Summary

This study presents the results from the EMAC simulations when different versions of the coupled ISORROPIA thermodynamic modules are used. The study is focused on the main inorganic aerosols (i.e., SO₄^2-, NO₃^-, and NH₄⁺), together with changes in the aerosol water and acidity. The authors conclude that the new version of ISORROPIA (i.e., ISORROPIA-lite) is computationally more efficient than the previous versions of the thermodynamic module (i.e., ISORROPIA II v1 and v2.3, both for stable and metastable modes) and is therefore a good replacement for 3D global simulations. The paper is well-written and well-organized, and the conclusions are useful in exploring the uncertainties of using different versions and setups of the ISORROPIA thermodynamic module in global models. However, the authors can address a few minor issues before the final publication in GMD to make the proposed parameterizations easier to understand for the reader.

General comments

The authors present the differences in simulating inorganic aerosols due to the various versions of the ISORROPIA thermodynamic module. Although the differences are minimal, the authors could provide some additional information on their findings. For example, the authors only state that the differences between ISORROPIA v1 and v2.3 are due to improvements in acidity calculations (Song et al. 2018), or the differences between ISORROPIA v2.3 and the lite version under the same conditions are, on average, less than 5%. Some additional sentences on the impact of these updates to the code on the simulated concentrations of inorganic aerosol components would be useful for the reader.

Considering that the gas-particle partitioning of semi-volatile species such as HNO₃ is very sensitive to the calculated acidity levels and aerosol water concentrations, the authors could discuss more about why these differences exist in the model among the different versions of ISORROPIA, providing additional global maps for the main inorganics and focusing particularly on regions where such differences (positive or negative) are important. This, along with a slightly more detailed technical description of the advances in thermodynamic calculations and the evolution of the ISORROPIA module, will help the reader understand and interpret the presented sensitivity simulations.

Finally, the authors present a comparison of EMAC simulations between ISORROPIA-lite and ISORROPIA II in stable mode. It is not clear why such a comparison is shown here,  especially taking into account previous works of the authors. Is it because these are the standard versions available now in the EMAC model? If the "comparison is done in an attempt to quantify the effects of using the metastable case in global atmospheric simulations,"  as stated in the manuscript, why didn't the authors just use the metastable mode of ISORROPIA II v2.3 to show that? Wouldn't a fair comparison between the two versions require them to be in the same (metastable) mode? Further discussion is needed to support this choice since the results of the different ISORROPIA aerosol modes (i.e., stable vs. metastable) are, indeed, expected to differ.

Specific comments

• In Sect. 4.1 (p. 14), the authors present the differences in SO₄^2- annual mean surface concentrations between ISORROPIA-lite and ISORROPIA II (in stable mode). Does ISORROPIA II directly impact the SO₄^2- concentrations in the model, e.g., through the formation of insoluble CaSO₄ and its precipitation out of the aerosol aqueous phase? Does the model also consider sulfate production in aerosol water? Does the difference in inorganics from ISORROPIA calculations impact cloud acidity in the model and, thus, the respective sulfate production? Please discuss. 

• In Sect. 4.1 (p. 14; l. 456), the authors state that the absolute differences between ISORROPIA-lite and ISORROPIA II for the fine NO₃^- are greater than those of coarse mode. Although this can be explained due to the different aerosol states used for ISORROPIA among the two simulations, it would be helpful to show which version of the thermodynamic model produces results that are closer to observed values. Can such a difference in the coarse aerosols also emerge through the assumption of kinetic limitations applied in the model during condensation of HNO₃ in the coarse mode? Although the parameterization is well documented in the literature, a somewhat more extended discussion would be useful for the reader.

• In Sect. 4.3 (p. 23), the authors state that the pH values are calculated based on instantaneous H⁺ and H₂O values estimated every 5 hours. Why specifically 5 hours? Does the model produce instantaneous outputs only every 5 hours by default (standard output), and that is the reason the authors use the most frequent model output? Does this, further, mean that a more frequent instantaneous output (e.g., hourly) would potentially produce more accurate pH results? Please discuss. 

• Karydis et al. (2021) showed that the metastable assumption produces more acidic particles in regions with high concentrations of mineral cations, such as downwind desert areas, and low RH values. As expected, almost the same results are presented here when comparing the ISORROPIA-lite (i.e., only in the metastable mode) and ISORROPIA II (in stable mode) simulations. It is not clear, thus, the added value of such a comparison here. Can you please discuss more?

• It is well established that NH₃ is the major buffer in most regions of the world. Therefore, if all NH₃ emissions were turned off, the thermodynamic system would definitely give unrealistic results, and as expected, aerosol particles would be extremely acidic. Maybe doubling or cutting in half NH₃ emissions would make more sense to explore potential differences in the responses on the two versions. It would also be advantageous to discuss the presence of non-volatile crustal species from sea salt and dust and how drastically they can change (increase) the aerosol pH in ISORROPIA-lite simulations compared to ISORROPIA v2.3 in the metastable mode. This would also give additional information on the impact of binary activity coefficient calculation between the two versions.

• Sect. 3.3: It would be very useful to also present the seasonal variation for the comparison of the main inorganics (where available) between observations and model predictions, not only the annual mean values. Additionally, you can present the evaluation of the other sensitivity simulations performed for this study, not only the ISORROPIA-lite. This would help the reader better understand the pros and cons of each assumption.

• Section 5, Page 25: It is not clear from the conclusions which version of ISORROPIA the authors propose to use for EMAC simulations. This section lacks an explanation as to why the stable mode was previously chosen for the model over the metastable mode, but now it is replaced with the metastable one. Is it only a matter of computational speed?  A more detailed discussion would help.

• Sect. 2.1, l.169 & Sect. 3.1 l. 216: The emissions of crustal ions such as Ca²⁺, Mg⁺, and K⁺, are calculated as a fraction of dust fluxes in the model. In what form these ions are emitted; totally or partially soluble/insoluble? Are these fractions directly inserted in ISORROPIA calculations? Do you also track in your model the different species upon the ISORROPIA call (e.g., CaSO₄)? In how many modes/sizes aerosol emissions are emitted in the model? Is ISORROPIA  called for every aerosol mode/size or only for accumulation and coarse, as presented in the manuscript? If yes, how do you define here the fine aerosol acidity? Please discuss.

Technical comments

• Page 2, l. 50: The transition from health-related issues to the climate impacts of aerosols is very steep.

• Figure 9: It would be easier for the reader to provide more details in the titles of the figures in the right column because negative pHs are acceptable values (not only for differences). A more detailed figure title can apply to all figures, especially when you show differences.

Summary

This EMAC study investigates differences in aerosol modeling results using ISORROPIA II v1, ISORROPIA II v2.3, and ISORROPIA-lite. Notably, disparities in major aerosol components between ISORROPIA II v2.3 and ISORROPIA-lite are consistently less than 10%. Moreover, the application of ISORROPIA-lite results in a notable 5% acceleration in EMAC's computational performance. Despite ISORROPIA-lite's limitation to supersaturated aqueous (metastable) solutions, the authors endorse it as a dependable replacement for the previous thermodynamic module in EMAC. The paper's content is sufficiently detailed, and with the code now accessible through a Zenodo private repository, the manuscript could be considered for publication once all reviewer comments have been addressed. It's important to note that I concur with the specific comments made by referee 1 and won't reiterate them here.

 

General Comments

1. Accuracy and Clarity:

To ensure accuracy and clarity, it's essential to avoid misleading statements. While the results from ISORROPIA-lite are promising, its restriction to the metastable aerosol state renders it too limited for global atmospheric chemistry applications. This limitation could lead to errors in radiative forcing estimates, particularly in the free troposphere with low humidity. What is really needed are codes that can capture the hysteresis effect of aerosols in order to improve aerosol radiative forcing effects. Therefore, the statement that "ISORROPIA-lite can be a reliable and computationally effective replacement of the previous thermodynamic module in EMAC" should be approached with caution, pending a thorough evaluation of its suitability for global applications.

2. Omission of References:

The omission of references to relevant thermodynamic codes commonly used within EMAC is a notable gap in the introduction, potentially impacting the manuscript's scientific credibility. It's crucial to acknowledge and cite widely accepted models, following established conventions in scientific publishing.

3. Consistency:

Ensure consistency in the spelling of "ISORROPIA-lite" and other acronyms throughout the text to maintain clarity and professionalism.

Specific Comments

1. Discussion of Activity Coefficient:

If tabulated activity coefficients are mentioned, it's crucial to provide a clear and comprehensive explanation or reference regarding their origin and relevance. This will ensure that readers fully understand their context.

2. Temporal Analysis:

In Table 2, where annual means of surface concentrations are discussed, it's worth noting that a 5% difference on an annual scale can translate to significantly higher variations when considering shorter timeframes, such as hourly averages, commonly used in air quality applications. To enhance the analysis, consider extending the statistical examination to at least daily values at a regional scale, focusing on selected networks. Relying solely on mean annual concentrations limits the scope of the analysis and its conclusions.

3. Computational Speed-Up Analysis:

The metric presented in Table 6 regarding computational speed-up should ideally encompass information about load imbalances within the system or undergo a more rigorous statistical analysis. To strengthen the analysis, consider running multiple iterations for each version to draw more robust and conclusive findings. As currently presented, the analysis is relatively weak, and its conclusions are somewhat limited.

4. Section 4 Focus on Surface Concentrations:

Section 4 predominantly concentrates on surface concentrations, which may not offer a comprehensive evaluation of the metastable effect as intended by the authors. Consider revising the analysis in Section 4 to include an assessment of the vertical integral (burden) and, at the very least, a comparison of zonal means. The current presentation may be misleading without these additional elements.

5. References and Errata:

Ensure that references are not duplicated, and address any missing errata. This will enhance the overall quality of the document and its accuracy.