This paper describes an improved direct measurement system of the photochemical ozone production rate (P(O₃)). The authors also performed a field observation of P(O₃) and evaluated the instrument and observation results using a detailed box model simulation.

O₃ pollution is a crucial problem for atmospheric environment, and a behavior of O₃ is very difficult, so direct measurements of P(O₃) such as this research are very important and valuable. However, this paper has major concerns about the evaluation by the box model simulation. So I cannot recommend this paper to be published in Atmospheric Chemistry and Physics in the present form. The authors should resolve the concerns by re-evaluation of the box model simulation or other ways.

Major comments:

• The results of the box model simulation

I suspect such high concentrations of NO₃ in the reference chamber. First, the value of JNO₃ in the reference chamber is about 90% of that in the reaction chamber, which is sufficiently high. Second, the rate constant of the reaction of NO₃ with NO is sufficiently large and the lifetime of NO₃ (at 298 K) is 1.6 s and very short in the presence of NO of 1 ppbv. There might be high concentrations of NO₃ if there are very large sources of NO₃, but in that case, the authors should mention the evidence. I think the NO₂ + O₃ reaction cannot be a large source of NO₃. I’m not sure about N₂O₅, but it is unlikely that there could be high concentrations of N₂O₅ under the temperature during the observation period.

On the other hand, I also suspect the reaction of NO₃ with VOCs as a major source of RO₂ in the reference chamber. The rate constants of the reactions of NO₃ with VOCs are not so large, and it is questionable that NO₃ and RO₂ concentrations change in minutes by the reactions of NO₃ with VOCs unless there are extremely high concentrations of VOCs.

For Figs. 9(b) and S10(b), photodissociation of OVOCs is a significant part of ROx sources in the reference chamber. Why such a large fraction in the absence of UV? The authors should mention the evidence.

I entertain doubts about the results of the present box model simulation. I think the authors should revalidate the appropriateness of the model thoroughly.

• Measurements of NOx using a chemiluminescence NOx monitor.

Did the authors use a commercially available chemiluminescence NOx monitor without further modification? If so, accuracy of NO₂ and NOx concentrations would be low because NOz such as HNO₃ and PANs interfere observed values of NO₂ and NOx. I think this interference could affect discussion in this paper, so the authors should evaluate the interference quantitatively. In the fresh air masses, the interference by descendent pollutants of NOx such as HNO₃ and PANs might be small, but that by HONO could be large instead.

Other minor comments:

Some mathematical formulae: The authors should use italic and roman letters correctly.

L65: at 424 nm → at < 424 nm (at less than 424 nm)

L72: are proportional to → affect ?

Fig. 1: Why do the authors use critical orifices instead of mass flow controllers? Is the temperature of orifices controlled to keep a constant flow rate?

Fig. 1: Are inner walls of the reaction and reference chambers coated with Teflon? If so, please indicate the kind of the Teflon coat.

Table S1: The authors should add standard deviation to average residence times.

Fig. S5: The regression lines have non-zero intercepts. These are significant? If so, why? The regression lines for ozone have negative intercepts. In this case, there are large losses of ozone in the high concentration of ozone? For NO₂, how about relative humidity in the experiment? Is there no loss of NO₂ at high relative humidity?

L218: low → high?

L226: Why the transmittivity of HONO in the reference chamber is lower than that of O₃? How about accuracy and precision of the actinic flux spectrometer?

L228: agree → agreement

Table 1: What is Ultem? There are no definitions in the text.

Table 1: The values of 0.019±0.011 should be shaded.

L272-278 (The airtightness of the reaction and reference chambers): It is hard to follow this section. I think the authors should explain using a schematic diagram for the experiment in the supplement.

L297-L300: For calibration of NO₂, it is not appropriate to perform calibration of NO₂ using a NO₂ standard gas because of low reliability. Calibration should be performed using a gas-phase titration method using NO and O₃.

Reference: The authors should put the list into alphabetical order.