This manuscript presents the development and application of an integrated online system for simultaneous measurements of gaseous and particulate reactive oxygen species (ROS). As ROS is a key driver of atmospheric oxidative potential and associated health impacts, the ability to measure gas-phase and particle-phase ROS synchronously is valuable for improving our understanding of atmospheric chemistry and exposure pathways. Overall, this work is within the scope of AMT, the developed instrument is well-characterized, and the four-season field observation validates the robust application of the instrument. However, several issues need to be addressed before publication.

1. The paper would be strengthened by a direct comparison table of the present instrument vs. earlier online ROS analyzers (e.g., the instrument by Huang et al., AE, 2016; King and Weber, AMT, 2013). Compare the key parameters: time resolution, LOD for gas and particle phase, reagent consumption, portability, and major artifacts. Also, clarify how the present system differs from the earlier instruments.
2. The authors did interference assessments with individual species. However, in the real atmosphere, these soluble species coexist, and do the authors also test with the mixture of all these soluble species (i.e., O3, NOx, SO2, and transition metals)?
3. All ROS concentrations are expressed as “H2O2 equivalents” using H2O2 standards. Given that ambient ROS comprises a mixture of peroxides, radicals, and possibly other oxidants, the authors should clarify the sensitivity of the DCFH probe to different types of ROS and discuss the possible limitations of using the DCFH assay.
4. The ROSp collection efficiency (91.2%) was derived from recovery experiments with three fractions (chamber liquid, wall rinse, backup filter). Was this efficiency constant over the range of ambient PM loads and compositions? High PM loading might alter wall loss or droplet growth efficiency. Also, please clarify the reproducibility of the efficiency measurement (e.g., triplicate tests) and whether it was rechecked after extended field use.
5. Line 134: “~0.13 s” residence time in the glass spiral absorber. How was this estimated? Provide the calculation or a reference.
6. Please provide a brief introduction about the field site used for instrument validation.
7. Line 344, ROSg and ROSp have the same seasonal pattern, why “whereas” is in the sentence?

This manuscript presents an integrated online analyzer for the synchronous measurement of gaseous and particulate reactive oxygen species, based on the DCFH-HRP fluorescence method, together with a four-season field campaign in Beijing. The synchronous dual-phase capability and the seasonal dataset are useful contributions. However, several issues regarding methodological rigor, the reliability of key performance metrics, and the interpretation of the field data should be addressed.

Major Comments

1. Definition and chemical selectivity of the measurement. The DCFH-HRP system responds very differently to different ROS species, and after wet-chemical absorption the instrument effectively quantifies only the water-soluble, DCFH-reactive subset of ROS. I suggest the authors state explicitly what operationally-defined quantity is measured, and revise the wording in the Abstract and Conclusions (e.g., "DCFH-reactive water-soluble ROS, as H₂O₂ equivalents") so the values are not read as total ROS.
2. Collection efficiencies (γ_p). Eq. 2.5 includes M_wall (wall-deposited material) in the numerator, but this fraction is recovered only by offline rinsing and does not enter the detection path during routine operation. The effective online efficiency would be closer to 81.72%, implying the reported ROSp values may be underestimated by ~10%. Please either (1) adopt M_col / (M_col + M_wall + M_filter) as the collection efficiency and revise the ROSp values throughout the manuscript accordingly, (2) demonstrate that wall deposits could be eventually detected, or (3) explicitly discuss the direction and magnitude of this bias.
3. Interference assessment relies largely on the literature. Section 3.3 discusses O₃, NO, SO₂, and Fe interferences almost entirely from prior studies, without tests on this instrument. Since the present configuration (pH 7.0, 10 µM DCFH, 2 units mL⁻¹ HRP) differs from those works, I recommend adding at least some direct interference tests. Also, in lines 315–318, you mention that "In fluorescence-based H₂O₂ detection, SO₂ can interfere with fluorophore formation through acid-catalyzed reactions. This interference can be fully suppressed by adding formaldehyde." Please clarify whether formaldehyde is actually used for SO₂ suppression in this study; if so, it does not appear in the flow schematic.
4. Correlation versus causation in the field interpretation. Several mechanistic interpretations in Section 4 rest on relatively weak correlations. Please provide p-values (or confidence intervals) and sample sizes, and consider tempering the mechanistic claims for relationships with |r| < 0.3.
5. Mismatch between calibration range and ambient concentrations. The calibration covers atmospheric equivalents of 2.88–14.41 ppbv (ROSg) and 0.29–1.49 µg m⁻³ (ROSp), yet essentially all seasonal-mean field values fall below or close to the lowest calibration point. Low-concentration linearity is unverified; reagent auto-oxidation and blank-subtraction errors contribute proportionally more at low signal, and the near-zero-intercept calibration becomes sensitive to small intercept errors. I recognize the campaign has concluded; nonetheless, it would be valuable to perform a low-range verification at the current instrument state and use any campaign QA/QC records to constrain drift.
6. Contradiction with previous Beijing observations. Huang et al. (2016), using a closely related instrument at a rural Beijing site, found ROS higher in winter than spring; the present study (urban Beijing) finds spring highest and autumn lowest. Please discuss the possible reasons for the opposite seasonal ranking.

Minor Comments

1. Eq. 2.3. The factor (T₀/T)(P₀/P) appears inverted; it should read (T₀/T)(P/P₀). Please correct and confirm whether the reported ROSg values were affected.
2. Table 1. The bolded optimal row for temperature effects is 37 °C, whereas the text (line 234) states 40 °C was selected. Please reconcile.
3. A short paragraph acknowledging the methodological limitations would be a useful addition in the conclusion section.