the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 11 Aug 2025
The ELK global emission inventory for the transport sectors
Abstract. The transport sectors, comprising land transport, shipping and aviation, are major contributors to climate change and have a detrimental impact on air quality, with adverse consequences for human health. The emissions from transport, already contributing 23 % of total anthropogenic CO2 emissions in 2019, are projected to continuously grow in the future, challenging the achievement of climate protection and pollution reduction targets. A major goal of transport research on climate and air quality is the accurate assessment of its impacts, which requires detailed emission data to drive atmospheric models and calculate projections for future scenarios. This paper presents the ELK global emission inventory for the transport sectors. The inventory is developed using a consistent bottom-up approach fed with a wide range of input data to model the transport fleets of land transport, shipping and aviation. It provides several major improvements over existing datasets, such as the explicit resolution of the emissions at the subsector level, the consideration of transport-specific quantities and emission species, and the quantification of the transport-related emissions from the energy sectors. The emission data is complemented by a quantitative uncertainty score, based on a detailed expert-judgement analysis along the modelling chain, from the activity data to the emission factors. The emission data is validated by comparing it with other, well-established global inventories, and biases are discussed and, where possible, explained in terms of the different assumptions and features of the underlying emission models. The ELK dataset is released under an open-source licence to encourage their use in the atmospheric modelling community.
- Preprint
(8141 KB) - Metadata XML
-
Supplement
(87418 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on essd-2025-454', Anonymous Referee #1, 16 Sep 2025
-
RC2: 'Comment on essd-2025-454', Flávio Quadros, 28 Oct 2025
The manuscript describes a new inventory of global atmospheric emissions from transportation in the year 2019, covering land transport, shipping, aviation, and refining of transportation fuel. The inventory uses up-to-date methods and data sources, yielding estimates that should be of comparable quality to the available more general global anthropogenic emission inventories. Due to the specific choice of methods and data sources, this dataset offers unique features that make it a valuable and useful addition to the emissions inventories already publicly available. These include a (semi-)systematic assignment of uncertainty scores to all emission species and subsectors with some spatial resolution, quantification of propulsion efficiency for aircraft, and estimates of emissions associated with refining of fuel used for transport. The manuscript adequately describes the methodology and discusses the results obtained. The data is properly formatted and documented.
In my opinion, some points would still benefit from further clarification, as per the specific comments below. Besides plenty of minor comments, I think some additional details on the methodology and additional discussion on some of the results that stand out more are needed. Additionally, the scope of the inventory needs to be stated more clearly in the Introduction, and the qualitative nature of the uncertainty assessment should be better indicated.Specific comments:
1. (line 68): The label "present day" is used for 2019, which by now 6 years ago. Is there any indication of how much things might have changed since then? E.g. what has been the recent % change per year, very broadly speaking?
2. (section 2.1): the methods used for non-road (rail) transport are not really discussed here, both activity data and emission factors.
3. (section 2.1.3): it is not (explicitly) addressed how are electric vehicles taken into account. This can have a larger impact in certain regions and for certain subsectors (such as rail).
4. (L200-203): are the stock numbers from 2018 instead of 2019? Where does the 2021 come from?
5. (L212-217): there are not enough details here for this method to be reproducible. Is there a separate publication that describes it? Especially if this is one of the improvements of ELK vs. the state of the art, it would be nice to mention things like what is the structure of the network, how was is trained, what is the source of the data, how well can it reproduce actual traffic data, etc.
6. (L218-219): is this improvement shown in your results or from previous studies? The evidence for it should be referenced here.
7. (L223-234): you describe this alternative method, show that it results in significantly different spatial distribution (Fig. S1), but then do not discuss what does that mean and this new method is simply discarded. "The spatial disaggregation for Europe is based on ... ULTImodel", except that is not was is used for the data being released or presented in the Results section, if I understood correctly.
8. (L279-280): what fraction of the fleet moving in 2019 (or fuel usage) does these 86,192 represent? I.e., what percentage is missing from the estimates due to unavailable identification.
9. (L280): What is the source of this assumption?
10. (L401-405): despite ELK being a global inventory, emissions from domestic navigation are only evaluated for select regions / river systems (understandably due to data limitations). Because of this, I think this caveat should be emphasized earlier in the text. The other mentioned systems are said to contribute less than 4% of tonnage, but that is from a webpage from 2012 which quotes statistics from even earlier. And crucially, is there any estimate of what is the share of transport from waterways besides those mentioned?
11. (L434-435): I am not an expert on this field, but is this assumption that conditions of waterways in China are closer to Europe than the US trivial? Also, my impression is that for everything else ELK does not seem to go for "a conservative approach" and instead aims for a "best estimate", so this line seems a bit odd.
12. (L452-482): This paragraph reads a bit weird because that you are speculating about the feasibility of applying a method that is not used here. Some of this would perhaps be better suited for a Discussion section rather than the Methods.
13. (section 2.3): just a comment that does not need addressing, but since you are going through the trouble of estimating transport-related emissions in the energy sector, I would like to suggest a similar improvement could be the estimation of emissions from airport activities (ground support equipment, electricity and HVAC for buildings, fuel storage, land traffic supporting the airport, etc.). There had been previous efforts in this regard, such as in Stettler et al. 2011.
14. (L524-526): what is the rationale behind the 5/7 split in the seasons? Are there any indicators that these two weeks with these weights might yield a good approximation of the annual average?
15. (L527-531): the definition of aircraft subsectors are maybe a little too arbitrary. For example, a Boeing 757 would be classified as a wide-body, even though it is a "single-aisle" in the literal sense. The A321XLR (which was not flying in 2019) would likewise fall under an unnatural classification.
16. (section 2.3.3, section 3.3.2): it is not specified, but I assume your activity data gives aircraft typecodes, which you then use to select emissions data from ICAO. As pointed out in Quadros et al. 2022, there is a large amount of uncertainty associated in the mapping of aircraft typecodes to specific engine models, with the emissions of some species varying by an order of magnitude depending on the choices made. So I think it is important to describe how was this issue was handled here and what are its implications with regards to the uncertainty of the emission estimates.
17. (L574-580): BC is estimated from smoke number, but is also a component of nvPM which is estimated from separate measurements. I am curious if there are instances when the mass of BC is greater than nvPMm. Is that something you checked?
18. (L578-580): are the (LTO) emission rates of nvPM adjusted for cruise conditions in any way?
19. (L597-598): great circle trajectories are known to underestimate emissions (Teoh et al. 2024, for example), I am surprised that you do not make any adjustments to the emissions in those cases. Considering you already have a database of flights with actual trajectories, would it be feasible to derive a factor to adjust flight lengths when you do not have trajectories?
20. (section 2.3.5): I am curious if you estimated what the overall difference in fuel usage (or emissions) is of considering wind. Since head and tail winds cancel each other out over all flights, the difference should mainly be from crosswind drift. Quadros et al. 2022 found a ~0.7% increase in fuel burn, but maybe this approach with wind rose distributions yields a larger impact.
21. (section 2.3.6): if you have "actual 3D flight paths" (L592), why use this approach of the average between optimal and constant cruise flight level? Could the actual cruise altitudes be used in these estimates?
22. (L637-639): just something to consider in the future, but maybe for air quality research a higher vertical resolution close to the surface would be beneficial, as you could get more precision to split how much emissions are inside/outside the boundary layer.
23. (L665-666): how much these 549 represent of global emissions? Is there any metric available that could put this into perspective, i.e. % of refineries or capacity worldwide?
24. (L652-655, L673-677): some 13-34% of refining emissions are excluded from the analysis, and refining excludes some 13-44% of all emissions related to fuel production. This limitation is something that I think should be highlighted more strongly. For example, when you say "quantification of the transport-related emissions from the energy sector" in the Abstract, it gives a misleading idea about the scope of this quantification.
25. (L719): I wish it was not the case, but at least for aviation I am aware that emissions are expect to increase significantly in the future.
26. (L10-11, L65-67, L87, L731-732, etc.): I appreciate the uncertainty assessment as a valuable contribution of this work and a novel feature of ELK, but I am not sure you can call it a "quantitative assessment" or "quantitative uncertainty score". It is a qualitative assessment which you happen to label with numbers 1-5. To me, a quantitative uncertainty analysis would involve estimating uncertainty in quantifiable terms, i.e. ±X kg of emissions for a confidence interval of Y%. The "Interpretation" column of Table 4 is almost completely subjective; you are assigning the numbers 1-5 using "lowest uncertainty" / "low uncertainty" / "medium uncertainty" / "relatively high uncertainty" / "highest uncertainty" as your criteria.
27. (L780-782): could a reference be provided for these indications? Border control might slow down traffic, but I would expect major border crossings to have relatively higher traffic than a similarly sized road elsewhere.
28. (L801-803): what does "leaving emissions of LCVs unconsidered" mean? You still have LCV emissions in Europe. Is it just that those emissions are calculated only from GDP without activity data?
29. (L807-808): how is volume inferred?
30. (L837-839, L845-847): tire wear PN is said to be in the 1010-1011 per vehicle-km range while a wider range of 109-1013 is given for brake PN. So it is not clear why the former was excluded due to high uncertainties but not the latter.
31. (L1009-1010): the scope of the inventory should be clarified earlier in the Introduction or Methods. Non-scheduled flights and these other categories are also transportation, but it is only here in the section about uncertainty that it is clarified that the inventory concerns exclusively scheduled civil aviation flights.
32. (L1068-1070): 51%/35% of movements, but presumably a much smaller percentage of CO2 emissions (and other species)?
33. (L1070-1072): several in-flight measurement campaigns have been done since 1997, maybe reference to a newer discussion could be cited here.
34. (Table S4): what does tirewear from rail mean?
35. (L1157-1159, Figure S26): compared to other inventories, CO2 estimates for W.NA are a third lower, C.NA estimates are half, and E.NA are about the same. Since all three are dominated by emissions in the US, I would expect the methods and data sources to be consistent across them. What could be driving this behavior?
36. (L1166-1169): "factor-5 difference" should be 115 times higher? Is there a reason why EDGAR8 has such low amounts of SO2?
37. (Figure 6): I understand Greenland, but is there a particular reason why there are no emissions in Montenegro and Kosovo?
38. (Figure 6): related to the lack of description on the methods for estimating rail emissions, why are there so much emissions in railways that are electrified? For example, ~100% of rail in Switzerland.
39. (L1208-1211, Figure S47): these differences are very large. What could be driving them? The uncertainty score is rated "2 (robust)" for N.Atlantic-Ocean, but CEDS has 4 times the amount of CO2 there.
40. (L1266-1279, Figure 14): the diurnal cycle analysis is shown prominently as a unique feature of this inventory, but I am missing what the use is of looking at it globally in UTC. Would it not make more sense to think of the "diurnal cycle" in terms of local time? Sunlight and temperature cycles, which are relevant for contrail formation, are going to be shifted according to local time.
Technical corrections:41. (Figure 2): the x labels are in different orders between (a) and (b).
42. (L558): typo in "for a plenty".
43. (L574): typo in "are build on".
44. (L625): weighted = averaged with equal weights?
45. (L784): "GDPpc" / "GDPpC" capitalization inconsistency.
46. (L892): maybe I am confused, but it stands to reason that stationary vessels might have emissions underestimated.
47. (L1027-1028): are flight path data available for Asia?
48. (Supplement): "tire" vs. "tyre" in the main text.
49. (Table S4): formatting of PM10 subscript.
50. (L1337-1338): you could also leave Russia and Japan in the figure and add the note to it.
51. (Data availability): just a suggestion, and there are tradeoffs, but a lot of storage could be saved by using compression on the netCDF files. Relatedly, there are some species for rail transport where you have 300 MB files that are entirely zeros, which could be omitted entirely.
Citation: https://doi.org/10.5194/essd-2025-454-RC2
Data sets
ELK - Global Emission Inventory - Land Transport Sector, 2019 S. Ehrenberger et al. https://doi.org/10.15489/d9dswthdix21
ELK - Global Emission Inventory - Shipping Sector, 2019 P. Banys et al. https://doi.org/10.15489/lhqawfes5755
ELK - Global Emission Inventory - Aviation Sector, 2019 C. M. Weder et al. https://doi.org/10.15489/86s8uwpxik95
ELK - Global Emission Inventory - Energy-for-Transport Sector, 2019 P. Draheim et al. https://doi.org/10.15489/gixadaq6ds98
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,262 | 76 | 18 | 1,356 | 42 | 31 | 28 |
- HTML: 1,262
- PDF: 76
- XML: 18
- Total: 1,356
- Supplement: 42
- BibTeX: 31
- EndNote: 28
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript presents a global emission inventory for the transport sectors of 2019, covering land transport, shipping, aviation, and transport-related emissions from the energy sector. Although this emission inventory is developed with detailed data and models, the spatiotemporal resolution is the same with some existing data products. Many studies on the source-specific transport emissions provided finer resolution data. The manuscript is too long to read.