Nault et al. describe O₃ production and its individual contributors in the Seoul Metropolitan Area based on airborne measurements with the NASA DC-8 aircraft during the KORUS-AQ campaign in 2016, as well as box model simulations using F0AM. The authors highlight three important aspects, which are the VOC reactivity, the production of HO_x and the branching ratio of alkyl nitrates. A particular focus is put on the impact of unmeasured (O)VOCs, affecting underestimated peroxy and alkyl nitrates, and in turn deviations in NO_x and radical sinks.

The paper is well written and interesting to read. I have some remaining questions and comments (see below). Once these are addressed, the paper would be a valuable contribution to the literature and I recommend it for publication.

Major Comments

Does “unmeasured VOC” refer to species that are neither measured, nor represented in the model?

Was Eq. (1) or Eq. (9) used to calculate P(O_x) throughout the study? Could you present a comparison between the results of the different approaches?

Lines 99 ff.: I have some questions regarding the calculations presented in the Supplement:

• Line 44 (Supplement) / Eq. S2: What about the reaction of CO with OH? HO₂ is formed without going through RO₂? Does this need to be accounted for? Depending on the location / altitude, I would expect that HO₂ could be up to a factor of 2-3 higher than RO₂.
• Figures S1b: It would be helpful to show the equation that presents the relationship between P(O₃) and P(HO_x) as well.
• Eq. S7 / Figure S1c: Do I understand correctly that Eq. S7 is used as a basis to create Figure S1c? It looks like that O₃ production is approximately halved when increasing the branching ratio α from 0 to 10%. However, this is difficult to understand when looking at Eq. S7. The rate constants for HO₂ and RO₂ with NO are similar (k(HO₂+NO) is a bit higher), and you assume that HO₂ ≈ RO₂. Therefore Eq. S7 could be simplified to P(O_x) ≈ (2-α) * k * [HO₂] [NO]. Shouldn’t P(O_x) decrease by only a few % for α=0.1? Maybe it could be clarified how Figure S1 is developed / what causes the large impact on O₃ production.

Lines 173 ff.: Airborne NO₂ measurements are a challenge, particularly in the presence of peroxy nitrates, because they can decompose in the instrument (where we usually find higher temperatures than those of the ambient air) (Reed et al. (2016), Shah et al. (2023)). Usually, this problem arises at higher altitudes, but if you expect large amounts of PNs this might have a bias on the NO₂ measurements. Was this investigated? How well does the measured NO₂ and the PSS calculated NO₂ agree? Maybe a comparison of measured and calculated NO₂ beyond the NO₂/NO ratio (e. g. in the Supplement) could strengthen your argument.

Lines 224: Why do you use the box model calculated HO₂ instead of the measurements? Maybe you could present a comparison of modeled and measured HO₂?

Line 238: Could this also include airport NO_x emissions?

Lines 272 ff.: Are these differences significant? What’s the uncertainty of the individual shares?  

Lines 311 f.: Does this mean that one go through the HO_x cycle produces only 1.53, instead of 2 O₃? Does this in turn mean, that only 1.53 NO molecules are involved? Could you explain the role of CO and HCHO in more detail?

Lines 337 – 352: This section is a bit hard to follow. Could you clarify how R(VOC) is determined? Is Eq. 11 needed to understand Figure 4? Maybe it would make sense to present Eq. 11 earlier in the text?

Line 466 ff.: Could you elaborate a bit further on how the competition between R8 and R9 relates to formaldehyde?

Lines 577 ff.: Are Figures 6(b) and (c) created using the box model or the observations?

Minor Comments:

Line 84: Is there a word missing? “One important subclass of VOCs are (?) aldehydes…”

Figure 3: The Figure caption mentions panel (c) instead of (b).

Line 341 / Figure 4b: Do you mean “α using Eq. 10”?

Line 568 f.: There seems to be something wrong in this sentence. Can you rephrase it?

Literature:

Reed et al. (2016) https://doi.org/10.5194/acp-16-4707-2016

Shah et al. (2023) https://doi.org/10.5194/acp-23-1227-2023

Nault et al. present an intriguing dataset which shows that the oxidized NOx budget measured during KORUS-AQ includes alkyl (AN) and peroxy nitrates (PN) that cannot be explained by the emissions and chemistry represented by common chemical mechanisms. The authors use these observations to show that a significant source of VOC reactivity [R(VOC)] is needed to explain observations of OH reactivity and potentially close the NOz budget. The authors show that this missing chemistry has an important impact on radical production and loss rates, and thus predictions of ozone formation. The authors assess potential sources and suggest that aldehydes from cooking and other oxygenated VOCs could explain this missing chemistry.

I found the study very interesting and the authors provide number of useful constraints to assess the role of understudied chemistry impacting the air quality in Seoul. I think this is a valuable contribution to characterization of urban air qualtiy. I have a few comments below that I hope will help to strengthen the discussion.

Major Comments

Lines 373 – 397: I think the comparison of F0AM R(VOC) to PSU measurements convincingly shows the missing chemistry in the model. In my opinion, the extrapolation to higher NOx using equation 11 is a bit speculative and not really necessary to make the authors point. While I appreciate that there is a lot of discussion about the uncertainty in this approach, I’d suggest leaning into the observations and comparison to previous studies as written at lines 400-410. To my eye, the missing R(VOC) and the modeled VOC distribution are not drastically different at low NOx and high NOx.

Discussion of PANs: The authors note that the model overpredicted PAN by a factor of 2 and that this could be related to the assumed background, dilution rates, and/or temperature Do the authors have a sense of the major cause for this discrepancy? My concern is that if this is mostly affected by temperature, then the higher PNs would also be affected and bias the model/measurement comparison shown in Fig 5a. Can the authors provide some sensitivity analyses to determine how much each factor might affect net PAN production?

I would also like to note that the discrepancy on PAN could also be due to uncertainties in the ethanol constraint. Since ethanol is a large component of the PAN budget (Fig 5c), it would be worth exploring an emissions sensitivity.

Finally, the GT mass spec reports PBzN. Since PBzN has a more limited number of precursors, it might be worth comparing the F0AM output to this species. This might help rule out the impact of processes that would affect most PNs (e.g., temperature) from those that are specific to individual PN species (e.g., emissions).

Minor Comments

Line 46: This is a bit contrary with sentence 33 - 34 where increases in O3 seem to be attributed to decreases in NOx in a NOx-saturated ozone production regime. Perhaps clearer to say "in altering" O3 production?

Line 61: from?

Line 62: “product” should be plural.

Line 84: add “are” before aldehydes

Line 165: Is there another description for the blocking period other than "blocking conditions"

Line 182: For each "aircraft observation" is this 1-min data, or some average of urban plume intercepts? The term "diurnal cycle mode" is unclear. Which species are being evaluated for convergence? Is this meant to be O3, HCHO, or other specific species? And is the model moved forward in time based on the aircraft observations until Obs = Model? I realize this is in Schroeder et al., but some small details here would be useful.

Equation 5: This equation represents primary HOx and photolysis of HOx reservoirs. Are there other sources that might matter in this formulation, or is this the majority of species that contribute to HOx production in the model? What about photolysis of higher-peroxides formed by RO2 + HO2? These are treated as a loss of HOx, but wouldn't their photolysis contribute to HOx as well?

Ultimately, I'm wondering what fraction of the HOx production term is represented by these measures. My understanding is that this can be derived from F0AM by summing up all the terms that produce OH and HO2 and excluding reactions that interconvert HOx (e.g., NO + HO2)

Lines 224-225: This seems out of place without also knowing that HO2 was measured onboard the aircraft. Perhaps clarify by noting that "HO2 calculated from F0AM, rather than aircraft measurements (Crawford et al. 2021), is used in the equation to determine the Ox and HOx budget." Is there a reason for using modeled HO2 as opposed to observations?

Line 264: Are there studies other than Woolridge that also show closure? It seems that the number of urban sites reported in that study are limited, and the disagreement observed in the key urban study (PIE in Boulder, CO) is attributed to poor inlet design. It would be helpful if other studies were cited here.

Lines 274-275: At any point, do the measurements (either in NOy or the sum of individual NOz components) begin to reach detection limits? May be useful to note which measurements are still above detection limit at this point?

Line 311 – 312: How exactly is y calculated? Is this a model simulation output (e.g., the yield of ozone over the full oxidation of the molecule) or is this inferred from the mechanism branching ratios? And how is this weighted for every VOC? Some details here (perhaps with an equation) would be helpful to understand this calculation.

Line 326: I presume that the units of the Ox/sumANs is ppb / ppb?

Line 337-339: It would be helpful to see an equation for how a_eff is calculated from the observations. It seems that the authors are not including the intermediates in the calculation of a_eff. Are the F0AM intermediates a negligible component?

Line 415: Cooking emissions are rich in a suite of long-chain aldehydes (e.g., Schauer et al. 2002) and so it may be better to discuss these emissions as a group. Coggon et al. (2024) recently showed that C2 – C11 aldehydes from cooking are present in ambient air but not well characterized in ambient datasets. It would be more encompassing to say “one possible group of missing VOCs are long-chain aldehydes from cooking and vegetative emissions, including nonanal.”

Line 418-419:  Did the PTR observe nonanal? It is also possible that the interference could be significantly impacted by a C5-aldehyde, which almost completely undergoes fragmentation to produce the isoprene signal (Buhr et al. 2002). Thus, if nonanal is not super abundant, other aldehydes (or cycloalkanes) could be contributing and not easily detected at the proton-transfer product.

Line 430: I suggest rewriting to say "nonanal and other long-chain aldehydes" may be an important PN precursors.

Line 457: The oVOCs discussed up until this point were mostly associated with aldehydes and emissions associated with cooking. What about solvent sources, such as VCPs? According to McDonald et al., these emissions may contain significant contributions from glycols and alcohols, which aren’t very well measured by PTR or GC.

Lines 481 – 486: I was confused in this section. At line 481, it reads as though the authors are calculating higher PNs from the model by subtracting PAN, but then in the following sentences it is noted that PAN was excluded from this analysis. Could this be clarified?

Line 514: PBzN is reported by the GT CIMS. Can these species be compared?

Line 515: It would be worth noting that PHAN formation may be overestimated in the MCM. Butkovskaya et al. (2006) present evidence showing that the RO2 radical formed from glycolaldehyde + OH (i.e., the PHAN precursor) decomposes to formaldehyde and CO2. This pathway could compete with other RO2 pathways and limit PHAN production. Other studies (e.g., Magneron et al. 2005) note that PN’s were not observed in glycolaldehyde + OH oxidation, and this decomposition pathway could be a possible cause.

Lines 535-547: Ethanol is in most VCPs. I would expand this statement about cleaning agents to include all VCP sectors.

Lines 592-594: Please point towards Fig 6c in this sentence.

Figure 5: The legend in Fig 5b is confusing. Some species refer to PNs while others refer to precursors of PNs, which I presume are then lumped together for that specific precursor (correct?). Perhaps clearer to say "monoterpene-derived PNs, isoprene derived PNs..." or some acronym (MONOPN, ISOPN...etc).

References:

Buhr, K., van Ruth, S., and Delahunty, C.: Analysis of volatile flavour compounds by Proton Transfer Reaction-Mass Spectrometry: fragmentation patterns and discrimination between isobaric and isomeric compounds, International Journal of Mass Spectrometry, 221, 1-7, https://doi.org/10.1016/S1387-3806(02)00896-5, 2002.

Butkovskaya, N. I., Pouvesle, N., Kukui, A., and Le Bras, G.: Mechanism of the OH-Initiated Oxidation of Glycolaldehyde over the Temperature Range 233−296 K, The Journal of Physical Chemistry A, 110, 13492-13499, 10.1021/jp064993k, 2006.

Coggon, M. M., Stockwell, C. E., Xu, L., Peischl, J., Gilman, J. B., Lamplugh, A., Bowman, H. J., Aikin, K., Harkins, C., Zhu, Q., Schwantes, R. H., He, J., Li, M., Seltzer, K., McDonald, B., and Warneke, C.: Contribution of cooking emissions to the urban volatile organic compounds in Las Vegas, NV, Atmos. Chem. Phys., 24, 4289–4304, https://doi.org/10.5194/acp-24-4289-2024, 2024.

Magneron, I., Mellouki, A., Le Bras, G., Moortgat, G. K., Horowitz, A., and Wirtz, K.: Photolysis and OH-Initiated Oxidation of Glycolaldehyde under Atmospheric Conditions, The Journal of Physical Chemistry A, 109, 4552-4561, 10.1021/jp044346y, 2005.