the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Climate impact of contrail cirrus from hydrogen combustion aircraft
Abstract. To mitigate the climate impact of aviation, combustion of hydrogen as a fuel is one possible future pathway. Hydrogen combustion leads to zero carbon emissions in the exhaust, representing a major step toward climate neutrality, although the non-CO2 effects, primary contrail cirrus remain uncertain. In this study, we simulate the climate impact, in terms of energy forcing, of contrail cirrus from hydrogen combustion aviation using a modified version of the Contrail Cirrus Prediction model (CoCiP).
Without soot in the exhaust from hydrogen combustion, contrail ice particles instead form on ambient aerosols entrained into the plume and on lubrication oil droplets in the exhaust. The formation of ice particles are modeled using an emulator developed from a theoretically based microphysical contrail formation model.
Following the Schmidt Applemann criterion, hydrogen combustion enables contrail formation at lower altitudes and higher temperatures than fossil jet fuel. However, we find a significant reduction in contrail energy forcing. This result holds across a wide range of assumptions, including different oil particle size distributions and properties, with a global average reduction of about 70 % using our base case assumptions. We conclude that hydrogen aircraft not only eliminates CO2 emissions but may also substantially reduce the climate impact of contrail cirrus, although the reduction and magnitude depend on engine design for lubrication oil handling.
- Preprint
(4682 KB) - Metadata XML
-
Supplement
(8159 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-3535', Anonymous Referee #1, 01 Oct 2025
-
RC2: 'Comment on egusphere-2025-3535', Anonymous Referee #2, 13 Oct 2025
The study provides an estimate of the radiative effect of contrail-cirrus originating from aircraft with hydrogen combustion. It thus makes an important contribution to assessing the overall climate impact of hydrogen-powered aviation beyond CO₂ emissions.
The authors make use of a microphysical model developed by Kärcher et al. (2015), originally designed to model contrail formation on soot and ambient particles from kerosene combustion. Since soot particles are absent in hydrogen combustion, the authors adapted the model to account for contrail formation on lubrication oil particles and ambient aerosol particles. They further developed emulators of the model outcomes and integrated them into the Contrail Cirrus Prediction Model (CoCiP) to estimate the energy forcing of contrail-cirrus.
The manuscript is well-structured, concise, and generally easy to follow. At some points, however, the study would benefit from a more thorough discussion of the physical plausibility of the results and the associated uncertainties of the applied modeling chain.Major comments:
- Since the role of lubrication oil particles is uncertain, the authors explore many possible scenarios. Their assumptions about oil emissions and particle sizes are based on measurements behind conventional aircraft and laboratory experiments, since no such measurements exist yet for hydrogen engines.
In my view, however, not all potential scenarios are fully covered.
a) The smallest geometric mean radius of oil particles investigated is 5 nm. However, volatile particle sizes can reach down to ~1 nm (e.g., Figure 3 in Yu et al. 2024). At present, it is unclear whether such small particles could form from evaporated oil at cruise altitude (and measurement devices may struggle to detect them). Still, this is one potential scenario missing in the analysis. For the same oil mass emission index, the particle number increases by more than a factor of ~100 when the size changes from 5 nm to 1 nm. Of course, smaller particles are more affected by the Kelvin effect, so not all may activate. But given the high supersaturations in hydrogen combustion, I can still imagine that the larger number of oil particles could produce ice crystal numbers exceeding those from conventional kerosene combustion. For completeness, such “worst-case” scenarios should be included.
b) On the other hand, the authors should also emphasize the “best-case” scenario more strongly. As discussed, the impact of oil could be significantly reduced—or even eliminated—if venting does not occur in the hot sections. In this ideal case, ice crystals would only form on ambient aerosols, which cannot be mitigated through technical measures. When estimating the radiative effect of contrail cirrus, the natural variability of aerosol properties should be represented.
Currently, the authors assume a constant ambient aerosol number concentration of 1000 cm⁻³. They argue that Bier et al. (2024) already investigated the effect of ambient aerosols, which is why the current study focuses on lubrication oil properties. However, Bier et al. (2024) only addressed the formation stage and did not estimate the radiative effect of contrail-cirrus.
In the current study, ERA5 reanalysis data for the year 2019 are used to represent background conditions for temperature, pressure, and relative humidity. Ideally, an external aerosol module should also be included to realistically capture the large variability of aerosol properties. At the very least, the assumed aerosol number concentration should be varied within realistic ranges to better assess how sensitive the radiative effect is to ambient aerosol variability. - Figure 1: I do not understand the physical basis of the activation fraction dependency on ambient temperature. The decrease in activation fractions of lubrication oil particles with increasing temperature seems plausible to me. However, the increase in activation fractions of ambient aerosols with temperature is less intuitive. In lines 347–350, you argue that with lower activation of lubrication oil particles, it takes longer to quench the plume supersaturation, which allows more ambient aerosols to be entrained and activated at higher temperatures. If this competition effect were the explanation, I would expect a clear dependence on the number of oil particles. Yet, the activation fractions of ambient aerosols appear very similar across all panels.
Could you check whether this temperature dependence still exists when the number of oil particles is set to zero? If it does, this would seem to contradict the findings of Bier et al. (2024), who showed that the number of ice crystals formed decreases with increasing temperature due to shorter-lasting and lower supersaturations.
It could be helpful to compare the ice crystal numbers obtained here with those reported by Bier et al. (2024) and to discuss possible reasons for any differences. In addition, Zink et al. (2025) also studied the competition between lubrication oil particles and ambient aerosols in hydrogen combustion. A comparison with their results might provide further insight. - The authors argue that the decisive factor for ice crystal formation is the number of particles present. However, this conclusion cannot be directly inferred from Figure 5. When the results are plotted as a function of geometric mean radius, it is not immediately clear how radius influences the oil particle number (for a fixed oil mass emission index) versus how radius affects activation suppression due to the Kelvin effect. I therefore strongly recommend plotting the results as a function of oil particle number Nₒᵢₗ (per flight meter). In this case, the highest Nₒᵢₗ would correspond to the smallest radius, and the lowest Nₒᵢₗ to the largest radius. In the second row, the two impacts of oil particle radius would then be disentangled, making it possible to directly see how many of the oil particles activate into ice crystals.
Minor comments:
- In the abstract, it is not immediately clear to the reader what your base case assumptions are. I recommend stating them briefly.
- Line 16: The clouds themselves are not artificial; they are artificially produced.
- Line 37-38: What about NOx emissions and the associated formation of nitric acid in hydrogen combustion cases? Could this contribute to volatile particle formation?
- Line 59-61: There is a dedicated simulation study on the potential role of lubrication oil particles in contrail formation (Zink et al., 2025).
- Line 67-68: Wasn’t it demonstrated by Ponsonby et al. (2024) that lubrication oil can act as condensation nuclei?
- Line 106: You use a value of 123 MJ kg-1 for the specific combustion heat. Others use 120 MJ kg-1 (Schumann, 1996, Bier et al., 2024). Could you clarify the source of this difference, and discuss its potential implications for your results?
- Line 159-162 and line 229-230: Ponsonby et al. (2025) state that the K15 model is not capable of realistically resolving competition effects among different particle types. Please review this statement and discuss in more detail how this influence your results.
- Line 232: Stating that the fit is “very good” is rather qualitative. Please quantify the errors between the original K15 model and your fit to provide a more precise assessment.
- Figure 1: Your fits occasionally fail to reproduce the sharp decrease of activation fraction to zero at high ambient temperatures. Homogeneous freezing temperatures of supercooled droplets are typically below ~235 K, yet in some extreme cases, your fits show activation fractions above zero up to 250 K. Would it be beneficial to introduce a hard cut at 235 K?
- Table 1: You provide the unit “nm” for the standard deviation, but it should be unitless, correct?
- Table 1: Your oil emission index is based on measurements behind kerosene engines and is given in mg (kg fuel)⁻¹. Care must be taken when applying this emission index to a different fuel type. The lubrication system is separate, and oil emissions are not necessarily related to fuel flow. For example, if both a kerosene and a hydrogen engine emit the same oil mass from their lubrication systems, the emission indices will differ due to the different fuel flows, even though the actual emitted oil mass is the same. This may not strongly affect your conclusions, especially since you also varied the oil mass emission index, but it would be helpful to include a brief clarification on this point.
- Figure 2 is based on the assumption that your model aircraft is present at each grid point at every timestep. Is that correct? In other words, no realistic air traffic is assumed, right? It is a valid approach to look at such potential contrail coverage. But a few more words on the used method would improve clarity.
- Line 425-428: CoCiP has been compared to measurement data for conventional kerosene combustion. It is not clear whether this validation also applies to scenarios with strongly reduced ice crystal numbers. Please clarify.
- I appreciate the discussion about the comparison of CoCiP and APCEMM. However, I am not happy about the sentence 'It is still uncertain if APCEMM or CoCiP is closer to the real world'. For any model, there is of course kind of uncertainty how close the results are to reality. But Akhtar Martínez et al. (2025) demonstrated that an important physical process is not well represented in CoCiP: the assumption of monodisperse ice particle sizes leads to the same sedimentation rate for all ice crystals, which produces a fallstreak and cuts off the lifecycle. As you note, APCEMM shows better agreement with large-eddy simulation studies.
This raises questions about the reliability of the radiative effect estimates in the present study. I recommend that this uncertainty be explicitly acknowledged, and I would also welcome a statement about it in the abstract. - Table S1: Should the unit of aerosol number concentration be cm-3?
- S3: Do the relative humidity values rh=0.1…..1 refer to relative humidity values over water? Only the combinations of Ta and rh are relevant, where the environment is supersaturated over ice (then persistent contrails develop). Would restricting to these values results in a better fit?
- Figures S5 and S6: Could you combine them into a single figure, where you plot the difference between the two approaches?
Technical Corrections:
- Line 71: It seems that something went wrong with the citations.
- Line 86: Just cite Kärcher et al. with \citet{}
- Line 207: It seems that something went wrong with the citations.
References:
Kärcher, B., Burkhardt, U., Bier, A., Bock, L., and Ford, I. J.: The microphysical pathway to contrail formation, J. Geophys. Res., 120, 7893–7927, https://doi.org/10.1002/2015JD023491, 2015.
Yu, F., Kärcher, B., and Anderson, B. E.: Revisiting Contrail Ice Formation: Impact of Primary Soot Particle Sizes and Contribution of Volatile Particles, Environ. Sci. Technol., 58, 17 650–17 660, https://doi.org/10.1021/acs.est.4c04340, 2024
Bier, A., Unterstrasser, S., Zink, J., Hillenbrand, D., Jurkat-Witschas, T., and Lottermoser, A.: Contrail formation on ambient aerosol particles for aircraft with hydrogen combustion: a box model trajectory study, Atmos. Chem. Phys., 24, 2319–2344, https://doi.org/10.5194/acp24-2319-2024, 2024Ponsonby, J., King, L., Murray, B. J., and Stettler, M. E. J.: Jet aircraft lubrication oil droplets as contrail ice-forming particles, Atmos. Chem.Phys., 24, 2045–2058, https://doi.org/10.5194/acp-24-2045-2024, 2024
Ponsonby, J., Teoh, R., Kärcher, B., and Stettler, M.: An updated microphysical model for particle activation in contrails: the role of volatile plume particles, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2025-1717, 2025
Zink, J., Unterstrasser, S., and Jurkat-Witschas, T.: On the potential role of lubrication oil particles in contrail formation for kerosene and hydrogen combustion, J. Geophys. Res., 130, https://doi.org/10.1029/2025jd043487, 2025Schumann, U.: On conditions for contrail formation from aircraft exhausts, Meteorol. Z., 5, 4–23, https://doi.org/10.1127/metz/5/1996/4, 1996
Akhtar Martínez, C., Eastham, S. D., and Jarrett, J. P.: Contrail models lacking post-fallstreak behavior could underpredict lifetime optical depth, EGUsphere [preprint], https://doi.org/10.5194/egusphere-2025-278, 2025Citation: https://doi.org/10.5194/egusphere-2025-3535-RC2 - Since the role of lubrication oil particles is uncertain, the authors explore many possible scenarios. Their assumptions about oil emissions and particle sizes are based on measurements behind conventional aircraft and laboratory experiments, since no such measurements exist yet for hydrogen engines.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,814 | 42 | 12 | 1,868 | 27 | 12 | 11 |
- HTML: 1,814
- PDF: 42
- XML: 12
- Total: 1,868
- Supplement: 27
- BibTeX: 12
- EndNote: 11
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The authors investigate the potential for hydrogen-fueled aircraft to reduce the climate impacts of contrails. They do so by developing an emulator for a sophisticated model of the early plume, incorporating the effects of lubrication oil and ambient particles. They consider a broad range of possible oil properties, and make reasonable assumptions regarding the performance of near-future hydrogen aircraft. Using the CoCiP model to simulate the mature plume and radiative forcing they find that hydrogen aircraft contrails are likely to produce reduced radiative forcing per flight meter compared to kerosene contrails, but that this is sensitive to the handling of lubrication oil.
The question posed by the authors is interesting and timely, given that there is significant uncertainty regarding the potential environmental benefits of a transition to hydrogen. However, the methods used are somewhat limited, and this in turn limits the degree to which the data can support the conclusions drawn. Given that the results may be strongly dependent on the choice of meteorological data, early plume representation and contrail model, it is striking that there is relatively little uncertainty quantification and no acknowledgement in the abstract of potential sources of error in the methods.
Despite this, I am broadly supportive of the manuscript’s publication pending these issues being addressed. The development of an early plume emulator is a particularly valuable methodological advance, and it will be interesting to see how these results compare to those from groups performing (eg) LES modelling of hydrogen-fueled contrails.
Major comments
The development of an emulator of the K15 model is a potentially significant advance for the community. However, quantification of the accuracy of this emulator is currently lacking. The authors state on lines 229 to 234 that the fit looks “very good” and that, even in the worst cases, “the fit is decent”; these are however fundamentally value judgements. I recommend that the authors pick a representative set of, say, 100 cases (including some extremes) and quantify the degree to which disagreement between the emulator and the parent model propagates to disagreement in their simulated impacts (i.e. contrail lifetime and EF, not just inputs to CoCiP).
On a similar basis, it would be useful to quantify the degree of (dis)agreement between results from CoCiP with the standard and K15 models when simulating fossil jet fuel contrails (lines 256-258). While S6 does show how the two models differ in terms of activation fraction, this difference is not propagated to an overall impact; furthermore the different formatting choices and lack of parity plotting between the results in S6 make it very difficult to establish whether there really are meaningful differences between the two.
I was surprised that homogeneous nucleation of water was not discussed. Given the potential for extremely high RH in the early plume of a hydrogen-fueled contrail, Zink et al. argue in their preprint (Contrail formation for aircraft with hydrogen combustion - Part 1: A systematic microphysical investigation) that homogeneous droplet nucleation (HDN) may be significant under some circumstances. It is not clear to me whether neglecting HDN is or is not likely to significantly change the conclusions of this study, but I would recommend that the authors at least clearly highlight whether or not this eventuality is covered by their modelling approach (since the base model by Kärcher et al. was developed for kerosene-based contrails and therefore does not appear to include HDN). Whether or not it is, I would also recommend at least a brief discussion of the implications of HDN for their study since this mechanism would be expected to be possible even in the absence of lubrication oil.
The authors make recommendations regarding engineering decisions which, while plausible, are not fully justified by this work. A recommendation that future engine design should avoid venting of lubrication oil into the main exhaust (line 472) seems premature given that only one aircraft/engine combination is assessed and given that there is as-yet almost no data on how difficult it might be to re-engineer an engine to avoid this (and therefore whether there may be efficiency penalties). I would therefore suggest that statements such as that in closing – that lubrication oil handling “should” be treated “as a critical parameter” – be instead tempered. A defensible statement might be that, for example, a 10 times reduction in lubrication oil emissions in the exhaust appears to reduce hydrogen contrail RF by 60% compared to a representative base case (this latter number read from figure 5 – clearly a better assessment is needed). Whether or not this is critical depends on balancing that against the alternatives, including the potential costs of doing so (which are not evaluated here).
Minor comments
The authors introduce a well-reasoned set of caveats in the discussion, which I found very encouraging. However, these caveats did not propagate to the abstract which surprised me. Given that the authors appear to have identified potentially significant limitations, I would recommend that the abstract include at least a statement to the effect that these findings need to be validated by both a more comprehensive modelling campaign (including e.g. a higher-fidelity aircraft representation, alternative meteorological/contrail models, and a larger dataset of flights) and of course measurements given that there are as-yet no fully hydrogen-fueled aircraft in service or testing.
It is often unclear whether the authors are referring to CoCiP (essentially a modelling approach, given that the original Fortran codebase is not – to my knowledge – publicly available) or pycontrails (a public, Python-based implementation of CoCiP). If the latter, as I expect given line 274, I would recommend that the authors specifically refer to “CoCiP as implemented in pycontrails”. For example, line 264 states that “CoCiP has a built in aircraft performance model”. From what I understand however, pycontrails includes an implementation of the Poll-Schumann model which it uses to initialize its implementation of CoCiP (but which can equally be used to initialize other models, such as a dry advection code).
Similarly, I was surprised to see that the gridded version of CoCiP was used (line 274) but that the paper describing that code (Engberg et al., 2025) was not cited.
A small request: the zonal plots (e.g. Figures 2, 3, 4) are rather confusingly oriented. It is more conventional for altitude to be the vertical axis, and latitude the horizontal. To ease reader interpretation I would recommend changing these figures accordingly.
There are some minor typographical, grammatical, and formatting errors throughout the document (e.g. malformed citation on line 71; “is” instead of “are” on line 196; “effects” instead of “affects” on line 195; missing space on line 74; incorrectly formatted citations on lines 86, 136, 138, 155, 207…). I would recommend that the authors review the document thoroughly to eliminate such errors.
Finally, I would suggest that the authors aim to make their assessment more quantitative. For example, section 4.2.1 compares different lubrication oil parameters but provides almost no quantitative assessment; almost all statements refer to high, low, small, large. It would be of great benefit to the reader to understand, for example, what the average EF per flight meter was across the 20 flights in the baseline case, and the percentage reduction (or increase) which was achieved when testing different factors. Such quantitative assessments enable subsequent researchers to compare their own models directly, and therefore provide outsized value in terms of advancing the field. If the authors could therefore augment their analysis with more quantitative evaluation (not just in 4.2.1 but throughout) I believe it would improve the utility of the manuscript for the community.