Liu et al. describes a new approach to inferring NOx emissions from cities by combining two previously published methods. The first, the 1D MISATEAM approach described in the Liu et al. 2022 reference, is a whole-city mass balance approach that divides space-based NO2 column data by wind speed and direction and finds the emissions that balance the transport and chemical removal of NO2. The second, the divergence-based approach described in the Beirle et al. 2019 reference, is also a mass balance approach in essence, but one which is applied to individual grid cells. There the difference between horizontal flux of NO2 into and out of a grid cell is taken to represent the sum of emissions and sinks in that grid cell, with the sink assumed to be the first-order chemical loss of NO2. By combining these two methods, this paper is able to use lifetimes and background NO2 columns derived from the whole-city analysis in the grid-cell level calculation.

This is an interesting evolution of our ability to directly constrain NOx emissions from space without use of computationally expensive models. The paper generally does a good job of evaluating the veracity of this method using synthetic data, which demonstrates that this method has good skill in recovering known emissions assuming no systematic biases. The uncertainty estimates are reasonable and justified, though I have one suggestion for an additional test. There are a few points that can be strengthened, which I will detail below. However, this is already a strong paper and I recommend publication after the points below are addressed.

- Point 1: the only limitation I saw in the validation with NU-WRF data was that possible systematic biases in the AMF were not tested. If I understood correctly, the synthetic NO2 columns used in the validation were an integration of the NU-WRF profiles without any AKs from the NO2 retrieval applied. Thus, this essentially assumes perfect AMFs. We know from Laughner et al. 2016 (https://doi.org/10.5194/acp-16-15247-2016) that AMF biases from the a priori profiles can lead to biases in the emissions and lifetime derived from methods similar to the 1D MISATEAM approach. I suspect that such biases would be fairly small in this case, as the MINDS NO2 retrieval used in this study does have reasonably high resolution a priori profiles (0.25 deg). But we also know from Valin et al. 2011 (https://doi.org/10.5194/acp-11-11647-2011) that even at ~25 km, chemical transport models don't capture the full nonlinearity of NOx chemistry.

I think that there is a straightforward way to test whether any AMF biases present in the NO2 retrieval are large enough to affect the 2D MISATEAM method. If you were to repeat the test where you derived emission by applying 2D MISATEAM to the synthetic NU-WRF columns, but this time apply MINDS AKs to the NU-WRF profiles rather than doing a simple column integration, then the emissions derived in this test should reflect the impact of an imperfect AMF. By comparing these imperfect AMF emissions against the emissions derived using the NU-WRF columns without AKs (that represent a "perfect" AMF case), that difference should reveal any systematic impact of systematic AMF biases on the 2D MISATEAM emissions.

- Point 2: there is one sentence at the end of Sect. 3.2 that could use additional justification - "The slopes of the linear regression lines in Fig. 5 decrease from 0.91 in 2019 to 0.85 in 2021. This can be attributed to the long-term trend of decreasing emissions in the US, primarily driven by the downturn trend in vehicular NOx emissions (McDonald et al., 2018)." My concern is that 2021 may still include effects from the COVID-19 pandemic. In Laughner et al. 2021 (https://doi.org/10.1073/pnas.2109481118), we see that metrics for traffic and commercial flights (globally as well as in Los Angeles and San Francisco specifically), remain well below their Jan 2020 levels at the end of 2020.

If this conclusion (that the 2018 to 2021 decrease in NOx emissions is part of the long term trend in the US) is an important part of your work, I'd strongly suggest looking at at least the Google mobility trends (https://www.google.com/covid19/mobility/) and possibly state/city level traffic metrics (e.g. CalTrans PEMS, https://pems.dot.ca.gov/) to check if the underlying traffic driving a substantial part of these emissions had returned to pre-pandemic levels to support this conclusion. If this conclusion isn't critical, then I would recommend adding a caveat that it could include some lingering effects of reduced traffic during the pandemic.

- Point 3: unless I misunderstood, it seems like you should be able to check for closure of emissions between the 1D and 2D MISATEAM results. That is, the emissions which could be output by the 1D MISATEAM algorithm as in Liu et al. 2022 should represent the total city emissions, and so should be approximately equal to the sum of the gridded emissions derived in the 2D MISATEAM approach. In particular, I wonder if this could be a useful quality check to allow you to expect this method to more cities around the world without needing to validate each city with synthetic NU-WRF data. It would be interesting to see if the cities listed in Table S1 that failed NU-WRF validation also have these two emission estimates (from 1D MISATEAM and this method) differ by more than their uncertainty.

- Point 4: it seems like the 2D MISATEAM method implicitly assumes that the background NO2 is the NO2 above the boundary layer. Otherwise, it doesn't make sense to me to use only the non-background NO2 in the calculation of chemical loss (Eq. 3). Is this true? If so, it would be good to explicitly state that assumption.

- Point 5: I was initially confused by the discussion of the lifetime uncertainty in Sect. 3.3 (lines 221 to 225). The way the uncertainty analysis was presented made me think that the lifetime in Eq. (3) was a single lifetime used for all cities, rather than having unique lifetimes for each city but that does not change in time. On a second read, I found the sentence at line 109 that indicated that the lifetime and background were calculated for each city. Still, it might be good to restate in Sect. 3.3 that the constant lifetime over several years is different for each city. Also, I assume that the reason only 14 cities could be used for the year-by-year lifetime standard deviation in the uncertainty analysis is that they were the only cities with enough good quality data to derive robust lifetimes separately for each year? If so, please state that and list which cities those 14 were. That will be useful documentation in case it is later found that those 14 cities aren't representative of the trend in lifetime for the 39 cities for which emissions were estimated.

This manuscript presents an improved top-down NOx emission estimate methodology using TROPOMI for select US cities and discusses the method validation and outputs. The improved methodology is the combination of two previously published and widely accepted methods, developed respectively by the two leading coauthors of the manuscript. While I like the manuscript is concise and generally well-written, my main concerns are lack of explanation in some key places and also the lack of details on the derived emissions.

1. The improved emission mapping algorithm, 2D MISATEAM, is the foundation of the paper. I found Section 2.2 as written does not provide a sufficient justification and motivation for it. Line 93-94 simply states that 2D MISATEAM “is capable of mapping NOx emissions over urban areas.”. This statement is not followed by any justification, making it a speculation. The authors need to provide more details on the precedent methodologies, namely 1D MISATEAN and 2D divergency method, as to their respective pros and cons that motivate the development of 2D MISATEAM and how the presented 2D MISATEAM method overcomes the shortcomings of the precedent methods.
2. The limitation of 2D MISATEAM and its applicability to outside of US should be discussed better. The paper uses the new methodology to large US cities based on population. If the community wants to adopt 2D MISATEAM to other countries/regions which have different population sizes than the US, what should they use to select the suitable places? Does it require that the city has a well-defined urban core with concentrated emissions so that it means certain assumptions in the shape of the urban plumes, etc?
3. The paper states that TROPOMI NO2 columns from May – Sep of each year during 2018-2021 were used to derive top-down emissions. It is not clear what’s the temporal time step of 2D MISATEAM when it derives top-down emissions. Does it apply to monthly-averaged TROPOMI NO2 using monthly-averaged winds to derive monthly mean emissions per city, or does it apply to May-Sep mean of those quantities and estimate May-Sep averaged emissions? Also, what determines the temporal resolution suitable for 2D MISATEAM? For example, if one wants to use to derive weekly or even daily emissions, assuming TROPOMI has plenty of good pixels for such a short time period, is there anything assumed within 2D MISATEAM that prevents such application from being successful?
4. The manuscript does not provide a good amount of details on the derived top-down emissions over the US cities and how they compare to the NEI inventory. In the abstract, the last sentence states that “there are noticeable differences in the spatial patterns of emissions in some cities” between the TROPOMI-derived and NEI inventory. I don’t find where the manuscript elaborated on this main point. Figure 1 is the only place I saw the spatial pattern with a city is presented, but that’s only for NYC. To prove the point for “some cities”, at least two more cities should be presented. Is the within-city spatial pattern resolution a key strength of 2D MISATEAM? What’s the key innovation in it that makes it outperform 1D MISATEAN and 2D divergency method in achieving this?

Minor comments:

• Line 73: the TROPOMI footprint changes over the study period. Specify the changes.
• Line 79: Specify what official product of TROPOMI NO2 is and provides a reference.
• Line 88: There is no cloud screening applied? Why?
• Line 245: The uncertainty in the derived NOx emissions is 47%. How does this affect the comparison with NEI? The abstract last sentence attributed all the discrepancy to the NEI. Will the uncertainty in the top-down emissions explain some of the discrepancies?