General comments

This preprint presents high-resolution air quality modelling at local scale/street level across Europe, performed with the uEMEP_v6 model, which downscales emission input and combines regional calculations with the EMEP model with Gaussian plume modelling of receptor points to obtain annual mean concentrations of NO₂, PM_2.5, PM₁₀ and O₃ in a very high resolution subgrid (down to 100 m resolution). Results presented are comparisons of EMEP and uEMEP model results country by country for Airbase monitoring stations across Europe, as well as sensitivity studies with respect to resolution, weighting of relative traffic distribution, proxies for residential combustion emissions, use of national emission and proxy data with higher detail than the EMEP data, and NO₂ chemistry schemes for the NO₂-NO_x-O₃ reactions. Model results are seen to significantly improve in comparison with measurements, for NO₂ and O₃ when the uEMEP model is applied whereas little improvement for PM_2.5 and PM₁₀ is gained using the downscaling approach.

The manuscript addresses the highly relevant scientific question concerning how to obtain high resolution air quality estimates at the urban scale in an operational way, using publicly available data. The different parts of the methodology are known, but the combination is novel and the uEMEP model tool is potentially extremely useful for air quality and population exposure studies across Europe. The manuscript is well-written and the structure and argumentation of the study is easy to follow with a few minor details that could be clarified (se Specific comments).  All model code for the uEMEP model and the visualization tools is publicly available for this study.

Specific comments:

When investigating annual mean values, a rotation symmetric approach for the Gaussion dispersion is used. This implies the assumption that the wind is equally distributed from all directons over a year. In reality these wind distributions will probably be quite different. Could the authors elaborate on if this rotation symmetric approach is more or less accurate in different locations? And what potential comparisons of the approach to applying actual annual met-data would show?

When the OSM data are applied across Europe, a weighting based on the Norwegian traffic data is used – can the authors elaborate on what this means for the distribution of local emissions in the rest of Europe?

The NO-NO₂-O₃ chemistry for annual mean values is based on a calculation including a frequency distribution of the concentrations of NO, NO₂ and O₃. It is a little unclear from the manuscript, if these frequency distributions are acquired from Norwegian stations only, or if all available measurements from Europe have been taken into account? The dependency on solar input in the photochemical reactions must mean that the frequency distribution will differ across Europe?

The results of the uEMEP calculations correspond to street-level, but the building configuration is not included, and common situations with development of street-canyon circulation vortices are therefore also not taken into account. Can the authors elaborate on what this means for the results at street-level in the large cities with tall and dense building mass?

Regarding PM_2.5: All annual mean concentrations for (NO₂ and) PM_2.5 increases when the uEMEP downscaling is applied compared to the EMEP model results. In two countries (Austria and Finland), the EMEP model is already overestimating the PM_2.5 concentration, and applying the uEMEP model only increases the overestimation. Wouldn’t the authors expect that downscaling using proxies would give a more precise result for the distribution, and thereby a more accurate replica of what is observed? Or is there an underlying risk, that uEMEP increases the concentrations in general?

Figure 10: is this results for the whole of Europe, i.e. a mean of all countries?

Figure 11 and 12: It would be good with a little more explanation in the Figure captions, e.g. a note whether this is all of Europe, or only Norway.

Figure 12: the conclusion that the correlation is clearly highest for power law index 1 is putting much trust in the decimals of the correlations. As the numbers are 0.567 (~0.57), 0.574 (~0.57) and 0.557 (~0.56), one could wonder how much the third decimal of the correlation estimate is worth in terms of accuracy?

In the discussion: “Though the problem remains that uEMEP does not take into account dispersion in street canyons, where a number of traffic site measurements are made, it is generally the case that the spatial representativeness of the uEMEP calculations is suitable for comparison with these measurements.” How does the authors know, that this is the case?

Technical comments:

Figure 14: the order of the components is NO₂, PM_2.5 PM₁₀, but in the text the order of discussion is NO₂, PM₁₀ and PM_2.5. Would be easier to read if the order is the same both places.

Line 315: all five methods are described in suppl,. but in the figure, there are results for 6 methods? Not easy to follow the names of the 6 methods in the text in the manuscript as they are not consistently defined (nor in the supplementary material).

Line 319: “Since the Romberg scheme is specifically designed to reflect measurements, providing the correct NO₂/NO_X ratio, this means that the chemistry schemes are overestimating the NO₂ contribution when applied to annual mean concentrations.” It is somewhat difficult to understand what is meant here?

Table 2: within an region – should be: within a region

Line 374: verses – should be versus

Figure S5: verses – versus  + include should be – included

This publication present a downscaled version of the EMEP model and its application to Europe for air quality pollutants. The topic is very interesting and certainly relevant with respect to the scope of the journal. The methodological approach is sound and the work is well structured and well presented. I list below a few comments.

Comments:

1. The introduction dives directly into the proposed uEMEP downscaling. I would suggest starting with a paragraph to provide some context and explain why we actually need such a downscaling.
2. While the Authors state (L107) that “uEMEP downscales only primary pollutants”, downscaled results for O3 are discussed and shown. A few lines to explain how secondary pollutants like O3 can be downscaled would clarify a possible confusion.
3. Orography variations can be important over two 0.1deg grid cells. Can the authors comment on the applicability of their Gaussian approximation to hilly cities?
4. A few remarks regarding Table 2:
   1. Caption: an region à a region
   2. Please add that the region of +/- 0.1 deg also represents 2x2 grid cells.
   3. If my understanding is correct, the source column should precise that only primary contributions are considered (e.g. primary traffic…) and that non-local EMEP includes all secondary contributions as well.
   4. It is unclear how the non-local EMEP is obtained from the 2x2 EMEP local fractions. If summed up, how is double-counting avoided? Indeed, the non-local fraction of one given cell reaching the neighbouring cell will be counted as non-local while it is actually local within the 2x2 area, right?
5. Can the Authors detail the 2x2 grid cell kernel calculation (L48-49)? For an emission located within a given EMEP cell, say near its SW corner, how are these 2x2 cells defined?

Technical remarks

1. L45: even though detailed in other references (listed), some explanations on how the EMEP local fraction is calculated would be helpful in this work.
2. L56: Can the authors comment on the height of the first vertical layer (50m). Does this impact the downscaled results?
3. L85: Is the split in tiles motivated by CPU gains only or does it improve the efficiency for other aspects?
4. L93: “Weights are based on Norwegian average road situations”. Can the authors comment on the validity of this assumption when applying it to other countries
5. L99: The choice of population as proxy to redistribute residential heating emissions is known to lead to important issues for some cities (as shown later in the document). Is this initial choice related to data availability?
6. L171 and L334: The spatial representativeness of some traffic stations, especially those located in street canyons is lower than 25 meters. Can the authors comment on their choice to keep those stations anyway for the comparison?
7. L169 and others: Note that the correlation is actually “r” not “r²”. The latter is referred to as the coefficient of determination. Please adapt the text or figures accordingly.
8. Figures 10 and 12: please explain which station or station averages are shown.
9. L229: “the” is doubled twice!
10. L311: Can the Authors comment on the robustness of this fitting for other years?