Droplet collection efficiencies estimated from satellite retrievals constrain effective radiative forcing of aerosol-cloud interactions

Beall, Charlotte M.; Ma, Po-Lun; Christensen, Matthew W.; Mülmenstädt, Johannes; Varble, Adam; Suzuki, Kentaroh; Michibata, Takuro

doi:https://doi.org/10.5194/egusphere-2023-2161

Preprints

https://doi.org/10.5194/egusphere-2023-2161

Preprints

06 Nov 2023

| 06 Nov 2023

Droplet collection efficiencies estimated from satellite retrievals constrain effective radiative forcing of aerosol-cloud interactions

Charlotte M. Beall, Po-Lun Ma, Matthew W. Christensen, Johannes Mülmenstädt, Adam Varble, Kentaroh Suzuki, and Takuro Michibata

Abstract. Process-oriented observational constraints for the anthropogenic effective radiative forcing due to aerosol-cloud-interactions (ERFaci) are highly desirable because the large uncertainty associated with ERFaci poses a significant challenge to climate prediction. The satellite-based Contoured Frequency by Optical Depth Diagrams (CFODD) analysis was previously proposed to support evaluation of model representation of cloud liquid to rain conversion processes because the slope of a CFODD, generated from joint MODerate Resolution Imaging Spectroradiometer (MODIS)-CloudSat cloud retrievals, provides an estimate of cloud droplet collection efficiency in single-layer warm liquid clouds (SLWCs). Here we present an updated CFODD analysis as an observational constraint for the ERFaci due to warm rain processes and apply it to the U.S. Department of Energy’s Energy Exascale Earth System Model version 2 (E3SMv2). Updates to the CFODD analysis include multiple changes to the SLWC detection algorithm for better consistency between MODIS-CloudSat observations and the satellite simulators, as well as the estimation of CFODD slopes using Random Sample Consensus robust linear regression. A series of sensitivity experiments shows that E3SMv2 droplet collection efficiencies and ERFaci are highly sensitive to the treatment of autoconversion, the rate of mass transfer from cloud liquid to rain, yielding a strong correlation between the CFODD slope and the shortwave component of ERFaci (Pearson’s R = -0.91). We estimate the shortwave component of ERFaci (ERFaci_SW), constrained by MODIS-CloudSat, by calculating the intercept of the linear association between E3SMv2 ERFaci_SW and the CFODD slopes, using the MODIS-CloudSat CFODD slope as a reference. When E3SMv2’s droplet collection efficiency is constrained to agree with the A-Train retrievals, ERFaci_SW is reduced by 13 % in magnitude, indicating that correcting bias in the ERFaci_SW due to autoconversion would bring E3SMv2’s total ERFaci (-1.50 W m^-2) into better agreement with the IPCC AR6 ‘very likely’ range for ERFaci (-1.0 ± 0.7 W m^-2). This study provides a new process-oriented observational constraint for ERFaci due to warm rain processes to reduce the uncertainty of climate predictions.

Received: 20 Sep 2023 – Discussion started: 06 Nov 2023

Download & links

Preprint (PDF, 1050 KB)

Supplement (2550 KB)

Download & links

Charlotte M. Beall, Po-Lun Ma, Matthew W. Christensen, Johannes Mülmenstädt, Adam Varble, Kentaroh Suzuki, and Takuro Michibata

Status: closed

RC1:
'Comment on egusphere-2023-2161', Anonymous Referee #1, 17 Nov 2023

This paper uses the CFODD (Contoured Frequency by Optical Depth Diagrams) to evaluate the ERF_ACI (effective radiative forcing due to aerosol cloud interactions) in a modern climate model. 13 perturbed parameter experiments are performed to evaluate the model ERF_ACI using the emergent constraints paradigm. The paper is timely and appears to have some interesting results. However, at the moment, there is a fundamental methodological issue that I cannot understand and therefore I cannot provide a full review until the methodology is clarified. Specifically, note Figure 3. There is something basic that I’m missing. Why are the ERF_ACI_SW values so small? I would expect negative values near 1 Wm^-2. Also why does ERF_ACI_SW differ between figures 3, and S7. Some of the ERF_ACI_SW are even positive in S7! I would think that you have 13 experiments so you should have 13 fixed values of ERF_ACI_SW. It is almost like the total ERF_ACI has been decomposed into components in some way, however I can find no description of this in the text. Until the methodology is further explained I can not complete this review. In addition I have a few specific comments below.

Additional specific comments:
Abstract: Abstract should be less than 250 words according to the ACP guidelines: https://www.atmospheric-chemistry-and-physics.net/policies/guidelines_for_authors.html
The following paper is very similar in scope and methods and should be cited in the introduction: Takahashi, H., A. Bodas-Salcedo, and G. Stephens, 2021: Warm Cloud Evolution, Precipitation, and Their Weak Linkage in HadGEM3: New Process-Level Diagnostics Using A-Train Observations. J. Atmos. Sci., 78, 2075–2087, https://doi.org/10.1175/JAS-D-20-0321.1.
Line 125: is there a reference for RANSAC? I think this is a python package but will not be familiar to the average reader.
Line 221: MODIS provides 3 different effective radius values from the 1.6, 2.1, and 3.7 micron channels. 2.1 micron is the ‘standard’ product. Which one are you using?
Line 260 and Figure 1: Is a SLWC defined as a cloud with optical depth > 0.3 and radar reflectivity > -30 dBZ? If this is correct can you mention this in the text. In this case I would assume that the identification of detectable cloud is nearly entirely limited by the radar reflectivity threshold whereas the optical depth threshold probably was the dominant limiting factor in the earlier study.
Figure 3: You should include an uncertainty range around the linear fit as gray shading. You should quote that uncertainty in the text.
Line 350: This idea of adjusting the estimate of the ERF_ACI based on this linear fit seems bizarre to me. The CNTRL experiment already agrees with the observed CFODD slope almost exactly. So to me the story of this paper is that this version of the model agrees very well with this particular metric which suggests that the auto conversion process is been simulated with relatively good fidelity.

Citation: https://doi.org/10.5194/egusphere-2023-2161-RC1
- AC1:
  'Reply on RC1', Charlotte Beall, 27 Nov 2023
  
  We thank the reviewer for bringing attention to the clarification needed for the ERFaci values in Fig. 3 and Fig. S7 and for the helpful specific comments. The ERFaci_SW values in Fig. 3 are smaller in magnitude than an expected value for total ERFaci because they reflect the ERFaci_SW calculated for only the cloud samples featured in the CFODDs (i.e., the SLWCs with 5 micron ≤ R_e < 18 micron). These SLWC samples represent ~8 % of the total simulated cloud fraction according to the MODIS simulator. The reason for the different ERFaci values between Fig. 3 and the subplots in Fig. S7 is that ERFaci values in each plot reflects the subset of SLWC samples in the MODIS R_e bin indicated by the plot title. If you sum the ERFaci_SW values for each simulation in Fig. S7 (a) (5 micron ≤ R_e < 12) and (b) (12 micron ≤ R_e < 18), the resulting values are equivalent to the values shown in Fig. 3 (5 micron ≤ R_e < 18). The positive ERFaci_SWvalues in the medium and large R_e bins partially offset the negative ERFaci_SW values in the smallest R_e bin, resulting in an ERFaci_SW that is smaller in magnitude than for the 5 micron ≤ R_e < 12 bin alone. This may be a result of compensating biases in E3SMv2 where the positive ERFaci_SW predicted for medium and large R_e bins is offset by simulating too many SLWCs in the small R_e bin. The point that calculation of ERFaci is made for SLWCs within the specified MODIS R_e range is mentioned on L255, but I will specify that ERFaci is calculated for SLWCs in the y-axis labels (e.g., ERFaci_{SW_SLWCs}), the figure captions and throughout the text.
  Responses to the reviewer’s specific comments are below:
  RC1: “Abstract: Abstract should be less than 250 words according to the ACP guidelines”
  Thank you for this reminder. The abstract will be cut to adhere to the word limit and modified to feature the linear fit uncertainty (e.g., “… is reduced by 13% ± x”), to the reviewer’s points about the uncertainty and agreement below.
  RC1: “The following paper is very similar in scope and methods and should be cited in the introduction: Takahashi, H., A. Bodas-Salcedo, and G. Stephens, 2021: Warm Cloud Evolution, Precipitation, and Their Weak Linkage in HadGEM3: New Process-Level Diagnostics Using A-Train Observations. J. Atmos. Sci., 78, 2075–2087, https://doi.org/10.1175/JAS-D-20-0321.1.”
  Agreed that this reference is highly relevant to our manuscript. This reference will be added to the introduction and in the results on sampling frequency MODIS R_e bin in Sect. 3.
  RC1: “Line 125: is there a reference for RANSAC? I think this is a python package but will not be familiar to the average reader.”
  Good point. The following reference for RANSAC will be added to L125:
  Fischler, M. A. and R. C. Bolles (1981). "Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography." Commun. ACM 24(6): 381–395.
  RC1: “Line 221: MODIS provides 3 different effective radius values from the 1.6, 2.1, and 3.7 micron channels. 2.1 micron is the ‘standard’ product. Which one are you using?”
  Yes, we are using the standard ‘Cloud_Effective Radius’ and ‘Cloud_Effective Radius_PCL’ products retrieved from the 2.1 micron channel. We will add this detail to L233.
  RC1: “Line 260 and Figure 1: Is a SLWC defined as a cloud with optical depth > 0.3 and radar reflectivity > -30 dBZ? If this is correct can you mention this in the text. In this case I would assume that the identification of detectable cloud is nearly entirely limited by the radar reflectivity threshold whereas the optical depth threshold probably was the dominant limiting factor in the earlier study.”
  Yes, this is correct. We will add the following to L260: “The criteria for SLWC sampling in A-Train observations and E3SMv2-COSP are provided in the Supplement Table 1 and include CloudSat reflectivity ≥ -30 dBZ, MODIS liquid COT > 0.3, and cloud top temperature ≥ 273 K.”
  RC1: “Figure 3: You should include an uncertainty range around the linear fit as gray shading. You should quote that uncertainty in the text.”
  Agreed. An uncertainty estimate for the linear fit will be added to Fig. 3 and the results of the comparison will likely support a discussion of “agreement”, which the reviewer points out in the following comment.
  RC1: “Line 350: This idea of adjusting the estimate of the ERF_ACI based on this linear fit seems bizarre to me. The CNTRL experiment already agrees with the observed CFODD slope almost exactly. So to me the story of this paper is that this version of the model agrees very well with this particular metric which suggests that the auto conversion process is been simulated with relatively good fidelity.”
  Agreed that the CNTL experiment CFODD slope agrees with the observed CFODD slope within uncertainty, and the point about agreement with what the linear fit predicts for ERFaci_SW will be modified with the fit’s uncertainty range, as suggested in the comment above. Comparing the CNTL-predicted ERFaci with the fit uncertainty range may show agreement within estimated uncertainty, in which case the text will be modified to reflect this.
  
  Citation: https://doi.org/10.5194/egusphere-2023-2161-AC1
  - RC3: 'Reply on AC1', Anonymous Referee #1, 03 Dec 2023
    
    Thank you for the authors explanation of the methodology. I look forward to seeing a revised manuscript that adequately address my previous comments. I have a few more major comments here:
    Minor Comment:
    I'm a little uncomfortable with the title phrasing of 'droplet collection efficiency'. The paper looks at a proxy for efficiency but doesn't quantify the collection efficiency in a direct way. Consider a more accurate title.
    
    Major comments:
    Abstract line 26: ‘When E3SMv2’s droplet 27 collection efficiency is constrained to agree with the A-Train retrievals, ERFaciSW is reduced by 13% in magnitude’ Consistent with my previous statement I don’t think this is the correct takeaway. I would state the model droplet collection efficiency is consistent with the observations lending credence to the estimate of the ERFaciSW in the model.
    While I appreciate the sentiment of using a process-oriented constraint instead of state variables to evaluate the representation of ERFaci I have a lot of concern of relying on a single metric to make any sweeping conclusions about the fidelity of the simulations. For example, while the current version of this model agrees with this particular metric it is telling (that like many models) it does not reproduce the historical temperature trend (Golaz et al., 2022). To address my concerns, I think the paper would benefit from some analysis of some state variables. At the very least it would be nice to see an analysis of the TOA reflected shortwave biases relative to EBAF for the 13 simulations. Furthermore, since you’ve run the MODIS simulator it would be easy to evaluate the CF and LWP as well. Finally, some discussion of these concerns, limitations of the study, and outlook for future research should also be added in the summary.
    
    Citation: https://doi.org/10.5194/egusphere-2023-2161-RC3
RC2: 'Comment on egusphere-2023-2161', Anonymous Referee #2, 01 Dec 2023

The paper discusses the use of satellite retrievals to evaluate warm rain processes and constrain the ERFaci in the Energy Exascale Earth System Model version 2 (E3SMv2). It applies an updated Contoured Frequency by Optical Depth Diagrams (CFODD) analysis, in both observations and simulations, to provide observational constraints for droplet collection efficiency. The autoconversion parameterization is modified in a series of sensitivity experiments to evaluate the simulated CFODD with respect to the observed one. This allows to estimate a more accurate simulated ERFaci that is driven by warm rain processes.

Main comments:
The manuscript provides an adequate description of the model and the satellite data that are used, including their limitations. I appreciate the authors' approach using CFODD to constrain the simulations, it is elegant and provide interesting results. Nevertheless, the CTRL simulation seems to agree well with the observations. This is a good result for the model and the control coefficients. Shouldn’t this be the main conclusion of the study? Is there a physical meaning for choosing other coefficients, in this model or others? it would be interesting for discussion.
The paper would benefit from improved organization and clearer, more concise writing. Specifically, Section 4 (with the note that there is no Section 5 between sections 4 and 6) is lengthy and could be subdivided into subsections. The latter part of Section 4 could potentially be formed into a subsection that addresses the limitations of the study. Additionally, the introduction section should concentrate on past studies and their relevance to the current research. Paragraphs that discuss the results and the methodology of the present study could be removed from the introduction. Furthermore, the abstract appears overly detailed. Consider revising to make it more concise while still providing essential information.
The manuscript frequently directs the reader to key figures in the supplement. It may be advantageous to consider incorporating some of these figures into the main manuscript to enhance accessibility.
Including a brief description of the CFODD method would be beneficial for readers who may be less familiar with such diagrams. This addition could provide clarity and context for those who may not have an in-depth understanding of the CFODD method.
One of the modifications to CFODD applied by the authors is reducing the cloud optical thickness (COT) to 0.3. As the authors mention, this adjustment has a substantial impact on cloud occurrence. It is worth considering whether this modification could also influence the slope of the CFODD by altering the weight of the bins with low COT (does the fitting account for data density?). Clarification is also needed on whether COT represents the in-cloud mean or the grid mean values. Specifying the resolution of the data is missing and it is unclear how products with different resolutions are combined.
Additional specific comments:
L266: There is a reference typo that needs correction.
L309: The statement about how optically thin layers can have high reflectivities near the cloud top is unclear. It would be helpful to clarify whether this refers to the small optical thickness of thin layers at the top of deeper clouds. The entire sentence needs clarification for better understanding.
L310: The term "strange features" needs clarification. Specify and elaborate on what exactly is being referred to.
L315: It would be beneficial to explicitly mention that you are describing Figure 2 in this section for better guidance to the reader.
L337: Clarify the meaning of "ERFaciSW strengthens." Provide a physical explanation, particularly in relation to its impact on cloud cover and albedo. This clarification will enhance the understanding of the reader.
L348: It is recommended to reword this sentence and provide a clear explanation that the observed slopes are utilized to constrain the simulated ERFaci. This sentence essentially describes the core of the study.
L350: Suggest marking -0.067 on the Figure for clarity and reference.
L351: Clarify the meaning of the total ERFaci of -1.5 in the context of the discussion. Mention again that it is the difference between the simulations (as stated in Line 249). Address why this value cannot be deduced directly from Figure 3.
L352: Specify whether "correcting" means changing the coefficients in the autoconversion parameterization.
L390: Why "however"? The linear relationship applies to both medium-large and small Re.
L460: Elaborate on the statement that "simulated reflectivity profiles near the cloud top are decoupled from the cloud below." Provide a more detailed explanation of what is meant by "decoupled." This is the only instance where "decoupled" is mentioned in the manuscript.

Citation: https://doi.org/10.5194/egusphere-2023-2161-RC2

Status: closed

RC1:
'Comment on egusphere-2023-2161', Anonymous Referee #1, 17 Nov 2023

This paper uses the CFODD (Contoured Frequency by Optical Depth Diagrams) to evaluate the ERF_ACI (effective radiative forcing due to aerosol cloud interactions) in a modern climate model. 13 perturbed parameter experiments are performed to evaluate the model ERF_ACI using the emergent constraints paradigm. The paper is timely and appears to have some interesting results. However, at the moment, there is a fundamental methodological issue that I cannot understand and therefore I cannot provide a full review until the methodology is clarified. Specifically, note Figure 3. There is something basic that I’m missing. Why are the ERF_ACI_SW values so small? I would expect negative values near 1 Wm^-2. Also why does ERF_ACI_SW differ between figures 3, and S7. Some of the ERF_ACI_SW are even positive in S7! I would think that you have 13 experiments so you should have 13 fixed values of ERF_ACI_SW. It is almost like the total ERF_ACI has been decomposed into components in some way, however I can find no description of this in the text. Until the methodology is further explained I can not complete this review. In addition I have a few specific comments below.

Additional specific comments:
Abstract: Abstract should be less than 250 words according to the ACP guidelines: https://www.atmospheric-chemistry-and-physics.net/policies/guidelines_for_authors.html
The following paper is very similar in scope and methods and should be cited in the introduction: Takahashi, H., A. Bodas-Salcedo, and G. Stephens, 2021: Warm Cloud Evolution, Precipitation, and Their Weak Linkage in HadGEM3: New Process-Level Diagnostics Using A-Train Observations. J. Atmos. Sci., 78, 2075–2087, https://doi.org/10.1175/JAS-D-20-0321.1.
Line 125: is there a reference for RANSAC? I think this is a python package but will not be familiar to the average reader.
Line 221: MODIS provides 3 different effective radius values from the 1.6, 2.1, and 3.7 micron channels. 2.1 micron is the ‘standard’ product. Which one are you using?
Line 260 and Figure 1: Is a SLWC defined as a cloud with optical depth > 0.3 and radar reflectivity > -30 dBZ? If this is correct can you mention this in the text. In this case I would assume that the identification of detectable cloud is nearly entirely limited by the radar reflectivity threshold whereas the optical depth threshold probably was the dominant limiting factor in the earlier study.
Figure 3: You should include an uncertainty range around the linear fit as gray shading. You should quote that uncertainty in the text.
Line 350: This idea of adjusting the estimate of the ERF_ACI based on this linear fit seems bizarre to me. The CNTRL experiment already agrees with the observed CFODD slope almost exactly. So to me the story of this paper is that this version of the model agrees very well with this particular metric which suggests that the auto conversion process is been simulated with relatively good fidelity.

Citation: https://doi.org/10.5194/egusphere-2023-2161-RC1
- AC1:
  'Reply on RC1', Charlotte Beall, 27 Nov 2023
  
  We thank the reviewer for bringing attention to the clarification needed for the ERFaci values in Fig. 3 and Fig. S7 and for the helpful specific comments. The ERFaci_SW values in Fig. 3 are smaller in magnitude than an expected value for total ERFaci because they reflect the ERFaci_SW calculated for only the cloud samples featured in the CFODDs (i.e., the SLWCs with 5 micron ≤ R_e < 18 micron). These SLWC samples represent ~8 % of the total simulated cloud fraction according to the MODIS simulator. The reason for the different ERFaci values between Fig. 3 and the subplots in Fig. S7 is that ERFaci values in each plot reflects the subset of SLWC samples in the MODIS R_e bin indicated by the plot title. If you sum the ERFaci_SW values for each simulation in Fig. S7 (a) (5 micron ≤ R_e < 12) and (b) (12 micron ≤ R_e < 18), the resulting values are equivalent to the values shown in Fig. 3 (5 micron ≤ R_e < 18). The positive ERFaci_SWvalues in the medium and large R_e bins partially offset the negative ERFaci_SW values in the smallest R_e bin, resulting in an ERFaci_SW that is smaller in magnitude than for the 5 micron ≤ R_e < 12 bin alone. This may be a result of compensating biases in E3SMv2 where the positive ERFaci_SW predicted for medium and large R_e bins is offset by simulating too many SLWCs in the small R_e bin. The point that calculation of ERFaci is made for SLWCs within the specified MODIS R_e range is mentioned on L255, but I will specify that ERFaci is calculated for SLWCs in the y-axis labels (e.g., ERFaci_{SW_SLWCs}), the figure captions and throughout the text.
  Responses to the reviewer’s specific comments are below:
  RC1: “Abstract: Abstract should be less than 250 words according to the ACP guidelines”
  Thank you for this reminder. The abstract will be cut to adhere to the word limit and modified to feature the linear fit uncertainty (e.g., “… is reduced by 13% ± x”), to the reviewer’s points about the uncertainty and agreement below.
  RC1: “The following paper is very similar in scope and methods and should be cited in the introduction: Takahashi, H., A. Bodas-Salcedo, and G. Stephens, 2021: Warm Cloud Evolution, Precipitation, and Their Weak Linkage in HadGEM3: New Process-Level Diagnostics Using A-Train Observations. J. Atmos. Sci., 78, 2075–2087, https://doi.org/10.1175/JAS-D-20-0321.1.”
  Agreed that this reference is highly relevant to our manuscript. This reference will be added to the introduction and in the results on sampling frequency MODIS R_e bin in Sect. 3.
  RC1: “Line 125: is there a reference for RANSAC? I think this is a python package but will not be familiar to the average reader.”
  Good point. The following reference for RANSAC will be added to L125:
  Fischler, M. A. and R. C. Bolles (1981). "Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography." Commun. ACM 24(6): 381–395.
  RC1: “Line 221: MODIS provides 3 different effective radius values from the 1.6, 2.1, and 3.7 micron channels. 2.1 micron is the ‘standard’ product. Which one are you using?”
  Yes, we are using the standard ‘Cloud_Effective Radius’ and ‘Cloud_Effective Radius_PCL’ products retrieved from the 2.1 micron channel. We will add this detail to L233.
  RC1: “Line 260 and Figure 1: Is a SLWC defined as a cloud with optical depth > 0.3 and radar reflectivity > -30 dBZ? If this is correct can you mention this in the text. In this case I would assume that the identification of detectable cloud is nearly entirely limited by the radar reflectivity threshold whereas the optical depth threshold probably was the dominant limiting factor in the earlier study.”
  Yes, this is correct. We will add the following to L260: “The criteria for SLWC sampling in A-Train observations and E3SMv2-COSP are provided in the Supplement Table 1 and include CloudSat reflectivity ≥ -30 dBZ, MODIS liquid COT > 0.3, and cloud top temperature ≥ 273 K.”
  RC1: “Figure 3: You should include an uncertainty range around the linear fit as gray shading. You should quote that uncertainty in the text.”
  Agreed. An uncertainty estimate for the linear fit will be added to Fig. 3 and the results of the comparison will likely support a discussion of “agreement”, which the reviewer points out in the following comment.
  RC1: “Line 350: This idea of adjusting the estimate of the ERF_ACI based on this linear fit seems bizarre to me. The CNTRL experiment already agrees with the observed CFODD slope almost exactly. So to me the story of this paper is that this version of the model agrees very well with this particular metric which suggests that the auto conversion process is been simulated with relatively good fidelity.”
  Agreed that the CNTL experiment CFODD slope agrees with the observed CFODD slope within uncertainty, and the point about agreement with what the linear fit predicts for ERFaci_SW will be modified with the fit’s uncertainty range, as suggested in the comment above. Comparing the CNTL-predicted ERFaci with the fit uncertainty range may show agreement within estimated uncertainty, in which case the text will be modified to reflect this.
  
  Citation: https://doi.org/10.5194/egusphere-2023-2161-AC1
  - RC3: 'Reply on AC1', Anonymous Referee #1, 03 Dec 2023
    
    Thank you for the authors explanation of the methodology. I look forward to seeing a revised manuscript that adequately address my previous comments. I have a few more major comments here:
    Minor Comment:
    I'm a little uncomfortable with the title phrasing of 'droplet collection efficiency'. The paper looks at a proxy for efficiency but doesn't quantify the collection efficiency in a direct way. Consider a more accurate title.
    
    Major comments:
    Abstract line 26: ‘When E3SMv2’s droplet 27 collection efficiency is constrained to agree with the A-Train retrievals, ERFaciSW is reduced by 13% in magnitude’ Consistent with my previous statement I don’t think this is the correct takeaway. I would state the model droplet collection efficiency is consistent with the observations lending credence to the estimate of the ERFaciSW in the model.
    While I appreciate the sentiment of using a process-oriented constraint instead of state variables to evaluate the representation of ERFaci I have a lot of concern of relying on a single metric to make any sweeping conclusions about the fidelity of the simulations. For example, while the current version of this model agrees with this particular metric it is telling (that like many models) it does not reproduce the historical temperature trend (Golaz et al., 2022). To address my concerns, I think the paper would benefit from some analysis of some state variables. At the very least it would be nice to see an analysis of the TOA reflected shortwave biases relative to EBAF for the 13 simulations. Furthermore, since you’ve run the MODIS simulator it would be easy to evaluate the CF and LWP as well. Finally, some discussion of these concerns, limitations of the study, and outlook for future research should also be added in the summary.
    
    Citation: https://doi.org/10.5194/egusphere-2023-2161-RC3
RC2: 'Comment on egusphere-2023-2161', Anonymous Referee #2, 01 Dec 2023

The paper discusses the use of satellite retrievals to evaluate warm rain processes and constrain the ERFaci in the Energy Exascale Earth System Model version 2 (E3SMv2). It applies an updated Contoured Frequency by Optical Depth Diagrams (CFODD) analysis, in both observations and simulations, to provide observational constraints for droplet collection efficiency. The autoconversion parameterization is modified in a series of sensitivity experiments to evaluate the simulated CFODD with respect to the observed one. This allows to estimate a more accurate simulated ERFaci that is driven by warm rain processes.

Main comments:
The manuscript provides an adequate description of the model and the satellite data that are used, including their limitations. I appreciate the authors' approach using CFODD to constrain the simulations, it is elegant and provide interesting results. Nevertheless, the CTRL simulation seems to agree well with the observations. This is a good result for the model and the control coefficients. Shouldn’t this be the main conclusion of the study? Is there a physical meaning for choosing other coefficients, in this model or others? it would be interesting for discussion.
The paper would benefit from improved organization and clearer, more concise writing. Specifically, Section 4 (with the note that there is no Section 5 between sections 4 and 6) is lengthy and could be subdivided into subsections. The latter part of Section 4 could potentially be formed into a subsection that addresses the limitations of the study. Additionally, the introduction section should concentrate on past studies and their relevance to the current research. Paragraphs that discuss the results and the methodology of the present study could be removed from the introduction. Furthermore, the abstract appears overly detailed. Consider revising to make it more concise while still providing essential information.
The manuscript frequently directs the reader to key figures in the supplement. It may be advantageous to consider incorporating some of these figures into the main manuscript to enhance accessibility.
Including a brief description of the CFODD method would be beneficial for readers who may be less familiar with such diagrams. This addition could provide clarity and context for those who may not have an in-depth understanding of the CFODD method.
One of the modifications to CFODD applied by the authors is reducing the cloud optical thickness (COT) to 0.3. As the authors mention, this adjustment has a substantial impact on cloud occurrence. It is worth considering whether this modification could also influence the slope of the CFODD by altering the weight of the bins with low COT (does the fitting account for data density?). Clarification is also needed on whether COT represents the in-cloud mean or the grid mean values. Specifying the resolution of the data is missing and it is unclear how products with different resolutions are combined.
Additional specific comments:
L266: There is a reference typo that needs correction.
L309: The statement about how optically thin layers can have high reflectivities near the cloud top is unclear. It would be helpful to clarify whether this refers to the small optical thickness of thin layers at the top of deeper clouds. The entire sentence needs clarification for better understanding.
L310: The term "strange features" needs clarification. Specify and elaborate on what exactly is being referred to.
L315: It would be beneficial to explicitly mention that you are describing Figure 2 in this section for better guidance to the reader.
L337: Clarify the meaning of "ERFaciSW strengthens." Provide a physical explanation, particularly in relation to its impact on cloud cover and albedo. This clarification will enhance the understanding of the reader.
L348: It is recommended to reword this sentence and provide a clear explanation that the observed slopes are utilized to constrain the simulated ERFaci. This sentence essentially describes the core of the study.
L350: Suggest marking -0.067 on the Figure for clarity and reference.
L351: Clarify the meaning of the total ERFaci of -1.5 in the context of the discussion. Mention again that it is the difference between the simulations (as stated in Line 249). Address why this value cannot be deduced directly from Figure 3.
L352: Specify whether "correcting" means changing the coefficients in the autoconversion parameterization.
L390: Why "however"? The linear relationship applies to both medium-large and small Re.
L460: Elaborate on the statement that "simulated reflectivity profiles near the cloud top are decoupled from the cloud below." Provide a more detailed explanation of what is meant by "decoupled." This is the only instance where "decoupled" is mentioned in the manuscript.

Citation: https://doi.org/10.5194/egusphere-2023-2161-RC2

Charlotte M. Beall, Po-Lun Ma, Matthew W. Christensen, Johannes Mülmenstädt, Adam Varble, Kentaroh Suzuki, and Takuro Michibata

Supplement

https://doi.org/10.5194/egusphere-2023-2161-supplement

Charlotte M. Beall, Po-Lun Ma, Matthew W. Christensen, Johannes Mülmenstädt, Adam Varble, Kentaroh Suzuki, and Takuro Michibata

Viewed

Total article views: 466 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
324	121	21	466	43	14	20

HTML: 324
PDF: 121
XML: 21
Total: 466
Supplement: 43
BibTeX: 14
EndNote: 20

Views and downloads (calculated since 06 Nov 2023)

Month	HTML	PDF	XML	Total
Nov 2023	188	64	8	260
Dec 2023	55	21	5	81
Jan 2024	28	6	2	36
Feb 2024	13	5	1	19
Mar 2024	26	18	1	45
Apr 2024	14	7	4	25

Cumulative views and downloads (calculated since 06 Nov 2023)

Month	HTML	PDF	XML	Total
Nov 2023	188	64	8	260
Dec 2023	55	21	5	81
Jan 2024	28	6	2	36
Feb 2024	13	5	1	19
Mar 2024	26	18	1	45
Apr 2024	14	7	4	25

Viewed (geographical distribution)

Total article views: 455 (including HTML, PDF, and XML) Thereof 455 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 26 Apr 2024

Short summary

Single-layer warm liquid clouds cover nearly one-third of the earth's surface, and uncertainties regarding the impact of aerosols on their radiative properties pose a significant challenge to climate prediction. Here, we demonstrate how satellite observations can be used to constrain Earth Systems Model estimates of the radiative forcing due to the interactions of aerosols with clouds due to warm rain processes.


Total:	0
HTML:	0
PDF:	0
XML:	0