Manuscript egusphere-2025-1423 by S.D. Harni et al. reports measurements and modelling data during a 3-month measurement campaign in different horizontal distances from a noise barrier. The effect of the noise barrier on particulate pollutants and nitrogen dioxide concentrations is quantified based on measurements and modelled concentrations of a Gaussian air quality model Enfuser that incorporates an obstacle detection and concentration reduction routine for simulating the effect of the noise barrier. An advantage of that model is that it can be informed with complementary datasets of observations, traffic flows and geodata. A caveat of the experimental design is the missing PNC measurement at 10 m and that BC was only measured at 20 m distance behind the noise barrier. The work appears well thought out and executed and forms a coherent study which fits well in the field of air quality research. The manuscript is clearly written but overall lacks transparency in model description and accuracy in data interpretation. I recommend publication if a number of smaller issues addressed below are resolved.

Specific Comments:

1.) Abstract: The reduction in pollutant concentrations through the noise barrier are described as being “significant”. The word “significantly” is used excessively in the Introduction. The excessive use of “significant” and “significantly” in this manuscript without presenting a check of statistical significance should be avoided. Examples are found on P15, L323; P18, L380, L385, L408, but there are more.

2.) Introduction (P2, L66-67): CFD models have been used to simulate street canyons and other built environments, the argument of not adopting LES should be more specific, i.e. with respect to fitness for the purpose of this study.

3.) Modelling framework (P8): I understand the brevity of the modelling framework section, by referring readers to the published model description of the Enfuser model. Nevertheless, it should be possible to understand the model features which are utilized in this study, without having to consult this publication. I suggest adding two pieces of information: a) how many receptors are used for the Helsinki metropolitan area with base resolution and the campaign area of 4x4 m2 (it should be meter square in the text); and b) describe the calculation of pollutant concentrations in the vertical – are they inferred from the SILAM CTM model? – which is relevant when you compare to the vertical profile of the drone measurements.

4.) Modelling of noise barrier (P10): the description of the reduction effect of the noise barrier on concentrations at receptors is adequate. Some more details on the precomputed obstacle detection should be given, such as the workflow of this routine and the datasets (topography, building heights, etc.) used. Equation (19) contains a duplicate comma. How is the concentration at the obstacle itself calculated or are the obstacles masked in the 2D concentration map?

5.) LDSA, BC and PNC modelling (P12): Please provide details of the proxy of PNC emission factor based on PM2.5 emission. The PNC emission factors are probably different for various emission sectors. A table with that information would be useful. Is there any size segregation of PNC in the model (the information should be placed here)? LDSA represents surface area concentration of particles deposited in the alveolar region of human lungs and depends on the size distribution of particles. The most common size range for particle surface area is in the range of 100–500 nm. Obviously, LDSA is simulated as passive tracer like the other particulate pollutants, with emission as a fraction of PM2.5. Please explain which algorithm or post-processing is used to calculate the LDSA concentration field. Assimilation of LDSA and PNC background – again, is this based on available concentration measurements of the respective component?

6.) P3, L298-300: “The gradients of large particle might also be due to larger particles having slower dispersion” – seems to contradict the findings of the cited study by Zheng et al. (2022), who noted that larger particles (coarse mode, > 1 µm aerodynamic diameter) are more affected by traffic-induced turbulence than smaller particles, which would indicate more efficient dispersion. A reference for the apparently slower dispersion of larger particles should be given here.

7.) Figure 6 and belonging text: PNC should also follow a logarithmic curve, as many studies have demonstrated. Is this not observed because of the missing measurement point at 10 m? At least, the linear dashed line in figure part D appears unrealistic. The NOx measurement data should be displayed as well, since NOx is not affected by the chemical conversion and rather behaves like a passive tracer.

8.) Drone measurements (P16): is the PNC at 15 m height in open area significantly lower than at 2 m height? The labels in the plots of Figure 7 are unclear, they should indicate that 2 m and 15 m are in the vertical.

9.) P18, L380-390: Add the percentage contribution of direct traffic emission to the modelled concentrations.

10.) P 20, L420-425: The model also underestimates high PNC peaks, which cannot be explained by the use of studded tyres.

11.) P23, L475-480: Compare the range of measured and modelled LDSA concentrations of this campaign to other LDSA measurements in Helsinki and other Finnish cities.

12.) Figure 11: it is very confusing for the reader that NB20m and O10m plots of BC are placed next to each other, because the comparison does not reveal the effect of the noise barrier – as it is the case for PM10, LDSA and PNC in the same figure. My suggestion is to either place a note of caution about this in the caption or even better, to move the BC plots to an Appendix figure.

13.) Conclusions: regarding the simulation of the noise barrier (P29, L589-601), it should be discussed that the reduction effect of the barrier might be different for particles than for gases, as particles may deposit on the vertical surfaces of the barrier.

Technical corrections:

• P19, L412: in the unit of LDSA, delete space in cm and add a space between µm and cm.
• P23, L470: replace “PM” by “PM10”.
• P17, L359-362: change to present tense. Same for P18, L392-393.

The manuscript describes a thorough study on the effect of a 6.5 m high noise barrier at a major highway on air pollutant levels behind the barrier. In my initial assessment, I raised multiple concerns that have been adequately addressed by the authors. The manuscript has improved significantly and I only have a few minor comments on the current version:

Line 46: brackets are missing for the citation of (WHO, 2021)

Line 55: do you mean “…at 5 m behind…”?

Lines 101-103: For how long did the passive samplers collect before being evaluated?

Line 106: The AE51 is from microAeth, not TSI

Line 110: According to the manufacturer’s website, the A30 is EN-16976 compliant, i.e. it should have a lower cut size at 10 nm, not 7 nm. Have the temperatures been adjusted to shift the cut size to 7 nm?

Line 117: Was the factor 0.66 consistent for all partector devices and more or less constant over time? Do you have an explanation for the rather large mismatch? Could it be due to different calibrations, e.g. the use of different lung deposition efficiency curves?

Line 135: C shows the precipitation and D shows the water layer height.

Line 203: “e” is not a variable and should thus not be in italic.

Line 356: I am not a native speaker, but “not considered outliers” sounds a bit awkward in this sentence. I’d suggest to rephrase to “…lowest and highest values considered for data evaluation…”

Line 422-423: Couldn’t it also be that the model underestimates the cleaning effect of precipitation? Especially April, but also May were significantly drier than March.

Line 470: PM should read PM10

Line 521: Do you mean “combustion source” with “burning source”

Line 575: Do you really mean “significant” in the sense of “statistically significant”? If so, you will need to provide a proof for the significance. Otherwise, I’d suggest to rephrase to “…and lowest for NO2”.

Line 604: I think the “sensors” were actually “samplers” (appears twice in this line).