Thanks for a prompt reaction! With all respect, we find that the reviewer’s methodological concerns are misguided, but we will welcome a less generic review that goes beyond the specifics of the Introduction. Our study is indeed statistical in nature and we use appropriate and well-tested methods. We also address various uncertainties and we discuss potential underlying processes and dynamics in various parts of the manuscript. Furthermore, we not merely assume a significant influence of volcanic forcing (from tropical eruptions or extra-tropical ones) on local to regional climate but test for it empirically (see section 3.2 in our article).

There are many variants of detection and attribution, ranging from looking at global temperature trends to single extreme events. The detection part deals with establishing the statistical significance of, e.g., the trend or the increased frequency of the considered type of event. The attribution part deals with identifying the physical cause and may involve modelling to estimate for example the ‘world without anthropogenic changes’ or the fingerprint in space/time of different candidate forcings. In this context it should be remembered that models have their own problems and model deficiencies may be a problem.

However, our situation is very different. We have many instances of an episodic forcing and want to isolate the effect of that forcing. For this purpose, the methods chosen in the present article are appropriate. The superposed epoch analysis (SEA; also known as composite analysis) distills the common impact of the episodic forcing and suppresses the noise. The noise includes random variability/weather that alters the impact of volcanic forcing to a varying degree.

SEA is a well-tested method – discussed in many texts on statistical analysis of climate data – used in many studies of a similar nature. None of these applications are without issues; other variables may be important and the situation may change over time. Often successful noise cancellation might be an issue if only few events enter the SEA. In this regard, we particularly emphasize the highly significant results in our study for SEAs with more than 20 events.

We will be happy to discuss the methodological choice in the revised version of the article, but we do not find it fair to dismiss the article on this ground alone.

The work by Brönnimann and Krämer (2016) is interesting but approaches the issue of volcanic forcing from a very different perspective. As we are studying the average effect of a large number of volcanic forcing events (eruptions), and not the effect of particular individual ones, we believe the otherwise very promising approach of Brönnimann and Krämer (2016) would be difficult to implement in our case. We will, however, cite Brönnimann and Krämer (2016) in the revised version of the article in the Introduction (as well as the other suggested additional references).

We do not find the statement about the purpose misleading. The entire purpose and aim of the article are, in fact, exactly to study ‘the impact from climate, with particular focus on the influence from volcanic forcing’ and do so ‘using superposed epoch analysis (SEA) appropriate for studying episodic events in non-stationary time-series’.

Reviewer comment: As the history of agricultural production in relation to climate is not my field of research, I am not familiar with this bibliography, and I therefore leave the task of critiquing the relevance of the references cited to more experienced reviewers. With my outside eye, I’m content simply to highlight the points that seem to me in need of improvement. In general, I found the paper interesting but I have major critics to formulate. I thus suggest the editor to considers a second round of review after a revised version is submitted. There is my view many sections that are out of the scope of the title. In fact, the relationship between volcanoes and wine production is barely discussed beyond the statistical significance. Just for that I think the paper needs major revision before be published.

Answer: Thank you for your comments. We want to emphasise, in the light of the criticism, that this is indeed an article of a statistical nature. Our main aim is to investigate the statistical relationships between volcanic forcing and climate and wine production in the region in question. Actually, we do discuss details of the connection between the volcanoes and wine:  we show that the mechanism goes through the climate (sections 3.2 and first half of 3 .3).

Reviewer comment: The authors state on line 89 that long-term trends are unreliable. How can short-term trends be reliable then?  I confess I find it hard to understand how fraud and resistance to taxation would not also skew high-frequency fluctuations.

Answer: We will slightly rephrase this in the revision. From previous research about tithes in general it has been shown that long-term trends contain larger biases than short-term variability. But that is not to say that fraud and resistance to taxation are not relevant for short-term variability. The SEA and the Monte Carlo method for significance is designed for identifying signals from an episodic forcing. We have many eruptions and “stacking” them helps to reduce the noise to a level where the signal stands out. Note that the SEA is based on the 5 years before and after each event and therefore not sensitive to the long-term trend (and for wine the trend is positive while the volcanic signal is negative). We cannot rule out the possibility that the respective error term is skewed towards a higher resistance to taxation in years of bad harvests. Such non-linearities are very difficult to assess and might similarly affect other proxy records. We are willing to discuss these issues in more detail in the revision. In the words by Rao et al. (2019): The compositing and averaging process serves as a filter that enhances the high-frequency response signal of interest while minimising noise. This technique also accounts for long-term drift, or low frequency variability that may be present. 

Reviewer comment:  Rather than conducting a simple correlation between the two series filtered for the solar cycle and corrected for the long-term trend, why didn’t the authors conduct a Student’s t-test? And why are there two correlation coefficients when only two series are being compared? Is it possible to show these correlations on a figure to see the data dispersion?

Answer: We suppose the is about last half of section 3.3 and Fig.  4 . Firstly, we want to point out that the 11-year filter length does not have anything to do with the length of solar cycles. Instead, we found that variations of time-scales shorter/longer than about 10 years are best captured by this filter and it is common to use uneven numbers for filters. Regarding the student’s t-test we instead use a stronger more robust test. A student’s t-test does not work when various assumptions are not met, such as the serial correlations. Our method allows determination of significance despite this.

Reviewer comment: Even if there is a very strong similarity between the two records (only about 50 % of the signal variability is correlated), what guarantee do we have that these cleaned signals are representative of a global climatic response? Is it not necessary to validate this representativeness with recent data? Is current grape production in these regions well correlated with summer temperatures, for example?

Answer: We guarantee a relationship between volcanic forcing and wine production by showing the clear and statistically significant response in the SEA. Our statistic testing verifies that this response is extremely unlikely to happen by chance. Subsequently, we try to explain the response that has to be transmitted through atmospheric processes. That’s why we look at temperature changes in the years of volcanic forcing and at correlations between wine production and climate reconstructions. Relationships between different parameters associated with wine production and summer temperature has been described in the literature (e.g., Labbé et al. 2019; Pfister et al. 2024).

What we correlate is only the local regional climate impact on wine production in addition to the effects the hemispheric volcanic forcing has to the same. We are not performing correlation analysis between volcanic forcing and regional climate or the wine production – we only perform SEA. We refer to literature, and can be even more explicit about it in the revision, showing that indeed present-day summer temperatures and wine production is strongly correlated in this region.

Reviewer comment:  Where do the eruptions listed in Table 1 come from? These eruption dates are not found in the paper by Toohey & Sigl, 2017. If they are extracted from the eVolv2K database, please use the correct reference and the version used. It is also important to mention whether all the eruption listed are stratospheric types or not.

Answer: We can make this clearer in the revision. The eruptions come from a spreadsheet available from the authors of the Toohey & Sigl paper.

Reviewer comment: Concerning the SEA method, how do the authors ensure that the response statistics to volcanic forcing are not driven by just a few major eruptions?

Answer: We use all eruptions to avoid having someone say ‘it is all driven by a few eruptions’, but we are already showing in the article that even when just a few of the strongest (exceeding certain thresholds) are used we get a similar signal – it’s just not as ‘clean’ as when we use many eruptions. We show results for only forcings larger than a threshold in Fig 2. The results are robust also to using forcings less than a threshold. We will expand the discussion in the revised manuscript. We intend to especially address the influencer of the Laki 1783 eruption in the revision.

Reviewer comment: When K years are taken at random, are those with a volcanic eruption included? Please clarify this point in the text, as including years with eruptions may induce a bias.

Answer: Yes, they are included and this is the correct procedure. The null-hypothesis is that there is no connection between the time of the eruptions and the wine. If there is a signal in the observations then not including years with observed eruptions would underestimate the signal in the surrogates and thereby the overestimate the significance. We can discuss this more in the revision.

Reviewer comment: Please show the distributions obtained by random draw and the original one in a supplementary document or appendix.

Answer: We are willing to include this in an Appendix pending on the editor’s decision.

Reviewer comment: As mentioned above, the SEA method can be drawn by a few major events. Would it be possible to make random draws for half of the volcanic events and to conduct a SEA for each draw to see if the statistical weight of each volcanic event is indeed homogeneous and not driven by deposition flux?

Answer: See our answer above. What the reviewer suggests would be possible to do but we are unsure whether this is necessary in the light of the sensitivity experiments already preformed.

Reviewer comment: The difference in response depending on the size of the sulfate flux is not discussed? Why would there be a preferential 2–3-year delay for high sulfate deposits? What is the underlying mechanism between a rapid and significant response for all eruptions and a delayed and less significant response for strong eruptions?  This is counter-intuitive and deserves further discussion.

Answer: We can expand the discussion in the revision about possible explanation behind this. It is actually not a delay of any signal but rather that the peak signal occurs later. A possible explanation could be for example factors like sea-ice feedbacks and/or physiological effects of the grapevine (which the existing literature supports).

Reviewer comment: It’s not clear whether the Monte Carlo approach has also been used for climate variables. The author should clarify this point.

Answer: The Monte Carlo approach using random years has been used for all SEA. For correlations we use the phase-scrambling method. These are both Monte-Carlo approaches. We will make this clearer this somewhere in the revision.

Reviewer comment: Figure 3: Indicate the year with the lowest p-value for each climate variable on the graph.

Answer: We will indicate the lag with the lowest p-value. This is mainly lag 1 or 2.

Reviewer comment: Why is there such a difference in response depending on temperature reconstructions?

Answer: We assume the reviewer is referring to the correlations in Fig. 4 and last half of section 3.3. This difference can actually be expected given the different input proxy data, the different methods to combine them and the way they have been “calibrated” to instrumental temperature. We are willing to include a new sentence or paragraph in the Discussion section about this issue along with references to literature about the issue. Two of the authors are actually co-authors to a study published as late as May this year touching on this.

Reviewer comment: It is quite puzzling to see the author tackles the long-term relationships between wine production and seasonal and annual temperature, precipitation and soil moisture when they stated at the beginning that long-term wine production is unreliable. Where is the logic? I also found this section quite long an in contracdiction with the title (see also below).

Answer: We are filtering the climate data in the same way as the wine data to allow for unbiased comparison and we focus on the high-frequency relationship. It is common to take that part of the data that seems worth working with. We only use data that has been linearly detrended to remove the (very) low-frequency. The linearly detrended data is then treated with the 11-year low-pass filter (see captions Fig. 4). That means, that we are emphasizing variability at decadal scale.

Reviewer comment: Why is De Bilt correlation not shown on figure 4?

Answer: Because De Bilt covers such a short period it is not relay comparable with the other climate series used.

Reviewer comment: The authors should explain better the reasons for filtering at 11 years. I imagine it has something to do with the solar cycle. Has the solar cycle been shown to influence agricultural production?

Answer: As pointed out above, it does not have anything to do with the solar cycle. We will add a sentence in the revision stating the reason for the filter length. Furthermore, the 11- years corresponds roughly to the time-scale of the SEA.

Reviewer comment: Why does Grevenmacher respond more strongly to volcanic forcing than Remich? This point is not discussed. Given the proximity of the sites, this seems surprising.

Answer: We can discuss this more in the revision.

Reviewer comment: I found the title a bit misleading as the discussion at the end of the paper is mainly focused on wine production and paleoclimate variables than with volcanic forcing. Either the discussion should better describe the title or the title should be broaden to match the discussion

Answer: We can add more about the role of volcanic forcing in the revision. However, we do already compare the influence on volcanic forcing on tree-ring data and wine production in the Discussion section.

Reviewer comment: Line 115: add kg/km2 for each category

Answer: OK, can be corrected.

Reviewer comment: Figure 1: for sulfate flux, units and scales are missing.

Answer: OK, can be corrected.

Reviewer comment: Line 142: “compare size of the considered time-series …” I don’t understand what the authors mean here by size

Answer: OK, can be clarified.

References:

Labbé, T., Pfister, C., Brönnimann, S., Rousseau, D., Franke, J., and Bois, B.: The longest homogeneous series of grape harvest dates, Beaune 1354–2018, and its significance for the understanding of past and present climate, Climate of the Past, 15, 1485–1501, 2019.

Pfister, C., Brönnimann, S., Altwegg, A., Brázdil, R., Litzenburger, L., Lorusso, D., and Pliemon, T.: 600 years of wine must quality and April to August temperatures in western Europe 1420–2019, Climate of the Past, 20, 1387–1399, 2024.

Rao, M. P., Cook, E. R., Cook, B. I., Anchukaitis, K. J., D’Arrigo, R. D., Krusic, P. J., and LeGrande, A. N.: A double bootstrap approach to Superposed Epoch Analysis to evaluate response uncertainty, Dendrochronologia, 55, 119–124, 2019.

Reviewer comment: One possible explanation for the observed heightened sensitivity of wine production to volcanic eruptions, compared to dendroclimatological data from Central Europe, is that the Mosel region may lie at the periphery of viable wine-producing areas. From this perspective, wine production in this region could be analogous to tree-growth in Torneträsk, Sweden, where trees at the limit of their growth conditions are highly sensitive to even minor temperature drops. Similarly, one might expect grapevines in the Mosel region to be particularly susceptible to adverse environmental conditions. Is this the interpretation the authors intended in their discussion? If so, it could be more explicitly stated. This could explain why the sensitivity in the northern region of Gravenmacher is higher than in Remich. According to this explanation, the sensitivity of wine production to volcanic eruptions in more temperate regions such as France, Italy, or Spain should be then comparatively lower than in the Upper Moselle region.

Answer: Yes, the reviewer is correct that the Mosel region is situated at the northern limit of viable wine-producing areas (which we mentioned in the manuscript) in the same way as the Torneträsk region in northern-most Sweden is situated close to the Arctic tree-line. Both the grapevines in the Mosel region and the trees in Torneträsk are thus, as the reviewer points out, very sensitive to temperature drops and presumably also to less sunlight. We can make this clearer and more explicit in the Discussion section in the revision. The higher climate sensitivity of Gravenmacher compared to Remich may also, rather than the small difference in latitude, be due to different soil conditions. Again, we intend to make this clearer in the Discussion section in the revision.

Reviewer comment: Figure 1 is particularly compelling. The absence of data during a portion of the Little Ice Age and the Maunder Minimum is unfortunate, as it limits the authors’ ability to assess sensitivity to more prolonged periods of extreme cold. Nevertheless, these time series offer valuable insights for testing the sensitivity of dendroclimatological data. Both series include years of nearly zero wine production, raising an initial question: were these very low production years attributable to climatic factors, or were they the result of sociopolitical events? If the authors can confirm that the drastically reduced production during these years was indeed climate-induced, it would be worthwhile to examine the climate reconstructions for those specific years. In this regard, while the study already includes a Superposed Epoch Analysis (SEA) of the impact of cold years on wine production and explores correlations between wine production and temperatures, presenting a scatter plot of reconstructed temperature versus wine production might also be informative. This could reveal whether a linear relationship exists across the high and middle-temperature ranges and whether the slope of this relationship changes at the cold-temperature extremes. Alternatively, as wine production cannot go below zero, it may be unable to capture the extreme cold years. One way or another, such a scatter plot could shed more light on the relationship between wine production and dendroclimatological data than just establishing an overall correlation.

Answer: We agree that it is unfortunate that the Late Maunder Minimum is missing in the Mosel region wine data. However, the last three decades of the 16^th century were as cold as the 1690s. We can point this out in the Discussion section in the revision. Regrading that the wine tithe data is zero in certain years, it is not certain that the grapevine harvest was zero or almost zero. From grain tithes, for example in the early modern Swedish Realm, we know that sometimes no tithes were collected during years with extremely low harvest (like only one-third of the average harvest). Thus, zero wine tithe indicates very low grapevine harvests but not necessarily none at all. We can include something about this uncertainty in the revised manuscript.

The reviewer shows an interest in seeing a scatter-plot of some wine series against some climate reconstruction series. This is very possible – but there are many such choices given the Figure 4. We would like to remind the readers that Figure 4 is an efficient summary of the many possible scatter-plots that one could think of – each scatter-plot has been concentrated to a single coloured square in Figure 4. We thank that is enough to show.

We have tested (it is in the manuscript line 165) that linear and non-linear correlation methods (that is Pearson vs. rank correlation methods such as Kendall and Spearman) all give similar results; this implies that the nonlinearity that is indeed present in the wine data (cannot go below 0) and various non-linearities present in climate reconstructions have not prevented the Pearson correlation that we use from rendering an accurate picture of how the series are correlated. It is all encapsulated in Figure 4.

Reviewer comment: I think the Urhausen et al. (2011) reference is incorrect. The correct one would be ‘Climatic conditions and their impact on viticulture in the Upper Moselle region’ published in Climate Change doi: 10.1007/s10584-011-0059-z

Answer: Thank you for pointing this out. The reviewer is correct and we will fix this error in the revision.

Reviewer comment: The manuscript is concerned with comparing two historical records of wine production in present-day Luxemburg with volcanic events in the period 1444 to 1786 AD and with a number of climate European climate reconstructions. A stacking of the wine production record centred at the years of large volcanic eruption forcing - in terms of sulfate deposition in Greenland - shows a clear decrease in production in the years immediately following the eruptions.

Answer: Thank you for an excellent summary of our article.

Reviewer comment: In general, the figures are lacking basic information about what is shown on the axes, what units are used, and what curves of different colours are showing. In particular, it needs to be stated clearly what is shown in Figures 1 and 2. Please make this basic information available.

Answer: Thank you for pointing this out. We will fix this in the revision.

Reviewer comment: Otherwise, I am probably more positive towards the approach taken by the authors than some of my co-reviewers. In my view, it is okay to try to find a statistically significant volcanic/climatic impact on wine production without having to include and discuss the many other society- and environmental-related indirect effects that must also impact wine production from the volcanic cooling events. I think it will be close to impossible to try to disentangle those effects in a statistically sound way, so then it is probably better to state clearly that they are not considered and then stick to what can be handled statistically, namely the NH volcanic forcing versus the wine production.

Likewise, I also buy the suggestion that short term effects of a climatic cooling may show up in a stack of events although the long-term wine production record may be unreliable. The way the wine production has been quantified has probably changed over time, but not every year. Therefore, many short-term events will be preserved and show up in a stack/SEA although the long-term changes will introduce some noise.

Answer: Again, thank you for your positive evaluation of our work. You have effectively underlined why the SEA analysis is strong.

Reviewer comment: In Figure 2, besides specifying what is shown on the axes and in what units, and what ‘forcing flux’ refers to and what units it has, it would be quite helpful to state how many events are included in each stack. For example, how many events are stacked with a volcanic NH forcing flux larger than 15? If NH forcing flux refers to the Greenland sulfate deposition (in kg per km2) associated with individual volcanic events as specified in table 1, then I can count 11 events in the period 1453-1786 AD. However, if we exclude the wine production data gab 1684-1741 AD, then we are down to 8 events related to volcanic eruptions of this magnitude. Is that correct? Please state number of events included for each category.

Answer: We already have information about the number of eruptions in line 115. We will also include the number of eruptions in the figure captions in the revised version of the manuscript.

Reviewer comment: The most significant of the volcanic events both in terms of sulfate deposition and in terms of impact on wine production appears to be related to the 1783 AD Laki eruption (would be worth to mention this eruption in the text). If I am correct in assuming that the x-axis in Figure 2 shows years before and after the maximum sulfate deposition in Greenland, how do the authors include 5-year wine production impact of the Laki eruption, when the production record ends in 1786 three years after the eruption? Like one of the other reviewers, I am concerned that the Laki eruption is so significant that it is dominating the stack of rather few large events. What do the stacks look like if the Laki event is left out?

Answer: We have already tested the effects of excluding the 1783 AD Laki eruption. We can without problems say that this eruption is not dominating the results. In the revised version of the manuscript, we will mention this. We note that also Reviewer#2 wants more information about the influence of large eruptions. Finally, we are aware of the edge-effects (i.e., after the Laki eruption), and we do not count these ‘impossible’ lags, and we will describe it in more details in the revised version of the manuscript.