Referee report:
Quantifying NOx emissions in Egypt using TROPOMI observations,
Anthony Rey-Pommier et al.
In response to the reviews the authors have submitted a substantially modified manuscript. I thank the authors for the extended replies to my comments. In particular, I am happy to see that the authors repeated the calculations using the vertical tropospheric column instead of the slant column.
Below I repeat the points from my first review (indicated with a >) and comment to the replies provided by the authors to these points (see the author response document). A couple of minor issues remain (see below), and I ask the authors to address these before the paper is published.
One more major point is on the chosen grid (resolution of 0.1 degree) and the interpolation methods used. However, adapting this first basic step in the calculation would imply repeating all calculations, which implies a very major delay. Furthermore, to my opinion the results of the paper, error estimates and discussion of all aspects involved are valuable and sufficient to be published in the form presented in the revised manuscript. I do not expect that improving the gridding approach would lead to fundamentally different results.
To conclude, for me the paper is ready to be published after my other (more minor) points have been addressed.
> In this interesting paper the authors use the continuity or mass closure equation to derive emissions from the
> TROPOMI NO2 column observations. The authors discuss all the relevant ingredients of the calculation and
> provide estimates of the uncertainty. I am in favour of publishing, but with substantial revisions in response to
> a large number of questions provided below.
> Why Egypt? I understand there are many cloud-free days over the desert. Does the method require entire
> regions to be cloud-free? Or could it be applied to France just as well?
Egypt is motivated by the number of overpasses without cloud cover. This motivation is fine for me.
About the chemistry: I appreciate the reformulated discussion of the different reactions and deposition mechanisms in section 3.1.
> Section 2.1: "We use TROPOMI NO2 retrievals from November 2018 to November 2020". Please provide
> details. Which version (versions) is used?
The version of the product is still not provided. Please add! The TROPOMI data product has changed significantly over time, see for instance v Geffen et al.(https://doi.org/10.5194/amt-15-2037-2022). The TROPOMI measurements are key to the paper, but the dataproduct and quality of the data is discussed only superficially. Instead of the Compernolle [83] reference: why not refer to Verhoelst, https://doi.org/10.5194/amt-14-481-2021?
> Section 2.1: "TROPOMI sounding are gridded for this study at a spatial resolution of 0.1 × 0.1". The
> authors mention that the resolution of TROPOMI is 3.5 x 5.5 km. So the choice of the grid is a bit
> disappointing (11x11 km). Why choose this resolution and not a higher one? Please provide details of how
> the gridding is done. Is this conserving NO2?
The motivation for the grid chosen ("due to technical constraints") is not satisfactory. It should not be an issue to perform the calculations at higher resolution. Also the averaging procedure is not very satisfactory. Partial overlaps between TROPOMI and the grid cell are not accounted for it seems. The authors claim "The observed plumes remain correctly resolved" but there are clearly plumes to be expected with smaller scales than 0.1 degree. It seems to me the authors are not fully using the information which is provided by the TROPOMI instrument.
One minor advantage of the chosen grid resolution of 0.1 degree is the match in resolution with the emission inventories.
As mentioned above, I do not think that changing (improving) the gridding approach would lead to fundamentally different results so I suggest to move ahead with the results as presented in the revised manuscript.
> Section 2.3: "Therefore, the CAMS OH concentrations are used". The resolution of CAMS is not very high,
> 0.4 degree. Given non-linearities and dependency on NOx, would the use of CAMS OH be a good choice?
> What are typical uncertainties, in particular those linked to the downscaling from 0.4 degree to 0.1 degree?
There are comments on the relatively small variability in OH which "means that CAMS concentrations are able to produce a realistic concentration gradient" I do not understand this argument. I would reason that the resolution is low which implies the model will not be able to resolve gradients in chemical concentrations (e.g. across the Nile delta), and therefore the range in OH may be largely underestimated. Also, it seems that Fig.5 suggest quite a wide range of concentrations/lifetimes.
(Note: One reason why I mentioned France as target area is because of the availability of the CAMS regional air quality forecasts at 0.1 degree, the same resolution as the grid chosen.)
> Section 2.4: "It is therefore necessary to remove the natural part of the atmospheric signal " We do not
> expect a lot of lightning and soil emissions over the desert. How large a signal is expected, why is removal
> needed, and how is this done?
I appreciate the new discussion on mechanisms responsible for background NOx observed in the TROPOMI data.
The authors write "Removing this natural signal is necessary for two reasons. Firstly, the natural part of the TROPOMI signal has to be removed from the emissions in order to interpret the results in terms of human activities. Secondly, the model uses quantities (wind, [OH], NOx to NO2 ratio, etc.) calculated in the lower troposphere. Applying it to a signal that is not entirely located in the troposphere would not make sense. "
It could be useful to the reader to add these sentences to the paper.
> Section 2.4: "We conduct this removal by subtracting the mean emissions over desert and rural areas from
> the mean emissions over urban and industrial areas. " Should "emissions" be "NO2 tropospheric column
> concentrations" here? Later in the paper there is a background emission term introduced. Why are
> background corrections not applied to the concentrations?
Thanks for the answer and the revised discussion in the paper, which I find satisfactory.
> Section 2.5: The CAMS emissions also seem to rely on EDGAR and will use similar approaches/assumptions
> and input datasets. Please comment on how independent or dependent these two datasets are.
Thanks for the extra clarification on the differences.
> Section 3.1, line 184: "Slant column densities are used as vertical densities" This does not make any sence to
> me, and should be a large and unnecessary source of uncertainty. The simplest approach to the air-mass factor
> would be a geometric path length of the incoming and outgoing light which depends on the viewing angles and is > 2.0.
> So, neglecting the air-mass factor can easily lead to 50% errors. Why is this better than using the air-mass
> factors from the retrieval?? Furthermore, the slant column will include (be dominated by) the stratosphere.
> Why not use the tropospheric column? As mentioned, the sink is modelled as concentration divided by lifetime.
> But this concentration should be the column in the lower troposphere only, otherwise it does not make sense?!
I am happy to learn that the AMF is now used (e.g. the tropospheric columns are used).
> Equation 3: What is the omega_NO2 in this formula. Is it the slant column from TROPOMI?
OK
> Section 3.2. The discussion focusses entirely on electricity consumption, motivaing that 13:30 is representative
> for the daily mean. However, I would expect that traffic (industry) is also a major source of NOx, and this has a
> distinct diurnal (seasonal) pattern. So the discussion seems to be over-simplified.
Thanks for the changes in the text in section 3.2, which is now more balanced.
> Line 258: The city of Riyadh has been extensively discussed by Beirle et al., 2019.
> A reference to this paper in section 3.3 should be added.
Thanks for adding this to the new manuscript.
> Line 263: sqrt(w^2) = w. The notation is a bit unclear.
OK
> Equation 7: I still have a conceptual difficulty with a "rural emission". Over the desert the estimated emission
> should be close to =0 and negligible compared to urban emissions, otherwise the methodology is flawed.
Thanks for modifying the text and equation. I am satisfied with the response.
> line 324: "limit the high inter-day variability due to changing wind patterns or differences between week days
> and week-ends". What is the real reason averaging over a month is needed? Winds change, but if the method
> is correct the emissions should be equal (assuming stationary sources).
I still think that the sentence, which was not modified, is misleading. In their response the authors focus on uncertainties, which makes more sense. I would suggest to rephrase the text of the paper accordingly.
> l 359: "Level B is therefore the one that leads to the best match between the lifetime calculated with Equation
> (2) and the lifetime calculated from line densities." What does this really prove? Does it really mean Level B is
> better? Due to the coarse resolution we may expect CAMS is biased in OH since it does not resolve the
> plumes.
The abstract mentions "It it also provides the location of the most appropriate vertical level to represent typical pollution sources in industrial areas and megacities in the Middle East region." I'm not convinced yet about this conclusion and think it is a rather bold statement. The correlations between CAMS and EMG are low for both A and B, and the slope in both cases is close to 1. It is valuable to see the comparison with the EMG method, but I wonder if the comparison alone can be used as evidence for the selection of an appropriate level.
> Figure 6: Before showing this, I would suggest the authors apply the method to Riyadh and compare with
> Beirle et al. (2019) to test the consistency of the results.
In their response the authors give numbers for the comparisons which are interesting and relevant, because they provide an alternative insight in uncertainties in the method. Beirle is mentioned in Sec 3.3, but the comparisons in terms of total emissions, lifetime, relative importance of the sink term vs transport term is still not provided. I suggest the authors copy some of the numbers from their response to the paper to highlight these substantial differences in the estimates.
> Table 1: I would suggest to replace "khab/km^2" by "10^3/km^2"
I am satisfied with the response.
> l 420: "It is also observed that TROPOMI NO2 column densities above this zone are relatively homogeneous"
> As demonstrated in several papers, there is a clear shipping signal in the TROPOMI data over oceans and
> seas, and I would expect TROPOMI to be rather inhomogeneous here?!
I am satisfied with the response.
> Figure 8: The unit is "kt" which I assume is 10^6 kg. But what is the time unit? Per hour, per day, per year? I'm
> a bit surprised by the big scatter for the weekly (daily) values averaged over the entire country?
I am satisfied with the response (unit is now kt/d). Maybe good to mention once in the text that "kt/d" means kilotons per day, to avoid confusion.
> Section 4.5, Covid-19. There is a nice review paper, https://doi.org/10.1525/elementa.2021.00176, which could
> be added here.
Thanks for adding this review paper
> l 488: "no significant changes in OH concentrations ". Does the CAMS system describe the change in
> emissions and concentrations observed resulting from the lockdown? If not, how would this impact the results
> (given the non-linearity of the chemistry)?
I think it would be useful if the response of the authors (no changes in emissions during lockdowns, but possibly some impact through the assimilation of satellite data) is also added to the text of sec. 4.5.
> l521: "TROPOMI-inferred emissions show an annual variability" I was wondering how much we can believe
> the seasonality in OH as modelled by CAMS? This seems to directly link to the seasonality of the sink term
> and, as a consequence, the emission estimate. Please discuss.
I am satisfied with the response.
> l551: "S-5P validation activities" Please add a reference
As above: I suggest Verhoelst et al.
> l 558: "For [OH]," The authors showed that OH is strongly height dependent, so it seems that the choice of the
> vertivcal level is a major uncertainty. Has this been accounted for?
I am satisfied with the response.
> Data availability: TROPOMI data is missing here.
I am satisfied with the response. |