Summary

This paper reports new measurements of organic nitrate concentrations and reactivity, alongside box modelling, in order to constrain the lifetime of alkyl and peroxyalkyl nitrates. The analysis uses measurements made during the ACROSS campaign in 2022, at a tower sit in a Rambouillet forest, 50 km southwest of Paris, France. The major result is an apparently very short (deposition) lifetime for these organic nitrate species over the forest, due to their low observed concentrations. The modelling applied to interpret the data is rather complex. Because no speciated monoterpene data was available, different MT mixtures were tested, including a cutoff correction to treat the effect of (frequent) thermal inversion periods on the MT mixture. The complexity of the modelling made the analysis a bit difficult to follow and given the goal of determining lifetimes, I wonder if it could be made simpler / clearer. Of larger concern are the surprisingly high inferred contribution of OH to nitrate formation at night (when OH contributions are typically considered to be low), and the 4 x model overestimation of XO2 concentration relative to measurements. Could there be biases in these OH and XO2 measurements that skew this analysis substantially?

General Comments:

1) You use the term “peroxycarboxylic nitric anhydrides” throughout for PAN. I understand this is a correct IUPAC name, but I think “peroxyacetyl nitrate” is more commonly used in atmospheric chemistry community – might make your paper clearer understood (and more likely successfully searched for) if you use the latter name?

2) Section 3.1: You refer to the measurement site as a tower in a clearing, which to me sounds like no trees directly around the tower, but later it’s clear that your inlets are partly shaded (you mention the difference in photolysis rates if you were to use the above-clearing radiation measurements). Were they in the woods? Or are the trees so tall that even faraway trees shade your inlets for substantial parts of the day? I think a photo or diagram of the site might help here. And sometime when you talk about inside or outside the clearing, do you mean on the tower above the treetops? Some clarifying language could help here.

3) in line109 you introduce the variable k^NO3, and in line 117 you say you will henceforth refer to this as k^BVOC because BVOC losses are dominant. But later you introduce additional k’s to describe the losses to various oxidants, so it’s confusing to lose the oxidant here. Why not k^NO3+BVOC, to be completely clear?

4) around line 120: was O3 measured on the same inlet line? (the diagram of relevant sampling instrumentation would avoid readers wondering things like this)

5) Because your OH and XO2 results are so surprising, I think the descriptions of the OH and XO2 measurements in section 3.2.2. are key. Can you add some discussion of potential interferences or reasons for non-zero backgrounds? And explain how the XO2 is determined subtractively from the NO converted channel minus no NO channel, and what the associated uncertainties are. Is there any possibility of additional reactions between the conversion reactor and measurement? (diagram helpful here too)

6) around  line 153, where you talk about the neglect of upwelling radiation which could cause an underestimate of 5-10%, it would be helpful to know a bit more what the canopy is like. (again, perhaps a diagram + photo figure would help here). How shaded is the container? Maybe add a plot comparing the top of tower radiation to that at the measurement container, and show the percentage difference over the daily cycle.

7) around line 165: At what height was the PTR-MS sampling? (again, diagram would help). Maybe discuss this in the context of your MT composition estimates – are you measuring at the same position as the reactivity or a different height, and might there be gradients?

8) around line 185: the limonene fraction sensitivity analysis is rather confusing. It’s not clear why the specific 3 mixtures are chosen, and in Figure 3B it doesn’t look like the fraction of limonene matters much. What matters is that most of the time, you actually use the delta-T cutoff and assume you don’t have any limonene at all in the reactivity mix.  What motivated the choice of delta-T = 1 for the cutoff? This feels strangely arbitrary, and in figure 3 you see that the cutoff case seems to apply to most of the data. Why not just omit limonene altogether and make your discussion less complicated?  Or at least, if the limonene does affect the low concentrations / daytime modelling, why present the 3 scenarios and not just your best estimate of the daytime vs. limonene fraction? This brings a lot of extra discussion that may confuse your main message.

9) towards the end of section 4.2: How did you partition XO2 into RO2 and HO2 to determine how to weight the RO2 / HO2 rate constants? Or you could just show the reaction as k_XO2 and explain how you estimated that aggregated rate constant.

10) In section 4.4 where you mention mixture 2 again, could just state the assumed mixture to avoid confusion (why 2?), and if you keep the delta-T threshold, remind the reader of how that works briefly here –e.g., “temperature inversions and resulting assumed removal of all limonene at the surface.”  Also, somewhere early in this section say that you report all errors as +/- 1 sigma, and then don’t repeat it after every number. At line 333, you mention mixture 2 again – state that at the outset somewhere that it applies to all further analysis, so you don’t need to keep interrupting the text with it. Readability gets compromised by these repetitions, and mixture 2 could get confused with phase 2, etc.

11) The very high simulated OH contribution to PANs at night is quite surprising, so it would be good to see how else you can corroborate this. Does your box model reproduce the measured high nighttime OH concentration, with a reasonable assumption of OH recycling from the O3 reactions? (I assume that has to be where it’s coming from). Or could there have been some other (spurious?) local source of OH?

12) Lines 378 – 384: the 4 times higher modeled XO2 than measured is also shocking. It doesn’t seem satisfactory to just say the origin is not known and then use the higher modeled value. Can you look more into what the highest individual XO2 concentrations are in your model and try to figure out whether there could be some erroneous pathways or some that might not be relevant to your site?

13) Lines 480-483: Why do you bring up HPAN here and not when you introduce all the other reactions? Also, the use of an “arbitrary loss term” is not well justified. Why not just omit HPAN formation from the model, if they have never been observed and you have to make them go away with an arbitrarily high loss term to get things right?

14) Mid-page 13:  You describe the differences in daytime vs. nighttime physical loss rates based on vertical mixing, but it sounds (here) like in your model you use one physical loss rate. At this point I’m wondering, why not account for the different daytime and nighttime loss rates by tuning them separately? But then on the top of p. 14 you describe separate lifetimes for daytime and nighttime. I’m a bit confused whether these refer to separately tuned parameters in a single model, or the optimized (full-diurnal-cycle) parameter in cases where you tune to match daytime or nighttime better. I suggest a careful edit of the last 3 paragraphs of this section (lines 484-527) to sharpen and clarify the description of what you did to obtain the day and night lifetime estimates. The table is a helpful addition, but the text is still a bit difficult.

15) Figure 5 now confuses me again about your MT scenarios. Limonene does not appear to be a fixed 20% of MTs during the day, when you are not switching off limonene. Is it 20% of emissions in your model, not concentration? You discussion made me think you had partitioned by concentration, but this looks like it contradicts that.

Minor/ technical corrections and suggestions:

Line 35: format both lifetimes the same, ie., either x +/-y, or xx – yy

Line 51: dependent, resulting in

Line 68-69: (SOA), thereby

Line 84: large variety of instruments were

At lines 105-106, you might already mention that no NO3 or N2O5 results were used in this study, because they were all below DL?

Line 124: suggest to mention briefly (even though you cite a reference) generally what NO measurements were corrected for.

Line 178: fractional contributions (a_i) from different monoterpenes as described in equation (2), where k_NO3+I is the rate coefficient

Line 238: Is NO3 photolysis to NO and O2 really an important atmospheric reaction?

Line 252: (eqn 3) error in superscript should be PNO3ANS

Line 282: missing “i” index under summation

Around line 296: I think it would be clearer if you laveled the fraction alpha^RO2+NO rather than just alpha^RO2

Line 344: daytime and nighttime from those

Line 392-393: MACR), methyl vinyl ketone (…, MVK), and a-pinene, after multiple

Line 397: In summertime forested environments, where

Line 408-410: For PAN, MPAN … depends on the concentration of NO, … XO2 designates the sum

Line 422: (R23) from 7.5 hours at 283 K to 40 minutes

Line 427: an important role, depending

Line 431: plots the measured mixing ratio for sumPANS (=PAN + MPAN + PPN + other PANs)

Line 434: campaign result in a …magnitude, from 15 hours at 279 K to 3 minutes at 314 K as shown in Figure 7B (i.e, remove the parenthetical)

Line 440-441: Figure 7), when measurements … available, have been modelled

Line 450-452: and the prescribed 20% limonene monoterpene mixture described above. Note that

Line 454: modelled day, 507 pptv

Line 460: frequencies thus determined were

Lines 534 and 535: There are 2 instances of using the word “reactant” where I think “VOC” would be true and more clear.

Figure 1: I think heat in a reaction scheme is typically indicated with a capital delta and not DELTA T. In the caption, what do you mean by OH formation in reactions of HO2 with OH? That reaction does not result in net OH formation.

Figure 2: Suggest to report NOx, isoprene and sum(MT) in ppbv rather than pptv. Also, could you put more of the traces on the same axes so that each individual plot is a bit taller and has a few more ticks on the y axis? For example, could combine sum MTs and isoprene on same panel, and perhaps NO2 and CH3CHO.

Figure 4: suggest to switch colors to red is hotter and blue cooler (more intuitive)

Figure 6 panels E and F: suggest to make y axis 0 – 1500 so you can see the measurements (even though this will push the without loss case offscale sooner – you’re not arguing that it’s a realistic case anyway, so it’s fine if it runs offscale.