Many thanks to the authors for this manuscript and for the opportunity to review it. 

The comments were uploaded in the form of a supplement.

In this paper, different methods are applied and evaluated for estimation of background concentrations from air pollutant measurements that are influenced by emissions from local sources. This allows to determine the contribution of local emissions to the total measured air pollutant concentration. The different methods are tested using measurements of nitrogen oxides, carbon monoxide, carbon dioxide and PM2.5 from four sites in Toronto, Canada.

The paper is generally well written, interesting and relevant and fits to the scope this journal. It should therefore be published in AMT. My only criticism is that the paper is unnecessarily long, The message and conclusions of the paper are relatively simple, it should be possible to shorten the paper without losing information. I enjoyed reading in the beginning and became tired in the results and discussions section because of repetitions and my attention waned. A shorter and more concise version of this paper would definitely improve readability.

There are repetitions that can and should be avoided. It is a bit tedious for me as a reviewer to list all the repetitions here. I would like to leave it to the authors to revise the text accordingly. Only two examples: a. The discussion on the different background situations depending on the proximity to pollution sources (e.g. page 17) can be more concise and shorter.  There is a repetition in the description of the measurement sites, e.g. the characteristics of Hanlan's Point is repeatedly mentioned. b. The conclusions and recommendation section includes repetitions from the results and discussions section, the conclusions chapter should focus on the given recommendations and can therefore be shortened.  

Additional comments:

Page 11, line 277: I think "multiple linear regression" would be the more common terminology (instead of multi-variable linear regression).

Page 12, Table 2: In the note on PM2.5 from the highway upwind background it is stated, that only periods were included when sensors were upwind, whereas all other sites were not restricted by wind direction or speed. What about the other pollutants at the highway upwind site? Were the measurements also restricted by wind direction and speed? From reading, I assume that the measurements at the highway downwind site were not restricted by wind direction and speed. Perhaps this should be made clearer in the text and why this was done (there are situations (southerly wind) where the down-wind location is more in line with background conditions). 

Page 14, Figure 2: What is the temporal resolution of the values used for the box plots. Please provide this information.

Page 15, lines 368/369: It is stated that CO2 baseline concentrations have at all sites been aligned during measurement preprocessing, it is, however, not described how this has been done. A short description of the applied method should be given.

Page 15, lines 373-375. The authors argue that the urban background location (Downsview) is not suit-able for background estimation when looking at minute or hourly data, but would provide a ‘reasonable estimate’ when looking at long-term averages (24 hours and longer). What does ‘reasonable’ mean, please clarify. Averaging does not eliminate systematic differences, but may reduce the differences. So whether averaging is helpful or not depends on the spatial concentration gradients that need to be resolved.

Page 15, lines 365-375: Discussion about diurnal variation of CO2. I'm missing some words on CO2 sinks in the vicinity of the sites and their potential influence on the measured CO2.  

Page 15, line 376-378: The authors interpret the minimum in afternoon PM2.5 at two of the sites as being from evaporation of ammonium nitrate around midday. It is not correct/accurate that ammonium nitrate is a precursor of PM2.5 that is emitted in the morning (it is a secondary constituent of PM2.5). Please correct the wording. Evaporation of ammonium nitrate might contribute to decreasing PM2.5 in the afternoon, however, the main reason for the lower concentration of PM2.5 is likely the lower stability of the lower troposphere and better vertical mixing.

Page 16, Figure 3, diurnal variation of PM2.5. There is a rather large difference between PM2.5 at the highway upwind bkg. Site compared to Downview and Hanlan's Point. This systematic difference seems to be in contradiction to the discussion on the homogeneous spatial distribution of PM2.5 in the manuscript. It seems that the measurements at the highway upwind bkg are biased low. The authors argue in lines 385 to 389 that because of the homogeneity of PM2.5, urban background sites provide rather generally a good estimation of background PM2.5. This is only true, when the measurements are of high quality and without a bias. This seems not to be the case for the highway upwind measurements. The authors should address this point.      

Page 18, Figure 4: Labels on y-axis of the density plots are missing. Should be included.

Page 19, Figure 5: It is surprising that the XGBoost bkg. and the highway upwind bkg. are always super-imposed. The authors mention this in the legend, but give no explanation. Why is this?

Page 25, lines 587-588. It is stated that measurements occurred in winter with low temperatures. This information should be given earlier in the text. This is also important  for the interpretation of the CO2 measurements. For measurements during the warm season, the uptake of CO2 from the biosphere and respiration of plants during the night can significantly complicate the analysis of CO2 measurements. The authors should briefly address this point.