| 09 Mar 2026
TANGO CO2 and NO2 Observations: Synergistic Usage to Improve Emission Quantification and Characterize Atmospheric Chemistry
Abstract. The Twin Anthropogenic Greenhouse Gas Observers (TANGO) mission, scheduled for launch in 2028, will observe CO₂, CH₄, and NO₂ emission plumes from more than 10,000 industrial facilities per year using two formation-flying CubeSats. Here, NO₂ plume structures exhibit substantially lower random noise than the corresponding CO₂ features, motivating a synergistic exploitation of both species for improved emission quantification and for enhanced characterization of atmospheric chemistry within plumes. Using large-eddy simulations in combination with the Integrated Mass Enhancement (IME) method, we assess NO₂-based masking of CO₂ plumes for emission rates in the range 2.0–12.5 Mt yr⁻¹. This yields CO₂ emission estimates with precisions between 18.5 % and 3.4 %, depending on the emission strength, and corresponding absolute biases that decrease from 15.3 % to 2.4 %. As an alternative approach, we analyze the observed CO₂/NO₂ ratio. By fitting an empirical model to measurement simulations of this ratio and subsequently reconstructing the CO₂ plume from NO₂ observations, we obtain a substantial reduction in the apparent noise of the reconstructed CO₂ plume. For the inferred emission rates, however, the precision remains largely unchanged. Consequently, despite reduced errors in individual pixel-level observations, plume reconstruction does not enhance the precision of CO₂ emission estimates, because it converts originally uncorrelated pixel noise into spatially correlated errors. Neglecting these spatial error correlations leads to a severe underestimation of the retrieval uncertainty. A key advantage of the empirical CO₂/NO₂ ratio model is its ability to characterize plume chemistry. Here CO₂ serves as non-decaying reference tracer. We demonstrate that an effective timescale for the NO → NO₂ conversion in emission plumes can be inferred for sources with CO₂ emissions > 5.0 Mt yr⁻¹. Application of the method to Environmental Mapping and Analysis Program (EnMAP) observations demonstrates its practical utility, confirming its applicability to real satellite data.
The manuscript by Bordsdorff et al. discusses the synergistic use of CO2 and NO2 observations for the upcoming TANGO Scout mission. The manuscript presents novel ideas on fitting the CO2-to-NO2 ratio to the 2D image and is, in general, well written. It uses the CO2/NO2 ratio to analyze plume chemistry. I have one major comment related to the validity of the model. All other comments are technical/minor.
Major comment:
My main concern of the manuscript is related to the model presented in Equation (1). As a reference, the paper by Meier et al. (2024) is given. This is indeed the same equation used in that paper; however, Meier et al. (2024) use this model to analyze the NO2-to-NOx ratio and not the CO2-to-NO2 ratio. In the context of Meier et al. (2024), I think it is clearly justified. The idea is that, after the initial period, NO and NO2 go to a steady state where the ratio is constant (or m0). See Figure 1b in that paper. Figure 1a shows how NO, NO2, and NOx behave. Note that they analyze this model to about 100 km from the emission source.
The manuscript here, however, uses the same model to analyze the CO2-to-NO2 ratio at distances of less than 16 km from the emission source. At first glance, I thought the equation was written the other way around. For example, NOx-to-CO2 would behave in this way, with NOx decaying exponentially and CO2 being a constant tracer.
After seeing Figure 8 by Krol et al. (2024), I sort of understand why this model works near the emission source. However, it certainly cannot work far away from the emission source, as NOx decays (and hence eventually also NO2). That is also clear from Figure 8 in Krol et al. (2024). I would assume the model works until a steady state is reached, but I do not directly see any theoretical justification for it.
Equation (1) may be a useful empirical fit for the simulated near-field plume, where NO is still being converted to NO2, but the manuscript does not provide sufficient physical justification for applying this form more generally to CO2/NO2. In particular, the interpretation of the asymptotic term m0 is unclear: approximate NOx partitioning equilibrium does not by itself imply a constant CO2/NO2 ratio farther downwind, especially as ongoing NOx loss increasingly affects the NO2 plume farther downwind. The authors should better motivate the validity range of Equation (1), clarify that it is an empirical approximation, and discuss where and why it may break down.
All in all, I think the authors should motivate the use of this model, discuss its validity, and potentially provide a theoretical justification. Also, how does the decay of NOx affect the interpretation of the parameters m0, m1, and tau of the selected model?
(As the model used is essentially 1D, many previous studies analyze line densities. This includes the already cited Meier et al. (2024) and Krol et al. (2024). I wonder why the line density approach was not discussed here. I think line densities would help to analyze the results. For example, based on Figure 2b, I am wondering if NO2 is already decaying after 11 km. Plotting CO2 and NO2 line densities could help visualize this.)
Technical/minor comments:
Title: I checked that AMT does not capitalize individual words in their titles. As a general suggestion, I would also recommend writing carbon dioxide and nitrogen dioxide instead of CO2 and NO2.
Abstract: With the same line of thought, CO2, NO2, CH4, and NO should also be defined in the abstract.
Abstract: In general, I believe the method names should be written in lower case. So, please change “Integrated Mass Enhancement (IME)” to “integrated mass enhancement (IME).” Note that in the manuscript, the method is defined quite many times.
Abstract and many other places, especially in the Results: In the English language, there is no space before “%.”
Abstract L3: “Here.” I think this is true in general, so you could maybe write “In general” instead of “Here.” This is also related to the discussion in line 37.
Abstract and in general: According to Table 1, the bias is always lower in the “CO2 with NO2 mask” method. Should it be reported?
Introduction: CO2 and NOx are defined, but please also define NO, NO2, CH4, CO, and O3. You can then also remove the definitions from the beginning of Section 2. In Section 3.1, you might want to define CO and C3H3 (propane?).
Introduction: GOSAT-GW and EnMAP should be defined. Note that you do not mention CO2M in the introduction but suddenly mention the mission in lines 520 and 523. Maybe a short mention could also be useful here?
L35: “representing the first global platform for co-located greenhouse gas and air pollutant monitoring.” I see what you mean here, but you might want to also highlight the resolution, as some could argue that SCIAMACHY was the first one.
L37–45: Would it be more accurate to write “estimate NOx emissions from NO2 observations” instead of “convert NO2 observations to NOx”? Also, I just wanted to note that there are many more studies than are cited here.
L59–60: I find that the approaches “first” and “second” go the other way around here.
L64: According to Table 1, the bias is always lower for the “NO2 plume masking” approach.
L124: I noticed that you have used a special environment for CO2 and NOx/CO2 here. Maybe some \chem should be used in general?
L152: “CO2 and NO2 denote retrieved column densities (mol m−2) after background subtraction.” This comment is slightly more major. I was wondering what the native units of the TANGO retrievals are. I was expecting to see ppm for CO2. Anyways, in my studies, I have noted that it is very important to remove the CO2 background in ppm; otherwise, you might run into problems later. Also, removing the background is a much more important step than is given credit here, and making small mistakes with the background definition can cause large errors in the final emission estimates. This is particularly the case with methods like IME that integrate over a large area.
L152: “The approaches require CO2 and NO2 observations on a common spatial grid.” If you were to use line densities, it would not be so strict.
Section 4.1: See the major comment above.
L161: As the equation ends with “,” I assume that this paragraph should start with “here” and not “Here.”
L162: “Asymptotic background value.” This concept is quite difficult for me to comprehend, as NOx (and eventually also NO2) decays and CO2 does not. I completely understand this concept for the NOx/NO2 ratio.
Equation (6): Please use “argmin” instead of “min.” Also, as your covariance matrix Sy is diagonal, it might be clearer to write the sum-of-squares function as sum{(y_i − F(s_i; x))/sigma_yi} instead of using matrices.
Equation (11): Note that this is essentially plume masking, as can also be seen from Figure 2c,d.
Equation (14): This equation probably should end with “.”
Section 4.5: I found this section quite short. Although this is not maybe the main part of the manuscript, you could connect your work to the masking done by Kuhlmann et al. (2019) and Varon et al. (2018). Both papers are already cited and use a “neighboring” approach.
Section 4.6: This comment is also slightly more major. Just to note that the IME method used by Santaren et al. (2025) is different from the one introduced by Varon et al. (2018), although they have the same name. Varon et al. (2018) do not have a similar distance concept as used here. They use L = sqrt(A) and LES simulations to calibrate the effective wind speed. For the version used by Santaren et al. (2025), and essentially this paper, you might want to check the derivation from Kuhlmann et al. (2024). Anyways, Equation (16), as presented, is not very intuitive, as it already mixes the theory with practical choices like pixel area. It might be more intuitive to write the integral directly and say that, fundamentally, emission = mass/lifetime, and then continue with E = mass*(U/L), and only after that explain how the integral is calculated by summing in practice.
L261–262: The way I read this is that you do not use masking for the “reconstructed” approach. I think this is quite a dangerous approach, especially with real observations, and is prone to systematic errors. As an example, if you make a minimal mistake in the background definition, you sum over the entire area and make a big mistake in the final emission estimate. Why not use both approaches at the same time? (Later, when looking at Figure 8, I suspect that you use Equation (11) to essentially mask the plume, but I am not sure. Anyways, I think it should be written clearly how the plume is selected.)
L296: There is some extra space after “(.”
Figure 2: As your model is essentially 1D, would it make sense to also illustrate line densities?
Figure 2: At first glance, I thought that you had used some plume masking for the panels in the middle, but the “masking” might actually come from Equation (11).
Figure 2: Note that at the plume edges, where your CO2 values are small, the NO2/CO2 ratio can be really large.
Figure 2: I do not understand why you want to create a CO2 reconstruction outside the essential plume area. As panel e is panel b times panel d, should you not also present your model for the full image?
Figure 2: Would it also make sense to analyze line densities here and show the fit in 1D? If your plume were 100 km long, how would you know where to stop?
Table 1: Would it make sense to analyze the “CO2/NO2 recon.” results using the “NO2 mask” approach? Or do you already do some plume masking?
Figure 8: You could define the red line in Figure 8. I assume it comes from Equation (11).
Figure 8: I am wondering if I have understood the integration correctly. Just by looking at this image, I would assume that you only integrate inside the red line.
Figure 8: You could mark the source location with “x.”
Figure 8: How do you make a CO2 reconstruction upwind from the source?
Figure 8: It would be nice to also see the CO2/NO2 ratio and the fit for these cases, as in Figure 2. You could also show the line densities.
L385: I think it would be safe to note that this is not the CO2/NOx ratio. In the MicroHH setup, you assume that 95% of NOx is emitted as NO.
L394–395: “Finally, we discuss the background ratio m0, which represents the CO2/NO2 ratio in the far downwind portion of the plume, where atmospheric processing has progressed to an approximate NOx equilibrium.” I still have some problems understanding this. When NO and NO2 are at steady state, NOx/CO2 decreases. See Figure 8 in Krol et al. (2024).
L510: You did not consider any errors related to wind speed. Errors in wind speed dominate the total error budget and can easily be more than 30% alone. In your IME method, the effect of wind speed is linear, so a 30% error in wind speed means a 30% error in the final emission estimate. You might want to also mention this.