the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 09 Dec 2024
Aerosol-cloud interactions in liquid-phase clouds under different meteorological and aerosol backgrounds
Abstract. We conduct a comparative analysis of aerosol-cloud responses in liquid-phase clouds under different aerosol and meteorological conditions based on simulations using the WRF-Chem-SBM model to improve our understanding of aerosol-cloud interactions. This study reveals that in relatively unstable but dry atmosphere, aerosols uplift cloud top height but have no significant impact on cloud thickness, while also suppressing precipitation in clean conditions (sea salt aerosol only). In relatively stable but humid atmosphere, aerosols significantly increase both cloud top height and cloud thickness. Although aerosols also suppress precipitation in clean conditions, they promote the occurrence of relatively intense precipitation by facilitating the development of deep clouds. Aerosols have both enhancing and weakening effects on cloud liquid water path (CLWP). The weakening occurs mainly through two mechanisms: 1) by increasing Nd in thin clouds within a dry atmosphere, leading to smaller droplet sizes, which enhances evaporation within clouds and thus reduces CLWP. 2) By lifting cloud top height, facilitating the transition of liquid-phase clouds into mixed-phase or ice-phase clouds. The enhancing effect becomes more pronounced in environments with a relatively high column-averaged relative humidity, and is also modulated by atmospheric stability: 1) under low lower tropospheric stability (LTS), aerosols cause a relatively brief, explosive increase in CLWP. 2) Under high LTS, aerosols lead to relatively persistent increase in CLWP. For the liquid-phase clouds in the study, aerosols affect cloud development but have no significant impact on cloud lifetime, and precipitation affects the short-term variation of Nd but does not change its overall trend.
- Preprint
(11688 KB) - Metadata XML
-
Supplement
(2722 KB) - BibTeX
- EndNote
Status: open (until 20 Jan 2025)
-
RC1: 'Comment on egusphere-2024-3662', Anonymous Referee #1, 02 Jan 2025
reply
This manuscript shows real-world simulations of a cloud-aerosol interaction case and compares it to observations. The simulations are sophisticated, e.g., they use bin microphysics. The simulations are potentially informative, but I had difficulty understanding the manuscript. E.g., several times, the authors claim that one of their figures shows some phenomenon, but when I look at the figure, I don’t see that phenomenon. Some instances are listed below. Maybe the figures need to be redesigned, for instance, split up and simplified. In addition, while there are lots of words describing what is (purportedly) plotted in the figures, a key issue of how the aerosol affects rainfall is not convincingly explained.
Abstract: Please state what type of cloud you’ll be exploring. Stratocumulus? Cumulus? Shallow clouds? Deep clouds?
Abstract: The abstract leads the reader to believe that the manuscript will discuss liquid-only clouds: “We conduct a comparative analysis of aerosol-cloud responses in liquid-phase clouds”. But only a few sentences later it talks about mixed-phase clouds “the transition of liquid-phase clouds into mixed-phase or ice-phase clouds.” Can the wording be clarified?
Line 48: “in non-precipitating clouds, CLWP first increases and then decreases with the increase in Nd.” This is somewhat similar to the result found by Chen et al. (2024)’s paper entitled “Magnitude and timescale of liquid water path adjustments to cloud droplet number concentration perturbations for nocturnal non-precipitating marine stratocumulus”.
Line 90: “resolutions of 12 km and 2.4 km”. A 2.4-km grid spacing won’t be able to fully resolve turbulent updrafts (see Fig. 4), and hence won’t activate aerosol accurately. Are subgrid updrafts parameterized in the model? If so, how? Can you make some comments on the accuracy of the simulated vertical velocity?
Line 113: “Sen experiment (counterfactual scenario)” I’ve never before heard the expression “Sen experiment”. Is “Sen” an abbreviation for “sensitivity”? If so, please say so.
Fig 5: How does the precipitation profile differ between the NT and T runs? Could you include the precipitation profile in Fig. 5? It looks like the T time period has a greater reduction in CLWP. Is this what one would expect given the theory described in Ackerman et al. (2004)’s paper entitled “The impact of humidity above stratiform clouds on indirect aerosol climate forcing”? (However, in the T case, the RWP decreases when the continental aerosols are omitted, which may differ from what Ackerman found.)
Line 274: “Continental aerosols have a significant impact on precipitation in ECO (Fig. 7). In the absence of continental aerosols, during the more unstable NT with stronger vertical motion, the rainwater path (RWP) in ECO (4.7 g·m⁻²) is much higher than during T (2.2 g·m⁻²). However, in the environment with high aerosol concentrations that includes continental aerosols, Nd increases, and cloud droplet sizes decrease to below the precipitation threshold, leading to precipitation being significantly suppressed in areas with high rainwater content (RWC) in the Sen experiment.” These sentences are hard to understand. Are continental aerosols included, or is this the Sen experiment? Which RWP is greater than which other RWP? In these sentences, why not include mention of the titles of the panels in Fig. 7 (Sen_RWC_NT, Sen_RWC_T, etc.)? In general, I don’t understand why Sen_RWC_T has *less* rain than Control_RWC_T (see Figs. 7 and 8). It seems counterintuitive, from the perspective of the Albrecht lifetime hypothesis.
Line 298: “As shown in Fig. 8, RH exhibits a relatively clear negative correlation with LTS over the four areas during NT and the following 12 hours. In contrast, LTS is relatively high, and there is no significant correlation between RH and LTS during T, suggesting that changes in RH during this period are mainly driven by horizontal variations 300 in temperature and water vapor content.” I don’t see the “clear negative correlation with LTS”.
Line 318: “Overall, while continental aerosols lead to a decrease in CLWP in some conditions, they generally have a positive impact on CLWP.” Do you mean *the presence of* continental aerosols?
Line 321: “The additional continental aerosols also accelerate a transition to ice-phase and mixed-phase clouds by elevating cloud top heights, for example, during the first 12 hours in areas A and B.” Where is ice plotted? By the way, in Fig. 8, I don’t even see a line for CWP. The legend for CWP is blank.
Line 324: “1) for low LTS conditions,such as in areas A, B, and C from 12:00 on the 2nd to 00:00 on the 3rd, continental aerosols cause a relatively brief and explosive increase in CLWP. 2) For high LTS conditions, such as during T in the four areas, continental aerosols lead to relatively prolonged high values of CLWP (around 10 hours) in ECO.” I don’t see this in Fig. 8. Rather, I see peaks in CLWP both between 12:00 on the 2nd to 00:00 on the 3rd, and also during period T.
Line 330: “In terms of precipitation, during the relatively dry NT, continental aerosols cause reductions in both the number concentration of raindrops (Nr) and RWP.” The RWP line is barely distinguishable from zero. It is hard to see.
Citation: https://doi.org/10.5194/egusphere-2024-3662-RC1 -
RC2: 'Comment on egusphere-2024-3662', Anonymous Referee #2, 08 Jan 2025
reply
Summary
In this study, the authors utilize WRF case study simulations to evaluate the influence of meteorology and aerosols on cloud properties. Simulations are set up to capture the eastern coast of China and the marine clouds off the coast. Four types of sensitivity cases are evaluated, testing the influence of anthropogenic aerosols (Control vs. Sen, which has anthropogenic aerosols removed) and of aerosol transport from the mainland over the ocean (T vs. no transport, NT). The Control simulation, which assimilates reanalysis information, is compared against satellite and surface observations and reanalysis to establish the skill of the WRF set up. Then the four types of cases are contrasted to evaluate meteorological (Control T vs NT) and aerosol (T vs. NT, Control vs. Sen) influences. Conclusions are drawn about the influence of aerosols in varied meteorological states and the influence of meteorology on the marine cloud systems.
The conceptualization of this study is interesting and the mission, to understand how meteorology and aerosols influence marine clouds, is an important one. However, there are several critical problems that need to be addressed before publication can be considered. Their analysis is not sufficiently robust to support their conclusions at this point. They have not done the necessary statistical testing and quantification of uncertainty needed for identifying aci signals above the noise of the complex system they are examining. The skill of the underlying Control simulations and whether they can reproduce observed aerosol and cloud properties is not robustly demonstrated (Section 3.1), impairing the subsequent sensitivity tests and comparisons conducted in the final sections (3.2-3.3). Their conclusions (Sections 3-4) infer causality from largely correlative and qualitative comparisons that are confusingly presented. While there could be potentially useful insights developed about meteorology-aerosol influence on these marine cloud systems, these analysis flaws impair that. It is also concerning that the simulation startup files and key outputs are not archived publicly for this paper, making their work irreproducible and unable to be evaluated by the community. While the introduction has good awareness of the literature, there is a lack of discussion of the results of this paper and how they relate to the literature. Because I am not confident as to the skill of the simulations or the robustness of any of the conclusions, as well as the potentially large task to improve the analysis to rectify these issues, I recommend rejection and resubmission. See below for more details.
General Comments
- There is a fatal lack of statistical analysis and uncertainties in this paper. The results appear to be largely qualitative comparisons that are correlative and not causal. Please perform statistical tests on all analysis (e.g., t-tests or non-parametric equivalents to evaluate distributions); establish the bias of the Control simulation against observations and propagate this uncertainty through to the sensitivity simulations; and provide uncertainty measures for all the comparisons you make (e.g., 2 standard error, interquartile range, etc.).
- Please also be clear about what you are comparing and what domains you are computing variable values (means?) for in your plots. While the red box and a, b, c, d regions are discussed, it is unclear when they are used and what the criteria is for their use. Across this paper, please include uncertainty bars and statistical quantification to demonstrate when differences between NT and T and between Control and Sen are significant (and at what confidence level). Otherwise, the figures are visually engaging and well executed.
- Provide simulation start files and output so that your results can be reproduced and evaluated by the community.
Detail Comments:
Lines 12-24 (Abstract): I don’t think you can make these claims without showing that the results are statistically significant (e.g., at 95% confidence). You also need to establish that these are causal linkages, not correlative, which is unclear from the analysis.
Figure 1 caption: define what is NT and T.
Section 2.3: It’s important that you have matched the satellite observations and simulations so that their sampling time and general retrieval criteria are similar. Very nice.
Line 157-158: How does not including thinner clouds influence your results? Please comment on the potential biases that this could introduce into the results.
Line 172-174: I don’t understand why you include the non-liquid clouds as zeroes… if you aren’t looking at the total, why do you need to add in this offset?
Figure S1 (and other profile comparisons): Please state whether the values shown in S1 are means, add uncertainties (e.g., 2SE, 25-75%), and indicate where the profiles are being computed over (what is the outer domain, the whole region?).
Line 189-190: You are suggesting that the Control agrees with the meteorological behavior (please show statistically, see above), which you note it must due to data assimilation. However, that says nothing about how well it does at getting the aerosol (and cloud) properties so remove the “in consequence” statement based on S1.
Figure 2 and 3: As far as I can tell, these are qualitative comparisons. It does look promising but it is hard to tell from this what the actual skill of the Control simulation is in aerosol and cloud properties (which I agree is crucial to establish in this section). I recommend that you focus on the Control to observation comparison in this section and remove the Sen column (see comment below). Instead, I strongly encourage you to make the last column in both these figures the difference map between Control and observations so that it is obvious what the differences are and how they may geographically vary. You can then use this to compute the bias between the Control and observations for these various parameters, which you can then propagate through the other comparisons later (i.e., to show that the meteorology or aerosol influence signal is larger than the bias in the Control simulations). I also strongly suggest that you do a statistical comparison between the observations and Control. For example, you could look at a coarser grid and test whether the value distribution for each coarse grid box is statistically similar (e.g., with a t test, if assuming they have a normal distribution, or a non-parametric equivalent) and what their r values are at 95% confidence. For Figure 3 (and Figure S2) please also include the land borders so easier to compare across figures.
Line 203-212, Section 3.1: I would recommend not discussing or introducing the Sen experiment until the next section. It’s clearer to really focus on establishing that Control is reliable by comparing to observations in this section. It’s confusing to be discussing Sen at the same time especially as its unclear at this point whether we can trust the underlying simulation. It would also be helpful to introduce more details of how you have done the Sen experiment (minimal in the methods) before discussing the Sen results. Specifically for this paragraph, you need to do a statistical comparison between the Control and Obs (to show the Control captures the key cloud properties) and the Control and Sen (to establish the magnitude of the difference and whether that signal is larger than the Control to Obs bias). Visually, the cloud properties look much more different from the Obs than the aerosol. This likely impairs the ability to evaluate cloud changes in the Sen experiments.
Line 219-220: I disagree with this statement. Please include more extensive evaluation including statistical assessment of how well the model can reproduce the meteorology, aerosol, and cloud parameters to establish the skill of the Control simulation. Because you are assimilating the meteorological information, that's reasonable to get right. However, you need to demonstrate statistical skill for the aerosol and cloud parameters to analyze aerosol and cloud later in the sensitivity experiments. Please separate out the Sen experiment discussion until after you have shown that the Control experiment has sufficient skill. It will help clarify this story and your conclusions a lot.
Section 3.2, 3.3, Conclusions: Unfortunately, without establishing the bias in the Control relative to the observations, it is unclear whether any of the sensitivity test comparisons are statistically significant and thus informative.
Figures 4-6: For all the profile and distribution (mean?) comparisons, please indicate what part of the domain they are over (the red box?). Please add some uncertainty measure, such as 2SE or 25-75% for these to show the statistically significant differences between simulations.
Line 223: To test this, please establish that the Control simulation sufficiently captures these behaviors first for both T and NT cases.
Line 226-227: From the Control or observations? Is this all the region or just the red box collapsed into profiles?
Line 233-235: Where is this shown?
Line 243-244: Do you show that these are causal linkages? Or are these correlations for the behaviors and cloud properties during NT and T cases? If the latter, they should not be discussed as causal. Please be clear of your comparisons and whether they are causally informative here and throughout.
Line 244-246: Please expand on what you mean here, I don’t understand the point you are making.
Line 246-248: It’s not clear this is causal to me, are you gathering this from correlation or comparisons between Control and Sen?
Line 250: Can you show that “similar to that near the source area.”?
Line 253: You discuss this as causal (“led”), is this the comparison between Control and Sen? Otherwise, correlative, please change this language and clarify.
Line 259-260: Please show the supersaturation to support this statement.
Line 260-262: Can you please rephrase? I’m not sure I understand what you mean here.
Line 262-265: Are these statistically significant effects?
Line 274: At what level is this statistically “significant”?
Line 280: Which figure shows that it’s unstable and dry? Please indicate.
Figure 7: This difference plot is very helpful. Please do something like this for all the other map comparisons (Control vs. Obs or Control vs. Sen). However, is this difference larger than the bias between Control and Observations? Please have some indication on here for what is a statistically significant difference. Can you compare the control and sen distributions for the levels in the gridded regions, as suggested before, to determine this?
Line 295-296: I don't understand the rationale here. Usually, you want to look at how the clouds are influenced by the environment, so you pick either a sub/surface level (925hPa, Wood 2012) or something above the cloud in the free troposphere (e.g., 700hPa) (e.g., Klein et al. 2017). Can you show that this metric is giving you useful information about how the cloud is being influenced by the environment?
Figure 8: This is a very hard figure to read... I suggest separating out more of these parameters so you can tell a clearer story. Also, if you are discussing correlations over time it would be informative to use lagged correlations (showing r when at some confidence level, e.g., 95%). These could help to test hypotheses of what is influencing the cloud parameters and which factor is comparatively more important. I noticed that the regions you are comparing (a, b, c, d) is marked in Figure S2. Please also mark them on the maps in the main text so it is clear. Also, please explain the rationale behind choosing these boxes as they seem widely dispersed and quite small. Why are these informative? Is there a flow between the boxes you are trying to capture?
Line 303-305: Please clarify what you mean here.
Line 305-307: Can you show this? Do you see that the critical supersaturation is modified in the updrafts?
Line 307-308: You say “(due to the transient and localized nature of supersaturation, RH is used to represent the overall supersaturation intensity in this environment)”. If this is how you are using this integrated RH quantity, you need to explain this at the beginning of the controlling factor discussion and demonstrate that it is really containing the supersaturation information you believe it is. Because you are integrating over the cloud as well (not just the sub-cloud layer) and weighting to where there is more moisture, I suspect that this is telling you about how juicy the cloud is, not about the moisture in the updraft that the aerosol is experiencing when lofted and activated.
Line 309: As noted above, if this is essentially the cloud liquid measure it likely does not inform you of the impact on aerosol activation. I suggest either looking at RH below cloud or, preferably, pull the critical supersaturations in the updrafts so you can get the activation. Please demonstrate that the RH metric has some relationship with the activation potential you are claiming here. What about the differences in aerosol composition and size that could also impact activation capability between the anthropogenic dominated Control and sea spray dominated Sen? How does that influence aerosol activation ability?
Line 327-329: I don’t see how you can disentangle the meteorology and aci effects to make this statement. Please show the work that supports this. One way to isolate the aci from the meteorology is if you show that the meteorology is the same in the Control and Sen experiments for the NT and T cases, separately. If you were able to establish that they are relatively similar (for the respective NT and T cases), then you may have some causal connection in the difference between Control and Sen that is informative of aci. Could you quantify the contributions of adjustments and Twomey effect using something like the Erfani et al. 2022 method? You could maybe then infer something from comparing the NT and T aci contributions. Please include uncertainties and statistics to show if the differences are significant.
Line 330: How do you know they are causing this precipitation change?
Line 340-342: Your analysis does not support this conclusion. See previous points about i) statistical testing, ii) biases in Control from obs being larger than Control-Sen differences, iii) decomposing Twomey and adjustment effects, and iv) controlling for meteorology. For the latter, you likely need to show that the meteorology is relatively unchanged between Control and Sen except for their cloud and aerosol, otherwise you can't distinguish the meteorology and aerosol influences.
Line 352-353: I disagree with this. Statistical analysis needs to be employed here to establish Control captures real world behaviors and the bias from observations. This bias will be essential to account for in your subsequent analysis.
Line 355: How are the aerosols disabled? Does this effect anything else about how aerosols are handled or just the number?
Line 355-357: These four different comparisons need to be discussed more clearly throughout so that it’s obvious when you are contrasting NT and T (not causal) and Control and Sen (theoretically more causal if you are only changing the aerosol and the meteorology is the same). I found this quite confusing in reading the text and evaluating the plots (which are also over varying domains).
Line 359-360: Where do you show this: “the atmosphere fails to enable full activation of aerosols during both periods”?
Line 365-367: This does not seem like a surprising conclusion, reference previous literature?
Line 369: Please show the representative areas on the maps in the main figure and make it clear the criteria for how you chose them and why they are representative.
Line 371: Where do you show this: “aerosols only affect clouds during their development stage without noticeably impacting their lifetime”?
Line 373: Please look at the supersaturation itself to support this: “by supersaturation, with low supersaturation limiting the full activation of continental aerosols.”
Line 375-376: Do you show this change in phase or are you inferring it? Please show work or cite literature.
Line 379-381: Do you show this somewhere? Please clarify what you mean here and indicate the supporting work.
Line 385: Please provide all the necessary setup files and the key outputs from your simulations at an archive (e.g., like the free Zenodo) so that your simulations can be reproduced by the community and evaluated.
Section 4 (or before): Please include some discussion of how your results relate to prior work in the literature and provide some contextualization of this study.
References
Erfani, E., Blossey, P., Wood, R., Mohrmann, J., Doherty, S.J., Wyant, M., O, K., 2022. Simulating Aerosol Lifecycle Impacts on the Subtropical Stratocumulus‐to‐Cumulus Transition Using Large‐Eddy Simulations. JGR Atmospheres 127, e2022JD037258. https://doi.org/10.1029/2022JD037258
Wood, R., 2012. Stratocumulus Clouds. Mon. Weather Rev. 140, 2373–2423. https://doi.org/10.1175/mwr-d-11-00121.1
Klein, S.A., Hall, A., Norris, J.R., Pincus, R., 2017. Low-Cloud Feedbacks from Cloud-Controlling Factors: A Review. Surveys in Geophysics 38, 1307–1329. https://doi.org/10.1007/s10712-017-9433-3
Citation: https://doi.org/10.5194/egusphere-2024-3662-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 121 | 32 | 6 | 159 | 17 | 2 | 3 |
- HTML: 121
- PDF: 32
- XML: 6
- Total: 159
- Supplement: 17
- BibTeX: 2
- EndNote: 3
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1