Review of

"Concept of risk-aware contrail avoidance strategies"

by A. Borella et al.

General impression and recommendation:

The authors use a flight planning tool in combination with numerical weather forecasts and a contrail prediction tool to study three concepts of contrail avoidance with increasing awareness of the risk involved. This is the risk of unintendedly damaging climate when the rerouting required for contrail avoidance leads to higher fuel consumption and emissions such that the increase of the CO2 climate impact is larger than the climate benefit by contrail avoidance. The three concepts are: 1) planning a cost-climate optimal flight, that is, the cost-optimal contrail avodoidant route and fly it; 2) as 1), but then use ensemble weather forecasts to estimate the bandwidth of the potential climate impacts and fly the cost optimal route only if the there is a (positive) climate benefit in almost all ensemble members; 3) optimize the route in all ensemble members (i.e., plan N flights), and combine each of these routings with the remaining members of the ensemble, then select from those flights that have low risk the one with the largest predicted climate benefit. 

To my feeling, this is a good strategy of research and the paper is a valuable contribution to the collection of ideas how to deal with the uncertainties caused by the contrail-avoidance vs. fuel consumption trade-off.

In the following I'll give some ideas for further improving the paper. All these are eventually minor comments.

1) My most important comment refers to the description of the "risk-optimal strategy" in Section 2. This description is not good and misleading. The reader gets a full understanding only in Section 6. What I understand from Section 2 is this: For each of N members of an ensemble forecast a climate cost-optimal route is computed. The uncertainty-based risk assessment mentioned in the previous par. ("risk-informed strategy") is then performed on each of the N flight trajectories (if this is correct, please state it clearly). A subgroup M

2) I would also like to suggest to the authors to consider the distribution/clustering of their "big hits". Admittedly, this is not central to this paper, but it is still an important topic that, to my view, has received too little attention in the past. I suggest that the data is screened for big hits, that is, extract the flights with the highest 2% of contrail warming and check where they appear spatially and temporally. I expect some clustering and this should make big contrail prediction easier and with lower risk. It could, however, lead to problems with airspace congestion if a big hit cluster is to be avoided completely. I believe you can here enhance the importance of your paper with little effort.

3) A point for the discussion section is this: The study uses the NAFC as the model region. Do you think that the risk-informed and -optimal methods can be applied to more congested air spaces like Europe. How should the nearby presence of cirrus clouds and other contrails be treated?

4) The expression "to reject any risk" (twice in the Abstract and also later in the text) is surprising. This sounds as you would think of a significance level of zero (that is, probability of an error of first kind is zero). In a system with random elements this is not possible. In aviation this would imply to stay grounded.

5) Line 22: I suggest to replace "contribution" by ERF. Contrails have the largest ERF, but perhaps not the largest contribution (to anything).

6) Line 24: replace "saturation" with "relative humidity". Saturation is 100% and does not exceed it.

7) Section 2, description of the "risk-informed strategy". This was difficult to understand, probably because the additional calculation of the cost-optimal route is mentioned at the beginning. The reader must think that this is the important step, but it is not, I believe. I suggest to present this method as follows: We start as in the risk-unaware method with a climate-cost-optimal routing, then we use the uncertainties to estimate a risk. If the risk is too large, we calculate the usual clost-optimal route and fly it.  

8) Section 3.2, last par: Correction of the humidity field is necessary since NWP models often underestimate RHi in ISSRs. Perhaps you should note explicitly that the problem that ISSRs are often not predicted at their actual position cannot be fixed easily and that it needs more data for assimilation.

9) Section 3.3, around line 207: It might be that the uncertainty is not weather-dependent, however, the efficacy factor itself is probably weather dependent. As the efficacy somehow measures the integrated effect of fast feedbacks over the course of weeks or months, it will certainly vary with the weather. I suggest to clarify this point. Otherwise, to simply assume a constant factor for this study is as a first step certainly in order.  

10) Section 4.1: I was surprised that flight A that exploits the jet stream needs more fuel then flight B that has strong headwinds, whereas both flights are quite similar in distance. Is this perhaps a consequence of the two different aircraft? Is the B767 more efficient than the B777? 

11) Section 4.2, 2nd par: Is there a (tentative) explanation for this surprising finding, namely that flight A is sensitive to wind shear direction, while flight B is sensitive to particle emissions? 

12) Section 4.2, around line 328: It is even worse, because the nominal estimates can be very different from the actual weather.

13) Section 4.2, end: Now it becomes clear what you mean with "zero risk". It can be zero because you use a finite sample. Anyway, I suggest a slight rewording, say "negligible risk" and add in brackets where it is mentioned first (with zero climate damages in our finite sample).

14) Section 6, line 446-447: The sentence is easier to read if the last part "when adopting..." is shifted after "number of rerouted flights". 

15) Sect. 7: I agree that "the estimation of the climate benefit of reroutings must not be reduced to only one deterministic modelling configuration". I am not sure, but isn't this what the MRV-system is just doing to "determine" the contrail impact of single flights? Do you know more and can you comment on this? I think, your statement is an important recommendation.

This paper introduces the concept of "risk-aware" contrail avoidance strategies. In such a strategy, a trajectory is optimized against a contrail forecast, and the resulting climate impact is re-evaluated by ensemble members to create a risk score that captures the likelihood that such a re-routing will result in a climate benefit or harm. Additionally, the authors present a “risk-optimized” approach, where multiple trajectories are created based on the differing ensemble members, allowing an operator to select the resulting trajectory that achieves the desired climate benefit at an appropriate risk level. The contributions in this manuscript are novel, appear scientifically and technically sound. An exploration of robust optimization strategies for contrail avoidance is currently limited in the literature on contrail mitigation, and thus I see this paper as a valuable and timely piece of work. In addition, the paper is very well written and organized. I recommend the paper be accepted, and only have very minor comments as noted below.

• The measure of risk used in this work is related to the fraction of ensemble forecast members that, when used as input to a contrail model, cause climate harm rather than benefit. The authors attempt to also capture model uncertainty by also adjusting model parameters and measuring how frequently these adjustments result in climate damage. While this presents a valid metric, the authors do not present any evidence that this risk score is well calibrated. For example, seems like it should be possible for no forecast ensembles to predict a climate harm, but for a reanalysis to later show a climate harm.  Calibrating this risk score ultimately seems like a very difficult task, because it is difficult for us to directly measure the climate impact of an individual contrail, and ultimately, the risk should be calibrated against observations rather than reanalysis products and models.  For this reason, I find it acceptable for such a calibration to not be within the scope of this study. However, I suggest that the authors place a discussion of this important limitation within either Section 1 or 2 of this paper.

• On lines 315-316, the authors state that, for Flight B, the climate impact is highly sensitive to the parameter controlling the enhancement of nvPM emissions, but the same is not true for Flight A. The authors should report the engine types assumed for each of the two flights. It should be noted that recent work has shown that CoCiP is less sensitive to nvPM emission indices when accounting for the activation of vPM emissions, especially in newer lean-burn engines. Further, more recent versions of pyContrails have models of vPM activation as experimental parameters.  I do not see it necessary for the authors to re-run their simulations with these experimental parameters, but I do recommend the authors include a reference to Ponsonby et al (https://acp.copernicus.org/articles/25/18617/2025/), and include a comment, possibly in Section 3.3, noting the above and that research in this space is evolving.

• In Section 6, the authors explore the concept of risk-optimized avoidance, where flights are optimized individually against a number of ensemble members. This concept is introduced in Section 2, where is it compared to by Simorgh et al, where risk is directly incorporated into the objective function of the optimization process. The authors’ approach to risk optimization is valid and has the advantage of being far easier to implement. However, missing from both Section 2 and Section 6 is a comment on the limitation of the author’s alternative approach. Namely, that by jointly optimizing against multiple ensemble members simultaneously, it may be possible to construct trajectories that achieve lower risk scores with similar operational costs. It is not clear a priori what advantage such a scheme would achieve, if any, and so this is topic that would warrant further research in a future study.

The following comment is entirely editorial. The authors may consider this comment in the revision of this work at their discretion:

• Section 5 of your paper shows that the majority of contrail warming may be avoided through trajectory optimization at a relatively low risk level. This is somewhat a corollary of the main results of this paper. I have a concern that the presentation of the current paper may lead some readers to reach a different conclusion. Specifically, in Section 4, the authors show an example of a flight with a relatively high risk level. Based on the statistics in Section 5, it appears that this higher risk flight is somewhat of an outlier. These examples are still highly useful to illustrate the merit of the authors proposed risk mitigation strategy.  To alleviate this concern, in Section 4, the authors may consider quantifying how likely it is to encounter flights like Flight A and Flight B. Further, the authors may consider switching the order of Sections 4 and 5, which would help provide more context to the reader for how typical Flights A and B are.