This manuscript estimates the NOx emissions from cities and large sources around the globe using TROPOMI NO2 satellite data. I appreciate that the manuscript is thorough and well-written.

I am concerned about the derived NO2 lifetimes by season and month. The magnitude of the seasonal cycle of emissions reported here are in disagreement (and sometimes strong disagreement) with previous literature. Please see Crippa et al., 2020 which shows that CO2 emissions (which can be used as a rough surrogate for NOx in this instance) don’t have a very strong seasonal cycle. Using data reported in Crippa et al., 2020… CO2 emissions are maybe 10% lower in summer in the US and China as compared to winter. In Spring and Fall, CO2 emissions are lowest, perhaps 15-20% lower than the winter peak. The summer-to-winter ratios reported in this manuscript, such as ~0.5 in New York City, ~0.4 in Chicago, ~0.25 in Wuhan are simply not reasonable. Even in Europe, Crippa et al., report ratios of ~0.8, and the values reported here are significantly lower.

My guess is that the EMG method is having trouble discerning the true NO2 lifetime during winter. This is a known issue and is why previous literature focus on summer time emissions (Lu et al., 2015; Goldberg et al., 2019). To prove whether your method is reasonable during the winter, I suggest two potential strategies: 1.) You can apply the method to a model simulation that has data during both the winter and summer; since the emissions in the model are known, the method should hopefully reproduce the seasonal cycle of the emissions that are input into the model (whatever they are). 2.) You can apply the method to power plants which report their NOx emissions for all seasons such as all the large power generation facilities in the US. This will hopefully provide insight on whether the  seasonal effects on the NO2 lifetime and therefore NOx emissions are being correctly accounted for.

With the aforementioned said, I think that Sections 4.1 and 4.4 are useful advancements to the literature. I recommend that Sections 4.2, 4.3 and 4.5 are excluded, and included in a follow-up manuscript that more rigorously evaluates the top-down NO2 lifetime by season.

All suggestions:

Line 39: in-homogeneous —> heterogeneous

Line 43: Missing comma between winter and residential

Line 62: states —> countries

Line 147: Availability of data in March and April?

Figure 1: Suggestion to remove lat/lon lines on the plot. The lines obscures a few cities.

Line 163: Wind direction likely similar between heights, but wind speed can be different, especially if comparing near surface to 1000 m. Maybe something to note.

Line 163: Might also want to mentioned ease of use of 100-m wind speed since it is a standard variable output in the ERA5 re-analysis.

Line 172: Can you clarify? Do you mean every three hours (0Z, 3Z, etc.)? I originally interpreted three hourly estimates to mean three different outputs each hour.

Line 173: Presumably the interpolation is location specific (Europe may use 12Z model data but China may use 3Z for example). Please clarify.

Line 207: earth —> globe

Line 213: Modify "low cloud coverage" to "few clouds". "Low cloud" could be interpreted to mean low in the atmosphere.

Line 318: Verhoelst et al., 2021 and Judd et al. 2020, which you already cite, show a low bias in polluted areas of ~20-30%. I see no problem with using this to scale the final estimates by the value, while also noting the large uncertainty in this conversion factor.

Line 336: Early studies, published in 2018 & 2019 used an algorithm that has since been re-processed. Also there was a different rotation in Goldberg et al., 2019; it was done manually.

Line 350: I am not convinced that these ratios should be this small. Traffic pollution and industrial/manufacturing pollution, is relatively constant year-round and represents a large fraction of emissions. If anything traffic and industrial emissions would be biased to be larger in the summer (traffic peaks in July and lowest in February: https://www.fhwa.dot.gov/policyinformation/travel_monitoring/tvt.cfm). NO2 lifetime, in theory, should be varying much more than you found. So I'm thinking the summer/winter ratios are an artifact of an erroneous lifetime fit. See Zheng 2018 as an example. They report that residential emissions in China (the sector causing the most intraseasonal variation in emissions) is ~5% of Chinese emissions. Even if this percentage is off by a factor of two and power genration varies much more than we suspected (Crippa et al. 2020 show this doesn't vary much by season in China), it would be very unlikely to get ratios larger/small than 0.8-1.2.

Line 381: I'm glad you are transparent with this, but these don't seem to be reasonable. NO2 lifetime should be shortest in summer, similar in Spring & Fall, and longest in Winter.

Line 391: Can't you go a step further and estimate OH for a few cities, perhaps from a model or previous literature, and see if the NO2 lifetimes you derive are approximately similar?

Line 415: Often, Saturday is different than Sunday… Saturday emissions can be quite larger than Sunday. See Goldberg et al., 2021 and Crippa et al., 2020 as an example

Line 420: Also see Goldberg et al., 2021

Section 4.5 and Figure 7: If seasonal emissions are biased high in winter, then these drops attributed to COVID will be overestimated. Sections 4.2 & 4.3 need to revised in order for me to have more confidence in the results presented in this section.

Section 4.6: Please more explicitly differentiate between random errors, which should mostly cancel out since you are averaging over many days of observations, and systematic errors, which would not cancel out.

Line 564: Perhaps may even want to say that re-gridding TROPOMI data to a resolution of 0.1 degrees (~10 km) might be better than a higher resolution (~1 km) for this specific purpose only.

References:

Crippa, M., Solazzo, E., Huang, G., Guizzardi, D., Koffi, E., Muntean, M., Schieberle, C., Friedrich, R. and Janssens-Maenhout, G.: High resolution temporal profiles in the Emissions Database for Global Atmospheric Research, Sci. Data, 7(1), 121, doi:10.1038/s41597-020-0462-2, 2020.

Goldberg, D. L., Lu, Z., Streets, D. G., De Foy, B., Griffin, D., Mclinden, C. A., Lamsal, L. N., Krotkov, N. A. and Eskes, H.: Enhanced Capabilities of TROPOMI NO2: Estimating NOx from North American Cities and Power Plants, Environ. Sci. Technol., 53(21), doi:10.1021/acs.est.9b04488, 2019.

Goldberg, D. L., Anenberg, S. C., Kerr, G. H., Mohegh, A., Lu, Z. and Streets, D. G.: TROPOMI NO2 in the United States: A detailed look at the annual averages, weekly cycles, effects of temperature, and correlation with surface NO2 concentrations, Earth’s Future, e2020EF001665, doi:10.1029/2020EF001665, 2021.

Lu, Z., Streets, D. G., de Foy, B., Lamsal, L. N., Duncan, B. N. and Xing, J.: Emissions of nitrogen oxides from US urban areas: Estimation from Ozone Monitoring Instrument retrievals for 2005-2014, Atmos. Chem. Phys., 15(18), 10367–10383, doi:10.5194/acp-15-10367-2015, 2015.

Zheng, B., Tong, D., Li, M., Liu, F., Hong, C., Geng, G., Li, H., Li, X., Peng, L., Qi, J., Yan, L., Zhang, Y., Zhao, H., Zheng, Y., He, K. and Zhang, Q.: Trends in China’s anthropogenic emissions since 2010 as the consequence of clean air actions, Atmos. Chem. Phys., 18(19), 14095–14111, doi:10.5194/acp-18-14095-2018, 2018.

The study "Variability of nitrogen oxide emission fluxes and lifetimes estimated from Sentinel-5P TROPOMI observations" 

by Lange et al. presents emission and lifetime estimates from TROPOMI NO2 measurements for a selection of source regions around the world,

and investigates weekly & seasonal cycles as well as the impact of recent lockdown measures.

The manuscript is clearly written, provides a detailed description of method & error calculation, and discusses the results adequately, 

also in context with previous studies.  

I recommend publication after dealing with the comments below.

- NOx: x should be subscript throughout the manuscript.

- Line 3: this resolution was not provided in the first months of operation.

- Line 7: "dessert" should be "desert"

- Lines 14&15: 2x "In the atmosphere". I would skip the first one.

- Line 123: Please specify that for the highest latitudes considered here, up to two overpasses are available. 

  Three overpasses only occur for higher latitudes.

- Lines 126-129 (up to "Seo et al."): This is off-topic and should be skipped.

- Line 135: Please explain why the difference in resolution is more extreme at the edges.

- Lines 145ff: Please explain why versions up to 1.03 are mixed in your analysis, but 1.04 is excluded.

  Shortly mention what has changed in 1.04 and refer to the discussion of the low bias of tropospheric NO2 

  in section 4.6.

- Line 153: Do you have a reference for 50% CRF ~ 20% GCF?

- Line 154: I am confused by the different time intervals. 

  In line 144: March 2018 to November 2020.

  In line 197: March 2018 to February 2020 general 

               January 2019 to November 2020 Covid 19

  Why is yet another interval chosen for Fig. 1? Why not March 2018 to February 2020?

- Line 155: The criterion "higher NO2 than their surrounding" would result in a quite different selection of hot spots visible in the mean map.

  I propose to skip the reference to Fig. 1 in this section, but add it section 3.1, where the selection of source regions is explained in detail.

- Caption Fig. 1: Add a reference to table A1 for further details.

- Line 162: Please quantify "low".

- Section 2.3: Using model input in order to correct for the NO2/NOx ratio is probably an improvement compared to just taking a constant value.

  However, I don't understand why the calculation in Eq. 3 was done for a fixed pressure, since the model would also provide the actual pressure

  at the altitude of interest (probably the same as chosen for the wind vector).

  The considered source regions include some elevated locations like Las Vegas, Medupi, or Colstrip (1 km asl!), where 950 hPa is not appropriate.

- Line 219: How far does the - somehow arbitrary - selection by visual inspection affects the generality of the results? 

  Could there be some "selection bias"?

- Line 224: What amount of TROPOMI observations is needed in order to get robust results from this method?

- Line 293: What is "dispersion" meaning in this context?  

- Line 294: or the effects of non-linearities in the NOx lifetime.

- Figure 5: This figure is very hard to read. I propose to

  - add one further panel, highlighting the seasonal dependence. E.g. show the lifetime for different locations as function of season.

  - arrange the 4 panels 2x2

  - enlarge

- Line 430: One important difference is that the tropospheric columns also contain the upper tropospheric background, which is removed in the EMG method by fitting the background B.

- Lines 507ff: This section is quite vague. One main reason for the low bias was identified as the FRESCO cloudheight which is biased low 

(Compernolle et al., https://amt.copernicus.org/articles/14/2451/2021/) causing high-biased AMFs. This was changed with v1.04 of the operational NO2 product.

Please extend this discussion accordingly.  

- Line 613: Please specify "short time periods": how many TROPOMI overpasses are needed?