This paper describes the updated version of the regional chemical transport model CHIMERE. Details of various improvements are introduced. It is shown that the new version has better model performance and lesser computational demand.

This manuscript seems to me as a technical document of the model. I think following information is important as a scientific paper describing the new model.

- What were the problems in the previous version?

- How were the problems in the previous version solved in the new version?

- How did the improvements in the new version contributed to better performance?

We hope to know which improvements contributed the most to the better model performance shown in Section 8.

I think the pollen modeling is one of new important improvements. How about the performance of this pollen modeling?

Specific comments:

Figure 1

I think this figure needs to be improved. Everything is not written by English words. I cannot understand some (computational?) words. The boxes are connected by allows. Some boxes just show the name of the routine, and others describe their roles (?). I do not understand why they are separated by boxes and connected by allows, and what their colors represent. And I do not understand what the allows within the two of the three boxes represent. I think the data transfer between CHIMERE and WRF is not well shown in this figure. I cannot imagine “the XIOS routine is called to write the results on disc” in this figure.

P3L13-15

What does “at least two are recommended (not recommanded)” mean? These three levels should be well represented in Figure 1.

Figure 2

Uppercase and lowercase letters (e.g. XIOS and xios) should be consistent among the figure, its title, and the main text.

P5L5-6

Does “v2020r3” include full outputs with nc4? What else contributed to the faster computation?

P5L15-16

Is it impossible to share processors between CHIMERE and WRF?

P6L22-25

I think the size here means the aerosol size. It should be shown.

P7L6-8

Readers may not understand these sentences if they do not know “operator splitting”, “twostep”, and “Melchior 2”.

P7L22

I do not understand a relationship between these four modifications and the Després and Lagoutière scheme described in the previous paragraph. In addition, these four schemes are relatively old. What is a motivation of these four modifications?

P7L23 and L26

Implementation of the Walcek (2000) advection scheme is duplicated.

P8L11

How are products from tracers due to chemical reactions treated in the model?

P9L3

“between” is duplicated.

P9L18-26

Is new CHIMERE compatible only with the USGS landuse? How about MODIS or other landuses which can be selected in WRF?

P10L6

What is Q0?

P10L10-11

“10m wind speed, the one of WRF being dependent on the boundary layer height scheme and not always satisfying”. What does this mean?

P12L7

Is P0=10000 Pa correct?

P13L7-8

“the multiplication is replaced in the calculation by its result 7.2868 1022”

I cannot understand it.

P13L15

What is “q”?

P14L12 and Figure 3

The Charnock constant is set to be 0.018. Why are roughness lengths calculated for three different Charnock constants displayed in Figure 3?

Section 5.4

Has this urban correction been newly implemented in this version of CHIMERE? What is the motivation? Is there any reference showing the default values used in this study?

P16L15-17

Which equation was finally used to calculate z0h?

Table 4

These values are also shown in Table 3.

Section 6.3.2

What kind of input information is necessary to consider influences of fire? I suppose influences of fire become evident only after fires. That means information of dates when fires occur is required. In addition, burned vegetation may recover after a certain period. How is such recover of vegetation modeled? I am curious how much this modeling on impacts of fires realizes improvements in performance.

This paper provides a technical description of the latest CHIMERE model updates and developments, including a comparison of the new and old model version against a number of observational datasets. The information provided here is useful for both the CHIMERE community and other chemistry-transport modeling groups, making it suited for publication in GMD.

While the general structure of the paper is good, the text should be be expanded in places to add depth and clarity to the described approaches. The below comments follow the chronological order of the manuscript and include both specific and general comments (bold), hopefully avoiding points already raised by the first reviewer. As a general comment, please make sure that acronyms are appropriately defined throughout the text. For the revised manuscript it would also be helpful if the line numbering was not reset for each (sub-)subsection.

Line 10 page 4: What is meant by “the chimere ranks”?

Line 13 page 4: Please define “xml files”.

Section 2.2 general comments:

The increase in run-time of 69.4 % when using the nc4 system seems very high, considering that then around 1/3rd of the total CPU time is spent on writing outputs. Did the old system somehow always include all possible outputs? It would be useful to mention in the text to what extent the XIOS_full scenario corresponds to the outputs that users would have typically requested using the old system. If the old system was also configurable,  a "nc4_slim" experiment should probably be added to Figure 2.

Figure 2 also shows that v2020r3 was slower by ~35 % due to things unrelated to I/O. Since this seems to be valid as a general conclusion, I think this should be mentioned/discussed in the text. Have there also been any experiments investigating the impact of model resolution on the performance gains, considering that, e.g., the time-step depends on the grid-size of the meteorology? The conclusion (Section 9) mentions that the computational cost of the model is 40 % less than previous versions, also for forecast runs. To my knowledge, the CHIMERE forecasts use a 10 km x 10 km horizontal grid, for which the I/O performance gains might have a different impact?

Lastly, it would be helpful if a short description was added about why one would want to output maximum or minimum sub-hourly concentrations. Is it somehow relevant for comparisons to observations, or to how certain health metrics are defined?

Page 5 Bullet point 2: Managing the optimal distribution of CPUs between the three different processes seems like a non-trivial task for users. Could a short description of the default, or recommended, resource splits be included?

Page 6 Bullet point 12: Could it be clarified if (and how) these limits impact the number of aerosol bin sizes?

Page 7 bullet point 2: Please define PPM. Is the PPM+W scheme somehow a combination of the PPM scheme and that of Walcek (2000)?

Page 7 bullet point 5: Crudely put, if the simple first-order advection scheme is too diffusive for practical use, why is it included as an option? A short motivation on why one would nevertheless want to use this scheme would be helpful.

Page 9 bullet point 5: Just a suggestion, but rephrasing “multiplicative spread” to “geometric standard deviation” would make sense to me when describing aerosol log-normal distributions.

Line 19 page 9: Please define USGS.

Line 22 page 9: I would suggest adding information about the Charnock (1995) ocean surface roughness calculations Table A1 (e.g., by replacing the 0 entries with “Charnock (1995)”, or by adding a footnote). It would also be helpful if the text here refers to Section 5.2, where the Charnock (1995) methodology is discussed in more detail.

Page 10 bullet point 2: Could the upgrade to a more recent scheme be specified in more detail, or somehow referred to? Currently it reads like a rather vague statement. A short description of how the vertical and horizontal resolution of the CHIMERE grid is treated with respect to that of the input meteorology would also be helpful (here or elsewhere in the text).

Page 10 bullet point 4: What is meant with the WRF 10-meter wind speed not always being satisfying? How was this diagnosed?

Table 3: At first glance, it is unclear what the “High cloud option for attenuation” parameter means, even though this is later discussed in Section 5.3. In Table 3 it currently reads as if it is an option that can be turned on or off. Perhaps the wording could be changed to “High cloud attenuation factor”?

Line 5 page 13: Eq. 9 writes the density in molecules per cubic centimeters, but the equation seems to be for molecules per cubic meter, as is also written in the text above.

Line 7 page 13: It would be helpful to specify that “the multiplication” refers to the calculation of air density (in molecules per cubic meter) from the atmospheric pressure and temperature fields. Currently it reads as if the resulting 7.2868e22 factor does not include the Ra = 287.04 J K^-1 kg^-1 term from Eq. 8.

Table 3: The relative humidity limit is set to 1.2 (rhmax), even though the table defines the unit to be in the range of 0 to 1. How should this be understood? Does rhmax also apply to the thermodynamic gas-aerosol calculations and the (coupled) WRF cloud microphysics schemes?

Section 5.4 general comment: The urban correction factors for wind speed and surface sensible heat flux seem like an important part of the model, which I think warrants a more detailed discussion and motivation. For example, which observed trends are being referred to, what are the implications for simulations across cities (i.e., how much do the results change), how representative are the parameter choices for other cities, etc. Is this correction factor normally applied only for urban-scale modeling? And if so, is it independent of model resolution?

Section 5.5 general comment: While a detailed overview of the turbulence variable calculations is useful, it is not clear how much of section 5.5 is a more detailed description of the turbulence calculations described in Menut et al. (2013) versus a description of new model updates. Is the iterative process by itself new?

Line 4 page 21: What does the sub-grid variability option refer to? The option to turn on the urban correction described in Section 5.4? If so, does this then imply that the urban correction factors are applied only to certain GNFR activity sectors, as seems to be suggested here?

Line 4 page 22: It would be useful if the densities and size distributions of each of the 5 pollen species are explicitly mentioned, for example in Table 6 or in Section 6.2.2. In section 6.2.2, it is also unclear if the diameter Dp = 32 µm and density of 800 kg m^-3 refers to grass, ragweed, or both. I agree with RC1 that the pollen updates represent a significant addition to the model, as is also evident from the detail in which the pollen implementation is discussed. Any kind of comparison or validation against observations, or even just a sensitivity comparison to see how the model results are changed, should therefore be included.

Section 7.1 general comment: In parts of the manuscript, aerosols are described using their log-normal size modes (Section 3.2), whereas in other parts they are described using their size bins (Section 8.1). A brief discussion on the different aerosol treatments in the model would be useful, and how that affects the settling velocity calculations based on the equations for particles with a fixed diameter.

Section 7.3 general comment: What is meant by the ISORROPIA module being more stable in the Fortran90 implementation? There are other CTMs still using the FORTRAN77 version, and it would be good to know what kind of instabilities were diagnosed using the CHIMERE model. For example, the GEOS-chem community described convergence problems which can lead to extreme non-linearities in the equilibrium solution (http://wiki.seas.harvard.edu/geos-chem/index.php/ISORROPIA_II). Did the CHIMERE implementation impact any of this behavior?

Line 4 page 33: On line 3 page 14 the current model top is said to be at 200 hPa, whereas here it is defined as 300 hPa. Was the default value changed from 200 to 300 specifically for the experiments in Section 8?

Section 8 general comments

The horizontal resolution of 60 km x 60 km seems quite coarse considering that, for example, 0.25 x 0.25 degree (~25 km x 25 km) ECMWF data was available for these simulations. Is the 60 km x 60 km grid representative of what a typical user would use?

As another general comment, I think it is important to add at least some form of discussion surrounding the comparisons to observations. For example, for the offline WRF configuration, the ozone bias change from -11,2 % to -1,34 % and correlation coefficient change from 0.56 to 0.66 relative to EBAS surface measurements (Figure 5), strikes me as very large. Which part of the model updates could/would have caused such a big change? Some of these discussion points could then also be added to the conclusion in Section 9 (I think ozone or PM10 bias changes by 5 to 10 percent points should absolutely be mentioned in the conclusion).

References

Menut, Laurent, et al. "CHIMERE 2013: a model for regional atmospheric composition modelling." Geoscientific model development 6.4 (2013): 981-1028. doi: https://doi.org/10.5194/gmd-6-981-2013

The paper has been submitted as model description paper, and describes CHIMERE chemistry-transport model (CTM) v2023r1.

The paper provides an overview of a changelog for last decades, describes technical aspects of running model including the syntax of  control files, used libraries, choices of configurable options, details of input data, classical equations used with slight variations in many other atmospheric models, descriptions of various modules and choices for them. At the end results of an evaluation of several 6 configurations (3 pairs of setups before and  after the update) are shown. The main conclusion is that the new version is faster, more flexible and more accurate than the previous one. 

 

The paper does not meet the GMD criteria for the model description paper:

"The papers should be detailed, complete, rigorous, and accessible to a wide community of geoscientists."

The paper touches too many aspects to give a description for in each of them that meets these criteria. 

 

The paper does not follow the required structure:

"The publication should consist of three parts: the main paper, a user manual, and the source code, ideally supported by some summary outputs from test case simulations"

 

CHIMERE is one of the leading European offline chemistry-transport models, so a publication on is of a major interest, therefore I suggest a major revision or "encourage to resubmit":

- move all the aspects of used formulations and technical implementations for different modules  to the model description. Such documents are normally rather long and can accommodate for  a necessary level of details. They can also be versioned and distributed together with the model.

- ways to run the simulation, details of configuration namelists and guidelines to chose specific   options, belong to the user manual

- the paper should give a concise and accessible description of the model, its modules and components and their interaction, and focus on several most important improvements in the new version and demonstration of their effect on the model performance. This way the above criteria can be met.

 General comments:

 

1. The manuscript is poorly written. Quite a few phrases have unclear meaning.  It has to be heavily edited to become understandable by a wide community of geoscientists. I have highlighted few such phrases for the first page only (see minor comments), but the problem exists for the entire text.

 

2. Introduction lists a lot of changes and updates, lists various modules, mechanisms and pieces of data provided with the model. Some table summarizing these changes would be of help, but bot in the introduction. Many of the modules are given with abbreviations and names etc without references. Each point there leaves more questions than gives answers. 

3. Introduction does not specify what "CHIMERE v2023v1" actually is. Is it just a CTM code, a code bound with a specific emission datasets, a bunch of setups for several different applications? Does it include initial and boundary conditions for all possible use cases? 

  

4. The paper describes the new state of the model without describing the old state, which makes it difficult to understand what has been improved, why and how significant each improvement is. Some table summarizing these changes would be of help.

  

5. The paper attempts to cover all the changes to the model, effectively not covering anything in sufficient details. It would be beneficial to highlight few main changes, describe in details why they were needed, what has been changed to what, describe use cases when they bring the most improvement and illustrate the improvements due to them separately. To mention several: Urban correction, subgrid dust emission, effect of fires on ozone deposition and dust emission, several implementation for ragweed pollen etc. Does each of them improve the model performance? How much? Those interested in complete list and chronology of changes during last couple of decades should be referred to a changelog.

 

6. Specific keywords, data formats, namelist syntaxes, files, variables, classical equations used to diagnose meteorological parameters, implementations of specific schemes/modules, ways to organize subroutines etc. should be  left to the model documentation, where they should be properly described. 

 

7. The conclusions essentially say that the new version is faster, more accurate and more flexible.  How does this information advances geoscientific model development in general?

 

Specific comments (1st page only):

 

p1.l5: "The impact of fires on wind speed, soil properties and LAI was added." Do fires affect the wind speed? 

 

p1.l7.: "..Europe with 60×60 km..." The model has been run as a part of a CAMS-REGIONAL ensemble with 10-km resolution for the whole Europe. The project makes a whole-year assimilated run at the same resolution. Why so coarse resolution for the evaluation of non-assimilated simulation?

 

p1,l12: "models have proven their interest ..."

p1.,l13: "These models have to be accurate and efficient, with at the same time a good accuracy and a low computational time."   Same message twice?

 

p1., l18: "the possibility to read emission fluxes from fires"  Hopefully, "fluxes from fires" not "reading from fires". What makes these emission fluxes different from other emission fluxes?

 

p1. l18: "the SAPRC chemical mechanism" Non-introduced abbreviation. 

 

p1.,l19: ".. version included ... new datasets for the chemical boundary conditions" Does the model include a full datast needed to run it? Can it be used with other datasets?

 

p1,l.20: What makes a hemispheric run so specific?

 

p1,l.21: Fast-JX is not introduced properly. What is "resuspension scheme"? 

 

p1,l.22: WRF abbreviation, no reference

 

p1,l.22: OASIS3-MCT abbreviation