The authors present analysis using three years of seemingly superb measurements from an excellent measurement site. The conclusions drawn from measurements of NO, NO₂, jNO2, O3, CO, and several VOCs is that there are “missing oxidants” that convert NO to NO₂ in the air, and that are not accounted for by past peroxy radical measurements. I recommend it be published after the following major and minor concerns are addressed:

1. The detailed model used is only as good and accurate as the inputs (i.e., compound concentrations constrained by measurements), and as impressive as the long-term dataset is, it does not include oxygenated VOCs. As such, it does not seem fair to expect that the models could accurately simulate the actual photochemistry given that it is likely not adequately constrained. Please include a discussion of the impact of unmeasured VOCs, especially oxygenated VOCs. Also please be clear what is meant by the term “missing” – are there reactions missing in the chemical mechanisms?
2. The analysis needs a more quantitative handing of the uncertainties. In particular, what is the uncertainty of the calculated quantity [NO₂]PSS-ext? (based on its constituent parts in equation III). For example, in line 343 of the manuscript. See also another comment below regarding the stated measurement uncertainties in Table 2 which require improvement. In numerous places it refers to older peroxy radical measurements and explains that those measurements are highly uncertain, especially at high RH. What are those uncertainties – both as stated in the original papers, and as concluded by the authors today?

Detailed comments

Abstract

Line 29 “…implying 18.5-104 pptV (25th-75th percentile) of missing RO2 radicals”  - the term “missing RO₂ radicals” is unclear. Please clarify as “…of RO₂ radicals missing from photochemical models”.

Line 32: “If the missing RO2 radicals have an ozone production efficiency equivalent to that of…” The term ozone production efficiency is traditionally defined as the number of ozone molecules produced per NOx molecule. Please use a more accurate and defined term for what you mean in the abstract.

Line 34 (same sentence): “then the calculated net ozone production including these additional oxidants is similar to that observed”

The term “net ozone production” is unclear. Do you mean net ozone production rate (ppb/hr)? or does it mean “net ozone produced”, which would be in # of molecules, or possibly mixing ratio (ppb)? Furthermore, it is confusing to refer to the “observed” ozone production rate, since nowhere in the abstract is it explained how that was “observed”. Does “observed” actually man “calculated based on measured quantities”? Please clarify.

Line 37 “and that measured and modelled RO2 are both significantly underestimated under these conditions.” This is the first reference in the abstract to measured RO₂ and as such is quite confusing. Later in the paper it becomes apparent that it is referring to past measurements of RO₂ at this site. Please clarify.

Body of manuscript

54: “Under very polluted conditions, where O3 is the only oxidant converting NO to NO2” – I disagree with that statement. There are plenty of very polluted conditions in which there are plenty of peroxy radicals present that also convert NO to NO₂ (e.g., Mexico City, Los Angeles…). This would be better phrased as “Under conditions in which O3 is the only oxidant converting NO to NO₂, …” and can clarify that perhaps the are referring to time periods with low sunlight and very high NO (I assume)

72: the equations would be much easier to read if more subscripts were added. i.e., rather than jNO₂[NO₂], write as j_NO2[NO₂]

86: “However, PSS-derived ROx concentrations are generally higher than both measured and

modelled values in rural conditions” – the wording can be tricky and sometimes confusing. The term “modelled” is confusing, since use of the PSS to derive ROx concentrations is in itself a simple model.

116-117: “However, more recent instruments use “cavity absorption phase shift (CAPS)”  - that should be attenuated rather than absorption, and probably wise to add “spectroscopy” or “spectrometry” afterwards.

124: “… the increase in HO2 wall loss on wet surfaces” – humid surfaces, not wet surfaces. “Wet” implies there is a fair amount of liquid water on the surface (rather than a possible thin layer of adsorbed water).

Lines 123 onward describe in detail the sensitivity of chemical amplifiers to humidity and specifics of the RO₂ being sampled. It appears that the main point of this section is to point out that these measurements are not perfect and subject to uncertainties. This is true of course, just as it is for measurements of all compounds. The resulting concentrations and stated uncertainties produced by chemical amplifiers ideally reflect the issues discussed in the text (RH dependence, dependence on organic nitrate and nitrite formation…). I recommend that this section describing RO₂ measurements by chemical amplifier conclude with a summary of the uncertainties of those measurements as described in the referenced papers. If the authors feel that the measurements are even more uncertain, they should state so explicitly. This might be especially important given that the peroxy measurements were made over 20 years ago.

The last sentence of the paragraph could easily be left off, since similar statements apply to all analytical measurement techniques: “It is therefore important to determine the optimal concentrations of reagent gas for each individual instrument as it could vary with what material has been used in the reactor”. Similarly, it is important for each chemiluminescence instrument to use the proper ozone concentrations and flow rates, and for HOx LIF instruments to operate with the correct laser settings, NO flow rates….etc.

141: “The production of O3 (P(O3)) can be calculated using equation (VI)” insert the word rate after production

157: “In regions where the net O3 production is negligible or negative” again this is ambiguous wording, especially in light of the above note regarding the same term “net O3 production” (line 34). Please define what is meant by “net O3 production” – the rate? The change in O3 concentration over time?

Line 159 and 177: O3 should be [O3], or written as “O3 concentration”

Line 180: define what is meant by “photochemical regime”.

181 onward, and Table 1: Although later in the text the authors do a good job evaluating the possible interferences in the Chemi-photolytic converter technique, it is noteworthy that all almost all of the NO₂ measurements from Table 1 were made with chemiluminescence and a photolytic converter. The only study that used cavity ring-down spectroscopy (Tadic et al. 2020) appeared to find agreement between ROx(PSS) and ROx(model).

181 – 189: “The large uncertainties associated with ROx measurements, especially at high humidities…” again, the authors really need to include the stated uncertainties from the chemical amplifier measurement papers themselves, and if they believe that the true uncertainties are higher, then they should state so. By how much higher would the uncertainties need to be to have agreement with ROx(model) or ROx(PSS)? Furthermore, is “high humidity” defined as greater than 80%, say, or greater than 50%? What is the range of humidity values observed during daytime at this site?

Table 2: The “accuracy” column is very confusing. For NO, NO₂, O3, CO, and CH4 an absolute mixing ratio is listed (e.g., 4.4 ppt), but for all the VOCs, a percentage is listed. The NO and NO₂ values undoubtedly need an accuracy listed in percentage, presumably determined largely by the calibration methods. Perhaps the 1.4 ppt and 4.4 ppt for the NO and NO₂ are actually the 1 sigma precision values? For what time averaging interval? The value for O3 seems erroneously low – 0.07 ppb! Please fix. The uncertainty of these measurements is crucial given their use in equations II and III.

Section 3.1.1: given the detailed treatment of the NO₂ measurement artefact, it would be useful to include either a spectrum of the blue LEDs or to simply state its spectral width (FWHM).

Line 261-262: “If NO2 is the product then it will be photolysed to NO with the same efficiency as NO2 in the ambient air” This does not seem correct, as for an interfering compound it’s a two-step process and thus the NO₂ formed will have less exposure time to the UV radiation (e.g., X --> NO₂ --> NO, rather than NO₂ --> NO). An interfering compound that is converted to NO₂ in the photolysis cell should have a lower efficiency at making NO than NO₂ does.

264: “Organic nitrates, HNO3, and NO3 do not photolyse at 385 nm and have therefore not been included in the evaluation of photolytic artefacts” Is this true for all organic nitrates?! There are many kinds – alkyl nitrates, hydroxy-alkyl nitrates, peroxy acyl nitrates…

Line 273: “making it highly likely that a significant fraction of HONO is lost on the manifold before the air is introduced to the NOx instrument due to the high surface reactivity of HONO (Pinto et al., 2014)” What is the manifold made of? Glass? Teflon? If it’s Teflon, then the quoted section seems like an overstatement. Have loss rates of HONO on surfaces been presented in other studes? Pinto et al 2014 appears to have little to say about surface losses and does not conclude that surface losses played a big role in that comparison study.

331: Both of the references which provided the RO₂+HO₂ measurements by chemical amplifiers (Hernández et al., 2001 and Burkert et al., 2001) were from 21 years ago. Do changes in background NOx and O3 affect the context of their inclusion in figure 2?

343: “Daily midday values of [NO2]PSS ext were calculated using equation III” What is the combined uncertainty of [NO2]PSS ext? Note that this is an important area where the uncertainties of the past chemical amplifier measurements can be addressed quantitatively, as it is part of equation III. This is a crucial area of revision.

Line 361: “the abundance of NO on …” although the term “abundance” is commonly used synonymously with “concentration”, I advise against it in this case as NO molecules were anything but abundant!

Andersen et al. use long-term (years) measurements of NO_x, O₃, organic compounds and associated parameters from a remote marine sampling location to evaluate understanding of radical chemistry affecting NO₂/NO ratios and ozone production. This topic is of wide interest as radical chemistry is central to understanding global oxidation processes, and many studies have failed to explain the observed NO_x partitioning in a variety of chemical environments. Strengths of the work are uniqueness of the dataset, and analysis using GEOS-Chem and a detailed chemical box model to evaluate the chemistry.

Overall, I think the paper is well written, provides an excellent review of and links to the prior work on this topic, and has interesting analysis. I think that the paper will deserve publication but that the authors should first consider a few important points concerning the limitations of the measurements and modeling analysis and how that might affect the way that the conclusions are stated.

General comments:

1) The primary conclusion of the paper is that the NO₂/NO ratio observations are consistent with the expected NO->NO₂ oxidants in the cleanest conditions, but more polluted air masses would require significantly more organic peroxy radicals or halogen oxides to explain the observed NO₂/NO ratios. This is first stated in the paragraph beginning on line 354. I am not convinced, however, that there is a clear difference in the behavior between the more pristine and more polluted air masses. In other words, it is not clear to me that one can say the cleaner data definitely are completely explained by the known chemistry whereas the more polluted data have a different behavior. I think a more thorough discussion of the uncertainties of each data point due to precision or artifact uncertainties would help the interpretation of the figures.

For example in Fig 3B while the scatter of data at NO2 < 20 ppt are hard to distinguish from the 1-1 line, I would not say by eye that the overall trend there is different than at the higher NO2 mixing ratios. In Figure 5, while enhancements in acetylene and ethane are associated with higher than expected NO2, the data with low acetylene and ethane do not cluster around a value of 1 for NO2_obs / NO2_pss, but appear to have significantly lower than expected NO2. I did not see discussion of the lower than expected NO₂ observations. In Figure 6 while the CO < 90 ppb data are centered around a value of 1 for NO2_obs / NO2_pss, many of the points are not close to one. Is the width of the histogram explained by the precision of the measurements or is it possible that some of the width here is also evidence for incomplete understanding of the chemistry?

2) As I understand it from this paper and Anderson et al. 2021, a potential positive artifact on the NO₂ measurement from the photolytic converter is assumed to be negligible (Anderson et al., 2021 state that measurements of zero air show 0 – 10 ppt of NO₂, which is assumed to be real NO₂ in the zero air). While I understand the problems/challenges with experimentally determining if there is a real surface artifact, I find it concerning that the potential for a positive artifact in the NO₂ measurement due to illumination of species on the walls of the photolytic converter is assumed to be zero. It is well documented that typically a positive NO signal of at least a few ppt will be generated by illuminating such converters (even quartz ones) even in the presence of synthetic, NOy-free air (e.g. Gao et al., 1994, Pollack et al., 2010, others). Can the authors please comment in the artifact section in some way on this? What would the impact be if there were a few ppt of fake NO₂ from the converter? Perhaps the lowest measured NO₂ could be used at least as an upper limit of such an artifact. Are there other upper limits that can be stated for such an artifact?

3) I suggest that the authors put a bit more emphasis/discussion on the good agreement shown in Fig. 7 between measured and calculated ozone tendency. It could be argued that this is more important than being able to reproduce the NO/NO₂ ratio, and therefore remaining uncertainties or discrepancies in observed vs calculated NO₂/NO are less important to resolve since the ozone tendency seems nicely explained.

Specific comments by line:

Line 60: Suggest defining RO₂ as ‘organic peroxy radicals’ rather than just ‘peroxy radicals.’

252: Recommend using the symbol s rather than defining the ACS acronym.

260: Can you state the width of the LED spectrum?

277: While GEOS-Chem may not show a coherent seasonal pattern for NOy, clearly there is a lot of real variability that is likely related to airmass origin, and higher NOy is probably related to pollution sources. PAN for example could matter. Could you comment on the origin of the variability in GEOS-Chem? Perhaps adding a timeseries of CO to Fig. S7 would be helpful.

292: The GEOS-Chem timeseries of PAN (S7) which seems to be routinely above 20 ppt would suggest that if GEOS-Chem has some skill here the PAN would be above this 6 ppt detection limit frequently, or always. Can you comment on this?

322: Since the calculation of RO2 is critical to the argument of the paper, it would be helpful to see more information about the relative importance of these measured RO₂ precursors. Is there any correlation between the calculated RO₂ and the pollution indicators? Do the authors think that the missing RO₂ sources could be due to VOCs that are not measured by the GC system at CVAO? If the air is of African origin and possibly influenced by biomass burning, can the authors comment on how sufficient the measured suite of VOCs might be in comparison to recent those reported in more recent papers with comprehensive measurements of biomass burning VOC emissions? Overall, I’m a bit unsure if ‘missing’ is the right word to use to describe the unaccounted for RO₂, or rather that we should expect there are a number of important organic compounds that were not measured.

460: I would say that the required additional factor for XO is higher than that of RO2 not because of the difference in rate coefficients, but because the measured/calculated XO is << measured/calculated RO2.

Figures

Fig1: please provide a colorscale and explanation. Does each point represent the calculated location of an air parcel 10 days prior to arrival at CVAO?

Fig2: Would be nice to mention in the caption the seasons of those campaigns.