Review of “Sunlight-absorbing aerosol amplifies the seasonal cycle in low cloud fraction over the southeast Atlantic,” by J. Zhang and P. Zuidema submitted for publication in Atmospheric Chemistry and Physics.

This manuscript builds on the authors' work published in 2019 and presents results from the analysis 8 months (July-October 2016, 2017) of observations of aerosols, clouds and environmental conditions to investigate the interactions of frequently occurring smoke particles and low-level cloudiness in the Southeast Atlantic. An important component of the analysis is a breakdown to subseasonal (monthly) timescales which allow consideration of the variability in boundary layer structure due to large-scale meteorological influences. The important results of this work quantify the influences that smoke, transported form the African continent, has on the seasonal evolution of cloud properties. The manuscript represents an important contribution to the field and is well -written. I do believe that there are a few minor issues and edits that should be made for improved clarity before the manuscript is published in Atmospheric Chemistry and Physics.

Minor Comments:

Lines 109-120 – I found this section, comparing the ACAOD, rBC mass and fine mode AOD, a little confusing to keep track. One source of confusion is he use of “These” in line 114. I think “These” refers to ACAOD and τ_af . I think then that the purpose is to use τ_af as a proxy for ACAOD such that one avoids the problems of ACAOD only being available when clouds are present. If that is the case, it would be helpful to also include joint histograms of τ_af and rBC. It might also be helpful to use τaf on the label axis for Fig. 1 (right most panel) and in the legend for Fig. 2.

Line 165 – It is not clear how the stratiform and cumuliform classifications shown in Fig. 3b are done. Are these based on the observer reports? A radar-based classification? Or something else? It would be helpful to clearly identify where these are coming from.

Line 180 – Here, in the discussion of Fig. 4a and later in the discussion of Fig. 5a, the cloud boundaries being higher or lower are mentioned. But this information is somewhat buried in the profiles of cloud frequency of occurrence. To help clarify these discussions, it would be helpful to provide statistics of the cloud base and top heights themselves.

Line 183 – Would suggest replacing “drizzles” with “precipitation.” The disdrometers are not very sensitive to drizzle sized droplets likely need larger precipitation drops to make a recording. If you decide to stick with drizzle, it should be “drizzle” rather than “drizzles.”

Suggested minor edits:

Line 74 – Add “rain gauge” after “tipping bucket.”

Line 80 – Add “the” before “Terra and Aqua.”

Line 113 – Add “compared to other months” after “lower.”

Line 114 – Add “a” before “3^o x 3^o”

Line 122 – What threshold are you using to define “high ACAOD?”

Line 125 – What is meant by “composite decisions?” Is this the determination of high or low smoke days? Would “composite classifications” be clearer?

Line 179 – It would help clarify this sentence to start with a phrase like, “For more smoky conditions…” It was not clear to me at first that this is what was being discussed.

Line 189 – “absorbing” should be “absorption.”

Line 190 – “extending through the night” suggests that shortwave absorption occurs through during the night. I think you mean to say that the warmer sub-cloud layer persists through the night when no shortwave absorption is taking place.  

Line 217-218 - The difference in the sharpness (difference from inversion base to inversion top) of the inversion is a combination of the sharpness of the individual cases, and the variability in inversion height (both base and top) across the composite population. Reporting the statistics of the inversion base, inversion top and thickness (in height, and change in temperature and water vapor) would be useful. Another way to separate the impact of the changing inversion base height is to normalize the height coordinate by the inversion base height.

Line 228 – Add “the” before “boundary layer”.

Line 234/235 – Add “the” before “Ascension region.”

Line 235 – “weaken” should be “weakened.”

Line 237 – “is correlating” should be “correlate.”

Line 248 – Same comment as that for line #217-218.

Line 249 – “an” should be “a.”

Line 274/275 – With the exception of the 12-18 LST cloud frequency profile, the plot does suggest greater cloud occurrence during the less smoky cases. Have you done a statistical test to determine that these profiles are not significantly different?

Line 289 – “Offshore” is used twice. Suggest removing the first instance.

Line 294 – It would be very helpful to the reader to note the latitude of Namibia here.

Line 326 – It would probably make sense to put square brackets around [2018] here.

Line 329 – It is not clear how the profile is “bench-shaped.” Is this a term used elsewhere? If so, please provide a reference.

Line 329 – It looks like the maximum in the longwave cooling profile is > 50K/day at cloud top near a height of 1.7 km? Is the 28K/day just for the smoke layer?

Line 330-331 – It is very difficult to see this difference in height of the maximum of LW versus SW cooling/heating. Adding some grid lines, or some other horizontal line to Fig. 8 would help the reader see this.

Line 336-340 – This last paragraph seems tertiary to the previous discussion and does not really add necessary support. I would suggest removing it.

Line 342 – Remove “the.”

Line 343 – Suggest “ We extend the work of ZZ19…”    rather than “This extends Zhang and Zuidema (2019)…”

Line 351 – Suggest “compared to” rather than “than.”

Line 360 and line 377 – Use acronym (LWP) for liquid water path.

This manuscript seeks to explain the role of absorbing aerosol on cloud structure throughout the biomass burning season in the southeast Atlantic as opposed to focusing on a single month or the biomass burning season as a whole using observations collected during 2016 and 2017 at Ascension Island.  The work is certainly a worthwhile contribution to the literature however there are some concerns that should be rectified first, as listed below. Many of the figures contain a lot of information, so much so, that it can become overwhelming.

Line 120: Is there any concern about subgrid box variability and sharp gradients with such a large grid box? Could this explain why figure 3a shows roughly the same cloud fraction for more and less smoky cases in September, but Figure 5a indicates more clouds (or thicker) with smoky conditions.

Paragraph beginning on Line 121, Page 4: This assessment seems difficult to follow, and subjective. The terms “mostly tracks”, “track each other well”, “tracks fairly well” are used to characterize three distinct periods and three different variables, yet it is hard to distinguish between these. The second half of July 2016 and July 2017 are indeed in excellent agreement, but the character of the agreement between the surface and free troposphere is very different during the other months and this is the point that I believe is trying to be made. This paragraph could benefit from some refinement, with connections made to what was learned from the histograms.

Line 140: A careful assessment should be conducted regarding the selected days for high and low smoke loadings with regard to the composites. A larger number of low smoke days, for example in July, could have a muted composite compared to the smaller number of high smoke days. Furthermore, the large scale circulation in the southeast Atlantic differed between the two 2016 and 2017 (as pointed out by the earlier shift in regime in 2016) so it should be established whether the difference in the large scale meteorology composites were due to more cases in one year over the other.

Lines 160-162: This sentence seems out of place.

Line 165: How is stratiform vs cumuliform determined? The surface observers?

Line 181: typo (should be bases, not based)

Line 223: typo (days WITH little smoke)

Section 7: Important details are missing regarding the RRTMG simulation. What exactly is being fed into RRTMG? Clouds must be represented somehow due to the peak in SW warming and LW cooling at the top of the PBL, and the kink in LW at cloud base around 750 m.  Is an angstrom exponent specified? Does SSA not vary by wavelength or humidity for simplicity? I also feel that more can be done with this type of simulation as the section seems somewhat incomplete. How does the change in height of the aerosol layer from July/August to September/October alter the heating structure and stability of the atmosphere?

Figure 1: Please add panel labels

Figure 2: 1) fine mode AOD should be in the legend box; 2) \tau_AC is used in the caption, but ACAOD is used everywhere else; 3) The amount of information in this figure is almost overwhelming. Consider making panel b its own figure and rearrange so that the winds are in separate panels from the other variables.

Figure 8: This figure should be split into 2 panels, one with observations, the other with the radiation output. Qv and theta should also be displayed with different colors or line styles. Variable names should be consistent with previous figures which use potential temperature