the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 04 Jul 2025
The ice supersaturation biases limiting contrail modelling are structured around extratropical depressions
Abstract. Contrails are ice clouds occasionally formed along aircraft flight tracks, responsible for much of aviation's climate warming impact. Contrails persist in ice-supersaturated regions (ISSRs), but meteorological models often mispredict their occurrence, limiting contrail modelling. This deficiency is often treated by applying local humidity corrections. However, model performance is also affected by synoptic conditions (such as extratropical depressions).
Here, composites of ERA5 reanalysis model around North Atlantic extratropical depressions enable a link between their structure and ISSR modelling. ISSRs are highly structured by these systems: at flight levels, the ERA5 ISSR rate is 91 % less in the dry intrusion – in descending upper-tropospheric air – than above warm conveyors – where air is lifted. The contrast also occurs in composites of in situ aircraft observations, showing the model reproduces the fundamental relationship. However, performance in modelling individual ISSRs also differs across the structures. Compared to the warm conveyor belt, the infrequent ISSRs in the dry intrusion are less well captured by ERA5, with a 20–25 % drop in diagnosing confidence (precision) and 13–19 % drop in comprehensiveness (recall). Scaling humidity beyond ISSR occurrence rate corrections is able to dramatically increase the recall with a small precision cost and high specificity, demonstrating the potential value of scaling approaches designed with different intentions. However, the failure to improve precision, or the performance in the dry intrusion, implies that there is a need to account for the synoptic weather situation and structure in order to improve ISSR forecasts in support of mitigating aviation’s climate impact.
- Preprint
(7236 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-2737', Anonymous Referee #1, 25 Jul 2025
-
RC2: 'Comment on egusphere-2025-2737', Anonymous Referee #2, 28 Aug 2025
Review of egusphere-2025-2737
General Comments
The manuscript reports a novel approach for improving the predictability of ice-supersaturated regions in the upper troposphere. The approach investigates ice-supersaturation in relation to low pressure systems over the North Atlantic and use ERA5 data to create a composite of thousands of lows, with consideration of humidity and wind structures around the composite low. To remove the underlying background signal, a second composite is created for the same position at the low but using synoptic cases that have occurred one year after each composite low. The difference between the two composites produces the modifications due to the low as output signal. The appearance of ISSRs is finally examined in the difference plots, and their relationship with air traffic patterns and IAGOS observational data is studied.
The authors determined score values of ISSR predictability against IAGOS in-situ observations of relative humidity with respect to ice (RHi). The study confirms that ISSR prediction is currently a big challenge. There is indication that lowering the RHi threshold to, e.g. 90% increases recall substantially, at only a minor loss to precision.
The study builds on a novel idea and the presentations of the approach and of the results are well structured and clearly written. The manuscript fits very well into the scope of the journal and deserves publication after a few minor revisions haven been implemented.
The following revisions or adaptions are requested:
1. In the methodology section on the scaling and morphology of ISSRs (lines 189 – 200), the authors describe that their composite approach of diagnosing ISSRs works best for a threshold of RHi > 90%, and this threshold was used in their entire manuscript. There is no further justification added for applying this threshold.
However, there are several observation-based justifications for using a threshold lower than 100% RHi to identify regions of cirrus and contrail-cirrus occurrence:
For contrail-cirrus, Li et al. (2023) showed from an analysis of the ML-Cirrus aircraft campaign, that contrail-cirrus occur most frequently at slight subsaturation with highest probability at 90% RHice (Figures 4 and 7). In the same analysis, PDF of RHice of natural cirrus also has its maximum at slight subsaturation (Figure 7). Similar results were reported from an earlier flight campaign by Kübbeler et al. (2011).
These arguments may help justifying the application of this lower threshold. Furthermore, the statement on line 430 that there is no suggestion that using a 90% RHI threshold produces a more-physically representative population of contrails should be adapted. Indeed, there is observational evidence that such a lower threshold is physically justified.
2. The study presented in the manuscript targets a better predictability of ISSRs in the context of flight rerouting for reducing the climate impact of aviation, see, e.g., abstract and lines 37 – 42. In the current discussion, uncertainty in predicting ISSR is identified as one major bottleneck for contrail mitigation concepts, which is targeted by the study presented in the manuscript. Another more fundamental criticism of contrail avoidance strategies is based on the question how to assess climate impact ‘saved’ by contrail avoidance against climate impact generated by additional CO2 emissions. It might be worthwhile mentioning this conflict of objectives, particularly since a recent study has shown that the increase of surface temperature in K per W m-2 of generated effective radiative forcing, the s-called ‘climate sensitivity’, is smaller for contrail-cirrus than for CO2 (Bickel et al., 2025). Furthermore, the timescales of the climate impacts of contrail-cirrus and CO₂ differ significantly (Borella et al., 2024). A short paragraph on these arguments may be added to the introduction section.
3. The connection between warm conveyor belts and ice clouds has been investigated in-depth by Wernli et al. (2016), who presented climatological distributions of in situ and liquid-origin ice clouds over the North Atlantic. This study may make a substantial contribution to the scientific background of Section 2 on low pressure systems.
Also for section 2, the influence of surface inhomogeneities on the distribution of water vapour an RHi is demonstrated for the regions over eastern North America, the North Atlantic and Europe by Petzold et al. (2020). The annual distribution patterns of water vapour an RHi shown there underpin the selection of an ocean area for the study presented here.
SPECIFIC COMMENTS
1| Line 62: Another very successful approach to adjusting humidity distributions to improve ISSR forecasts has recently been published, see Arriolabengoa et al. (2025).
2| Figures 4 to 7: The height level of the composite plots should be mentioned in the figures or figure captions. The reference to Figure 3 does not help because there, the composites are shown for two height levels.
3| The adequate reference to the performance characteristics of the IAGOS humidity sensor is Neis et al. (2015).
4| In the conclusions, the authors discuss the approach of balancing the under-representation of ice-supersaturation in ERA5 via an over-active saturation adjustment when natural clouds form. Here, the work presented by Arriolabengoa et al. (2025) might also be mentioned.
5| Since IAGOS data are heavily used for the verification of the method, it might be appropriate adding the IAGOS Data Policy statement to the acknowledgements. This statement is available here https://www.iagos.org/data-policy/.
MINOR ISSUES:
Line 400: There is one ‘the’ too much, please remove.
References
Arriolabengoa, S., Crispel, P., Jaron, O., Bouteloup, Y., Vié, B., Li, Y., Petzold, A., and Plu, M.: Modeling and verifying ice supersaturated regions in the ARPEGE model for persistent contrail forecast, EGUsphere, 2025, 1-34, 10.5194/egusphere-2025-1499, 2025.
Bickel, M., Ponater, M., Burkhardt, U., Righi, M., Hendricks, J., and Jöckel, P.: Contrail Cirrus Climate Impact: From Radiative Forcing to Surface Temperature Change, Journal of Climate, 38, 1895-1912, 10.1175/jcli-d-24-0245.1, 2025.
Borella, A., Boucher, O., Shine, K. P., Stettler, M., Tanaka, K., Teoh, R., and Bellouin, N.: The importance of an informed choice of CO2 -equivalence metrics for contrail avoidance, Atmospheric Chemistry and Physics, 24, 9401-9417, 10.5194/acp-24-9401-2024, 2024.
Kübbeler, M., Hildebrandt, M., Meyer, J., Schiller, C., Hamburger, T., Jurkat, T., Minikin, A., Petzold, A., Rautenhaus, M., Schlager, H., Schumann, U., Voigt, C., Spichtinger, P., Gayet, J. F., Gourbeyre, C., and Krämer, M.: Thin and subvisible cirrus and contrails in a subsaturated environment, Atmospheric Chemistry and Physics, 11, 5853-5865, 10.5194/acp-11-5853-2011, 2011.
Li, Y., Mahnke, C., Rohs, S., Bundke, U., Spelten, N., Dekoutsidis, G., Groß, S., Voigt, C., Schumann, U., Petzold, A., and Krämer, M.: Upper-tropospheric slightly ice-subsaturated regions: frequency of occurrence and statistical evidence for the appearance of contrail cirrus, Atmospheric Chemistry and Physics, 23, 2251-2271, https://doi.org/10.5194/acp-23-2251-2023, 2023.
Neis, P., Smit, H. G. J., Rohs, S., Bundke, U., Krämer, M., Spelten, N., Ebert, V., Buchholz, B., Thomas, K., and Petzold, A.: Quality assessment of MOZAIC and IAGOS capacitive hygrometers: Insights from airborne field studies, Tellus Series B-Chemical and Physical Meteorology, 67, 28320, 10.3402/tellusb.v67.28320, 2015.
Petzold, A., Neis, P., Rütimann, M., Rohs, S., Berkes, F., Smit, H. G. J., Krämer, M., Spelten, N., Spichtinger, P., Nédélec, P., and Wahner, A.: Ice-supersaturated air masses in the northern mid-latitudes from regular in situ observations by passenger aircraft: vertical distribution, seasonality and tropospheric fingerprint, Atmospheric Chemistry and Physics, 20, 8157-8179, https://doi.org/10.5194/acp-20-8157-2020, 2020.
Wernli, H., Boettcher, M., Joos, H., Miltenberger, A. K., and Spichtinger, P.: A trajectory-based classification of ERA-Interim ice clouds in the region of the North Atlantic storm track, Geophysical Research Letters, 43, 6657-6664, 10.1002/2016gl068922, 2016.
Citation: https://doi.org/10.5194/egusphere-2025-2737-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 595 | 62 | 15 | 672 | 7 | 16 |
- HTML: 595
- PDF: 62
- XML: 15
- Total: 672
- BibTeX: 7
- EndNote: 16
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Review of
The ice supersaturation biases limiting contrail modelling are structured around extratropical depressions
by OGA Driver, MEJ Stettler and E Gryspeerdt
Overview:
The authors of this paper had a clever idea, that is indeed a novel contribution to the complex problems of contrail prediction. The idea is that contrails and ice supersaturation often appear in certain dynamical regimes, and rarely or not at all in others. Early papers with such findings, like Kästner et al. (1999) and Immler et al. (2008) might be quoted here. The authors consider ice supersaturation in relation to lows over the atlantic and use ERA5 data to make a composite of thousands of lows, considering in particular the humidity structure and the winds around the composite low. In order to enhance the S/N ratio, another clever idea is to make a second composite with synoptic cases that happen one year after each composited low, thus they transform a random selection of weather patterns into a kind of featureless average weather. To extract the modifications that are due to the low, the two composites are substracted and a signal appears.
The appearance of ISSRs is then inspected in the difference plots (the signal), and the relation to air traffic patterns and observational data from IAGOS studied.
IAGOS data later serve as the truth to derive score values of ISSR prediction by comparing the ISSRs in the composites to the IAGOS data mapped to the composites. The result confirms that ISSR prediction is currently a big challenge. There is indication that lowering the RHi threshold to, e.g. 90% increases recall substantially, at only a minor loss to precision.
The paper closes with the statement that forecasting of ISSRs must be improved, and that regarding the structure of the underlying weather could help.
I agree fully with this statement.
The paper is well written and I could not spot any typographical error. I am impressed.
The paper is clearly worth of publication, in particular because of its novel ideas. But I think, a little more work should be invested to make a couple of issues more clear.
Abstract: The message transported in the abstract is not so clear to me. I understand that, if composites of depressions in ERA5 lead to a clear distinction in ISSR occurrence between the dry intrusion on the one hand and the warm conveyor on the other hand. But then, to my surprise, the authors state that the same ERA5 data have problems to capture the very infrequent occurrence of ISSRs in the dry intrusion. This sounds like a contradiction.
L 32: the meaning of the half sentence "insofar as the contrail population..." is not clear to me.
L 47: what do you mean with "coupled to the wider meteorology"?
LL 59-61: Regarding the detrimental effect of saturation adjustment, see also Sperber and Gierens (2024).
Fig. 1: I find the colour bar not very helpful. If it is meant to clearly indicate RHi>100%, a clear change in colour would be good. Furthermore, regions with quite dry air have a similar blue to much moister regions.
L 108: "upper atmosphere" sounds to me as levels close to TOA.
Section 3:
Why is this section called "Methodology" instead of "Methods"?
Section 3.1, first paragraph: I have some questions about the compositing method. First, as a Low has a time-scale of, say, 2 days, it may appear in several subsequent 6-hour periods. How did you treat this? Is a low taken only once for the composite (e.g. in order to avoid multiple counting of the same system), or is every 6-hour slot taken as independent of each other? Second, as the 6-hour snapshots find the lows in different phases of their development, doesn't this smooth the features that you intend to find? Third, the actual pressure on a model level is determined by the surface pressure. As this varies across the system, the surface pressure structure is somehow imprinted on the upper model levels, that is, there is a certain pressure variation as well across the system. How large is it and how do you think it impacts your analyses?
Further questions to methods: ISSRs are frequently connected to cirrus clouds. How is this treated in your method? As you don't mention cloud clearing, I suppose clouds are part of the ISSRs shown and analysed later. If so, are they considered or at least mentioned in the analysis?
L 145: I suggest to delete "empirical". The M&K paper has a lot of theory involved to derive the formula, but, as far as I remember, no new measurements, so the word "empirical" is not appropriate here.
LL 164 ff: Do I understand that correctly: You take the 2009 air-traffic for each day at 5, 11, 17, 23 UTC and the corresponding contrails over the next hour?
Section 3.2, L 187: There are "non-local" approaches that I would like to bring to your attention, for instance by Duda and Minnis (2009a,b) and recently by Wang et al. (2025).
Section 4:
Figure 3: I come back to my question whether you have applied cloud-clearing. I am a bit surprised on the quite strong ISSR frequency at mid-level north of the centre of the low. Is ist possible that we simply see water or mixed-phase clouds there? Inside a water cloud, the water vapour is clearly supersaturated, but this is usually not considered in studies of ice supersaturation. At the higher level, I am suprised to have the signal circling around the low and that nothing is to be seen circling into the anticyclonal direction. However, this is what I would expect from past work on ISSRs in relation to atmospheric dynamics (e.g. Spichtinger et al. 2005, Wilhelm et al. 2022).
Figure 4 and sect. 4.2.1 Have you tried to separate flights in east- and westbound flights? As you explain, the signal southeast of the low is probably mainly due to eastbound flights. Why is there no signal from westbound flights?
Figure 4 and 5: The air traffic and contrail maxima at the western and eastern edges of these figures are explained by traffic close to America and Europe. Now, in a composite there is no real America and Europe because these continents get smeared out by the composition. Your red box in figure 2 extends far into America such that I wonder why there is a structure at all in the counterfactuals of figs 4 and 5. Perhaps an indication of where your 4000km * 4000km box is on average (longitude) would help to clarify this issue.
Figure 6 and sect. 4.2.2: To my opinion, the discussion is incomplete here. It is difficult to believe that a contrail located in the dry intrusion is something like a big hit. Furthermore, the signal in fig. 6 may mainly show noise, since panels a and b show similar forcings. I suggest to use the background (i.e. panel c) to estimate a standard deviation of the signal and to compare that with the size of the signal. You may also try t-tests or similar (if a t-test is inappropriate) to test the Null-hypothesis that there is no statistical difference between 6a and 6c.
Sect. 4.3: Please indicate whether the IAGOS flights are vertically summed up or whether only flight paths at about 260 hPa are selected.
Section 5:
Eq. 4 and adjacent text: The use of specificity (or recall on non-ISSR cases) is not convincing. It has a similar flaw as accuracy has. Assume that I predict always non-ISSR and assume that TN is about 0.9 (this number does not actually matter here). As I always predict non-ISSRs, FP=0, since I never predict a positive outcome. That is, TN/(TN+FP)=1. In Figure 8f) the results are >0.9, so the real prediction of non-ISSRs is not far away from the grotesque prediction of always non-ISSR. I don't believe that this makes sense.
Section 5.2: This is a very interesting section, but it can be improved. I assume that the discussion is exclusively for the composite weather around this average low. But I am not sure because Fig. 8 also shows bars for the counterfactual. Perhaps, these two artificial situations, one with a composite low and one with an average weather without a low, can be separated more clearly here. A question that arises is whether the artifical average weather is something that has anything to do with any real weather that might occur. Another issue with this section is the use of relative quantities. I suppose that ISSRs in dry intrusions are close to impossible, that is, if there are any at all they should be very rare. In this case, ratios result from division by quite small numbers and the result is inflated therefore. I suggest to consider the absolute values as well. Perhaps a large relative error in the dry intrusion is negligible against a small error over the WCB.
Section 6:
LL 396 ff: I am not sure, what you want to say. I understand that 53 and 54% of ISSRs are found in the 50% of the study area where the lows occur. Do you want to say, that accordingly 47 and 46% of ISSRs are in regions whithout a low in the middle? And that therefore lows are not the only cause of ISSRs? It would be nice if that were stated more clearly.
References:
Duda, D., P. Minnis, 2009a: Basic Diagnosis and Prediction of Persistent Contrail Occurrence Using High-Resolution Numerical Weather Analyses/Forecasts and Logistic Regression. Part I: Effects of Random Error. J. Appl. Meteorol. Climatol. 48, 1780-1789.
Duda, D., P. Minnis, 2009b: Basic Diagnosis and Prediction of Persistent Contrail Occurrence Using High-Resolution Numerical Weather Analyses/Forecasts and Logistic Regression. Part II: Evaluation of Sample Models. J. Appl. Meteorol. Climatol. 48, 1790-1802.
Immler, F., etal, 2008: Cirrus, contrails, and ice supersaturated regions in high pressure systems at northern mid latitudes. Atmos. Chem. Phys. 8, 1689–1699.
Kästner, M., etal, 1999: Influence of weather conditions on the distribution of persistent contrails. Meteorol. Appl. 6, 261–271.
Sperber, D., K. Gierens, 2023: Towards a more reliable forecast of ice supersaturation: concept of a one-moment ice-cloud scheme that avoids saturation adjustment. Atmos. Chem. Phys. 23, 15609–15627. doi.org/10.5194/acp-23-15609-2023
Spichtinger, P., etal, 2005: A case study on the formation and evolution of ice supersaturation in the vicinity of a warm conveyor belt's outflow region. Atmos. Chem. Phys., 5, 973-987.
Wang, Z., etal, 2025: Machine learning for improvement of upper-tropospheric relative humidity in ERA5 weather model data. Atmos. Chem. Phys., 25, 2845–2861. doi.org/10.5194/acp-25-2845-2025.
Wilhelm, L., etal, 2022: Meteorological Conditions that Promote Persistent Contrails. Appl. Sci. 2022, 12, 4450. doi: 10.3390/app12094450