General comment

The paper reports a modelling study of the contribution of shipping to NO2 and O3 concentrations in the Mediterranean area. In addition, deposition is included in the analysis. Results of different models are compared among themselves and with measurements in some specific stations. The topic is interesting and the paper generally well written. It has elements of novelty and I believe that it could be published after a revision step necessary to clarify some aspects and put results in a better perspective, see my specific comments.

Specific comments

Lines 31-32. It is not clear this sentence. The other two models use different meteorological input?

Lines 41-43. I suggest to mention the recent work of Contini et al (Atmosphere 2021, 12, 92.) that gives a global overview of the effects of shipping on air quality and health.

Line 140. Please remove etc. if authors want to add something it is better to do it explicitly.

Line 243. Actually, looking at the map in Fig. 1 it seems that it is included also the major part of eastern Mediterranean.

Section 2.5. It should be mentioned how these stations have been chosen and if a certain threshold of distance from the coast or from the main routes of ships. This because it is known that the impact of the emissions from ships to air quality is strongly depending from the distance from the harbours/routes and I see some stations that are quite inland, especially in Northern Italy. A discussion on this should be provided even because I believe that the impact of shipping on such stations would be really small.

Another aspect that should be clarified and it is partially correlated to the previous point is if the emission dataset used include emissions of ships at berth. Several studies indicated that in EU harbours the emission at berth lead to the majority of the impact on local air quality in port cities, see for example Merico et al. (Atmospheric Environment 139 (2016) 1e10). Considering the use of low sulphur fuels at berth, this phase is particularly relevant for nitrogen oxides and could also lead to local exceedances of air quality standards. If neglected it could be present an underestimation of the impacts.

Line 336. A correlation with R=0.06 is not weak, rather it is a total absence of correlation.

Lines 410-415. I believe that the results here are also comparable with those obtained with CAMx in the central/eastern part of Mediterranean area reported in Merico et al (Transportation Research Part D 50 (2017) 431–445).

Lines 420-421. I would add or near the harbours.

Figure 6. Please use the apex for m3 as in the other figures.

Line 446. The absence of negative values in the tagging method is a consequence of how the method is formulated rather than a relevant result. Could this lead to problems in evaluation titration of O3?

Section 3.1.5. Regarding O3. There are several experimental evidences, some of them also in the papers that I already mentioned in my previous points, that emission of NO from ships could lead to a local reduction of O3 concentrations, especially in the spring/summer period. At larger distances instead there could be an increase. There are also some hypothesis that models could catch this behaviour more or less efficiently according to the spatial resolution of simulations. Could this be an issue in your results considering that outcomes ranging from negative to positive impacts for O3 were observed? Also Figure 16 shows that relevant differences are observed especially in the negative part.

Line 656. I would not say air pollution considering that only NO2 and O3 are considered in this work.

Lines 714-719. This part is a little vague. It could be useful to understand if there is any possibility to understand if one of the model performs better than the other. In addition, it should be mentioned how to use results from the different models with different resolution, averaging the results?

The manuscript of Fink et al., 2022, investigates the effect of shipping emissions on NO2 and O3 in the Mediterranean Sea. Results of five different models are compared which used the zero-out method. In addition results of the tagging method in LOTOS-EUROS for NO2 are presented.

The topic of the manuscript is of high relevance as shipping emissions are an important source of pollution in the Mediterranean Sea. In the current state, however, I can’t recommend a publication in ACP. To my opinion the following major points need to be adapted:

1) The biggest issue of the manuscript is the used terminology. By definition, contributions can not be calculated by the zero-out method (at least not for non-linear species), but by the LOTOS-EUROS Tagging method. I recommend to use the terminology (e.g. potential impacts and contributions) of the „Source apportionment to support air quality management practices“ from FAIRMODE (https://fairmode.jrc.ec.europa.eu/document/fairmode/WG3/European%20guide%20SA_3.1_online.pdf). The clear terminology is important as the different methods (tagging, zero-out) focus on two different scientific questions. Zero-Out shows the change of e.g. ozone in case of an emission reduction. Tagging gives the contribution to ozone for the ‘reference state’. Due to the different aspects of the two methods it is important to have a clear terminology to avoid any misunderstandings. A lot of literature exists on this topic for further reading, e.g.:

Wang et al., 2009, Grewe et al., 2010, Emmons et al., 2012, Butler et al., 2017, Clappier et al., 2017, Mertens et al., 2020, Belis et al., 2021, Thürkow et al, 2021, Rieger & Grewe, 2022

2) The paper is very long. The authors tried to explain some of the (large) differences of the model results, but many differences remain unclear. As an example, the model results for Ox look very different (e.g. compare EMEP with e.g. CAMx). Also the results for deposition differ largely (as noted by the authors), but there are no further analyses. On p39l606 the authors note that this could be due to different dry deposition velocities, but the velocities itself are not analyzed. I understand that due to the multitude of effects and the differences between the models itself it is almost impossible to find a reason for the large spread between the models. However, in this case I suggest to reduce the amount of information (e.g. also the length) of the paper by presenting the most important findings only. This could be for example the impacts of the shipping on NO2 and O3 as simulated by the different models (and a short chapter to deposition). Also, for example, the time-series of the different models at the different stations (e.g. Figs 2 – 4) are interesting, but also very lengthy. The figures and their discussion could be moved to the supplement and ‘only’ the summarizing evaluation could be presented in detail.

In addition, I suggest to better highlight/focus on what we can learn from a policy point from this study? Where are open questions? What did you lean from the multi-model study which should be considered in follow up studies? Should more things be harmonized? Where do models need to be improved?

3) I like the idea that the different models were applied in their ‘default configuration’ and only resolution and anthropogenic emissions are prescribed. However, the description of the models differ strongly in their level of detail. Some examples:

- For CAMx no information about the biogenic emissions are given;

- Information about sea salt emissions are only given for CMAQ, EMEP and CAMx;

- Information about dust emissions are only given for CMAQ and EMEP.

Similarly, the description of dry- and wet-deposition differs (for example EMEP in Sect 2.1, for all other in Sect 2.4). Information about lightning NOx are missing completely.

I suggest to give the same amount of information for all models in the same level of detail. I would also suggest to expand Table 1 with details about the chemical mechanism, the used dry deposition scheme, biogenic emissions etc. Finally, I further suggest to add tables with total emissions (especially for biogenic sources) for each model to the supplement. It would also be nice to see all computational domains in the supplement.

In addition, I noticed that for all models the figures for NO2, O3 etc. (e.g. Fig. 7) show slightly different geographical regions. This contradicts with Fig. 1 and the information about a common domain. As example, Fig 7 (e) does only partly show the Po Valley while Fig. 7 (b), (c) and (d) show the Po Valley completely. For better comparability the same geographical region should be displayed for all models (and of course should be used for calculating mean values, frequency distribution etc.).

4) I suggest to replace the color scales. The rainbow color scale can be misleading. In addition it is problematic for people who are colorblind. You can check your plots for example with a ‘CV Simulator’ on you phone. Also some of the labels at the figures are very small. I suggest to use at leas the same font size as in the figure caption.

Minor comments:

p6l155: Is there something missing in this sentence (boundary conditions from Mozart44 output were activated?). But more importantly, if CAMx OSAT output is available why not discuss this in the manuscript? To my opinion the paper would benefit from including OSAT results.

P8l233: I am not familiar with LOTOS-EUROS, but does this mean that the model time step is 1 hour or should it read ‘hourly model output' ?

p10l257: This does mean that the NMVOC split was not adjusted to the chemical mechanisms of the individual models, right? No lumping of species were performed?

P10l264f: The part about the VOC emissions is unclear to me. Please rephrase. Thanks!

P20l402: You mention the longer lifetime of NO2 for CAMx and CHIMERE. I wondered if HNO3 mixing ratios of the models differ. Please add figures in the supplement and discuss them shortly.

Fig 6: Please don’t use the tagging results for calculating mean impacts. Tagging and zero out give something different (see main point 1).

p24l448: See also (1) – To my opinion the main reason zero-out gives different results (and results from different sensitivity simulations do not add up) is the non-linearity of the chemistry. Of course other factors also lead to differences.

P26l495: Please provide figures of the different boundary conditions in the Supplement.

P32l516ff: I don’t understand this sentence. How should a split of the emissions lead to high concentrations over sea and low concentrations over land? I guess the main reason is the low dry deposition over sea, right? (as well as the overall higher land emissions).

Figure 18: The label for the subplot should be contribution frequency distribution? Please check also for all other figures.

P46l679ff: Please see main point (1) above. There is also a lot of literature discussing zero out vs. tagging which could be cited here.

P47l691: Is there any answer on the question of and how the different dry deposition can explain the model differences?

Technical comments:

I found some typos and missing spaces etc. Please double check the manuscript. Some examples:

p3l88 differences, p4l103 % by

p13l334 – Should be Table 3?

p36l580ff (and throughout the whole manuscript): I suggest to replace ‘output’ with model results or similar

p46l661: The output was quantified? I guess it should read the differences of the model results was quantified or the impact of shipping simulated by the different models was quantified.

P46l673: the maps display – In my opinion the model results display (please check also the manuscript for similar wording as the term ‘maps’ have been used quite often)

Literature:

Belis, C. A., Pirovano, G., Villani, M. G., Calori, G., Pepe, N., and Putaud, J. P.: Comparison of source apportionment approaches and analysis of non-linearity in a real case model application, Geosci. Model Dev., 14, 4731–4750, https://doi.org/10.5194/gmd-14-4731-2021, 2021.

Butler, T., Lupascu, A., Coates, J., and Zhu, S.: TOAST 1.0: Tropospheric Ozone Attribution of Sources with Tagging for CESM 1.2.2, Geosci. Model Dev., 11, 2825–2840, https://doi.org/10.5194/gmd-11-2825-2018, 2018.

Clappier, A., Belis, C. A., Pernigotti, D., and Thunis, P.: Source apportionment and sensitivity analysis: two methodologies with two different purposes, Geosci. Model Dev., 10, 4245–4256, https://doi.org/10.5194/gmd-10-4245-2017, 2017.

Emmons, L. K., Hess, P. G., Lamarque, J.-F., and Pfister, G. G.: Tagged ozone mechanism for MOZART-4, CAM-chem and other chemical transport models, Geosci. Model Dev., 5, 1531–1542, https://doi.org/10.5194/gmd-5-1531-2012, 2012.

Grewe, V., Tsati, E., and Hoor, P.: On the attribution of contributions of atmospheric trace gases to emissions in atmospheric model applications, Geosci. Model Dev., 3, 487–499, https://doi.org/10.5194/gmd-3-487-2010, 2010.

Mertens, M., Kerkweg, A., Grewe, V., Jöckel, P., and Sausen, R.: Attributing ozone and its precursors to land transport emissions in Europe and Germany, Atmos. Chem. Phys., 20, 7843–7873, https://doi.org/10.5194/acp-20-7843-2020, 2020.

Thürkow, M., Pültz, J., and Schaap, M.: A mitigation study for air pollution management across Germany for NOX (NO + NO2) with the LOTOS-EUROS CTM – Part I: Comparing the labeling and brute force technique for source attribution., EGU General Assembly 2021, online, 19–30 Apr 2021, EGU21-5862, https://doi.org/10.5194/egusphere-egu21-5862, 2021.

Rieger, V. S. and Grewe, V.: TransClim (v1.0): a chemistry–climate response model for assessing the effect of mitigation strategies for road traffic on ozone, Geosci. Model Dev., 15, 5883–5903, https://doi.org/10.5194/gmd-15-5883-2022, 2022.

Wang et al., 2009 https://doi.org/10.1029/2008JD010846