General comments

The proposed manuscript deals with the estimation of return levels (or high quantiles) in regions affected by tropical cyclones.

The paper focuses on using a synthetic database of cyclones to validate the proposed method, developed in other papers, namely the STM-E approach.

The paper is well organized with clear motivations, sensible application of the method, and an important issue in tropical areas.

Specific issues

Near L.200, the sentence " There is no obviously no spatial dependence between the size of STM and its location" is quite unclear to me. The figure clearly shows that the locations of STM are unevenly distributed in space, in particular, the points are located on the boundary of the domain, although there is no discussion about this repartition.

Figure 6, 7,8,9: Quite difficult to read... What do the black boxplots correspond to? I think if the whole database is considered, the 500-year return level should be only one value per location, with associated CI?

There is no clear dependence of the depth here, nor is it included in the model, so maybe the authors should comment on why they include this covariate in the figures, or if it is a perspective for future work.

In the paragraph starting L290, the authors claim that they study the relationship between STM and Exposure in Figure 11, but the figure only shows two distributions for maximal and minimal STM, maybe a more quantitative assessment with parametric models would help understanding the (absence of) relation.

The authors look at a very high quantile (500-year return level), while it is of more common practice to estimate the 100-year return level: the conclusions may be rather less clear due to the uncertainties linked to estimation of such high quantiles.

There is constant under-estimation of the return levels as seen in the figures and table 1. Is there a way to decide if it does come from the STM part of the model or the Expose part? Maybe the authors could provide keys to understanding where do the limitations come from. 

Technical corrections

Figure 10: the 0 on the x-axis should be a 1,200? Again, how do you obtain de box-plot here?

Table 1 may be more clear if the number of observations is replaced by the corresponding years.

Overall review

The paper presents a novel approach to predicting extreme wave conditions, the methodology appears to be rigorous and the study of high quality. The subject will be of interest to readers and is recommended for publication following consideration of the comments below.

General comments:

The main test is to compare 500 year return values predicted from 200 years of data. Was there a particular reason why the 200 years of data was selected?  I would have thought that a smaller length of data, say 50 years typically the length of record available for cyclone, would provide a sterner test and may reveal larger differences between STM-E and single location analysis. That said, the paper is valid and worthwhile as it stands. 

Specific comments:

• P2 line 32  'Return level'. Most of the paper uses the term 'Return value' and, for clarity I think it would be better to be consistent throughout
• P6 Figure 3   From the size of SWH it would appear that the SWAN model is using Wu (1982) wind drag formula which is linear with wind speed. Later versions of SWAN use Zijlema et al (2012) where the wind drag coefficient rises to a maximum and falls at high wind speeds. It would be worth specifying the version of SWAN used.
• P7 line 142   I don't understand 'longitude-latitude transects with arbitrary orientation'. Please clarify.
• P9 lines 202-203  Please consider reprasing this sentence to avoid a double negative. It is a bit confusing.
• P11 line 229-231  If I understand correctly, you have modelled 1971 cyclones so would it not be better to count back rather than randomly sample? It would be a more robust way of estimating the 500 year return value.
• P14-15 Figures 6 to 9   There is a lot of information included in the graphs and they are hard to interpret. It is difficult to compare black with blue and red. They need to be bigger which may require splitting into smaller subsets.
• P14 line 271   Should this say decreasing rather than increasing? Either way perhaps this sentence needs some clarification
• P15 figure 11  Red and blue show a consistent difference in shape for all sample sizes. Is this important and is it common to other locations?
• P18 line 313  Is this the case for both MLE and PWM? It doesn't look that way.

Typos:

• P6 line 116  remove 'on'
• P7 line 145  Insert 'be' between 'should' and 'sufficiently'
• P9 line 199  An extra 'no' has been included. Please use the most appropriate of 'no obvious' or 'obviously no'
• P18 line 331 'that' should be 'than'