The authors present an interesting work merging retrievals from several satellites to investigate enhancement ratios of CO₂, CO, and NO₂ from a large number of cities around the world. The methodology used to calculate the enhancement ratio appears robust, building off work previously presented in the literature, and applied to newer satellite datasets. Given the growing access to greenhouse gas measurements from space, I find the work an important study as the scientific community strives to find the appropriate analyses and interpretations to deploy with such global measurements.

The manuscript is generally well written and appropriate references are included throughout, however, I find the overall presentation of the work could be improved in the areas described below.

• The methodology section would benefit greatly from the inclusion of more figures to illustrate the various steps of the analysis and aid the reader’s understanding. I find this particularly important since several satellite products are used, each with distinct spatial coverage and overall data handling. For example, the differences in spatial coverage between the various satellite are not shown, nor is the colocation of OCO-2 and TROMPOMI/OCO-3 SAM data (section 3.1), or the masking that results from taking only using data from the sparser production (section 3.3.3). Similarly, the plume extraction presented in section 3.2 is not shown, nor is the impact of smoothing on the TROPOMI/OCO-3 data (section 3.3.1). At a minimum, a figure showing the different spatial sampling of the three satellites should be included.
• There appear to be two distinct methods used for the enhancement calculation depending on which product is used in the analysis; a cross section taken downwind of the city (OCO) and the enhancement over the entire bounding area for the city (TROPOMI). The manuscript jumps around between these various methods, which is difficult to follow. I suggest clarifying the these two complementary methods within the text or with some sort of figure summarizing the methods.

More specific comments are below:

• Representativeness of enhancement ratios – Due to a limited number of overpasses (e.g. 1 for Toronto), how representative do you think some of these ratios are to the timescale of the inventories? I believe a brief discussion concerning this is warranted since the number of overpasses per city varies, while it appears conclusions about the validity of inventories are drawn from all cities regardless of the number of overpasses used in the ratio calculation.
• Do you expect the biosphere to impact your CO₂ or CO enhancements, and subsequently your ratios?
• Lines 191-194 – How are the secondary sources identified?
• Lines 194- 199 – What explains the discrepancy between the MERRA-2 winds and the plume direction, and the resulting variability in overpass retention rates? Is it errors in the wind direction and/or issues with the automatic filtering scheme since manual inspection and correction is required?
• Figure 1 – How are the red points assigned to the city? Without more information or geographical details about the city and location of the overpass, it is hard to interpret.
• Line 273 – What is allowed to change in the bootstrap? Are the anomalies of each species within each overpass resampled?
• Lines 284-285 – “As we are correcting the enhancement ratios, the effect of dispersion cancels out in the ratio.” This sentence is unclear to me. What dispersion?
• Figure 2 – The color scale in the ‘Background’ plot is hard to interpret using the color bar. I understand this the same color bar as the observed data which has a larger spread of values, but I cannot tell if it is all one value for the background or not. This might be an issue with my screen, but it is something to think about.
• Figure 4 & Section 3.4 – It was not until I saw Figure 4 that I realized that a single regression is calculated for all anomalies across multiple days. I would suggest explicitly stating this in the text since, in contrast, Figures 1 & 2 are for single days. In this analysis, are the number of anomaly data points that same for each overpass? If not, is there potential for a subset of overpasses with higher density of data (i.e. more points) to drive the combined regression?
• Figure 5-7 – The number of overpasses that each measurement ratio is based upon should be included in the figure. It seems that this information is at least partially available in the appendix, so another option would be to include a reference to those tables in the caption.
• Figure 5 – I cannot see where the “light blue areas” are on top of the measured ratios. I see black error bars, so perhaps this is a typo.
• Lines 355-358 – Is the comparison for cities both in this work and Lama et al. shown anywhere, in something like a figure or table? If not, I believe this should be added, possibly to the appendix. If this comparison is buried into one of the figures, a reference to it should be added to the text.
• In the introduction, you mention the TIMES scaling of CO₂ by Nassar et al. Do you apply those scale factors to account for the diurnal variability of CO₂, since your measurements are based on afternoon overpasses? You mention this correction (the magnitude of which is not included) in section 5.2 for LA and Indianapolis, however, it is not clear if these corrections are applied to the ratios shown in Figures 5 &6.
• Lines 425-429 – You present the NO₂ emission estimates based on ratios with and without the NO₂ lifetime correction. The reader should not assume the un-corrected emission rate is accurate, right? If this is the case, I would suggest you state that because as written it is not clear which value for NO₂ is the one the reader should remember.
• Line 455 – underestimations of which species?
• Table 2 – What metric is used to discriminate a “poor” linear relationship? Low correlation coefficient, low R?
• Tables 2, 3 – You should include how many TROPOMI overpasses are used to generate the NO₂:CO enhancement ratios for each city. 

Overall comments:

The paper is excellently written and covers a subject of much interest in the scientific community. The methodology employed appears sound and the obtained results are adequately discussed. However, the way in which uncertainties are discussed and represented could be improved. Here I think of a two specific  issues.

• In order to derive the emission ratios, using total column dry air-column mole fractions from satellite sounders one needs the know to what extent the observed variations in clean vs. urban air emissions are translated into the satellite signal. For NO2:CO ratios this was done by Lama et al. (2020) using CAMS simulated profiles onto which the total column averaging kernel was applied. Here the anomalies are divided by the surface averaging kernel values (I assume it corresponds with the value in the total column averaging kernel array that corresponds to the lowest altitude layer although this is not that clearly stated). Using these column averaging kernel values to look into aspects that pertain to a specific partial column, comes with its own set of uncertainties that need to be discussed. For instance the column averaging kernel values correspond to specific airmass layers with specific dimensions. These vertical dimensions could differ between sounders (and retrieved species), the Planetary Boundary Layer under consideration etc., and with that, biases could be induced into the analysis. This is particularly relevant for NO₂, which features relatively low kernel values near the surface, but also CO which features an equally strong gradient near the surface (an additional figure showing typical total column averaging kernel profile shapes would be a useful addition to the paper).
  
  Related to this, dividing the anomaly by the surface averaging kernel will only yield correct results if there is no a priori contribution to the anomaly (i.e. the a priori used in the satellite retrieval may not differ between what is considered background and urban air). This needs to be verified and clearly stated.
• Secondly, some of the obtained emission ratios are derived from a very limited set of data. Foremost among these is Toronto with only one OCO-2/3 overpass. This begs the question to what extent this affects the overall uncertainty. Valid indicative information in this regard would be to perform the analysis for better sampled stations on an overpass per overpass basis and see what the obtained range of results would be and whether or not there is a temporal/seasonal aspect to the variability.

Specific comments:

• Line 103: add references for each of these products
• Line 188: concerning ‘points away’: Is there a degree threshold?
• Line 194: how are these corrections implemented? Could the misdirection of the plume also be related to the injection height of the emissions?
• Line 228: Do you use certain fixed criteria to do this manual selection?
• Line 291: assuming equal area?
• Line 295: 20% of the uncertainty itself, the initial emission ratio, the final emission ratio, the correction?
• Line 357-364: It is stated that the differences between this study and Lama et al. prior to the lifetime correction falls within 25% for most sites. Could you discuss these differences in light of your (and Lama et al.)  uncertainty estimates? An additional figure would also help the discussion. The authors also point to the difference in sampling rate (each overpass vs. a limited number of days in Lama et al.) but this aspect alone seems inconclusive to describe the particularly large discrepancies at some sites.
• Line 448: Are we referring to lifetime corrected or uncorrected data? This is sometimes unclear in this section of the document. The agreement is sometimes different by several factors. ‘Good’ might therefore be nuanced.
• Figure 1: It is unclear to me what the red area points to. It seems that this would be the downwind area that is considered for the analysis. But it also features some upwind areas?
• Figures 6: Particularly Delhi features a correction that is larger than a factor 6. Could you elaborate why this particular city stands out?