An observational study of the effects of aerosols on diurnal variation of heavy rainfall and associated clouds over Beijing–Tianjin–Hebei

Zhou, Siyuan; Yang, Jing; Wang, Wei-Chyung; Zhao, Chuanfeng; Gong, Daoyi; Shi, Peijun

doi:https://doi.org/10.5194/acp-20-5211-2020

Articles | Volume 20, issue 9

https://doi.org/10.5194/acp-20-5211-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/acp-20-5211-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 20, issue 9

Research article

|

05 May 2020

Research article |

| 05 May 2020

An observational study of the effects of aerosols on diurnal variation of heavy rainfall and associated clouds over Beijing–Tianjin–Hebei

Siyuan Zhou, Jing Yang, Wei-Chyung Wang, Chuanfeng Zhao, Daoyi Gong, and Peijun Shi

Download

Final revised paper (published on 05 May 2020)
Preprint (discussion started on 16 Nov 2018)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'Referee comment', Anonymous Referee #1, 01 Dec 2018
- AC1: 'Reply to RC1', Siyuan Zhou, 22 Feb 2019
RC2: 'Review', Anonymous Referee #2, 04 Jan 2019
- AC2: 'Reply to RC2', Siyuan Zhou, 22 Feb 2019

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

AR by Siyuan Zhou on behalf of the Authors (27 Feb 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (08 Mar 2019) by Philip Stier

RR by Anonymous Referee #1 (11 Mar 2019)

Suggestions for revision or reasons for rejection

The authors have done a lot of work redoing the paper, and I applaud their effort.

I have two critical concerns with this paper:
1. Moisture and CDNC/AOD covary in the sample. No multivariate analysis is undertaken and it is possible that covariability between CDNC/AOD and cloud properties or rain statistics are really just moisture variability covarying with precipitation and cloud.
2. The calculation of CDNC here does not acknowledge the numerous sources of retrieval error that may covary with precipitation.

My central concern- and what may be a critical error in the analysis- is that moisture and pollution covary. The authors reveal this in a later part of the paper. To me this potentially means that all their univariate analysis of rain statistics and cloud statistics as a function of CDNC and AOD are spurious and may just be aliasing onto meteorology. This may not be the case once variability in meteorology is better accounted for, but the authors need to state that moisture and aerosol covary in this region right away and then do all of their analysis in a multivariate way (for example, showing variables as 2D PDFs in the space of moisture and CDNC/AOD/AI instead of 1-D PDFs in the space of just CDNC/AOD/AI).

The authors have added CDNC, but their approach to this does not incorporate knowledge of the various sources of systematic error in CDNC retrievals from MODIS that we know of (Daniel P. Grosvenor et al., 2018). The authors need to expand their discussion of CDNC, clarify exactly how they are calculating this quantity, and discuss how known covariances between cloud habit, rainfall rate, and CDNC retrieval error might impact their results. This will substantially impact section 5.1 of the MS, but needs to be done as there has been shown to be substantial covariability between cloud morphology and CDNC retrieval error(D. P. Grosvenor & Wood, 2014).

The incorporation of OMI aerosol index is interesting,and supports their work. I do not know anything about these products, but it is less clear how spurious covariability between these retrievals and rainfall could occur than with MODIS AOD/CDNC. I suggest that they incorporate these measurements into their evaluation of rainfall statistics.

Overall, I think the paper has potential in the use of ground-based data to get around shortcomings in remote-sensing of rain, but some work has to be done on the analysis to acknowledge the information we have about CDNC retrievals and to work up from the assumption that meteorological variability is the leading order driver of cloud and precipitation variability- as opposed to putting it in as an afterthought.

The paper has some difficulty with grammar, but generally the intent of the authors is understandable and I am assuming writing will be tuned up by the ACP proofing staff.

Ln 36: Do the authors mean this paper, or a recent paper they have published?

Ln 47: Correlation and causation are not the same thing. Changes in cloud statistics covary with the proxies for aerosol used in this study.

Ln 51: It is unclear what this sentence means.

Ln 47: It is unclear what these numbers mean. What was the change in aerosols (AOD/CDNC?)

Ln 84: What is moisture supply? Is this the RH at some pressure level? I see this isn’t described until line 188.

Ln 134: I would say that it tends to be before heavy rain fall events, however, I find Fig. 2 a little confusing- where did LST 6-13 go? Where there never any rainfall events in that time range? That seems unlikely. If I linearly interpolate between the last point and the first point I can see that there a fair number of rainfall events occurring at 10:30 LST, although not the maximum of the PDF- so I would say ‘tends’ and not ‘always’.

Ln 167: What band is used to retrieve reff? It makes a big difference (Daniel P. Grosvenor et al., 2018). Are low (80%) cloud fractions removed? From looking ahead I infer that they are not. This is also important to deal with heterogeneity effects, which are critical since the authors are looking at precipitation (see restrictions applied in (Bennartz, Fan, Rausch, Leung, & Heidinger, 2011)).

Ln 217: As noted in the previous revision, it is unclear if CDNC can be used to stratify into clean and polluted cloud to learn anything about rain. This is because cloud top heterogeneity affects the retrieval of CDNC from MODIS (see Fig 16 of D. P. Grosvenor and Wood (2014)). For example, higher cloud top heterogeneity could go in step with higher rain rates and higher CDNC. The authors need to discuss this source of spurious covariability.

Ln 238: I think this section is compelling, especially combined with table 2, which shows that high AOD and CDNC tend to lead to the same change in sign for changes in precipitation characteristics. I do recommend that the authors look into the difference in log intensity since the distribution of intensity is clearly non-normal. Can a similar table be made for the AI from OMI and SO4/BC AOD? That would be nice to compare across.

Ln 302:
I think that this section looking at cloud changes sorted by CDNC could be removed. The retrievals of CDNC, LWP, CF, CER and COT are all dependent on each other and retrieval errors in one will affect all the others. I am not sure how much this analysis adds to the primary results of the paper, which is (in my reading) the covariability between precipitation statistics and CDNC/AOD. However, I understand that this analysis helps the authors support their hypothesized mechanism, so I think that if the retrievals in this section are heavily caveated in that none of the variables are truly independent retrievals then that will be acceptable.

Ln 377: In the last part of the section the authors discuss how moisture and CDNC covary in their sample. This means that most of the paper up till now may be looking at spurious covariability. We would expect that meteorology (in this case- moisture) to dominate driving cloud property variability. This also means that the results relating to rainfall are potentially just looking at times when it is more or less moist.

To clarify- if I believe that aerosol has no effect on rain rates, based on what the authors have just told me it seems perfectly reasonable to suspect that meteorology drives everything and that CCN-proxies are just covarying with moisture. Obviously this is an empirical study and they can’t fully rule out the possibility that CCN proxies are covarying with some unknown predictor, but the authors need to start out binning by CCN proxy and their meteorological parameter at the beginning instead of introducing it now.

Ln 416: Or estimated inversion strength and BC covary.

Ln 431: Or estimated inversion strength and SO4 covary- or SO4 is bright and leads to enhanced cloud fraction through misdetection.

Fig 8. Something looks off in this plot and Fig. 6 (top left). The CF doesn’t seem to have a very smooth distribution. Why are there so many completely overcast 1x1° days with no transition?

Bennartz, R., Fan, J., Rausch, J., Leung, L. R., & Heidinger, A. K. (2011). Pollution from China increases cloud droplet number, suppresses rain over the East China Sea. Geophysical Research Letters, 38(9), n/a-n/a. doi:10.1029/2011gl047235
Christensen, M. W., Neubauer, D., Poulsen, C., Thomas, G., McGarragh, G., Povey, A. C., . . . Grainger, R. G. (2017). Unveiling aerosol-cloud interactions Part 1: Cloud contamination in satellite products enhances the aerosol indirect forcing estimate. Atmos. Chem. Phys. Discuss., 2017, 1-21. doi:10.5194/acp-2017-450
Grosvenor, D. P., Sourdeval, O., Zuidema, P., Ackerman, A., Alexandrov, M. D., Bennartz, R., . . . Quaas, J. (2018). Remote Sensing of Droplet Number Concentration in Warm Clouds: A Review of the Current State of Knowledge and Perspectives. Reviews of Geophysics, 56(2), 409-453. doi:10.1029/2017rg000593
Grosvenor, D. P., & Wood, R. (2014). The effect of solar zenith angle on MODIS cloud optical and microphysical retrievals within marine liquid water clouds. Atmos. Chem. Phys., 14(14), 7291-7321. doi:10.5194/acp-14-7291-2014
Twohy, C. H., Coakley, J. A., & Tahnk, W. R. (2009). Effect of changes in relative humidity on aerosol scattering near clouds. Journal of Geophysical Research: Atmospheres, 114(D5), n/a-n/a. doi:10.1029/2008JD010991

Hide

RR by Anonymous Referee #2 (27 Mar 2019)

Suggestions for revision or reasons for rejection

I thank the authors for the work they have put into revising the paper, it is much improved over the previous version, although there are a number of things that I am still concerned about.

The linking of aerosol properties to the diurnal variation of aerosol remains a nice result to base this study around, but the correlation between cloud and aerosol properties shown in later parts of the work is subject to a number of meteorological covariations. Just controlling for the humidity has been shown to be insufficient to isolate an aerosol effect (Quaas and Boucher, 2012). With this in mind, Section 5 is useful in enumerating the cloud properties that are correlated to CDNC, but I am not sure it demonstrates a causal role of aerosol. I am not sure that it is vital to explain the rainfall results from the first part of the paper. Perhaps this section could be presented relationships between cloud and aerosol properties, rather than evidence of an aerosol impact.

Main points:
I like the use of the CDNC as an aerosol proxy, but I think that the authors need to do more to demonstrate that it is applicable in this case. It is not just a drop-in replacement for AOD and has a number of specific biases.

The CDNC calculation is only applicable in adiabatic clouds. In general, convective clouds are not adiabatic and a precipitating cloud is unlikely to be. This may not be a large factor in this study, but it should be considered and factored into the discussion/conclusions. There are a number of other potential biases that may affect this work. Grosvenor et al. (2018) is a good summary.

Additionally, is the CDNC calculation done using the 1 by 1 degree mean data? As a non-linear calculation, this can strongly impact the ``retrieved'' CDNC. The MODIS L3 product includes a cloud optical depth-effective radius joint histogram which can be used to better calculate the CDNC (e.g. Quaas et al., 2008; Grandey and Stier, 2010)

As the CDNC is calculated from the CER and COD, it is not clear to me that the CDNC-CER and CDNC-COD relationships are useful. I have similar concerns about the CDNC-LWP relationship, which is difficult to interpret even in low-level liquid clouds (e.g. Sato
et al, 2018; Gryspeerdt et al., 2018b). In this case, I think that Fig. 7 is just reproducing the assumptions used to calculate the CDNC and LWP (if the CER and CDNC are known, then the LWP is uniquely identified - at least at a pixel level).

Finally, is it clear what biases in the CDNC might be caused by the addition of thin overlying ice cloud? The authors are considering very complex situations where this may be an issue in a way that it is not for studies of liquid clouds.

Minor points:
L188 - What are the advantages of the AH here compared to a more common meteorological value such as specific humidity (easily available from ERA-Interim)?

L218 - Absorbing aerosol index is dependent on the altitude of the aerosol layer. What is assumed in this work?

L406 - How does decreasing the supersaturation reduce the strength of the freezing process? The supersaturation over ice is higher than over liquid. There are some studies which have noted an aerosol relationship to observations of mixed phase and ice cloud that might help explain this result (Chylek et al., 2006; Zhao et al., 2018; Gryspeerdt et al., 2018)

L446 - There have been many studies looking at the impact of BC on precipitation, perhaps they might be helpful in interpreting these results (e.g. Fan et al., 2008; O'Gorman et al., 2011). The role of BC is expected to change with altitude. Is it clear that the BC here is all in the boundary layer?

L462 - This might be due to a change in LWP/cloud depth with changing AOD. However, it is not clear if that change in cloud depth could be attributed to aerosols.

Figures - The text on may of the plots is very small.

References:
Boucher, O., and J. Quaas (2012), Water vapour affects both rain and aerosol optical depth, Nat. Geosci., 6(1), 4–5, doi:10.1038/ngeo1692.

Chylek, P., M. K. Dubey, U. Lohmann, V. Ramanathan, Y. J. Kaufman, G. Lesins, J. Hudson, G. Altmann, and S. Olsen (2006), Aerosol indirect effect over the Indian Ocean, Geophys. Res. Lett., 33(6), L06806, doi:10.1029/2005GL025397.

Fan, J., R. Zhang, W.-K. Tao, and K. I. Mohr (2008), Effects of aerosol optical properties on deep convective clouds and radiative forcing, J. Geophys. Res., 113(D8), doi:10.1029/2007JD009257.

Grandey, B. S., and P. Stier (2010), A critical look at spatial scale choices in satellite-based aerosol indirect effect studies, Atmos. Chem. Phys., 10(23), 11459–11470, doi:10.5194/acp-10-11459-2010.

Grosvenor, D. P. et al. (2018), Remote Sensing of Droplet Number Concentration in Warm Clouds: A Review of the Current State of Knowledge and Perspectives, Rev. Geophys., doi:10.1029/2017RG000593.

Gryspeerdt, E., T. Goren, O. Sourdeval, J. Quaas, J. Mülmenstädt, S. Dipu, C. Unglaub, A. Gettelman, and M. Christensen (2018a), Constraining the aerosol influence on cloud liquid water path, Atmos. Chem. Phys. Discuss., 1–25, doi:10.5194/acp-2018-885.

Gryspeerdt, E., O. Sourdeval, J. Quaas, J. Delanoë, M. Krämer, and P. Kühne (2018b), Ice crystal number concentration estimates from lidar–radar satellite remote sensing – Part 2: Controls on the ice crystal number concentration, Atmos. Chem. Phys, 18(19), 14351–14370, doi:10.5194/acp-18-14351-2018.

O'Gorman, P. A., R. P. Allan, M. P. Byrne, and M. Previdi (2011), Energetic Constraints on Precipitation Under Climate Change, Surv. Geophys., 111, doi:10.1007/s10712-011-9159-6.

Quaas, J., O. Boucher, N. Bellouin, and S. Kinne (2008), Satellite-based estimate of the direct and indirect aerosol climate forcing, J. Geophys. Res., 113, 05204, doi:10.1029/2007JD008962.

Sato, Y., D. Goto, T. Michibata, K. Suzuki, T. Takemura, H. Tomita, and T. Nakajima (2018), Aerosol effects on cloud water amounts were successfully simulated by a global cloud-system resolving model, Nat. Commun., 9(1), doi:10.1038/s41467-018-03379-6.

Zhao, B., Y. Gu, K.-N. Liou, Y. Wang, X. Liu, L. Huang, J. H. Jiang, and H. Su (2018), Type-Dependent Responses of Ice Cloud Properties to Aerosols From Satellite Retrievals, Geophys. Res. Lett., 45(7), 3297–3306, doi:10.1002/2018GL077261.

Hide

ED: Reconsider after major revisions (04 Apr 2019) by Philip Stier

AR by Siyuan Zhou on behalf of the Authors (15 May 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (19 Jun 2019) by Philip Stier

RR by Anonymous Referee #1 (20 Jun 2019)

RR by Anonymous Referee #2 (08 Jul 2019)

Suggestions for revision or reasons for rejection

The authors have done a significant amount of work improving the manuscript. The precipitation part of this paper is interesting and I think worthy of publication. However, some of my concerns remain, particularly in regards to the detection of causal relationships and the use of the CDNC.

Aerosol-cloud-precipitaiton relationships

I understand that it is difficult to control for meteorological covariations and systematic biases when performing these observational studies and I appreciate the extra section on the impact of humidity that the authors have included. Section 3.3 (and 4.2) is a useful addition to the paper, especially given that the paper is dealing with a difficult subject. However, it seems that the results from section 3.3 do not make it to the abstract and conclusions, which still give the impression that aerosols have been shown to cause these changes in cloud and precipitation properties. Where these results from section 3.3 are mentioned (e.g. in the abstract, L49-51), it is almost as an afterthought, failing to point out that the impact of moisture changes is the same as the impact of sulphate, which means that the aerosol effect cannot be isolated from the impact of humidity.

It has previously been shown that using reanalysis moisture variables cannot completely control for meteorological covariations between aerosol and precipitation (Boucher and Quaas, 2012). In addition, several of the relationships examined within section 4 are known to be affected by meteorological covariations: AOD-CF (Quaas et al., 2010; Grandey et al., 2013) and AOD-CTP (Gryspeerdt et al., 2014). Where these relationships are discussed in this paper (e.g. L521 onwards), they appear to be used as evidence for an aerosol effect on cloud, despite these known issues.

This is not a large change to the paper, mostly just in the abstract and conclusions (e.g. changing sentences such as L547 - '... the different roles of [aerosol] in modifying the diurnal ...' to '... the different relationships of [aerosol] to the diurnal ...'). I also would suggest that the discussion in section 5, particularly around the relationships between AOD and cloud properties, is modified to reflect the previous results.

CDNC retrieval

My fourth point in the previous review mentioned that it is not clear that the CDNC-CER and CDNC-COT relationships are useful, as they are not independent. The response suggests that these have been removed, but investigation of the CDNC-CER relationship is still present and compared with the AOD-CER relationship (e.g. L360, L521, L532). There may be a justification for including the CDNC-CER relationship (I am not clear that it has to be removed), but could the authors clarify whether they are removing it or not?

It is good to include the reference to the Grosvenor paper and a discussion of the uncertainties. However, I am not sure the impact of the uncertainties on the results is addressed. In particular, the CDNC retrieval is for adiabatic clouds, and it is not clear that it can be applied to convective clouds. Clouds that are raining are by definition non-adiabatic. How does this affect the results in this work? Might it reduce the significance of the results?

Section 4 starts referring to the CDNC as CCN. Is this appropriate? CDNC depends on CCN and updraught, presumably there is a strong variation in updraught between convective clouds which weakens the link between CDNC and CCN?

The average values for the retrieved CDNC (L391) are very high (more than 2000 cm-3). For an optical depth of 100, this would mean that all of the effective radius retrievals are smaller than around 6um. Is this correct?

Following my third point in the previous review, it is not clear how the CDNC is calculated. Are the 1 by 1 degree mean values of CER and COT used? If so, this should be noted, as it may cause an underestimate in the CDNC compared to other studies (e.g Quaas et al., 2008; Grandey et al., 2010) which use the level 2 data or the L3 joint histogram 'Cloud_Optical_Thickness_Liquid_JHisto_vs_Eff_Radius' to calculate the CDNC.

Minor points

L442 - Ackerman (2000) is a usual reference for the semi-direct effect.

L470 - 'absorbing aerosol effect' -> 'absorbing aerosol results'? I am not sure that an effect is specified (to me, it would require some kind of process being suggested)

Section 4.2 - repeated references to section 5.1, when presumably 4.1 is meant?

L500 - Relative changes are still subject to systematic biases, which are the more important source of error in a study like this.

L510 - I understand that these datasests are the best available. However, if they are insufficient to demonstrate a causal effect of aerosol on cloud or precipitation properties, then they are insufficient. In that case, this must be noted.

L520 - Or sampling differences between the AOD and CDNC?

L539 - What happened to section 6?

Fig. 8 - The units for the x-axis are "%" only

There are still some spelling/grammar issues (e.g. black carbons - L44 and in the new sections), but they don't appear to make to too difficult to understand the paper and could be corrected in proof reading.

References

Ackerman, A. S. (2000), Reduction of Tropical Cloudiness by Soot, Science, 288(5468), 10421047, doi:10.1126/science.288.5468.1042.

Boucher, O., and J. Quaas (2012), Water vapour affects both rain and aerosol optical depth, Nat. Geosci., 6(1), 45, doi:10.1038/ngeo1692.

Grandey, B. S., and P. Stier (2010), A critical look at spatial scale choices in satellite-based aerosol indirect effect studies, Atmos. Chem. Phys., 10(23), 1145911470, doi:10.5194/acp-10-11459-2010.

Grandey, B. S., P. Stier, and T. M. Wagner (2013), Investigating relationships between aerosol optical depth and cloud fraction using satellite, aerosol reanalysis and general circulation model data, Atmos. Chem. Phys., 13(6), 31773184, doi:10.5194/acp-13-3177-2013.

Gryspeerdt, E., P. Stier, and B. S. Grandey (2014), Cloud fraction mediates the aerosol optical depth-cloud top height relationship, Geophys. Res. Lett., 41, 36223627, doi:10.1002/2014GL059524.

Quaas, J., O. Boucher, N. Bellouin, and S. Kinne (2008), Satellite-based estimate of the direct and indirect aerosol climate forcing, J. Geophys. Res., 113, 05204, doi:10.1029/2007JD008962.

Quaas, J., B. Stevens, P. Stier, and U. Lohmann (2010), Interpreting the cloud cover aerosol optical depth relationship found in satellite data using a general circulation model, Atmos. Chem. Phys., 10(13), 61296135, doi:10.5194/acp-10-6129-2010.

Rosenfeld, D. et al. (2016), Satellite retrieval of cloud condensation nuclei concentrations by using clouds as CCN chambers., P. Natl. Acad. Sci. USA, 113(21), 582834, doi:10.1073/pnas.1514044113.

Hide

ED: Publish subject to minor revisions (review by editor) (12 Jul 2019) by Philip Stier

AR by Siyuan Zhou on behalf of the Authors (21 Jul 2019) Author's response Manuscript

ED: Reconsider after major revisions (30 Aug 2019) by Philip Stier

Dear Authors

Thank you for providing the revised manuscript “An observational study of the effects of aerosols on diurnal variation of heavy rainfall and the concurrent cloud changes over Beijing-Tianjin-Hebei” in response to the earlier requests for revisions.

Previous marks given by the reviewers were not sufficient to reach the standard required for publication in ACP. Having re-assessed the reviewers’ comments and conducted my own assessment, I have now concluded that a significant number of outstanding issues need to be addressed before reaching the required standard, even after the performed minor and major revision iteration. In addition, your latest revision corrected a key parameter in the analysis, retrieved cloud drop number concentration, from 2000cm-3 to a new value of 63cm-3, which seems unrealistically low for such a polluted region (c.f. Grosvenor et al., 2018).

In summary, I have to conclude that this manuscript has not yet reached the standard required for publication in ACP and will reconsider the manuscript for a final time after major revisions although you may also want to reconsider a resubmission after completion of the requested changes.

Please consider each of the items below in your response.

Kind regards,

Philip Stier
Co-Editor
Atmospheric Chemistry and Physics

Major issues:

Response P2:
A major issue of the manuscript remains that it is predicated on the assumption that the derived aerosol-cloud relationships are causal. Some caveats have been inserted in response to critical reviewer comments, e.g. related to relative humidity, but the overall wording of the manuscript remains fairly one-track and I agree with the reviewer that this appears to be more an afterthought. In a revised version of the manuscript, it is critical to separate the description of the derived relationships (e.g. for higher AOD the onset of precipitation occurs…) from a causal attribution (e.g. aerosols delay the onset of precipitation). Causal attribution needs to be presented and discussed in the context of confounding variables. Potential limitations of all causal attributions should be explicitly discussed. This critical analysis needs to be pulled through to the abstract and conclusions.

Response P3:
The use of CDNC as surrogate for CCN has severe limitations and it is not sufficient to simply declare the issue of updrafts out of scope of this study. Aerosol activation is controlled by CCN and updraft velocity. The region of interest is highly polluted so this is very likely to be an updraft limited regime (c.f. Reutter et al., ACP, 2009). Assuming this is the case, any presented correlation with CDNC with cloud properties could be interpreted as correlation of updraft velocity with the respective cloud property or precipitation – which are of course be expected, even under constant CCN. There are good reasons to believe that this could significantly contribute to or even dominate your presented correlations. If this analysis is retained, its reliability needs to be demonstrated. This cannot be ignored in the analysis and discussion.

Response P3:
You have now corrected your retrieved CDNC values by more than an order of magnitude from 2000cm-3 to a new value of 63cm-3, which seems unrealistically low for such a polluted region (c.f. Grosvenor et al., 2018). It appears neither of these values have been evaluated or critically analysed. This needs to be done thoroughly for this data to be used in the manuscript.
Given the significant structural uncertainties involved it is not appropriate to conclude on upper or lower bounds of the total effect from a single modelling study. Even if structural errors were negligible there exist a large number of parametric uncertainties that have not been sampled. The only dimension you sample is the perturbation strength so this need so be clear. However, given the extreme range of values chosen, these perturbations are likely to act as on/off switch for some processes, such as autoconversion. The implications should be discussed.

Figures & Tables
All figures and tables need to be fully self-explaining and captions need to state the used data sources.

Hide

AR by Siyuan Zhou on behalf of the Authors (10 Oct 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (26 Jan 2020) by Philip Stier

RR by Anonymous Referee #2 (03 Feb 2020)

Suggestions for revision or reasons for rejection

I thank the authors for the changes they have made. The conclusions
are better at separating out the observed relationships from the
hypothesised aerosol effects.

The discussion section could be moved earlier in the paper (before the
conclusions section) - this would be a more usual structure and would
help give the conclusions more prominence.

Following my previous comment on calculating the CDNC, the gridbox
mean values are not really suitable for calculating the CDNC as the
calculation is highly non-linear. Doing the calculation using the
joint histogram does not require that much more work, but will produce
better results. This may not change the results much, but it might be
a good idea for future studies at the very least. I have attached a plot of the mean CDNC in March-April-May over the study region, calculated from the MODIS L3 histograms for the period 2006-2010 (the data I had easily available). The plot on the left shows the mean CDNC across the nearby region.

The english in the paper could still use some improvement, although
the meaning is usually clear. I have not gone through the paper again
to highlight all the changes, but they seem to be primarily in the new
sections.

Other comments:
L307 - The SH is defined correctly earlier as the specific humidity,
rather than the ``water vapor content''

L434 - The cases of heavy rainfall using CDNC seem more extreme - then
at L440 - the change in rainfall intensity is not significant. Please
clarify this

L639 - 'accumulate more water in the cloud' - Is this a reference to
some kind of Albrecht (1989) type effect? If so, it might be good to
state it (although an earlier start to precipitation does not match
this). If not, it would also be good to state it (and explain what it
is).

L654(and associated section) - CDNC is not CCN. As the editor has
pointed out, in an updraft limited regime, variations in CDNC come
primarily from changes in updraft. Just stating that you are using
CDNC to represent CCN does not make CDNC actually represent CCN.

L694 - Following from my comments in previous reviews - The
relationship is not completely meaningless, just difficult to
interpret. Also, this sentence does not make much sense to me - you
state that this comparison cannot be done, then use it to infer the
role of aerosols. I would suggest you leave this sentence out, you
have plenty of other results without relying on this.

L853 - It is not clear the large scale updraft is a good measure of
the in-cloud updraft relevant for droplet formation.

Referee Report: PDF

Hide

ED: Publish subject to minor revisions (review by editor) (10 Feb 2020) by Philip Stier

AR by Siyuan Zhou on behalf of the Authors (19 Feb 2020) Author's response Manuscript

ED: Publish as is (23 Mar 2020) by Philip Stier

Short summary

Aerosol–cloud–precipitation interaction is a challenging problem in regional climate. Our study contrasted the observed diurnal variation of heavy rainfall and associated clouds over Beijing–Tianjin–Hebei between clean and polluted days during the 2002–2012 summers. We found the heavy rainfall under pollution has earlier start time, earlier peak time and longer duration, and further found the absorbing aerosols and scattering aerosols play different roles in the heavy rainfall diurnal variation.