Review History


All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.

Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.

View examples of open peer review.

Summary

  • The initial submission of this article was received on November 26th, 2021 and was peer-reviewed by 2 reviewers and the Academic Editor.
  • The Academic Editor made their initial decision on March 1st, 2022.
  • The first revision was submitted on March 30th, 2022 and was reviewed by the Academic Editor.
  • A further revision was submitted on April 7th, 2022 and was reviewed by the Academic Editor.
  • The article was Accepted by the Academic Editor on April 8th, 2022.

Version 0.3 (accepted)

· Apr 8, 2022 · Academic Editor

Accept

Thank you for implementing the suggested modifications. I spotted one typo: "descried" on line 310. Perhaps there will be an opportunity to fix this for the final version.

[# PeerJ Staff Note - this decision was reviewed and approved by Dezene Huber, a PeerJ Section Editor covering this Section #]

Version 0.2

· Apr 4, 2022 · Academic Editor

Minor Revisions

Given the change in focus away from bird migration in environmental science, which I had recommended, I have not attempted to send out the submission for review again (which would almost certainly take much too long and yield little benefit), but instead read through it myself and left a number of recommended wording changes, comments, and questions in the attached PDF. Please view the PDF using Acrobat Reader so that all my annotations are shown as intended. If these remaining small problems are fixed, I will recommend publication.

Version 0.1 (original submission)

· Mar 1, 2022 · Academic Editor

Major Revisions

Please address all the reviewers' comments, particularly, the concerns about transparency and repeatability.

[# PeerJ Staff Note: It is PeerJ policy that additional references suggested during the peer-review process should only be included if the authors are in agreement that they are relevant and useful #]

[# PeerJ Staff Note: Please ensure that all review comments are addressed in a rebuttal letter and any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate. It is a common mistake to address reviewer questions in the rebuttal letter but not in the revised manuscript. If a reviewer raised a question then your readers will probably have the same question so you should ensure that the manuscript can stand alone without the rebuttal letter. Directions on how to prepare a rebuttal letter can be found at: https://peerj.com/benefits/academic-rebuttal-letters/ #]

·

Basic reporting

Thanks for the submission and for sending me the manuscript (MS) by Liu et al. on Imputations of Geese titled “Bird-CGCNImp: a causal graph convolutional network for multivariate time series imputation”
I find the approach is experimental and creative. I am not much against it, but the current MS falls short on a few relevant science points:
-Whatever authors propose and defend, unless we see the data and the code with ISO compliant metadata, this work is not repeatable or transparent. It currently does not meet latest principles of science, or best professional practices. It can easily be fixed though as all data can now fully be made open access and with open source code models with ISO standard documentation (metadata), online, in GBIF and in Movebank even. I stress that specifically because work on these two species and the flyway is already full open access and available with 24 years of remote field work data. See GBIF and then Solovyeva et al. 2021. So why not here ? Not sharing data is not defendable whatsoever and it violates any best professional practices and collegiality. Before that is not clearly done I do not approve of the review or publication.
-Imputation is just one of many methods. Personally, I would use the method of predicted ecological niches instead, or in parallel and to compare. Authors brought it down to a very narrow and self-fulfilling prophecy, just as the endless data filtering to get rid of so-called outliers (as often done in such telemetry works). From the MS and figures, authors seem to believe birds fly on a straight line connecting dots, which is highly dubious. How about krigging, Least Cost paths, or Circuitscape or Marxan approaches instead? One certainly needs a CERTAINTY and CONFIDENCE in the approach, outcome and maps; done how ? Anyways, if authors stick with imputations, please state citations on the issue, namely Jerome Friedman, classification trees and boosting, and many forest inventory and remote sensing references. Much literature and expertise sit there that was not used here. It should.
-The way how to best approach the gap filling should be like a scenario, an hypothesis, or a re-analysis of the existing data. That should be made clear and pursued that way.
-I like Figures 5, 6 and 7 for the concept but those are currently way too small and too selective to be useful or convincing. We need it for all the data, not just some examples.
-let’s agree that the biggest topics - research design, representative sampling of the tagged species, technical fault patterns in the transmission, and impacts of anesthesia - are not mentioned. They must, as those do overrule the data and policy question one wants to purse, e.g. where are the ‘true’ dots, where are the flyways, habitats and do we have broadfront migration or any mixing ? Getting stuck with a line is just not appropriate for a cohort or population inference. The latter is presumably our all policy goal.
-the title and concept of “Bird-CGCNImp” reads odd, and is odd. It’s not needed that way and should be improved. Same is true for ‘causal graph convolutional network’. That’s hardly English, not a term, and nobody speaks that way (in England, or the English speaking/reading world).
-re ‘Identifying bird migration trajectories and discovering habitats is very important for conserving species diversity.’: That’s 100% untrue. Can you show me an example in China, or Malta or anywhere else ? We see flyway declines all over and for many years, see Jiao et al. 2016. The protected areas are way too tiny, paper parks, and do not help. So, this phrase above sets up a strawman; not needed but better to be honest.
-the Literature references are poorly formatted, e.g. first names, abbreviations, and widely incomplete for topics
-Finally, the problem we have in flyways and in China is a ruthless habitat conversion and loss scheme; as it is found globally now. Connecting the bird dots on a line for policy helps little. Happy to be shown any other.

If authors can fix those items, specifically data shared and all data they have at hand, as well as conservation progress, then I would be happy to re-read their update perhaps.
Thanks, best regards
Falk Huettmann

Experimental design

The article has no typical research design

Validity of the findings

As stated in section 1, authors assume linear movements, which is not justified.

Additional comments

Thanks.

I am happy to support creative and experimental work, but the MS - as presented - fails on some basic
science issues, namely transparency and repeatability.

Reviewer 2 ·

Basic reporting

In this paper, the authors propose a causal GCN for bird migration trajectories motivated by the existence of both the attribute correlation and the temporal auto-correlation dependencies. The proposed method is novel and the application of bird migration trajectories shows its effectiveness.

Experimental design

In the proposed Bird-CGCNImp method, the authors establish an end-to-end multi-task model to capture both attribute correlation and temporal auto-correlation dependencies. Besides, the authors also notice the noise in the actual sampling process and design a total variation reconstruction regularization term to improve the imputation accuracy. The authors clearly present the proposed method's motivation and technology details.

Validity of the findings

In the experiment, the authors conduct the experiment on one public time-series imputation benchmark and two real-world bird migration trajectory datasets. The experimental results are promising. The authors also demonstrate the effectiveness of each component in Bird-CGCNImp by ablation study. My suggestion of the experiment is that the authors should repeat the experiment several times and report the mean and standard deviation of the metric so that the experimental results would be more convincing.

Additional comments

Missing citations:

Suo Q, Yao L, Xun G, Sun J, Zhang A. Recurrent imputation for multivariate time series with missing values. In2019 IEEE International Conference on Healthcare Informatics (ICHI) 2019 Jun 10 (pp. 1-3). IEEE.


minor:

Some of the citations are incomplete. For example, the following citations miss the journal name:


Nazabal, A., Olmos, P. M., Ghahramani, Z., and Valera, I. (2020). Handling incomplete heterogeneous
481 data using vaes.

Yoon, J., Zame, W., and Schaar, M. (2017). Multi-directional recurrent neural networks : a novel method
498 for estimating missing data.

All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.