Chi et al. (2025) used the high temporal and vertical resolution of ground-based remote sensing observations to study the hourly evolution of HCHO and NO2 columns (and near-surface concentrations) over several cities in Japan. They further classified the chemical conditions (e.g., NOx-sensitive, VOC-sensitive, and transitional) based on thresholds derived from a chemical transport model, which was vetted against in situ observations. The novelty of this work lies in its application of Pandora observations to ozone-sensitivity diagnosis. However, the paper overlooks several critical components that are necessary for reasonably attributing ozone sensitivities. Moreover, this reviewer is not entirely convinced that the model used to determine the thresholds has reasonable accuracy, or that some of its significant discrepancies with observations could result in a wide range of ambiguous possible outcomes. Therefore, I do not recommend the publication of this paper in the present form.

Major comments:

Light, light, light, A central but often under-emphasized point in tropospheric ozone research is the dominant role of photochemistry. Virtually all ozone-related studies begin by noting that ozone formation is a multifaceted problem driven by interactions among its precursors (primarily NOX and non-methane VOCs) under sunlight. Yet over the past two decades, many studies have disproportionately emphasized the roles of NOX and VOC while devoting relatively little attention to the availability of light (photolysis rates) itself, which is the fundamental driver of the chemical processes governing both ozone production and loss.

Beyond the well-known limitations that HCHO and NO2 do not fully represent VOCR and reactive nitrogen, it is crucial to recognize that both the magnitude and sensitivity of ozone production rates depend strongly on geophysical factors that are independent of FNR. Among these factors, photolysis rates and water vapor are dominant controls on atmospheric oxidative capacity, driving numerous reactions relevant to ozone formation (e.g., Kleinman et al., 2001). The use of FNR reduces this inherently multidimensional and nonlinear chemical system to only two dimensions, thereby concealing key variability of the light and water vapor.

To illustrate this limitation, I perturbed photolysis rates over polluted regions during the KORUS-AQ campaign using the observationally constrained F0AM model (Souri et al., 2023; 2025a; 2025b). Photolysis frequencies were multiplied by factors of 0.5 (dim), 1.0 (default), and 2.0 (bright), producing three corresponding sets of PO3 isopleths. Systematic increases in photolysis enhanced both net PO3 and its sensitivities to NOx and VOCs, as evidenced by more compact isopleths in the bright-light scenario (each contour shown corresponds to 3 ppbv hr-1). This demonstrates that identical FNR values can correspond to substantially different sensitivity regimes depending solely on available sunlight. Moreover, in situ observations show no meaningful correlation between FNR and photolysis rates, as the latter depend on solar zenith angle, altitude, surface albedo, column ozone, and particles.

Link to the figure: https://drive.google.com/file/d/1055kbTHJ01fdjA2WcFbWokGeuExicxj4/view?usp=sharing

This is a fundamental point at which FNR falls apart.

The manuscript attempts to classify air parcels into sensitivity regimes even during periods of weak photochemical activity (e.g., winter, early morning, late evening). Such classifications are utterly nonsensical. Under low-photolysis conditions (an entirely separate dimension of ozone chemistry, as discussed in Chatfield et al. (2010) and Souri et al. (2025)), ozone production becomes largely insensitive to perturbations in NOX and VOCs.

This figure from Souri et al. 2025 clearly shows this using five variables derived from TROPOMI and our PO3 parameterization across two seasons in Los Angeles. During December, January, and February, FNR values are low due to abundant NO2 relative to HCHO, implying a VOC-limited regime. However, the derivatives of PO3 with respect to HCHO and NO2 are substantially muted due to weak photochemistry, making PO3 largely unresponsive to precursor concentrations. In contrast, summer conditions show markedly stronger sensitivities. The same logic extends to diurnal cycles, including differences between early-morning, late-evening, and afternoon Pandora observations.

The link to the figure: https://drive.google.com/file/d/1h2j8Wb0z9vVX-sb0FHVnEpe7-mAJaHns/view?usp=sharing

These findings underscore that discussions of PO3 magnitude or precursor sensitivity are of limited relevance during winter or under low-light conditions. This issue is not addressed in the current manuscript. Although the manuscript discusses diurnal variability in HCHO and NO2, which is interesting, its interpretation implicitly assumes that the sun remains constant throughout the day. Additional processes with intense diurnal cycles, such as HONO chemistry and water-vapor-dependent OH production, further complicate the picture. Even without including these processes, the incompleteness of FNR with respect to photolysis rates alone is sufficient to undermine its interpretive value.

If the authors wish to pursue publication in the current journal or another one, the following steps are highly recommended:

i) Restrict interpretations of FNR and ozone sensitivity to periods with active photochemistry, such as summertime and midday/afternoon hours.

ii) Discuss the dependence of photolysis rates on altitude, emphasizing that photolysis frequencies can be at least 20–50% higher between 2–4 km altitude compared to the surface, substantially altering the magnitude of PO3 sensitivities.

iii) Include a dedicated limitations section addressing uncertainties and assumptions inherent in FNR. Key sources of error have been quantified in Schroeder et al. (2017), Souri et al. (2020, 2023), and Jin et al. (2017). This section should explicitly note that FNR is insensitive to both water vapor and photolysis rates, making FNR-based chemical classification incomplete and, in specific contexts, obsolete. Please quantify the errors of GEOS-Chem, the errors in Pandora, and other assumptions to find a set of threshold values. Will the errors be so large that the clear definition of the regimes becomes compromised?

Alternative approaches, including joint inversion/data assimilation frameworks (https://acp.copernicus.org/articles/20/9837/2020/) or data-driven parameterizations of PO3, such as those presented in Souri et al. (2025) (https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1679/ ), offer more physically grounded paths forward.

The specific method is less important than the need for the community to recognize the intrinsic limitations of FNR and move toward more comprehensive representations.

Why exclude the primary source of HCHO? I found it puzzling that the authors are so focused on separating primary from secondary formaldehyde. The title of the manuscript (which I think is unclear) appears to suggest that the goal is to study secondary formation pathways. However, the manuscript does not provide a strong justification for why this focus is necessary. Primary HCHO also affects both net ozone production rates and their sensitivities to NOx. Because HCHO is reactive, even directly emitted HCHO can be an important source of HO2, accelerating the ROx–HOx cycle and enhancing ozone formation per molecule of NOx. In many regions, the primary source of HCHO, which is a regulatory concern, can substantially contribute to ozone production. Ignoring it weakens the motivation of the study.

Ozone titration by NO is not the issue. The manuscript repeatedly brings up the influence of NOx on ozone through titration in NOx-saturated environments, but this is not the central point of defining chemical sensitivity regimes. Classification of ozone chemistry is about determining whether radical termination occurs through the loss of ROx or through the loss of NOx, and how the relative amounts of NOx and VOCs shift that balance. When NOx is extremely high relative to VOCs, the radical chain that drives ozone formation is suppressed. As a result, the efficiency of producing ozone per molecule of NOx becomes very low, meaning low P(O3)/NOx. Ozone titration itself is not particularly relevant to chemical sensitivity or to regulatory interpretation.

Titration is rapid, local, and temporary, and it is easily reversed during the day through photolysis. What matters are the processes that are not easily reversible, such as the formation of H2O2, HNO3, or organic nitrates. In NOx-saturated regimes, adding more NOx removes OH and reduces the radical pool through HNO3 formation. This is the central mechanism that determines chemical sensitivity, not the short-term titration of ozone by NO.

What do we learn from validating the model? The authors validated GEOS-Chem against in situ and ground-based remote sensing data, but found substantial biases (especially for ozone). But I found it to be highly disjointed from the rest of the paper. What do they imply for the rest of the manuscript? Are these errors so significant that we cannot confidently determine the chemical conditions? Are they indicative of a missing source or mechanisms? What is the connection of the analysis to the rest? If the goal is to build confidence in the model, it needs to be compared with a large set of modeling efforts to understand how far it falls short.

Storytelling. Another issue is that the authors pick and choose what to show in the primary draft versus the supplementary material. I think all stations should be included in the primary draft. There should be more emphasis on the evolution of NO2 and HCHO, and less on the implications of ozone sensitivities arising from the factors above. The study also ends a bit prematurely. What do we learn from Pandora that we cannot do using satellites? How important is it to look at the vertical components for regulators? Are we expecting to see changes in emissions or composition at higher altitudes (changes in aviation)?

Specific Comments:

Line 27: In the presence of sunlight. That’s a critical knob missing from the analysis.

Line 40. FNR is only a subset of significant (and more robust) indicators such as H2O2/HNO3 or Ox/NOz. You may want to say that FNR gained popularity because it can be accessed from remote sensing satellites. However, if H2O2/HNO3 were not limited to limb sounding, it would be preferred over FNR, as it directly explains the chemical loss of peroxy radicals over the chemical loss of NOX.

Line 40. NMHC/NOx is a very crude ratio. It must be reactivity-weighted VOCs over NOx.

Line 42. FNR is not the most precise indicator. Please introduce the HOx-ROx cycle notion and the robustness of H2O2/HNO3.

Line 45. A proxy for VOCs reactivity, not VOCs.

Line 48. I don’t understand how the primary source of HCHO can be misleading. HCHO is reactive so it has an influence on PO3 by itself.

Line 50. More recently? HCHO/NO2 was introduced 25 years ago.

The whole paragraph from L35 onward should be significantly improved. It is unclear what message is conveyed. Please begin by discussing the controlling factors of the HOX-ROX cycle and where FNR fits into this picture. Unfortunately, Sillman’s and Kleinman’s pioneering works have not been recognized here:

Sillman, S. and He, D.: Some theoretical results concerning O3-NOx-VOC chemistry and NOx-VOC indicators, J. Geophys. Res., 107, 4659, https://doi.org/10.1029/2001JD001123, 2002. 

Sillman, S., Logan, J. A., and Wofsy, S. C.: The sensitivity of ozone to nitrogen oxides and hydrocarbons in regional ozone episodes, J. Geophys. Res., 95, 1837–1851, https://doi.org/10.1029/JD095iD02p01837, 1990. 

Kleinman, L. I., Daum, P. H., Lee, Y.-N., Nunnermacker, L. J., Springston, S. R., Weinstein-Lloyd, J., and Rudolph, J.: Sensitivity of ozone production rate to ozone precursors, Geophys. Res. Lett., 28, 2903–2906, https://doi.org/10.1029/2000GL012597, 2001. 

Line 54. Please remove “some”.

Line 59. Some issues have prevented people from using Pandora. Aside from the lack of high spatial coverage, there are lingering issues with Pandora HCHO retrievals that may need to be disclosed here. In addition, in very high SZA, the airmass Pandora observes may not be fully representative of the overhead columns (the homogeneity assumption used in the application of geometric AMF gets violated).

L79. What do you mean by oxidant levels? Ox? Ox is immune to NO titration because O3+NO2 won’t change under titration. When we label a regime as NOx-saturated, it means NO2+OH is taking place. Titration is not a concern because once the sun is out or the air mass ages outside of super emitters, the titration effects go away, and NO2 will contribute to O3. In other words, Ox will remain the same in both conditions, but the partitioning between NO2- and NO-O3 changes. This isn’t important for regulations. What matters for regulations is the formation of H2O2 and HNO3, which can be mostly lost in the system (not always).

Line 135. “a sensitivity run” is ambiguous.

Figure S1 is confusing because the signs are the opposite. It suggests that reducing VOCs should lead to higher ozone levels.

Section 3.2.1. Supplementary figures should be used only for supplementary information. This section is deliberately designed to discuss model comparisons vs. surface observations, necessitating the inclusion of figures in the primary draft.

Section 3.2.1. It’s interesting to see HCHO being higher in Oct than in July. What is contributing to it, given that the model couldn’t see this tendency? Could this be caused by different air masses or primary anthropogenic HCHO levels? It’s also interesting that the observed columns show an opposite trend, given that most of HCHO is confined in the first few kms. Are we confident in the accuracy of the HCHO measurements?

Line 253. I disagree with this statement. Surface measurement is designed to monitor the surface, and there is nothing inherently wrong about it. Columnar data represents something else. I don’t think the discrepancies indicate that each domain is inferior. They represent different regions.

Line 260. You have cited this paper elsewhere, and it may be highly relevant. We examined the individual tendencies shaping HCHO concentrations (Fig. 7): https://www.sciencedirect.com/science/article/pii/S1352231023003552

Figure 3. Interestingly, PBLH did not change much throughout the seasons. Is it expected over Japan?

Line 295. Could anthropogenic VOC oxidation also contribute to the enhancement?

Line 300. This isn’t a strong argument. If by “exceedances” you meant high ozone episodes, those could have been caused by factors such as regional background, vertical contributions, and other precursors. I agree that high ozone exceedance usually is associated with high HCHO levels because large HCHO concentrations are indicative of favorable meteorological conditions to form ozone, but it is not the only factor.

L302. This is precisely why I think it is not interesting to study surface ozone sensitivity. Every parcel within the PBL is important for ozone production rates. This is why Souri et al. (2023; 2025a, 2025b) focused on the PBL region rather than the surface.

Section 3.2.3. Unfortunately, only one station is shown here. Also, it’s essential to include the trajectory characteristics somewhere in the primary draft. This could be a major comment, but I mention it here: the location of trajectories is one thing, and when it occurs more frequently is another. This part needs to be elaborated to understand how often they happen before delving into where they are coming from.

Line 343. This tendency was observed in Jin et al. 2017 and Figure 6 in https://acp.copernicus.org/articles/23/1963/2023/

Figure 7. What caused a significant drop in June in Sapporo?

On MAX-DOAS plots: it may be worth adding AKs to show where the sensitivity of the retrieval to the absorber is the largest.

Please see the attached file.