The manuscript presents the high O3 episodes in two heatwaves in Germany, attributing the O3 levels to NOx and VOC sources within and outside Germany using the TOAST system in the WRF-Chem model. The manuscript is very well written, well-structured, and easy to navigate. The tables and figures are relevant and to the point, well-supporting the discussion of the results. There are few points that should be addressed before the manuscript can be published in ACP.

• The emissions are based on the TNO-MACC III for the year 2015. Are the emissions used for the 2018 simulation kept the same or somehow scaled? Information on 2018 anthropogenic emissions is lacking.

• There is no mention on the dust emissions or if dust is used at all in the simulations, ie from the boundaries. This might be important to consider since some periods are associated with southerlies with pressure systems starting from Africa. Dust can alter the radiation in the atmosphere, which can impact the chemical reactions such as O3 formation. Using an online-coupled model has the advantage to account for these feedbacks and not using dust emissions or particles from boundaries can partly be a reason for biases in temperature and O3 levels.

• It would be good to describe why the NOx and VOC attribution simulations use different sets of source regions. Is it a simplification based on the complexity or computationally as VOCs included many lumped species?

• Temperature plot in Figure 1 does not show overestimations at nighttime as suggested in the manuscript (Line 216).

• Figures 5-8. Add that the dots represent the observed O3 concentrations.

• In general, it would be beneficial if the contributions are not only given in absolute terms but also in relative terms (percentages) so the reader can see if the impact is large or small.

Attribution of surface ozone to NOx and VOC sources during two different high ozone events

by Lupascu et al.

The manuscript is written in a clear language and well structured. The relevant literature is presented in the introduction. The limitations of the approach are described and discussed. The results are presented and discussed with reference to previous studies. However, the figures reporting the results should be improved because are difficult to visualize. A discussion about possible steps to deal with the model underestimation compared to observations would enhance the impact of the study.

Specific comments

Line 70: OH radical

Line 75: There are two Butler et al. citations in the reference list. The authors should distinguish them with a letter.

Line 178: Please clarify the abbreviations of the VOC regions used in Figures 7 and 8

Line 210: Please, explain what temporal resolution is used for the comparison (hourly, daily, other?)

Line 228: Please, comment if there are significant differences in the performance indicators between urban and rural sites.

Line 306: More work is needed to understand the causes of the underestimation

Line 319: the influence of stomatal resistance on the model performance should be further explored

Section 4.3.1: The authors should discuss more the stratospheric ozone intrusion observed in Northern Germany in this section.

Line 359: What do you mean by small? Please, quantify

Line 368: What to mean by polluted air? polluted with O3 or with other pollutants?

Line 381: Please, report the share here

Line 427: In the figures can be observed diurnal cycles but not easy for the reader to compare the exact timing between O3 from NOx and VOC

Line 472: In Figure 9 total O3 and the one deriving from GER and ROE are plotted, not clear to which one your refer when you speak of O3 from biogenic sources. Please, be more specific

Results: it would be useful to add a figure summarizing the overall results where both absolute and relative source contributions are reported.

Figures 5 to 8: it is a good idea to combine the graphs of all the German regions in one single figure. However the details are hardly visible and the text/ scales are too small. It is necessary to find a way to summarize the info. My suggestion is to show only the average values for each region in the combo figure and to connect the plots with a line to the relevant region to ease the visualization (reading codes is awkward).

Conclusions: include a statement on what actions should be taken to deal with the overall model underestimation of observations.