The manuscript reports on comprehensive observations of bubbles in the upper 4m of the ocean during high wind conditions, with a focus on void fractions at 2m and 4m depth. Bubbles play a crucial part in air-sea gas exchange and are relevant to the transmission of optical and acoustic signals.  The observations are consistent with a nearly permanent bubble layer in the top 1-2m and occasional deeper bubble plumes associated with Langmuir convergence zones. The conclusions drawn from this unique data set are solid but are asking for further in-depth analysis. The authors rightly point out that the bubble presence is not the most relevant parameter for air-sea gas exchange parameterizations, but bubble flux. The additional information on size distribution could help in addressing the related issue of bubble plume age. There is a companion paper which might just do this. The manuscript is well written and the topic is suitable for publication in Ocean Science.

Scientific issues:

Lin 334, Figure 4:  Splitting the data into 3 cases seems very arbitrary. Give more detail on how the data were assigned to a particular case. For example, why are the Case 2 data with 2m void fraction < 7 x 10 ^-7 not part of Case 3. If there is no objective way to stratify the data, then present them as one case and a fit through all data – or best to remove this figure.

Ln352: What is the area of the camera FOV?  Please quantify “very  few “. How does the observed number of breaking waves compare to the breaking rate reported in the literature? (E.g. there are several studies that quantify the active breaking rate from breaking crest length distributions obtained from video imagery). If the observed breaking rate is much lower than expected, it might indicate that the freely drifting buoy somehow avoids breaking waves., similar to wave buoys often missing the peak of a steep crest, or by the buoy being located preferentially in the convergence region of Langmuir circulations. If so, the bubble statistics would be biased.

Ln405: Can you exclude the possibility of measurement saturation at high void fractions? Turbulence models assume TKE injection in breaking waves to a depth comparable to the significant wave height. It is surprising that bubbles would not get injected to 2m depth when Hs is 3 to 5m. Does this imply that void fractions in breaking waves are <10^-4? Or does it mean that the breaking layer is << Hs?

Section 3.3.2:  Your analysis shows that void fraction does only poorly stratify with wind history. In almost all cases the median void fraction for rising/falling winds is within 1 standard deviation of the opposite situation. Similarly, wave age does not affect void fraction. Both results are not surprising. Void fraction is ultimately determined by wave breaking, which in turn is not directly linked to wave age or wind history. But is well correlated with the saturation of the wave field (see work by M. Banner and colleagues).

Ln 517:  Can the (relative) age of a bubble plume be determined from the bubble size distribution? If this is done in the companion paper it should be mentioned here. If not, the analysis should be included here.

Ln594-619: The explanation of higher void fractions during decreasing wind speeds being the result of higher gas saturation is plausible. Another explanation is similarly plausible: during falling wind speeds the wave field is not equilibrated, resulting in a higher breaking rate (bubble production) associated with wave saturation at a prior time.

Technical issues:

Ln 24: define R_Hw

Ln 159: change to:  “Studies by Salter et al (2014) and Slauenwhite & Johnson (1999) observed….”

Ln294: The standard deviation of the measurement depth is not dimensionless (add units to both instances).

Figure 3: Change y-axis label to “Void fraction” on panel b and c

Figure 7: Specify in caption “….split by wind speed (bottom row) and wind-wave Reynolds number (top row).”

Figure9, caption Ln123:  “There were no bubble camera data…”

Czerski et al. present bubble and void fraction measurements collected under moderate to relatively high wind speed. Bubble and void fraction measurements are difficult to obtain, therefore the data set is valuable to advance our understanding of bubble evolution and quantifying their roles in gas transfer, sea spray, and turbulence. The contents are of interest to the readers of ocean science and the manuscript is generally clear. Below are my suggestions.

Panel (a) of Figure 3 indicates that bubble ~ 12 microns could go down to 6 m at a wind speed of 18 m/s. I recall backscatter data at ocean station Papa (by Svein Vagle) showing that bubbles of around 20 to 30 microns go down to more than 10 meters at a wind speed of ~10 m/s. I believe a figure showing the data is in the textbook "Chemical Oceanography and the Marine Carbon Cycle" by Steve Emerson (I apologize that I do not have the book with me right now and do not have the figure number). Do bubbles of 12 microns dissolve quickly in the water column and do not exist below that depth? Or is there any limitation in the observation that bubbles deeper than 6 m cannot be detected?

Also for Figure3: I would suggest revising the ylabel for panels b and c as “void fraction”. The current ylabels seem to be more appropriate as titles.

Regarding the results in Figure 4: Figure 10 shows that the void fraction at 4 m could lag behind that at 2 m. Is that already considered in the correlation presented in Figure 4?

Currents in Figure 6 are useful to explain void fractions in Figure 5. I would suggest the author also try to connect Figure 6 to explain Figure 3. For example, the downward current at around 18:30 is strong than at around 18:10, but the increase in the void fraction is smaller. Is it because there is no breaking wave observed at 18:30?

Although this is mentioned in the text, I would suggest adding that the measurement is at 3.8 m in the caption of Figure 6.

Figure 9: I would suggest the authors clarify if the rising/falling wind means a sustained period of rising/falling wind, like in Liang et al. 2017 JGR-Oceans referenced in the manuscript. Or do the authors include substantial periods when the wind is fluctuating?

The data in Figure 9(a) show that void fraction is higher at falling wind than at rising wind until about 20 m/s. This could be interesting results and I have different thoughts from the paragraph starting from line 456. Breaking waves are bigger at the rising wind, but Langmuir turbulence is stronger during the falling wind. I wonder if there is a possibility that void fraction at this depth is primarily due to Langmuir circulations at wind speed < 20 m/s and gets more contribution from breaking waves when wind speed > 20 m/s.

The same argument could be used to explain that void fraction is mostly larger during falling wind than during rising wind at 4 m. However, I could not make sense of why the void fraction is the largest when wind speed is at 10 m/s and 12.5 m/s. Is there any sampling error there?

Regarding the presentation of Figure 9. I would suggest stating in the caption that panel a is from data at 2 m and panel c is from data at 4 m.

Figure 9(b) is referenced after Figure 9(c). Perhaps the order of the panels could be changed.

Both “parametrization” and “parameterization” are used. I would suggest the authors pick one of them.