Linear Contrails Detection, Tracking and Matching with Aircraft Using Geostationary Satellite and Air Traffic Data
Round 1
Reviewer 1 Report
See the two documents in the zip file
Comments for author File: Comments.zip
Author Response
We thank the reviewers for careful and thorough reading of this manuscript and for encouraging us to refine this paper. We have made edits in response to reviewers’ comments that we believe have improved the paper. This document summarizes our response and revisions. Any reference numbers within this document refer to the reference list at the end of this document, not the reference list in the paper. In addition to this response letter, a document showing the additions and removals from the original document have been included.
Please find attached both documents combined as one pdf file.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Review of the article “Condansation trails detection, tracking and matching with aircraft using geostationary satellite and air traffic data”
The manuscript describes a linear contrail detection method based on satellite and wind retrievals. It provides a clear description and an example of some of the cases, complications, and limitations of the technique. In general the method is clearly described with a logical structure, despite some the written style needing some attention.
It would be important to adapt the title to acknowledge that it is only linear contrails that can be tracked with this methodology. In this respect, we would also ask the authors to specify the conditions at which an evolving contrail stops being identified by the study and for how much longer human visual detection would still identify it as a contrail.
Can you please explain in terms of the methodology used, what does your study do that ACTA does not?
It would be extremely important for the manuscript to provide an estimate of the aereal cover, ideally for a period of at least a month in order to allow contrasting this value with previous satellite estimates. The authors mention the difficulties in finding data to validate their technique. This is a very important aspect in terms of the study’s relevance and applicability, and given that the CoCiP model was used in the algorithm’s training, I would strongly suggest to apply the model to the days and cases treated in the study. This would also provide an estimate of which portion of the contrail’s lifetime corresponds to its linear stage. It is also strongly recommended to improve the contrail formation conditions by applying the Schmidt Appleman criterion instead of a temperature threshold.
Ln 77 delete “s” from “corresponds”
Ln 85 replace “are” for “is”
Ln 86 replace “assess from” by “account for”
Ln 153 and 154 replace “proposes’ by “produces”
Ln 168 replace “These” by “this”
199 replace “provides” by “provide”
201 replace “cirruses” by “cirrus”
221 define COCO
273 replace “have” by “ that has”
278 replace “use” by “used”
Fig 6, the colours are impossible to see on the labels, please use thick lines for the advected traces and thinner for the flight track. Please also avoid capital letters for no reason.
308 please explain the reason to choose the -25 degrees C threshold and why you don’t use the Schmidt-Appleman criteria.
It is not clear why the parallel/or not criterion is necessary if wind direction is taken into account, which would definitely result in most cases in a non-parallel situation. Can you please explain why this is the case?
368 Here it would be most reasonable to assume that using the SAC criteria would largely improve the likelihood of identifying the correct altitude and wind speed, please provide results about this approach, otherwise a blind guess based on distance seems to be extremely poor for heavy trafficked regions, and would give the reader that the study’s approach is missing a very basic physical proven approach which has been validated in various studies previously. Please also include the temperature differences amongst ECMWF’s data and assess the differences with respect to the SAC temperature for each altitude.
467 Has MILP been defined before?
513 add “one” only valid…
Fig 14 The legend colours seem to be wrong.
585 replace “as long” for “as well”
I believe that the specific corrections suggested are enough in terms of readability, but a quick edit by a native speaker would not hurt.
Author Response
We thank the reviewers for careful and thorough reading of this manuscript and for encouraging us to refine this paper. We have made edits in response to reviewers’ comments that we believe have improved the paper. This document summarizes our response and revisions. Any reference numbers within this document refer to the reference list at the end of this document, not the reference list in the paper. In addition to this response letter, a document showing the additions and removals from the original document have been included.
Please find attached both documents combined as one pdf file.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Dear Authors,
I appreciate your work during the revision stage so far. Thanks for detailed replies and I believe several section are clearer.
Your reply explains over three pages how the detection algorithm works. This is quite interesting and must also be part of the paper.
You mention "The main interest of the study is on the tracking and matching methodology. The detection method was not the aspect we spent the more time on, and we believe that this detection method could be largely improved by using a better dataset. It is however important to note that the purpose of this paper was not to become the new state of the art in contrail detection. It is rather to demonstrate that the use of a detection model, even if imperfect, allows us to detect and track a certain number of contrails for which we will be able to identify the aircraft responsible for their formation and to track them during the portion of their lifetime where they maintain a recognizable shape according to our detection model.
"
I think it is not sufficient to state that you focused on the tracking. The detection is equally important. You write that the detection could be improved in the future. In this case it is even more important, to describe this part in some detail. In my opinion the text additions of the current revision are not sufficient. The original version didn't explain at all, that your algorithm is limited to line-shaped objects. Hence, there is the clear need to explain the detection in much more detail.
A second issue:
In the reply to Reviewer 2, you mix up basic contrail facts.
"We would like to thank the reviewer for the suggestion, and we believe that in a situation where weather forecast provides reliable data on the relative humidity field, this method would indeed provide a much better filtering than the temperature filtering we used. We even used this filter at the first steps of the study. However, as highlighted in Gierens et al [GMR20] and Agarwal et al [AME+22], relative humidity from ERA5 and other reanalysis products sometimes predict supersaturation areas wrong. Using Schmidt Appleman Criterion as a way to filter the flights that might have produced the contrails could mean that we incorrectly remove some candidates which are eventually the right match. "
The formation of a contrail depends much more on temperature than on relative humidty and is well-predictable. Only persistence (as it depends on RHi) is not easily predictable.
You find evidence, that NWP T and RHi do not predict a contrail in cases where you detect one. Hence, you question the NWP data quality(, which is okay). How safe is the wind prediction in such a case? Can you rely on it?
I believe the language quality is very good.
Author Response
We would like to thank once again the reviewer for his helpful comments which we believe helped improve significantly the quality of the paper.
Please find attached the document describing all the modifications we made for round 2.
Author Response File: Author Response.pdf
Reviewer 2 Report
The authors have thoroughly addressed the previous review. I believe the article should be published in its current version.
In the future, the authors might like to seek help from a native speaker.
Author Response
We would like to thank once again the reviewer for his helpful comments that helped increase significantly the quality of the paper. We also thank you for recommending the paper for publication.
The other reviewer requested a few last changes, about the description of the detection method. Please find the resulting modifications attached.
Author Response File: Author Response.pdf
Round 3
Reviewer 1 Report
I think the manuscript has benefitted a lot from including the new parts on the detection. And you did a good job of describing how CoCip is applied to generate a training data set. Thats exactly what I wanted to have included in the manuscript.
I understand that the DeepLearning approach is a bit like a black blox and not much can be said on that.
I recommend publication and I am looking forward to new results and developments from the author group in the near future!
I spotted several typos. I strongly recommend that all authors carefully read the text again.
In the manuscript with highlighted changes:
line number 220: "are added"
240: "there are"
264: "that", not "which" (https://www.dictionary.com/e/that-vs-which/)
274, 324, ...
I am sure there are more occasions.
Kind regards
Quality of language is good. But typos and some sloppy sentences still exist (see comments above)
