Representativeness of TROPOMI tropospheric NO<sub>2</sub> columns near Asian airports constrained by IAGOS airborne measurements

Lim, Hyunkwang; Fujinawa, Tamaki; Sugita, Takafumi; Inomata, Satoshi; Tanimoto, Hiroshi; Mahnke, Christoph; Petzold, Andreas; Bundke, Ulrich; Kanaya, Yugo; Kasai, Yasuko; Sato, Tomohiro O.; Park, Sang Seo; Kim, Minseok; Lee, Yun Gon; Jeong, Ukkyo

doi:10.5194/egusphere-2026-4968

Preprints

https://doi.org/10.5194/egusphere-2026-4968

Preprints

11 Sep 2026

| 11 Sep 2026

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

Representativeness of TROPOMI tropospheric NO₂ columns near Asian airports constrained by IAGOS airborne measurements

Hyunkwang Lim, Tamaki Fujinawa, Takafumi Sugita, Satoshi Inomata, Hiroshi Tanimoto, Christoph Mahnke, Andreas Petzold, Ulrich Bundke, Yugo Kanaya, Yasuko Kasai, Tomohiro O. Sato, Sang Seo Park, Minseok Kim, Yun Gon Lee, and Ukkyo Jeong

Abstract. Tropospheric NO₂ data retrievals from the satellite-borne TROPOspheric Monitoring Instrument (TROPOMI) are sensitive to the vertical profile assumed in the air mass factor (AMF) calculation, especially in polluted lower-tropospheric environments. This study investigated the use of aircraft and ground-based profile information to evaluate this sensitivity near five major airports. We combined TROPOMI tropospheric NO₂ vertical column densities (VCDs), land-based Pandora observations, and airplane-based In-service Aircraft for a Global Observing System (IAGOS) descent profiles for selected airport–Pandora pairs in Japan, South Korea, and Thailand from March to July 2025. Before applying profile-based AMF recalculations, we examined whether the TROPOMI NO₂ field around each Pandora site was representative of that around the corresponding airport. The airport- and Pandora-region time series were generally coherent, although their mean NO₂ levels differed, indicating that temporal consistency and absolute spatial representativeness need to be considered separately.

The lower part of the a priori NO₂ profile obtained by the TM5-MP model was then replaced with IAGOS-derived profiles to recalculate TROPOMI AMFs and VCDs. The IAGOS-profile recalculation reduced the overall low bias relative to Pandora data and improved several validation metrics, but the response differed by site. Improvements were most evident where the IAGOS descent profiles, Pandora site, and satellite-observed NO₂ field sampled similar environments. In contrast, weaker or degraded performance occurred when the aircraft profiles represented an air mass different from that influencing the Pandora-acquired column. The Pandora-scaled-profile case produced the closest agreement with Pandora results, but the comparison is more suitable as a profile-sensitivity test rather than an independent validation. Residual diagnosis showed that the remaining differences after the Pandora-scaled-profile recalculation were not mainly controlled by horizontal NO₂ heterogeneity, pixel sampling density, mean quality assurance value, or the magnitude of the AMF correction. A slant column density (SCD)-equivalent closure analysis further suggested that profile replacement alone cannot fully reconcile TROPOMI and Pandora columns, because the recalculated VCD still depends on the TROPOMI effective tropospheric SCD and other non-profile retrieval inputs. These findings demonstrate the value of IAGOS and Pandora lower-tropospheric profile information for diagnosing TROPOMI NO₂ AMF uncertainty near airports, while underscoring the need to account for slant-column retrieval uncertainty, cloud and surface effects, and spatial representativeness.

Received: 18 Aug 2026 – Discussion started: 11 Sep 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

Status: open (until 24 Nov 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on egusphere-2026-4968', Anonymous Referee #1, 12 Oct 2026 reply
Reviewer’s comment on
“Representativeness of TROPOMI tropospheric NO₂ columns near Asian airports constrained by IAGOS airborne measurements”
This manuscript employs NO₂ vertical profiles acquired from IAGOS in-situ and Pandora sky-scan measurements as a priori profiles for TROPOMI NO₂ retrievals, and examines how the air-mass factor (AMF) changes when using observation-adjusted vertical profiles instead of the operational TM5-MP model input. These AMFs are then used to recalculate TROPOMI NO₂ tropospheric column densities, which in turn have been evaluated against the reference Pandora tropospheric column measurements. It has been a long-standing question how well the model a priori inputs represent the real-world NO₂ vertical profiles, and the associated uncertainties in satellite NO₂ column retrievals are not fully understood and deserve attention. Both IAGOS and Pandora measurements of the lower-tropospheric NO₂ vertical profiles are unique but, more importantly, critical observational constraints that can be used to facilitate our understanding of current uncertainties in satellite retrievals and to advance the retrieval accuracy. However, the results presented in this study do not persuasively demonstrate the value and significance of IAGOS and Pandora profiles. This study should better highlight the value of IAGOS (and Pandora) profiles in terms of how they can benefit satellite retrievals, while acknowledging that NO₂ columns are extremely heterogeneous in the urban atmosphere and that the vertical profile can be even more variable. I encourage the authors to consider additional comprehensive analyses that explain the factors that can limit the utility of these measurements and show how these observations can be properly incorporated to improve satellite retrievals. Altogether, I believe that this study aligns with the scope of this journal and addresses the scientific question to which the community is seeking an answer, and is worth being published after proper revision.

Major comments:
Abstract should be revised for better clarity and coherence. Current abstract is not easy to understand because they do not follow a logical flow. It reads better when revisited after one has read the whole manuscript, but the abstract should be comprehensible by itself without the context from the main manuscript.

Introduction is quite verbose. I suggest streamlining it by narrowing the scope to subjects that are directly relevant to this study. Specific suggestions are listed in the “Minor and technical comments” section.

I strongly suggest restructuring Sects. 4.1 and 4.2:
Why did the authors use the agreement between TROPOMI column NO₂ near the Pandora station and near the corresponding airport as a benchmark for how well the IAGOS profiles represent what Pandora measured? It would be better to compare TROPOMI NO₂ averaged along the line of sight of the Pandora sky scans and along the IAGOS flight tracks. The IAGOS flight tracks shown in Figure 1 do not appear to remain within the vicinity of the airport, not even within a 20 km radius. This can change the results shown in Sect. 4.1.

It sounds a little weird to me when the day-to-day variability of TROPOMI NO₂ is referred to as “temporal variability”. I’d suggest avoiding the term “temporal variability” (throughout the manuscript, but particularly in Sects. 3.1 and 4.1), since what the analysis really needs to emphasize is how closely the two regions (near the airport and near the Pandora station) coincide with each other (e.g., “similar” or “coherent”). What readers would expect from “agreement of temporal variability”, especially for NO₂, is a high covariance between two different, temporally well-resolved (e.g., semi-continuous) time series such as Pandora direct-sun observations, based on a comprehensive analysis (e.g., incorporating meteorological conditions).

The day-to-day covariance of areal-averaged TROPOMI column NO₂ (e.g., Fig. 2; Sect. 4.1) can provide partial context on how the two regions are correlated with each other. However, I’m not quite convinced that the day-to-day covariance between normalized areal-averaged TROPOMI column NO₂ over the two regions can provide meaningful context on how vertical profiles would behave. There are too many factors that can change the vertical profile shape, even if the column densities agree with each other. Moreover, absolute column densities have a direct association with vertical profile shape; hence, the absolute differences between the column densities can significantly alter the vertical profile shape.

I would suggest revising Sects. 4.1 and 4.2 by adding a comprehensive analysis that physically explains the differences between airport-Pandora pairs, focusing on how the results can support interpreting the difference between Pandora tropospheric columns and IAGOS- and Pandora-scaled TROPOMI columns (e.g., Sect. 4.3~). In particular, these two sections should clearly exhibit how IAGOS vertical profiles agree with or differ from the vertical profiles near the Pandora station, and explain when and why. For example, the authors could compare Pandora and IAGOS vertical profiles with TROPOMI-areal averages while considering the local wind field. Such an analysis will help readers to understand how the environments generally differ within each pair and how and when they are correlated, and it will address the concern about “how two stations more than 20~30 km apart in an urban environment can be meaningfully compared”. This will add invaluable context to later discussions on evaluating the accuracy of IAGOS- and Pandora-scaled TROPOMI products (e.g., Sect. 4.3~), which in turn adds significant value to this study.

Section 4.3 is rather trivial, since an areal-averaged (e.g., pixels within 10 km) AMF will blur out the potentially-varying scattering weights, which in turn will generally follow what vertical profiles would force. Moreover, there is no detailed analysis or interpretation of how the AMF behaves differently for each airport-Pandora pair. I think this should be discussed as a supporting detail in the interpretation of the differences against Pandora (Sect. 4.4).

The discussions in Sect. 4.4 would benefit from a more detailed analysis. For example, instead of just trying to explain how the IAGOS- and Pandora-scaled TROPOMI evaluation against Pandora differed among the airport-Pandora pairs, the authors should provide comprehensive supporting details. I would suggest constraining the environmental conditions so that the IAGOS profiles can represent what the Pandora sky scans measure, and then evaluating how the TROPOMI column retrieval can be improved by employing IAGOS profiles under ideal conditions.

Section 5 lacks detail, especially on how the spatial heterogeneity, sampling/quality, and AMF correction tests were conducted. There should be a clear and interpretable description of how these tests were designed, and the results should be presented explicitly (i.e., Table 3 and Figure 7 are insufficient).

Minor and technical comments:
General comments:
I suggest that the authors review the manuscript to improve paragraph coherence and overall English usage. Specific comments are listed below, but I encourage the authors to refine the manuscript beyond what is explicitly stated. Language editing by a native English speaker would be helpful in improving the readability of the manuscript.

Specific comments:
Abstract
Line 31: The word “representative” can be misleading. Maybe “correlate with” is what you intended?
Lines 32-33: “The airport- and Pandora-region time series” – do these refer to areal-averaged TROPOMI NO₂? It’s hard to understand the following phrase in line 34 “indicating that temporal …”, and what this sentence is trying to emphasize. I wonder if it is necessary to state it in the abstract.
Lines 36-37: Unable to understand this sentence. Please clarify.
Lines 39-40: “Pandora-scaled-profile case” – does it mean TROPOMI NO2 recalculated with Pandora profiles as an input? It might be worth stating the definition once before using this term in the abstract. What do you mean by “closest agreement with Pandora results”? Compared to what? What is meant by “comparison is more suitable as a profile-sensitivity test rather than an independent validation”? Why?
Lines 40-41: Then, what is the main contributing factor to the residual differences?
Line 47: This contradicts the previous sentence, since you’ve stated that the spatial representativeness is not the dominant factor (line 41: “not mainly controlled by horizontal NO₂ heterogeneity, pixel sampling density”).

Introduction
Line 66: May need to add references supporting the sentence.
Lines 68-69: It would be nice to have acronyms that are commonly used names(e.g., GEMS, TEMPO, UVNS). Also, please place the corresponding reference next to each instrument.
Lines 74-75: The AMF rather represents the effective path length of the slant column measurements by definition, and is affected by “the sensitivity of the satellite measurement to …”.
Paragraphs between lines 82 and 115: I think these three paragraphs are not really relevant to this study. Since the Introduction is already very long and verbose, I’d suggest removing these.
Line 117: Remove “finite”. May use “over pixel footprints” instead?
Line 119: Remove “scale” since it adds unnecessary complication.
Line 122: Pandora makes both direct-sun and MAX-DOAS (sky-scan) measurements, and both types of measurement data have been used for satellite evaluation. I suggest stating the Pandonia Global Network (PGN) here since NDACC was also stated.
Line 125: Please add more recent studies as references.
Lines 125-127: This sentence adds verbosity to the Introduction. I don’t think you need to specifically introduce every detail of each study throughout the Introduction section.
Lines 127-128: Likewise, to be more concise – “TROPOMI generally captures the temporal variability of tropospheric NO₂, but it” can be replaced with “satellite NO₂ retrievals”.
Lines 138-139: How can the comparison of column amounts be (or can be considered as) the limitation in satellite retrievals? Please revisit the causality.
Lines 142-144: Please add reference(s).
Line 145: “from an idealized point column” is unclear. Was it your intention to say, “the vertical column directly above the instrument”?
Line 146: “sensitive” -> “representative”?
Paragraphs between lines 131 and 163: I feel that the subject is jumping around in these three paragraphs. Why don’t you change the order of the first and second paragraphs?
Line 156: The AMF is not more “sensitive” to the lower-tropospheric NO₂ distribution than to that at other altitudes. AMFs are strongly affected by the lower-tropospheric NO₂ distribution because it is highly variable (the magnitude of its variability is much higher than in the FT).
Lines 156-159: I think the airborne imaging spectrometer measurements are irrelevant to what either this paragraph or the manuscript is trying to address.
Line 169: “position” -> “profile shape”
Lines 172-173: Why would IAGOS profiles provide an observational basis for testing …, particularly in the presence of Pandora observations? How can Pandora observations benefit? Please add a narrative supporting this statement.
Line 174: Does the study utilize only the Pandora MAX-DOAS (sky-scan) mode observations? Or does it also use direct-sun observations?
Line 177: In association with one of my major comments – I’d suggest refraining from using “temporal variability”.
Line 179: : “results in airport environments” -> “results near the airports”
Line 188: “airport-Pandora representativeness” – this phrase is understandable, but rather confusing. Can there be any alternatives?
Line 190: What are “Pandora-scaled residuals”?

Data and study domain
Figure 1: Please add the TM5-MP model grids to this figure – it is important to demonstrate that all the relevant data analyzed in this study fall within the same a priori model grid. Also add the viewing azimuth of the Pandora sky scans.
Line 215: What is meant by “airport-scale”?
Line 237: viewing directions -> viewing elevations
Lines 239-241: Do you mean that the Pandora sky-scan products (i.e., vertical profiles for the Pandora-scaled TROPOMI NO₂ retrieval and the tropospheric columns used to evaluate satellite NO₂ columns) are used in this study? Is there a specific reason why the tropospheric columns from the sky scans are used to evaluate the satellite columns instead of the direct-sun total columns?
Sect. 2.4: Since Pandora sky-scan data (profiles, tropospheric columns) are used as one of the key elements in this study, I’d suggest adding a more detailed description of the characteristics of Pandora sky-scan observations (e.g., reliability, known strengths and weaknesses). In particular, Pandora profiles are sensitive to the lowest-available (unobstructed) elevation angle for the sky scan, viewing azimuth, and station altitude. I suggest adding a description of how the Pandora sky-scan data have been screened, so that only the quality-assured data are used in this study.

Method
Line 263: How was the 10 km radius determined as an averaging range?
Lines 273-275: Did you apply a time correction based on the Pandora time series for the Pandora-airport pairs with relatively low correlation (e.g., KIX, NRT)?
Lines 270-271: Please revisit this statement in accordance with my major comment (e.g., “Pandora observations represent temporal variations …”).
Lines 275-277: This might need to be a little more descriptive. How can you scale the IAGOS profiles based on the diurnal variation of the Pandora tropospheric columns? This is straightforward only under the premise that all of the NO₂ is below 3 km altitude (i.e., how did you account for NO₂ above 3 km altitude while scaling?). A mathematical description might help.
Lines 284-285: Generally speaking, it is not favorable to set negative values to zero, since it can positively bias the measurement.
Lines 287-288: How did you treat the profile below the lowest available altitude? Extrapolated to the surface?
Lines 326-329: Are these two sentences duplicative?
Lines 329-336: These sentences are unnecessary. Instead, explicitly state what the GCOS is and what the attainment goal is. In particular, please use a plain-sentence to describe the set goal threshold (e.g., “smaller than 20 % of the Pandora NO₂ VCD or 1e15 molec. cm^-2, whichever is greater”).

Results
Figure 2: Please add the airport-Pandora pair names to the figure (not only in the caption). Panel (f) should state that it displays areal-averaged TROPOMI NO₂ in the scatterplot, either in the figure or in the caption.
Lines 363-364: “spatial representatives of the Pandora-site region” representing what? Please clarify. This sentence should be better elaborated, since it does not describe the subject clearly. E.g., what is the “subsequent Pandora-based evaluation”? Is it the interpretation of IAGOS-scaled TROPOMI NO₂ against Pandora?
Lines 379-380: Can you please clarify what this sentence is trying to state? How can the satellite NO₂ field at one point provide information for “interpreting temporal changes” in the satellite NO₂ field at the other point? How is it relevant to the discussions in later sections?
Figure 4: Please state the airport-Pandora pair names in the figure, not only in the caption. It would be helpful to include the distance between the airport and the Pandora station in parentheses. I suggest adding an additional row of figures showing the differences (or the differences between vertically normalized profiles (i.e., profile shape)) between the profiles.
Line 390: Section 3 already states that the vertical profiles are reconstructed with IAGOS and Pandora only up to ~ 3 km altitude.
Line 410: There is no Sect. 4.5 in this manuscript.
Line 433: Try to be more explicit. E.g., “a larger fraction of NO₂ in the lower troposphere”.
Figure 6: How were the TROPOMI and Pandora data co-located? Do “A” and “B” have a specific meaning in the %GCOS metric shown in the figure? If “A” refers to near-airport TROPOMI pixels (operational for (a), IAGOS-scaled for (b), and Pandora-scaled for (c)), and “B” refers to near-Pandora TROPOMI pixels, please show the two groups in separate scatter plots and add a relevant discussion explaining the differences in terms of how spatial disparity influences the results.
Figures 5-7: Please use consistent colors for each airport (-Pandora pair) in the figures.
Lines 457-459: The tropospheric column NO₂ retrieval algorithm for Pandora sky-scan observations utilizes only the 60° and 75° zenith-angle scans, and the vertical profiles get most of the information from the lower-elevation scans. Moreover, what satellite retrievals take is a vertical profile shape and is not necessarily tied to the absolute column density. Therefore, the better agreement of Pandora-scaled TROPOMI with the Pandora tropospheric columns should rather be viewed as a result of tighter spatial co-location.

Discussion
Figure 7: What do you mean by TROPOMI “effective” SCD (y-axis)? Why did you describe it as “effective”?
Lines 514-515: This sentence is puzzling. How are these metrics derived?
Lines 534-535: This statement can only be true if a single TROPOMI pixel has been targeted and the vertical profile has been acquired at that pixel with commensurate accuracy and spatial representativeness (i.e., the NO₂ field should be homogeneous). This study uses a TROPOMI pixel average over a 10 km radius, which is not closely co-located with the Pandora profile measurement, and the NO₂ field cannot be regarded as homogeneous over that much of an area in the study region.

Reply
Citation: https://doi.org/10.5194/egusphere-2026-4968-RC1

Viewed

Total article views: 449 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
262	125	62	449	59	52

HTML: 262
PDF: 125
XML: 62
Total: 449
BibTeX: 59
EndNote: 52

Views and downloads (calculated since 11 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	212	98	32	342
Oct 2026	50	27	30	107

Cumulative views and downloads (calculated since 11 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	212	98	32	342
Oct 2026	50	27	30	107

Viewed (geographical distribution)

Total article views: 449 (including HTML, PDF, and XML) Thereof 449 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 22 Oct 2026

Short summary

Satellite measurements of nitrogen dioxide can be biased when the assumed vertical distribution of pollution does not match reality. We combined satellite, aircraft, and ground-based observations near major Asian airports and found that using aircraft-measured profiles generally reduced this bias. However, improvements varied by location, showing that spatial representativeness and other retrieval uncertainties also remain important.


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

Representativeness of TROPOMI tropospheric NO2 columns near Asian airports constrained by IAGOS airborne measurements

Viewed

Viewed (geographical distribution)

Representativeness of TROPOMI tropospheric NO₂ columns near Asian airports constrained by IAGOS airborne measurements