sciences for processes in the ocean. Some relevant studies might still be included: The herring catch on the west coast of Scotland declined remarkably between 1585 and 1597 which Parry (1978) interpreted this as an escape of this cold-sensitive species from the cold water masses advancing southwards. The detailed temperature reconstruction by Dobrovolny et al. (2010) for Western and Central Europe since 1500 CE was overlooked. It contains seasonal temperature that are explained in more detail by the recent synthesis by Pfister and Wanner (2021) that is published on 6^th September. These data show that the very severe winter of 1600 preceded the Huynaputina eruption in contrast to the cold winter in 1601. Likewise, just spring 1600 was very cold. On the other hand, the summer 1601 was very cold. 

Overview

The manuscript fits the scope of the journal well with integration of climate modeling, paleoclimatology, and historical climatology to address the effect of the Huaynaputina (Peru) eruption in 1600 CE on northern hemisphere cooling via the North Atlantic subpolar gyre and ocean-atmospheric feedbacks. The SPG is hypothesized to be the cooling mechanism that led to low temperature anomalies in Europe and Russia in the early 17^th century. While the results are not conclusive, the authors have established a research course to investigate the relationship between volcanic eruptions and important climate shifts that have affected humans. The methods and assumptions of the work are clearly outlined and the authors’ interpretation of the results is in accord with their analysis. The supplemental materials make the research reproducible with extensive presentation of the historical observations used in the methods. I found the paper well-structured and written with few technical errors.

General Comments

The cluster of volcanic activity in the late 1500s would seem to make it difficult to determine if the Huaynaputina eruption seeded the SPG slowdown or if the eruption was the final push over a threshold given the background state of the atmosphere after multiple VEI 4 eruptions. Perhaps this is why an SPG shift can occur in some simulations without a volcanic forcing in 1600. However, the combined use of model simulations, paleoclimate reconstruction, and historical climatology helped to better target an initial seed to the SPG slowdown but unfortunately the data are inconclusive at this point. That said, this is a fine contribution demonstrating how these data sources can be integrated to elucidate the mechanism driving climatic change circa 1600.

Further analysis and discussion of the North Atlantic Oscillation (NAO) and Artic Oscillation (AO) would be helpful to disentangle how the background climate state and internal variability might contribute to an SPG shift. Previous research suggests an interaction between NAO and volcanic eruptions (e.g., Ortega et al. 2015) with a positive NAO emerging after strong volcanic eruptions. NAO+ would lead to stronger westerlies in northern Europe resulting in warmer and wetter winters, meaning the SPG would likely not slowdown. Of course, there are many NAO reconstructions out there to choose from including several recent reconstructions from Ortega et al. 2015, Cook et al. 2019, and Hernandez et al. 2020. The research could benefit from a more comprehensive treatment of NAO/AO.

Specific Comments

L70 - North Atlantic Oscillation (NAO) was not significantly affected by the eruption but it could be foundation to understanding the background state of climate leading into an SPG shift.

L143-48 – Previous research is showing that no volcanic forcing is need to produce SPG shifts depending on the background climate state. Okay, so what was NAO, or the Artic Oscillation (AO), doing when the SPG shift occurred? How would you distinguish intrinsic variability from a volcanic forcing of an SPG shift?

L208 – Is the 1550-1590 baseline period suitable to calculate anomalies when it includes multiple eruptions?

L229- the reference period here changes from the reference period for the reconstructed anomalies. Please justify the change in reference period. Or is this a typo?

L275 – there also appears to be a lack of agreement between NTREND and the simulations. Why might this be?

L282- in Figure 6, it appears that the NVOLC reconstruction has much more annual variability and different spectral properties than the simulations. What is causing this discrepancy?

L297-301 – NAO does play a major role in setting winter conditions in Europe. So, what was the state of the NAO during the period of analysis?

L308 – Could the shifts in ice break of dates be connected to NAO and AO? Some of the reconstructions of NAO (Cook 2019, Ortega 2015, Hernandez 2020, etc.) show shifts that might correspond to the ice break up regime shifts.

L376-385- If additional simulations do not result in determining what the initial seed for SPG slowdown is, what model improvements would be needed to better model what the climate proxies and historical records appear to show?

Technical Edits

L31 – add space between number and m - “4,850m”

L140 – missing hypen “Moreno Chamarro”

Summary: The focus of the study is to disentangle the role of the external volcanic forcing and of internal climate variability in the cooling surrounding the strong volcanic eruption in of the Huaynaputina eruption 1601. The authors analyse different proxy data sets, including air,temperature, sea-ice extent in the Arctic-North-Atlantic, historical ice break-up dates at several Baltic ports and ensemble of simulations with the Max-Plank-Institute Earth System model.

The conclusions is that the attribution of all aspects of climate change around this date are very difficult to disentangle. Internal climate variability may be large, and although the attribution of the whole temperature evolution around those decades is compatible with the effect of volcanic eruptions, internal processes may play also an important role, even before the eruption.

Recommendation: I enjoyed reading this paper very much, and I like to congratulate the authors at this point. Although, as the authors acknowledge, the study is not conclusive, the authors have tried to use all data sets available to them and have conducted a very thorough, objective and candid analysis. On the other hand, it is very well written, provides an exhaustive background on the physical mechanisms and on the historical evidence, also a proof of a well functioning collaboration between climatologist and historians. Perhaps the significance of this and similar studies goes beyond what the authors let on: this type of events can occur at any time, regardless of whether they are produce only by volcanic forcing or by a combination of volcanic eruption and internal variability. This, it is important to understand these past events.

This manuscript is one of the best that I had a chance to evaluate. My recommendation is to publish it - I have just a few minor comments that the authors may want to consider.

1) My most general comment is directed to the quantification of the magnitude of internal climate variations in this context. The study does compare simulations with and without external volcanic forcing for this period, but here a more general question remains open: what is the largest magnitude of multi-year cooling in a long control simulation in this region ? Do large periods of multiyear cooling, comparable to the years following the 1601 eruptions, also appear in simulations without variations in external forcing ?

2) line 34 'The VEI 6 1600 Huaynaputina eruption' I think VEI has not been defined at this point in the manuscript.

3) line 145 ' Of the 8 total members of the SPG-shift ensemble, 6 had volcanic forcing' . How many simulations with volcanic forcing do not show a SPG-shift ?

Review for the manuscript

The 1600 Huaynaputina Eruption as Possible Trigger for Persistent Cooling in the North Atlantic Region

by Sam White et al.

Submitted for publication in Climate of the Past

General

The paper investigates the climatic and historical context of a tropical volcanic eruption at the beginning of the 17^th century using a combination of proxy, historical and modeling evidence. The authors initially hypothesize a prominent influence of the state of the North Atlantic Subpolar Gyre (SPG), initiating a sustained cooling over Europe. Authors conclude that the SPG could have an important role. However, their results remain inconclusive taking into account their combined evidence using different reconstructed metrics (e.g. Baltic Sea Ice, North Sea Winds).

The basic structure of the manuscript is not in an optimal shape. The reader is confronted with different parts of a classical paper, resulting in a mix of background information, methods, data sets and conclusions presented in the different sub-chapters. This might be sourced in the fact that it is an interdisciplinary research paper. However, some crucial statements and physical mechanisms contradict in different sections of the paper.

Therefore I suggest that i) the manuscript should be completely re-written ii) the proxy- and historically derived hypotheses should be clearly formulated in the beginning and (statistically) tested by a more comprehensive suite of available CMIP6 simulations [ e.g. those CMIP6 model simulations with appropriate ocean models and according horizontal and vertical resolution – in the present manuscript only one Earth System Model is used ] and iii) the different parts and disciplines should be conceptually better coordinated.

In the following I provide some suggestions how to re-structure the manuscript.

Specific

Introduction/Basic Concept

The introduction should contain the basic background information to understand the concept and eventually the conclusions of the paper. Therefore it is of ultimate importance to be conceptually sound and also introduce the main concepts elaborated in the manuscript. For instance, the SPG is not introduced at all. Also, the model used is never explained related to the fact how the model is capable to realistically simulate the SPG. For this a dedicated methods and data section is necessary where the different components of the paper are comprehensively explained. In the current version the first part is a mere repetition of results already published elsewhere (Morene-Camaro et al., 2017) even including the same set of figures (cf. Figure 1) that is already published.

This also relates to the 2^nd important concern: The authors only use one model from the CMIP6 suite. For their period under consideration a larger number of simulations, also for ocean models, is available in the Earth System Grid Federation (ESGF) platform. Since results for MPI already have been published and this very mechanism might be evident only in the MPI model, the question is whether all of the CMIP6 models, or at least those with similar horizontal resolution show a similar response. This is even more important when comparing model derived results to real-world derived hypotheses to test the robustness of the model derived results and to finally discriminate between I) internal vs. forced changes and ii) model-vs-model intrinsic variability related to the individual structure of the Earth System Model.

Methods/Hypotheses

The basic setup of the authors using a combination of different disciplines to address a certain questions is a good asset. However, the potential and per-requisites should be formulated conceptually more sound. For instance, in the present setup the hypotheses should be derived based on proxy and/or historical evidence. In a second step a potential physical mechanism should be motivated explaining the initially formulated hypothesis (e.g. changes in SPG and its impact on European temperatures). In a third step this should then be tested in the model world, most preferentially using a suite of comprehensive Earth System Models simulating this period and using state-of-the art statistical tests (for instance Boot Strap methods using control simulations to derive reference climatic states). In the present version this concept is reversed and the initial hypotheses are derived from the climate model. In general, this is also possible but it is of ultimate importance to state this clearly and also present a way of how this (set of) hypotheses is falsified.

The lack of a sound statistical testing scheme and a careful inspection of the different conclusions derived in specific parts of the manuscript results e.g. in the following contradicting statement:

Moreno Chamarro et al. (2017b) found consistencies at multidecadal scales between simulations with a weakened SPG and reconstructed changes in several geophysical variables of the North Atlantic after ca. 1600 CE. The study did not conclude that the late 16 th -century volcanic cluster was necessary for the SPG shift, which was instead mainly attributed to intrinsic variability of the simulated climate system. Sensitivity simulations of the period 1593-1650 with no volcanic forcing yielded SPG shifts similar to those in the volcanically forced simulations. [ cf l. 141 ff ]

vs.

1. This study has examined high-resolution proxies and historical observations to investigate whether the 1600 Huaynaputina eruption triggered persistent cooling in the North Atlantic region by initiating a regime-shift of the North Atlantic subpolar gyre toward a persistent weak phase in the early 17 th century,as shown by paleoclimate model simulations. [ cf. l. 371 ff. ]

The conclusion derived from the model analysis showed that the shift might be simply due to intrinsic or internal climate variability. However, in the conclusions authors sate the volcanic eruption triggered the regime shift initiated by the volcanic eruption. Moreover, the paleoclimatic model simulations only relate to the MPI-ESM model the authors used for their investigations.

An important information that was also never mentioned in the manuscript is that a number of volcanic reconstructions is available that have already been used for simulating the impact of volcanic eruptions on climate (e.g. Crowley and Unterman (2013); Gao et al. (2008), Toohey et al. (2016)). Especially the strength of larger tropical eruptions can vary up to a factor of two within the change in aerosol optical depth (AOD), the most important radiative physical moment in the stratosphere in the context of explosive volcanic eruptions. This should and could be taken into account by including a 2^nd ESM simulation (e.g. the CCSM4 CMIP6 model used the Gao et al. 2008 data set in contrast to the Toohey et al. 2016 volcanic data set used in the present simulation). Integrating a 2^nd set of simulations would help to better assess the impact/change of the SPG on the climate in Europe in the different Earth System Models.

Statistical Tests

The general setup of the manuscript would greatly benefit by implementing a statistical test scheme with a clear formulation of a Null hypotheses that is falsified by an appropriate statistical test. Especially in the virtual world of the Earth System model this could be (quite easily) achieved. An option is for instance to design a test in the context of a bootstrap method: The null hypothesis is that the SPG has no influence on European temperatures. The nominal level can be set even to a two-sided test with 5 % . The test can now formally be carried out using sub-samplings of the different trajectories of the SPG in terms of block bootstrap by using control simulations. The test should be applied to the canonical pattern between the state of the SPG and European temperatures. If the sub-sampling leads not to statistically significant negative deviations of European temperatures in the presence of a shift in SPG, then the null hypothesis cannot be rejected. Eventually, these tests should be carried out for simulations with and without volcanic forcing.

Physical mechanisms

The authors mention a couple of (important) physical mechanisms that might support their hypotheses.

A first example relates to the NAO: at several places in the manuscript (cf. l. 69; l. 297) the authors mention the North Atlantic Oscillation as physical mechanisms explaining part of the temperature variability and being important also in the context of volcanic eruptions citing different authors. I wonder why the authors do not briefly explain the main mechanism suggested for the NAO in the first winter after volcanic eruptions (so called mid-Winter warming in Europe because of a positive state of the NAO, Kirchner, 1999; Zambri et al., 2017;). What mechanism is giving rise to such a response ? How robust is such kind of response and what effect does it exert on the winter temperatures in Europe ?

A second mechanism mentioned in this context relates to changes in blocking frequencies that are believed to be larger in the aftermath of volcanic eruptions and/or are an important mechanism explaining cold and very cold winters over (western) Europe (l. 66 ff). The authors argue that the blocking is independent to changes in the NAO. This is a bit surprising, because the NAO is the leading mode in Europe’s winter variability and changes in the blocking should also effect the state of the NAO.

A third mechanisms relates to the role of sea ice (l 220 ff.). First, also sea ice concentrations can show a spatially heterogeneous pattern, especially when the entire North Atlantic region including Greenland is taken into account. In this context changes in the NAO can lead to dipole patterns with anomalous high sea ice around Greenland, and low sea ice over western Europe and vice versa. If this is not the case at least it should be motivated which canonical Circulation-sea ice patterns could lead to a spatially homogeneous response and/or whether direct radiative changes caused by volcanic eruptions could compensate or offset dynamically induced dipole patterns.

A last mechanism I would like to mention here relates to the direct vs. indirect effects of volcanic eruptions on climate:

The summer cooling is, by contrast, absent in the no-shift ensemble, which comprises mainly simulations without volcanic forcing (8 out of 12), the two ensembles show minor differences in oceanic variables such as the barotropic stream function and winter sea-surface temperature in the North Atlantic, which are weakly impacted by the volcanic forcing in the short term. Larger differences in these variables between the ensembles emerge over the following decades, particularly after the 1610s, in association with the SPG slowdown, as shown in Figure 3.[ l 162 ff. ]

Here the authors even state that the summer cooling is absent in those simulations without volcanic forcing. Therefore the question remains as to whether a change in SPG is really necessary to initiate the sustained cooling. Also, a more objective formulation how the shift is quantified would be necessary to test if the deviation from the mean state is large enough to speak of a regime shift.

Synthesis with proxy and historical reconstructions

In its present form the different chapters are not integrated into a consistent manner. Although this step is usually the most demanding, authors should at least indicate that their approach in comparing reconstructions with the world of the climate model is not a state-of-the art approach. For instance, forward models would be at hand to directly simulate respective proxies. In its present form the authors just use the information directly from the climate model without any further (advanced) processing. This represents an additional source of uncertainty in their conceptual framework.

The study investigates the likelihood of a slowdown in the SPG around the onset of the LIA being linked to the 1600 Huaynaputina volcanic eruption. In order to resolve this issue, the authors attempt to integrate evidence from model-based simulations of past climate conditions with proxy-based paleoclimatic reconstructions and historical records. Despite the inconclusive results, the study highlights both the advantages of adopting an interdisciplinary approach as well as the challenges and limitations of bringing together and interpreting various sources of information.

General comments:

In my opinion, the multi-disciplinary nature of this study represents a considerable strength of this work. In general, the manuscript is well written and the findings are presented in a clear and logical manner. The evidence is interpreted objectively and the authors clearly acknowledge the limits of the analysis. From the presented results, conclusions are drawn to the extent that the simulation, reconstruction and limited observational data from the period allow. However, the unconventional structure is rather confusing since the introduction, methods description and some of the results are all blended together, and this also makes it somewhat difficult to distinguish for example what was done in previous studies and what represents original analysis. The authors should therefore seriously consider whether restructuring the manuscript in a more conventional format would be beneficial.

Currently, a large part of the discussion is dedicated to discussing the historical / societal impacts of cold conditions at the end of the 16^th and during the early 17^th century. Greater focus on integrating and discussing the results of the modeling, proxy and historical datasets in more detail would be helpful.

Another important point is recognizing and acknowledging discrepancies between model-based simulations with proxy-based reconstructions, which has consequences for understanding uncertainty and the overall reliability of these data sources. This issue is highlighted for example by Figure 6, which shows poor spatial agreement between modeled and reconstructed temperatures. Model simulations are often associated with high uncertainty particularly in relation to post-volcanic cooling and, for example, over-estimation of the magnitude of post-volcanic cooling by some models has been known to occur (e.g. Chylek et al., 2020; Hartl-Meier et al., 2017). Better understanding of some of the shortcomings of these datasets and limitations in their utility within the context of this study could be achieved by exploring a broader set of model simulations or model types to help disentangle the possible influence of model bias and a more detailed examination of the proxy-based temperature reconstructions would also be helpful in this regard.

It is also necessary recognize the potential importance of background climate conditions in modulating the (cooling) response of the North Atlantic to large volcanic eruptions based on the state of the climate system. In relation to this point, the role of internal variability and specifically the potential role of the North Atlantic Oscillation (NAO) in the initiation of SPG weakening and cooler conditions in the north Atlantic sector remains a subject of debate (e.g. Trouet et al., 2009; Lehner et al., 2012). For this reason, some type of examination and discussion of the modes of atmospheric variability in the north Atlantic within this context would be helpful.

One obvious limitation is that most of the presented evidence for the SPG shift is either indirect / circumstantial or entirely model-based. Although the study provides a compelling narrative characterizing anomalously cold conditions in the early 17^th century, a certain leap of faith is currently required to link an SPG mode shift to these changes. In any case, more information would be required to clarify the relationship between the eruption, short-term and long-term cooling and how these events and changes relate to the state of the SPG. Ultimately, there are limits to the answers that modeling can provide and additional more direct proxy data would likely be required to better understand the dynamics of oceanic circulation and atmospheric dynamics during this period to more precisely pin down the timing, duration and extent of the purported SPG slowdown. Perhaps then it would be possible to confirm or refute the attribution of the observed longer-term cooling in the early 17^thC, and by extension the initiation of an SPG slowdown, to a volcanic trigger.

Specific comments:

L63-72: While it may perhaps be possible for such changes to occur without invoking substantial changes to atmospheric dynamics in the North Atlantic, the background state of the atmosphere, internal variability and the role of the NAO cannot be discounted a priori, particularly as these factors may act to modulate the response of the climate system to a large volcanic event.

L89: The phrase ‘possibilities for adaptation’ seems a bit vague and it is not clear what this refers to. Please specify / clarify this point.

Figure 2: For easier interpretation of the figure, it may be clearer to also state in the panel sub-headings that the plots are showing temp. / Sv. anomalies.

L233: It is not clear whether this implies that only a 30-yr segment length was used or a range of segment lengths (30-yr+) was examined. If it is the former case, please remove ‘minimum’ to avoid confusion. Otherwise, please specify the range of segment lengths utilized.

Figure 5: Please specify in the figure caption what the purple dots in top-left plot represent.

L203-210: What was the size of the reconstructed grid cells? Which instrumental dataset was used for calibration? How were the chronologies merged and how was the reconstruction performed (e.g. PCA, nesting), etc.? In general, more detailed information about the development of the spatial reconstruction is needed here (or at least in supplementary materials).

Figure 6: How does the NVOLC reconstruction compare with N-TREND (and model output) over the investigated period? Currently, only NVOLC is compared to model output, whereas N-TREND is only used for illustration and is not compared to NVOLC or the modeled temperatures. The highly anomalous cooling in SE Europe in the NVOLC reconstruction (Fig. 6a) is rather suspicious and I wonder how robust this feature is. According to Supplementary Figure S3 in Guillet et al. (2017), most of northern Europe and parts of western / southwest Europe calibrate well, whereas calibration / verification statistics are very weak for NW, central and especially eastern and SW Europe. Consider that poor spatial representation of reconstructed temperatures may cause disagreement with modeled temperatures in some areas. Likewise, specific limitations of the model may also lead to disagreement. Such considerations should be acknowledged and discussed.

L286: Why is the NVOLC v2 reconstruction shifted by +0.5 K?

L350: I suggest that a more appropriate term to use in this context would be ‘support’ rather than ‘appear to confirm’.

L368-370: So, considering the timing, might this in fact suggest that the Huaynaputina eruption is rather unlikely to be the cause of the SPG slowdown?

L371-385: Another possibility could be that a pronounced shorter-term cooling impact of the Huaynaputina eruption was ‘superimposed’ on the longer-term cooling trend, which may have been initiated prior to 1600 (either in response to the cluster of late-16^th century volcanic eruptions or otherwise). Evidence for volcanic-induced short-term cooling is on firmer ground as the results are consistent with this type of response to the eruption (Fig. 6c) and this is also consistent with the duration and magnitude of inferred NH cooling responses to large (tropical) eruptions more generally based on proxy reconstructions (e.g. Esper et al. 2015) and modeling of surface air temperature. In contrast, the mechanism for initiating longer-term cooling / SPG slowdown and attribution of such changes to a particular volcanic event is highly uncertain and rather problematic.

L376-377: One could argue that it is uncertain whether this issue could be definitively resolved through modelling alone.

L395-397: I would recommend reformulating this sentence considering that, based on extensive paleoclimatic evidence, the occurrence of cool (wet) summers during this period is actually not in question. Therefore, rather than ‘confirming’ this, it would be more appropriate to state that this study provides further support and a broader context for such conditions at that time.

Minor / technical comments:

L95: consider ‘… activation (or lack thereof) …’

L186/380/384: ‘an SPG’ rather than ‘a SPG’

Fig.4 legend / L233 / (L303): ‘ice break-up timing data’ instead of ‘date data’

L223: consider ‘obtained’ instead of ‘gained’

L238: consider ‘recorded’ rather than ‘left’

L239: ‘latter’ rather than ‘later’?

L243: Suggested wording adjustment: ‘These observations were recorded in areas with flat terrain …’

L266: ‘did or did not turn back’ or alternatively ‘ships could pass or were forced to turn back’

L267: ‘turn back during a voyage’ or perhaps ‘terminate a voyage’?

L268: ‘cold conditions’ or ‘cold temperatures’ / ‘dangerous sailing conditions’ or ‘danger posed by sailing conditions’

L280: Change ‘NTREND’ to ‘N-TREND’. Also, something is missing here - consider: ‘… in each year over the 1601-1609 period …’?

L287: should the range be 1593-1650 instead of 1593-1640?

L290: Consider: ‘The analysis in Figure 7 indicating the ice …’

L293: ‘can occur’ rather than ‘can happen’

L303: please change ‘wrt’ to ’w.r.t.’

L308: ‘detected in’ rather than ‘detected at’?

L362: ‘as a Possible Trigger’?

L372: remove ‘has’?

L375: ‘any role of’?