Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

 

Comments:

 

The motivation for the paper needs to be more clearly stated. The introduction should explain the importance of the problem, why the detection of anomalies in ocean-drifting vessels is crucial, and how the proposed approach addresses a gap in existing research.

 

The conclusion should clearly state the research gap and clarify why this topic is a novel contribution to the field. The conclusion should also summarize the main achievements of the research, highlighting improvements in performance, methodological innovation, and the practical significance of the proposed framework.

 

All parameters used in the experiment must be clearly presented, preferably in tabular form for readability and reproducibility.

 

Additionally, please include training loss curves for the models. These metrics are crucial for understanding the model's convergence behavior and overfitting potential.

 

Finally, it's necessary to clearly explain the rationale for using a 50/50 split between training and test data. Often, other ratios like 70/30 or 80/20 are more common, so explaining the rationale for choosing a balanced split will reinforce methodological clarity.

Author Response

please see the enclosed file

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

-In the manuscript, the authors compares CNN and LSTM, but the design choices like window size = 10, number of layers, unit counts, training regime require deeper justification.
Also some issues like; the CNN architecture is described generally, but not enough architectural details (kernel sizes, padding choices, receptive field sizes) are provided to allow reproducibility. The LSTM model’s performance advantage in certain regions is described, but the manuscript does not analyze why temporal dependencies vary geographically, nor why specific anomalies benefit LSTM over CNN.

 

-Although Section 2 provides a comprehensive discussion of related methods, the manuscript should explain the novelty of the proposed pipeline in comparison to existing trajectory anomaly-detection approaches. Currently, the proposed method appears to be a reasonable integration of known techniques, but the manuscript does not clearly highlight what is fundamentally new about the approach, aside from combining multiple components. A clear novelty statement is needed.

 

-The bootstrap analysis shown in Figure 7 is interesting but lacks key details like; why was a block size of 2 chosen? What assumptions justify the use of block shuffling? Also, Bootstrap distributions should be tied back to sample size; the sample of only 10 experiments may be insufficient for stable inference. More explanation is needed to ensure statistical validity.

 

Author Response

please see the enclosed file

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The author responded to all questions

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have made significant changes in the revised manuscript. I have not furthur concerns.