the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 25 Sep 2025
On the Weather Impact of Contrails: New Insights from Coupled ICON–CoCiP Simulations
Abstract. Contrail forecasts typically neglect feedbacks with the atmosphere. Here, we investigate the contrail-weather interaction using a two-way coupling of the Contrail Cirrus Prediction model (CoCiP) with the global non-hydrostatic numerical weather model ICON. ICON includes a new two-moment cloud ice microphysics scheme that enables skillful predictions of ice supersaturation, validated against radiosonde observations and compared with ECMWF forecasts. The CoCiP model uses a new method to limit the uptake of ambient ice supersaturation when many contrails form. Radiative effects of contrails are calculated using the ecRad radiation scheme within ICON. The models are coupled using the YAC coupler to exchange atmospheric and contrail state variables after each ICON time step. The coupled system results are broadly consistent with offline CoCiP simulations, but captures additional feedbacks. The significance of the computed contrail effects is tested by comparison to numerical noise perturbation or twin experiments of the results of two forecasts differing by small random factors in the initial values. The instantaneous radiative forcing (RF) by the contrails exhibits slightly higher global mean values and a more nonlinear dependence on optical depth than previous standalone CoCiP estimates. Contrails induce a butterfly effect that reduces weather predictability after a few days. Hence, contrails are predictable – but only for a finite period. The global mean forecast simulations reveal short-term atmospheric impacts of contrails, including warming at flight levels, as expected. Effects on surface temperature and precipitation appear regionally random, with negligible global mean values before the butterfly effect dominates the results.
- Preprint
(6704 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- AC1: 'Comment on egusphere-2025-4512, Zenodo address updated', Ulrich Schumann, 08 Oct 2025
-
RC1: 'Comment on egusphere-2025-4512', Anonymous Referee #1, 30 Oct 2025
Comments on "On the Weather Impact of Contrails: New Insights from Coupled ICON–CoCiP Simulations" by Ulrich Schumann and Axel Seifert
This manuscript documents the two way coupling of a contrail plume parameterization to a weather prediction model and simulates a few regional test cases. In general this manuscript is an appropriate and comprehensive treatment. It should be publishable in Atmospheric Chemistry and Physics subject to minor revisions.
It would be nice to highlight some of the major points a bit more clearly. I don't think the 'butterfly effect' analogy really works. To me the two major highlights are that (1) weather forecasts in high traffic regions might be improved by including contrails and (2) the efficacy for contrails on surface warming is low (or undetectable) in a weather context. Note that this does not necessarily carry over to the climate context (e.g. the perturbations to surface temperature are smaller than weather 'noise', but may very well be detectable when averaged over a long period of time.
A series of detailed specific comments and clarifications I would like to see addressed are noted below.
Page 1, L18: Contrails only reduce predictability if a model does not account for them. Better to say something like contrails affect the atmosphere and may increase predictability if their effects are simulated. Can you be specific about the effects (beyond ‘warming the upper troposphere’)
Page 1, L20: can you be quantitative about the effects?
Page 1, L21: Abstract. The ‘butterfly effect’ refers to chaotic systems with propagation of small scales perturbations up to larger scales. “Negligible global mean” values before the butterfly effect dominates is just saying that you cannot predict a daily global mean of contrail Ts or Precipitation signal given the noise in your methods. If you were able to reduce the noise (e.g. with a long simulation) you probably could determine regional mean results. It just means you cannot give global means given your configuration and ability to run the model.
Page 2, L58: The principle of the butterfly effect would imply that any perturbation should change the state of the atmosphere over time. Maybe better to say looking for systematic effects?
Page 3, L78: BOA —> SRF (surface)
Page 4, L95: Variance of what? The variance of temperature is just the seasonal cycle…
Page 5, L123: What about cloud fraction? Is there no cloud fraction in ICON at 26 or 13km? If not, is that realistic for contrail simulation to assume uniform properties in a grid box of that size?
Page 5, L145: So there is a cloud fraction in ICON. How does that interface with CoCiP? That should be explained somewhere….
Page 6, L149: but sub-visible Cirrus are real. So why would this be a non-linear bias? Shouldn’t these clouds be included?
Page 7, L183: what is the ICON time-step for the cloud physics?
Page 7, L184: I assume subscripts 1 and 2 correspond to contrails and no-contrails?
Page 7, L209: is this a double call to the radiation in the same simulation?
Page 15, L363: only one line is visible in Figure 5d: suggest they need to be more differentiated from the other data in the plot panel.
Page 18, L420: what is the cause of the non-zero radiative interaction? Does it affect the results for large Tau?
Page 19, L434: please remind the reader of what the change is in a sentence here. What findings? The difference in particle scattering?
Page 24, L495: I see high RF (red) scattered all over the maps in Figure 11.
Page 24, L497: what is the noise level at this time? Might be good to plot the ‘noise’ as a function of time. (E.g. regions without air traffic, or with no contrails in the CoCiP RF areas.
Page 25, L530: If you area average figure 13 differences (or raw values of precip and then take a difference) do. You see any change in averaged precipitation over space (and time).
Page 29, L591: How do you define significance and can you show it on Figure 15? It is not obvious if the variability is large and the blue and red are not statistically different. I think you are right, but define it a bit better please.
Page 31, L613: Does ICON have a gravity wave drag scheme, or are you talking about resolved scale temperature perturbations. If large scale, how do you define gravity waves? Just a temperature perturbation? Or that it propagates (horizontally and/or vertically) as a ‘difference’ between two simulations? I’m not really sure that is a gravity wave, but it is plausible. Maybe a bit more explanation.
Page 31, L621: what is DKE? ∆KE?
Page 36, L731: Figure 15 shows no significant surface warming, and in Figure 16 it is not significant (only the twin experiment).
Page 36, L751: can you comment on what effects you think dominates in the simulations? I’m not sure 8 is relevant: any small forcing over time will probably show a small difference: time can overcome the heat capacity.
Page 36, L754: But it shows the efficacy of contrail RF on the surface is low, and due to the other effects above, might actually not be significant. That’s actually important I would say, and deserves more investigation.
Citation: https://doi.org/10.5194/egusphere-2025-4512-RC1 - AC2: 'Reply on RC1', Ulrich Schumann, 12 Nov 2025
-
RC2: 'Comment on egusphere-2025-4512', Anonymous Referee #2, 31 Oct 2025
Review of: “On the Weather Impact of Contrails: New Insights from Coupled ICON–CoCiP Simulations”
By: Ulrich Schumann and Axel Seifert
Recommendation: Minor Revision
Overview:
The manuscript reports the method used that allows a two-way interaction between a contrail model (CoCiP) and a numerical weather prediction model (ICON). The tests conducted demonstrate that the ICON model can generate realistic ice supersaturation without adjustment while the CoCiP model is able to predict contrail properties consistent with observations and reasonable radiative forcing. The feedback of contrails to weather is further explored. While a warming at flight altitudes is found, the changes of surface temperature, precipitation, and kinetic energy do not appear to be systematic.
The manuscript presents a detailed documentation of the method, with model tests that address the performance of each component. The exploration on the possible weather impact of contrails is also new and insightful. My only concern would be the interpretation of certain figures or results related to the comparison to the twin-experiment being a bit unclear, which may require more explanation or clarification.
Main comments:
- The interpretation related to the twin ICON runs is a bit unclear. From Section 1 (lines 64-67) and 2 (lines 103-107), the role of the twin ICON runs is to provide a reference for the effect of numerical noise and serve as a basis to determine if the changes due to inclusion of contrail feedback are significant (i.e. the changes are significant if they are systematically different from the changes due to initial noise perturbation). However, in the latter discussion in Section 4, the wording in certain sentences or paragraph appears to confuse with this original aim. For instance, in the discussion of Figure 12 (line 518-521), “The lower panel in this figure show that contrails also impact precipitation” is stated and only at the end a comparison to the twin experiment is given to state that “the contrail effects are not much different from the random disturbances…” In my opinion, the presentation for surface temperature in lines 517-518 (“…contrails indeed change the surface temperature… but hardly in a systematic and statistically significant manner.”) is clearer and the readers should be reminded of this fact. For Figure 17 (similarly for Figure 18 as well), while the authors stated that the deviations in both the two-way coupled and twin experiments are amplified by the nonlinearity of the dynamics (lines 645-647), it is not mentioned that the contrail effect on KE or KEw, at least with this model set up and resolution, cannot be distinguished from those caused by random disturbances significantly. This is an important value of the twin experiment, and the readers should be better informed.
- Another related question concerns the almost identical signals in Figure 12 for the left and right panels. The geographical distribution and spatial patterns of the signals are very similar for the experiment with contrails and for the twin experiment. In particular, the signals in regions with low traffic density (Figure 9) or low solar optical depth of contrails (Figure 10), e.g. precipitation over the southern part of the Indian Ocean (30-110E, 30-50S) or southeastern Pacific (near 90W, 40S), remain similar. I am therefore curious if the two-way coupling introduces numerical noise in regions without contrails (e.g. due to the interpolation error of the coupler (YAC)?), or if there is certain random component in the ICON model (e.g. stochastic parametrization or random parameters?) that may explain this similarity? Or if the authors would suggest that the similarity is the result of the contrail effect being amplified in regions that are sensitive to any perturbation, such that large changes show up in similar locations. (Similar to the explanation given for Figure 13, but extending to regions where contrails seems distant.) I think this question may be meaningful for the discussion in this manuscript.
Below are specific comments or technical corrections:
Lines 17-18: Are contrails “inducing” butterfly effect or their effect cannot be distinguished from the typical butterfly effect of initial perturbations?
Line 40: The full name of ICON can be introduced already here, or even earlier in the abstract.
Figure 1: The variable “SRF-rad” should be “BOA-rad” in the figure.
Line 100: “rms” can be introduced here rather than line 260.
Line 114: Is it correct that the fine grid, 2-day simulations are only used for results in Section 3, and the results in Section 4 is only based on the coarse grid, 10-day simulations?
Line 119: Is “τ_cirrus” the same as TACI? It is not found in Figure 1.
Lines 130-154: This part is quite technical, which is great for modellers. The authors may consider summarizing the most important points for general readers and move the details to an appendix. However, this is just a suggestion.
Lines 290-291: Could the authors add a reference to the table or figures that support “the 4-8% higher mean RHi” in the model? I cannot find this number.
Lines 268-270: When compared to IFS, are the radiosonde RHi computed with the ice saturation pressure function used in the IFS (as in line 325?)? It is previously mentioned in line 239 that the ICON formulation is used. Would that affect the score of IFS?
Lines 274-275: Is this conclusion derived from the fact that ICON scores are higher than IFS scores? Since ICON and IFS have many more differences than just the microphysics scheme, is a comparison between ICON simulations with 1-moment and 2-moment microphysics schemes be more suitable to justify this statement?
Line 296: “where” -> “are”/“were”
Line 306: Would “near” instead of “at” the tropopause be more suitable? RHi agrees exactly at around 9 km, but show very large differences above and below.
Figure 4: There is an extra “a)” in x-axis of the first panel. If possible, the three x-axes can use the same limits (up to 1.6)
Line 378: Figure 6 b to d show ageing contrails. Do the authors mean “form” or “persist” in this sentence?
Line 407: Figure 7c is not exactly “corresponding” to 7 a and b (only ICON results are shown)
Lines 417-421: Are the references to 7a and 7b correct in this paragraph? To me, the stepwise change is visible in 7b, but not 7a. Also, when background clouds are excluded, does it correspond to 7a? The terminology of “added cirrus” and “background cirrus” used in the paragraph is not the same as that used in the figure caption (“with or without other cirrus”), which is a bit confusing. The authors may consider using consistent names here.
Lines 476-477: Apart from these signals, there are also localized maximum of the increase in solar optical depth of contrails in Figure 10 panel bottom right in the Eastern US. Is there any explanation for these signals?
Figure 11: For the twin experiment panels, the top right number is “18” instead of “48”. Are they typos or do they actually correspond to forecast hour 18 instead of 48?
Lines 500-505: Related to the main comments, would this “near-random” displacements of clouds be a possible explanation for the similarity between the experiment with contrails and the twin experiment?
Line 520: Are the numbers 0.5 mK and 0.4mm/day the average over the 4 simulations initialized at the 4 different dates? They appears to be inconsistent with the numbers shown in Figure 12.
Line 530: The authors may consider adding a short summary of the major take-home messages for Section 4.1 here. Since numerous aspects are discussed, it may be a good place to remind the readers the major conclusion from this part of the analysis.
Line 547: missing “.” after “error bars”
Line 552: “these two ICON simulations” <- which two are referred to? Combining with the previous paragraph, it seems to be the one-way and two-way simulations. However, the top panel of Figure 14 show the one-way and the twin experiment.
Line 555: Similarly, “two ICON runs” may be better clarified to avoid confusion. (This time should be one-way and two-way simulations.)
Lines 588-589: With reduced humidity, is the age longer or shorter? I am confused by the phrase “only 2% longer”.
Lines 616-617: Could the authors provide further explanation of what exactly are “numerical disturbances near the resolution limit”?
Figure 17: For the DKE twin (long dashed grey lines), are there only 5 days shown instead of 10?
Lines 659-660: As the contrail results is similar to that for the initial disturbances, does it mean that contrails do not significantly foster or slow down the “loss of memory” (or predictability) compared to initial value error?
Lines 731-732: Figure 15 does not show surface warming nor the evolution beyond 5 days. Figure 16 also show no significant change in T2m, but warming at 11km. The conclusion has to be better justified.
Lines 738-750: I assume that this part is suggesting different possible causes of the weak surface warming by referring to other studies. I would therefore suggest changing the first sentence “…the weak surface warming is a consequence of…” to “…the weak surface warming may be explained by…”.
Lines 755-759: This paragraph seemingly lacks evidence in the manuscript. Apart from one sentence in lines 612-613, there is no clear support for the mechanism described here. In particular, the disturbance may be related to numerical methods (lines 771-776). Therefore, I suggest rephrasing this part, with this study providing insights on the possible impact of contrails on dynamics in weather timescales, but requires further investigation to confirm.
Citation: https://doi.org/10.5194/egusphere-2025-4512-RC2 - AC3: 'Reply on RC2', Ulrich Schumann, 12 Nov 2025
- AC4: 'Comment on egusphere-2025-4512', Ulrich Schumann, 12 Nov 2025
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 997 | 64 | 14 | 1,075 | 11 | 7 |
- HTML: 997
- PDF: 64
- XML: 14
- Total: 1,075
- BibTeX: 11
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The reference
Schumann, U.: Note on the CoCiP-libRadtran contrail-cirrus radiative forcing model, 2025.
is now uploaded to a new Zenodo address:
https://zenodo.org/records/17241725