The paper investigates the impact of windstorms on Finland's electricity grid. The authors present a new classification method for identifying windstorms, which is based on the impacts (here: power outages) rather than on meteorological aspects. They analyze the different windstorm classes in terms of their meteorological properties and impacts, focusing on regional and seasonal differences.

The paper covers an interesting and relevant topic. It is well written and clearly structured. Apart from some comments below, I can recommend the paper for publication and I feel it will provide a useful contribution to the field.

1. When I read the paper, I was quickly confused about how you differentiate between extratropical cyclones and windstorms. Only when I got to Section 3.2 on page 7, I was sure that windstorms are those extratropical cyclones that cause high power outages. Maybe it would be good to define this as early as possible.
2. Section 3.2
   1. How many cyclones did you analyze in total - in addition to the 92 windstorms?
   2. Line 187: You state that you exclude dates if the same windstorm caused power outages during consecutive days. Which dates did you exclude in these cases - the whole storm or just one of the days? And which day did you keep? Please clarify.
3. Section 3.3: What are the borders of the "Outside Finland" domain?
4. Figures 3 and 4: Would it make sense to use the same color coding for windstorms and extratropical cyclones in both figures?
5. Section 4.4.1: How exactly did you compute the correlation coefficient between ERA5 and NDP or between observations and NDP given their different spatial scales?
6. Section 5
   1. Could your classification method be applied to any other country in Europe given that the impact data is available? Or would one need a comparable country size, cyclone statistics, ...?
   2. Could your method be used to estimate the impacts of windstorms on power outages under climate change conditions?

The study of Láng-Ritter et al. investigates extra-tropical cyclones and their impact on the electricity grid in Finland beween 2005 and 2018. The authors analyse 92 selected cyclones, which they define as windstorms, where power outages in Finland can be registred.

The sudy compares a set of cyclones with windstorms (impact in Finland) by means  of selected cyclone criteria. The authors use a classification approach where the location and propagation direction of the cyclone into the target region is used.

The manuscript is well written and structured. The reader can nicely follow.

I have only one bigger comment: in my opinion the authors slightly exaggerate the value of their study. The manuscript gives insights in parts of charachteristics of extra-tropical cyclones and windstorms over Finland which is defenitely interesting and relevant. The authors write in the conclusion that the aim of the study is to improve preparedness for future windstorm risk, provide tools for forecasting windstorm impacts, etc.

There are defenitly further steps to do to achieve these goals.

I suggest to clarify it in the conclusion.

(see also my last minor comment)

Please find below further minor comments which I recommend to consider before publication.

L 5: The impact is purely on power outage, correct? Can you include it to avoid misunderstanding.

L. 45: MSLP is typically used for mean sea level pressure.

L. 98: The 3 s wind gust is computed every time step. Isn’t this the internal model time step instead of the time resolution of the data, which is 1h?

L. 139: is it possible to see a clear relationship with NDP and wind storms? Can it happen that one failure which is related with a wind storm leads to a different numbers of NDP dependent on the location of the failure within the electricity grid. Are there more and less vulnerable locations in the grid? This would include further random effects in the relationship.

L. 186: is your comparison of days with failure and the occurrence of windstorms fully subjective? When do you decide that there is no cyclone around (case 2), line 187) and when is a cyclone near enough to potentially have impact on the outage?

Are there objective (transparent) criteria to decide? Can you convince the reader that this validation works well, e.g. by adding meaningful examples to the supplement?

L. 196: how is the cyclone classified if it passes both boxes F_N and F_S? This would be exactly the blue example in Fig. 2-2.

You explain it in line 209, that the blue one is marked with F_N. Is it the box where the cyclone enters first? Can you add this information to the definition of the box classification?

L. 236: how is it possible that strong winds do not effect Finland but the cyclone leads to an outage?

L. 260: F_N-NW has the shortest lifetime and F_N-SW the longest. How can this be understood since the difference of the classes is only the more northerly or southerly propagation? Probably the cyclones originate in different locations.

Fig. 4: you are not using all classes out of theoretically 8 classes. You should mention that. I am wondering how reliable the box plot is, e.g. for F_S-NW and F_S-SE including 5 and 4 cases, respectively.

L. 315: Fig 6a and b show one case each which has its origin in a very southerly position (around 10°N). This is still the North Atlantic, isn’t it?

Please reformulate.

L. 353: you have not analysed the wind direction but the propagation of the cyclone. It has not to be the same. It is interesting that both SE classes (F_N, F_S) show a huge difference in NDP. Do you have an idea.

Probably your statement is true, that the classes contain 4 and 1 events which is much too small in order to draw clear conclusions.

L. 379: can you exactly define the threshold for medium and strong correlation?

L. 396: I agree the argument that soil condition can influence tree fall. But Fig. 9 agrees with the general relationship between number of storms and number of NDP (largest number of storms and NDP in autumn, second largest of both in winter).

L. 421: doubling of word “types”

L. 462: The goal you are highlighting here, is a nice motivation for upcoming studies. But you are not addressing it with your current study.

Your classification approach allows to distinguish between cyclone characteristics of windstorms (impact cyclones in your definition) in the different regions and propagation directions and cyclones (without impact).

This is a nice way to learn about these cyclone properties.

To learn about risks and to perform impact forecast, you would need an impact model. This can be more or less complex but needs at least a description of a relationship between meteorological parameters (e.g. gust, soil temperature, precipitation, …) and impact (NDP in your case).

You are not analysing when a characteristic leads to impact, when not, and when characteristic leads to no impact (false alarm).

This would go into the direction of understanding risk.

I agree, further research is needed to achieve this aims.

The manuscript presents an interesting an thorough analysis of the impacts of windstorms on Finland's electricity sector. Although in principle arbitrary, their classification shows relevant discrimination power into the impacts, especially if analysed jointly with other factors such as seasonality and environmental conditions. I consider the analysis to be quite thorough and the presentation of the results is very clear, in particular all the graphics have been developed quite carefully.

I think this manuscript provides a relevant contribution to impact studies driven from a good understanding of the meteorology of windstorms, and could promote further studies focused on other regions affected by ETCs and other types of impacts. 

Some minor comments I think the authors should address are:

General:

Would the authors consider that there could be a benefit in redefining the large domain to reduce the influence of the O category cyclones? There is no discussion in sections 4.1 or 5

For figs 5 and 6, I would suggest a reconsideration for the colours, since these are nor colour-blind friendly (most common type of colour blindness can’t distinguish between red and green), but even for a standard sighted person I find it hard to distinguish the different tones in Fig 6, so maybe combine with different symbols?

Minor comments on text

• Line 8 (abstract): it currently read “northern part of a country”, but it should be of THE country, as this is not generic
• Lines 74-76, sentence starting with “Furthermore” though this might become clearer once the reader has covered the whole manuscript, this description is very confusing so early on in the text. I suggest the authors should review it and focus on the key aspects.
• Lines 107-108: On the issue of weather ERA5 overestimate or underestimates winds and gusts, the literature is a lot more nuanced. I suggest the authors take some time to look for example at the references below. There is specific literature that has focused on the performance for ETCs in particular:
• https://rmets.onlinelibrary.wiley.com/doi/10.1002/joc.8339 --> “ERA5 shows a good skill for wind speed with normalized mean bias (NMB) of −0.7% and normalized root-mean-square error (NRMSE) of 14.3%, despite a tendency to overestimate low winds and underestimate high winds”
• https://confluence.ecmwf.int/display/CKB/Windstorm+footprints%3A+Product+User+Guide --> “It was found that wind gust from reanalysis (ERA-Interim and ERA5) underestimates measured wind gust on average”
• And more generally there is a distinction between performance onshore/offshore and for low winds/high winds, and with topography and land use features

      https://link.springer.com/article/10.1007/s00382-020-05302-6

      https://rmets.onlinelibrary.wiley.com/doi/pdf/10.1002/qj.3616

      https://wes.copernicus.org/articles/9/1727/2024/

      https://www.sciencedirect.com/science/article/pii/S2352484723015603

      https://asr.copernicus.org/articles/17/63/2020/ -->  “ERA5 is very skilled, despite its low resolution compared to the regional models, but it underestimates wind speeds, especially in mountainous areas”

https://ges.rgo.ru/jour/article/view/3328/761 -->  “The assessments revealed a

systematic error at most stations; in general, ERA5 tends to overestimate wind speed over forests and underestimate it over grasslands and deserts.”

 I would suggest that the authors review their statement and add a bit more detail.

• Line 396: If the values are not normalised by number of storms, then the conclusion is not as direct as there are also more storms on autumn than winter. This  should at least be mentioned.