This paper looks to address the 3-d radiative transfer effects of urban landscapes on NO2 retrievals. The authors use monte carlo simulations from MYSTIC for a simplified urban landscape in Zurich to examine the impacts of buildings AMF calculations (so called 3D-box AMF).  This is a very interesting study with significant implications to airborne retrievals. The only part the authors don't seem to address is how they would account for such 3-d effects in actual airborne retrievals. Many of the assumptions they make in this study would likely not be applicable to real world retrievals. Maybe this will come in a later paper, but it would make the paper stronger to explain how this could be translated into actual retrievals. 

General Comments: 

1. The authors note that 50% of the NO2 sensitivity is from outside the ground pixel for a nadir viewing geometry. They also note that the urban canopy module in MYSTIC currently only supports Lambertian reflections. Could the authors elaborate on how the Lambertian assumption would effect their results? Would it be a safe assumption that accounting for specular reflection you would have less light scattering in from outside the ground pixel?

2.  The simple assumptions about albedo seem to not be very realistic. It would be nice to do simulations with more reasnoable albedos or at least provide some discussion on the possible errors from making these assumptions about the albedo. 

Minor Comments: 

Pg 1 Ln 18. "by fuel combustion by traffic, heating systems...", suggest changing to "by fuel combustion, traffic, heating systems..."

Pg 2 Ln 53. Please spell out MYSTIC completely, the Monte carlo code for the phYSically correct Tracing of photons In Cloud atmospheres

Pg 8 Ln 186 "The reflectance was set to 0.1", please clarify that you are referring to surface reflectance, not top of atmosphere reflectance 

Pg 15, Fig 10 It is curious that the building in the bottom right of the figure shows now shadowing effects despite the sun being to the west. Can you explain why this building is not effect by shadows effects. Even if it was in the shadow of the building to the west, one would still expect and impact on SCDs. 

Pg 15 Line 316 "with 19% and 24% for simulations without aerosols and without and with buildings, respectively". This statement is a bit confusing. I suspect you mean 19% from outside the optical path with no aerosol/no building and 24% from outside the optical path with no aerosol/with building, but it is not clear 

Pg 15 Line 317 should be "The effect becomes more.."

This paper describes the influence of 3D radiative transfer and shadowing on airborne NO2 retrievals, using a case study over Zurich. The paper is well-organized, concise and easy to follow. I have a couple of general and specific comments. After those are addressed I would recommend it be published in AMT. The study results are interesting but I am not sure how the results can be transferred over to a practical application in retrievals; however, it will be a good reference for the impacts of 3D radiative transfer on these kinds of measurements.

General comments:

I find the description of the motivation to be unconvincing. The authors point to possible 3D effects as being responsible for the discrepancy between high resolution airborne NO2 maps over urban areas and city-scale urban models, and point to Figure S1 for an example. First, to drive home the need for this study, it would be good to see the motivating figures in the main paper instead of the supplement. To me, the observed and modeled NO2 maps look so very different that I doubt the source of the differences is entirely or even primarily 3D radiative effects. I would suspect issues with the model like inaccuracies in mixed layer height, transport, emissions, and chemistry, or issues with surface reflectance and profile shapes in the retrievals. For instance, there are what look like three plumes in the southwest corner of the map which actually look quite well-represented. Why are these represented fairly well but other plumes are not? It may be that 3D effects are the main reason for observed/model discrepancies, but the current example is unconvincing.

The study’s motivation would be more convincing if: 1) The new calculations were actually applied to APEX data to calculate new AMFs, perhaps at the end of the paper, and the new maps showed better agreement with the model or improvements in resolution, or 2) a simulation were done using the GRAL NO2 columns to simulate airborne measurements that use 1D radiative transfer, and these simulations were found to show significant smearing. Even if this is not possible, I would suggest leaving out the example figure, and be more nuanced in the motivation, i.e., along the lines of “we explore 3D effects as a possible contributor to smearing…”. The later results will show whether or not they are significant.

My second general comment is that it would be nice to see some discussion of the practical implementation of these calculations. Is it too computationally intensive to use for actual campaign AMF calculations? How would a more realistic albedo field change the results?

Specific Comments:

Line 3: It’s unlikely that the entire cause would be 3D radiative transfer effects. Should qualify with wording like “3D radiative transfer effects may contribute to this discrepancy due to….”

Line 152: Can you give an estimate (maybe in results section) about how much the results would change for shorter wavelength fitting windows? Shorter wavelengths are used in almost all other airborne and satellite instruments that measure NO2, so this would be most interesting for the majority of readers who do not use APEX data.

Technical Comments:

Some mixing of verb tenses in paper. For example, in abstract say “We compute..” then next sentence says “We found…”

Line 63: I’m a bit confused at the wording of this sentence. Should it be “The model, however, is able to …”?