Windthrow Mapping with Sentinel-2 and PlanetScope in Triglav National Park: A Regional Case Study

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study presents a well-structured and timely investigation into the use of multi-temporal high-resolution satellite imagery (Sentinel-2 and PlanetScope) for detecting windthrow damage and estimating timber volume in alpine forests. The research is relevant, methodologically sound, and clearly presented. However, several aspects need to  be  clarified, elaborated, revisied.

1. Lines 35-36 The results were robust, with deviations in the calculated timber volume ranging from 5% to 7%” should be clarified. Please specify these values refer to absolute deviations or relative errors.
2. In the Introduction section, the transition between paragraphs could be made smoother. Consider adding a sentence at the end of the first paragraph (Lines 50–51) to more effectively connect the general problem of forest disturbances to the specific method of remote sensing).
3. Pages 5–6: The descriptions of Sentinel-2 and PlanetScope are clear. However, the pre-processing steps (cloud masking, mosaicking) need to be described in more detail. Which specific algorithms or tools were used?
4. Pages 6–7: The use of multiple datasets is appropriate. However, the forest stand map from 2021 might introduce temporal bias. Although this is acknowledged later, it should also be mentioned here.
5. Pages 7–8: The NDVI-based differencing method is standard and well-explained. However, the manual thresholding approach (Lines 221–227) is subjective. Please justify the chosen threshold range (0.05–0.10) with references or through sensitivity analysis.
6. Lines 221-222: The statement “a threshold between 0.05 and 0.10 was manually defined by visual interpretation” is not sufficiently reproducible. Please provide (i) the exact NDVI-difference histogram, (ii) the final threshold value(s) used for each sensor, and (iii) an inter-operator agreement metric (e.g., Cohen’s κ) if more than one interpreter was involved.
7. Lines 251-256: The 2021 stand-level volume map is systematically biased by 2018-2021 bark-beetle salvage. Quantify this bias by comparing the 2021 map with the 2023 post-storm field sheets for a random subset of 30 parcels. Report the mean % difference and incorporate this uncertainty into the volume-error bars in Fig. 11.
8. Pages 8–9: The methodology is logical, but the workflow in Figure 6 is not fully explained in the text. Please refer to the figure and briefly describe the steps.
9. Lines 274-284: The agreement percentages (77 % S2; 81 % PS) are calculated on a per-pixel basis without considering spatial autocorrelation. Recalculate the 95 % confidence intervals using a block-bootstrap (e.g., 500 iterations with 1 ha blocks) and add the corrected intervals to Table 3.
10. Table 3: The terms “Overestimation” and “Underestimation” should be more clearly defined in the context of false positives and false negatives.
11. Pages 12–13: The comparison between S2, PS, and DOF is well done. However, the large discrepancy between SFS estimation (>60,000 m³) and DOF (80,000 m³) in Table 4 should be discussed.
12. Line 261: “provided from” should be “provided by” or “obtained from”.
13. Lines 364–369: The advantage of PlanetScope over Sentinel-2 is well argued. However, the cost and accessibility of PlanetScope compared to freely available Sentinel-2 could be discussed more comprehensively.
14. Lines 379–384: The temporal mismatch and resolution issues are appropriately acknowledged. Also, consider discussing the impact of topography on optical imagery in alpine regions. The statement that “coarser-resolution data … may not fully capture small or fragmented storm areas” provides no quantitative estimate. Generate a size-class distribution (0–0.1 ha, 0.1–0.5 ha, >0.5 ha) of omitted gaps and relate omission error to gap size via logistic regression.
15. Line 392: “emphasises” should be “emphasizes”.
16. Lines 466-467, 475-476: The reference list contains duplicates (Refs. 8 & 12). Please merge and update them.
17. Some figures and tables are referenced but not provided in the text (e.g., Figure 3, Figure 5, Figure 6 and Table 2). Please ensure all are included and numbered correctly.
18. 5 x-axis: The label “2025-01” extends beyond the study period.
19. Figure 7 and 8: These are useful for visual comparison. However, the captions should be more descriptive. For example, specify what the bars in Figure 7 represent.
20. All units of measurement in the text should be in superscript. Please check them all.

Comments on the Quality of English Language

The language is acceptable but requires moderate editing to meet the requirements. I recommend professional proofreading or language editing to address grammatical inaccuracies, improve sentence structure, and ensure terminological consistency in oder to enhance the clarity, readability.

Author Response

Dear Editor and Reviewer,
We thank the reviewer for the careful and constructive evaluation of our manuscript. We also apologize that the resubmission took nearly the full 10-day window; we used this time to perform substantial additional analyses and revisions. The detailed comments and suggestions have been extremely valuable in improving both the scientific rigor and clarity of the paper. In revising the manuscript, we addressed all points by clarifying methodological details, strengthening justifications with additional analyses (e.g., threshold sensitivity and block-bootstrap procedures), improving figure captions and table content, and correcting minor language and formatting issues.

These revisions have enhanced the reproducibility and transparency of our workflow and improved the readability of the manuscript for a wider audience. We are confident that the paper has gained scientific value and clarity thanks to the reviewer’s thoughtful feedback. In the following, we provide point-by-point responses to all comments, which are also indicated in the revised manuscript as “R1 num.” for ease of reference.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The study examines the detection of windthrow in alpine forests using Sentinel-2 and PlanetScope imagery, combined with timber-volume estimates and official sanitary logging records. While the integration with forestry records is valuable, the technical contribution is minimal. NDVI differencing with a visually chosen threshold is a well-established method and cannot be presented as innovation. The work does not advance beyond approaches already demonstrated in many disturbance mapping studies over the past two decades.

The reliance on a single NDVI threshold, selected by visual inspection, makes the method subjective and not reproducible. No systematic benchmarking is attempted against alternative indices or change metrics, even though Sentinel-2 offers SWIR and red-edge bands that are better suited to disturbance detection. The timing of image selection also undermines the analysis: the first post-event Sentinel-2 image is taken more than a month after the storm, when salvage logging and phenological change confound attribution. Without a careful sensitivity analysis, the reported accuracies cannot be trusted.

Accuracy assessment is incomplete. The paper reports area overlaps and percent differences but omits standard metrics such as precision, recall, F1, or IoU, and does not provide uncertainty bounds. The parcel-level volume comparison is similarly weak, based on outdated stand inventory data and without error propagation. The claim of deviations within 5–7 percent is not substantiated by rigorous statistical analysis.

Figures and tables are serviceable but do not resolve key uncertainties. Thresholds, registration quality, and terrain effects are insufficiently documented. References are uneven and partly redundant. The writing requires careful editing for grammar, consistency, and clarity.

Overall, the manuscript lacks originality, methodological rigor, and reproducibility. The central approach is too simplistic to merit publication in Remote Sensing. I recommend rejection.

Comments on the Quality of English Language

Comments on English: understandable, but in places imprecise and repetitive. A professional language edit would be required if the manuscript were to be considered further.

Author Response

Dear Editor and Reviewer,

Thank you for the detailed and constructive review. We apologize for using nearly the full 10-day response window; this time was devoted to additional analyses and revisions. Your feedback helped us improve methodological rigor, uncertainty analysis, documentation, and clarity. Below we provide a point-by-point response and indicate the associated changes.
Although your recommendation was to reject the manuscript, in the intervening days we have substantially strengthened the study by (i) benchmarking additional spectral indices (e.g., NDRE, NDMI) alongside NDVI, (ii) adding comprehensive statistical evaluation (precision, recall, F1, specificity, IoU, Cohen’s κ), (iii) conducting threshold-sensitivity analyses and block-bootstrap resampling to quantify uncertainty, and (iv) improving figures, tables, and methodological documentation for transparency and reproducibility.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have demonstrated a strong commitment to improving the manuscript, with substantial improvements in methodology, analysis, and presentation. The revisions have enhanced the manuscript's scientific quality, transparency, and relevance. While the authors have addressed most comments thoroughly, two areas could be further strengthened:

1. Comment 7 (Bias quantification):

Although the authors justified the use of only six parcels due to data accessibility, a more detailed discussion on how this limited validation may affect the generalizability of the volume estimates would be beneficial. Future work could explicitly recommend broader access to official forest data for more robust validation.

2. Comment 9 (Block-bootstrap):

While the method was correctly applied, the authors could briefly discuss the implications of the revised confidence intervals on the interpretation of sensor performance, especially in the context of operational forest monitoring.

Author Response

Dear reviewer,

We appreciate your careful evaluation and valuable suggestions. The manuscript has been revised to reflect your feedback, and a point-by-point response to all comments is included in the accompanying file.

Sincerely,

Authors

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The revised manuscript is improved compared to the previous version. The authors have addressed several earlier concerns by making thresholding more objective, adding sensitivity analyses, benchmarking NDVI against alternative indices, and expanding the accuracy assessment with bootstrap confidence intervals. Figures and tables are clearer, and the integration of PlanetScope with official logging records is documented more transparently. However, I still see important issues that need to be resolved before the paper can be considered for publication. 

First, the contribution remains primarily a regional case study. The methods are established and widely used in the literature, and while the workflow is well executed, the novelty is modest. The authors should scale back claims of a “robust operational framework” and clearly frame this as a demonstration of applicability in alpine conditions rather than as methodological innovation.

Second, the validation is limited. The parcel-level comparison is based on only a handful of plots, which does not provide a sufficient statistical basis. The authors need to be explicit about this limitation in both the Results and Discussion, and avoid overinterpreting the 5–7% deviations as definitive evidence of accuracy. At present, the validation reads stronger than it actually is.

Third, the manuscript is still longer than necessary and overloaded with routine analyses presented in full detail. Several tables and figures (for example, the threshold sensitivity curves and large sets of performance metrics) could be streamlined or moved to supplementary materials, with the main text focusing on the most important results. This would improve readability without sacrificing transparency.

Finally, the discussion should more clearly situate the study in the context of existing windthrow research. Many previous studies using high-resolution imagery reported higher accuracies; the authors should acknowledge this and explain more directly what their study adds, beyond being another case example in the Alps.

In its current form, the paper shows technical competence but needs substantial revision in framing, interpretation, and presentation. I recommend major revision, with attention to positioning the work as a regional application study, tightening the manuscript, and tempering claims about generalizability and novelty.

Author Response

Dear Reviewer,

Thank you for your careful and constructive guidance. We have followed your recommendations throughout. We have shortened and reorganised the manuscript into a clearer, more focused form, moving routine materials to the appendix. We have revised the Abstract and Highlights, updated sections of the Results, and refreshed the Discussion and Conclusions with a softer, more cautious tone. We explicitly acknowledge the study’s limitations while emphasising its practical value for forestry operations, which practitioners have indicated is needed. We believe this version aligns closely with your directions. We have also prepared a document that answers all comments point by point.

Sincerely, 
The authors

Author Response File: Author Response.docx