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.
The reviewer has no further comments.
[# PeerJ Staff Note - this decision was reviewed and approved by Xiangjie Kong, a PeerJ Section Editor covering this Section #]
After reviewing the manuscript and the response letter, I believe the author has responded well to my feedback. I have no additional comments. Thank you.
The experimental design is appropriate and clearly described. No further concerns.
The findings appear valid and sufficiently supported. No additional comments.
**PeerJ Staff Note:** Please ensure that all review, editorial, and staff comments are addressed in a response letter and that any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.
I have reviewed the manuscript and the response letter, and I believe the author has made good revisions to my remarks. I currently have no further comments. Thank you.
-
-
The author has corrected most of my comments and suggestions. I appreciate the team's efforts. However, you should check your English spelling.
- Spelling error line 312: "5 Disscussion", should be corrected to “5 Discussion”
- Section 3.3 BO-Prophet Prediction Model line 174
The formulas are not yet numbered. In addition, three formulas in this section contain parameters that are not clearly explained. For example, the error term εt\varepsilon_tεt in the Prophet formula is not described. Likewise, subsequent formulas lack explanations for parameters such as a(t), δ, and γ.
- In the section on LSTM neural networks from lines 240 to 245. The formulas numbered from 1 to 6 still lack explanations: Ct, Wc, and bc
- The section: Acknowledgement has the sentence “thank AI (DeepSeek) for its assistance in English translation work” → may cause debate about the journal's acceptance level.
The experimental design is adequate, and the authors have incorporated the additional information as requested.
The validity of the findings is also satisfactory.
**PeerJ Staff Note:** Please ensure that all review, editorial, and staff comments are addressed in a response letter and that any edits or clarifications mentioned in the letter are also inserted into the revised manuscript where appropriate.
**PeerJ Staff Note:** It is PeerJ policy that additional references suggested during the peer-review process should only be included if the authors agree that they are relevant and useful.
**Language Note:** When you prepare your next revision, please either (i) have a colleague who is proficient in English and familiar with the subject matter review your manuscript, or (ii) contact a professional editing service to review your manuscript. PeerJ can provide language editing services - you can contact us at [email protected] for pricing (be sure to provide your manuscript number and title). – PeerJ Staff
Dear Editors & Authors,
I am really pleased to review this manuscript submitted to PeerJ Computer Science. In this study, the author proposed a workflow for InSAR monitoring of deformation in mining areas and applying Prophet models for prediction. The experimental results are relatively reliable, and this work also has certain engineering significance. I found some issues while reading the manuscript, and the following remarks can be used as references for the author to revise the manuscript.
1. Abstract, Line 13-14: This sentence seems to have not been finished yet, and it does not fully correspond to the following sentence about the work of this study. The limitations of traditional surface deformation monitoring methods have been mentioned here, and it is necessary to incorporate the limitations of traditional deformation prediction methods.
2. Section 3.2, Line 115-125: In this paragraph, the author briefly introduces the deformation inversion process of SBAS-InSAR technology. It would be helpful if the author could indicate which software environment was used in this study, such as SARScape or GAMMA.
3. Section 3.3-3.4: Here I suggest briefly mentioning the programming environment and which important packages or libraries were used.
4. Section 4.2, Line 234-242: Much of the content about SBAS-InSAR settings in this paragraph is repetitive with section 4.1, and it is recommended that the author simplify it.
5. Section 5.2, Line 279-289: Is there only one layer of the LSTM model used for comparison in this study? This usually does not fully leverage the advantages of deep learning models. It is recommended to include the number of layers in parameter retrieval. Generally speaking, stacking 1-3 layers in RNN is sufficient for comparison. In addition, since stacking RNNs can result in a longer time required to determine the hidden_size of different layers, it is recommended to determine a value from powers of 2 such as (16, 32, 64, 128); The epoch is also similar, as there may be fluctuations in training, so epochs that are too close may not necessarily be easy to determine the optimization of the model.
6. Section 5.3.1: Here, I would prefer to see not only the comparison between the predicted results of the test set and the monitoring results within the monitoring time range but also preferably the display of future extrapolation results.
7. Section 5.1, Line 266: Please check the cross-references here.
8. In the acknowledgments, the author should not only thank GenAI for its translation assistance but at least the storage and distribution structure of remote sensing data sources should be mentioned.
Best wishes,
- The manuscript focuses on the monitoring and forecasting of ground subsidence caused by coal mining, through the integration of SBAS-InSAR technology and machine learning models: BO-Prophet and LSTM. The study area is the Yinying Coal Mine in Shanxi Province, China – a region strongly affected by mining-induced subsidence and associated environmental risks.
- The article is clearly presented, using professional English throughout most of the process.
- In the introduction of the paper, the Literature references do not fully present the research related to land subsidence prediction. Machine learning or deep learning models are just listed without specific analysis or comparison. The authors should synthesize and analyze more clearly to highlight the choice for their research.
- Line 13–14: “observation, This study develops...” should be corrected to: “observation. This study develops...”
- Line 26: “...monitoring results.Correlation analysis...”
Missing space after the period → Should be: “...monitoring results. Correlation analysis...”
- Check and correct the error: “Error! Reference source not found”
- Figures 8 and 12: They should include a coordinate grid overlay for better spatial reference.
The manuscript presents original research that fits within the scope of the journal. The research question is clearly stated and addresses a relevant gap by integrating spatial error analysis into machine learning-based subsidence prediction. The investigation was conducted to a high technical standard, combining SBAS-InSAR processing with BO-Prophet optimization. Although the methodology is described in sufficient detail, some comments are worth clarifying below:
- While the SBAS-InSAR technique is well-established for monitoring ground deformation, the manuscript does not explicitly justify its selection over other InSAR time-series methods such as PSInSAR or hybrid approaches. A brief comparison highlighting the advantages of SBAS for the specific mining environment (e.g., distributed scatterers, rapid deformation, vegetation coverage) would enhance the methodological rationale.
- The manuscript also does not provide any clear explanation as to why no ground-truth data (such as GPS, leveling, or field benchmarks) were used to validate the SBAS-derived deformation results. Instead, RADARSAT-2 imagery was used as the sole independent validation dataset. However, only field-based ground truth data can truly validate the accuracy of SBAS results, particularly when the goal is to forecast deformation. It would be more convincing to verify SBAS accuracy using field measurements before selecting subsidence points for modeling.
- I recommend the authors read the following study, which demonstrates the use of ground-based data for validation of mine subsidence monitoring: http://www.potopk.com.pl/Full_text/2020_n1_v1_full/IM%202-2020-v1-a20.pdf
**PeerJ Staff Note:** The PeerJ's policy is that any additional references suggested during peer review should only be included if the authors find them relevant and useful.
- Regarding the forecasting model, why were only six points (P1–P6) selected for the 7-step forecast? This requires a clearer explanation, including the geotechnical or geological characteristics of the selected points (e.g., soil type, slope, rock conditions). Restricting the analysis to only six points limits the representativeness of the entire dataset, especially given that SBAS provides a deformation time series for hundreds or thousands of coherent points across the area.
- If the authors had used more points for modeling, this would have enabled spatially continuous forecasting and potentially the generation of a forecast subsidence map, rather than point-based predictions.
- In addition, the paper lacks consistency in how it reports the number of forecasting points.
In Lines 279–280, it is mentioned that 10 high-coherence points with 27-time steps were used, yet in Section 5.3.1, only six points (P1–P6) are used and analyzed. This discrepancy should be resolved and clarified.
- The rationale for selecting Area B as the target forecasting region is also not fully convincing. According to Figure 6, Area D actually shows the highest subsidence magnitude (up to −121 mm), whereas Area B displays moderate deformation. A more detailed justification is needed, potentially including subsidence characteristics (e.g., deformation rate, geology) across all areas, to support the selection of Area B.
- The manuscript presents a technical integration of SBAS-InSAR with BO-Prophet for mine subsidence prediction. While this review does not formally assess the impact or novelty, the inclusion of spatial error analysis in deformation prediction represents a significant methodological contribution. The use of multi-source satellite imagery (Sentinel-1A and RADARSAT-2) and standard evaluation metrics (MAE, RMSE, R²) support the statistical robustness of the results. However, a more explicit presentation of choices such as the number of points included in the prediction, the number of training steps, and the number of testing steps would have made the paper more meaningful.
- The data and code are fully provided. Overall, the conclusions are clearly presented, remain closely linked to the original research question, and are appropriately limited to supporting results.
Line 15-16: spacing problem in BO Prophet(Bayesian Optimization-Prophet) models
Line 134: spacing problem in unique characteristics(Chang et al. 2025; Chérif et al. 2023)
Line 258:There is no need for spacing in the unit (121 mm)
• What gap did you identify in the literature, and how does it relate to your motivation or the key finding that led to the selection of the SBAS-InSAR BO-Prophet model?
• Why not use Sentinel-1B or descending Sentinel-1A? The rationale for data combination is lacking.
• There is no mention of coherence threshold selection, decorrelation handling, or atmospheric correction.
• Do you find it the best optimization method?
• Not much discussion on hyperparameter optimization.
• Study is done on only one site (Yinying mine). No cross-site validation or suggestion on broader applicability is discussed.
• It remains unclear whether the model's performance is specific to the Yinying mining area or if the approach generalizes well to other geological contexts.
• No cross-validation against independent ground truth (like GNSS leveling data) is provided, limiting the confidence in claimed prediction accuracy
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.