Zinke et al. described the combination of aerosol eddy covariance fluxes and aerosol number size distribution from a plunging jet to estimate the emission fluxes of sea spray between 0.015 and 10 um. The method has been described by Nilsson et al., (2021). The results are interesting and useful for improving sea spray source functions, but the authors have included too many details and unnecessary comparisons, making it difficult to follow the main narrative. The paper is written by well-known experts and represents their continuing work on sea spray aerosol. I recommend that the paper be published after addressing some issues:

• The claim of bridging the gap between in-situ and laboratory estimates is a bit of an oversell. To me, bridging the gap means identifying a sea spray profile in ambient marine environments that agrees well with chamber-produced sea spray. However, the representativeness of the chamber in mimicking the true ocean remains unclear. The study simply rescales the number size distribution by DMPS to EC measurement.
• The paper is lengthy, and I suggest that the authors restructure it. In many cases, I had to go back and forth between the Supplementary Figures and the main text.
• The method used in the study was first described by Nilsson 2021. In the original paper, merging DMPS and EC in the range of 0.3 to 0.5 um suffers from large uncertainty. How was this uncertainty dealt with in this study?
• The temporal resolution of DMPS and EC is quite different; how did the authors merge two datasets with such different resolutions?
• Line 37: The main reason for the discrepancy is the method used to obtain sea spray source functions. For example, Liu et al., use particles > 0.5 um as a proxy for sea salt, which is not representative of sea spray source functions.
• Line 48: Even different labs produce different sea spray source functions.
• Line 349: What are the ending-heights of HYSPLIT?
• Line 349: The language could be more concise; everyone knows that Figure S7 is in the supplementary materials.
• Line 352: Is it a real seasonal trend or just two different days?
• Line 354: I do not understand why the authors spend so many words explaining the difference in chlorophyll-a concentrations that come from different seasons.
• Line 358: Is it 186030 minutes or 186 hours and 30 minutes?
• Line 407: This is probably the most prominent reason; sea spray with different size ranges is barely comparable.
• Figure 4: Why are the first and last bins so similar for different U10 or ReHw?
• Line 445: The authors compared their size-resolved emission fluxes with Ovadnevaite 2014. Although the trend is similar, the shape is quite different. I don’t think it ‘agrees very well’.
• Line 456: Since jet flow rates are different, comparing number concentrations is meaningless.
• Line 458: The impact of water temperature is expected to be minor compared to the jet flow rate.
• Line 470: The first few bins of WELAS are unlikely to represent real sea spray and should be excluded.
• Line 594: If I understand correctly, EC measurements were also impacted by sea spray production history, as evidenced by the impact of short, long fetches, and shallow waters. But for sea spray simulation chambers, seawater is purely local; this might introduce additional uncertainty when merging two datasets.

Knowing the rate of sea spray production under different conditions is important for understanding the marine aerosol budget and marine clouds. This paper presents an ambitious set of observations: 1) direct sea spray flux measurements by the eddy covariance (EC) method for diameters of 0.25 to 2.5 micron from a coastal tower; 2) chamber flux measurements using seawater collected nearby from ships for diameters of 0.015 to 10 micron.

The results and discussions are mainly structured over two main aspects:

1. Assessment of previously proposed relationships between sea spray production and its drivers, namely wind speed, wave height, and wave Reynolds number. The authors found that qualitatively, those previous parameterizations are generally consistent with this dataset
2. Combining the EC flux and chamber flux to generate a sea spray production parametrization that covers the entire range of the chamber measurements (0.015 to 10 micron). Here the authors found that applying the air entrainment approach to the chamber data leads to best agreement with the EC fluxes (though for some sizes the difference is still large, e.g. half a decade), and so chose this as the basis for their new parametrization.

Overall, I think this is a very rich dataset, and I applaud the authors’ efforts in trying to bring together EC and chamber measurements. I can see their argument that this may be a useful approach to determine the specific fluxes of spray components that cannot be measured fast enough to apply EC. However, I think the paper can be improved significantly. My main recommendations/comments are:

1. (abstract/intro/conclusions) What are the key uncertainties/unknowns in sea spray production prior to this work, and what are the key improvements in our understanding after undertaking this ambitious work to couple EC & chamber fluxes? I feel that currently such key messages are largely absent.
2. The authors can be more constructively critical, both when comparing against previous observations/parametrizations, and also when comparing their EC fluxes with scaled up chamber fluxes. For example, that the scaled fluxes and EC fluxes have rather different size distributions (e.g. Figure 6) are mostly undiscussed.
3. If the whitecap approach (assuming 100%) is found not to be appropriate for scaling up chamber fluxes, what are the implications? Should the recommendation be to always measure the bubble fraction in the chamber? Or one shouldn’t use the whitecap method at all?
4. The measurement uncertainties (bias and random) as presented may be underestimated. In terms of bias, given the very low flow rate of the OPC, I’d be very surprised if the high frequency flux loss is as small as the authors claim to be. I urge the authors to address this by a) closely examine the N’W’ cospectrum and look at high frequency flux loss, and b) verify the response time of the OPC. In terms of random error, fluctuations in aerosol number not due to sea spray production (e.g. anthropogenic influence/transport) will significantly increase the noise in the measurement. The authors can estimate random error by shifting N’ and W’ by a large lag (e.g. a few minutes) and then compute the covariance – the stdev in that null covariance then approximates the random error.

Further specific comments

Abstract.  Specify the size range of EC flux measurements

Section 2.1 the aerosol flux tower was on the island, not on the ships, right?  The first paragraph in section 2 is a bit confusing, and it’s not clear what the ships contribute to the measurements at this point.  It’s only later that I realized the chamber measurements were done on the ships.

Line 114. Specify whether ¼’’ is inner or outer diameter

Line 137. A bit more info here would be useful for each, i.e. despiking (what sigma?), rotation (double rotation? Planar fit?), detrending (linear?), lags (how long?)

Line 164-167.  I’m not sure if that’s true.  Coastal areas are often subjective to strong horizontal gradients in aerosol concentrations, which could contribute to the low frequency part of the signal.  Have authors looked at parameters such as dN/dt and their impact on the flux?  I suspect that for this sort of coastal regions, a lot of the variability in N isn’t due to vertical transport, and could contribute to both random and systematic errors in flux.

Line 178-185.  Why are there two estimates, 20.7% and 13.5%?  what’s the difference? Is the first the ‘total’ attenuation?  Given the laminar flow and very low Reynolds number, I would’ve expected even greater flux loss.

There are two main sources of attenuation that needs to be considered: 1) OPC only outputs 1 Hz data, and thus all flux above the Nyquist (0.5 Hz) is not captured; 2) attenuation related to the response of the instrument. It’s not clear from A2 that both effects have been corrected.   Both of these terms are also dependent on the dominant scales of eddies (and so dependent on wind speed).

Have the authors looked at the cospectrum of aerosol flux (N’w’) and compare it vs cospectrum of heat to see how much high frequency aerosol flux they might be missing?  E.g. plot the bottom panels of S2 on a semi-log scale and see how quickly the cospectrum goes towards 0 at high frequency.

Also ‘constant with size’ is repeated 3 times.

Line 212.  This is a very hand-wavy guess of the random uncertainty.  It could be of roughly the right magnitude, but the text should reflect that the ’10-20%’ number is not based on actual analysis of the aerosol data.  

Section 2.6  I think the spectral analysis can be done much better. Please see comments towards the end of the document.

Figure 3. does the temperature dependence in the wave Reynolds number help to explain more variability?  In a recent paper about gas exchange (doi: 10.3389/fmars.2022.826421), we found that computing the wave Reynolds number at a constant number (e.g. 20 deg C) actually leads to more variability explained than computing it at the in situ temperature.

Line 658. Here the authors say the response time is 0.6 s, but on line 675, 0.3 s is stated.  The actual response time is of course dependent on the flow rate and setup. Have the authors determined the response time of their own system for these campaigns?  It’s probably not easy to determine a sub-second response time when the instrument only outputs 1 Hz data.

Line 705. H2O flux is also substantially dampened when sampling down a tube.  See e.g. Fig 9 in www.atmos-meas-tech.net/9/5509/2016/    Thus the H2O related Webb correction may be even smaller than what the authors have presented.

Figure S2. Within the inertial subrange, the slope should be -5/3, not -2/3. Also, inertial subrange is typically considered to be >1 Hz.  At a height of 10 m and wind speed of 5 m, this translates to a fz/U of 1*10/5 = 2.  Currently the red lines are drawn at lower frequencies than the initial subrange.

Bottom panel, it’s not clear what the two different colours indicate.

Line 565. The size distributions from the EC flux and the chamber flux are clearly different, with the EC flux falling more quickly with increasing size above ~0.4 um. The authors haven’t commented much on this discrepancy. Which measurement is more realistic?

Figure 6. is ‘source flux’ the mean flux derived from EC from both campaigns?  If so, indicate the mean wind speed and temperature here, and mention that the mean wind speed from the two cruise campaigns (6.4 and 6.6 m/s) are fairly similar to the Martensson et al 2003 and Salter et al 2015 parametrizations.

What about temperature & salinity dependence in chamber? Some of these might have been presented previously; if so perhaps a sentence would be sufficient.

Line 589. It’s unclear how the Reynolds number based parameterization was done for different wind speeds. What was the wave height used?

Line 591. I don’t understand how the parametrizations (wind speed or RHw dependent) are so much higher than the EC measurements even over the same size. Weren’t the parametrizations developed based on these same observations?

Figure S12: check caption ??

Appendix B doesn’t bring much to the main arguement. Could be move to supplement.  Figures S10 and S11 may be suitable for the main paper

The paper by Zinke et al. reports complex experiments in reconciling sea spray production and corresponding fluxes to shed new or more light into the sea spray source function development. It is a very valuable contribution and deserves publication in EGUSphere, but more work is needed to address some important comments or omissions.

The paper looks like an attempt to summarise very large experimental effort, like PhD thesis, however, I am not suggesting splitting it into several papers, because the value would be significantly diminished, and some parts of the paper do not constitute standalone paper.

The overarching comment would be the lack of judgement throughout the paper. The authors are too complacent about similarity, comparability, or consistency between different studies. If everything is consistent and repeatable, why publish in reputable journals? The authors are encouraged to make judgements and assertions which is the only way of making progress and encouraging debate.

Major comments

Line 76. Better justification of performing SSA source function experiments in the Baltic Sea should be provided given the fact that Baltic Sea has much lower salinity in comparison to the global ocean.

Line 97. Given the complex bottom topography of the Baltic Sea it is important to assess wave state vs wind speed to make the measurements scalable/consistent to the global ocean. Lower salinity of the Baltic sea can be accounted for when using Reynolds number as long as the wave breaking pattern is the same – or is it? Or any other metrics demonstrating consistent wave breaking.

Line 106. Was the 12m height of the EC measurements sufficient to avoid coastal surf-zone? Where was flux footprint relative to the mast distance from the surf-zone? I trust all was good, but should be supported by numbers.

Line 117. 1.2lpm flow rate through the 5m 1/4" sampling line cross-section produces lag time of ~4 seconds (if I am correct). How was this lag accounted for in the EC measurements?

Line 138. I wonder if arithmetic average is appropriate here, given the fact that aerosol distributions (and atmospheric parameters in general) are log-normally distributed in nature (due to fundamentals) . Median and range would be the correct parameters. I keep noting this issue for many years now, but that does not get widely accepted by excusing for historical legacy. Can it be partially for this reason that 30min average EC fluxes are so noisy?

Line 304. Was the assumption that the entire area was covered in bubbles tested at least visually? It is very difficult to have consistent 100% coverage without foam build-up. Foaminess would also depend on the sea water organic matter. All in all, assumption of 100%  coverage is very uncertain and, even more so, incorrect in the first place. More discussion is needed on this topic and then why not to assume 30-50% with the plotted uncertainty range which would be much closer to the truth. Later in conclusions the authors note that the coverage is more like under 20%, but that is too late and too striking.

Figure 3. Was the Reynolds number taking into account different salinity and temperature, resulting into different viscosity of water of the Baltic Sea vs global ocean? Same question for Figure 4.

Line 406. Was it really comparable? I beg to differ. Sometimes it feels that the authors are too focused on consistency, comparability and so on. Differences should not be afraid or dismissed, they may inform about some important overlooked aspects or mechanisms.

Line 410. Baltic sea salinity is 5 times lower than the open ocean and, therefore, has a profound effect on the number flux. I do not see that the authors took full account of salinity effect other than just mentioning it. If they did, please note that where appropriate.

Line 422. In power law relationships correlation coefficient is a poor metrics as it is almost always high due to high-end values. Maybe 95% confidence interval of the power coefficient would be more informative?

Line 486. Flux and particle concentration are two different concepts or measures. Flux was demonstrated to depend on the wind speed as it should be. The ambient concentration would only be dependent on the wind speed by the power law in fully developed/equilibrated boundary layer. It takes 1-2 days to fill the boundary layer with corresponding fluxes and the Baltic Sea is too small for the established air mass or synoptic scale processes. Overall, it just manifests that the wind dependent fluxes resulted in little changes in ambient concentration emphasizing that environmental impacts of small water bodies will be limited (going back to justification of the Baltic Sea experiments).

Section 3.5 What do we learn from the scaling and what is the purpose of sections 3.5&3.6 other than every chamber experiment will produce different scaling and the result? Maybe I missed something, but that should be clearly explained if I did.

Figure 7. Fig 7b can be easily predicted from 7a as the mass is dependent on the particle distribution >1um. Why mass is important as it varies massively moving towards super-micron range?

Table 4. Geometric sigma, especially of Mode 1 is way too large to be described by one mode. The fundamental rationale is that each log-normal mode is defined by individual production mechanism and typically does not exceed 1.4-1.5 (refer to Hinds, Aerosol Technology Textbook).

Line 601. Log-linear relationship is effectively power law, why confusing readers with different mathematical or plotting outcomes when there should be enough of distinguishing linear versus non-linear effects.

Minor comments

Line 376. Previous section reported 475+150=625 half-hourly fluxes.

Line 401. What was the power law coefficient?