The authors performed a metoocean analysis of extreme hydrodynamic and weather conditions for providing thresholds to define critical design parameters for offshore installations at the Gulf of Finland.  The analysis was extended to the entire Baltic Sea in a second part of the study. Although I have no doubts about the importance of safeguarding offshore infrastructures through metocean analysis, I don't see novel findings in terms of scentific knowledge. I also have some major concerns regarding the dataset and uncertainty in the analysis, which are listed below:

1. One of my main concern is about the scientific novelty in this study. The authors ultilized existing datasets and peformed extreme value analysis to derive the results. However, the manuscript reads like a technical report rather than a scientific paper due to lack of novelty either in the methods or findings. Please clarify this in your revision.

2. It seems that the wave modeling and hydrodynamic modeling is decoupled (lines 85-95), and provided by different data sources. I have a serious concern on this, since nearshore areas are strongly affected by interactions between waves and currents, and water levels. I strongly believe that a full coupling between waves and hydronamics should be considered for nearshore areas. Please provide an assessment on the uncertainty related to this and justify why a decoupling is sufficient.

3. Since LNG infrastructures such as FSRU are built very nearshore, which is subject to changes in seabed morphology by erosion or deposition. This might not be a problem in the study area  that is dominated by hard substrate (nevertheless, the authors should justify in the study why impact of sediment transport and morphological change are not considered in the study area). However, for sandy coasts such as the southern and southeastern Baltic, morphological change in nearshore may have significant impact on local metocean conditions. This would lead to large uncertainty in the results of the entire Baltic Sea for very nearshore areas, which should be clearly pointed out in the manuscript. 

4. The authors applied some correction factors to align better model data and observation data in the local study area, e.g. for wind speeds, significant wave heights and wave periods. However, it is unclear whether the same correction factors should be applied for other parts of the Baltic Sea, and the potential errors caused by ignorance of such local corrections. The authors did not address this uncertainty in the second part of the study for the entire Baltic Sea. 

This is my first review of ‘Metocean study for the planning of the FSRU terminal: a case study at the Estonian coast of the Gulf of Finland’ by Raudsepp et al., in which the authors present extreme value statistics for metocean variables, based on model data and observations in the Baltic Sea. The study is motivated by the vulnerability of marine infrastructures – especially the Paldinski FSRU terminal. As the first reviewer pointed out, as it is now, the manuscript finds itself somewhere in between a report and a manuscript. The scope of the ‘state of the planet’ is such that a more report like submission may do, but below I make some suggestions for improvements – mainly focusing on reducing the uncertainties in the Baltic wide analysis (see comment 9) which I think should be the main focus of the manuscript (see comment 1). I consider this to be a major revision that requires some dedicated work, but at the same time I think the revisions can be quite straightforward to implement.

1) I would suggest flipping the manuscript upside down. I think the readers would find it much more interesting to learn about Baltic Sea wide extreme values, rather than this one very specific location off the coast of Estonia. I would therefore change the title to be something like ‘Extreme winds, sea level, waves, and currents in the Baltic Sea’ and start with section 3.5 and use the FSRU site as a validation case study → if possible, I would also consider adding more station observations (e.g. from the Copernicus in-situ TAC) for validation.

2) In relation to the previous comment, framing the study around offshore infrastructure makes sense, but falls a bit short if is not taken to the engineering level (computing actual expected loads to structures, which would allow structural engineers to directly use these results). I think such work is out of scope for the present study, which is why I suggest recasting the study along the lines of comment 1 – also in the abstract/introduction.

3) The authors do not provide much quantitative details on the basis by which the higher resolution regional wind products were dropped out (apart from the general statements about the biased extreme values) – was this decision based on the Paldinski site alone? Although it is a bit of work, I would ask the authors to use CERRA data in addition to the ERA5. CERRA5 has in many occasions been shown to outperform ERA5 and I believe it should allow reducing the uncertainties in the Baltic wide analysis.

4) Overall I think it is a bit misleading to list the different datasets in table 1 and then not use them. However, linked to the comment 9, I would encourage the authors to reconsider using those datasets (and CERRA) in order to narrow down the uncertainties in the extreme values.

5) L148 – which model output was used to fill the tide gauge data-gaps?

6) L164-166. The authors mention the rather large (15% to SWH and 5% to periods!) corrections to wave properties, but there is no figure or reference to back this up. These corrections are quite large so perhaps the authors could add a supplementary figure with a bit more careful description on how these numbers were decided. These percentages seem to be quite round so the reader is left wondering if they are really coming out of a fit, or rather based on expert judgment?

7) L176 – to my understanding the Peaks Over Threshold method requires some way of ensuring that the same event is not considered multiple times. How did the authors do this?

8) section 2.6 – did the authors apply the site specific corrections for winds and waves to the full regional dataset? I am assuming not, but it would be good to explicitly state that.

9) The wind speed return values seem low and the spread is very large – almost to the point that is becomes useless. The error has also a very peculiar round maximums, possible around points where ERA5 has assimilated (or not?) some actual observations. I think the authors should make their utmost to reduce that uncertainty range and provide explanation for the peculiar shapes. For example, it seems very unlikely to me that moving a few kilometers east to west in the region north of Gotland island, one would have close to the same return period (22 m/s, which, btw seems very low) but a difference in the confidence interval that ranges from 1 m/s to well over 10 m/s (colormap is saturated).

I had a look at some other publications, for example this one https://www.sciencedirect.com/science/article/pii/S2212094721000530 who show similar maps of return values. Their colormap is not the easiest, but it looks to me like many models produce much higher values over the Baltic for 30-year return values than what the authors show for 50-year return values. Especially in coastal areas, I wonder how much of this is due to the ERA5 resolution.

I would ask the authors to check if the other datasets produce similar uncertainty range to the return period values and/or if the peculiar shapes could be attributed to the assimilation methods used in ERA5 (would be an important result I think). Use of CERRA dataset might help with robustness. I also encourage the authors to explore other statistical methods to narrow down the confidence interval, although I don’t have direct suggestions here.

Some of the other variables show similarly variable uncertainty fields, some of which might be understandable and linked to strong gradients, but for example the one for maximum SLA is rather difficult to grasp. Why is the relatively low return period in the central Baltic so uncertain (by well over a meter)? Again, I think it would be important that the authors review their statistical methods and aim to narrow the uncertainty range and/or explain where the uncertainty comes from.

10) Data Availability statement. I would ask the authors to share the resulting return periods (i.e. data that can be used to reproduce figures 2-3 and especially 4), their uncertainties, along with codes to reproduce them from original data. This could be done via Zenodo repository. Sharing the data and the codes is a standard today, and allows the interested readers to understand what has been done exactly (especially since the manuscript has no equations), and since the resulting data can be very useful, and is not very large, I see no reason why the authors should not share it.