Review of «A dynamical adjustment perspective on extreme event attribution» by Laurent Terray (wcd-2021-40)

General Comments

The manuscript is well-written, follows a clear narrative and the conclusions are supported by the analysis and literature cited in the paper. I especially appreciate the comprehensive literature review and that the observation-based results are frequently compared to results from modelling studies and vice versa. The study complements existing work in the field and provides added value to the understanding of past extreme events and long-term changes in extreme indicators. It shows that the dynamical adjustment method is a practical and versatile approach that can even be considered for rapid attribution of extreme events.

I was wondering whether there is similar year-to-year variability in the contribution of the dynamic component to changes in TXx and TNx as there is on a daily basis for the specific events. Below I also added a comment about this point. Maybe the author has already done some analysis in this direction that he can share.

Overall, I am happy to recommend publication and I only have minor comments and questions to the manuscript, which are given below.

Minor comments

line 37: The second approach of extreme event attribution is introduced as the “process-based or storyline approach”. To me, the term «storyline» might need a little more explanation in this context as I think I would not name every process-based study a storyline. Both are not probabilistic and aim to understand the driving factors. However, following e.g. Shepherd et al. 2018, storylines also explore different plausible climates, which is not done in Wehrli et al. 2019 or the present study. I am aware that it is probably not possible to make a clear distinction between a process-based or storyline study in every case. It might help if the author could share his definition of a storyline.

l. 60: I would leave away the word “tropical” as the heatwaves examined in Wehrli et al. 2019 were not really tropical events even if parts of the regions examined for Australia and South Africa can be classified as tropical/sub-tropical climate.

l. 98: It would be nice to briefly mention what method was used in Horton et al. 2015 as the study will be referenced also later in the manuscript.

l. 278: Do you mean “more persistent” instead of “intense”? The first spell in mid-December looks more intense to me than the second.

l. 397: Could this trend in TX due to dynamics be related to one extreme event in the recent years such as the Russian heatwave? Would the maps look different if you left out 2010? And is there a lot of year-to-year variability in the contribution by dynamics?

l. 401: Do you have a hypothesis why the thermodynamic component trend is overall cooling TNx? Would you expect this to be due to the local or the remote contribution associated with internal variability?

Reference:

Shepherd, T.G., Boyd, E., Calel, R.A. et al. Storylines: an alternative approach to representing uncertainty in physical aspects of climate change. Climatic Change 151, 555–571 (2018). https://doi.org/10.1007/s10584-018-2317-9

Details, typing errors, etc.

To ensure reproducibility of the study the location and extent of the SLP regions that were used for the two events in 3.1 and 3.2 still need to be specified . I only found the numbers for the TX regions in the tables.

l. 103: Do you mean TN maxima?

l. 115: Later in the manuscript (e.g. Table 1) an underscore is used before the version number in 20CR_V3

l. 220: The period chosen for the cold European winter is not exactly two weeks but 17 days

l. 237: Do you mean Fig. 1b?

l. 254: insert “the” before amplitude

Fig. 1: I find it hard to distinguish the one contour line that is thicker. Line thickness could be increased or the zero SLP anomaly line could be highlighted in a different colour e.g. violet.

l. 269: typo, should be 2009-2010 early winter

l. 270: Shouldn’t the numbers in the brackets be -3.07 °C and -2.04 °C?

l. 271: 20CR_V3

l. 336 I would close the first bracket after “anomaly”

l. 381: For the TX, TN and the SLP domain it should say °E instead of °W.

l. 387: I think the bracket saying “smallest” should be omitted as you are using the largest anomalies whether it is TX or TN.

l. 444: Was it not mitigated by around one third from -3.07°C to -2.04°C (instead of 50%)?

This is a nice paper which applies the existing method of dynamical adjustment to investigate some aspects of European climate change and extreme weather. The paper is sound and well written, and I am supportive of acceptance after the points below have been considered. My main comment is that the paper could be improved by the addition of some uncertainty analysis. 

1. The paper is aimed at the event attribution problem, ie in quantifying the role of climate change in extreme weather events. This is very clear in the abstract and the introduction. However, this specific method instead quantifies the circulation contribution, and any climate change effects are left in the residual. This is still very useful, but it could be made clearer early on that this method alone cannot make statements about the role of any particular external forcing, such as greenhouse gases or aerosols. 

2. Given that the main aim of the paper is to quantify the role of atmospheric circulation, it seems the method could be improved by adding an estimate of the uncertainty in this quantification. Even very close analogues in surface pressure will likely have differences in temperature due to large-scale effects not captured by the surface pressure. Hence, to complement the mean effect of circulation as used here, an uncertainty range could be given. More rigorous statements could then be made about the residual terms, in particular as to whether they are within the uncertainty range of the dynamical contribution or not. 

3. One term is labelled RES_ADV and frequently discussed as representing advection, but this is never tested. Changes in advection will indeed contribute to this term but it is not clear that other processes do not contribute. This seems especially likely in summer, when advection is relatively less important for temperature variability and other factors such as radiation or adiabatic heating anomalies may play an important role (eg Pfahl and Wernli, Quinting and Reeder). Could this term be re-named, or the role of advection tested (eg fig 1f does look consistent with the changing nature of advection from the warming Arctic...). 

https://doi.org/10.1029/2012GL052261

https://doi.org/10.1175/MWR-D-17-0165.1

4. Around Figure 1 there is speculation that the RES_INT anomalies reflect the role of regional SST features. This discussion could be informed by the use of uncertainty analysis as in point 2 above. In particular, are these anomalies outside the range of uncertainty in the dynamic term?

5. Has the author tested for trends affecting the estimated dynamical contribution? It seems possible that for a given case the selected analogues may not span the period evenly. Eg, if the analogues happened to sample the most recent decades only, they would then not give a representative sample of the temperatures over the whole period. Hopefully this is unlikely, but could perhaps happen when there is a lack of SLP variance in the early, data-poor, period of the reanalysis?

6. Some of the results in section 4 are interesting and some further discussion might be helpful. eg why is there a 'thermodynamic' cooling trend in fig 5e, and what mechanisms underlie the strong dynamical trends shown here. Could these be consistent with internal variability associated with AMV, as described for example by Sutton and Dong (2012)?

7. On the Russian heatwave: the text suggests this has been 'mainly linked' to La Nina. I agree this is very likely a factor (see also Drouard and Woollings 2018, GRL), but even so this statement feels a bit strong.