The paper by Sugawara et al. presents observation of noble gases in the stratosphere and discusses the effect of kinetic fractionation in explaining observed vertical gradients. The paper is well written and presents interesting data with a solid interpretation of the effects controlling the vertical gradients and in particular the effect of gravitational settling caused by kinetic fractionation of different isotopes of the noble gases. Overall, I found the paper well written, the data of high quality and the interpretation thorough. I found the section 3.3. about the two-dimensional model somewhat difficult to follow, but this may also be caused by my lack of expertise in this area. The sections 3.4. and 3.5. could be sharpened by deriving clearer conclusions, e.g. on magnitudes of changes that could be detectable. I also suggest summarizing clearer conclusions in the abstract. I recommend the paper for publication after addressing the issues mentioned here and some modifications as detailed below.

Specific comments:

1. 160 and following: what is the rf gas which was used?
2. 172: please clarify how the difference was calculated: is this the mean absolute difference (MAD)?
3. 201: how was the CO2 mole fraction corrected for gravitational separation, how large is this effect? Please briefly explain here, even if you reference other literature.
4. 245: In the case of 29N and 28N the delta-values and the normalised data values are equal. Neverthe, when comparing to normalised values, I suggest to denote the value with a subscript n.
5. 242-257: would it make sense to summarize these values in a table?
6. 436: I think it would be good to distinguish more clearly between molecular diffusion and eddy diffusion. In my understanding, the latter would certainly not lead to any kinetic separation and should be distinguished more clearly.

Section 3.4.: I understand that the overall conclusion from this is that an impact of a changing Brewer-Dobson circulation on the tropospheric isotope ratio would be hard to detect. The authors have focussed on a slowing-down of the BDC, however, models consistently predict an increasing BDC. It might be worthwhile discussing the detectability of an increasing BDC.