The authors developed an observation-based method to investigate the ozone production efficiency and ozone production sensitivity to precursors for two persistent ozone episodes in Guangdong, China, based on the hourly surface O₃, PM_2.5, CO and NO₂ data at 77 stations in Guangdong during the period 2018-2019. They also performed a box model constrained by ambient conditions observed during the two episodes for comparison. They find 67% of the station-days exhibit ozone formation sensitivity to NOx, which differs from other previous studies which suggested that limiting VOC emission rather than NOx would be more effective in reducing ozone in Guangdong, and these results are in semi-quantitative agreement with the results calculated by box model. The authors make some arbitrary assumptions and simplifications in derivations of VOC and OH concentrations, which is a major weakness of the current work. I had a number of specific comments for the authors to consider and address before publication.

Specific comments:

1. Specific information on the ratios of VOC/CO that is used in this study for the derivation of VOC is better added in the SI. Also, the uncertainties due to this treatment of VOC on OH concentration should be discussed.
2. Line 80-81, the author uses the same way to evaluate the leftover VOC as to evaluate the leftover CO in the following day. However, VOCs can continue to be oxidized by OH and NO₃ in the afternoon and at night. How great are the effects of this neglect of the depletion of leftover VOC on the derived VOC and OH concentrations in the following day?
3. Line 81-82, the authors state that the oxidized VOC are estimated from the observed ratios of HCHO, CH₃CHO, and ketone to CO in Wang et al., 2016, and other OVOCs are not included. What is the basis for this treatment? Besides, HCHO, CH₃CHO, and ketone are not only photochemical products of VOC oxidation but also from direct anthropogenic emission. How do the authors deal with the difference in emission-related origin of OVOC among different locations? Given that OVOCs typically make large portion of OH reactivity, the estimations of OVOCs are crucial for the simulation of OH concentrations. How large are the uncertainties of these assumption on the predicted OH concentrations?
4. Line 85-112, the derivation of OH concentrations and calculation of oxidized VOC and NOx in this study are all based on the Lagrangian condition assumption, which rarely exists in the real atmosphere, so the authors make a selection criterion to filter out days satisfy the quasi-Lagrangian condition. What is the basis of this selection criterion?
5. Line 122-126, why the product of the average OH at noontime and the mean NOx in 13:00-16:00 can be used as the hourly NOx emission rate between 08:00 and 13:00?
6. Line 311-315, these sentences are totally a copy of the sentences in line 78-83, and do not provide any useful information, and I would like to suggest the authors to delete these sentences, and provide more useful information about the uncertainty analysis.
7. I would like to suggest the authors can review other literatures reporting the ozone production efficiency in PRD areas to strengthen the discussion.
8. Considering that all the VOCs data are estimated based on the ratios of VOC/CO in the emission inventory, I think it is not appropriate to name it an observation-based method.

Observation-based Analysis of Ozone Production Sensitivity for Two Persistent Ozone Episodes in Guangdong, China

In this paper by Song et al., the authors use 2 episodes in Guangdong, China and a large number of measurement sites in the vicinity to construct an observation-based method (OBM), utilizing the measurements of various pollutants from said sites alongside a box model based on the CB05 chemical mechanism, with the purpose of determining ozone production efficiency (OPE) from NO_x and VOCs. They conclude that the area is under a NO_x limited regime, indicating that limiting NO_x emissions is the optimal strategy to reduce ozone formation in the area, contrary to previous studies.

While the paper does have its strong points – the analysis is thorough, the English used is clear and appropriate – it is not without shortcomings, many of which reviewer #1 covered. The paper merits publication based on the rigor of its analysis, but not that of the conclusions. As such I recommend the paper for publication only after the following points have been addressed and the discussion strengthened.

Science comments

1. As the authors mention two episodes are not enough. In addition, they are well into the ozone season in the fall, which could further bias the results. For example, biogenic emissions of VOCs are going to be significantly less than what they would be during the summertime, which could tip the balance of the OPE. A section should be added to discuss the potential differences between summer and fall months. The box model the authors have developed can be used, driven with meteorological variables from the observation sites during different seasons (if available), to investigate
2. Based on (1), the usage of CO to VOC ratios, while a valid strategy for anthropogenic emissions completely neglects possible biogenic impacts and thus is better suited towards the urban sites much more than the rural sites. In addition to the current analysis, it would be of value that the authors also conduct the same by splitting the sites in rural and urban which would be more representative
3. I second reviewer’s #1 comment about the NOx quasi steady state. This would only apply from 13:00 to 16:00. Using the average OH value for the early day is not accurate
4. The calculation of OPE assumes that the only real sink of NO_x is the ozone formative chemistry. However, NO_x is also lost to other processes and in an area like deposition and nitrate formation. The deposition is briefly mentioned towards the end, but some additional discussion and/or an estimate of the magnitude of the effect should be provided. Given the close proximity of ports in the area and therefore the likely high emissions of SO₂ and subsequent sulfate formation, the additional NO_x sinks could be of an important magnitude. I do realize that such an analysis would be out of the scope of the paper, and I do not require that authors conduct it, but some additional discussion on the matter is warranted, given the number of assumptions already used. On that note, particularities of Guangdong should be added to the introduction e.g., nearby ports, major highways, nearby agricultural activities etc.

Editorial comments

1. The timeseries of meteorological parameters is more suited for the SI. Use the diurnal profiles instead in the manuscript, so the reader can directly go back and forth with the diurnal concentrations to clearly see the dilution effect due to the PBL
2. While I do understand why Figure 8 was added, and it holds a lot of valuable information, it would be best to either omit it or add it to the SI. Figure 9 is more appropriate, and it would be even better if you turn it into 2D plots with variable marker sites
3. The combined site isopleth could use some polishing; fill out the contours. Also, I very strongly recommend that you make isopleth for each of the observation site clusters from Figure 1. This also feeds into point 2 from the science comments above