Two issues should be addressed and clarified before publishing the paper. Please reflect and comment on both issues.

1) It is not clear how "double counting" of the street traffic emissions is avoided. The emissions Ei are used in CHIMERE (eq. 1) as well as in MUNICH (eq. 2). This might lead to double counting and additional (too high) urban background contributions in CHIMERE caused by the street that is considered in MUNICH. On the other hand, if the street traffic emissions are only included in MUNICH, then the effect of neighboring streets is not included as contribution to the background concentration for the street considered. It is not clear how this is accounted for.

2) Secondly it is not clear if or how the O3-NO-NO2 chemistry is modelled differently in MUNICH compared to CHIMERE. This is important because the time scales might be different.

Modelling air pollutant concentrations inside cities and especially the enhancements within street canyons is a difficult problem, especially when not only considering primary but also secondary pollutants. Regional chemistry-transport models are by far too coarse to resolve the large concentration gradients in cities, and dispersion models, while being able to achieve a high resolution, are unable to account for the complex chemistry involved in the formation of secondary air pollutants.

The paper compares and evaluates two different approaches for downscaling from a regional chemistry transport model (CHIMERE) to the street level, which have been developed in the past years to overcome these issues. The first is a simple downscaling method redistributing the emissions and corresponding concentrations within a CHIMERE grid cell according to sub-grid scale information on emission sectors / land-use fractions ("subgrid method"). The second method is a more complex one-way coupling with a street-network dispersion and chemistry model MUNICH, that requires explicit information on street segments and intersections and on buildings for each grid cell of CHIMERE.

The comparison of two well-established methods presented in this paper is a valuable contribution to the literature, as it reveals specific strengths and weaknesses of the respective methods. The comparison is based on carefully designed model simulations ensuring a consistent setup with identical inputs making the comparison meaningful.

My only concern is that the study is presenting a fairly lengthy analysis of a rather trivial fact, namely, that the differences in concentrations at individual stations estimated by the subgrid and MUNICH methods strongly depend on the differences in local emissions at the respective road segment. This is a trivial observation because the "subgrid method" treats all streets within a grid cell in the same way, whereas the MUNICH approach accounts for the different emissions in each street segment separately. Another factor potentially affecting the differences is the street width:building height aspect ratio, but Figure 7 suggests that this factor is much less important. A more detailed (and again very lengthy) analysis then suggests that there must be an influence but that it is hardly discernable against the emissions influence.

I recommend to (substantially) shorten the respective sections and possibly put some results in an appendix. This would make the paper more easily accessible and attractive.

I would also welcome additional information on the specific implementation of the two methods:

The "subgrid method" seems to require multiplying the number of species in the chemistry scheme by the number of sectors considered, which seems very expensive. Please clarify whether really all species (including secondary) are treated n times (where n is number of sectors) and whether this is done in the whole 3D grid of CHIMERE or only in the lowest few layers. If this is done only in the lowest layers, how is the transition to the upper layers handled, where each species is represented only once?

Similarly, for the "MUNICH method" it seems that the number of species (per CHIMERE grid cell) has to be multiplied by the number of street segments in that cell. Within a 1 km x 1 km grid cell this number must be quite large, making the scheme potentially very expensive. Again, it is unclear to me whether all species of the chemistry and aerosol schemes are treated in this way or only a subset. And is this done for all CHIMERE cells in Paris or only for the handful of cells containing the measurement stations? Please clarify!

Finally, it should be briefly explained how wind along the different street segments is computed. I assume that the wind in segments perpendicular to the mean flow (from WRF) is lower than in segments oriented along the mean flow. I assume that the orientation of a street canyon relative to the mean flow is another factor potentially contributing to differences between the two methods, because the "subgrid method" cannot take this into account. Please comment on this point and eventually add a corresponding analysis.

Apart from this, I consider the publication acceptable subject to the following points and corrections.

Minor points

P1, L13: "difference" would be better than "bias" in this case, because we don't know which one is the truth

P1, L19: Use past tense, i.e. "assessed" rather than "assess"

P2, L36: "transfer" would be better than "transpose"

P2, L43: Delete "globally". You could use "based on first principles" instead of "deterministic"

P2, L49: "virtually requested" -> "required"

P3, L78: change "it splits" to "which splits"

P4, L97: What does the acronym SSH stand for?

P9, L200: Change "strong stable" to "highly stable"

P9, L201: change "physically relevant" to "realistic"

P11, Fig. 3: Assumption that traffic profile is representative for anthropogenic heat flux in February is fairly questionable, since heat release from heating is likely much larger in this month than traffic.

P12, L251: The friction velocity threshold was not reduced by increased to 0.1 m s-1.

P12, L263: Replace "About CHIMERE/MUNICH" by "For CHIMERE/MUNICH"

P17, L363: replace "analyze" by "analysis"

P17, L380: replace "queuing" by "congestion"

P21, L402: Again, I propose using the term "difference" instead of "bias" in this context, since the term bias implies that we know what the truth is (which we don't)

P21, L403: replace "close emissions" by "comparable emissions"

P22, L415: The sentence " Figure 7 shows the between our two approaches" is missing a noun.

P22, L416: Replace "confirmed" by "confirm"