General comments:

The paper titled “Elucidating ozone and PM_2.5 pollution in Fenwei Plain reveals the co-benefits of controlling precursor gas emissions in winter haze” by Lin et al. evaluates the status of ozone and PM2.5 pollution in a typical megacity of the Fenwei Plain, one of the most polluted regions in China, which reported a general trend of increasing secondary pollution (ozone and SOA) in winter haze, and the causes of this trend and the possible measures in controlling the complex pollution by O3 and PM2.5 were further studies and discussed. With this, the authors claimed that the co-benefits of reducing NOx and VOCs simultaneously in reducing ozone and SOA, that would be also suitable for other polluted regions of China suffering ozone and PM2.5 currently. The manuscript was well written and presented clearly. Therefore I recommend the publication of Lin et al. work after some issues were properly revised and improved.

Specific and technical comments:

1. Method, more details in the calibration of PTR-MS should be provided. In addition, What kinds of VOCs species were used in the standard mixture? Please list the VOCs species that calculated from the kinetic rate constant, and the uncertainty on the calculated VOCs should be discussed.
2. Line 109-112, it is better to provide more details for NR-PM2.5 monitored by an AMS which usually measured NR-PM1. I note that a novel PM2.5 was firstly equipped with AMS for the winter campaign in 2014 (Elser et al., 2014), It is unclear for the other winter campaigns.
3. Line 121-122, why the reduction in NO₂ for the observation sites was not used? Which would be more precisely than the satellite image.
4. Line 144-146, please list the VOC/VOCs information that used as input data for box model. I note that HCHO was not used to constrain the model, how about the other OVOCs? Considering the OVOCs was also from secondary formation. In addition, I am concerns on the model performance in the ozone simulations, as the majority of alkanes was unavailable in the model if only the VOC/VOCs measured by the PTR-MS. As least, the authors should provided more details in the performance of the box model and the analysis in the uncertainty.
5. Line 200-202, I do not agree that the secondary formation could be the major source of formaldehyde, as the measured and modelled formaldehyde showed different diurnal pattern. The similar level may suggest large uncertainty in the modelled formaldehyde.
6. Line 245-247, the significant reduction in primary fossil fuel OA (77%) from 2012-2014 to 2019-2021 could be expected, due to the implementation of the clean air act in 2013. The more magnitude of reduction in cooking OA (84%) is interesting, more evidence should provided and discussed here.

Review on “Elucidating ozone and PM2.5 pollution in Fenwei Plain reveals the co-benefits of controlling precursor gas emissions in winter haze” by Lin et al.

The study looks into the pollution patterns, sources, and formation mechanism of PM2.5 and ozone. The analysis of 7 years data reveals the severity of air pollution. The author found that increased ozone was due to the constantly elevated reactive OVOCs and the reduced NO₂, and then stimulated the increase of particle pollution. A 0-box model was applied to investigated the co-benefits of reducing NOx and VOCs simultaneously in reducing ozone and SOA. Finally, the atmospheric implication helps for developing cost-effective mitigation policies in the future. The results are important for the scientific community to increase their understanding of the O₃-PM_2.5 interaction. The paper is well written, and the literature is broadly cited. I recommended a minor revision before publication.

Comments:

1. Line 40: the font size needs to be constant
2. Line 115: More detail of the 5 datasets should be provided in this paper, including the sampling time in a year. Since the PM2.5 and OA obviously vary in different seasons, the correction or uncertainty analysis should be presented in the paper.
3. Line 134-136: What is the advantage of developed random forest model compared to previous de-weathered RF model? mean meteorological variables ? To average the meteorological data at a specific time point during each year?
4. Line 142-148: The detailed species of VOC/OVOCs, instead of only top 10 species, should be provided, since these precursors influence the results of the 0-box model. The input data are averaged diurnal profiles or the total time series? If the averaged diurnal data were applied, how the uncertainties of the model change? More importantly, the verification of the 0-box model is missing. For example, the comparison between measured and predicted concentration of ozone, which was the main object in this study.
5. Line 167-170: It seen that only 1-2 days difference found between the two period may not always support the authors’ conclusion. What about the ozone exceedance in each year during 2015-2021? Or what about the uncertainty/standard deviation for the averaged data?
6. Line 175: The reference or topographic map for the topography favoring the build-up of air pollutants should be provided.
7. Line 177-178: The BLH highlights very obvious difference between afternoon and other time period. The author could correct the PM2.5 by BLH to verified whether the reduced emission appears or not.
8. Line 191: (Sect. 21) ? or Sect 2.1
9. Line 233-234: Figure 4？
10. Line 242-244: More information of OA source apportionment by the PMF in this study (2021) should be provided to make the conclusion robust, referring to Feng T, et al. Atmos. Chem. Phys., 2023, 23: 611-636.
11. Line 246: How do the authors explain the large reduction of COA? Dose it mean that the cooking frequency is lower than before or the cooking method change or other reasons?
12. Line 290-291: The self-reaction between peroxy radicals can produce SOA, but the reaction between NOx/NO3 and peroxy radicals has the same effect. The branch ration between these two pathways can influence the level of increasing or decreasing SOA formation. More explanation could add here.

Comments for Figures

1. Figure 2: Deep and light color were applied for figure a/c and b/d, but detailed legends should be provided.
2. Figure S2: The x-axis labeling is ambiguous. 2015-2021?