Review of van der A et al., 2023

Van der A et al. discuss in their manuscript a new version of the DESCO NOx emission data set, based on TROPOMI satellite NO2 measurements. The manuscript discusses a valuable data product and certainly deserves to be published after addressing the points mentioned below in a major revision. In particular the structure and the internal logic of the manuscript should be improved. I think this manuscript could be considered within the scope of ACP, but for my taste, it would better fit into AMT, since it is mainly discussing a new data product.

General points: 

- The discussion about uncertainties, in particular for DESCO, but with the comparisons with other data sources also their error estimation, are out of place (see below), and not very extensive. Some numbers are given, but I would like to know, how the authors received these numbers.

- The DESCO algorithm uses a Kalman filter, so I think the description of DESCO should mention the influence of the a priori to the results.

- Cloud filters are mentioned briefly when the TROPOMI measurements are introduced. However, I would be interested in the fraction of filtered measurements, and if any bias due to the fact that one needs to do cloud filtering are expected.

- In some of the figures, the font size should be increased for better readability.

- The term NOx is very important in this manuscript and thus appears quite often in the text. Please make sure that it is spelled consistently. I think it should be with a subscript x (also in the title).

- The section "Discussion" is very strangely structured. Most of the information here appears for the first time, which is not adequate for such a section (in my opinion). The parts about Belgrade, error estimations and the comparison description should be moved to the corresponding sections above - I really was desperately looking for some of these information while reading the manuscript for the first time.

- The usage of the term CAMS is not very clear throughout the manuscript. It is mentioned that there are two different CAMS products used (CAMS-REG and CAMS-TEMPO), and I think the manuscript should be precise enough to tell the right product everywhere in the text.

- The usage of the "other" data sources than DESCO is not clear to me. Are they considered to be a benchmark, DESCO is compared with? Or are these data known to be uncertain and DESCO should be used to improve such data? Both approaches are used in different parts of the manuscript, even though, these approaches are contradictory.

Specific points:

L27: "Knowledge of NOx emissions...": Isn't this sentence redundant to the previous one? If not, please explain why NOx emissions are important for climate studies.

L29: Regulations of NOx concentrations and emissions are not just present in Europe (e.g. https://www3.epa.gov/region1/airquality/nox.html). I see that the focus of this manuscript is on Europe, but then the sentence should be formulated in a different way.

L44: Suggestion to rephrase: "observation based" -> "based on observations"

L48: I think, the authors are only referring to satellites in nadir geometry here (which are suited best to derive sources), so this should be mentioned. Further, I stumbled a little about the term "total concentrations"; do the authors mean "total columns" instead?

L59: "For a full three dimensional description...": Out of curiosity: How good is the vertical resolution of typical NO2 (nadir) satellite measurements, and does that vertical resolution allow comparisons to 3d models?

L65: "Interesting aspects ...": Are these aspects discussed in this paper? Otherwise I would skip this sentence, or make it more clear that these are important issues which are beyond the scope of this paper.

L96: "The most important updates are ...": For my taste this sentence posed a lot of questions, which were answered in the next paragraphs, so I would suggest to either omit this sentence or refer to the following text.

L100: Is CHIMERE an acronym? If so, please define. Same for HTAP and ECMWF later on.

L107: Maybe I'm not the expert here, but I don't understand what is meant by "a range of maximum 150 km" in this context. Is it the maximum path the trajectories are allowed to travel? Similar for the temporal resolution "(maximum 7.5 minutes)": Is that the maximum time the trajectories are calculated?

L113: Suggestion: "to the simultaneously" -> "to simultaneous"

L118: Please define TM5-MP

L119: I don't understand the sentence "The corresponding averaging kernels..." Please rephrase.

L138: "this new software will be used in future DESCO studies.": I'm confused: Is the newest version of the "software" used in this study or is it only planned to be used in future studies?

L170: Figure 1 is only briefly mentioned in the text, but not further discussed at all. I think it very interesting to finally see some results, but the authors could at least discuss a little that the results are plausible (point out bigger cities or ship routes).

L188: "the agricultural emissions, which are also excluded in DECSO.": How does DESCO exclude agricultural emissions? I think it was only mentioned that biogenic and anthropogenic emissions are separated?

L208: I did not understand the reason why CAMS-REG data was not included into Figure 2, while it was mentioned in the text that the total emissions for the examined region matches well with DESCO.

L221: I think that Figure 2 is really hard to read: The interesting thing here are the differences between DESCO and LRTAP and NEC. So the authors could consider to add a (relative or absolute?) difference plot here, where also the uncertainties of each emission could be added. At the moment, I don't know if the agreement is good or not (only Italy pops out, as mentioned in the text). But are the differences of e.g. France and the Czech Republic notable or not?

L214: "since these emissions are weighted by the land fraction": Why are there then any ship tracks visible in Figure 1? The weighting with land fraction should be mentioned in Section 2, in my opinion.

L229: Are Madrid and Rome really megacities? In my understanding this is for cities of more than 10 million inhabitants in the greater area. I would suggest to rather use the term "large city". Further, these three cities are not the largest in Europe (as mentioned in the text) - at least London is within the region of interest and considerably larger in population.

L230: It is stated that the country total emissions are similar for DESCO and CAMS-REG (even though this was not shown before), but in all cases absolute emissions are higher for the cities in DESCO. To my understanding this also means, that DESCO needs to be smaller compared to CAMS-REG, in order to have comparable country total emissions. So at what places is DESCO smaller than CAMS-REG for these countries? Are there systematic differences? Is this information interesting to improve the CAMS-REG data?

L250 and following: Again, there are huge differences between DESCO and CAMS. Sometimes the trends in the time series seem to be anti-correlated (e.g. August to November 2019: DESCO shows a decreasing trend, while CAMS shows an increase). This should be discussed. Further, I find it really hard to identify the mentioned "3 low emission periods". Maybe these periods could be marked in Figure S1? Additionally, it was said that the biogenic, time dependent signal was removed from the DESCO data. So in Figure S1, only the anthropogenic data is shown, but still has a time dependency? 

L276: "The difference in annual emissions [...] are relatively small ...": I do not agree with this statement at all. For 2019, DESCO is 1.1 Gg(?) higher than CAMS, and 2 Gg(?) higher than E-PRTR. This is 20%-40% relative difference, which I would never call "relatively small"!

L323: Table 1 gives only units in the caption, I would prefer to see them also at least in each column heading. However, the unit (N)Gg/year is a bit strange, since yearly values are shown here. Should this not rather be (N) Gg, since it is already multiplied by the time of one year?

Section 3.3: For most of the discussed LPS, there is little discussion about the comparison between DESCO and the other data sources, apart from the Maritsa Iztok facility. I would like to learn more about the (paritally quite substantial) differences, and why there is more month-to-month variability in DESCO compared to CAMS-TEMPO.

L312: Why are differences of 20% a "reasonable agreement" in this study? As mentioned above: A discussion of error estimates would be very important here.

L336: "... while CAMS is higher for rural regions.": I think I missed this information in the main part of the manuscript?

L347: "very similar on average." I would disagree on that statement given the larger descrepancies for the cities and LSPs. Maybe that is meant to be for the country emissions, but that should be stated. 

L353 Keppens and Lambert -> Keppens and Lambert (2023)

L361: "only gives mainly clear results": This is an awkward formulation, please rephrase.

L362: "improvement can be gained by providing the emissions on a higher resolution.": So why has this not been done in this study? I'm a little disappointed to learn in the end of the manuscript, that there have been studies with better resolution than the one I just spent some time with.

L372 and following: The URLs to the websites should all be included as clickable links within the PDF. Some of them are clickable, but broken (but the plain text works). It would be even better if the data sets would have DOIs, but that is most likely out of control for the authors.

Please see uploaded PDF.