Dear authors,

Thank you for the detailed responses to all reviewer comments and careful revisions of the manuscript. It is my pleasure to let you know that I consider your manuscript acceptable subject to small technical corrections that do not need further iteration with myself or the reviewers.

I'd like to add one comment along the lines of reviewer #2 who was worried that the overall message of the manuscript may be too optimistic:

It is certainly true that all calibrations were conducted under real-world conditions, but all calibrations (and training of the RF model) were conducted at one single location. Whether the same data quality can be achieved at other locations under different conditions (with different meteorology and pollutant source mixtures) is a key question that was addressed only to a limited extent. In that sense, adding the analysis of sensor performance at the Allegheny County Health Department in the revised manuscript was important. Nevertheless, in my view it will be crucial to extend this type of evaluation to more locations in the future to demonstrate the overall performance of a sensor network in a city.

Furthermore, even in the revised manuscript I was still fairly confused about the availability of the individual sensor units and how they were used for training and testing. I think it would be very useful to add a figure to the supplementary material showing the availability of all sensor units (e.g. as horizontal bars, with the horizontal axis being time) and highlighting the periods used for training and testing e.g. with different colors.
It is very confusing, for example, to say (P13, line 3) that "any data remaining after training were used to test model performance, provided there were at least 48 hours of testing data" and in the next sentence that "the size of the testing data set varied from 1.4 to 15 weeks". A period of 1.4 weeks is certainly more then 48 hours. How to these numbers relate?

A few small additional comments:
- Last sentence in abstract: To better emphasize the importance of high quality sensors along with machine learning, I suggest changing this sentence to:
"From this study, we conclude that combining RF models with carefully controlled state-of-the-art sensors as in the RAMP monitors appears to be a very promising …"

- Page 14, line 13: I was also confused about the permutation. Of course we know what "permutation" means, but it is not clear whether values or variables are permuted or randomly shuffled. I think it would be clearer to change "(i.e., randomly shuffled)" to "(i.e., values of this variable are randomly shuffled)"

- Page 16, line 31: I think the project was conducted in Portugal not in Italy. Aveiro is certainly in Portugal.

- Page 19, Line 28: Change "Figure 13" to "Figure 15".

- Page 20, Line 16: I guess it should be "limit levels OF SO2"