The manuscript entitled “Insights into the abnormal increase of ozone during COVID-19 in a typical urban city of China” by Kun Zhang et al. explored the drivers of elevated ozone concentration during COVID-19 lockdown. The manuscript provides valuable information for understanding ozone chemistry under rigorous emission reduction measures and efficiently directing ozone mitigation in the future. I would recommend publication if my following concerns are well addressed.

General comments:

(1) VOCs were measured by a PTR-TOF-MS in this study. However, this method cannot measure alkane and most alkene species, which will underestimate the ozone production and could mislead the diagnosis of ozone sensitivity regimes. Therefore, some uncertainty analysis regarding this deficiency is necessary.

(2) Temperature and solar radiation increase rapidly from Pre-lockdown period to Full-lockdown period, which could significantly contribute to the increase in ozone concentration during Full-lockdown period. This influence is not fully considered in the manuscript. Relevant analysis is also suggested to be included.

Specific comments:

Line 28: “the observed O3 “should be changed into “the increase in the observed O3”.

Line 34-35: Here, the authors describe that the changes in precursor emissions (or NOx/VOCs ratio) contributed 2.4 ppbv to the O3 increase, which is inconsistent with 5.1 ppb in lines 27-28. Please double check it.

Line 58-59: also include actinic flux in meteorological conditions and cite the papers (1, 2).

Line 175-176: The influence of RH on ozone is very complicated. Higher humidity is conducive to OH production and thus likely increase O3 production. I suggest to add some references about RH influence here and simulate the influence of RH on ozone by box model.

Line 177: please explain why RH>70% can be an indicator of adverse weather conditions.

Line 192: What does the r² represent? You should state out it in Section 2.4.

Line 195: I think “O_3,Obs” should be “O_3,Normal” here.

Figure 3 and Figure 12: In figure 3, O_3,Normal during Full-lockdown period is higher than that during Pre-lockdown period by 12 ppb. However, the corresponding value is only 2.4 ppb. Please explain this inconsistency.

Figure 3: My understanding is that the deweathered method normalizes the influence of meteorological factors on the difference between the same periods in different years. Were meteorological factors between different periods also normalized? The authors should clearly explain this in Section 2.4. This is important to figure out the influence of meteorological factors on ozone increase during Full-lockdown period compared to pre-lockdown period.

Line 214-215: I suggest to at least give some evidences that the decrease in VOC is due to the decrease in industrial activities and traffic volume. Besides industrial activities and traffic volume, solvent usage is also an important source of VOC.

Line 266-268: The expression is ambiguous here. Acetaldehyde and formaldehyde don’t belong to aromatics.

Line 273-274: “As for alkene, this could be explained by their chemical reactivities, which led to the fast degradation after emission.” I don’t agree with this statement as aromatics tend to have similar chemical reactivity as alkenes.

Line 277-278: This could also be due to enhanced solar radiation and temperature from January to March.

Line 281-282: “This result suggests that the VOCs in Partial-lockdown should produce less O3 than that in Pre-lockdown, and Partial-lockdown period”. This is misleading. First, the former “Partial-lockdown” should be Full-lockdown. Second, the MIR that you calculate here refers to the ability of VOC species composition to produce ozone, is it correct? Thus, I suggest to further explain the concept of MIR and change this sentence into “This result suggests that VOC species composition in Full-lockdown is more conducive to ozone formation………”. In addition, please specify each dot represent 1-hour average or 24-hour average in Figure note.

Line 304: “Feb 14^th ” should be “Feb 1^st”.

Line 325-327: I suggest to explain the reason why OVOC kept stable among the three cases, which is inconsistent with the remarkable difference in measured OVOC among the three periods as you shown in Figure 4.

Line 330-355: PTR-TOF-MS is unable to measure alkanes and most alkenes, which could influence the diagnosis of ozone sensitivity to precursors. Lower VOCs concentrations lead to more VOC-limited regime. I suggest to provide uncertainty analysis about it.

Figure 2: The legend of different parameters at the top of the Figure should be placed in corresponding sub-panels. Besides, the legend of ozone and TVOC is not given at present.

Section 2.2: How were photolysis frequencies been considered in the model? Were they constrained by photolysis measurements or calculated by a radiative transfer model (e.g., TUV)? If they were calculated, what about the uncertainty compared to the real condition? And what would be the influence on the afterwards data analysis?

Line 375-376: “underestimation of ozone sensitivity to alkanes and alkenes” should be “underestimation of ozone production from alkanes and alkenes”.

References:

1. Wang et al., The impact of aerosols on photolysis frequencies and ozone production in Beijing during the 4-year period 2012–2015. Atmos. Chem. Phys. 19, 9413-9429 (2019).
2. Wang et al., Exploring the drivers of the increased ozone production in Beijing in summertime during 2005–2016. Atmos. Chem. and Phys. 20, 15617-15633 (2020).

This study targets an important question, what causes the ozone increase during lockdown despite substantial decrease in anthropogenic emissions? By applying some statistical approaches, the authors decouple the effects of changing meteorology and emission on ozone formation, and reported that changes in emissions causes a 5 ppb increase in ozone during the lockdown, where changes in meteorology conditions only increase ozone by 0.5 ppb. Further, it is shown that the ozone formation shifts from a VOC-limited regime before lockdown to the conjunction of NOx- and VOC-limited regime, which increase ozone formation. Overall, the scope of this study fits the journal. I recommend publication after major revisions.

Major Comments

1. Several statistical methods are applied in the study, but it is not clearly stated why they are selected? For example, why Sen’s slope is used rather than a simple linear regression? There is a myriad of machine learning algorithms, so that the rationale behind each selection should be discussed. For example, Sen’s slope is a robust slope and less susceptible to outliers. Further, current description of deweathered model lacks details. What does the model do? If I understand correctly, it takes several parameters as inputs and use random forest to predict O3 concentration, right?
2. The interpretation of O3,met and O3,emission is confusing, partly because of lack of details in describing the stats methods. To the reader, the difference between observed O3 and weather-normalized O3 represents the influence of changing emission, as weather-normalized O3 takes into account the variation in O3. The difference in observed O3 between different years does not represent the influence of emission, because the meteorology between different years is different. Such interpretation will fundamentally change the conclusion on this manuscript as well as the conclusions from the box model. Please clarify.
3. The discussions on ozone formation potential (OFP) can be reconstructed in a more meaningful way. Mainly, it should be clearly stated that OFP does not indicate O3 concentration. With this premise, there is no need to discuss “consistency” or “inconsistency” between the two (Line 282). In other words, OFP is not helpful to answer the O3 question in the manuscript.
4. The reliability of the box model results is compromised by the fact that the modeled O3 during full lockdown (29 ppb) is lower than that during partial lockdown (32ppb), which contrasts the observation.

Minor Comments

1. Line 167 is a confusing sentence.
2. Line 257. “Vary”, not “varies”.
3. Line 281. It is “full-lockdown”, not “partial-lockdown”.
4. Line 299. What does “AOC” represent?

The subject manuscript by Zhang et al. presents an analysis of ambient measurements from the Yangtze River Delta assessing the observations of increased ozone (O₃) levels during the COVID-19 lockdown period. Results are presented for three periods: pre lockdown, full lockdown, and partial lockdown and are compared with the same time periods in 2019. The authors seek to understand the relative importance of precursor volatile organic compounds (largely grouped by compound class), nitrogen oxides (NOx), ambient reactivity/oxidation capacity, and meteorology in determining ozone levels.  The authors motivate the study well, and adequately convey the importance of measurements and analysis in this region. However, the methodology is not sufficiently described to support the results and conclusions, and the results and conclusions are somewhat difficult to follow as written. Specific questions, comments and suggestions on these points follow below. It is recommended that this manuscript be reconsidered for publication after major revisions.

Technical Comments:

line 34, lines 173-176: Using “supposed to” in this sentence does not necessarily reflect the complexity of O₃ formation. The prior sentence suggests that during the full lockdown period, the region shifted to a NOx-limited regime. Thus, it may be expected that a greater decrease in NOx relative to VOCs would lead to a decrease in O₃. However, there is also the role of NOx titration, in which decreasing NOx can lead to an increase in O₃. Later in the manuscript, this phrasing is repeated and it suggests that the authors are not necessarily referring to the NOx regime in the abstract, but the influence of meteorology (specifically T, RH). Again, this is not clear in the abstract, and oversimplified as written. What is the mechanism by which RH affects O₃ formation? How sensitive is O₃ to RH?

Similarly to the use of “supposed to”, the authors need to clarify “improper” decline (line 202) and “abnormal” increase of O₃ (line 203).

PTR-TOF-MS measurements (p. 5): Does the Jensen et al. companion paper address losses in the inlet and to the filter? How might these losses affect the results of the analysis presented here? The authors do not need to provide all of the details presented in Jensen et al., but should summarize the main findings, including any limitations, that are relevant to the analysis presented in this manuscript.  

Trend analysis: The authors state that the MK non-parametric test is recommended by the WMO. The authors should provide some additional detail here. What does the WMO recommend this test for? Under what conditions? What are the limitations/requirements for applicability in the context of this work? How is serial correlation applicable to the PTR-TOF-MS measurements of individual VOCs? What details of Pathakoti et al. and Alhathloul et al. are relevant here?

Deweathered model: While details of the VOC measurements are somewhat lacking, and more so for the trend analysis, this section is entirely lacking of sufficient detail (and is not, as the authors note on line 191, described in section 2.4). The authors should consider that the information in the manuscript needs to be sufficient such that the results can be reproduced. Further, it is difficult to assess the robustness of the results when sufficient details about the methodology are not provided. What are the uncertainties of the approach? Are data available for all parameters over all time periods? How are missing data handled? How were the model parameters determined (number of tress, minimal node size, and number of samples)? How sensitive are the results to these model parameters?

In general, in the results and discussion, it is often difficult to follow whether the authors are describing results between the three periods in 2020, or between given periods in 2019 and 2020. It is recommended that the authors try to more clearly differentiate these comparisons.

line 214: TVOC dropped to 22.19 ppb from what mixing ratio?

lines 218 and 234: The authors use “interesting” in these sentences, but it is not clear what is interesting about the observations as presented. The higher mixing ratios due to lower boundary layer heights (line 218) is a common observation, and the lower values of transportation-associated VOCs (line 234) during the lockdown is expected (and has been reported previously).  

In line 233, is the decreasing trend based on the Z-score or Q value? Were these metrics consistent? Why or why not? It might be useful to include the Z-scores and Q values for all compounds in the SI.

line 224-226: Can the authors be more quantitative about how many of the measured VOC species shown exhibited this U-shape pattern and then explicitly list those VOCs that didn’t? It is a little contradictory to say “most” and then “except for several”.

line 269: Is it expected that biogenic emissions would be the dominant source of O₃ in this region?

lines 273-275: This discussion of alkenes is not clear as written. The chemical reactivities of compounds is tied to their oxidation formation potential (and is not independent of).

line 279: How are the MIR values calculated? What are they dependent on? Is it expected that the MIR would be reflective of the different NOx/VOC regimes? I’m not sure this is the case.

line 286: Was NOx eliminated (which suggests some chemical/physical removal)? Or were the emissions reduced to a greater extent that VOC emissions?

line 326: What is the relevance of the stable OVOCs across the lockdown periods in the context of emission sources/anthropogenic activity? Is there any offset between emissions and chemistry during this period?

Editorial Comments:

In general, there is somewhat inconsistent introduction and use of abbreviations. It is suggested that the authors check the manuscript to insure all abbreviations have been defined in the abstract and first (and only first) appearance in the manuscript.

line 33: suggest to replace “conjunction” with “boundary”

line 49: I believe “Jan” should be “Feb”.

line 104: suggest to replace “inhaled”, maybe “air samples were drawn into a 3 m tube” or “air was sampled through a 3 m tube”

line 123: I believe “0-zero” should be “0-D”

Eq 1: The Zhu et al. reference should be replaced by Geyer et al. 2001 (10.1029/2000JD900681), which is where this representation was introduced previously.

line 150: “indicating” should be “indicate”

line 151, line 260: “specie” should be “species”

lines 153-154: Meteorology, emissions, and chemistry (though not entirely independent) can all affect ozone. It is suggested that this be rewritten as: “influenced by meteorological conditions, emissions and/or chemistry”. The authors could then indicate that emissions and chemistry are being treated together and separately from meteorology.

line 155: “was” should be “were”