In their study „Intraseasonal modulation of Sea Surface Temperatures in the North Tropical Atlantic by African Easterly Waves“, the authors investigate how meridional wind anomalies at time scales of 2-10 days in the Mauritania upwelling region impact the surface ocean and mixed layer heat budget in the tropical North Atlantic. They find that these wind fluctuations that are associated with African Easterly Waves (AEWs) modulate sea surface temperatures through changes in atmospheric fluxes (in particular, latent heat flux and shortwave radiation) as well as through changes in vertical mixing. The study is primarily based on model output, but a substantial part of the manuscript focuses on the validation of the model output with in-situ and satellite observations as well as reanalysis data. Overall, the study shows some interesting results and, in my opinion, could be a useful contribution toward a better understanding of air-sea interactions in the Mauritania upwelling region. However, I find several major concerns which should be considered before publication.

Major concerns:

1. The validation of the model output makes up a substantial part of the results and half of the shown figures (Figs. 1-5). While I agree that it is important to examine how well the model reproduces SST, horizontal surface winds, and vertical profiles of horizontal winds, this takes up space that could be used for more in-depth analysis (see below for some suggestions).
2. It would be useful to include a figure showing a map with the inertial period at each grid point and the period of peak wind variability to assess how close the observed wind variability is to inertial period and in which regions this relation is most prominent. This could strengthen the authors claim of near-inertial currents having an influence on mixed layer dynamics which is an interesting question that could be examined further this way.
3. The description of the AEW index in section 4 is not quite clear to me. It would be more useful to include a formula to clearly describe which variable and which region is used.
4. In my opinion, Fig. 7 is not very convincing. Why did the authors choose only one year to show the relationship between SST and meridional winds? Why are 2–10-day meridional wind fluctuations so similar to the original meridional wind time series while SST shows substantial differences when bandpass-filtered? More discussion would be helpful here.
5. In Figs. 8 and 9, a very locally defined AEW index (basically at the mooring location) is used to examine large-scale changes in winds and SST over the entire tropical North Atlantic. How useful is this regression in case of such a rapidly changing index?
6. It seems trivial that the 3–5-day fluctuations have a stronger impact on ocean surface variables because of the vertical pattern of the wind fluctuations (i.e., closer to the surface; Fig. 5) compared to the 6-9-day fluctuations?
7. The authors should provide a more thorough discussion of the mixed layer heat budget based on the model output and available studies using in-situ measurements (Foltz et al., 2003; Hummels et al., 2014). This could strengthen the claim that vertical mixing plays a role.
8. What is the temporal variability of the relation between AEWs and SST over the examined time period? How large are year-to-year changes that make it hard to quantify a more distinct relation between AEWs and SST?

Minor points:

• The region in the tropical Atlantic north of the equator has been named tropical North Atlantic (TNA) in most publications. I think this should be changed in the manuscript which often uses “North Tropical Atlantic”. In fact, even the authors sometimes refer to this region as tropical North Atlantic (e.g., line 125). It should be uniform throughout the manuscript.
• Lines 17-19: In which region are AEWs close to the inertial periods? This could be shown. See major concern 2
• Lines 23-26: I don’t think key points are required for Ocean Science? Ignore this comment if this has changed.
• Lines 111-113: I wonder how useful it is to validate the model output with a reanalysis product (ERA5) that is used to initialize the atmospheric model? Wouldn’t a comparison with independent data be more useful?
• Line 234: Which region is meant here? This should be defined clearly. The way it is written here is too vague.
• Figure 5e: How to distinguish between near-surface wind variability in the 3–5-day band from AEWs and the African westerly jet? Or is there interaction between these?

Specific comments:

Abstract:

• Line 14: Please define PIRATA or keep it more general in the abstract. For instance, by saying “moored surface buoys”.

1 Introduction:

• Lines 42-44: African Easterly Waves propagate from east to west.
• Line 51: I assume the authors mean “zonal wavelengths”?

2 Data and methodological approach:

• Line 88: Do the authors mean “air-sea” instead of “air-heat”?
• Line 124: Here the surface air temperature at 1m above sea level is meant, correct?
• Line 125: Please also mention the more up to date reference for the PIRATA buoy network: Bourlès et al. (2019)

3 Evaluation of the coupled model:

• Line 154: Typically, this upwelling region is referred to as “Mauritania upwelling”.
• Lines 157-159: It seems that the model also underestimates the magnitude of the Atlantic Cold Tongue. Here, a more quantitative comparison could be useful to validate the model output.
• Figure 1: Why using the time period 2007-2021 and not the full time series since 2001? I don’t see an explanation for using the shorter time period. In Figure 2 the full time series is used.
• Lines 172-174: But satellite SST data are provided as daily averages (i.e. some of the high-frequency variability is averaged out), whereas ERA5 and PIRATA data are available at higher frequencies (3-hourly and hourly). It would be interesting to look at an exemplary season and compare the time series of ERA5, PIRATA, and model output. Because even the PIRATA buoy north of the Cape Verde islands which is closest to the high SST STD off Africa shows reduced variability compared to regions outside of the high variability area.
• Figure 2: It is interesting that ERA5 and OISST produce the same climatology but very different standard deviation (as a function of calendar month). What could be the reason(s) for this? Larger swings in OISST around the same mean values in both products?
• Lines 193-194: I don’t follow this. Why does higher SST STD in OI-SST imply biases from satellite measurements? It seems that most of the model output validation simply depends on whether the comparisons are really between comparable variables (skin temperature vs. SST)?
• Lines 208-209: The winds north of the equator (10°N-15°N) do not cross the equator. The southeasterly trade winds cross the equator and are deflected to the right north of the equator. Please clarify this sentence.
• Figure 4: It should be noted (and discussed why) that the model exhibits the highest deviations from all other products during July to September (Fig. 4a) which is the time of the year when AEWs are investigated.

4 Ocean surface response to AEWs:

• Figure 7b: I believe “2-10jrs” is the French version of “2-10 days”. Please replace.
• Lines 289-290: It says 2015 in the text, but 2001 on the x axis in Figure 7. Please clarify which period is shown here.

5 The ocean mixed layer heat balance:

• Lines 373-375: Shouldn’t a deepening of the mixed layer depth imply warming and not cooling of the mixed layer?
• Lines 386-387: Is this really significant from a statistical point of view? Otherwise, the authors should be careful with using the phrase “significant”.

References:

• Bourlès, B., Araujo, M., McPhaden, M. J., Brandt, P., Foltz, G. R., Lumpkin, R., et al. (2019). Pirata: A sustained observing system for tropical Atlantic climate research and forecasting. Earth and Space Science, 6(4), 577–616. https://doi.org/10.1029/2018EA000428
• Foltz, G. R., Grodsky, S. A., Carton, J. A., & McPhaden, M. J. (2003). Seasonal mixed layer heat budget of the tropical Atlantic Ocean. Journal of Geophysical Research, 108(C5), 3146. https://doi.org/10.1029/2002JC001584
• Hummels, R., Dengler, M., Brandt, P., & Schlundt, M. (2014). Diapycnal heat flux and mixed layer heat budget within the Atlantic Cold Climate Dynamics, 43, 3179–3199. https://doi.org/10.1007/s0038201423396

Review for manuscript #egusphere-2025-4429 “Intraseasonal modulation of Sea Surface Temperatures in the North Tropical Atlantic by African Easterly Waves” by Mendy et al.

Summary:

A coupled climate model is used to investigate the role of the ocean’s response to high-frequency (1/10-1/2 day^-1) atmospheric forcing focusing on the mixed layer heat budget of the eastern tropical North Atlantic. The authors suggest that the atmospheric forcing cause fluctuations in sea surface temperatures exceeding ±0.5°C through modulation of air-sea fluxes and vertical mixing in the ocean.

Formal review:

The paper addresses relevant scientific questions that fall within the scope of Ocean Sciences. Foltz et al (2025) recently highlighted the needs for improved understanding of upper ocean physical processes and their forcing in the tropical oceans. This study could potentially contribute to a better understanding of the role of high-frequency atmospheric variability, such as African Easterly Waves, in setting sea surface temperature in the tropical North Atlantic. However, the current version of the manuscript contains a large number of errors that make the reported results untrustworthy. E.g., all time series used in the study are band pass filtered for the 1/2-1/10 day^-1 although daily time series are used that only resolve periods of 2 days. This filtering reduces variability in the frequency range from 1/4 to 1/2 day^-1, even though this variability is considered to be the factor that most strongly impacts sea surface temperature. In the model evaluation section, observations are compared to model output that are falsely interpreted and conclusions drawn from the comparison are invalid. Similarly, the limited explanation of the methods used suggests that the linear regression analysis has been misinterpreted and that the figures presented are unreliable, while a statistical uncertainty analysis is completely lacking.

Below, I am detailing my remarks with reference to the individual line numbers of the manuscript and provide some suggestions for revision.

Detailed remarks:

Line 70, quasi-inertial waves: What are these? Please explain this term.

Line 88, air-heat fluxes: Please explain this term.

Line 128, use of log wind profile: The coefficients used for the log wind profile and the assumed height of the anemometers of the PIRATA buoys are critical to the model evaluation section. How do you justify the values used and what may be their uncertainty?

Lines 85-109, description of the regional coupled model: I would suggest the authors provide more information on the ocean model used. E.g., which vertical mixing scheme is employed in the ocean model? This is relevant because vertical ocean processes do seem to be important.

Lines 131-132, Butterworth band-pass filter: The authors use daily time series of winds, SST and model output (mixed layer heat budget terms) in this study. This implies that the lowest period that is resolved by the data is 2 days (Nyquist frequency is 1/2 day^-1). It makes no sense to band-pass filter this data to “retain variability in the 2-10-day period (line 132)”. When doing so, variability in the 2-4-day period range will be damped. Instead, the data simply need to be high-pass filtered with a cut-off period of 10 days to retain high-frequency variability. The use of a band-pass filtered data for the analysis in section 3 to 5 and the model evaluation makes the results questionable, because variance in the 2-4-day period band is lost in all time series. Apart from this, additional information about the filtering methodology such as one-sided (does not preserve phase) or two-sided (preserves phase) filtering and filter order needs to be added to the section to ensure reproducibility and interpretability.

Lines 132-143, identification of AEWs: While I think that this section nicely motivates the use of regression to analyze the effect of EAWs on the oceans’ heat balance, a section should be added that details the use of the regression analysis. Detail should include (1) whether explanatory or predictor or time series were normalized, (2) processing of the time series used as input and (3) a detailed evaluation of statistical significance.

Lines 154-155, Southeast of the Equator, the Atlantic cold tongue …: Why is the cold tongue located southeast of the Equator?

Line 166-168, comparison of SST variability: Here, you are comparing daily averages of model and ERA5 SST with a satellite SST product (OISST) that is not a daily average, but measures SST whenever there is a satellite over that specific region. OISST thus retains diurnal variability in their data set. To me, the comparison of SST variability presented here does not make sense. You are comparing daily averaged with SST taken at specific times during a day. To compare model output to OISST, the SST from the time of day of satellite overpasses (probably less than a minute) need to extracted from the model and compared.

Line 168-174, discussion about skin temperature measured by satellites: I find this section highly speculative. The OISST product, as you state in the lines 120-122, is a product from satellite data, ship and buoy data (Huang et al, 2021). It uses satellite skin measurements but adjusts and blends them with in situ data, including PIRATA temperature measurements, so that the final product represents a bulk SST field rather than true skin temperature. This should be clearly stated here. There have been numerous studies comparing OISST with independent data sets. Can any of these previous studies support this discussion, here? As noted above, OISSTs are not daily averages. 

Lines 193-196, biases in satellite measurements: I find this hard to follow. Before, in line 168-174, it is argued that skin temperature is causing elevated variability. Here, it is argued that biases in the interpolated data are causing the elevated variability. How is any of this justified? OISST uses temperature measurements from PIRATA buoy data in their data sets. Why is OISST so much different from the data it uses? Again, PIRATA temperatures from the time of day of satellite overpasses need to extracted and compared to OISST data.

Lines 199-232 and Figures 3 and 4, model evaluation of 10-meter wind: Again, I have difficulties in believing any of the analysis presented here. In Figure 3 and 4, it is shown that PIRATA winds are very different from ERA5 winds, on average and in the magnitude of 2-20-day variability. However, ERA5 uses the data assimilated and processed in the ECMWF's Integrated Forecast System (IFS) (Hersbach et al., 2020). The IFS in turn draws data from the GTS, including 6 hourly PIRATA buoy wind data (e.g. see ECMWF global data monitoring reports). As stated in Johns et al. (2021), PIRATA wind have much larger weight for the ECMWF forecast compared to other data such as satellite retrievals. So, I do not understand why PIRATA wind data should be different from ERA5 wind data at the mooring position as suggest in the analysis presented in this section. Could the differences shown here arise from the log wind profile scaling used in this study for the PIRATA wind data that is different from what ECMWF uses? To me, there should be no difference between ERA5 winds and PIRATA winds as they are fully assimilated in the reanalysis and heavily weighted. I would even go as far as saying that the comparison here is unnecessary, as you are comparing the same data. However, why should there be differences in the order of 1 m/s (which is in the order of 15-20% of the wind magnitude) in average winds during July through September between the two data sets as shown in Figure 3? Were the same time intervals used for ERA5 and PIRATA wind averages, i.e. were the periods when no data was available from PIRATA excluded from ERA5 data averaging? I would think that this is most likely the cause the apparent differences in the two data sets. In general, the disagreement between ERA5 winds and PIRATA winds reported in this manuscript challenges the validity of the methods used in the model evaluation. Details of the data treatment for the comparison should be presented here, along with a detailed explanation of why the two data sets should be different. Last not least, ASCAT winds are again satellite measurements done at a certain time during the day. Short-term (e.g. diurnal) variability of these measurements should be much larger than daily averages from ERA5 and the model. I wonder, why ASCAT winds compare relatively well with daily model averages?

Line 251, observations: Which observations are you referring to?

Line 260, Figure 5: The figure contains non-english text which should be removed. Furthermore, I can not make out mean zonal winds in the figures. I would suggest to add a separate subplot showing average winds. What does Uz and Vz stand for? Is vertical shear of horizontal velocity shown here? In the caption it says that velocity is shown.

Line 274, North tropical Atlantic: It is either tropical North Atlantic or northern tropical Atlantic.

Line 281, mean meridional wind: What is meant by mean? What is averaged?

Lines 283-285, location of index: While I agree that location of the reference index is somewhat arbitrary in the sense that the pattern will look similar, I would expect statistically significant regression patterns to appear at different locations in Figure 8, 9 and 10 if the location of the index was altered, e.g. placed further offshore. If this is the case, the sentence written here is misleading and should be altered to facilitate understanding of the methodology.

Line 293-295, … a notable degree of correspondence …: I find this formulation rather unprecise. Are these two time series significantly correlated or not? By eye, I would think they are not significantly correlated.

Line 297-299, regression analysis: As stated above, the regression analysis should be detailed in section 2. 

Line 307, Figure 8: It is unclear to me how to quantitatively interpret the results of the linear regression analysis presented in the figure. This is due to the fact that details of the calculation are lacking. E.g. where the explanatory time series normalized? How was that done? What is actually shown in the plots, the linear regression slopes or the full linear regression of the filtered time series? How are 95% significant regressions indicated? Why are there units mentioned in the top panels. The units of a regression analysis should be different from those mentioned in the upper panels.

Line 319, +/- 0.5°C: How was this number determined? 

Line 328, “… where evaluated in the model”: What does that mean?

Line 330: All variable used in the equations must be introduced in the text.

Line 344, OLR: Why is outgoing longwave radiation (OLR) used? For the heat budget, net longwave radiation, i.e. the difference between outgoing and incoming longwave radiation would be much more meaningful.

Lines 356-357, cooling rate and OLR regression slopes: How do you interpret the numbers you have exemplarily selected here? When the mixed layer cools, there is less heat loss due to outgoing longwave radiation? What does this tell us? Furthermore, the units of the regression slopes should be different. E.g. cooling rate should be -0.2°C/day/(m/s). However, these numbers are extremely high and would require very large fluxes to sustain (~-250 (W/m²) per 1m/s of wind change). So, I again wonder how the EAW index was treated.

Line 360, Figure 9: Again, I do not agree with the units shown for the regression or the regression slope (whatever is shown here). It should be (°C/day)/(m/s) unless the AEW index was normalized somehow. Using the correct units would also make it easier to interpret the results of the regression analysis. How are statistically significant correlation slopes indicated? Also, I find it hard to identify clear patterns in the plot because they are so busy and small. The figure should be revised to fix this.

Lines 367-380: Again, the units of the parameters (lines 375-376) discussed in this section are wrong and the magnitude of the numbers presented seem to be too high. Results of the regression analysis should also be quantitatively compared to observations such as in Hummels et al (2020) and Foltz et al (2020). 

Line 369, Vertical mixing primarily controls …:  I would appreciate if this statement would be supported by a thorough discussion and a plot showing that.

Literature:

Foltz GR, Eddebbar YA, Sprintall J, Capotondi A, Cravatte S, Brandt P, Sutton AJ, Morris T, Hermes J, McMahon CR, McPhaden MJ, Looney LB, Tuchen FP, Roxy MK, Wang F, Chai F, Rodrigues RR, Rodriguez-Fonseca B, Subramanian AC, Dengler M, Stienbarger C, Bailey K and Yu W (2025) Toward an integrated pantropical ocean observing system. Front. Mar. Sci., 12. http://doi.org/10.3389/fmars.2025.1539183

Hersbach H, Bell B, Berrisford P, et al. (2020) The ERA5 global reanalysis.Q J R Meteorol Soc., 146, 1999–2049. https://doi.org/10.1002/qj.3803

Huang, B., C. Liu, V. Banzon, E. Freeman, G. Graham, B. Hankins, T. Smith, and H. Zhang (2021) Improvements of the Daily Optimum Interpolation Sea Surface Temperature (DOISST) Version 2.1. J. Climate, 34, 2923–2939, https://doi.org/10.1175/JCLI-D-20-0166.1.

Johns, William, S. Speich, M. Araujo and lead authors, 2021: Tropical Atlantic Observing

System (TAOS) Review Report. CLIVAR-01/2021, 218 pp. https://doi.org/10.36071/clivar.rp.1.2021