Global variability in the detectability of power plant NO<sub>2</sub> plumes from space

Huang, Ruizhe; Wang, Sherrie

doi:10.5194/egusphere-2025-6008

Preprints

https://doi.org/10.5194/egusphere-2025-6008

Preprints

06 Jan 2026

| 06 Jan 2026

Status: this preprint is open for discussion and under review for Atmospheric Measurement Techniques (AMT).

Global variability in the detectability of power plant NO₂ plumes from space

Ruizhe Huang and Sherrie Wang

Abstract. We present the first global, data-driven analysis of power plant NO₂ plume visibility from space. Using TROPOMI observations over 6,000 of the world’s highest-emitting power plants and hourly CEMS data for 500 U.S. plants, we develop an automated algorithm that labels plumes and attributes them to their sources with 98 % accuracy. We then train a machine learning model to predict plume detectability from environmental, meteorological, and observational variables (F1 score > 0.65, AUC > 0.8). Out of 25 variables, we find that NO_x emission rate, surface albedo, wind speed, and sensor zenith angle jointly explain much of the detection variability. An hourly NO_x emission rate of ≈ 400 kg/h corresponds to a 50 % detection probability on average, but detection rates vary from < 20 % to > 60 % under different combinations of these conditions. These results provide the first empirical quantification of the physical and environmental factors that govern NO₂ plume visibility in satellite data, establishing a foundation for models to use similar predictors as auxiliary variables when quantifying emission rates from plume appearance.

Received: 02 Dec 2025 – Discussion started: 06 Jan 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Ruizhe Huang and Sherrie Wang

Status: open (extended)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2025-6008', Anonymous Referee #1, 08 Mar 2026 reply
Summary
The paper "Global variability in the detectability of power plant NO2 plumes from space" by Huang and Wang presents a plume detection algorithm and trains a NN to check for plume "detectability", i.e. whether a plume is visible from satellite (TROPOMI) measurements or not.

With this NN, the most important input features driving "detectability" can be identified.

This is an interesting approach and helps to understand which conditions need to be fulfilled for successful plume detection and emission estimation from satellite measurements.
Overall, the study is written well, except that proper references and acknowledgements are lacking.

The method is comprehensible, but one major drawback is the usage of 10m winds, which are inappropriate even directly at the power plant due to stack height. Finally, while "detectability" is interesting, the overall goal is "quantifiability", and the study does not provide information on this.
I recommend publication on AMT after dealing appropriately with the comments below, which require major revisions.
General remarks
The study severely lacks proper references to literature and datasets. Several concrete examples are provided below. But the list below is not complete. So make sure that every dataset used in this study is properly referenced with key publications as well as working links to the datasets.

The overall procedure for investigating plume detectability is comprehensible and explained well. Selected features mostly make sense, with one important exception: the 10m winds are not representing horizontal transport within the boundary layer.

I see that modifying the input wind fields implies a complete re-analysis of this study. But it would be the way to go to get the best results of the presented methodology.

Plume detectability as defined in this study could just be put in words as "there is an enhanced TROPOMI pixel somewhere downwind from a known power plant". I see the motivation for this approach, and I find the study of driving features for this "detectability" legitimate and worth to be published on AMT. However, the authors should avoid raising wrong expectations. They should clearly state (when defining the plume detection algorithm) that this "detectability" does not require a clear plume over several TROPOMI pixels which would be necessary for some of the emission estimation methods. They should also state (e.g. in 5.2) that this plume detectability does not provide direct information about how far and how meaningful emission estimates could be derived from these satellite measurements.

Additional comments
Line 13: A reference to "M et al." is meaningless. I expect the authors refer to the paper by M. Crippa et al.

Line 31: I expect that the most important difference between the four examples in Fig. 1 is the surface albedo, which is not mentioned here.

Line 50: The "NO2 Window" has not been explained yet; should be skipped, or explained.

Lines 52-55: Repetition.

Fig. 1:
How can you have exactly 400.0 kg/h for four different sources? Is this a kind of "default" value in the database? Please check the distribution of values within your emission database.

The legend is hard to read. In particular, "0" looks like "8" if not zoomed in. Please choose a different font without the dot inside the "0".

The data source ("Satellite Image Source: Esri, i-cubed, USDA, USGS, AEX, GeoEye, Getmapping, Aerogrid, IGN, IGP, UPR-EGP, and the GIS User Community") is not helpful. Please clarify in the manuscript how and from where the satellite images have been derived, and give proper reference.

Section 2:

This sections starts with an explanation on what has been done with the power plant data, without stating where the data is coming from first (is it the EPA data introduced later in 2.3?). This section needs to be restructured such that the used input data is introduced, shortly described, and appropriately referenced and acknowledged first.

TROPOMI NO2: references to TROPOMI (Veefkind) and the NO2 product (e.g. Geffen) are missing.

Line 80: This only holds for nadir geometry!

Line 81: Note (and mention) that even if you downloaded the TROPOMI data from the NASA server, it is an ESA instrument, which also needs to be acknowledged.

Line 81: Which processor version has been used?

Line 87: provide reference to qa value and 0.75 threshold.

Line 144: Please specify if this equals the "viewing zenith angle" provided in the TROPOMI L2 data. Note that viewing zenith angle also determines the size of the ground pixel, which will likely have strong impact on plume detectability!

Line 145: How strong does the sensor altitude vary? I would expect a very low variation here that probably has no significant impact on the observation quality. Rather, I suspect that this codes in fact a latitude dependency for the NN.

Line 157: Why have 10m winds been chosen? I suspect from convenience (as this is provided in the TROPOMI L2), but ERA5 data is used as well anyhow. 10m winds are already inappropriate at the source due to the high power plant stacks and the plumes' buoyancy. For describing transport to the next TROPOMI pixel ~5 km away, typical boundary layer winds are relevant, not those at the ground! I see this as a severe weakness of the study and I would expect better results for appropriate wind fields.

Line 161: Solar zenith angle also affects actinic flux which is key in photochemistry, affecting photolysis of NO2 directly. In addition, photochemistry affects levels of O3 (needed for conversion of NO to NO2) and OH (sink of NO2).

Line 164: TOA is strongly related to solar zenith angle and basically codes seasonality.

Line 194: Given the number of features, I consider a sample size of 100 to be quite low for the hyper-parameter tuning.

Section 3.1.2: How many of the initial 500/6000 power plants are left after filtering for interferences?

Fig. 7: Since the total emissions are an obvious and direct cause for plume detectability, I would suggest to swap x and y axes in this plot.

Line 436: It is the first time for me to read about the sensor zenith angle - in the TROPOMI data product, there is a viewing zenith angle provided, typically abbreviated as VZA. The authors might have reasons to use the term sensor zenith angle instead. But in context of TROPOMI, the sensor zenith angle alias viewing zenith angle must not be abbreviated as SZA, as this is a standard abbreviation for the solar zenith angle.

Line 437: The rise of the AMF is not really an argument here, as this is corrected for in the retrieval. In addition, pixel size increases with VZA, which should cause a decreasing detectability with VZA. Please extend the discussion here, which might be not fully conclusive; I have no good explanation for the initial increase, instead I would have expected an overall decrease with VZA.

Line 471: But the cited studies provide emission estimates, not just binary information about plume detectability.

Line 481: See Line 437.

Line 515: Vertical wind shear could be extracted from ERA5.

Fit. E1: This figure is hard to read due to the colormap (light read could be ~0 or ~1). Please use a more appropriate colormap here.

Table E1: From all listed features, I am surprised by the low number for cloud fraction. By filtering for qa>0.75, large cloud fractions are removed. But the "clear" pixels are often still affected by clouds, and I would expect that detectability really depends on cloud fraction. It would thus be quite interesting to see a figure similar to Fig. 9 for cloud fraction. Please extend the discussion accordingly.

Reply
Citation: https://doi.org/10.5194/egusphere-2025-6008-RC1

Ruizhe Huang and Sherrie Wang

Viewed

Total article views: 498 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
265	214	19	498	14	18

HTML: 265
PDF: 214
XML: 19
Total: 498
BibTeX: 14
EndNote: 18

Views and downloads (calculated since 06 Jan 2026)

Month	HTML	PDF	XML	Total
Jan 2026	181	113	12	306
Feb 2026	25	32	2	59
Mar 2026	59	69	5	133

Cumulative views and downloads (calculated since 06 Jan 2026)

Month	HTML	PDF	XML	Total
Jan 2026	181	113	12	306
Feb 2026	25	32	2	59
Mar 2026	59	69	5	133

Viewed (geographical distribution)

Total article views: 489 (including HTML, PDF, and XML) Thereof 489 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 31 Mar 2026

Short summary

We analyzed satellite data from 6,000 power plants to determine when nitrogen dioxide plumes are visible from space. Using artificial intelligence, we found that detectability depends heavily on wind, ground brightness, and viewing angle—not just the amount of gas emitted. Consequently, identical emissions can be seen or missed depending on local conditions. This work helps scientists improve global pollution estimates by accounting for the environmental factors that hide plumes.


Total:	0
HTML:	0
PDF:	0
XML:	0

Global variability in the detectability of power plant NO2 plumes from space

Viewed

Viewed (geographical distribution)

Global variability in the detectability of power plant NO₂ plumes from space