Review of the manuscript: “Potential of TROPOMI for understanding spatio-temporal variations in surface NO2 and their dependencies upon land use over the Iberian Peninsula” by Petetin et al.

The manuscript describes the analysis of NO2 levels over the Iberian Peninsula for different land use categories using satellite observations and ground-based data. The study focuses on the analysis of weekly and seasonal variability and theri dependence on land use properties. This is a quite nice and novel approach to combine land use and air pollution information. The manuscript is suitable for publication once the authors address the following comments:

1. L4-5 “(considering the TrC-NO2 PAL product recently developed using a single TROPOMI 5 processor version, thus ensuring consistency over the time period)”-> Maybe this is not necessary in the abstract, you can mention the pal product but maybe avoid using 20 words on this.
2. Figure 2. Would it be possible to label the cities somehow in the figure? Not everybody is necessary familiar with the location of the cities. Also, for the administrative border: did you try white? This black on dark blue is a bit confusing. These are not a deal-breaker, just a suggestion.
3. Figure 3. Maybe you could remove the administrative borders here and leave only the city borders? Since all these administrative data are available at different resolution, you see several different lines overlapping but not precisely, and it looks a bit confusing to me.
4. Figure 4-5. It is not completely clear to me how do you average in time the data and why the number of data is reduced: can you clarify? Also, for Fig. 4 why do you use a discrete color scale when your values are not discrete? Maybe you could change that.
5. 3.4.1. I think it would be useful to have a map (even just in the supplement) of the land use data (or urban cover fraction) you use for the analysis to see how they relate to the distribution of NO2 over the study area. Also, it is not clear how you sampled the TROPOMI data according to land use data. How the large(r) TROPOMI pixel size is combined with the high resolution land use data? In the manuscript you write: …averaged over cells of different urban cover fractions.” What you exactly mean by that? Please clarify.
6. L291 “it is worth mentioning that” is redundant to me, you could maybe remove it. Overall, in the manuscript there are some quite long discussions that could be shortened. If you manage to shorten for example some sentences in the results and conclusions, it would help in getting the main message across.
7. Figure 7. The anthropogenic emissions show the largest weekend effect for the lowest bin of urban cover fraction. Can you explain that?

In their paper the auhors present comparisons between gridded NO2 tropospheric column datasets derived from TROPOMI and surface observations, emission estimates and other relevant indicators like land use. Results are presented for the Iberian Peninsula. The approach and methodology is clearly not new. What is new is the degree of detail in the analysis of these comparisons and presentation of the results. I appreciated the detailed investigations of the correlations between the surface and satellite data, contrast between cities and countryside linked to different sources, weekly and seasonal variability. The figures and the text in the paper are of high quality. To my opinion this paper is an interesting information resource for environmental agencies and scientists considering to apply satellite data for local air pollution applications. As such I would be in favour of publication, after the authors have addressed my comments provided below.

Comments:

line 17: "ranging from -40 % in summer to +60 % in winter ". Is this kind of variability as expected?

line 21: "this change started in 2018". But TROPOMI data also started in 2018, so how can a change be deduced from TROPOMI data?

line 33: Surface monitors can be influenced by very nearby sources and therefore they may not represent the average of a larger region. I wonder if the authors see this as a limitation?

line 47: "improvement of spatial resolution of about a factor of 16" In diameter, or area?

line 50: It may be useful to mention the estimate of shipping emissions, which is quite relevant for Iberia, surrounded by major shipping routes.

line 76: I wonder if table 1 is really relevant for the paper. The topic is mapping NO2 over Iberia, not to discuss detailed pixel properties of TROPOMI. The table could be removed and summarised by providing only min-mean-max dimensions in the text. 

line 105: "using a conservative method" Please provide more details about the gridding method.

line 130, GHOST: "More details on the quality assurance filtering are given in Appendix C." Again, adding this table/appendix is a bit beyond the scope of the paper to my taste. Is there a report or publication on GHOST which can be referred to and which contains this information? 

line 136: The choice to look at the d1max, daily mean and overpass value (often close to the lowest values) covers the diurnal range of values, is linked to legislation and therefore I understand this choice. Comparing TROPOMI with a daily mean, however, does not seem very appropriate, given the lifetime, photochemistry and very different meteorology at night. A possible additional choice would be a window (e.g. of 6 hours) around overpass, or focussing on day-time values.

line 191: "snow and ice at the surface". Why not? As long as snow and clouds can be reliably distinguished the retrieval should be straightforward. 

Figure 2: Please mention the averaging period in the caption.

line 233 etc.: These summer-winter sampling differences and impact on the yearly mean may be (partly) overcome by computing monthly-mean values, and then averaging over the months. As long as the sample is big enough such seasonally-varying sampling numbers should not be a real problem.

Fig. 5, caption: It would be helpful for the reader to repeat what d, dop, d1max mean.

line 246: Is it really TROPOMI noise, or actual spatio-temporal variability in NO2? Is it possible to distinguish these two? Same question for lines 254-255.

line 254: i.e. 3-4 months

line 257: d1max seems to have a much lower PCC!

Fig. E4: It is difficult to understand what is plotted here.

line 357: "Interestingly, this mean TrC-NO2 weekend effect progressively decreases when focusing on largest industrial point sources, " I was wondering if this is really an effect of the large industries continued activity over the weekend, or if this is caused by a relatively larger contribution to the column in comparison to the inflow from residential areas.

line 377: "slightly stronger discrepancies over least urbanized areas". Reasons for this could be the relatively larger importance of NO2 higher in the atmosphere, NO2 from other sources and importance of inflow from elsewhere. 

line 391: "results highlight an increase of the weekend effect over the last two decades" This is surprising. Please try to explain: could it be linked to social factors (people working less in the weekend)?

line 411: "discrepancies" This word has a negative meaning. Maybe "difference" is better to use. There may be several good reason why the detailed seasonal dependence may differ at the surface compared to the column even if the observation is perfect.

line 415: "such a broad flat minimum over most urbanized areas was not expected given that road transport in many cities is known to be substantially reduced in August when many people go on holidays," When I look at the curve it seems that the biggest difference occurs in April-May where the relative satellite column is higher than surface observations. June-August looks quite consistent at the surface and from space. A more focussed discussion for April-May would be interesting.

line 432-435: I would challenge the authors to find a more detailed explanation for the shift from August to May. A shift in chemistry (linked to the downward trend in NO2) or anomalous meteorology are factors to look at. Could surface ozone provide additional clues?

Table 4: TROPOMI column data between 2018 and 2021 could be added here.

line 449 etc: Is coverage really an issue to worry about? I do not find a clear answer to this in the paper. Even with 30% coverage there are 10 observations in a month at a given location. The variabilty in concentrations (linked e.g. to meteorological variability) and sampling should be compared with the amplitude of signals to be picked up. In combination with spatial or temporal averaging, as presented in the figures in the paper, it seems that sampling+observation noise is not a problem, and spatial/weekly/seasonal/trend estimates seem all very reliable.

line 464: "lies in its observations over rural areas and seas."

I found the analysis of shipping signals underrepresented / missing in the paper. A separate section on this (figure) would be good.

line 496: "discrepancies" Again, I'm not sure if this is the correct word. Surface concentrations are influenced by different factors than column amounts.

line 502: It would be useful to add some extra references to methods estimating surface concentrations (e.g. Cooper, https://doi.org/10.1088/1748-9326/aba3a5) and emissions from satellite column observations. These approaches + references are not mentioned sufficiently to my opinion.