4-D Aviation Emission Inventory Data of China Estimated with QAR Data in 2023 (4D-AEID-2023)

Lu, Binbin; Shi, Lingkun; Wang, Chun; Sun, Huabo; Guo, Qinli; Wang, Xinbo; Wang, Xuhui

doi:10.5194/essd-2026-530

Preprints

https://doi.org/10.5194/essd-2026-530

Preprints

12 Aug 2026

| 12 Aug 2026

Status: this preprint is currently under review for the journal ESSD.

4-D Aviation Emission Inventory Data of China Estimated with QAR Data in 2023 (4D-AEID-2023)

Binbin Lu, Lingkun Shi, Chun Wang, Huabo Sun, Qinli Guo, Xinbo Wang, and Xuhui Wang

Abstract. Accurate aviation emission inventories are essential for environmental impact assessment and mitigation, yet existing estimates predominantly rely on aircraft performance models and standardized assumptions, introducing significant uncertainties. Here, we present a high-resolution, four-dimensional aviation emission inventory for China's civil aviation sector in 2023, built directly from Quick Access Recorder (QAR) data covering 4.52 million flights (~96.8 % of national commercial movements). By coupling second-level fuel-flow measurements with observed flight-state and meteorological parameters via the Boeing Fuel Flow Method 2 and a dynamic thermodynamic correction, this dataset yields temporally and spatially explicit emission estimates for CO₂, NOₓ, SO₂, PM, CO, and HC across all flight phases. In 2023, China's civil aviation consumed 28.4 Mt of fuel, emitting 89.3 Mt CO₂, 634.6 kt NOₓ, 28.4 kt SO₂, 8.46 kt PM, 44.3 kt CO, and 6.35 kt HC. Monte Carlo analysis demonstrates that direct physical observations substantially constrain inventory uncertainty, reducing coefficients of variation (CV) for major pollutants to 2 %–10 % and shifting the dominant error source from activity-level assumptions to emission index parameterizations. High temporal resolution reveals pronounced phase-specific variations: taxiing accounts for 23.7 % of HC and 19.4 % of CO emissions due to low-thrust incomplete combustion, while the brief high-thrust acceleration segment preceding climb contributes 18.0 % of total PM. Spatially, emissions are highly concentrated, with the top 10 % of grid cells accounting for 74.2 % of national CO₂ emissions, exhibiting significant clustering (Global Moran's I = 0.3389, p < 0.05) along major corridors. Providing second-level and high resolution, this dataset serves as an observation-based benchmark for refining existing inventories and assessing near-airport air quality. The dataset is publicly available at https://doi.org/10.5281/zenodo.20828076 (Lu et al., 2026).

Received: 05 Jul 2026 – Discussion started: 12 Aug 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

Preprint (PDF, 1975 KB)

Supplement (251 KB)

Download & links

Binbin Lu, Lingkun Shi, Chun Wang, Huabo Sun, Qinli Guo, Xinbo Wang, and Xuhui Wang

Status: open (until 18 Sep 2026)

Post a comment Subscribe to comment alert

RC1:
'Comment on essd-2026-530', Anonymous Referee #1, 16 Aug 2026 reply
Overall, this manuscript is built on a highly valuable QAR dataset. The combination of nationwide coverage, second-level trajectories, and engine-level fuel-flow measurements has clear potential to substantially improve aviation emission inventories and to reveal operational features that are difficult to resolve using traditional approaches. However, the current manuscript does not yet fully realize this potential. Several critical steps between the QAR observations and the estimated emissions are either insufficiently documented or methodologically questionable, particularly the implementation of BFFM2, the treatment of humidity and PM, and the uncertainty analysis. In addition, key data-processing procedures, engine matching, and the construction of the final “4D” inventory are not described with sufficient transparency. As a result, some of the strongest conclusions regarding improved accuracy, reduced uncertainty, and newly resolved emission processes are not yet adequately supported. These methodological issues should be resolved before the inventory and its derived insights can be reliably interpreted.
The implementation of BFFM2 requires careful verification. Based on the equations currently presented, measured in-flight fuel flow appears to be mapped directly to the EEDB-derived reference EI relationship, followed by atmospheric correction of the EI. The intermediate conversion from actual in-flight fuel flow to equivalent reference fuel flow is not shown, and Mach-number correction is also absent. If this conversion was not performed, the fuel-flow input used to obtain the reference EI would not be consistent with the standard test-bench conditions underlying the EEDB, particularly during cruise. The authors should explicitly present the complete BFFM2 calculation chain and confirm whether the reference fuel-flow correction was applied. If it was omitted, the resulting NOx, CO, and HC estimates would need to be reassessed.

The humidity treatment is not physically clear from the current description. Temperature and pressure alone are insufficient to determine actual specific humidity. The manuscript should identify the actual humidity information or assumption used.

The full-flight PM methodology is incomplete. The atmospheric correction equations are provided for NOx, CO, and HC, but no corresponding procedure is given for converting EEDB PM values from LTO reference conditions to actual full-flight conditions. This is particularly important because the manuscript reports a national full-flight PM inventory and interprets phase-resolved PM behavior in terms of temperature and pressure effects. The authors should provide the full PM calculation framework, clarify its applicability outside the LTO regime, and justify the physical interpretation of the resulting cruise and climb PM emissions.

The Monte Carlo analysis should be reframed as uncertainty propagation under prescribed input distributions rather than as an assessment of model accuracy. The resulting CO2 CVs of approximately 2%, 15%, and 30% essentially reproduce the fuel-flow uncertainties assigned a priori to QAR, ADS-B, and SFD. Likewise, the SFD systematic underestimation is also partly imposed by setting the trajectory-distance factor to a mean of 0.93. These results demonstrate how the assumed error budget propagates through each framework, but they do not independently establish which inventory is more accurate.

The LTO/CCD altitude definition in Figure 6 needs immediate verification. The manuscript labels LTO as below 3000 m and CCD as above 3000 m, whereas the conventional ICAO LTO boundary is 3000 ft, approximately 915 m. If 3000 m was actually used in the calculations rather than being a labeling error, the reported LTO/CCD distance and emission fractions would need to be recalculated. This also affects the interpretation of the reported low-altitude emissions.

The QAR preprocessing and reconstruction procedure needs substantially more documentation. The manuscript should report the fractions of flights and second-level observations that are retained as original measurements, reconstructed by short-gap interpolation, and reconstructed using the similarity-based method. For the similarity-based reconstruction, the matching criteria, reconstruction algorithm, validation error, and contribution of reconstructed data to the final inventory are necessary to evaluate how much of the inventory is genuinely observation-based.

The QAR-schedule fusion and engine assignment are critical to the entire EI calculation but are currently treated as a black box. The external schedule database is not identified, and aircraft registration alone does not uniquely identify an individual flight. The manuscript should provide the database source and version, the complete linkage procedure, the origin of the ENGTYPE field, the mapping from ENGTYPE to EEDB engine identifiers, and diagnostics for matching quality. In particular, the fractions of flights assigned to an exact engine variant, an engine-family fallback, or an unresolved category should be reported, together with the sensitivity of emissions to ambiguous assigments.

The SO2 assumptions appear internally inconsistent. The nominal inventory uses a fixed EI of 1 g/kg, as reflected by the direct correspondence between 28.4 Mt of fuel and 28.4 kt of SO2, whereas the Monte Carlo analysis assumes a mean fuel sulfur content of 600 ppm with a standard deviation of 150 ppm. These assumptions do not appear to represent the same nominal sulfur condition under a simple sulfur mass balance. The point-estimate and uncertainty assumptions should be reconciled, and the sulfur-to-SO2 conversion should be stated explicitly.

The PM fit for the CF6-80A engine deserves specific verification. The reported parameters and are extreme outliers relative to the other engine models and imply an exceptionally strong dependence of EI on fuel flow. The underlying EEDB anchor points, units, fitted curve, and influence of this engine-specific relationship on the national PM total should be shown. This is particularly important given that the reported PM inventory is substantially higher than most previous estimates.

The 13.5% difference between the QAR-derived fuel consumption of 28.4 Mt and the CAAC-derived estimate of 25.03 Mt requires a more rigorous validation. The current explanation based mainly on accounting definitions is plausible but not sufficient, especially because the QAR dataset covers only 96.8% of reported commercial flight movements yet already produces a larger national fuel total. The comparison should first harmonize system boundaries and then examine possible contributions from scope differences, duplicate records, fuel-flow integration, unit conversion, and other systematic effects.

The unusually low CO estimate also warrants closer examination. The reported 2023 full-flight CO total of 44.3 kt is lower than the 46.1 kt reported by another 2023 study for the LTO phase alone. This is too substantial to dismiss as a generic difference in taxi or idle treatment.

Several interpretations extend beyond what the inventory itself demonstrates. In particular, CO2 emissions below 3000 m should not be directly interpreted as near-surface air-quality impacts, and CO2 is not generally the principal pollutant of concern in airport local-air-quality assessments. These statements should be limited to emission location and potential relevance, while conclusions about local air quality or exposure should be reserved for pollutants such as Nox and PM or supported by a dedicated dispersion analysis.

Reply
Citation: https://doi.org/10.5194/essd-2026-530-RC1
EC1: 'Editorial comment on essd-2026-530', Yuqiang Zhang, 19 Aug 2026 reply

The manuscript repeatedly describes the product as a four-dimensional aviation emission inventory, and the Data Availability section states that the dataset is provided as GeoTIFF files at 0.1° × 0.1° resolution for CO₂, NOx, SO₂, PM, CO, and HC. However, based on my inspection of the files currently available on Zenodo, the GeoTIFF files appear to contain only two-dimensional spatial fields, rather than explicit temporal, vertical, or phase-resolved dimensions.
Could the authors please clarify whether this is intentional? If the full four-dimensional dataset is available through additional files, layers, metadata, or a separate data structure, these should be clearly described in the Data Availability section and accompanying documentation. If the public dataset only provides 2-D gridded outputs, the authors should revise the terminology throughout the manuscript to avoid overstating the dimensionality of the released dataset. Given that the “4-D” nature of the dataset is a central claim of the manuscript, this issue is important for assessing the dataset’s completeness, reproducibility, and usability.

Reply

Citation: https://doi.org/10.5194/essd-2026-530-EC1
RC2: 'Comment on essd-2026-530', Anonymous Referee #2, 30 Aug 2026 reply

I agree with all the comments provided by Referee #1 and would like to add the following comments:
I agree that methodology presented in the manuscript has a certain degree of novelty. However, I have several concerns, ranging from minor to major regarding publication of the manuscript in its current form.
Major

1. Line 38-39: I am not fully convinced by the statement that a “high-resolution aviation emission inventory serves as a critical link connecting aviation activities with atmospheric environmental responses.” The manuscript should provide a clearer justification for why such high spatial and temporal resolution is necessary and what specific scientific or practical advantages it offers. This point is particularly important because the conclusion states that “the inventory provides second-level temporal resolution and nationwide coverage, offering a basis for near-airport air quality assessment and for evaluating the assumptions underlying existing lower-resolution aviation emission inventories.” I suggest that the Introduction be revised to better align with this more specific motivation. For example, Yim et al. (2015), which is cited in the manuscript, employed an atmospheric model with a horizontal resolution of approximately 4° × 5°. When an emission inventory with a resolution of 0.1° × 0.1° is used in such a coarse-resolution model, much of the benefit of the fine spatial resolution may be lost through model-grid aggregation. Therefore, the manuscript should more clearly explain under what applications the proposed high-resolution inventory provides meaningful added value. If the primary advantages of the inventory are improved estimation of total emissions and/or its applicability to regional- or local-scale air quality assessments, particularly near airports, I find these motivations reasonable. However, the current Introduction does not articulate these points sufficiently clearly and should be revised accordingly.

2. Lines 100 and 143–144: The manuscript states that the dataset covers approximately 4.52 million flights operating to and from civil airports across China in 2023 and later refers to a comprehensive dataset of civil aviation operations in mainland China. However, the exact spatial scope of the dataset is not sufficiently clear. From a data-coverage perspective, it appears that the inventory may be intended to represent aviation emissions over mainland China, but this should be stated explicitly. In particular, the authors should clarify how flight trajectories and emissions were treated at the geographical boundaries of the study domain. For example, for international flights departing from or arriving at airports in mainland China, were emissions included only while the aircraft was located within the defined mainland China domain, or were the full flight trajectories included? Similarly, how were flights crossing the domain boundary handled?

3. Line 168: Does the term “dynamic environmental correction model” refer to the methodology described in Section 2.3.2, “Atmospheric Correction of Emission Indices”? If so, I suggest using consistent terminology throughout the manuscript to avoid potential confusion.

4. Line 214, Equation (5): Please check Equation (5). It appears that the exponent should be 0.51 and 1.65. It is currently unclear whether this is only a typo in the manuscript or whether the equation was implemented incorrectly in the actual calculations. This should be carefully verified, as it may directly affect the resulting emissions estimates.

5. As also noted by RC #1, I believe that several parts of the manuscript go beyond what can be directly supported by the presented analysis. For example, the discussion beginning around Line 298 appears to attribute the differences between the emission inventories to specific causes that have not been explained in the study. While these explanations look plausible, they remain hypotheses unless authors apply the stated assumptions to the QAR data. Subsequently, the authors should reconstruct the emission inventory accordingly and then compare the resulting estimates with the CAAC inventory. Unless this statement may be removed from the document.

6. A similar issue appears in Lines 329-330. In this case, appropriate citations may help support “general” validity of the proposed explanation. However, there should still be a clear distinction between “general” statement and what has been presented by the results of the current manuscript.

7. Is NOx mass NO2 equivalent or nitrogen mass?

8. The dataset currently provided is incomplete. As also raised by the editor, the authors should either revise the manuscript to clearly reflect the actual scope of the available dataset or provide the complete dataset described in the manuscript.

Minor

1. There are few instances throughout the manuscript where spaces are either missing or duplicated (maybe more than indicated below, please check!)

a. Space

Line 90, 100, 172, 295, 440, 467

b. Double space

Line 355, 442, 553

2. Table 3 (also there is no space here)

Font size are different within the same table.

Tabe 4 -> no space here again in Emission Index(NOx/others)

3. As a reader, I found that there are several points where the terminology or intended reference has to be inferred from the surrounding context. I encourage the authors to review the manuscript with this in mind and define terms and references more explicitly where needed. For example, the “typical flight” denoted in the caption of Figure 5 presumably refers to the B737-800 fight described in Line 422.

Reply

Citation: https://doi.org/10.5194/essd-2026-530-RC2

Binbin Lu, Lingkun Shi, Chun Wang, Huabo Sun, Qinli Guo, Xinbo Wang, and Xuhui Wang

Supplement

https://doi.org/10.5194/essd-2026-530-supplement

Data sets

High-Resolution Gridded Aviation Emission Inventory for China (2023) B. Lu et al. https://zenodo.org/records/20828076

Binbin Lu, Lingkun Shi, Chun Wang, Huabo Sun, Qinli Guo, Xinbo Wang, and Xuhui Wang

Viewed

Total article views: 40 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
30	6	4	40	10	4	4

HTML: 30
PDF: 6
XML: 4
Total: 40
Supplement: 10
BibTeX: 4
EndNote: 4

Views and downloads (calculated since 12 Aug 2026)

Month	HTML	PDF	XML	Total
Aug 2026	30	6	4	40

Cumulative views and downloads (calculated since 12 Aug 2026)

Month	HTML	PDF	XML	Total
Aug 2026	30	6	4	40

Viewed (geographical distribution)

Total article views: 23 (including HTML, PDF, and XML) Thereof 23 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 01 Sep 2026

Download

Preprint (1975 KB)
Metadata XML

Short summary

Airplane emissions are usually estimated, not measured. We used flight data recorders from 4.52 million Chinese flights in 2023 to track real fuel burn and pollution, second by second. Aviation in China burned 28.4 million tonnes of fuel, emitting 89.3 million tonnes of CO₂ unevenly: just 10 % of flight paths produce 74 % of emissions, and brief ground idling causes disproportionate pollution. This precise map helps target routes and airport operations where emission cuts matter most.


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