Stability dependent increases in liquid water with droplet number in the Arctic by Murray-Watson and Gryspeerdt 2021

Under the background of global warming and Arctic amplification, the aerosol-cloud relationship is becoming increasingly important in the polar region. However, this relationship is hard to quantify in the high latitude due to a limited number of observations in the Arctic. Using MODIS and AMSR-E satellite retrievals, this work studies the effects of aerosols on liquid Arctic clouds over open ocean under different metrological conditions (specific humidity, LTS). Particularly, they focus on the relationship between cloud droplet number concentration (Nd) and liquid water path (LWP). Overall, this paper is well written and easy to follow. However, there are still some major deficiencies in the current version and need to be carefully handled. For example, it is concluded that there is a positive LWP-Nd relationship under high LTS condition. However, if we take a further look, the LWP-Nd relationship is not clear and strong enough when Nd falls into [50,200], which is very common based on the analysis in this paper. In addition, the authors did not check whether this relationship differ by season. Therefore, this paper should be accepted only if my concern has been well addressed.

Please find my specific comments as below.

Major comments:

Section 2-Materials and Methods:

1) What are the original spatial resolutions for AMSR-E and ERA5 data sets? In addition, it is highly recommended to generate a flow chart or schematic diagram to show how you filtered out cloud data to reduce retrieval bias.

2) Give the unique environment of Arctic, it is very challenging to obtain accurate cloud information over the Arctic. Therefore, there should be more discussion on uncertainties in satellite-retrieved cloud properties in the Arctic.

Reference:

A radiation closure study of Arctic stratus cloud microphysical properties using the collocated satellite-surface data and Fu-Liou radiative transfer model: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2016JD025255

Figure 1: I would believe that the data samples are highly variable across regions and seasons, given the filtering steps mentioned in Section 2. Can you provide a sperate spatial map to show the number of samples by season?

Section 3.2: There is no explanation on how you separated the positive and negative sensitivity regions. Are there enough and comparable samples in both regions?

Line 158-162: The explanation on ocean-air temperature gradient and LTS is not clear enough. In general, smaller ocean-air temperature gradient occurs with melting ice in summer, thereby increases the atmospheric stability. In comparison, there is lower atmospheric stability in autumn due to larger ocean-air temperature difference. It is also valid for spring.

In addition, why do the positive sensitivities occur under high LTS conditions?

References:

Covariance between Arctic sea ice and clouds within atmospheric state regimes at the satellite footprint level: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1002/2015JD023520

Thicker clouds and accelerated Arctic sea ice decline: The atmosphere‐sea ice interactions in spring: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2019GL082791

Line 163-170: Have you tried to test this relationship the specific humidity in other vertical levels? Are you expecting any differences?

Line 191-193: Here, you mentioned that very low Nd conditions are relatively rare in the Arctic and therefore have little impact on the linear sensitivity. Based on Nd histogram in Figure 4, Nd between 50 and 200 (roughly) are more common. However, the LWP-Nd relationship is not clear when Nd falls into [50,200] under high LTS conditions, regardless of high or low q750. It seems to me that there is no any significant LWP-Nd relationship. How do you explain this?

Figure 4: Are these relationships different by season?

Minor comments:

Line 15: “As the LTS is projected to decrease in a future, warmer Arctic…” Change it to “As the LTS is projected to decrease in a warmer Arctic”

Line 15: “As the LTS is projected to decrease in a future, warmer Arctic, these results show that aerosol increases may produce lower cloud water paths, offsetting their shortwave cooling effect.” If this is the case, what is the overall implication to Arctic climate?

Line 22: “These smaller droplets increase cloud albedo and lead to a shortwave cooling effect.” Are you talking about shortwave cooling effect at TOA?

Line 43-52: It is better to move this paragraph to the beginning of the Introduction section as it highlights the importance of Arctic clouds and aerosol-induced changes to cloud radiative effects.

Line 56-57: “Coopman et al. (2016) found that if meteorology is not accounted for, the magnitude of the Arctic clouds response to aerosol is artificially increased by a factor of three.” What do you mean by “artificially increased”? Please be specific.

Equation (1): I can’t understand that why the breakdown of LWP-AOD relationship by Nd is helpful. As you mentioned that aerosol retrieval is very limited in the Arctic, we will still need Nd-AOD ratio to derive LWP/AOD relationship, right?

Line 132: “This may be due to a potential negative bias in the MODIS data due to retrieval errors (Gryspeerdt et al., 2019).” A negative bias in which variable?

Line 154:” The r2 values of the correlation between the sensitivity is higher” What correlations did you refer to? Please be specific.

Line 233-244 and Line 264-266: It would be important to further link the changes in LWP with Arctic sea ice to demonstrate the large-scale impact of liquid cloud on Arctic climate.

References:

Evidence of Strong Contributions From Mixed-Phase Clouds to Arctic Climate Change: https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2018GL081871

The climate response to increased cloud liquid water over the Arctic in CESM1: a sensitivity study of Wegener–Bergeron–Findeisen process: https://link.springer.com/article/10.1007/s00382-021-05648-5

Cloud Phase Changes Induced by CO2 Warming—a Powerful yet Poorly Constrained Cloud-Climate Feedback: https://link.springer.com/article/10.1007/s40641-015-0026-2

Review of “Stability dependent increases in liquid water with droplet number in the Arctic” by R. Murray-Watson and E. Gryspeerdt

Overview

This paper uses satellite retrievals and reanalysis data to investigate the effects of aerosol on Arctic liquid clouds. The authors observe a positive response in LWP to Nd in high latitudes, which contrasts with previous studies that show a weak correlation between the two. The paper is well motivated, with a clearly identified gap in current knowledge and comparison to results from similar studies throughout. The authors did a good job of linking statistics of Nd, LTS, q750, and LWP to physical processes such as precipitation suppression, droplet evaporation, and cloud-top entrainment. While I do not think any further analysis is required, additional analysis could be useful to better understand the indirect effects in this region and some of the discussion could be improved.

Main Comments

1. My primary comment is on the choice of meteorological factors considered in this study. While the authors did a good job of justifying the use of LTS, free tropospheric moisture, and MCAO as meteorological indices by referencing their use in previous studies, I think more discussion of other potential meteorological influences on the LWP-Nd relationship would be beneficial. In the results, LTS is shown to be the most significant of the chosen metrics in predicting whether the LWP-Nd relationship is positive or negative. However, the r² is still fairly low at 0.39. What other factors, especially those mostly independent of LTS, could be influencing the LWP-Nd relationship, and could there be a better predictor than LTS?
2. Figure 4 shows the most interesting results of this study. I spent quite a bit of time contemplating this figure and I think the discussion of the figure could be improved. First, the authors may want to point out explicitly that LWP begins to decrease with Nd at high Nd, high LTS, and low q750. To a lesser extent, the low LTS & low q750 panel shows the same thing as high LTS & low q750, namely, an initial increase in LWP with Nd followed by a decrease in LWP. The big difference is that at high LTS the peak in LWP is around 100/cm3 whereas at low LTS the peak in LWP is at 20/cm3. This difference leads to the interesting patterns in panel h. So, the question to me seems to be why precipitation suppression (which is driving an increase in LWP) at low LTS ends so early. Is it because at low LTS precipitation is weaker for a given Nd? Or is it that the drying effects of mixing are much stronger for low LTS and so precipitation suppression is less evident? The latter seems more likely. As written now, there is no discussion of precipitation in explaining panels g and h.
3. Finally, the authors discuss moisture inversions frequently, but I’m not convinced that they need to be invoked in order to explain anything in this study. For example, moisture inversions are discussed in lines 209-211. But can’t the clouds in the Arctic have higher LWP at low LTS for the same reasons as discussed in lines 203-206? And generally, I’d think that more moisture above cloud top should reduce evaporation, regardless of whether it is in the form of an inversion or not.

Minor Comments

1. Line 12: LTS isn’t necessarily the driving force behind spatial variations in LWP response, just the strongest of the metrics studied here. I’m uncomfortable with “driving force” given that the R2 value was somewhat low even for LTS.
2. Line 23: Smaller droplets lead to smaller coalescence rates.
3. Line 25: See also Williams and Igel (2021) who argue that smaller droplets radiatively cool cloud top more quickly, generating turbulence, etc. as already stated.
4. Section 2: Can the authors mention somewhere that they’re using sunlit times only?
5. Line 130: Caption for Fig 1 states panel (b) is JJA not all seasons. Figure 1: Panel (b) says “AMSR-E all seasons” but caption reads “AMSR-E June, July, and August”. Please double check everything for consistency.
6. Line 172: I’m not sure what was meant by this sentence. The “small influence” seems at odds with “a strong response”.
7. Lines 219-220: Not sure what is meant by the background Nd state.
8. Section 5: The authors might remind readers that they’ve only analyzed liquid clouds and not mixed-phase clouds.

General comments

The article Stability dependent increases in liquid water with droplet number in the Arctic studies the aerosol-cloud interactions in the Arctic. The authors used space-based observations from MODIS of liquid water path and cloud droplet concentration. They retrieved the sensitivity of LWP with Nd for different regimes of meteorological conditions and especially LTS and the specific humidity (meteorological parameters being inferred from reanalysis). The Nd-LWP relationship is especially high when the atmosphere is stable, and the impact of humidity is highly dependent on the stability regime. The authors used a variety of method to explain their results, I appreciated the effort made with the joint histogram to retrieve P(LWP|Nd) which I found very interesting and should be used more often in the future to study the aerosol-cloud interaction. 

The research in the article is within the scope of ACP and present some novel concepts to study aerosol-cloud interaction, the conclusions are interesting and concern a topic that is currently undergoing a lot of interests by the scientific community. The method and materials are presented and support the discussed results. The overall presentation is well structured, clear, and I appreciated the Figures which are very precise and understandable, especially Figure 4 that I found fascinating a lot of results and discussion can be inferred from it. Nevertheless I have some comments that need to addressed before I can recommend a publication in ACP, especially concerning the number of data points that are used in the study to support the statistical analysis, the conclusions, and the figures. I think some figures could be added (maybe as supporting information). Also I have a major concern concerning the uncertainty from the observations. All the comments are detailed in the following sections.

Major comments

1) Section 2, Materials: There is no mention about the uncertainty in the retrievals. The authors briefly discussed about it in section 4, but there is no mention about the uncertainty in tau_c and r_e. The errors from this variables can propagate and lead to large uncertainty in LWP and N_d and maybe offset the signals described by the authors. Did the authors looked at this issue? I think it should be properly addressed in the manuscript. 

    

2) Section 2, Method: Both N_d and LWP depend only on re and tau_c, therefore the two parameters are not independent. I am skeptical about how robust the study is. I think it is robust, but I also think that the two parameters not being independent should be mentioned in the article and the potential issues that might result from that.

    

3) l.108, "This stringent filtering is not applied to LWP retrievals". I do not understand how the filter is applied to Nd and not LWP. I thought the colocalization of LWP and Nd was made at the pixel level and one value of LWP requires one value of Nd (especially when retrieving the sensibility). I am wondering if there is something I did not understand in the method. Is it during the spatial resolution averaging (from 1km to 25km)? I am not sure I understand why filtering on one parameters and not the others. Can the author explain ?

    

4) l. 92: Is it fair to consider γ constant, consider the different conditions in the Arctic? Is γ really constant in spring and fall? Can the authors comment on that?

    

5) Section 3: There is no mention on the number of points that the analysis is based on, it is an important information to make sure that the study is statistically robust. I advice the authors to add a plot or supporting information to show that there is enough points, especially to retrieve the sensitivities. For example, in Figure 3, I am skeptical that there is enough points per bin to be robust. Also Figure 3 would benefit to have the 95% confidence interval on the calculation of the sensitivity, or at least a discussion about it. 

Minor comments

1) Title: I find the title of the article confusing and not clear, even after reading the article I am still not convinced by it. A change in the title would benefit to make it more direct.

2) There is a lack of quantification in the study, the authors often refer to trend when describing the plots. Also, quantification is needed in the abstract and the conclusion to make the study more robust and appealing for readers.

    

3) For example, the authors described the different variations to be "stronger" (l.190, l.198), "more pronounced" (l. 195)...". Quantification is needed here to support the text and observations. 

    

4) l.22, Twomey, 1977: The publication from Twomey is about the change in albedo rather than the change in cloud droplet size. Therefore the publication is more suited for the next sentence.

    

5) l.26: The authors do not mention the effect described by Stevens & Feingold (2009), about the delayed and the offset of the precipitation. Is it because it does not necessarily apply to Arctic clouds? I would mention it in any case.

    

    Stevens B, Feingold G. Untangling aerosol effects on clouds and precipitation in a buffered system. Nature. 2009 Oct 1;461(7264):607-13. doi: 10.1038/nature08281. PMID: 19794487.

    

6) l.42 "The warming effect of clouds have been linked sea ice loss...": A "with" is missing "been linked with sea ice loss".

    

7) l.79: MODIS has issue in the Arctic and often underestimate the cloud top temperature (see Fig. 1 from Tietze et al. (2011)). Can the authors comment on that?

    

    Tietze, K., Riedi, J., Stohl, A., and Garrett, T. J.: Space-based evaluation of interactions between aerosols and low-level Arctic clouds during the Spring and Summer of 2008, Atmos. Chem. Phys., 11, 3359–3373, https://doi.org/10.5194/acp-11-3359-2011, 2011.

    

8) l.107, "tau_c greater than 4". This condition filters a lot of clouds I guess. Can the authors estimate the number of clouds that is filtered out, to understand how representative is it to Arctic clouds?

    

9) l.148, "Mann Whintey U test": I am not familiar with this test, can the authors explain why it is suited for this case?

    

10) l. 154: I am confused about the r^2 parameter, does it refer to the sensitivity of LWP with Nd or LWP with LTS? If the latter, do the authors expect LWP to be linear with LTS, this is not necessarily the case. Same comment about Tsurface and MCAO. Also, I do not understand how mean LTS is taken into account here. 

    

11) l.156, "LTS is used...": Is LTS also correlated with Tsurf and MCAO? If not, the two other parameters have strong sensitivities and might be considered? I am wondering if the authors did the analysis considering different bins of Tsurf and MCAO. 

    

12) Figure 2: The description of r^2 is missing in the caption?

    

13) l.176, "assumed a linear sensitivity of LWP to N_d": The authors are considering logarithms on Equation 1, so why do they mention it is linear?

    

14) l.185, "specific humidity bins": q_{850} has a low correlation and does not seem to be the most important parameter from Figure 2. I am wondering if the authors looked at Figure 3 and 4, considering Tsurf or MCAO instead of q_{850}.

    

15) l.199, "Figure 4 (f) shows that the difference between...": I do not see it in the Figure. Can the authors rephrase and explicit, (again, maybe quantification would help) ?

    

16) Section 4: There is already a lot of discussion in section 3, I would suggest to either move all the discussion from section 3 to section 4, or have one section "Results and Discussion".

    

17) l.358, "Reassessing the Effect of Cloud Type on Earth?s Energy": There is a "?" in the title, change to "Reassessing the Effect of Cloud Type on Earth's Energy".