The authors present a very relevant and interesting analysis into quantifying the impact of low frequency cycles on fatigue damage. 

While the subject is clearly presented, the results are not convincing.  The following aspects are to be clarified or added before the article can be published.

1) The low cycle region of the SN curve for  welded steel is conventionally taken with the slope m = 3 as given in Eurocode 3, DNV GL C 203 and IIW standards. The knee point is usually 2e06 or higher cycles.  Therefore the impact of LFFD would be limited to m =3 or low slope portion of the bi-linear SN curve only and have a smaller contribution relative to the high slope segment of the SN curve.  Reference Larsen, G C. Thomsen, K that you quote also states the impact with the small slope only has a 1% increase due to LFFD. In this light Figure 3 needs to be better explained as to how much impact to the Miner sum is obtained when including LFFD.

2) Figure 4:  Can you show that the 0 load response of the strain sensors have no frequency components (that is noise)?

3)  If the stress cycle has a period of several hours or days, then the mean wind speed would have signficant changes during that period and the conventional method of fatigue damage accumulation cannot be applied. How is fatigue damage accumulation to be made over different mean wind speeds considering LFFD?  Does it require a non-stationary statistics process to compute this?

4)  Usually for offshore structures, it is the welded joints that have the lowest fatigue life. It appears these are not consdiered at all in the present work and therefore is is unclear if LFFD has any impact on design life. The stress at the welded joint is significantly increased due to local stress gradients in different directions. The impact of the stress gradients can be much higher than the increase in loading due to LFFD.  Can an analysis be shown as to how much reduction in lifetime is present at a welded joint due to LFFD ?

5) In figure 5, is the m = 5 slope also at cycles less than 2e06, that is, the minimum slope m = 5?  If this is the case, the using Haibach rule, the higher slope of the SN curve would be 9 and would result in higher partial safety factors.  Can you quantity what is the impact of the LFFD in the usage of partial safety factors (PSF) in the fatigue life assessment? Does the inclusion of uncertainty due to LFFD result in a significant increase in the PSFs?  This assessment is needed to understand its impact in the design process.

6) Figure 6 is unclear. How is this damage presented to be considered over the lifetime of the structure as the mean wind speed is not a constant over a day or a week and therefore it is unclear how the lifetime of the structure can be evaluated without actually measuring the damage until failure.

Clarifications to the above are required before the article can be accepted.

The authors present a very interesting contribution on the effect of low-frequency response on the fatigue damage accumulation of offshore wind turbine foundations however, I would like the following points to be addressed before the article is considered for publication:

1) In Figure 5a it is not completely clear what data is plotted. The legend provides information only about the black, blue and red points but nothing is mentioned about the yellow curve, which is not even discussed in the text. Moreover, it would make it easier if you could use a different color for line showing the average of converged values for all sensors.

2) In line 246 the authors observe that "the share of cycles lasting more than a day is nearly absent in both FA and SS.", which is a bit contradicting with the results shown in Figure 6. According to latter, this statement is valid only for m=3 and partly for m=4, while a considerable percentage (>10%) of those cycles is present for m=5.



3) When looking at the results of Figure 6, one can draw the following conclusion: The main contribution of the of the low-frequency response to fatigue is due to variations in the mean wind speed, which in turn results in variations of the thrust force. The latter is certainly correlated to a few SCADA variables (power, rpm etc.) and therefore the LFFD as well. The authors have not explored at all these insights, which could further help in estimating the LFFD using SCADA data alone. Please provide these plots of the LFFD with some metric of the wind speed variation (std, Dirichlet energy, etc.)



4) Following up on the previous comment, the contribution of the different sources of variability should be further explored and quantified. Namely, the low-frequency response, whose cycles are smaller than a day, seems to be owed mostly to wind speed variations. On the other hand, the contributions of longer cycles is owed to both wind speed and wind directions changes. These changes can be well quantified using the available SCADA data and related to the LFFD. These are very substantial insights that the authors should explore and provide the corresponding plots.



5) The discussion between lines 286 and 291 seems to be a bit contradictory to the findings presented in section 4.1 and the results shown in Figure 7. The sensors perpendicular to the dominant wind direction, meaning the ones closer to the SS direction, are the ones that seem to have lower LFFD factors, while the ones aligned with the dominant wind direction have higher LFFD factors.



6) The existence of data from different turbines should be used to validate the insights from points 3) and 4) by exploring the data patterns between LFFD and SCADA for all four turbines (T1-T4).



7) A proofread would help improve the language in some parts of the manuscript.