North Atlantic freshwater events influence European weather in subsequent summers

Oltmanns, Marilena; Holliday, N. Penny; Screen, James; Evans, D. Gwyn; Josey, Simon A.; Bacon, Sheldon; Moat, Ben I.

doi:https://doi.org/10.5194/wcd-2021-79

Preprints

https://doi.org/10.5194/wcd-2021-79

Preprints

06 Dec 2021

| 06 Dec 2021

Status: this preprint was under review for the journal WCD but the revision was not accepted.

North Atlantic freshwater events influence European weather in subsequent summers

Marilena Oltmanns, N. Penny Holliday, James Screen, D. Gwyn Evans, Simon A. Josey, Sheldon Bacon, and Ben I. Moat

Abstract. Amplified Arctic ice loss in recent decades has been linked to increased occurrence of extreme mid-latitude weather. The underlying dynamical mechanisms remain elusive, however. Here, we demonstrate a novel mechanism linking freshwater releases into the North Atlantic with summer weather in Europe. Combining remote sensing, atmospheric reanalyses and model simulations, we show that freshwater events in summer trigger progressively sharper sea surface temperature gradients in subsequent winters, destabilising the overlying atmosphere and inducing a northward shift in the North Atlantic Current. In turn, the jet stream over the North Atlantic is deflected northward in the following summers, leading to warmer and drier weather over Europe. Our results suggest that growing Arctic freshwater fluxes will increase the risk of heat waves and droughts over the coming decades, and could yield enhanced predictability of European summer weather, months to years in advance.

Received: 22 Nov 2021 – Discussion started: 06 Dec 2021

Download & links

Marilena Oltmanns, N. Penny Holliday, James Screen, D. Gwyn Evans, Simon A. Josey, Sheldon Bacon, and Ben I. Moat

Status: closed

RC1:
'Comment on wcd-2021-79', Anonymous Referee #1, 06 Jan 2022

North Atlantic freshwater events influence European weather in subsequent summers, by Oltmanns et al

This manuscript examins the impact of summertime NAO events on European weather in subsequent summers. The paper is framed as being the impacts of North Atlantic freshwater events on European weather but the index that is used to depict these "freshwater events" is actually the summer NAO. The paper demonstrates some interesting connections between the summer NAO and the following summer weather. At this point, I'm unsure as to whether this manuscript is acceptable for publication. I have a number of comments on the analysis as outlined below. Overall, I'm giving a recommendation of major revisions to allow the authors to respond to these. My major concerns are that the direct link between the summer NAO and the freshwater events is unclear to me. This may be because I'm not an oceanographer and I haven't read the authors previous papers, so I hope that one of the other reviewers will be able to assess this aspect. I have some other concerns about the statistical methods used and the choices made for the scatter plots as outlined in my comments below.

General comments:

(1) The link to freshwater anomalies and the role of low frequency North Atlantic ocean variability. I am not an oceanographer, so I hope that one of the other reviewers will have the expertise to comment on this. The link between the summer NAO index and the freshwater anomalies was a bit lost on me. My understanding of lines 84-92 is that the authors are assuming that the temperature anomalies associated with the summer NAO are due to freshwater anomalies because they find that the cooling is not strongly related to surface fluxes, wind driven Ekman transports, Ekman pumping and re-emergence of SST anomalies from previous years, so by a process of elimination they conclude that it's freshwater anomalies. But I don't see how the role for other ocean circulation anomalies such as the AMOC or advective heat convergence due to circulation anomalies produced by things other than the wind driven Ekman transports has been eliminated. The atlantic ocean circulation exhibits variability on long timescales which can be a driver of the NAO and vice-versa (e.g., Zhang et al 2019, Review of Geophysics, 10.1029/2019RG000644 and references therin). It's not clear to me (a) whether it can really be concluded that the SST anomalies are related to freshwater inputs and (b) whether such low frequency variability in the ocean circulation has been appropriately taken into account. The NAO index is being described as a "freshwater index" (l99) but I'm not sure how appropriate this is and I'm not sure that much would be lost by instead referring to it as the NAO index and focussing on the impact of the summertime NAO on the climate in subsequent years.

(2) Detrending: It's stated at line 53-55 that regionally averaged trends were subtracted from the air temperatures to remove the greenhouse gas effects. It doesn't really seem appropriate to me to remove the linear trend from one field but not others. The NAO index that is used clearly has a linear trend in it (Figure 1a). I'd suggest detrending everything or detrending nothing. I'm not arguing that the NAO trend seen in Figure 1 is greenhouse gas forced or that this trend should necessarily be removed, but it just doesn't seem appropriate to me to remove the trend in one field and not in the others. Is the detrending also done on the SSTs? It doesn't make much sense to me to remove the trend from the surface air temperature but not the SSTs.

(3) For the scatter plots, the regions where the correlation is significant at the 95% level is used for the spatial averaging. This seems like cherry picking to me. Of course the correlations look good because you've chosen them to be that way. It would make more sense to choose a physically motivated region or it would seem to be less cherry picking if a regular spatial region such as a rectangle were chosen. The result is that in Figure 2b there is a correlation of 0.98, which seems quite unbelievable to me, but maybe it isn't if you are just averaging over regions where the correlation is high.

(4) Has autocorrelation been accounted for when calculating the significance levels? If not, I think it should be. Clearly each year is not independent and there is some low frequency variability and autocorrelation, as apparent in the NAO index (Figure 2a).

(5) It is argued that this work reveals new potential to enhance the predictability of European summer weather, but I think for the impacts on European summer weather the results have only been presented in the form of regression coefficients. To make this more relevant for predictability, it might be worth showing the variance explained.

Comments by line number:

Figure 2: It seems like it would be interesting to have the regression maps for F_C as well as F_M. You use the regions based on the regression onto F_M for both F_M and F_C, so it would be good to see whether the regression map for F_C has a similar spatial pattern to that for F_M or not.

Figure 2 caption: F_M and F_C are only defined in this figure caption. Given their central importance, I think they should also be defined in the text. Furthermore, it would be worthwhile making clear the motivations for this naming convention. It's not very intuitive where the choice of "F_M" and F_C" comes from and I think it would help readers to follow if you make that clear. In the end, I realized that this corresponds to "melt-driven' and "circulation-driven" events and I'm overall just very confused about how this distinction can be made just on the basis of the NAO index, which relates to my general comments above. I think this needs to be made clearer throughout the manuscript.

Figure 3 caption: Maybe explain a bit more what the "absolute dynamic topography" is. Is it just sea surface height?

l138: "expansion of the cold anomaly" - perhaps be clear about what this "expansion' is relative to? Is it relative to the previous summer?

Section 4.5: It might be worth making it clear at the beginning here that this is now back to looking at the observations, since in the previous section the focus was on model simulations.

Figure 11 caption: The referencing to the panel labels is messed up in the caption.

Typo's/wording suggestions:

l62: "this index" --> "the NAO index"

l68: suggest "smaller values" --> "more negative values" because the magnitude of the NAO index isn't smaller.

l84: "Fig. 2d" --> "Fig. 2c" (I think d is showing salinity, not temperature)

Figure 4 caption: "The thick contours show the 95% confidence levels" --> "The thick contours encompass regions that are significant at the 95% confidence levels"

l158: "SST-forced" is a bit unclear. Suggest "Simulations performed with prescribed observation-based SSTs".

Citation: https://doi.org/10.5194/wcd-2021-79-RC1
- AC1: 'Reply on RC1', Marilena Oltmanns, 16 Mar 2022
  
  The comment was uploaded in the form of a supplement: https://wcd.copernicus.org/preprints/wcd-2021-79/wcd-2021-79-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/wcd-2021-79-AC1
RC2:
'Comment on wcd-2021-79', Anonymous Referee #2, 26 Jan 2022
General comments

The relationship between the Atlantic ocean and the summer climate over Europe is investigated. It is argued that events releasing freshwater into the North Atlantic subpolar gyre are followed by a persistent cooling (warming) over the subpolar (western subtropical) gyre. Such modified SST is linked to warm and dry conditions over western Europe in the next two subsequent summers.

While the overall mechanism seems realistic, the overall presentation of the results is very confusing. I did not understand the link between the SST anomalies analyzed and the freshwater release in the manuscript. Similarly, I was not able to understand many of the analyses presented and the conclusion seems highly speculative. I believe a large amount of work is needed to publish this work in a scientific paper.

Specific comments

The Arctic sea ice loss is presented in the introduction and is mentioned in the abstract. But can Arctic sea ice loss release freshwater in summer in the right location? After in the manuscript, L93-98, the Greenland ice sheet melting is mentioned, then the authors say that the scope of the paper is not about understanding the origin of the freshwater. I guess that the introduction and abstract need to be reformulated to have a more balanced picture of the processes releasing freshwater during summer.

In many parts of the manuscript (for instance L59, or legend of Fig. 1), it is argued that a mass balance was used to infer the freshwater release from the SST observation. A reference is given, but can the authors present how this is done. The link between SST and SSS is not obvious and the present paper relies a lot on these previous findings. A presentation of these previous results would improve the manuscript.

In the interpretation, the authors discuss some sharper SST front between the Gulf Stream and the cold anomaly (L110). The location of the North Atlantic Current is also given by a thick arrow Fig. 3. After, in many parts of the manuscripts (L136-140, or L131) are discussed some shifts of the North Atlantic current. However, the SST anomalies in Fig. 2c show large scale SST anomalies rather than sharp fronts. The North Atlantic current is not well located in Fig. 3. I suggest the authors mention a modification of the SST gradients inducing modification of the lower tropospheric baroclinicity. The investigation of the link with the Gulf stream of North Atlantic current would require showing the mean location of the currents with more accuracy, and I am not sure it is needed to explain the large scale atmospheric response.

The manuscript is not based on a quantification of the freshwater released but use the NAO time series from July and August as the starting time series. Why not using the freshwater itself from ERA5? Why not using SSS which could be more related to the freshwater flux. The authors argue that the time series of summer NAO and freshwater are correlated but what does it mean? Can the authors at least suggest some hypothesis behind this statistical relationship? What are the correlations and their p-value? Similarly, the authors used other indices for the freshwater release are used when investigating climate model simulations. The choice of these indices is not well justified, and it seems that different processes are assessed when using different time series, and the link with the freshwater release remains unclear.

The authors chose to subsample their time series so that they have a large relationship between the summer NAO and the SST anomaly in the following winter. In particular, they chose an arbitrary threshold (0.5) and exclude part of the data (one year that seems to be 2014, represented by the yellow point). I do not believe the relationship obtained are representative of the data, as the subsample is somehow selected to have a large relationship. Similarly, later in the manuscript, the time series are again subsampled to build another index in Fig. 8. I am not sure about the interest of doing this.

How the SST impacts the atmosphere and land surfaces in summer is not well discussed or investigated. The impact of the SST on the baroclinic instability and storm tracks are relevant for winter, but in summer other processes might dominate, such as the impacts of the soil moisture or the impact of tropical Atlantic and the intrusion of moist air from the Mediterranean region. In Figs. 4ab only the few wind vectors are shown over the ocean, and it is difficult to see any shift of the jet stream as argued in L136-145. What are the SLP, geopotential height, zonal wind or streamfunction anomalies? Can the Fig. 4ab be extended to include most of Europe and the Mediterranean region? Similarly, when using model results (section 4.4 and Fig. 6), the SLP or the wind is never shown.

The authors argue that ‘’the large-scale dipolar circulation anomaly is reproduced by SST-forced simulations, supporting that it is driven by the ocean (Appendix B)’’ L121-122. What do the authors mean by dipolar? Why are the authors present the results in appendix? When looking at the appendix B, another index is used to characterize the freshwater events (why not using the index built on the NAO??), based on SST. Such regression may reflects here the impact of ENSO on the Atlantic ocean, or the impact of tropical Atlantic, and this cannot be interpreted as an impact of the Subpolar Atlantic SST.

The sea-level anomalies are interesting and show a large band of anti-cyclonic eddies (Fig. 3c). Can the authors discuss these small scale structures? What is the link between the fresh water release and the sea-level anomalies? Maybe a spatial smoothing would be needed to see the large scale structure suggested in the text.

The authors should try to reduce the number of figures and appendices, or better summarize their results. I found the appendix not always relevant. For instance appendix A does not help to understand the surface mass balance and the link with the salinity shown in the main manuscript.

Technical details and other comments:

L35: is the NAO defined as the first or second rotated EOF of monthly 500-hPa geopotential height?

L40 : why not only use HadISST to avoid discontinuity in the dataset used?

L47: can the authors specify if u is the module of the wind or the zonal wind?

L49-55 : can the author specify the boundary condition used for SST and sea ice, as well as the external forcing (for the two experiments). How are generated the initial conditions?

L55:’’we subtracted regionally averaged trends from the air temperatures, both in ERA5 and the model output’’ -> I do not understand what are mean regionally averaged trends. How are the regions defined? I believe that it is important that all trends be removed before calculating the regression. Are the SST trends removed as well?

Figure 1 : The regressions shown are regressions of the SST and SSS variation from summer to winter onto the summer NAO. Are the variations calculated from the previous winter (n-1) to summer n? Or from summer n to next winter n? I do not understand why the authors investigate the SST and SSS variations and not the actual SST and SSS anomalies.

Figure 1 : I do not understand what are the SSS results? Are they from SSS observations? Can the authors provide more details on the method used to retrieve the SSS?

Figure 2 : I believe Fig. 2b shows the regression and not the correlation.

L66 : The authors find also warming in the western Atlantic at 30°N. Can the authors explain the link between the freshwater flux in the subpolar gyre and the SST anomalies in the subtropical region? It seems that the atmosphere is forcing a large part of the signal, with the so-called tripole pattern as a response to the NAO (Czaja and Frankignoul, 2002).

L64: ‘’the relationship between the negative summer NAO and the seasonal surface freshening is approximately linear’’ -> Can the authors explain how this was assessed and analyzed in the data?

L74 : I think that directional t-tests are not justified here. Do the authors mean ‘one-tailed test’? The sign of the regression of the variables studied is not obvious and only two-tailed test are needed here. Can the authors explain?

L86-87 : ‘’After evaluating the surface fluxes, wind-driven Ekman transports, Ekman pumping and re-emergence of SST anomalies’’ Can the authors explain better where and how these processes are evaluated?

L89-92: I believe this needs to be better explained. Does the authors assume a perfect density compensation to deduce the SSS? It does not explain how the entrainment below the mixed layer and the re-emergence are evaluated then.

L95 : I do not see a pronounced seasonality of the […] surface freshening in Fig. 1d. Can the authors explain what is the seasonality of the surface freshening and how the anomalies observed reinforce this seasonality?

L93-98 : I would rather link the freshening with P-E, and I do not understand well the hypothesis that runoff from Greenland is dominant here.

Figure 3b : The SLP anomalies are huge. Maybe hPa are Pa?

L110: I do not understand what the authors mean with ‘’ after stronger relative to weaker freshwater events ‘’. What not just say ‘’after the large freshwater events’’?

Figure 3c: the data used for the ADT need to be presented in the method section.

Figure 3c: the thin black arrow shows the flow implied by the ADT anomaly. What does it mean? I am surprised that such flow is not geostrophic... Can the authors explain how the arrows are computed?

L123-125 : ‘’most negative NAO summers are followed by a positive NAO in the subsequent winter’’ -> This statement is not supported by the results presented so far, as the regressions shown in observation are built using only 8 winters.

L130 : ‘’the northward shift of the North Atlantic Current is obscured by the southern Ekman flow’’ -> Many studies argue that the heat flux is dominant in driving the SST anomalies during the NAO, while the Ekman flow drives weak anomalies (Deser et al., 2010). A more accurate presentation of the terms driving the SST anomalies is required to support this statement.

L133-134 : ‘’an increasingly sharpened SST front all across the eastern boundary of the North Atlantic (Fig. 3d)’’ -> Can the authors describe where are these fronts in the eastern Atlantic in Fig. 3d? The SST in the second winter looks similar to that in the first winter, but weaker.

L141: ‘’it shields the regions to the south from the moist air over the Atlantic’’ I do not understand this statement. Can the authors reformulate?

L146: ‘’the regressions […] are […] characterized by steep slopes and high correlation’’ Note that the correlation is never shown in figures, so that the authors may provide hear some number to support this statement. The authors should note that with 8 points, the threshold for a significant correlation is 0.707 for a p-value at 5%. Therefore high correlation does not necessarily mean a significant relationship. I would remove this comment, and I would only comment the level of statistical significance and not the amplitude of the correlation.

Fig. 4ef and Fig. 5ab, the values for P-E are huge. Can the authors check if the map shows correlation and not regression.

Fig. 5cd: I do not see why the authors show these figures… I would remove them.

L154-157: I do not understand why another index for the freshwater release is used. I do not understand what it is? For instance, L156-157 ‘’we map the SST each summer onto the observed pattern obtained from freshwater event. The mapping is obtained from a least square fit of the SST […] to the SST pattern obtained from the freshwater events (Figs. 6a and 6b)’’. How are the freshwater event defined? Does the author mean that the time series is obtained using a projection onto a spatial pattern over a specified region (that needs to be defined)?

L159-160 ; ‘’we find that the observed and simulated atmospheric response agree qualitatively well’’ Can the authors explain the difference between Fig. 4 and Fig. 6. It seems that the SST anomaly is different in Fig. 6, with a clear SST tripole, with large impacts on the P-E. It seems that the simulation and models show a feedback between warming over the continent and soil moisture decrease. Did the authors use detrended data? If not, the authors might see here the impacts of global warming over the summer continents.

L173 : ‘’we […] refer to these freshwater events as circulation-driven events’’. The authors argue that explaining the origin of the freshwater anomalies is beyond the scope of the paper. Therefore, I suggest changing the name of these events, otherwise, the authors should justify how the circulation explains the freshwater anomalies.

L182-183: ‘’we exclude events, that are preceded by another strong circulation-driven freshwater event, for which F_C is larger than 0.2’’. Can the authors justify the choice of 0.2 here? Did they test other values?

L184 : ‘’correlation with the SST gradient’’ I do not see any figure of the correlation SST gradient in Fig. 8, but some regression of the SST.

Fig. 10c and 10d :I do not see any level of statistical significance. Does it mean that the SSS or SST in winter is not related to the temperature in summer?

Figure 7: how are the outlier defined in the box plot of Figure 7. I believe that showing the 5% and 95% percentile would be useful, as observation seems to be at the edge of the simulated distribution concerning precipitation minus evaportation. The authors should also note that the observation never lies in the interval defined by the first and third quartiles.

Figure 10a : How is T_summer defined? What region? Is it surface air temperature over land only?

Figure 11: This figure does not add much to the manuscript.

Appendix C: the results are not discussed in the main text. Note sure that the appendix C is needed, the authors may summarize the results by one or two sentence in the main manuscript with ‘’not shown’’.

References

Deser, C., Alexander, M. A., Xie, S. P., & Phillips, A. S. (2010). Sea surface temperature variability: Patterns and mechanisms. , , 115-143.

Czaja, A., & Frankignoul, C. (2002). Observed impact of Atlantic SST anomalies on the North Atlantic Oscillation. , (6), 606-623.
Citation: https://doi.org/10.5194/wcd-2021-79-RC2
- AC2: 'Reply on RC2', Marilena Oltmanns, 16 Mar 2022
  
  The comment was uploaded in the form of a supplement: https://wcd.copernicus.org/preprints/wcd-2021-79/wcd-2021-79-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/wcd-2021-79-AC2
RC3:
'Comment on wcd-2021-79', Anonymous Referee #3, 05 Feb 2022

This study investigates a relationship between freshwater anomalies in

the North Atlantic and summer European climate up to several years

later. The proposed mechanism involves cooling over the subpolar region

and warming over the subtropical region that increases the meridional

temperature gradient, leading to enhanced baroclinicity that alters the

atmospheric circulation. The physical relationships are plausible and

there are some interesting implications for predictability. However, I

find the approach and manuscript quite confusing, and I believe major

revisions would be required before publication.

Main points

1) I am not clear on whether the analyses actually address the role of

freshwater events on European climate. The authors spend quite a bit of

time establishing that the relationship between the NAO index and the

freshwater events in the period studied is robust and useful (mainly

based on previous studies), and hence that the NAO index can be used as

a proxy for freshwater anomalies. The justification/explanation comes

back in several places throughout the manuscript, perhaps drawing more

attention to it than the authors intended. However, I did not fully

follow many some aspects of the justification (e.g., a number of other

possibilities are eliminated in L86-87, but the explanation is quite

brief and as far as I can tell, only focuses on Ekman processes ). The

main question I was left with was, why not just use an index of

freshwater anomalies? Perhaps there is an obvious answer here, but it

didn't come through to me in the manuscript, and makes statements like

L102-103 quite unsatisfying.

2) In general, it would be extremely helpful to clarify what this study

is about and to choose an analysis strategy that directly addresses the

problem. The idea of circulation-induced versus melt-driven freshwater

events in section 4.5 came as a surprise to me. In fact, I only realized

that F_M and F_C (introduced earlier) are related to this, but had spent

quite a bit of the manuscript until then puzzled by the names. Is it

really the NAO index that's used to discriminate between these types of

events? These ideas should probably be introduced in section 1, as they

seem to motivate quite a bit of the study. Interestingly, section 1 as

written seems more focused on sea ice loss and the origin of summertime

freshwater, but later, the manuscript states that this isn't the focus

of the study.

Other points that may or may not be relevant once the main comments are

addressed:

3) If the negative NAO index is kept: It's quite confusing to talk about

more negative or more positive values of the negative NAO index. I think

it's fine to flip the NAO index, but perhaps the text should just talk

about higher or lower values of the NAO. Also, I don't think the NAO

index was detrended, but 2m temperatures were detrended. What is the

reason for this? If trends are kept in, then the autocorrelation needs

to be accounted for in subsequent statistical analyses.

4) Some of the oceanography concepts could be better explained for the

non-oceanographers, and the same goes for the atmospheric concepts.

e.g., L89 "the mass increase, implied by the cold anomaly,..."; L112-115

connection between poleward vorticity transport and momentum tranfer

from STJ to EDJ, L138-140 is there some relevant theory for the time

scales behind the delay in the shift of the North Atlantic Current?

5) L 148 "succesfully extracts..." Perhaps related to my general

confusion about F_C and F_M, I don't have a good feel for how downstream

effects from other drivers and IV would influence F_M, so this statement

is difficult to understand.

6) L224: This first line of the conclusions is not representative of the

main message of this study, is it?

Technical points:

-L61: "well-correlated" should be quantified if the NAO is kept

-Fig 1a is encapsulated in Fig. 2a - maybe don't need both?

-L84: Fig 2d is SSS?

-L127: the increase in sea level height is just in the subtropical gyre?

Citation: https://doi.org/10.5194/wcd-2021-79-RC3
- AC3: 'Reply on RC3', Marilena Oltmanns, 16 Mar 2022
  
  The comment was uploaded in the form of a supplement: https://wcd.copernicus.org/preprints/wcd-2021-79/wcd-2021-79-AC3-supplement.pdf
  
  Citation: https://doi.org/10.5194/wcd-2021-79-AC3

Status: closed

RC1:
'Comment on wcd-2021-79', Anonymous Referee #1, 06 Jan 2022

North Atlantic freshwater events influence European weather in subsequent summers, by Oltmanns et al

This manuscript examins the impact of summertime NAO events on European weather in subsequent summers. The paper is framed as being the impacts of North Atlantic freshwater events on European weather but the index that is used to depict these "freshwater events" is actually the summer NAO. The paper demonstrates some interesting connections between the summer NAO and the following summer weather. At this point, I'm unsure as to whether this manuscript is acceptable for publication. I have a number of comments on the analysis as outlined below. Overall, I'm giving a recommendation of major revisions to allow the authors to respond to these. My major concerns are that the direct link between the summer NAO and the freshwater events is unclear to me. This may be because I'm not an oceanographer and I haven't read the authors previous papers, so I hope that one of the other reviewers will be able to assess this aspect. I have some other concerns about the statistical methods used and the choices made for the scatter plots as outlined in my comments below.

General comments:

(1) The link to freshwater anomalies and the role of low frequency North Atlantic ocean variability. I am not an oceanographer, so I hope that one of the other reviewers will have the expertise to comment on this. The link between the summer NAO index and the freshwater anomalies was a bit lost on me. My understanding of lines 84-92 is that the authors are assuming that the temperature anomalies associated with the summer NAO are due to freshwater anomalies because they find that the cooling is not strongly related to surface fluxes, wind driven Ekman transports, Ekman pumping and re-emergence of SST anomalies from previous years, so by a process of elimination they conclude that it's freshwater anomalies. But I don't see how the role for other ocean circulation anomalies such as the AMOC or advective heat convergence due to circulation anomalies produced by things other than the wind driven Ekman transports has been eliminated. The atlantic ocean circulation exhibits variability on long timescales which can be a driver of the NAO and vice-versa (e.g., Zhang et al 2019, Review of Geophysics, 10.1029/2019RG000644 and references therin). It's not clear to me (a) whether it can really be concluded that the SST anomalies are related to freshwater inputs and (b) whether such low frequency variability in the ocean circulation has been appropriately taken into account. The NAO index is being described as a "freshwater index" (l99) but I'm not sure how appropriate this is and I'm not sure that much would be lost by instead referring to it as the NAO index and focussing on the impact of the summertime NAO on the climate in subsequent years.

(2) Detrending: It's stated at line 53-55 that regionally averaged trends were subtracted from the air temperatures to remove the greenhouse gas effects. It doesn't really seem appropriate to me to remove the linear trend from one field but not others. The NAO index that is used clearly has a linear trend in it (Figure 1a). I'd suggest detrending everything or detrending nothing. I'm not arguing that the NAO trend seen in Figure 1 is greenhouse gas forced or that this trend should necessarily be removed, but it just doesn't seem appropriate to me to remove the trend in one field and not in the others. Is the detrending also done on the SSTs? It doesn't make much sense to me to remove the trend from the surface air temperature but not the SSTs.

(3) For the scatter plots, the regions where the correlation is significant at the 95% level is used for the spatial averaging. This seems like cherry picking to me. Of course the correlations look good because you've chosen them to be that way. It would make more sense to choose a physically motivated region or it would seem to be less cherry picking if a regular spatial region such as a rectangle were chosen. The result is that in Figure 2b there is a correlation of 0.98, which seems quite unbelievable to me, but maybe it isn't if you are just averaging over regions where the correlation is high.

(4) Has autocorrelation been accounted for when calculating the significance levels? If not, I think it should be. Clearly each year is not independent and there is some low frequency variability and autocorrelation, as apparent in the NAO index (Figure 2a).

(5) It is argued that this work reveals new potential to enhance the predictability of European summer weather, but I think for the impacts on European summer weather the results have only been presented in the form of regression coefficients. To make this more relevant for predictability, it might be worth showing the variance explained.

Comments by line number:

Figure 2: It seems like it would be interesting to have the regression maps for F_C as well as F_M. You use the regions based on the regression onto F_M for both F_M and F_C, so it would be good to see whether the regression map for F_C has a similar spatial pattern to that for F_M or not.

Figure 2 caption: F_M and F_C are only defined in this figure caption. Given their central importance, I think they should also be defined in the text. Furthermore, it would be worthwhile making clear the motivations for this naming convention. It's not very intuitive where the choice of "F_M" and F_C" comes from and I think it would help readers to follow if you make that clear. In the end, I realized that this corresponds to "melt-driven' and "circulation-driven" events and I'm overall just very confused about how this distinction can be made just on the basis of the NAO index, which relates to my general comments above. I think this needs to be made clearer throughout the manuscript.

Figure 3 caption: Maybe explain a bit more what the "absolute dynamic topography" is. Is it just sea surface height?

l138: "expansion of the cold anomaly" - perhaps be clear about what this "expansion' is relative to? Is it relative to the previous summer?

Section 4.5: It might be worth making it clear at the beginning here that this is now back to looking at the observations, since in the previous section the focus was on model simulations.

Figure 11 caption: The referencing to the panel labels is messed up in the caption.

Typo's/wording suggestions:

l62: "this index" --> "the NAO index"

l68: suggest "smaller values" --> "more negative values" because the magnitude of the NAO index isn't smaller.

l84: "Fig. 2d" --> "Fig. 2c" (I think d is showing salinity, not temperature)

Figure 4 caption: "The thick contours show the 95% confidence levels" --> "The thick contours encompass regions that are significant at the 95% confidence levels"

l158: "SST-forced" is a bit unclear. Suggest "Simulations performed with prescribed observation-based SSTs".

Citation: https://doi.org/10.5194/wcd-2021-79-RC1
- AC1: 'Reply on RC1', Marilena Oltmanns, 16 Mar 2022
  
  The comment was uploaded in the form of a supplement: https://wcd.copernicus.org/preprints/wcd-2021-79/wcd-2021-79-AC1-supplement.pdf
  
  Citation: https://doi.org/10.5194/wcd-2021-79-AC1
RC2:
'Comment on wcd-2021-79', Anonymous Referee #2, 26 Jan 2022
General comments

The relationship between the Atlantic ocean and the summer climate over Europe is investigated. It is argued that events releasing freshwater into the North Atlantic subpolar gyre are followed by a persistent cooling (warming) over the subpolar (western subtropical) gyre. Such modified SST is linked to warm and dry conditions over western Europe in the next two subsequent summers.

While the overall mechanism seems realistic, the overall presentation of the results is very confusing. I did not understand the link between the SST anomalies analyzed and the freshwater release in the manuscript. Similarly, I was not able to understand many of the analyses presented and the conclusion seems highly speculative. I believe a large amount of work is needed to publish this work in a scientific paper.

Specific comments

The Arctic sea ice loss is presented in the introduction and is mentioned in the abstract. But can Arctic sea ice loss release freshwater in summer in the right location? After in the manuscript, L93-98, the Greenland ice sheet melting is mentioned, then the authors say that the scope of the paper is not about understanding the origin of the freshwater. I guess that the introduction and abstract need to be reformulated to have a more balanced picture of the processes releasing freshwater during summer.

In many parts of the manuscript (for instance L59, or legend of Fig. 1), it is argued that a mass balance was used to infer the freshwater release from the SST observation. A reference is given, but can the authors present how this is done. The link between SST and SSS is not obvious and the present paper relies a lot on these previous findings. A presentation of these previous results would improve the manuscript.

In the interpretation, the authors discuss some sharper SST front between the Gulf Stream and the cold anomaly (L110). The location of the North Atlantic Current is also given by a thick arrow Fig. 3. After, in many parts of the manuscripts (L136-140, or L131) are discussed some shifts of the North Atlantic current. However, the SST anomalies in Fig. 2c show large scale SST anomalies rather than sharp fronts. The North Atlantic current is not well located in Fig. 3. I suggest the authors mention a modification of the SST gradients inducing modification of the lower tropospheric baroclinicity. The investigation of the link with the Gulf stream of North Atlantic current would require showing the mean location of the currents with more accuracy, and I am not sure it is needed to explain the large scale atmospheric response.

The manuscript is not based on a quantification of the freshwater released but use the NAO time series from July and August as the starting time series. Why not using the freshwater itself from ERA5? Why not using SSS which could be more related to the freshwater flux. The authors argue that the time series of summer NAO and freshwater are correlated but what does it mean? Can the authors at least suggest some hypothesis behind this statistical relationship? What are the correlations and their p-value? Similarly, the authors used other indices for the freshwater release are used when investigating climate model simulations. The choice of these indices is not well justified, and it seems that different processes are assessed when using different time series, and the link with the freshwater release remains unclear.

The authors chose to subsample their time series so that they have a large relationship between the summer NAO and the SST anomaly in the following winter. In particular, they chose an arbitrary threshold (0.5) and exclude part of the data (one year that seems to be 2014, represented by the yellow point). I do not believe the relationship obtained are representative of the data, as the subsample is somehow selected to have a large relationship. Similarly, later in the manuscript, the time series are again subsampled to build another index in Fig. 8. I am not sure about the interest of doing this.

How the SST impacts the atmosphere and land surfaces in summer is not well discussed or investigated. The impact of the SST on the baroclinic instability and storm tracks are relevant for winter, but in summer other processes might dominate, such as the impacts of the soil moisture or the impact of tropical Atlantic and the intrusion of moist air from the Mediterranean region. In Figs. 4ab only the few wind vectors are shown over the ocean, and it is difficult to see any shift of the jet stream as argued in L136-145. What are the SLP, geopotential height, zonal wind or streamfunction anomalies? Can the Fig. 4ab be extended to include most of Europe and the Mediterranean region? Similarly, when using model results (section 4.4 and Fig. 6), the SLP or the wind is never shown.

The authors argue that ‘’the large-scale dipolar circulation anomaly is reproduced by SST-forced simulations, supporting that it is driven by the ocean (Appendix B)’’ L121-122. What do the authors mean by dipolar? Why are the authors present the results in appendix? When looking at the appendix B, another index is used to characterize the freshwater events (why not using the index built on the NAO??), based on SST. Such regression may reflects here the impact of ENSO on the Atlantic ocean, or the impact of tropical Atlantic, and this cannot be interpreted as an impact of the Subpolar Atlantic SST.

The sea-level anomalies are interesting and show a large band of anti-cyclonic eddies (Fig. 3c). Can the authors discuss these small scale structures? What is the link between the fresh water release and the sea-level anomalies? Maybe a spatial smoothing would be needed to see the large scale structure suggested in the text.

The authors should try to reduce the number of figures and appendices, or better summarize their results. I found the appendix not always relevant. For instance appendix A does not help to understand the surface mass balance and the link with the salinity shown in the main manuscript.

Technical details and other comments:

L35: is the NAO defined as the first or second rotated EOF of monthly 500-hPa geopotential height?

L40 : why not only use HadISST to avoid discontinuity in the dataset used?

L47: can the authors specify if u is the module of the wind or the zonal wind?

L49-55 : can the author specify the boundary condition used for SST and sea ice, as well as the external forcing (for the two experiments). How are generated the initial conditions?

L55:’’we subtracted regionally averaged trends from the air temperatures, both in ERA5 and the model output’’ -> I do not understand what are mean regionally averaged trends. How are the regions defined? I believe that it is important that all trends be removed before calculating the regression. Are the SST trends removed as well?

Figure 1 : The regressions shown are regressions of the SST and SSS variation from summer to winter onto the summer NAO. Are the variations calculated from the previous winter (n-1) to summer n? Or from summer n to next winter n? I do not understand why the authors investigate the SST and SSS variations and not the actual SST and SSS anomalies.

Figure 1 : I do not understand what are the SSS results? Are they from SSS observations? Can the authors provide more details on the method used to retrieve the SSS?

Figure 2 : I believe Fig. 2b shows the regression and not the correlation.

L66 : The authors find also warming in the western Atlantic at 30°N. Can the authors explain the link between the freshwater flux in the subpolar gyre and the SST anomalies in the subtropical region? It seems that the atmosphere is forcing a large part of the signal, with the so-called tripole pattern as a response to the NAO (Czaja and Frankignoul, 2002).

L64: ‘’the relationship between the negative summer NAO and the seasonal surface freshening is approximately linear’’ -> Can the authors explain how this was assessed and analyzed in the data?

L74 : I think that directional t-tests are not justified here. Do the authors mean ‘one-tailed test’? The sign of the regression of the variables studied is not obvious and only two-tailed test are needed here. Can the authors explain?

L86-87 : ‘’After evaluating the surface fluxes, wind-driven Ekman transports, Ekman pumping and re-emergence of SST anomalies’’ Can the authors explain better where and how these processes are evaluated?

L89-92: I believe this needs to be better explained. Does the authors assume a perfect density compensation to deduce the SSS? It does not explain how the entrainment below the mixed layer and the re-emergence are evaluated then.

L95 : I do not see a pronounced seasonality of the […] surface freshening in Fig. 1d. Can the authors explain what is the seasonality of the surface freshening and how the anomalies observed reinforce this seasonality?

L93-98 : I would rather link the freshening with P-E, and I do not understand well the hypothesis that runoff from Greenland is dominant here.

Figure 3b : The SLP anomalies are huge. Maybe hPa are Pa?

L110: I do not understand what the authors mean with ‘’ after stronger relative to weaker freshwater events ‘’. What not just say ‘’after the large freshwater events’’?

Figure 3c: the data used for the ADT need to be presented in the method section.

Figure 3c: the thin black arrow shows the flow implied by the ADT anomaly. What does it mean? I am surprised that such flow is not geostrophic... Can the authors explain how the arrows are computed?

L123-125 : ‘’most negative NAO summers are followed by a positive NAO in the subsequent winter’’ -> This statement is not supported by the results presented so far, as the regressions shown in observation are built using only 8 winters.

L130 : ‘’the northward shift of the North Atlantic Current is obscured by the southern Ekman flow’’ -> Many studies argue that the heat flux is dominant in driving the SST anomalies during the NAO, while the Ekman flow drives weak anomalies (Deser et al., 2010). A more accurate presentation of the terms driving the SST anomalies is required to support this statement.

L133-134 : ‘’an increasingly sharpened SST front all across the eastern boundary of the North Atlantic (Fig. 3d)’’ -> Can the authors describe where are these fronts in the eastern Atlantic in Fig. 3d? The SST in the second winter looks similar to that in the first winter, but weaker.

L141: ‘’it shields the regions to the south from the moist air over the Atlantic’’ I do not understand this statement. Can the authors reformulate?

L146: ‘’the regressions […] are […] characterized by steep slopes and high correlation’’ Note that the correlation is never shown in figures, so that the authors may provide hear some number to support this statement. The authors should note that with 8 points, the threshold for a significant correlation is 0.707 for a p-value at 5%. Therefore high correlation does not necessarily mean a significant relationship. I would remove this comment, and I would only comment the level of statistical significance and not the amplitude of the correlation.

Fig. 4ef and Fig. 5ab, the values for P-E are huge. Can the authors check if the map shows correlation and not regression.

Fig. 5cd: I do not see why the authors show these figures… I would remove them.

L154-157: I do not understand why another index for the freshwater release is used. I do not understand what it is? For instance, L156-157 ‘’we map the SST each summer onto the observed pattern obtained from freshwater event. The mapping is obtained from a least square fit of the SST […] to the SST pattern obtained from the freshwater events (Figs. 6a and 6b)’’. How are the freshwater event defined? Does the author mean that the time series is obtained using a projection onto a spatial pattern over a specified region (that needs to be defined)?

L159-160 ; ‘’we find that the observed and simulated atmospheric response agree qualitatively well’’ Can the authors explain the difference between Fig. 4 and Fig. 6. It seems that the SST anomaly is different in Fig. 6, with a clear SST tripole, with large impacts on the P-E. It seems that the simulation and models show a feedback between warming over the continent and soil moisture decrease. Did the authors use detrended data? If not, the authors might see here the impacts of global warming over the summer continents.

L173 : ‘’we […] refer to these freshwater events as circulation-driven events’’. The authors argue that explaining the origin of the freshwater anomalies is beyond the scope of the paper. Therefore, I suggest changing the name of these events, otherwise, the authors should justify how the circulation explains the freshwater anomalies.

L182-183: ‘’we exclude events, that are preceded by another strong circulation-driven freshwater event, for which F_C is larger than 0.2’’. Can the authors justify the choice of 0.2 here? Did they test other values?

L184 : ‘’correlation with the SST gradient’’ I do not see any figure of the correlation SST gradient in Fig. 8, but some regression of the SST.

Fig. 10c and 10d :I do not see any level of statistical significance. Does it mean that the SSS or SST in winter is not related to the temperature in summer?

Figure 7: how are the outlier defined in the box plot of Figure 7. I believe that showing the 5% and 95% percentile would be useful, as observation seems to be at the edge of the simulated distribution concerning precipitation minus evaportation. The authors should also note that the observation never lies in the interval defined by the first and third quartiles.

Figure 10a : How is T_summer defined? What region? Is it surface air temperature over land only?

Figure 11: This figure does not add much to the manuscript.

Appendix C: the results are not discussed in the main text. Note sure that the appendix C is needed, the authors may summarize the results by one or two sentence in the main manuscript with ‘’not shown’’.

References

Deser, C., Alexander, M. A., Xie, S. P., & Phillips, A. S. (2010). Sea surface temperature variability: Patterns and mechanisms. , , 115-143.

Czaja, A., & Frankignoul, C. (2002). Observed impact of Atlantic SST anomalies on the North Atlantic Oscillation. , (6), 606-623.
Citation: https://doi.org/10.5194/wcd-2021-79-RC2
- AC2: 'Reply on RC2', Marilena Oltmanns, 16 Mar 2022
  
  The comment was uploaded in the form of a supplement: https://wcd.copernicus.org/preprints/wcd-2021-79/wcd-2021-79-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/wcd-2021-79-AC2
RC3:
'Comment on wcd-2021-79', Anonymous Referee #3, 05 Feb 2022

This study investigates a relationship between freshwater anomalies in

the North Atlantic and summer European climate up to several years

later. The proposed mechanism involves cooling over the subpolar region

and warming over the subtropical region that increases the meridional

temperature gradient, leading to enhanced baroclinicity that alters the

atmospheric circulation. The physical relationships are plausible and

there are some interesting implications for predictability. However, I

find the approach and manuscript quite confusing, and I believe major

revisions would be required before publication.

Main points

1) I am not clear on whether the analyses actually address the role of

freshwater events on European climate. The authors spend quite a bit of

time establishing that the relationship between the NAO index and the

freshwater events in the period studied is robust and useful (mainly

based on previous studies), and hence that the NAO index can be used as

a proxy for freshwater anomalies. The justification/explanation comes

back in several places throughout the manuscript, perhaps drawing more

attention to it than the authors intended. However, I did not fully

follow many some aspects of the justification (e.g., a number of other

possibilities are eliminated in L86-87, but the explanation is quite

brief and as far as I can tell, only focuses on Ekman processes ). The

main question I was left with was, why not just use an index of

freshwater anomalies? Perhaps there is an obvious answer here, but it

didn't come through to me in the manuscript, and makes statements like

L102-103 quite unsatisfying.

2) In general, it would be extremely helpful to clarify what this study

is about and to choose an analysis strategy that directly addresses the

problem. The idea of circulation-induced versus melt-driven freshwater

events in section 4.5 came as a surprise to me. In fact, I only realized

that F_M and F_C (introduced earlier) are related to this, but had spent

quite a bit of the manuscript until then puzzled by the names. Is it

really the NAO index that's used to discriminate between these types of

events? These ideas should probably be introduced in section 1, as they

seem to motivate quite a bit of the study. Interestingly, section 1 as

written seems more focused on sea ice loss and the origin of summertime

freshwater, but later, the manuscript states that this isn't the focus

of the study.

Other points that may or may not be relevant once the main comments are

addressed:

3) If the negative NAO index is kept: It's quite confusing to talk about

more negative or more positive values of the negative NAO index. I think

it's fine to flip the NAO index, but perhaps the text should just talk

about higher or lower values of the NAO. Also, I don't think the NAO

index was detrended, but 2m temperatures were detrended. What is the

reason for this? If trends are kept in, then the autocorrelation needs

to be accounted for in subsequent statistical analyses.

4) Some of the oceanography concepts could be better explained for the

non-oceanographers, and the same goes for the atmospheric concepts.

e.g., L89 "the mass increase, implied by the cold anomaly,..."; L112-115

connection between poleward vorticity transport and momentum tranfer

from STJ to EDJ, L138-140 is there some relevant theory for the time

scales behind the delay in the shift of the North Atlantic Current?

5) L 148 "succesfully extracts..." Perhaps related to my general

confusion about F_C and F_M, I don't have a good feel for how downstream

effects from other drivers and IV would influence F_M, so this statement

is difficult to understand.

6) L224: This first line of the conclusions is not representative of the

main message of this study, is it?

Technical points:

-L61: "well-correlated" should be quantified if the NAO is kept

-Fig 1a is encapsulated in Fig. 2a - maybe don't need both?

-L84: Fig 2d is SSS?

-L127: the increase in sea level height is just in the subtropical gyre?

Citation: https://doi.org/10.5194/wcd-2021-79-RC3
- AC3: 'Reply on RC3', Marilena Oltmanns, 16 Mar 2022
  
  The comment was uploaded in the form of a supplement: https://wcd.copernicus.org/preprints/wcd-2021-79/wcd-2021-79-AC3-supplement.pdf
  
  Citation: https://doi.org/10.5194/wcd-2021-79-AC3

Marilena Oltmanns, N. Penny Holliday, James Screen, D. Gwyn Evans, Simon A. Josey, Sheldon Bacon, and Ben I. Moat

Viewed

Total article views: 1,304 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
827	440	37	1,304	21	18

HTML: 827
PDF: 440
XML: 37
Total: 1,304
BibTeX: 21
EndNote: 18

Views and downloads (calculated since 06 Dec 2021)

Month	HTML	PDF	XML	Total
Dec 2021	151	65	2	218
Jan 2022	83	23	5	111
Feb 2022	76	22	4	102
Mar 2022	57	25	9	91
Apr 2022	38	18	0	56
May 2022	26	9	1	36
Jun 2022	20	4	1	25
Jul 2022	36	42	1	79
Aug 2022	23	16	0	39
Sep 2022	6	11	0	17
Oct 2022	20	10	2	32
Nov 2022	23	13	0	36
Dec 2022	12	10	0	22
Jan 2023	22	21	1	44
Feb 2023	23	8	0	31
Mar 2023	23	9	0	32
Apr 2023	9	12	0	21
May 2023	15	6	1	22
Jun 2023	13	15	0	28
Jul 2023	13	14	0	27
Aug 2023	6	12	0	18
Sep 2023	17	20	0	37
Oct 2023	17	8	1	26
Nov 2023	25	9	0	34
Dec 2023	17	2	0	19
Jan 2024	18	4	2	24
Feb 2024	9	11	1	21
Mar 2024	14	17	2	33
Apr 2024	15	4	4	23

Cumulative views and downloads (calculated since 06 Dec 2021)

Month	HTML	PDF	XML	Total
Dec 2021	151	65	2	218
Jan 2022	83	23	5	111
Feb 2022	76	22	4	102
Mar 2022	57	25	9	91
Apr 2022	38	18	0	56
May 2022	26	9	1	36
Jun 2022	20	4	1	25
Jul 2022	36	42	1	79
Aug 2022	23	16	0	39
Sep 2022	6	11	0	17
Oct 2022	20	10	2	32
Nov 2022	23	13	0	36
Dec 2022	12	10	0	22
Jan 2023	22	21	1	44
Feb 2023	23	8	0	31
Mar 2023	23	9	0	32
Apr 2023	9	12	0	21
May 2023	15	6	1	22
Jun 2023	13	15	0	28
Jul 2023	13	14	0	27
Aug 2023	6	12	0	18
Sep 2023	17	20	0	37
Oct 2023	17	8	1	26
Nov 2023	25	9	0	34
Dec 2023	17	2	0	19
Jan 2024	18	4	2	24
Feb 2024	9	11	1	21
Mar 2024	14	17	2	33
Apr 2024	15	4	4	23

Viewed (geographical distribution)

Total article views: 1,249 (including HTML, PDF, and XML) Thereof 1,249 with geography defined and 0 with unknown origin.

Country	#	Views	%

Cited

Latest update: 25 Apr 2024

Short summary

The Arctic is currently warming twice as fast as the global average. This results in enhanced melting and thus freshwater releases into the North Atlantic. Using a combination of observations and models, we show that atmosphere-ocean feedbacks initiated by freshwater releases into the North Atlantic lead to warmer and drier weather over Europe in subsequent summers. The existence of this dynamical link suggests that European summer weather can potentially be predicted months to years in advance.


Total:	0
HTML:	0
PDF:	0
XML:	0

North Atlantic freshwater events influence European weather in subsequent summers

Viewed

Viewed (geographical distribution)

Cited

1 citations as recorded by crossref.