the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 07 May 2025
Why observed and modelled ozone production rates and sensitives differ, a case study at rural site in China
Abstract. Ground-level ozone (O3) pollution has recently become of increasing concern in China, traditional models often fail to accurately predict the net O3 production rate (P(O3)net) and O3 formation sensitivity (OFS) due to missing reactive volatile organic compounds (VOCs) and their complex reactions. Therefore, we conducted a field observation of P(O3)net and OFS using a P(O3)net (NPOPR) detection system based on a dual-channel reaction chamber technique at the Guangdong Atmospheric Supersite of China in Heshan, Pearl River Delta in autumn of 2023. The in-situ monitoring data were compared with results from a zero-dimensional model incorporating the Master Chemical Mechanism (MCM v3.3.1). We tested the model performance by incorporating parameterization for 4 processes including HO2 uptake by ambient aerosols, dry deposition, N2O5 uptake, and ClNO2 photolysis, and found that the discrepancies between the modelled and measured P(O3)net did not change evidently, the maximum daily P(O3)net differed by ~44.8 %. Meanwhile, we found the agreement of OFS assessment results between the direct measurements and the modelling was lower in the P(O3)net rising phase (08:00–09:00, 63.6 %) than in the P(O3)net stable/declining phase (10:00–17:00, 72.7 %). The only approach to fill the gap between observation and computation was to add possible unmeasured reactive VOCs, especially oxygenated VOCs (OVOCs) in box model, this was true for both P(O3)net and consequent OFS, highlighting the importance of quantitative understanding the total reactivity of VOCs in O3 chemistry.
- Preprint
(1112 KB) - Metadata XML
-
Supplement
(2779 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-1618', Anonymous Referee #1, 06 Jun 2025
-
RC2: 'Comment on egusphere-2025-1618', Anonymous Referee #2, 22 Jul 2025
Zhong and colleagues present measurements and detailed analysis using constrained box model approaches of in-situ ozone formation at a field site in the PRD, using a newly-developed direct measurement of ozone production rates, alongside measurements of various atmospheric chemical species / photochemical parameters.
The paper presents an extensive exploration of the measurements, assessing the NOx- and VOC-dependence of the measured and modelled ozone formation, and relating this to (e.g.) missing VOC species. The approach is logical and largely well described (although I have some significant suggestions for clarifications – below), and the work represents a good advance in approaches to analysis of these new measurement approaches / data and could make a valuable contribution; in particular the assessment of ozone production “gaps” vs model with co-reactant concentrations / conditions (NOx/VOC sensitivity)
My principal concern is that the degree of accuracy (and maybe precision) of the measurements may be over estimated, and that the analysis of these – still relatively new – measurements is taken further than the data uncertainties really justify; that the data are over-interpreted.
I think this is certainly the case for the model, considering the VOC coverage (correctly) identified and uncertainties in e.g. knowledge of HONO (not measured directly), and I would ask the authors to consider carefully if the measured P(O3) is really good to 10% accuracy – and hence if quite such an extensive set of analysis is warranted. I’m conscious that lots of things will more-or-less co-vary diurnally within the measurement uncertainty (concentrations, j, T, O3…) – is it really possible to extract missing reactants at a few % accuracy from within the combined measurement and model uncertainties ?
Showing more raw data – maybe 2-3 individual days plotted in detail so the measured and modelled data can be made out – would help the reader understand the sensitivities of the various metrics, alongside the “integrated” plots of sensitivities. This might help focus the manuscript also, as the story would be clearer with fewer analyses (and fewer SI figures) which would then also give confidence that the degree of analysis is appropriate and the data not over-interpreted.
I recognise the method is published in the Hao et al (2023) paper, but would encourage the authors to include more discussion of measurement uncertainties here, and at the least a detailed justification of the statement “…measurement uncertainties around 10%” (L79). this should include, separately, accuracy, precision, and selectivity/bias – the impact of wall artefacts on the measured P(O3), which may vary with conditions (j, RH, VOC/NOx levels), and would best appear around L120.
Corrections / Comments
L47 and following – it would be useful to distinguish between NO titration of O3 – ie NOx/O3 PSS shifts – and net production of Ox (which is what we really mean by ozone production). Several points in the text later (eg L134) there is reference to titration reducing ozone production – I’d argue that this is PSS shift, not a change in the ozone production chemistry, and a different terminology might help.
L77 please acknowledge / include the pioneering work of Brune and colleagues (Cazorla et al., 2012) as the first “modern” MOPS system developers (I realise this is referenced later).
L79 Measurement uncertainties – I do not think 10% is realistic – see general comments above
L100 is there much emission / chemical heterogeneity around the site ? e.g. on the timescale of NOx PSS (1 min+) or HONO PSS (10-15min+) ?
L131 Explain how the VOC addition amounts were determined / apportioned between the two species
L202 E10 – I do not follow how the net P(O3) is equal to P(O3) multipled by the change in P(O3) divided by the (natural log of) change in X.
L234 not sure “stronger” photochemical reactions is right word – do you mean higher photolysis rates ? An alternative explanation for the (slightly) lower concs might be greater solar heating / higher BLH / more dilution on the hotter/sunnier/higher P(O3) days – evidence in the BLH data ?
L305 give the missing P(O3) as a % also – maybe 24 hour mean. Statistical test – I assume P level of 0.05 not 0.5 ?
L325+: Is there really sensitivity in the correlations to identify particular causes ?
L345 is it valuable to include all of D1-D4 – cut straight to the final case, D4 ?
Fig S13 – It is very hard to see the change in PO3 from the added NO / added VOCs – suggest show a zoom in on a polluted / non-polluted day in addition so the data can be seen. The uncertainty ranges look very small on this figure ?
L440ish: Ozone regime – there is only one data point in the afternoon showing a VOC limited regime (14:00). Is there really a shift from VOC to NOx to VOC to NOx limited through the day – can the data really show this ? I am conscious that there are not many days going into these averages. How is “transition regime” defined for the measurements ? The explanation (L442): P(O3) measurement is not affected by NOx/O3 titration (or PSS) – rather it measures change in the net Ox production.
Fig 6 – please show the mean diurnals for PO3 (from the measurements) for the three regimes identified. Not sure that the rapid changes in emissions can be the explanation – the model is constrained to the observed concentrations, so it has this “built in”. A greater challenge in this regard may be the P(O3) measurements which average over an hour effectively ? Isnt it more that you have already shown that the model (not unexpectedly) has bias from missing VOCs and this is reflected in these analyses also ?
L489/Fig S15 – what are cases E1-E3 ? Not mentioned previously and I cannot find a definition / description of these. I cannot follow L490-L505 as these cases/ scenarios are not defined
Conclusions – is there a comment to make on the impact of different chemical mechanisms (eg L65+) on the model/measurement agreement ?
Conclusions – do you wish to add an overall comment vs the NOx- vs VOC- control on O3 observed at the site, and implications for policy vs reducing ozone formation rates ? this might usefully also remind the reader of the distinction between the in situ formation rate (focus here) and the total level experienced (integration of local formation and upwind chemistry / advection).
Model Approach Clarifications
-A summary table of the different scenarios, A-E, would be very helpful – I think this is referred to but I cannot find in the SI ?
-Model observation constraint: It would be helpful to explain how the constraint to observations was implemented – we have model outputs on an hourly basis but how frequently where the constraints applied to the model and did concentrations evolve freely in between (shorter model integration timestep) ? If the model species are not in balance a “saw tooth” effect can result in the simulated concs at the higher model time resolution between observation constraint points – was this the case and if so impacts on the P(O3) which is in effect averaged over this period ?
-Approach to HONO – tucked away in the SI, the use of MARGA measurements of soluble nitrite for gas-phase HONO is mentioned. I appreciate that some approach was needed, but the sensitivity of the model results to this assumption is needed – eg what shift in P(O3) does a 20% change in HONO concs (or whatever is reasonable) result in. If this is a large shift – are the subsequent analysis all valid, i.e. can you be confident in pushing the model explorations so far ? What is the time resolution of the MARGA data vs the observed temporal variation of e.g. NOx ?
-Lots on the deposition velocities but how well do we know the boundary layer height BLH ?
L172 – is the HO2 uptake process an irreversible loss (in the model) ?
L174 – is the N2O5 uptake an irreversible loss – if I follow some scenarios included recycling via ClNO2 – not quite clear how the condensed phase processed were simulated ?
Minor Points
Please define OFS where first used (abstract, L61)
Various places – ppb etc are mixing ratios not concentrations
L165 het loss can be important for HO2 removal globally – but wont HO2 + NO dominate under the conditions of these BL measurements ?
Citation: https://doi.org/10.5194/egusphere-2025-1618-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 787 | 59 | 22 | 868 | 51 | 23 | 37 |
- HTML: 787
- PDF: 59
- XML: 22
- Total: 868
- Supplement: 51
- BibTeX: 23
- EndNote: 37
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript presents a comprehensive investigation of net ozone production rate (P(O₃)ₙₑₜ) and ozone formation sensitivity (OFS) through the integration of in situ field observations using a novel dual-channel reaction chamber system (NPOPR) and detailed box model simulations based on MCM v3.3.1. The study is of high relevance and scientific value, particularly in addressing long-standing issues of underestimation in modeled ozone production. The work also has practical implications for improving model-based OFS diagnosis and VOC pollution control strategies. However, several major issues must be addressed in the manuscript.
Lines 29-31: You mentioned “the only approach to fill the gag was to add unmeasured VOCs” appears too strong. It implies that no other explanations or methods could be relevant, which may not be justified. Consider softening this to reflect that unmeasured OVOCs were the most effective compensating factor in this study, rather than the only one possible.
Line 43: “precursor” should be “precursors”.
Line 46: “equation” should be “equations”.
Lines 205-206: It is unclear what magnitude of precursor perturbation was applied, please provide a more explicit description of the model configuration used for the sensitivity analysis.
Lines 278-281: The confidence interval of 68.3% is relatively conservative, please provide additional analysis.
Line 306: The reported p-value (P < 0.5) does not indicate statistical significance, and the analysis doesn’t hold.
Lines 305-307：The statement that the mechanisms added in Case D1 “are not the main cause” of the bias may overstate the conclusion. The remaining discrepancy could still be partly due to uncertainties in those mechanisms, parameterization, site-specific variability and transportation etc. A more cautious wording would improve clarity and avoid giving a false sense of certainty.
Lines 328-329: The statement that “P(O₃)net_Missing increases significantly at higher O₃ precursor concentrations” (based on r² = 0.4-0.5) may overstate the strength of the relationship. A moderate correlation should not be equated with a strong or significant increase unless supported by statistical testing.
Lines 389-391: You use an empirical relationship between kOH and P(O₃)net to get kOH_Missing, and then adjust VOC concentrations to match this value. However, this method assumes a direct linear relationship without showing how real chemical reactions support this assumption. Please explain why this approach is reasonable, and whether it reflects actual atmospheric chemistry.
Lines 396-397: In Case E2, ethylene was amplified to 5.9-85.6 times the original concentration, far exceeding the limit emission levels in the conventional urban atmosphere. The lack of emission inventories or observational data support may cause the simulation results to deviate from reality. More discussion on the rationality of these magnifications is required.
Line 485: It is not necessary to add legends to every subgraph in Fig.6. Simplification can be considered.
Lines 519-521: The conclusion repeatedly emphasizes the role of OVOCs in O₃ formation and compensation, yet it assumes these are mainly anthropogenic in origin. As noted earlier, many OVOCs are also formed secondarily. It is recommended that you should distinguish between primary and secondary OVOC contributions or clearly state the limitation of their current attribution.