Evaluating the EPICC-Model for Regional Air Quality Simulation: A Comparative Study with CAMx and CMAQ

Lou, Mengjie; Wu, Qizhong; Wang, Wending; Chen, Huansheng; Cao, Kai; Fan, Xiaohan; Liang, Dingyue; Yu, Fen Fen; Zhang, Jiating; Wang, Wei; Wang, Zifa

doi:10.5194/egusphere-2026-3428

Preprints

https://doi.org/10.5194/egusphere-2026-3428

Preprints

29 Jun 2026

| 29 Jun 2026

Evaluating the EPICC-Model for Regional Air Quality Simulation: A Comparative Study with CAMx and CMAQ

Mengjie Lou, Qizhong Wu, Wending Wang, Huansheng Chen, Kai Cao, Xiaohan Fan, Dingyue Liang, Fen Fen Yu, Jiating Zhang, Wei Wang, and Zifa Wang

Abstract. This study presents a systematic evaluation of China’s independently developed EPICC-Model for regional PM_2.5 and MDA8 O₃ simulations against established international models, using WRF meteorological fields and a multi-source integrated emission inventory. Results highlight the strengths of the EPICC-Model in several aspects: it achieves relatively high spatial consistency with measurements for PM_2.5, with an annual index of agreement (IOA) of 0.80, and accurately captures pollution patterns in heavily polluted North China. It also demonstrates improved performance in simulating summer O₃ peaks, reducing maximum biases by more than 20 μg m^-3, primarily through enhanced heterogeneous HONO formation and nitrate photolysis pathways that elevate OH concentrations, and it incorporates the CB6r5 mechanism to better represent biogenic VOC oxidation. The model exhibits the highest hit rate (45.6 %) for forecasting moderate PM_2.5 and moderate O₃ pollution events and successfully reproduces persistent pollution episodes. However, all models share common limitations, including insufficient capability in reproducing heavy pollution episodes, systematic underestimation of SO₄²⁻, and uncertainties in SOA-related OC simulations. Future improvements should focus on refining secondary aerosol chemistry, emission inventories, and boundary layer representations. This study has not only demonstrated the performance of the EPICC-Model against international models but also provides guidance for improving regional and global air quality models.

Received: 11 Jun 2026 – Discussion started: 29 Jun 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 22038 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (22038 KB)

Supplement (11661 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

12 Oct 2026

Evaluating the EPICC-Model for regional air quality simulation: a comparative study with CAMx and CMAQ

Mengjie Lou, Qizhong Wu, Wending Wang, Huansheng Chen, Kai Cao, Xiaohan Fan, Dingyue Liang, Fenfen Yu, Jiating Zhang, Wei Wang, and Zifa Wang

Geosci. Model Dev., 19, 9651–9676, https://doi.org/10.5194/gmd-19-9651-2026,https://doi.org/10.5194/gmd-19-9651-2026, 2026

Short summary

Mengjie Lou, Qizhong Wu, Wending Wang, Huansheng Chen, Kai Cao, Xiaohan Fan, Dingyue Liang, Fen Fen Yu, Jiating Zhang, Wei Wang, and Zifa Wang

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2026-3428', Anonymous Referee #1, 07 Jul 2026

Please see attached review.

Citation: https://doi.org/10.5194/egusphere-2026-3428-RC1
- AC1: 'Reply on RC1', Qizhong Wu, 19 Aug 2026
  
  The point-to-point responses is at the attached supplement.
  
  Citation: https://doi.org/10.5194/egusphere-2026-3428-AC1
RC2:
'Comment on egusphere-2026-3428', Anonymous Referee #2, 28 Jul 2026
This manuscript provides a comparison of simulated regional PM2.5 and MDA8O3 fields generated by the Emission and Atmospheric Processes Integrated and Coupled Community Model (EPICC-Model) with results from the CAMx and CMAQ modeling systems. Operational model evaluation statistics for all three models, benchmarked against a set of surface observations, are presented to facilitate an assessment of their performance across average spatial and temporal scales. Comparison of EPICC performance results alongside those from established and widely used models such as CAMx and CMAQ is a primary motivation for this study. As expected, statistical metrics averaged over broad spatial (domain-wide) and temporal (seasonal) scales reveal comparable performance among the models, which are fundamentally based on similar process representations.
Establishing and documenting this baseline for EPICC performance relative to other models is valuable for prospective users – thus, a manuscript of this nature will be of interest to the journal readership. Equally critical for the adoption and advancement of a modeling system is the explicit demonstration of its unique strengths and a clear articulation of how its formulation and structural features support its intended applications – an aspect of the manuscript that could be improved further. This could be accomplished by enhancing the clarity of diagnostic evaluation for the EPICC model, thereby highlighting the influence of distinct attributes such as process representation, numerical schemes, and the integration with meteorology and emissions, on the simulation of PM2.5 and O3, relative to CAMx and CMAQ. The following comments and suggestions are offered that may help improve the uniqueness and usefulness of the manuscript.
Throughout the manuscript the EPICC model is referred to as “independently developed”, but it is not apparent to me what that implies? Was the code development independent and if so, independent of what? Is the science in the model not based on other experiences and knowledge accumulated over decades of international research? Looking at the EPICC code, the overall structure of the model bears resemblance to other contemporary systems including CAMx and CMAQ, share similar names for routines, and often also borrows process routines and numerical approaches. Though the individual routines acknowledge derivative work and include some history of the code development, given the similarities between many aspects of the EPICC model structure and code with other existing open-source models, some clarity on what “independently developed” is intended to convey would be useful for the readers.

L31: “maximum biases” should perhaps be “maximum bias”.

L40: some indication of what specific aspect of boundary layer representation improvement is suggested would be useful. As written, the statement is too broad and does not convey much.

L47: “As a typical secondary pollutant, PM2.5 is influenced by primary emissions” is contradictory and awkwardly stated. The sentence should be rewritten to convey that ambient PM2.5 is composed of constituents that have both primary emissions and are secondarily produced.

L48-49: There are more seminal references that establish these underlying processes regulating PM2.5 accumulation and distributions in the troposphere and should probably be included.

L53-54: Again, the dependence of O3 formation on NOx and VOC and on meteorological factors has been well established dating back several decades (e.g., Hagen Smit and Fox, 1956) and should be reflected by the citations.

L54: This statement (use of "despite") implies some dependence on O3 and PM levels which may not always occur because of different trends in their precursors. Perhaps the statement should just convey that in many Chinese urban clusters PM2.5 levels have reduced but MDA8O3 has increased.

L64-68: References for these models should be provided here.

L83: Structural differences in context of models is often used to describe differences in model structure - architecture, code, design, and not necessarily process differences. Suggest changing "structural differences" to "differences”.

Table 1: Though based on Fast-J, Binkowski et al. (2007; https://doi.org/10.1175/JAM2531.1) is perhaps a more appropriate reference for the photolysis approach in CMAQ.

L139-140: To my knowledge there is nothing that specifies 14 layers for CMAQ or CAMx. Why did the authors also not configure CMAQ and CAMx with the so-called "superior" 20-layer configuration or the default WRF layer structure?

L142-143: Though not employed in the current study, it should be noted that for consistency, CMAQ applications typically employ WRF fields also simulated using ACM in WRF.

L151-153: Given that models are continuously evolving, it is important to qualify these statements with the specific version of the models used in the inter-comparison. Though the CMAQ version used here may have been configured with CB6r3, newer versions do include CB6r5 and variants.

L164-166: Default LBC profiles for CMAQ are only provided to facilitate model setup and initial testing. By definition, the LBCs are dependent on the geographical extent and location of the limited area model application. Thus, it is a bit surprising that the authors did not consider harmonizing the LBCs across the 3 modeling systems. Acknowledgement of this somewhat fundamental difference in configuration setting should be provided.

L265: Is the Kz_min setting too low relative to other models or what is physically expected - if the latter what is the basis to support the value? What fraction of the time during the January-May period are stable conditions encountered?

L278-279: This sentence is vague as it covers everything possible and thus not insightful – which process may be the dominant in driving the noted overestimation?

Figure 1 caption: please clarify if daily PM2.5 implies daily-average PM2.5?

L296-297: please clarify what “lower background concentrations” imply. Is the background low or is it that ventilation is strong in the summer and thus concentrations are lower?

L299-301: Based on Figure 1 alone it is difficult to say whether compared to the other two, one model does better in striking better balance between accurately simulating extreme pollution and sustaining strong annual performance, since these are regional average values. Would the same conclusion also hold at individual locations?

L309: Reduced reaction activity during the cooler months arises from both decreased photolysis as well as lower reaction rates associated with thermal chemistry.

Do the scatter plots in Figure 2b show the same data as in 2a, i.e., average across all sites or is each point the MDA8O3 at individual observation location?

L317-319: References should be provided for the heterogeneous HONO formation and nitrate photolysis pathways. To my knowledge, though uncertain, the nitrate photolysis primarily occurs in marine environments (e.g., Ye et al., 2017; Reed et al 2017; Kasibhatla et al., 2018; Shah et al., 2023; Sarwar et al., 2024; Sarwar et al., 2025) or environments impacted by marine air masses, since halide ions are highly surface active and promote the surface propensity of coexisting nitrate anions (DOI: 10.1039/d1ea00087j). How then does it influence inland environments, especially since the large scale O3 in EPICC is expected to be modulated by the MOZART LBCs? Did the MOZART simulations also include nitrate photolysis? How different are the OH concentrations across the 3 models?

L325: Is the isoprene chemistry representation in cb6r3 and cb6r5 that different to impact the ozone forming potential as speculated? It would be useful to further discuss the differences in the isoprene chemistry leading to the noted O3 differences.

L328: An overly deep PBL should result in more dilution and systematic low (negative) bias as opposed to positive bias. I am not sure I understand the rationale here?

L336: “well-controlled absolute bias despite its systematic underestimation” is confusing and somewhat contradictory. If the model is systematically low biased, how can its absolute bias be well controlled?

L345-346: How does one quantify the better balance between overestimation and underestimation. Based on the analyses presented, it is hard to conclude this in any compelling manner. At best the performance is comparable across 3 models with one model performing better than the other two under certain conditions.

L404: It is not apparent to me what “multi-centered and scattered distribution pattern” implies? Is there a specific physical process that drives the Springtime distribution? This discussion would benefit from additional clarification.

L408: Please elaborate on the reasons why coarse vertical resolution reduces mixing? Also, what altitude O3 are the near surface precursors being mixed with? Why does the EPICC layer structure facilitate "better" mixing representation and under what conditions? What prevented CMAQ and CAMx from also being configured with the same layer structure (and implicitly same meteorology fields) as EPICC?

L414-417: The authors suggest that regional O3 levels and differences between models are primarily controlled by photochemical production. How does one rule out the role of large-scale transport as represented by the differing boundary conditions? Also, given that the layer structure of the native meteorological fields and that of the 3 models is different, and also different amongst the CTMs, I do not see how it can be claimed that identical meteorological fields were used? The starting WRF output is the same but it's translation to the grid structures of the individual CTMs (Table 1) is likely different, resulting in not "identical" meteorological fields driving the CTM calculations.

L417: Perhaps "nation-wide O3" is better usage than "national O3”.

L425: What shifts southward - peak values or just relative abundance? What processes drive this seasonal change in distribution and how does the ability of a model to capture it link back to its process formulation?

L430-431: Can just examination of a single IOA metric be regarded as "comprehensive evaluation”? Suggest replacing “comprehensive evaluation” with just “evaluation”.

L486: The underestimation of SO4 may also contribute to the overestimation of NO3 assuming NH3 emissions are well constrained.

L510-511: Since one would expect artificial dilution of emissions over grid cell volumes to likely result in underestimation of concentrations of a primary species, does the BC overestimation imply overestimation of BC emissions or could the representativeness of the monitor locations also influence the noted overestimation?

L570-571: It is not clear to me what either "consistency in species simulations" or "stable performance of RMSE and NMB" implies? Are the authors suggesting that EPICC simulated PM composition is similar to the other models? Both RMSE and NMB in Figure 7 exhibit some spread across models for several species - it is not apparent to me what stable performance of these metrics implies?

L576-579: Full references to the documents mentioned here should be provided.

L582: If heavy and severe AQI are dictated by significant wind-blown dust (WBD) events and if WBD emissions are not included in the models, what is the value of conducting this AQI comparison, since it should be expected that lower accuracy is expected at higher observed AQI. In other words, if the models are not configured for an application, what is the value in "evaluating the applicability" for graded air quality forecasts. Do the measurement locations examined in Figure 6 also report crustal contribution?

L591-596: This is a potentially useful finding but based on the results presented it is somewhat speculative. Based on comparisons of concentrations of secondary inorganic aerosol constituents by the 3 models for selected such episodes, can the authors demonstrate that EPICC indeed predicts different and more accurate "rapid growth" of secondary inorganic aerosols?

L696-700: How does one quantify spatial progression from central to south China to assess which model simulated this progression "best". Figure 11 only enables a qualitative and subjective assessment.

L700: Would be useful to elaborate on how excessively strong aerosol-radiation feedback led to O3 overestimation in southeastern China and on which day?

L701: Visually while CAMx and CMAQ appear to underestimate relative to the EPICC fields, both CAMx and CMAQ also appear to better capture the gradients in O3 - high O3 at monitors interspersed with monitors with low O3 while EPICC depicts more regionally elevated O3.

L706-708: In my assessment Figure 11 in its current form does not conclusively support the claim conveyed in this statement that EPICC outperformed the other two models in capturing the spatiotemporal patterns.

L725-726: I am not sure what "unified WRF fields" implies. Also as noted earlier, given the different layer structures, it is also difficult to ascertain whether the 3D fields are identical or even consistent across models. Thus, probably should be reworded to: "using meteorological fields derived from the same WRF simulation”.

L731: “seasonal turning point” is vague – sentence should be reworded.

L745-746: No comparisons with observed boundary layer heights were presented to support the statement that ACM2/YSU schemes underestimate mixing layer heights. The statement should be adequately qualified.

L800: It is stated that this work “provides a clear pathway for improving next generation of atmospheric CTMs”. I feel this is a somewhat lofty claim as no new or specific process or structural deficiency was identified. There may certainly be specific enhancements identified for the EPICC system, but I am not sure the analysis identified a “clear path” for next generation of models. Suggest deleting this sentence as the study and results do not substantiate it.

L802-804: In my assessment this study at best provides a comparison of operational evaluation metrics of 3 models. There is no specific analysis presented to support air pollution control or policy making. Thus, I feel this statement should be reworded or deleted.
Citation: https://doi.org/10.5194/egusphere-2026-3428-RC2
- AC2: 'Reply on RC2', Qizhong Wu, 10 Sep 2026
  
  Thank you very much for the reviewer's careful and constructive review of our manuscript, and the point-to-point responses is at the attached supplement.
  
  Citation: https://doi.org/10.5194/egusphere-2026-3428-AC2

Peer review completion

AR – Author's response | RR – Referee report | ED – Editor decision | EF – Editorial file upload

AR by Qizhong Wu on behalf of the Authors (10 Sep 2026) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (14 Sep 2026) by Ulas Im

RR by Anonymous Referee #1 (18 Sep 2026)

Suggestions for revision or reasons for rejection

General Comment
The revised draft manuscript reflects important updates in response to reviewer comments from the first round. I thank the authors for their thoughtful consideration and thorough responses to each comment. Most of the author edits in this round are appropriate and satisfactory.

I have identified some more minor revisions that would help to improve the manuscript.

Specific Comments
1. Section 1, line 70: It is appropriate to use (Ramboll, 2024) as a reference for CAMx, but as a User’s Guide it is considered grey literature. It would be better to include a published reference like (Emery et al., 2024). That reference is cited elsewhere in the revised manuscript but there is no listing for it in the Reference section.

2. Section 2.1, lines 150-151: “whereas CMAQ and CAMx use 14𝜎𝑝 layers.” This specification of sigma-p is not correct for CAMx and is not entirely correct for CMAQ. CAMx uses an absolute terrain-following height coordinate (rather than normalized height sigma-z coordinate) and CMAQ uses the hybrid normalized pressure coordinate system directly from WRF. I suggest simply removing the sigma-p variable from the sentence and wherever it is used throughout the document when referencing the CMAQ and CAMx vertical coordinates. Since the EPICC-Model algorithms are largely based on CAMx, does it also use the same vertical coordinate as CAMx? If so, I suggest also removing the reference to sigma-z for EPICC-Model.

On a related note, the authors’ response to my original comment about inconsistent number of layers was: “Regarding the different vertical layer structures used by the three models, the primary reason is that EPICC-Model adopts a terrain-following height coordinate, whereas CAMx and CMAQ employ terrain-following sigma-pressure coordinates. Due to differences in vertical coordinate definitions and internal numerical implementations, it is not feasible to maintain a completely identical vertical layer structure among the three models.” This argument is based on an incorrect understanding of the CAMx vertical coordinate and how CAMx and CMAQ accommodate the WRF layer structure, as I point out above. Therefore, using this argument for 20 vs. 14 layers is invalid. I appreciate that the authors’ response above was not included in the manuscript.

3. Section 2.1, lines 151-155: Should there be an explicit mention of all 3-D transport treatments for EPICC-Model and CMAQ in the text like there is for CAMx?

4. Section 2.1, lines 159-169: The flow of this paragraph could be improved by moving the rationale for the use of CB05 chemistry in CAMx after the chemistry descriptions of EPICC-Model, CMAQ, and CAMx (at line 182). Also, I suggest dropping the sigma-p reference on line 160 per my comment above.

5. Section 2.1, lines 186-187: This statement remains imprecise. I suggest something like: “CMAQ and CAMx also support using global models like MOZART to provide initial and boundary conditions, but simpler sources of boundary conditions were used for CMAQ and CAMx.” Then provide the rationale for those choices.

6. Section 3.1, lines 390-391: “In spring, elevated PM2.5 concentrations occurred in several geographically separated regions.” I think the authors are referring to observed conditions, that could be clarified by replacing “occurred” with “were measured” (or observed).

7. Section 3.2, lines 554-555: “and the elevated fraction mainly results from underestimation of the total measured-component concentration.” I do not understand this sentence, please revise to make the point more clearly.

Technical Corrections
Be sure to define all acronyms and abbreviations at the time of their first occurrence.

Use consistent past or present verb tense. Many sentences within single paragraphs shift between past and present throughout the manuscript.

Section 3.1, Line 383: Replace “As shown in Figure 3” with “As shown in Table 3”.

Section 3.4, Line 709: The use of the European form of “centre” is inconsistent with the use of “center” throughout the rest of the paper.

Hide

RR by Anonymous Referee #2 (21 Sep 2026)

Suggestions for revision or reasons for rejection

I thank the authors for their careful consideration of my comments on the earlier draft of the manuscript and for their response and revisions to the manuscript. There are a few additional minor clarifications which I feel would be helpful.

1) L129-136: Though the authors highlight the EPICC system to be modular plug-in architecture, the evaluation itself does not assess alternate process representations or numerical schemes. If those assessments have been performed, it may be useful to include relevant citations for such. Also, if alternate approaches are available in the plug-in architecture, it would be useful to state the rationale for the choices in EPICC configuration in Table 1.

2) Table 1: In the photolysis row and associated discussion, what does “Full science” imply?

3) L149-151: In their response to the choice of differing layer structures employed in the various models, the authors noted that these represented “operational configurations actually implemented in their respective routine air-quality forecasting systems”. Are additional details or references available on these air quality forecasting configurations that interested readers can be pointed to? Even though the impacts of layer structure on the model comparison for surface level pollution are not discernible in the results presented, some readers are likely to be curious about why the models against which EPICC is compared employed a seemingly degraded layer structure/vertical resolution? Also, perhaps it should be noted that even though not employed in this study, harmonizing grid resolutions across models is desirable in any future intercomparison study.

4) L185-196: To my knowledge the CMAQ boundary condition processor does include options to extract limited area LBCs from model outputs on a larger domain – say hemispheric CMAQ simulations, not just the profile file as suggested in this discussion. The impact of LBC specification on simulated surface concentrations is not only confined to areas in vicinity of the boundary but across large geographic domains, since specification of free tropospheric LBCs is influential due to rapid lateral transport at these altitudes and subsequent mixing to surface (see for example: https://doi.org/10.5194/acp-17-12449-2017; https://doi.org/10.5194/acp-18-3839-2018; https://doi.org/10.1021/acs.est.8b02496). Consequently, differences in LBC specification could be contributing to the noted differences between model O3 and PM2.5 at the surface and should be acknowledged in this discussion. Also, it would be useful to provide some comparison of O3 and PM2.5 values at altitudes of the top of the nominal daytime PBL (say ~2km) along the dominant inflow boundary for the 3 models. This may help better gauge the impacts of the different LBC specifications.

5) L359: “daytime particulate nitrate photolysis” is somewhat redundant. Could probably just say “particulate nitrate photolysis”.

6) L642: Please clarify if manganese catalyzed SO2 oxidation is associated with aerosol water?

7) L841-844: Perhaps it is a matter of framing, but in my view representation of natural dust, boundary layer mixing, secondary aerosol formation pathways, and emission input data are all areas of ongoing active research and model development. Thus, “priorities for future model enhancements” may be better wording than “priorities for future model development”.

Hide

ED: Publish subject to minor revisions (review by editor) (22 Sep 2026) by Ulas Im

AR by Qizhong Wu on behalf of the Authors (28 Sep 2026) Author's response Author's tracked changes Manuscript

ED: Publish as is (29 Sep 2026) by Ulas Im

AR by Qizhong Wu on behalf of the Authors (30 Sep 2026) Manuscript

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2026-3428', Anonymous Referee #1, 07 Jul 2026

Please see attached review.

Citation: https://doi.org/10.5194/egusphere-2026-3428-RC1
- AC1: 'Reply on RC1', Qizhong Wu, 19 Aug 2026
  
  The point-to-point responses is at the attached supplement.
  
  Citation: https://doi.org/10.5194/egusphere-2026-3428-AC1
RC2:
'Comment on egusphere-2026-3428', Anonymous Referee #2, 28 Jul 2026
This manuscript provides a comparison of simulated regional PM2.5 and MDA8O3 fields generated by the Emission and Atmospheric Processes Integrated and Coupled Community Model (EPICC-Model) with results from the CAMx and CMAQ modeling systems. Operational model evaluation statistics for all three models, benchmarked against a set of surface observations, are presented to facilitate an assessment of their performance across average spatial and temporal scales. Comparison of EPICC performance results alongside those from established and widely used models such as CAMx and CMAQ is a primary motivation for this study. As expected, statistical metrics averaged over broad spatial (domain-wide) and temporal (seasonal) scales reveal comparable performance among the models, which are fundamentally based on similar process representations.
Establishing and documenting this baseline for EPICC performance relative to other models is valuable for prospective users – thus, a manuscript of this nature will be of interest to the journal readership. Equally critical for the adoption and advancement of a modeling system is the explicit demonstration of its unique strengths and a clear articulation of how its formulation and structural features support its intended applications – an aspect of the manuscript that could be improved further. This could be accomplished by enhancing the clarity of diagnostic evaluation for the EPICC model, thereby highlighting the influence of distinct attributes such as process representation, numerical schemes, and the integration with meteorology and emissions, on the simulation of PM2.5 and O3, relative to CAMx and CMAQ. The following comments and suggestions are offered that may help improve the uniqueness and usefulness of the manuscript.
Throughout the manuscript the EPICC model is referred to as “independently developed”, but it is not apparent to me what that implies? Was the code development independent and if so, independent of what? Is the science in the model not based on other experiences and knowledge accumulated over decades of international research? Looking at the EPICC code, the overall structure of the model bears resemblance to other contemporary systems including CAMx and CMAQ, share similar names for routines, and often also borrows process routines and numerical approaches. Though the individual routines acknowledge derivative work and include some history of the code development, given the similarities between many aspects of the EPICC model structure and code with other existing open-source models, some clarity on what “independently developed” is intended to convey would be useful for the readers.

L31: “maximum biases” should perhaps be “maximum bias”.

L40: some indication of what specific aspect of boundary layer representation improvement is suggested would be useful. As written, the statement is too broad and does not convey much.

L47: “As a typical secondary pollutant, PM2.5 is influenced by primary emissions” is contradictory and awkwardly stated. The sentence should be rewritten to convey that ambient PM2.5 is composed of constituents that have both primary emissions and are secondarily produced.

L48-49: There are more seminal references that establish these underlying processes regulating PM2.5 accumulation and distributions in the troposphere and should probably be included.

L53-54: Again, the dependence of O3 formation on NOx and VOC and on meteorological factors has been well established dating back several decades (e.g., Hagen Smit and Fox, 1956) and should be reflected by the citations.

L54: This statement (use of "despite") implies some dependence on O3 and PM levels which may not always occur because of different trends in their precursors. Perhaps the statement should just convey that in many Chinese urban clusters PM2.5 levels have reduced but MDA8O3 has increased.

L64-68: References for these models should be provided here.

L83: Structural differences in context of models is often used to describe differences in model structure - architecture, code, design, and not necessarily process differences. Suggest changing "structural differences" to "differences”.

Table 1: Though based on Fast-J, Binkowski et al. (2007; https://doi.org/10.1175/JAM2531.1) is perhaps a more appropriate reference for the photolysis approach in CMAQ.

L139-140: To my knowledge there is nothing that specifies 14 layers for CMAQ or CAMx. Why did the authors also not configure CMAQ and CAMx with the so-called "superior" 20-layer configuration or the default WRF layer structure?

L142-143: Though not employed in the current study, it should be noted that for consistency, CMAQ applications typically employ WRF fields also simulated using ACM in WRF.

L151-153: Given that models are continuously evolving, it is important to qualify these statements with the specific version of the models used in the inter-comparison. Though the CMAQ version used here may have been configured with CB6r3, newer versions do include CB6r5 and variants.

L164-166: Default LBC profiles for CMAQ are only provided to facilitate model setup and initial testing. By definition, the LBCs are dependent on the geographical extent and location of the limited area model application. Thus, it is a bit surprising that the authors did not consider harmonizing the LBCs across the 3 modeling systems. Acknowledgement of this somewhat fundamental difference in configuration setting should be provided.

L265: Is the Kz_min setting too low relative to other models or what is physically expected - if the latter what is the basis to support the value? What fraction of the time during the January-May period are stable conditions encountered?

L278-279: This sentence is vague as it covers everything possible and thus not insightful – which process may be the dominant in driving the noted overestimation?

Figure 1 caption: please clarify if daily PM2.5 implies daily-average PM2.5?

L296-297: please clarify what “lower background concentrations” imply. Is the background low or is it that ventilation is strong in the summer and thus concentrations are lower?

L299-301: Based on Figure 1 alone it is difficult to say whether compared to the other two, one model does better in striking better balance between accurately simulating extreme pollution and sustaining strong annual performance, since these are regional average values. Would the same conclusion also hold at individual locations?

L309: Reduced reaction activity during the cooler months arises from both decreased photolysis as well as lower reaction rates associated with thermal chemistry.

Do the scatter plots in Figure 2b show the same data as in 2a, i.e., average across all sites or is each point the MDA8O3 at individual observation location?

L317-319: References should be provided for the heterogeneous HONO formation and nitrate photolysis pathways. To my knowledge, though uncertain, the nitrate photolysis primarily occurs in marine environments (e.g., Ye et al., 2017; Reed et al 2017; Kasibhatla et al., 2018; Shah et al., 2023; Sarwar et al., 2024; Sarwar et al., 2025) or environments impacted by marine air masses, since halide ions are highly surface active and promote the surface propensity of coexisting nitrate anions (DOI: 10.1039/d1ea00087j). How then does it influence inland environments, especially since the large scale O3 in EPICC is expected to be modulated by the MOZART LBCs? Did the MOZART simulations also include nitrate photolysis? How different are the OH concentrations across the 3 models?

L325: Is the isoprene chemistry representation in cb6r3 and cb6r5 that different to impact the ozone forming potential as speculated? It would be useful to further discuss the differences in the isoprene chemistry leading to the noted O3 differences.

L328: An overly deep PBL should result in more dilution and systematic low (negative) bias as opposed to positive bias. I am not sure I understand the rationale here?

L336: “well-controlled absolute bias despite its systematic underestimation” is confusing and somewhat contradictory. If the model is systematically low biased, how can its absolute bias be well controlled?

L345-346: How does one quantify the better balance between overestimation and underestimation. Based on the analyses presented, it is hard to conclude this in any compelling manner. At best the performance is comparable across 3 models with one model performing better than the other two under certain conditions.

L404: It is not apparent to me what “multi-centered and scattered distribution pattern” implies? Is there a specific physical process that drives the Springtime distribution? This discussion would benefit from additional clarification.

L408: Please elaborate on the reasons why coarse vertical resolution reduces mixing? Also, what altitude O3 are the near surface precursors being mixed with? Why does the EPICC layer structure facilitate "better" mixing representation and under what conditions? What prevented CMAQ and CAMx from also being configured with the same layer structure (and implicitly same meteorology fields) as EPICC?

L414-417: The authors suggest that regional O3 levels and differences between models are primarily controlled by photochemical production. How does one rule out the role of large-scale transport as represented by the differing boundary conditions? Also, given that the layer structure of the native meteorological fields and that of the 3 models is different, and also different amongst the CTMs, I do not see how it can be claimed that identical meteorological fields were used? The starting WRF output is the same but it's translation to the grid structures of the individual CTMs (Table 1) is likely different, resulting in not "identical" meteorological fields driving the CTM calculations.

L417: Perhaps "nation-wide O3" is better usage than "national O3”.

L425: What shifts southward - peak values or just relative abundance? What processes drive this seasonal change in distribution and how does the ability of a model to capture it link back to its process formulation?

L430-431: Can just examination of a single IOA metric be regarded as "comprehensive evaluation”? Suggest replacing “comprehensive evaluation” with just “evaluation”.

L486: The underestimation of SO4 may also contribute to the overestimation of NO3 assuming NH3 emissions are well constrained.

L510-511: Since one would expect artificial dilution of emissions over grid cell volumes to likely result in underestimation of concentrations of a primary species, does the BC overestimation imply overestimation of BC emissions or could the representativeness of the monitor locations also influence the noted overestimation?

L570-571: It is not clear to me what either "consistency in species simulations" or "stable performance of RMSE and NMB" implies? Are the authors suggesting that EPICC simulated PM composition is similar to the other models? Both RMSE and NMB in Figure 7 exhibit some spread across models for several species - it is not apparent to me what stable performance of these metrics implies?

L576-579: Full references to the documents mentioned here should be provided.

L582: If heavy and severe AQI are dictated by significant wind-blown dust (WBD) events and if WBD emissions are not included in the models, what is the value of conducting this AQI comparison, since it should be expected that lower accuracy is expected at higher observed AQI. In other words, if the models are not configured for an application, what is the value in "evaluating the applicability" for graded air quality forecasts. Do the measurement locations examined in Figure 6 also report crustal contribution?

L591-596: This is a potentially useful finding but based on the results presented it is somewhat speculative. Based on comparisons of concentrations of secondary inorganic aerosol constituents by the 3 models for selected such episodes, can the authors demonstrate that EPICC indeed predicts different and more accurate "rapid growth" of secondary inorganic aerosols?

L696-700: How does one quantify spatial progression from central to south China to assess which model simulated this progression "best". Figure 11 only enables a qualitative and subjective assessment.

L700: Would be useful to elaborate on how excessively strong aerosol-radiation feedback led to O3 overestimation in southeastern China and on which day?

L701: Visually while CAMx and CMAQ appear to underestimate relative to the EPICC fields, both CAMx and CMAQ also appear to better capture the gradients in O3 - high O3 at monitors interspersed with monitors with low O3 while EPICC depicts more regionally elevated O3.

L706-708: In my assessment Figure 11 in its current form does not conclusively support the claim conveyed in this statement that EPICC outperformed the other two models in capturing the spatiotemporal patterns.

L725-726: I am not sure what "unified WRF fields" implies. Also as noted earlier, given the different layer structures, it is also difficult to ascertain whether the 3D fields are identical or even consistent across models. Thus, probably should be reworded to: "using meteorological fields derived from the same WRF simulation”.

L731: “seasonal turning point” is vague – sentence should be reworded.

L745-746: No comparisons with observed boundary layer heights were presented to support the statement that ACM2/YSU schemes underestimate mixing layer heights. The statement should be adequately qualified.

L800: It is stated that this work “provides a clear pathway for improving next generation of atmospheric CTMs”. I feel this is a somewhat lofty claim as no new or specific process or structural deficiency was identified. There may certainly be specific enhancements identified for the EPICC system, but I am not sure the analysis identified a “clear path” for next generation of models. Suggest deleting this sentence as the study and results do not substantiate it.

L802-804: In my assessment this study at best provides a comparison of operational evaluation metrics of 3 models. There is no specific analysis presented to support air pollution control or policy making. Thus, I feel this statement should be reworded or deleted.
Citation: https://doi.org/10.5194/egusphere-2026-3428-RC2
- AC2: 'Reply on RC2', Qizhong Wu, 10 Sep 2026
  
  Thank you very much for the reviewer's careful and constructive review of our manuscript, and the point-to-point responses is at the attached supplement.
  
  Citation: https://doi.org/10.5194/egusphere-2026-3428-AC2

Peer review completion

AR – Author's response | RR – Referee report | ED – Editor decision | EF – Editorial file upload

AR by Qizhong Wu on behalf of the Authors (10 Sep 2026) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (14 Sep 2026) by Ulas Im

RR by Anonymous Referee #1 (18 Sep 2026)

Suggestions for revision or reasons for rejection

Hide

RR by Anonymous Referee #2 (21 Sep 2026)

Suggestions for revision or reasons for rejection

Hide

ED: Publish subject to minor revisions (review by editor) (22 Sep 2026) by Ulas Im

AR by Qizhong Wu on behalf of the Authors (28 Sep 2026) Author's response Author's tracked changes Manuscript

ED: Publish as is (29 Sep 2026) by Ulas Im

AR by Qizhong Wu on behalf of the Authors (30 Sep 2026) Manuscript

Journal article(s) based on this preprint

12 Oct 2026

Evaluating the EPICC-Model for regional air quality simulation: a comparative study with CAMx and CMAQ

Mengjie Lou, Qizhong Wu, Wending Wang, Huansheng Chen, Kai Cao, Xiaohan Fan, Dingyue Liang, Fenfen Yu, Jiating Zhang, Wei Wang, and Zifa Wang

Geosci. Model Dev., 19, 9651–9676, https://doi.org/10.5194/gmd-19-9651-2026,https://doi.org/10.5194/gmd-19-9651-2026, 2026

Short summary

Mengjie Lou, Qizhong Wu, Wending Wang, Huansheng Chen, Kai Cao, Xiaohan Fan, Dingyue Liang, Fen Fen Yu, Jiating Zhang, Wei Wang, and Zifa Wang

Supplement

https://doi.org/10.5194/egusphere-2026-3428-supplement

Mengjie Lou, Qizhong Wu, Wending Wang, Huansheng Chen, Kai Cao, Xiaohan Fan, Dingyue Liang, Fen Fen Yu, Jiating Zhang, Wei Wang, and Zifa Wang

Viewed

Total article views: 318 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
186	103	29	318	86	36	42

HTML: 186
PDF: 103
XML: 29
Total: 318
Supplement: 86
BibTeX: 36
EndNote: 42

Views and downloads (calculated since 29 Jun 2026)

Month	HTML	PDF	XML	Total
Jul 2026	49	28	5	82
Aug 2026	20	12	7	39
Sep 2026	61	27	9	97
Oct 2026	56	36	8	100

Cumulative views and downloads (calculated since 29 Jun 2026)

Month	HTML	PDF	XML	Total
Jul 2026	49	28	5	82
Aug 2026	20	12	7	39
Sep 2026	61	27	9	97
Oct 2026	56	36	8	100

Viewed (geographical distribution)

Total article views: 291 (including HTML, PDF, and XML) Thereof 291 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 31 Oct 2026

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (22038 KB)
Metadata XML

Short summary

This study compares the model performance of the Emission and atmospheric Processes Integrated and Coupled Community Model with the two air quality models in China. The EPICC-Model excels in simulating summer ozone peaks, accurately captures pollution characteristics in highly polluted areas, and better reproduces persistent compound pollution processes. The study reveals common issues among the models and directions for improvement, providing a basis for optimizing global air quality models.


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