the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Benchmarking GOCART-2G in the Goddard Earth Observing System (GEOS)
Allison B. Collow
Peter R. Colarco
Arlindo M. da Silva
Virginie Buchard
Huisheng Bian
Mian Chin
Sampa Das
Ravi Govidaraju
Dongchul Kim
Valentina Aquila
Abstract. The Goddard Chemistry Aerosol Radiation and Transport (GOCART) model, which controls the sources, sinks, and chemistry of aerosols within the Goddard Earth Observing System, recently underwent a major refactoring and update to the representation of physical processes. This paper serves to document code changes that were included in GOCART 2nd Generation (GOCART-2G) and establishes a benchmark simulation that is to be used for future development of the system. The code refactoring increases flexibility such multiple instances of an aerosol species can be run and interact with radiation and cloud microphysics, in addition to the output of multiple wavelength aerosol optical properties in support of data assimilation. From a science perspective, a new radiatively active tracer, brown carbon, was added to distinguish smoke from other sources of organic aerosol thereby improving optical properties entering the radiative calculations. A four-year benchmark simulation was evaluated using in situ and space borne measurements to develop a baseline and prioritize future development. A comparison of simulated aerosol optical depth between GOCART-2G and MODIS retrievals indicates the model captures the overall spatial pattern and seasonal cycle of aerosol optical depth but overestimates aerosol extinction over dusty regions and underestimates aerosol extinction over northern hemisphere boreal forests, requiring further tuning of emissions. This MODIS-based analysis is corroborated by comparisons to MISR and selected AERONET stations. Despite the underestimate of aerosol optical depth in biomass burning regions in GEOS, there is an overestimate in the surface mass of organic carbon in the United States, especially during the summer months.
- Preprint
(2826 KB) - Metadata XML
-
Supplement
(1861 KB) - BibTeX
- EndNote
Allison B. Collow et al.
Status: open (until 05 Oct 2023)
-
RC1: 'Comment on gmd-2023-129', Anonymous Referee #1, 04 Sep 2023
reply
Collow et al. introduce and evaluate the GOCART-2G aerosol module based on a variety of observation-based data sets. The evaluation is informative and yields interesting results. But I think that especially the introduction, the conclusion section, and the abstract should be improved. Many key results are presented clearly in the figures and the main text, but they should also be summarized and discussed in the conclusion section. The summary section should ideally contain not simply a summary, but a synthesis of the results from the evaluation, for example with respect to results from the evaluation with MODIS and IMPROVE (compare my main comment #6 below). Ideally, the introduction and discussion should provide some context that may help the reader to better understand the results. Several key results could perhaps be discussed in the context of existing literature. A few points also deserve further discussion in the main text. AOD over the major source regions of anthropogenic aerosol in Europe, North America and South Asia in northern hemisphere spring and summer is lower in the GEOS-GOCART-2G model simulation compared to Aqua MODIS NNR data. Overall, the AOD difference between GEOS-GOCART-2G and Aqua MODIS NNR is smaller over South Asia compared to North America and Europe. There is also an interesting point about differences between Aqua MODIS NNR and Terra MODIS NNR, which I think could be further discussed in a few sentences, if possible in the light of existing literature. I do not suggest additional analysis within the framework of this manuscript. Instead, some of the suggestions that the authors made could be explained a bit better, and there could be additional attempts at synthesizing the findings. There should also be a better description of specific open issues that arise from the evaluation in the conclusion section, and perhaps the authors could mention more concrete ideas either for further investigation of these issues or for addressing them. The explanations of technical aspects require clarifications.
Main comments:
1. Abstract: I recommend to cut all the details on code changes and instead summarize key results from the evaluation (see specific comments below). I appreciate the technical work that went into the code refactoring and the new features, and I understand how very important this type of model development is. But I find the information in the abstract hard to follow and possibly of limited value for the wider readership. I think it is enough to mention that the code has been refactored in the abstract and to explain some of the improvements in a revised section 2.3. The rest should be left to documents such as user guides and more comprehensive technical documentation.2. Introduction: I suggest to re-write the introduction to better motivate and explain the model development and the evaluation. In order to motivate the model development, I suggest to clearly explain in which contexts this new module will most likely be used, and which shortcomings motivated the additional development. I suggest to provide specific scientific background that prepares readers to better understand and appreciate the interesting results of the model evaluation. I find that at the moment, the introduction contains a lot of fairly general background information without a clear link to the results from this specific study. I recommend to use the introduction to motivate the model development steps described in the manuscript, to motivate why specific steps were taken for the model evaluation, and to prepare the readers for understanding the specific results. In case you added diagnostics or substantially revised them, this could also be mentioned and motivated. If you think this will help the reader, you could also explain and mention some open issues that your code refactoring addresses (although I recommend to focus on motivating the scientific aspects and the evaluation, even in case the evaluation used a standard package and did not include additional diagnostics compared to previous publications on GOCART). I think that instead of providing general and/or historical background, the introduction should serve to motivate and explain the rest of the manuscript. You could also motivate Section 4.2.3 on stratospheric AOD.
3. Please revise Lines 176-183 for clarity. Avoid jargon and explain advantages in order to motivate the changes. At the moment, it is not clear to me, how the sentence starting in line 183 is linked to the preceding sentences. I think it is linked, but I am not sure.
4. The discussion section should put a much stronger focus on synthesizing and discussing the results from the evaluation.5. Can you explain which model biases are inherited? Can you provide clues based on existing literature? It would be nice to point out where the results from this very informative and comprehensive evaluation add to existing knowledge and where they confirm previous insights. I think you could use the introduction section to summarize known issues and the conclusion section to point out where your evaluation has yielded new insights or may lead to new ideas.
6. Based on Fig.6, AOD biases for Europe, North America and South Asia, which are major source regions of anthropogenic aerosol, appear to all show a seasonal cycle, with smaller biases in winter. Fig. 16a for North America shows a seasonal cycle for sulphate in the IMPROVE data, but much less for GEOS-GOCART-2G. Do you have any thoughts on this and/or can you find information on this in the literature? Could the lack of a seasonal cycle in Fig. 16a be part of the reason for the seasonal cycle of the bias Fig. 6e? Is this a known issue? And does it apply to Europe and South Asia as well? Figure 18a suggests that for Europe the answer might be yes, but that it is not limited to sulphate (as you noted in the text). Do you know whether this seasonal AOD bias is linked to a seasonal bias of precipitation? And/or is there a known seasonal bias in one of the source terms? Unless this is already understood, you could perhaps suggest to investigate potential links between biases in meteorological variables and biases in AOD and/or to investigate the seasonal cycle of source and sink terms in a follow-up study. This may or may not help to explain the seasonal cycle of the AOD bias. All in all, I very much like the wealth of different diagnostics and how the authors present them. I find that the outcomes provide a very good motivation for this manuscript. But I would nevertheless like to encourage the authors to spend even more effort on trying to synthesize results in order to derive ideas and conclusions from their data analysis, and where this is useful, also to put their results into the context of the existing literature. Especially with respect to sulphate, it may also be interesting to speculate on potential effects of biases on ERFari+aci, although whether to include such speculation is a matter of taste and should be decided by the authors. I am not sure how relevant this aspect is for the GOCART applications.
Specific and other general comments:
1. I suggest to write GEOS model instead of simply GEOS throughout the text. I suggest to also add the word model to the end of the title. Alternatively, the authors could consider writing "Goddard Earth Observing System model (GEOS)" each first time the acronym occurs in the title, the abstract and the main text. But I think that writing GEOS model would be much more accurate than GEOS when referring to the GEOS model, because GEOS obviously involves other activities apart from modeling or data assimilation.Did you set up a case and evaluate it with existing diagnostics, add new diagnostics to a set of existing diagnostics, and/or are you also providing a new or an updated tool or framework for model diagnostics? If you actually provide a new or an updated tool or framework, you could make this visible by slightly changing the wording in the abstract and then explain it in the main text. Are the new or revised diagnostics part of the preprocessing software mentioned in the code availability section? If yes, you could for example stress this in the discussion section.
Abstract:
Line 14f: ", which controls sources sinks and chemistry" -> sources and sinks of what? What about aerosol physics, deposition, wet deposition, etc.? I suggest to replace this statement by something like ", the aerosol component in the Goddard Earth System (GEOS) model" or ", an optional aerosol component in the Goddard Earth System (GEOS) model" or ", an optional aerosol component for the Goddard Earth System (GEOS) model", or simply ", which is part of the Goddard Earth System (GEOS) model". I included the GEOS because the acronym is used in the abstract but not explained.
Line 16-19: The benchmark case is mentioned again in line 21. I suggest to shorten this because the rest is of little interest for the wider readership. You could replace "This paper ... From a science perspective, a" by "A".Line 16: I think that this is neither the right place for documenting code changes nor that the manuscript does a fair job at actually documenting these code changes. Perhaps, "outline" would have been a better word than "document". But I think this sentence should be cut.
I think you should mention additional key results from the model evaluation in the abstract
1. Introduction:
Line 31-34: I find this sentence confusing because the first part sounds like GOCART is used in a traditional ESM context while the second part does not mention coupled models. I suggest to directly mention the GOCART-2G aerosol module and explain in which GEOS model applications GOCART-2G will be used (data assimilation, forecasting, ...?) in order to motivate your study. Optionally, you could also clarify whether GOCART is used for estimating ERFari+aci or for climate projections, whether it participates in CMIP and/or in AeroCom and/or whether there are plans to do so. Please also explain what setup you are using here and motivate this choice. I think the introduction should serve to motivate and explain this particular study. The statement "[a]s general circulation models strive to take a comprehensive Earth-system approach" hints at traditional ESMs, and I am not sure this will be the main application for GOCART-2G. Instead, the authors could for example briefly provide some background on applications of aerosol reanalyses products.Line 35-52: This seems like a very general and somewhat arbitrary background on aerosol modules. I do not understand what some of these points have to do with your results and how this either helps to motivate your study or else how some of this scientific background helps readers to understand the results from your model evaluation. I suggest to focus on the issues which are most important for your evaluation and to explain them a fashion that ensures that this becomes clear.
Line 52-64: This sounds like a history of GOCART. I think that readers may instead be interested in what future applications you envisage for your module. Can and will this be used only for data assimilation or also climate projection, short term forecasts, etc.? See also my point regarding lines 31-34 above.
I suggest to revise the introduction to motivate and explain this study and to provide scientific background that prepares the reader for understanding the results.
2 GOCART aerosol module in GEOS:
Line 166: Introduce GEOS FPLines 176-183: Please try to explain advantages for the user. Please try to avoid expressions such as "multiple instances" or "child" or else explain them. Personally, I very well understand the meaning of the expressions "multiple instances" and "child" and I also understand your goals. But I still feel like I do not quite understand what you are actually trying to say because I lack the (model specific) background to link your goals and your technical explanation.
4 Evaluation of GOCART-2G:
Line 300: Please explain this point in some more detail, reminding readers of the different overpass times, and, and as far as possible, also try to interpret the differences.Line 328: For South Asia, the difference between satellite and model AOD depends on whether Terra MODIS NNR or Aqua MODIS NNR is used, again suggesting the influence a diurnal cycle. Please explain this point in some more detail, and as far as possible, try to interpret the differences in the light of the different overpass times. Please consider linking your discussion to existing literature such as https://doi.org/10.5194/amt-11-4073-2018.
Line 349: Could you please briefly elaborate on your comment regarding emissions from smaller scale sources?
Line 449: Has planetary boundary height in the GEOS model version that is used here been evaluated with observations? If there is a reference, you could cite it.
5. Discussion:
Lines 455f: I suggest to omit this sentence, or else explain again what HEMCO is and what the advantage of this step is. I don't understand the meaning of "As part of the new species".Line 457-459: You cloud simplify this by saying something like "we added brown carbon and simplified the addition of new species." At the moment, I have two other comments regarding these lines:
Line 457: I understand your point. But this is the results section. And why do you repeat a point in line 457 that you have made in line 453?
Line 459: Is there any physical explanation why ash should be an instance of dust?Line 471-482: Please omit. This sounds like the introduction to another manuscript.
I encourage the authors to put a much stronger focus on synthesizing and discussing the results from the evaluation in the discussion section.
Technical comments and suggestions:
Line 211: with -> inLine 331: decent agreement -> decent agreement with respect to the annual cycle
(AOD is underestimated in all four regions in Fig. 7)Line 411: GEOS -> GEOS-GOCART-2G
Line 432: do not assimilated -> did not assimilate?
Citation: https://doi.org/10.5194/gmd-2023-129-RC1 -
RC2: 'Comment on gmd-2023-129', Anonymous Referee #2, 10 Sep 2023
reply
The comment was uploaded in the form of a supplement: https://gmd.copernicus.org/preprints/gmd-2023-129/gmd-2023-129-RC2-supplement.pdf
-
RC3: 'Comment on gmd-2023-129', Anonymous Referee #3, 26 Sep 2023
reply
This manuscript documents the GOCART-2G implementations, and a multi-year benchmarking simulation results in the Goddard Earth Observing System (GEOS) and compares the aerosol results with observations from multiple sources (e.g., satellite and ground networks). The main purpose is to provide a reference for this new aerosol package version. The paper is generally well-organized and clearly written. The scope of the study fits well with the GMD journal. In particular, I like the detailed descriptions of the model development, but some important information, such as comparisons with the previous version or peer model results, is surprisingly not included. Together with the relatively weak motivations, it is difficult to understand the significance of this new development and its implications to climate model development at large. I would suggest a major revision before publication.
General comments:
- Since this is the second generation of the Goddard Chemistry Aerosol Radiation and Transport (GOCART) model, it is a bit surprising that the authors did not show any comparison between the current and the previous versions. What are the improvements? Anything unfortunately gets worse? What are the driving factors for these changes? These are perhaps most important results to document in a paper like this. I strongly recommend these results be added in the revised version.
- As I mentioned above the details about the code changes and developments are highly appreciated. But a large fraction of these details seems belong to supplementary. In the main body, it would be more useful to provide the rationales of why such model developments are needed. Is it because it helps reduce the model biases? Or it adds some important processes that were not represented in the previous generation?
- More importantly, GOCART as a bulk aerosol scheme simplifies size distributions and lacks microphysics compared to modal and sectional schemes. What are the main reasons GOES continues not to update to those more advanced aerosol treatments? What are the trade-offs? Such information, if persuading, could motivate this study much better than the current version.
- Extending from #3, how does this bulk aerosol scheme perform relative to peer models, like CMIP6? Comparing to other models is as important as to the observations in this context. I suggest the authors add other model results besides the previous GOCART results.
- Many CMIP6 models appear to have too strong aerosol indirect forcing. This has become a major concern of the current model development. It would be helpful to include more details about how aerosol-cloud interactions are handled in this study and what is their aerosol indirect effect.
Minor comments:
L82: Glad that brown carbon is added. Can you show some results of the brown carbon bleaching effect? Is a 2-day e-folding decay good match with the observations?
L108: Change to SO42-.
Table 1: It is good to know that GOCART-2G can run with different emissions sources and dataset resolutions. Can you briefly illustrate the choices in this table? Why not other sources?
Section 4.1: Add a table summarizing global budget, burden, lift times. Otherwise, it seems missing information under the title of “aerosol mass budget”.
L324-325: Any results support this hypothesis that the negative bias is due to biomass burning aerosol?
Section 4.2.2: do we need to consider the sampling differences between the coarse model grid spacing and the site observations? If so, any attempt to reduce this potential bias?
Supplement: add a table to list all the input variables needed for another model to include GOCART-2G and/or the interface configurations. This will make this paper a better documentation for others trying to port GOCART-2G.
Figure 8 capture: what do you mean by “the one-to-one line plus or minus one of the one-to-one line”? Isn’t 1:1 minus 1:1 zero? Please clarify.
Figure 9: in panel (d), the two lines besides the 1:1 line are so far from the colored area. They are not useful at all. Perhaps, replace these lines with a standard deviation labelled with R and bias?
Figure 13: are there quality control (QC) flags in the OMPS-LP data? Have these QC flags been applied? Otherwise, limiting the comparison to where the satellite data are more reliable makes much more sense.
Citation: https://doi.org/10.5194/gmd-2023-129-RC3
Allison B. Collow et al.
Allison B. Collow et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 280 | 75 | 11 | 366 | 24 | 3 | 3 |
- HTML: 280
- PDF: 75
- XML: 11
- Total: 366
- Supplement: 24
- BibTeX: 3
- EndNote: 3
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1