First review for ACP-2022-615

Summary

This work examines surface-based observations at the Eastern North Atlantic (ENA) site on Graciosa Island in order to characterize the seasonal cycle and budget of boundary layer moisture. The manuscript is well written and the authors have plenty of history and skill in working with these observations but I struggle to understand some of the important decisions that are made in the analysis process. My main concerns are related to the fundamental assumption of a well-mixed boundary layer at ENA and the severe limitations in selecting data to include only the extremes of a particular weather regime while also claiming to present an encompassing depiction of moisture variability at ENA. Both of these concerns are explained in my General Comments.

General Comments

1. The authors assume a well-mixed boundary layer and justify their assumption by citing Albright et al. (2022) but the Albright paper was focused on the EUREC4A boundary layer, which is much closer to the Equator and actually located within the classical trade wind region. The ENA site often experiences decoupled boundary layer conditions and mid-latitude cyclone disturbances, invalidating this assumption that might be important for the conclusions presented in this work. The authors acknowledge that the assumption is potentially inaccurate but useful (lines 60-61). This manuscript should make more of an effort to justify this assumption. For example, this validation could take the form of showing that the decoupling index from ENA sondes is similar to EUREC4A or is generally low. Sonde profiles could also be shown, similar to Figure 2 in Albright et al. (2022). This could be included in the supplemental material, as the validation of the mixed-layer model for ENA is not really the focus of the manuscript.

Albright, A. L., S. Bony, B. Stevens, R. Vogel, 2022: Observed sub-cloud layer moisture and heat budgets in the trades. Journal of the Atmospheric Sciences, DOI 10.1175/JAS-D-21-0337.1

2. The manuscript employs a strict definition of “marine conditions”, discriminated solely on the basis of surface wind direction measured at the ENA site, which immediately discards 70% of the available observations. While the desire to eliminate effects of the island on observations is reasonable, given the goals of understanding the marine boundary layer, I wonder if the wind direction limitation also limits the analyses to certain weather regimes and biases the conclusions. Is the boundary layer either “not marine” or strongly affected by the land surface if the wind is from the west (wdir=270°)? There is likely some extra aerosol loading from the island’s natural and human activity but does that significantly affect the moisture budget over such a short distance from shore? I would like to see some discussion of this potential issue either in Section 2.2 or Section 6 or both.
3. This manuscript seems to be more about moisture budgets during marine stratocumulus-topped boundary layers than simply “marine conditions”. I think a change to the manuscript title is appropriate in order to accurately advertise the analyses according to the targeted weather/cloud type. The first paragraph of the discussion section also mentions, “we have examined the factors that control boundary layer moisture at the ARM ENA site on a seasonal and daily temporal scale using 5 years of ground-based observations and reanalysis data” but the really only the fully-overcast stratocumulus time periods with a particular surface wind direction were analyzed.
4. Many of the more interesting analyses and results are presented in the latter parts of the manuscript but there are so few samples due to inclusion of only 10 (hand-selected?) days. The conclusions formed from work explained in this manuscript would greatly benefit from an increased pool of data.

Specific Comments

60: Are decoupled conditions common at ENA? See my General Comment 1 for more.

144: This may well limit cases to marine environments but other environments exist at ENA so any conclusions are only for the marine state. Please be sure to make that clear throughout the paper. Only 30% of cases are “marine”?

148: Why such a strict requirement? Is this cloud fraction computed from hourly data so there are 24 values per day? When the manuscript says, “In the following discussion only boundary layer clouds with cloud fraction from the ceilometer greater than 0.99 were selected…”, does it mean all remaining analyses in the manuscript or just the brief remainder of Section 2 and Section 3?

178: “a stronger contribution of the free troposphere to the total PWV in summer compared to winter”. Examining Figure 3b, I find it hard to see whether the fractional contribution from z>3km to the total is higher in summer or winter. Certainly, the raw values of PWV(z>3km) are higher in the summer months. Maybe some clarification in 178 is appropriate to distinguish if you mean relative contribution or simply that the annual cycle of PWV(z>3km) peaks in the summer months, which was already stated in line 175.

Figure 3: The annual cycle in PWV would be easier to see in Figure 3a if the y-axis limit was reduced to 5 cm. Also, why are the lower y-limits not 0 in all cases?

Figure 3: It would be nice to show how many observations are used in each month as the top panel of this figure. Annual cycles in wind direction and cloud type could play a role in the interpretation of these results.

207: This actually appears to be a very small dataset of only 304 points! At this point in the manuscript, no ERA5 data are yet used, right? So why have the analyses been averaged up to hourly resolution? This is likely an overly-strict limitation that throws away too much valid data. As I understand it, Figure 4 is showing only 304 out of an initially-available 52608 points, only about one half of one percent of the total data. I realize that radiosondes aren’t released every hour, but the cloud fraction>0.99, wind direction, and weak precipitation requirements are likely overly strict and therefore likely to bias the conclusions of these analyses.

219-220: Many of the cases, even for relatively thick (>500 m) clouds, have adiabaticity considerably greater than one, with some as high as 150% adiabatic. Please provide more discussion and validation of the cloud boundary argument. If these cases simply appear to be superadiabatic due to uncertainties in cloud boundaries, that uncertainty will likely also affect the cases that appear sub-adiabatic. Can these “superadiabatic” cases result from the microwave radiometer seeing elevated large rain drops instead of only smaller cloud drops? More discussion is needed here because this occurs for a large portion of the limited dataset even for relatively thick clouds.

230: Does the liquid potion (ql) also include rain water mixing ratio? I assume it does from equation 1 but it would be helpful to be explicit about it in line 230.

254: The MBL height is likely not constant during the 12 hours containing a given PBL measurement. Would it be better to use a moving polynomial interpolation?

Line 273: If you’ve limited the analyses to when wind direction indicates “marine conditions”, why are there any cases with v>0? Disagreement between ENA observations and ERA5?

276: Why omit the extremes? What evidence do you have that these extremes are simply instrument noise instead of real events that should be included in the analyses?

301: Air density is not constant in the boundary layer. Is there confirmation that the vertical change in atmospheric density matters much less than changes in PBL height? Or do you mean that the profile of air density is relatively constant in time?

387: How were these cases chosen? The manuscript says “The selected days displayed persistent boundary layer cloudiness and at times precipitation” but how were they identified? By eye? Some thresholds on cloudiness?

397: Why are daily averages used? Do the winds not change at all during each of the 10 days? These cloud base winds come from ERA5, right?

400: So there are 345 mesoscale chunks during the 10 selected days? The chunks from within a given day have the same amount of temporal averaging but there are different temporal averaging times between the 10 days?

433: In line 428, the manuscript states that “moist and dry neighboring columns … were compared to those of its preceding and following neighbors. For clarification, are all of the 345 mesoscale perturbations considered but only 143 columns had durations longer than 10 minutes or was there a requirement for Figure 10 that only moist columns that were neighbored by dry columns be considered?

Figure 11: There are really few cases here. I think the filtering is too strict. How can we know that these 5 cases constitute a representative sample the full population for PWV perturbations from 0.8 to 1.0 cm?

469: Please remind the reader what papers have proposed and promoted this mechanism.

473: Why might the doppler lidar retrievals be invalid in a given layer? Could these observational limitations imposed by the doppler lidar result in a biased view of the vertical motion?

Figure 12: It would be simple and helpful to add uncertainties to (c) and (d) by subsampling/resampling techniques, which would strengthen the claims about differences between the two distributions.

538: Other papers have suggested this, too, right? They should be cited here.

Technical Corrections

238: It would be best if you used either “3^rd and 4^th” or “third and fourth”.

418: Both the LWP and precipitation increase when columns are moist so you do not need the “respectively”.

418: Parentheses are for clarification and references, not the exact opposite of what was just stated. You could remove the content of the parentheses and the sentence would be much clearer to the reader. Sentences like, “Moister columns correspond to regions of increased liquid water path and precipitation.” and “Increased moisture is associated with increased liquid water path and precipitation” are understood more easily. Additionally, in the preceding paragraph and in the following sentence, parentheses are used for clarification instead of opposition.

436: A dryer is a household appliance for making wet close dry. You want “drier”. I confuse these two often, too

474: Why is “downdrafts” in parentheses?

This paper analyses the variability of boundary layer moisture at the ARM ENA site using ground-based observations and ERA5 reanalysis data. The authors compute mixed-layer water budgets at monthly timescales and analyze the respective contributions of the different terms, and then assess mesoscale moisture variability on sub-daily timescales and their relationships to updrafts, liquid water path and precipitation. The analyses seem sound, and the paper is well written and quite easy to follow. My major concerns are that (i) many assumptions are not justified or discussed, (ii) I did not understand if the authors just focus on stratocumulus or also shallow cumulus conditions, (iii) the story and sequence of analyses is hard to anticipate, and the novelty and connection among different sections is not really clear, and (iv) uncertainties are not systematically quantified.

These comments, along with some more minor comments, are addressed in more detail below.

Major comments:

1. Justification & discussion of assumptions: For many of the assumptions made here, I miss a justification or discussion. For example, are the boundary layers analyzed really well-mixed up to the boundary layer top? For stratocumulus conditions I’d assume this is the case, but for decoupled shallow cumulus conditions, only the sub-cloud layer is well mixed. (See also comment 2 regarding the cloud regimes below). Could the authors demonstrate the well-mixedness of their boundary layers, and discuss how different cloud regimes might affect the budgets?

I understand that ERA5 data is necessary to complement the observations, but did the authors check whether the moisture profiles are consistent with the radiosonde and Raman lidar profiles? E.g., are the moisture profiles consistent enough that ERA5 can be used to computed the gradient? Even if this can be checked only for a limited data sample, it would greatly increase confidence in the approach. Similarly, for the local tendency and the PBL heights, could the Raman lidar profiles be used to check the hourly variability and a potential diurnal cycle in the terms that would be missed with the twice daily radiosondes?

2. Cloud regimes: As alluded to in Sec. 1 and 2, both stratocumulus and shallow cumulus conditions are frequent at ENA. But these cloud regimes are associated with coupled vs. uncoupled boundary layers, which is a relevant difference for this study. Except for Sec. 5, which clearly addresses stratocumulus conditions, only in L147 a cloud criterion is mention: “In the following discussion only boundary layer clouds with cloud fraction from the ceilometer greater than 0.99 were selected (total of 3580 hours).” So does this mean that the entire discussion of the budgets focuses on stratocumulus conditions with ~100% cloud cover? And does the 0.99 threshold refers to hourly values? Please clarify, and discuss more prominently.

3. Story, structure of the paper and novelty: I found it hard to anticipate the story and the structure of the paper from the abstract and the introduction. For example, Sec. 3.1 on the cloud adiabaticity seems rather peripheral and came quite unexpectedly. And at first, I expected just an analysis of the water vapor budget, and then I realize the paper focuses on the total water budget, and also constrains the mass budget. So it would have helped me if the story of the paper was clearer and if the reader’s expectation was a bit more guided.

In a similar spirit, I missed the coherence and connections between the different Sections. How do the monthly mixed-layer budget analyses connect to the analyses of the mesoscale variability? Could you construct mixed-layer budgets also on shorter timescales to connect to the monthly budgets and understand how the terms contribute differently at different timescales?

I would also recommend the authors to highlight more explicitly what is new in the paper. How does the novel retrieval used, which better separates cloud and drizzle contributions to TLWP, affect the robustness of your analyses and conclusions? What are the novel insights gained with the ENA data here compared to previous ML budgets? In L541 it is mentioned that ‘the results presented herein are useful for future observational and modeling studies on low clouds conducted at the ARM ENA site’, but can you be a bit more explicit?

4. Uncertainty quantification: I missed a quantification of the uncertainty of the different terms in equations (1) and (3). Also, can you briefly say in L133 how the uncertainties of the retrievals are estimated?

Minor comments:

• Retrievals section 2.3: I think this section could be written a bit more concisely. From one paragraph to the other, it seems to jump from one algorithm (with unfamiliar acronym for me) to another. Also, I think it could be worthwhile to present the comparison of SPARCL and MWRRET (L141 onwards) in an appendix, not to depart from the main story of the paper too much.
• Use of commas: I missed a lot of commas throughout the manuscript, e.g.:
  • L113: were derived, cases that ....
  • L134: For ten selected cases of weakly precipitating marine stratocumulus clouds, vertical profiles of water
  • L143: Traditionally, the total liquid water path retrieved by radiometers is assumed to represent the cloud water path. However, in the presence
  • ... Please check the entire manuscript carefully.
• References: I’m not sure which program (if any) the authors use for the references, but it’s not consistent and sometimes erroneous. E.g. Zhou and Bretherton 2019 is cited differently in L381 and L432. Also, the reference ‘Shultz and Stevens, 2018 is not correct (L379), or (Zheng et al. Lamer et al. 2019; ...) in L44. Please improve throughout the manuscript.
• L276: I understand that strong precipitation rates can introduce strong peaks, but this is the intermittent nature of rain, and I guess not a measurement error. Can you understand from your data how such strong precipitation rates are locally balanced? It would be very interesting to analyze this.
• L308: please specify what kind of filter is applied here.
• L343: “The seasonality of the large-scale advection term is also the factor that determines the seasonality of the overall budget.“ --> Please clarify how you get to this statement. From Fig. 7 it seems that the seasonality in the LHF or precipitation terms is also very large.
• L345ff: Does the magnitude of the monthly residuals depend on how much data was used per month? I.e. if only a few days of data could be used, it might not be surprising that the monthly budgets don’t balance well. Please clarify.
• L389: How do you interpolate 10-min profiles over 1 minute? Do you downsample the data?
• Figures:
  • 11: Please specify what normalized height refers to in the right panel
  • 12: the figures are far too small and can hardly be read..
• L511: I do not understand this sentence, please clarify: “The lower mid-tropospheric humidity during the winter months, together with turbulence (Ghate et al. 2021) point towards turbulence being the primary controlling factor rather than water vapor in determining cloudiness in the region.”
• L532: “Moist and dry patches present differences in vertical velocity with dry regions displaying more frequent downdrafts than moist regions immediately below the cloud base.”  from Fig. 12 I’d say the opposite, please clarify.

Typographical suggestions:

L45: that that the water ...

L229 / L263: the total water mixing ratio is once written in normal and once in bold font – please harmonize.

L234: please use proper math formulation for the averaging brackets.

L279: maybe add after ‘... the entrainment rate (see Sec. 4.2)‘

L303: ...balanced by local change in the boundary layer height (?)

L321: large-scale turbulence -->  do you mean subsidence instead of turbulence? And previously, largescale was written without ‘-‘, please make it consistent.

L322: that what reported --> than what was reported ..

L514: hear --> heat