Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This is a valuable contribution to GLOF hazard assessment with important practical implications for infrastructure planning in Tibet. The methodology is generally sound, though the susceptibility assessment could be more rigorously validated. The identification of cascading/clustering effects is a significant finding that advances beyond single-lake assessments. With the recommended revisions, particularly regarding methodological transparency and discussion of limitations, this manuscript would make a strong contribution to the geohazards literature.

Comments for author File: Comments.pdf

Author Response

GeoHazards

Re: revision on manuscript #geohazards-4452106

Dear Editors and Reviewers:

We would like to express our gratitude for the valuable comments and suggestions raised by the reviewer on our manuscript:

Evolution of Glacial Lakes and GLOF Risks to Transportation Routes in the Southeastern Tibetan Engineering Corridor

We sincerely thank the editors and all reviewers for their valuable comments and suggestions. These constructive inputs have been instrumental in significantly improving the quality of our manuscript. In the following, we have provided detailed responses and revisions to each point. We hope that the revised manuscript now meets the expectations of the editors and reviewers. Should any further issues or feedback arise, we would welcome continued guidance on our paper. Once again, we are deeply grateful for your time and valuable insights.

Authors’ response to reviewer’s comments

Reviewer 1:

This manuscript presents a relevant study on glacial lake evolution and GLOF susceptibility assessment in the southeastern Tibetan engineering corridor. The research addresses an important practical concern for critical infrastructure (highways G219/G318, Sichuan-Tibet Railway, and hydropower transmission projects) in a region experiencing rapid climate change. The authors have compiled a robust dataset using multi-temporal satellite imagery and developed a susceptibility assessment framework based on historical GLOF events.

Major Comments

Methodological Justification and Transparency

The susceptibility assessment methodology requires stronger justification. While the authors compare indicators between stable and outburst lakes (Figure 2), the statistical significance of these comparisons is not reported. Please provide:

Statistical test results (e.g., t-test or Mann-Whitney U test) with p-values for each indicator comparison
Justification for the specific weight assignments (V=0.4, Dparent=0.25, Aparent=0.25, Elevation=0.1). How were these weights derived? Were they based on expert judgment, statistical analysis, or a combination?

——We thank the reviewer for pointing out the methodological deficiencies in terms of justification and transparency. We fully accept the reviewer's criticism and have systematically supplemented and improved both issues in the revised manuscript.

(1) Regarding statistical testing: We have performed the Mann–Whitney U non‑parametric test to evaluate the significance of differences between outburst lakes (n = 12) and stable lakes (n = 257) across all indicators. The complete results are summarized in Supplementary Table S2. The test results show that lake volume (U = 1568.0, Z = 4.702, p < 0.001), distance to the parent glacier (U = 2537.0, Z = 2.248, p = 0.025), and parent glacier area (U = 2350.5, Z = 2.713, p = 0.007) differ significantly between the two groups, whereas elevation (p = 0.304), glacier tongue slope (p = 0.344), slope near the lake (p = 0.084), air temperature (p = 0.612), and precipitation (p = 0.302) do not reach statistical significance. These results have been described in detail in Section 2.3 of the revised manuscript.

(2) Regarding weight assignment: The weights of the four indicators were determined based on a comprehensive consideration of statistical discriminability, physical contribution, and literature conventions. In terms of statistical discriminability, the Mann–Whitney U test results provide a quantitative basis for weight allocation—indicators with higher discriminative power have stronger capacity to indicate outburst events and are therefore assigned higher weights. In terms of physical contribution, lake volume directly determines the potential flood magnitude, distance to the parent glacier and parent glacier area control the likelihood and magnitude of ice avalanches entering the lake, while elevation primarily affects the potential energy conversion process of outburst floods and plays a relatively indirect role. Meanwhile, the weight assignments also drew upon literature conventions in regional GLOF susceptibility assessments and were locally adjusted based on the characteristics of historical outburst events in the study area. Based on the above analyses, the weights for lake volume, distance to the parent glacier, parent glacier area, and elevation were set to 0.4, 0.25, 0.25, and 0.1, respectively. To test the robustness of the weight assignments, we have also added a sensitivity analysis (Supplementary Table S6), which shows that when individual weights vary within ±20% of the baseline values, the number of high and very high susceptibility lakes remains stable, indicating that the assessment results are not sensitive to reasonable variations in weight assignments and that the weighting scheme is robust.

The above revisions have been made in Section 2.3, the caption of Figure 2, and Supplementary Tables S2 and S6 of the revised manuscript. We thank the reviewer again for the valuable comments.

The authors acknowledge the difficulty of validating GLOF susceptibility assessments. However, given that 12 historical GLOF events were used to inform the indicator selection, it would strengthen the paper to:

Present a confusion matrix showing how well the assessment criteria classify these 12 historical events
Perform a leave-one-out cross-validation to demonstrate the robustness of the weighting scheme
Explicitly state that the assessment is based on susceptibility (hazard) and is not intended to predict timing

——We thank the reviewer for raising this critical issue. We fully agree that since 12 historical GLOF events were used to inform the indicator selection, the classification performance of the model must be validated. In response to the reviewer's requirements, we have made the following revisions in the manuscript:

(1) Regarding the classification performance of historical events. In Section 4.1 of the revised manuscript, we have supplemented the back‑testing results. After applying the evaluation system to the four historically documented GLOF lakes within the study area, Guangxie Co, Lumu Co, and Anjue Zhuqulong glacier lake were classified as having "very high" susceptibility, and Zhuxigou glacial Lake was classified as having "high" susceptibility. This result supports the validity of our susceptibility modeling. Meanwhile, in the limitations section of the Discussion (Section 4.5), we have explicitly acknowledged this limitation and noted that further validation will be improved with the accumulation of more GLOF records in the future.

(2) Regarding the clarification of the assessment scope. We have clearly stated in multiple sections of the revised manuscript that this assessment is based on susceptibility, aiming to evaluate the intrinsic sensitivity of lakes to outbursts rather than to predict the specific timing of such events. Specifically, in Section 2.3 (Methods), we clarified that the assessment is for "GLOF susceptibility"; in Section 4.1 (Validation), we noted that the back‑testing serves only to validate classification performance; and in Section 4.5 (Limitations), we explicitly stated that this study focuses on susceptibility rather than comprehensive risk or temporal prediction. In addition, to align with the reviewer's suggestion in MC7, we have uniformly revised "risk levels" to "susceptibility levels" throughout the text, and changed "Risks" to "Hazards" in the article title to more accurately reflect the scope of our research.

The above revisions have been made in Sections 2.3, 4.1, and 4.5, as well as in the title of the revised manuscript.

The cascading/clustering amplification effect (Section 4.3) is one of the most novel contributions of this study. However, the analysis is somewhat qualitative. Consider:

Quantifying the amplification effect for each cluster (e.g., volume ratios, peak discharge estimates)
Providing a simple conceptual model or schematic that illustrates the chain-reaction mechanism

——We thank the reviewer for acknowledging the novelty of the cascading/clustering amplification effect. We fully understand the reviewer's expectation for a more quantitative analysis of this effect. However, accurate peak discharge estimation requires detailed hydrodynamic modelling, including breach development processes, dam material parameters, and downstream channel cross‑section data—all of which depend on extensive field investigations and laboratory testing. At the current stage, rough estimates based solely on existing data would carry considerable uncertainty and could even be misleading for hazard prevention and early‑warning decision‑making. Therefore, adhering to a rigorous scientific approach, we have identified quantitative hydrodynamic simulation as a direction for future in‑depth research and have explicitly emphasized this in the Discussion section (Section 4.4) of the revised manuscript.

Regarding the reviewer's suggestion for a conceptual model or schematic diagram, we have provided a schematic conceptual model in Figure 15 of the revised manuscript, which intuitively illustrates the chain‑reaction mechanism of cascading lake clusters—namely, the process by which an upstream lake outburst propagates floodwater downstream, successively triggering failures of additional lakes, and ultimately producing a cumulative amplification effect in flood magnitude.

In addition, regarding methodology, we have added Section 2.4 (Identification of Cascading Lake Clusters) in the revised manuscript, where we define in detail the three criteria for identifying cascading clusters (spatial connectivity, distance threshold, and manual image interpretation), thereby providing a reproducible methodological basis for the cascade analysis rather than leaving it as a purely qualitative description confined to the Discussion.

In summary, through the methodological definition in Section 2.4 and the conceptual model in Figure 15, we have provided clear methodological support and a visual schematic for the qualitative analysis of cascading amplification effects, while positioning quantitative hydrodynamic simulation as a clearly defined direction for future research. We believe this approach addresses the reviewer's concerns while avoiding potentially misleading conclusions that could arise from insufficient data.

The Discussion section would benefit from a more explicit acknowledgment of limitations:

The temporal resolution (1990-2020 with 10-year intervals) may miss interannual variability and rapid changes
The volume estimation formula (Zhang et al., 2023) has inherent uncertainties not discussed
The study focuses on susceptibility rather than risk (which would require exposure and vulnerability assessment)
The possible misclassification of lakes due to seasonal variations in water level

——We thank the reviewer for suggesting that we more explicitly acknowledge the limitations of our study. We fully agree that a candid discussion of the study's shortcomings helps readers to correctly interpret the results. To this end, we have added a new section (Section 4.5, Limitations) in the revised manuscript, in which we respond to each of the four aspects raised by the reviewer. A brief summary is provided below:

First, regarding the temporal resolution, we acknowledge that the 10‑year interval of our lake inventory from 1990 to 2020 may indeed fail to capture interannual variability and short‑term rapid changes, particularly for small lakes that respond quickly to climatic fluctuations. We have explicitly stated this limitation and recommend that future studies employ higher temporal‑resolution remote sensing monitoring to address this issue. Second, regarding the uncertainty in the volume estimation formula, we note that the volume estimation was based on an empirical formula calibrated on 47 Himalayan lakes (Zhang et al., 2023). While this formula provides a reasonable approximation, it inevitably carries uncertainties when applied to lakes with different geometric characteristics, and future bathymetric surveys are needed to refine the estimates. Third, regarding the focus on susceptibility rather than full risk, we clearly state that this study did not incorporate exposure and vulnerability assessments of downstream infrastructure (e.g., bridge/culvert types and conditions, embankment resilience, and traffic exposure), and that future studies integrating these components would contribute to a more comprehensive risk assessment framework. Fourth, regarding the potential misclassification of lakes due to seasonal water‑level fluctuations, we have explained in the limitations that, to minimize this effect, we preferentially selected images from the summer–autumn months (June–October), when lake levels are relatively stable and lake boundaries are most discernible; however, residual seasonal variability cannot be entirely ruled out.

The above content has been fully presented in Section 4.5 of the revised manuscript. We thank the reviewer for the valuable suggestions, which have significantly enhanced the rigor and transparency of our study. The complete revised text is as follows:

“4.5. Limitations

Several limitations of this study should be acknowledged when interpreting the results. First, the temporal resolution of our lake inventory (1990, 2000, 2010, and 2020) may not capture interannual variability or short-term rapid changes in lake area, particularly for small lakes that respond quickly to climatic fluctuations. Future studies with higher temporal resolution would help to resolve this issue. Second, the lake volume estimation was based on an empirical area–volume relationship [60] calibrated on 47 Himalayan lakes. While this formula provides a reasonable approximation, it inevitably carries uncertainties, especially when applied to lakes whose geometric characteristics differ from the calibration dataset. Future bathymetric surveys of representative lakes in the study area would help to refine volume estimates and reduce associated uncertainties. Third, this study focuses on GLOF susceptibility rather than comprehensive risk, the latter of which would require systematic assessments of exposure and vulnerability of downstream infrastructure, including bridge/culvert types and conditions, embankment resilience, and traffic exposure. Future studies integrating these components would contribute to a more complete risk assessment framework. Fourth, seasonal water-level fluctuations may lead to lake misclassification, particularly for small, shallow lakes whose water boundaries differ substantially between dry and wet seasons. To minimize this effect, we preferentially selected imagery from the summer–autumn months (June–October), when lake levels are relatively stable and lake boundaries are most discernible; nonetheless, residual seasonal variability cannot be entirely ruled out. Finally, while the cluster/cascade analysis provides a valuable qualitative assessment of flood amplification effects, quantitative simulation of cascading outburst dynamics would further strengthen hazard prediction capabilities.”

While the study mentions global warming in the introduction, there is limited analysis of the climate drivers of lake expansion. Consider

——We thank the reviewer for this suggestion. We agree that although the Introduction mentioned global warming, the discussion of climatic drivers was indeed insufficient. We have made targeted additions in both the Introduction and the Discussion (Section 4.2), as detailed below:

(1) In the Introduction, we have added background information highlighting that the warming rate of the Tibetan Plateau over recent decades has been approximately twice the global average, emphasizing the southeastern Tibetan engineering corridor as a climatically sensitive region. We now explicitly state that "how glacial lakes in this region respond to this accelerated warming process, along with their spatiotemporal evolution patterns and potential hazard risks, remain poorly understood." This revision positions climate warming as the broad background and research motivation of our study rather than its core content, which we believe is more logically rigorous. The specific revision is as follows:

“It is noteworthy that the warming rate of the Tibetan Plateau over recent decades has been approximately twice the global average, making it one of the most cli-mate‑sensitive regions on Earth [20,21]. In the southeastern Tibetan engineering corridor, this accelerated warming has driven rapid glacier retreat [16]. However, how glacial lakes in this region respond to this accelerated warming process, along with their spatiotemporal evolution patterns and potential hazard risks, remain poorly understood.”

(2) In the Discussion (Section 4.2), we have added the statistical methods and results of the climate trend analysis, using the Mann–Kendall test and Sen's Slope estimator to analyze trends in temperature and precipitation from 1970 to 2022. The results show a highly significant warming trend (Z = 6.65, p < 0.001, Sen's Slope = 0.045 °C/yr), while the precipitation trend does not reach statistical significance (Z = 0.24, p ≥ 0.05). On this basis, we explicitly state that "the observed lake expansion is primarily attributable to regional warming, which accelerates glacier and snow melt and thereby supplies increased meltwater to glacial lakes; no statistically significant contribution from precipitation changes to lake expansion was identified." This conclusion directly addresses the research gap raised in the Introduction, while also clarifying the role of climate analysis in this study—as a background driver of lake evolution rather than the core objective of the paper.

We believe the above revisions effectively address the reviewer's concerns and strengthen the analysis of climatic drivers. The relevant changes have been made in the Introduction and Section 4.2 of the revised manuscript. Should you have any further suggestions, we welcome your continued guidance.

The manuscript would be strengthened by more concrete recommendations for infrastructure protection:

——We thank the reviewer for this suggestion. We agree that more concrete recommendations for infrastructure protection would significantly enhance the applied value of the paper. To this end, in the newly added Section 4.4 (Recommendations for Monitoring and Early Warning), we have specifically proposed practical measures for infrastructure protection. For example, for sections of National Highways G219 and G318 that lie within 10 km of potential outburst pathways (Supplementary Table S3), we recommend reinforcing bridges and culverts along these routes to enhance their resilience against potential flood impacts. Where conditions permit, diversion channels or debris‑flow barriers could be constructed to mitigate the effects of outburst floods. In addition, we recommend coordinating with local transportation authorities to develop targeted emergency response plans. The above revisions have been made in Section 4.4 of the revised manuscript. The complete revised text is as follows:

“4.4. Recommendations for Monitoring and Early Warning

Based on the susceptibility assessment results and the identification of clustered lake channels, we propose the following recommendations for future monitoring and early warning. First, the 14 glacial lakes with high expansion rates and very high susceptibility identified in Section 3.2.2 (Figure 10) should be included in the priority monitoring list, among which Guangxie Co (Lake 4) and Lakes 8, 11, and 12 (Figure 11) are of particular concern due to their proximity to National Highways G318 and G219 and the active geological hazards (landslides, ice avalanches, and moraine dam erosion) in their surrounding areas [70]. For in situ observations, we recommend installing automated water-level sensors in the priority lakes to continuously monitor water-level fluctuations [71]. Water depth is one of the primary indicators of lake stability, and anomalous water-level changes may serve as important precursors to outburst events. For cascading lake clusters, sensors should be deployed in both upstream and downstream lakes simultaneously to capture the dynamic chain-reaction processes. Such systems are structurally simple, low-cost, and can be powered by solar energy, making them suitable for long-term operation in remote areas. For remote sensing monitoring, we recommend acquiring high-resolution optical and SAR satellite im-agery at monthly or sub-monthly intervals to track changes in lake area, glacier retreat dynamics, and the development of hazard precursors such as ice avalanches, moraine dam erosion, and landslide activity [72]. Remote sensing data can complement in situ sensor data, providing important evidence for the early identification of outburst events. For predictive modeling, the historical GLOF records and lake inventory data compiled in this study can provide valuable prior information for constructing probabilistic prediction models. Field-observed water-level and area-change data can be dynamically assimilated into these models to enable updated assessments of outburst risk. For sections of National Highways G219 and G318 that lie within 10 km of potential outburst pathways (Table S3), we recommend reinforcing bridges and culverts along these routes to enhance their resilience against potential flood im-pacts. Where conditions permit, diversion channels or debris-flow barriers could be constructed to mitigate the effects of outburst floods. In addition, we recommend coordinating with local transportation authorities to develop targeted emergency response plans.”

Figure 6 caption: The y-axis label for (a) is "lakes/year" but the caption reads "various sizes of lakes." Please clarify what size categories are included.

——We thank the reviewer for this comment. Upon checking, we note that the original Figure 6 (now renumbered as Figure 7 in the revised manuscript) contains four subplots. Panel (a) shows the interannual variation trend in the total number of glacial lakes from 1990 to 2020; panel (b) shows the interannual variation trend in the total area; and panels (c) and (d) present the expansion rates of lakes across different size classes in terms of number and area, respectively. The phrase "various sizes of lakes" in the original caption was intended only for panels (c) and (d). The size classes for panels (c) and (d) are defined as follows: < 0.02 km², 0.02–0.05 km², 0.05–0.1 km², 0.1–0.2 km², 0.2–0.5 km², 0.5–1.0 km², and > 1.0 km².

This is a valuable contribution to GLOF hazard assessment with important practical implications for infrastructure planning in Tibet. The methodology is generally sound, though the susceptibility assessment could be more rigorously validated. The identification of cascading/clustering effects is a significant finding that advances beyond single-lake assessments. With the recommended revisions, particularly regarding methodological transparency, and discussion of limitations, this manuscript would make a strong contribution to the geohazards literature.

——We thank the reviewer for the strong affirmation of the value of our study. We sincerely appreciate the reviewer's recognition of the contributions of this work: its academic value for GLOF hazard assessment, its practical significance for infrastructure planning in Tibet, and the important finding of cascading/clustering effects that advances beyond traditional single‑lake assessments. We fully accept the reviewer's recommendation that methodological transparency and the discussion of limitations need to be strengthened. In response to these comments, we have made systematic improvements in the revised manuscript: we have added statistical testing for indicator selection (Mann–Whitney U test with complete statistical results), clarified the basis for weight assignment and added sensitivity analysis, included a new methodological subsection on cascading cluster identification (Section 2.4), supplemented back‑testing of historical outburst events, added a dedicated limitations subsection in the Discussion (Section 4.5), and systematically harmonized the terminology throughout the text by explicitly defining "risk" as "susceptibility/hazard" to accurately reflect the scope of our study. We believe that through these revisions, the manuscript has been substantially improved in methodological rigor, result reliability, and applied value, and now meets the publication standards of this journal. We thank the reviewer again for the valuable time and constructive comments, which have greatly improved our work.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

This is a useful paper on the geohazards associated with lake floods to transport routes in southeast Tibet. It examines ‘spatiotemporal changes’ via historical records and satellite imagery to help ensure safe operations of these transport routes and project.

Now this is a very worthy goal but, as presented, it remains of local use as it stands. This is evident from the first of your ‘implications’. I do not want to denigrate the work and results but rather suggest that it should be more outgoing. In particular, to suggest more explicit future work and emphasise this to show other readers what might be important as both triggers and consequences. I come back to this.

First of all I would like to see a re-submission but which has proper geolocation of the places mentioned and particularly of the illustrations. The maps are indicative rather than informative. Your table 1 does give 12 of the glof events as studied but the location is hidden as a decimal lat/decimal long value as the third item. Attribution under the FAIR data principles would be much more useful so I suggest (so that anyone can locate them easily) a format, decimal lat, decimal long, as a tuple [dLL] such that the first Gebuma Co (which I could not find in Google Earth) as [29.752,96.566]. (see Wikipedia decimal latitude longitude.) Then please place this value (it does not need the [ ] here for each row as a row header. This make the data significant for searching, a numbered id may be useful in a report but is not FAIR compliant. It would help if you included all the sites you refer to or have used here. Similarly, you can then supply on each Image and in the cation/metadata the appropriate location value. Incidentally I could not download the supplementary data from the site given. Any geolocation data should be included here too in the format just mentioned.

Regarding the significance. You do mention, lines 362-3, about prior studies and say ‘can provide a more comprehensive understanding’. Even if you do not include any comparison you might give this as a ‘future work’ and perhaps give the significance and data associated with them. This might show other workers in other areas some methodology resulting from your work.

You also mention, lines 504-606, future detailed field investigations’. Can you be more specificas to the type of observation? Eg repeat satellite images, water level sensors in the possible cascade to signify the event. These could be simple pressure transducer systems as the water depth is likely to be the main event. You also mention modelling, would a Bayesian approach be useful? You hasve priors, field results could be fed into this and help to produce a waring system that could be installed simple and relatively cheaply. Some warning systems in Nepal, (as far as I can remember, have not worked as well as they could because of lack of maintenance (and or money). Forecasting and prediction could be an important output from your work – so don’t hide it!

Author Response

GeoHazards

Re: revision on manuscript #geohazards-4452106

Dear Editors and Reviewers:

We would like to express our gratitude for the valuable comments and suggestions raised by the reviewer on our manuscript:

Evolution of Glacial Lakes and GLOF Risks to Transportation Routes in the Southeastern Tibetan Engineering Corridor

Authors’ response to reviewer’s comments

Reviewer 2:

Comments1: This is a useful paper on the geohazards associated with lake floods to transport routes in southeast Tibet. It examines ‘spatiotemporal changes’ via historical records and satellite imagery to help ensure safe operations of these transport routes and project.

——We thank the reviewer for acknowledging the value of our research topic and for candidly pointing out the limitations of the original manuscript. We fully understand the concern that the study may appear too regionally focused—if the results remain confined to serving National Highways G219 and G318 alone, it would indeed be difficult to attract broader readership. In the revised manuscript, we have attempted to enhance the generalizability and broader applicability of our study through the following improvements:

First, at the methodological level, we have strengthened the transferability of our approach. In the revised manuscript, we have supplemented the detailed criteria for cascading cluster identification (Section 2.4), including DEM‑based flow direction extraction, distance thresholds, and manual image interpretation. We have also added statistical testing (Mann–Whitney U test) for indicator selection, and incorporated multi‑formula comparison and sensitivity analysis for volume estimation.

Second, at the data level, we have improved data accessibility and FAIR compliance. We have uploaded the complete 1990–2020 glacial lake inventory for the study area (in Shapefile format) to the Mendeley Data open repository, accessible via a private link during the review process and to be publicly released with a DOI upon acceptance. This not only allows other researchers to directly use our data, but also provides a comparable data standard for glacial lake studies in other regions.

Third, in the Discussion section, we have clearly identified specific directions for future work. We have added Section 4.4 on monitoring and early‑warning recommendations, proposing a monitoring framework applicable to high‑mountain regions globally, and have identified hydrodynamic modelling and quantitative risk assessment as priority research directions in Section 4.5.

We believe that through these revisions, our study can offer a transferable methodological framework for GLOF hazard research and infrastructure planning in other high‑mountain regions worldwide. We hope these improvements meet the reviewer's expectations, and we welcome any further guidance the reviewer may provide.

Comments2: First of all I would like to see a re-submission but which has proper geolocation of the places mentioned and particularly of the illustrations. The maps are indicative rather than informative. Your table 1 does give 12 of the glof events as studied but the location is hidden as a decimal lat/decimal long value as the third item. Attribution under the FAIR data principles would be much more useful so I suggest (so that anyone can locate them easily) a format, decimal lat, decimal long, as a tuple [dLL] such that the first Gebuma Co (which I could not find in Google Earth) as [29.752,96.566]. (see Wikipedia decimal latitude longitude.) Then please place this value (it does not need the [ ] here for each row as a row header. This make the data significant for searching, a numbered id may be useful in a report but is not FAIR compliant. It would help if you included all the sites you refer to or have used here. Similarly, you can then supply on each Image and in the cation/metadata the appropriate location value. Incidentally I could not download the supplementary data from the site given. Any geolocation data should be included here too in the format just mentioned.

——We thank the reviewer for the professional comments. We have made comprehensive and systematic improvements to the data geolocation, accessibility, and reproducibility of our manuscript according to your requirements. The specific revisions are as follows:

(1) Reorganization of Table 1 to highlight location information using the [dLL] format.

We have redesigned Table 1 by moving the decimal latitude and longitude coordinates to the first column as row headers and removing the original ID column. The location information is now presented in the decimal tuple [dLL] format, making it more prominent and easier for machine reading and searching, fully complying with the FAIR data principles.

(2) Explanation regarding the inability to locate Gebuma Co in Google Earth.

The reason why "Gebuma Co" cannot be found in Google Earth is that this lake has now almost completely retreated. To facilitate verification by the reviewer, we have supplemented the manuscript with comparative images showing the lake before and after retreat (see Figures 1 and 2 below) to illustrate its current status.

(3) Correction of coordinate deviations for all glacial lakes in Table 1.

Upon checking, we found that the coordinates of some glacial lakes in Table 1 did have minor deviations, mainly due to errors introduced during coordinate format conversion. For example, the correct centroid coordinates of "Gebuma Co" should be [29.754, 96.557]. We have reviewed and corrected the coordinates of all 12 historical GLOF events in Table 1 one by one to ensure their accuracy.

(4) Supplementation of geolocation information in all figures.

In the revised manuscript, we have added decimal latitude and longitude coordinates ([dLL] tuples) to each subplot in Figure 10, labeled in the lower‑right corner of each subplot. The location information for Figures 11, 12, and 14 can be referenced through the index in Figure 10, achieving complete geolocation coverage for all figures in the manuscript.

(5) Explanation regarding the Supplementary Materials.

We have uploaded a supplementary file containing the susceptibility assessment results and location information for each glacial lake ("Supplementary document.docx") at the time of submission. The reviewer's previous inability to download this file may have been because the link in the manuscript is a general template generated by the journal system, which typically becomes active only after formal publication; during the review stage, supplementary materials are accessible only through the review system. In addition, considering that numbering helps readers quickly locate and cite specific lakes in the text, we have retained the first column as the ID number in the supplementary file, with the second and third columns containing the decimal latitude and longitude coordinates. We believe this arrangement achieves a balance between convenience and the FAIR principles and is reasonable.

(6) Upload of Shapefile data to a public repository to enhance reproducibility.

To further improve data transparency and reproducibility, we have uploaded the Shapefile spatial data from this study to the Mendeley Data open repository. During the review period, the data can be accessed via the following private link: https://data.mendeley.com/preview/xt5t28zym8?a=7d529408-94bd-41b2-888b-25cfc56073e0. The dataset includes glacial lake inventory vector data for four epochs and the susceptibility assessment results, and will be publicly released with a DOI upon formal acceptance of the paper.

We hope that the above revisions fully clarify the data location and transparency issues of this study. Should you have any further comments, we welcome your continued guidance.

Figure 1. Gebuma Co glacial lake image on 27 January 1988

Figure 2. Gebuma Co glacial lake image on 22 October 2024

Comments3: Regarding the significance. You do mention, lines 362-3, about prior studies and say ‘can provide a more comprehensive understanding’. Even if you do not include any comparison you might give this as a ‘future work’ and perhaps give the significance and data associated with them. This might show other workers in other areas some methodology resulting from your work.

——We thank the reviewer for this suggestion. We fully agree that a more detailed elaboration of the research significance and future work directions would significantly enhance the reference value of the paper and facilitate the extension of our findings and methodology to other regions.

(1) Regarding research significance, we have clarified the applied value of this study in the newly established Section 4.4 (Recommendations for Monitoring and Early Warning). The proposed monitoring and early warning framework—including the establishment of priority lake lists, the deployment of water-level sensors, multi-source remote sensing collaborative monitoring strategies, and the conceptual approach for constructing probabilistic prediction models based on historical inventory data—can serve as a transferable methodological reference for glacial lake hazard research in other high-mountain regions.

(2) Regarding specific future work directions, we have provided detailed responses in Section 4.4 to the reviewer's questions about future field investigations: (1) for in situ observations, we recommend installing automated water-level sensors (e.g., pressure transducer systems) in priority lakes for continuous monitoring of water-level fluctuations; for cascading lake clusters, sensors should be deployed in both upstream and downstream lakes simultaneously to capture chain-reaction dynamics; (2) for remote sensing monitoring, we recommend acquiring high-resolution optical and SAR satellite imagery at monthly or sub-monthly intervals to track lake area changes and the development of hazard precursors such as ice avalanches and moraine dam erosion; (3) for predictive modeling, the historical GLOF records and lake inventory data compiled in this study can provide prior information for constructing probabilistic prediction models, and field-observed data can be dynamically assimilated into these models to enable updated assessments of outburst risk, thereby establishing a relatively low-cost early warning system. The above content has been fully presented in Section 4.4 of the revised manuscript, and the revised text is as follows:

“4.4. Recommendations for Monitoring and Early Warning

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

Jin et al. constructed a multi-temporal lake inventory of the southeastern Tibetan engineering corridor from 1990 to 2020 in 10-year intervals, monitored lake evolution over this period, characterized the glacial lake outburst flood (GLOF) risks of lakes larger than 0.01 km², and identified hazards to infrastructure if various lakes experienced a GLOF. They constructed the inventory using Landsat scenes (and possibly Sentinel-2 as well). The authors thoroughly analyze their multi-temporal inventory. In addition, they perform a novel assessment of looking at connected at-risk lakes, where a GLOF in one lake could trigger downstream GLOFS. They identify three such clusterings, which provide improved hazard assessments from the commonly used, single-lake methods.

The writing is clear, and the visuals are compelling. My main critique is that some key details are missing. These, as well as other general comments, are outlined below. I have also provided some line-specific comments.

General Comments

General Comment 1:

How were the types of lakes determined, and were lakes allowed to have their classification change with time? For example, as a glacier retreats, can a supraglacial lake become a proglacial lake? More details should be provided in Lines 138-140. In particular, I am interested in how glacier retreat is handled. For example, if the RGI v6.0 were used to classify a lake from 2020 as a supraglacial, it might be a pro-glacial lake due to the retreat of the glacier between 2000 (the RGI typically maps glaciers for 2000 +/- 10 years) and 2020.

General Comment 2:

More details are needed about how these four variables in the GLOF metric were combined (Lines 206-213). Were the normalized tiers multiplied or added? If so, were each metric weighted equally, or was a high risk from one metric (e.g., lake volume) more impactful than from another metric (e.g., elevation)? It strikes me that these metrics should not be equally weighted – a high risk for some variables would be more impactful than for others.

Has this risk metric approach been used before? If so, make sure to cite that(those) paper(s). If not (and even if it’s been used before), validation is needed. In the Discussion (Line 358-360), it is stated that validation is not needed because these variables were chosen from past GLOFs. However, as I note in the Specific Comment for Lines 167-171/Fig. 1, it is not obvious from the analysis of historical GLOFS that these four variables are predictive. And even if they are in isolation, it’s not obvious that their combination should be predictive. If this is a new metric – and even if it’s been used before – I strongly believe validation is needed. Would this metric have predicted the lakes that had historical GLOFs as high or very high risk shortly before their GLOFs?

General Comment 3:

The statistical analyses for climate variable trends (Section 4.2) need to be more robust. I gather from Fig. 12 that regression was used. Was it OSL regression? If so, that would likely not be the most appropriate method for a climate time series if there is autocorrelation or if the residuals are not normally distributed. Were the conditions for OLS regression checked before using (if used)? If not – or if any of these conditions failed, a safer bet would be to use Sen’s Slope analysis for the trend and a Mann-Kendall test to determine the significance of the trend. This combination of Sen’s Slope analysis & Mann-Kendall test is the best choice for trend analyses of climate time series.

Also, when interpreting any trends in the data, the significance of the trends must be considered. The discussion of a moistening trend (Lines 403-404), I suspect, would not be significant given how small the trend is (<1% per year), compared to the larger interannual and multi-year variability. The significance of any trends should be noted in the text, and only statistically significant trends should be considered in the discussion.

Specific Comments:

Line 134: What band ratios were used?

Lines 167-171/Fig. 1: How were the four key variables (lake volume, elevation, distance to the pargnet glacier, and parent glacier size) determined? Looking at Fig. 1, I’m not sure I’d pick out those four variables as the only ones (or even the key ones) that differ. In fact, elevation looks pretty similar between each group, whereas precipitation looks distinct between the two groups. Both the sentence about the distinct variables (Lines 167-169) and the sentence about other variables not being notably different (Lines 169-171) should be expanded to provide the reasoning for these determinations, since looking at the plots, I am not sure I’d say either claim is true.

Fig. 3/Fig. 4. Fig. 7/Fig. 8: Are these maps/plots for glaciers in 2020 or for another year in the study period? The figure caption should note for what year(s) the lake data are for these maps.

Fig. 5: I’d included the interval over which these changes/rates were determined. I suspect it’s from 1990 to 2020, but that detail should be added in the caption.

Lines 271-274/Fig. 6 Caption: ‘Interannual variation’ strikes me as being off. For interannual variations, I’d expect year-on-year differences. Here, however, the variation is shown between 4 annual measurements, each spaced 10 years apart. It’d be more accurate to call it ‘Variation in …’.

Lines 284-286/Fig. 8b: The order of lake types in the x-axis of Fig. 8b seem off from the text in Lines 283-286. The text states that 2 lakes are supraglacial, but the figure indicates that there are over 10 supraglacial lakes and only 2 proglacial lakes.

Line 358-360: Validation of the GLOF hazard metric in this study seems warranted (see General Comment 2).

Line 403-404: See General Comment 3 about statistical significance. To say there is a trend of increased precipitation, that trend must be statistically significant. From Fig. 12, I suspect that is not the case for precipitation. But if [proper] statistical tests indicate it is the case, then that should be noted in the text (e.g., … an increase of x mm per decade (p < 0.05).

Line 423/Fig. 12: In the figure, it looks like the data spans to 2022, but the figure caption says it stops at 2020. Also see General Comment 3 about the analysis for trends.

Section 4.3/Line 231: A very cool analysis. Should ‘clustering analysis’ be written as a ‘cluster analysis’ if a cluster analysis was used? Or was an analysis used that clustered lakes, which is being called here a ‘clustering analysis’?

Line 487-488: Has it been shown in the main text that proglacial lakes contributed the most to lake-area growth? If so, I missed that.

Line 509/Supplementary Materials: The satellite scenes used to build the inventory should be listed somewhere for reproducibility. They are typically located in the Supplementary Materials if not included as a table in the Methods section.

Supplementary Materials/Data Availability: Are there shapefiles (or geopackages) of the multi-temporal lake inventory from the Supplementary Materials that could be shared? It’s great to have the data in tables, but it would be cumbersome to use the inventory for a follow-up study. The impact of this study could increase with the availability of these data in geospatial format (as well as in tables).

Author Response

GeoHazards

Re: revision on manuscript #geohazards-4452106

Dear Editors and Reviewers:

We would like to express our gratitude for the valuable comments and suggestions raised by the reviewer on our manuscript:

Evolution of Glacial Lakes and GLOF Risks to Transportation Routes in the Southeastern Tibetan Engineering Corridor

Authors’ response to reviewer’s comments

Reviewer 3:

Jin et al. constructed a multi-temporal lake inventory of the southeastern Tibetan engineering corridor from 1990 to 2020 in 10-year intervals, monitored lake evolution over this period, characterized the glacial lake outburst flood (GLOF) risks of lakes larger than 0.01 km2, and identified hazards to infrastructure if various lakes experienced a GLOF. They constructed the inventory using Landsat scenes (and possibly Sentinel-2 as well). The authors thoroughly analyze their multi-temporal inventory. In addition, they perform a novel assessment of looking at connected at-risk lakes, where a GLOF in one lake could trigger downstream GLOFS. They identify three such clusterings, which provide improved hazard assessments from the commonly used, single-lake methods.

——We sincerely thank you for the thorough and constructive evaluation of our manuscript. We are truly encouraged by the positive recognition of our work, particularly the acknowledgement that our multi-temporal lake inventory, the novel cascading/clustering risk assessment, and the clear writing and compelling visuals represent valuable contributions to the field. We fully agree with the reviewer's primary critique that some key methodological details are missing from the original submission. In the revised manuscript, we have systematically addressed this concern by supplementing all the requested details, following the reviewer’s specific General Comments and Line-specific Comments. Specifically:

General Comments

General Comment 1:

——We thank the reviewer for raising this important methodological issue. We fully agree that the classification of lake types and their potential changes over time require clear explanation. Below we respond to this comment in detail.

(1) Definition of lake types. In this study, the three types of glacial lakes are defined based on their spatial relationship with the parent glacier, following the classification criteria of Zhang et al. (2015). Since the classification of glacial lake types is already a relatively common background knowledge in this field, we did not provide a detailed description of each type in the main text in order to keep the manuscript concise, but instead supported these definitions by citing relevant references. We kindly ask for the reviewer's understanding on this matter.

Proglacial lakes: lakes whose water bodies are in direct contact with the glacier terminus, located beyond the glacier margin and impounded by moraine or bedrock.

Supraglacial lakes: lakes situated entirely on the glacier surface, with their boundaries completely surrounded by glacier ice.

Detached lakes: lakes that no longer have direct contact with any glacier, but are still supplied by glacial meltwater and located within 500 m of the glacier terminus (the distance threshold refers to Zhang et al., 2015).

(2) Lake types are allowed to change over time. We fully agree that as glaciers retreat, the hydrological connection between a lake and its parent glacier evolves. Accordingly, in our inventory, the type of an individual lake can transition (e.g., from supraglacial to proglacial, and subsequently to detached). The temporal changes in lake‑type counts in our inventory are as follows: in 1990, 14 proglacial, 0 supraglacial, and 169 detached lakes; in 2000, 28 proglacial, 2 supraglacial, and 213 detached lakes; in 2010, 28 proglacial, 1 supraglacial, and 216 detached lakes; and in 2020, 25 proglacial, 3 supraglacial, and 229 detached lakes.

(3) Handling of glacier retreat across different epochs. In determining lake types for each epoch, we did not directly use the glacier boundaries from RGI v6.0. Instead, we used RGI v6.0 as a reference base map, overlaid it with Landsat imagery from each epoch, and manually interpreted the spatial relationship between each lake and its adjacent glacier through visual inspection to determine its type. This approach ensures that the classification reflects the actual glacier–lake configuration at each specific time step, avoiding the systematic bias that would arise from using a fixed‑baseline (~2000) RGI dataset to classify lakes in earlier (1990) or later (2020) epochs.

(4) Feasibility of manual classification. Given that our study area is relatively small and the total number of glacial lakes is limited compared to the broader Tibetan Plateau, each lake was carefully examined and classified manually rather than through automated algorithms. This manual approach is both feasible and reliable in our case, as the manageable number of lakes allowed for thorough visual inspection of each individual lake.

Regarding this comment, our revision in Section 2.2 of the original manuscript is as follows:

“Notably, the present study permits shifts in lake classification for an individual lake across different periods. For lake type identification at each time step, classification was not directly implemented using glacier outlines from the RGI v6.0 dataset. Instead, taking RGI v6.0 as a reference, we overlaid multi-temporal Landsat imagery and manually interpreted the spatial relationship between each lake and adjacent glaciers to determine lake type.”

Zhang, G.; Yao, T.; Xie, H.; Wang, W.; Yang, W. An inventory of glacial lakes in the Third Pole region and their changes in response to global warming. Glob. Planet. Change 2015, 131, 148–157. https://doi.org/10.1016/j.gloplacha.2015.05.013.

General Comment 2:

——We thank the reviewer for raising this critical issue. We fully agree that the combination method of the four indicators, the derivation of weights, and model validation are essential methodological details that must be clearly addressed. In response to the reviewer's comments, we have made the following revisions and additions:

(1) We have explicitly defined the combination formula for the four indicators. In the original manuscript, we only stated the weights of the four indicators without specifying how they were combined. We have now added the weighted linear summation formula in the Methods section:

where S is the GLOF susceptibility score of the glacial lake; T_V is the normalized tier value of the lake volume indicator; T_D is the normalized tier value of the distance to the parent glacier indicator; T_A is the normalized tier value of the parent glacier area indicator; and T_E is the normalized tier value of the elevation indicator.

(2) The weight assignments were not subjectively determined, but were instead based on the statistical differences between outburst and stable lakes for each indicator (Figure 2), combined with the contribution of each indicator to the potential magnitude of GLOF hazards. Specifically, lake volume shows the most pronounced discriminative power (outburst lakes are generally 1–2 orders of magnitude larger than stable lakes) and directly determines the potential flood magnitude, thus receiving the highest weight (0.4). Distance to the parent glacier and parent glacier area also exhibit clear differences between the two groups, each receiving a weight of 0.25. Elevation shows relatively weaker discriminative power and therefore receives a lower weight (0.1).

(3) We have supplemented model validation. We fully accept the reviewer's criticism. The original statement that "validation is not necessary" was indeed inappropriate. We have completely removed this expression in the revised manuscript and added a discussion of the validation results. To validate the effectiveness of our susceptibility assessment framework, we applied the evaluation system to the historically documented GLOF lakes within the study area. The results show that Guangxie Co and Lumu Co were classified as "very high" susceptibility, Anjue Zhuqulong glacier lake as "very high" susceptibility, and Zhuxigou glacial Lake as "high" susceptibility. All four historically documented GLOF lakes within the study area were correctly identified as either "high" or "very high" susceptibility, preliminarily confirming that our assessment framework can effectively distinguish lakes prone to outburst. The revised Section 5.1 is presented below:

“Validating GLOF susceptibility assessments remains challenging due to the com-plex and stochastic nature of GLOF triggers [30,43,59]. Nevertheless, to test the per-formance of our assessment framework, we back-tested it against historically docu-mented GLOF lakes within the study area. The results show that Guangxie Co, Lumu Co, and Anjue Zhuqulong glacier lake were all classified as having "very high" suscep-tibility, while Zhuxigou glacial Lake was classified as "high" susceptibility. All docu-mented historical outburst lakes were correctly rated as either "high" or "very high" susceptibility, supporting the capability of our framework to distinguish lakes prone to outburst. It should be noted, however, that given the limited number of historical GLOF events in the study area, this validation remains a back-testing exercise; fully independent statistical validation will require the accumulation of more GLOF records in the future. In addition, assessment indicators and methods inevitably vary across studies due to regional differences [32,43,54]. Nonetheless, two previous GLOF suscep-tibility assessments partially overlap with our study area, and comparing our results with these studies provides a more comprehensive regional perspective on GLOF risks.”

(4) We have supplemented methodological references. This weighted multi‑indicator approach has been widely used in GLOF susceptibility assessments. While we have locally adjusted the indicator selection and weight assignment according to the characteristics of the southeastern Tibetan region, the basic framework draws on established practices from previous studies (Aggarwal et al., 2017; Zhang et al., 2023; Gaikwad et al., 2024).

Aggarwal, S.; Rai, S. C.; Thakur, P. K.; Emmer, A. Inventory and recently increasing GLOF susceptibility of glacial lakes in Sikkim, Eastern Himalaya. Geomorphology 2017, 295, 39–54. https://doi.org/10.1016/j.geomorph.2017.06.014.

Gaikwad, D.; Tyagi, A.; Tiwari, R. K. Hybrid FR-AHP approach for GLOF hazard assessment in the Himalayan region. Remote Sens. Appl. Soc. Environ. 2024, 101437. https://doi.org/10.1016/j.rsase.2024.101437.

Zhang, M.; Zhang, G.; Yu, J.; Wang, W.; Xu, F.; Rinzin, S. Potential threats of glacial lake changes to the Sichuan-Tibet Railway. J. Glaciol. 2024, 70, e76. https://doi.org/10.1017/jog.2024.40.

General Comment 3:

——We thank the reviewer for pointing out the deficiencies in the statistical methods used for the climate trend analysis. We fully accept the reviewer's suggestions and have thoroughly revised this section.

(1) We have adopted the Mann–Kendall test and Sen's Slope estimator. We no longer use OLS regression; instead, we employed the Mann–Kendall non‑parametric trend test to assess statistical significance and the Sen's Slope estimator to quantify trend magnitudes. This combination imposes no assumptions on data distribution and is robust to outliers and autocorrelation, making it a standard method for trend analysis of hydro‑climatic time series. The statistical methods are now clearly described in the revised text (Section 4.2).

(2) We have explicitly reported the significance of the trends. The revised manuscript now includes the test statistics (Z values) and p‑values for each trend. The temperature trend is significant at the 0.001 level (Z = 6.65, p < 0.001), with a Sen's Slope of approximately 0.045 °C/yr. The precipitation trend does not reach statistical significance (Z = 0.24, p ≥ 0.05), with a Sen's Slope of only 0.03 mm/yr. These statistical results have also been added to the caption of Figure 12.

(3) We now discuss only significant trends in the Discussion. Since the precipitation trend is not statistically significant, we have removed the original statement that "annual precipitation also shows an increasing trend" and instead explicitly state that interannual precipitation variability is primarily attributable to natural variability rather than a long‑term trend. When discussing climatic drivers, we attribute glacial lake expansion solely to the significant warming trend and do not treat precipitation change as a statistically significant contributing factor. The revised text is as follows:

“Climate change has been widely recognized as a significant background factor influencing the evolution of glaciers and glacial lakes in the southeastern of the Tibetan Plateau [23,24,25]. Based on the ECMWF Reanalysis v5 (ERA5), we obtained the annual average temperature evolution process for the study area from 1970 to 2022 (Figure 12). To detect long‑term trends in the climatic time series, we employed the Mann–Kendall non‑parametric trend test to evaluate statistical significance and the Sen's Slope estimator to quantify trend magnitudes. The results reveal a highly significant warming trend in air temperature (Z=6.65, p<0.001), with a Sen's Slope of 0.045 °C/yr. In contrast, the precipitation trend does not reach statistical significance (Z=0.24, p ≥ 0.05), indicating that interannual variability, rather than a long‑term directional change, dominates the precipitation regime. Climate change exerts direct influences on the hydrological balance of glacial lakes. The observed lake expansion is primarily attributable to regional warming, which accelerates glacier and snow melt and thereby supplies increased meltwater to glacial lakes [34,58,67]. No statistically significant contribution from precipitation changes to lake expansion was identified in this study.”

The above revisions have been made in Section 4.2 and the caption of Figure 12. We thank the reviewer again for the valuable suggestions on statistical rigor.

Specific Comments:

1.Line 134: What band ratios were used?

——We thank the reviewer for pointing out this missing detail. We agree that explicitly reporting the band ratios used is essential for methodological reproducibility. The following information has been added to Section 2.2 of the revised manuscript:

(1) Specific band ratios. For Landsat 5 and 7 (TM/ETM+), we used the ratio of the green band (band 2) to the shortwave infrared band (band 5), i.e., TM2/TM5. For Landsat 8 and 9 (OLI), we used the corresponding ratio of the green band (band 3) to the shortwave infrared band (band 6), i.e., OLI3/OLI6. These two ratios have corresponding physical relationships across different sensors, ensuring consistency across the multi‑temporal dataset.

(2) Rationale for selection. Water bodies have relatively high reflectance in the green band and very low reflectance in the shortwave infrared bands. In contrast, ice, snow, and bare rock exhibit much higher reflectance in the shortwave infrared bands than water. Thus, the green/SWIR ratio effectively enhances water body information while suppressing interference from non‑water features.

(3) Manual correction. All automatically extracted lake boundaries were manually inspected and corrected on a case‑by‑case basis to ensure boundary accuracy.

The above revisions have been made in Section 2.2 of the revised manuscript, as follows:

“Utilizing the Google Earth Engine (GEE) platform, the research employed optical re-mote sensing data from Landsat 5–9 satellites with cloud cover below 30% (Table S1). For glacial lake identification, we applied the band ratio method, which exploits the spectral differences between water and other land cover types. Specifically, for Landsat 5 and 7 (TM/ETM+), we used the ratio of band 2 (green) to band 5 (SWIR1), i.e., TM2/TM5; for Landsat 8 and 9 (OLI), we used the ratio of band 3 (green) to band 6 (SWIR1), i.e., OLI3/OLI6. These band combinations effectively enhance water bodies while sup-pressing the reflectance of ice, snow, and bedrock [50–52].”

Lines 167-171/Fig. 1: How were the four key variables (lake volume, elevation, distance to the pargnet glacier, and parent glacier size) determined? Looking at Fig. 1, I’m not sure I’d pick out those four variables as the only ones (or even the key ones) that differ. In fact, elevation looks pretty similar between each group, whereas precipitation looks distinct between the two groups. Both the sentence about the distinct variables (Lines 167-169) and the sentence about other variables not being notably different (Lines 169-171) should be expanded to provide the reasoning for these determinations, since looking at the plots, I am not sure I’d say either claim is true.

——We thank the reviewer for raising this critical issue. We fully understand your concern: visual comparison of the boxplots in Figure 2 alone does indeed make it difficult to determine which indicators are statistically significant. To address this, we have made the following improvements in the revised manuscript:

(1) We have added statistical testing to support variable selection. We performed the Mann–Whitney U non‑parametric test to evaluate the significance of differences between outburst and stable lakes for all candidate indicators. The results are summarized in Supplementary Table S2. The test results show that lake volume (p < 0.001), distance to the parent glacier (p = 0.025), and parent glacier area (p = 0.007) differ significantly between the two groups, whereas elevation (p = 0.304), glacier tongue slope (p = 0.344), slope near the lake (p = 0.084), air temperature (p = 0.612), and precipitation (p = 0.302) do not reach statistical significance.

Table S2. Comparison of indicators between stable and historically outburst glacial lakes

Indicator	All lakes (n = 257) Mean ± SD	Outburst lakes (n = 12) Mean ± SD	Mann-Whitney U	Z	p-value	Significance
Lake volume (×10⁶m³)	3.56 ± 13.72	4.71 ± 4.20	1568.0	4.702	< 0.001	***
Distance from the parent glacier (m)	1786.43 ± 2217.68	581.08 ± 828.92	2537.0	2.248	0.025	*
The area of the parent glacier (km²)	3.76 ± 9.97	5.68 ± 8.32	2350.5	2.713	0.007	**
Elevation (m)	4350.20 ± 441.69	4539.04 ± 412.57	2830.5	1.027	0.304	n.s.
Glacier tongue slope (°)	17.43 ± 9.30	14.15 ± 5.36	2839.0	0.947	0.344	n.s.
Slope gradient near the lake (°)	29.79 ± 14.75	31.62 ± 4.58	2616.0	1.729	0.084	n.s.
Temperature (°C)	-4.10 ± 3.09	-4.48 ± 3.82	2939.5	0.507	0.612	n.s.
Precipitation (mm)	714.89 ± 52.78	664.09 ± 103.68	2863.5	1.032	0.302	n.s.

Note: *** p < 0.001, ** p < 0.01, * p < 0.05, n.s. = not significant (p ≥ 0.05). SD = standard deviation. The Mann–Whitney U test was used due to the non-normal distribution of the data and the small sample size of the outburst lakes (n = 12)

(2) Regarding the relationship between the visual appearance of elevation and the statistical results. The reviewer observed that "elevation looks pretty similar between each group." This visual observation is consistent with our statistical results—the p‑value for elevation is 0.304, indeed not reaching statistical significance. We included elevation in the assessment framework not based on its statistical discriminative power, but rather on its widely recognized physical importance in the literature (higher‑altitude lakes are typically situated in steeper terrain, where the substantial elevation drop would significantly amplify flood velocity and reach in the event of an outburst; high‑altitude regions are also more sensitive to climate warming, with accelerated glacier retreat and more frequent ice avalanches indirectly increasing the likelihood of dam failure). Therefore, we assigned it a lower weight (0.1).

(3) Regarding the relationship between the visual appearance of precipitation and the statistical results. The reviewer astutely observed that "precipitation looks distinct between the two groups." This observation is correct—from the boxplot in Figure 2h, the mean precipitation of outburst lakes does appear slightly lower than that of stable lakes. However, the Mann–Whitney U test shows that this difference does not reach statistical significance (p = 0.302), indicating that the observed difference is relatively small. Moreover, the precipitation data used in this study are derived from ERA5 reanalysis, which has a relatively coarse spatial resolution (approximately 0.1° × 0.1°) and may suffer from systematic biases in valley glacier regions, making it difficult to accurately capture precipitation characteristics at the local scale of glacial lakes. In contrast, elevation has a clearer physical meaning as an indicator and can be directly and accurately extracted from the DEM. Therefore, although a visual difference in precipitation appears to exist, we ultimately selected elevation as the fourth indicator based on considerations of both statistical significance and data reliability.

(4) We have clarified the variable selection workflow. In the revised manuscript, the selection workflow is as follows: first, we perform the Mann–Whitney U test on all candidate indicators to identify those with statistically significant differences between outburst and stable lakes (volume, distance to the parent glacier, and parent glacier area); then, on this basis, we include elevation as a fourth indicator to reflect its widely recognized physical importance in the literature. This combined strategy ensures the scientific validity and rationality of the final selected variables. Accordingly, we have added relevant text in the manuscript to enhance methodological transparency.

Detailed statistical results (including sample sizes, means, standard deviations, U values, Z values, and exact p‑values) are summarized in Supplementary Table S2. We hope that, through the addition of statistical testing and the revision of the relevant text, the reviewer's concerns regarding the variable selection process have been fully addressed.

Fig. 3/Fig. 4. Fig. 7/Fig. 8: Are these maps/plots for glaciers in 2020 or for another year in the study period? The figure caption should note for what year(s) the lake data are for these maps.

——We thank the reviewer for pointing out this omission. We fully agree that explicitly stating the data year in the figure titles helps readers to quickly understand the temporal reference of the figures. Accordingly, we have revised the titles of the corresponding figures (Figures 3, 4, 7, and 8) by clearly adding the temporal information within the subplot titles.

Fig. 5: I’d included the interval over which these changes/rates were determined. I suspect it’s from 1990 to 2020, but that detail should be added in the caption.

——We thank the reviewer for pointing out this omission. The reviewer's judgment is entirely correct—the change rates and area changes shown in Figure 5 were calculated based on the period from 1990 to 2020 (i.e., the 30‑year interval of the entire study period). We have added the temporal information accordingly in the corresponding position of the figure caption.

Lines 271-274/Fig. 6 Caption: ‘Interannual variation’ strikes me as being off. For interannual variations, I’d expect year-on-year differences. Here, however, the variation is shown between 4 annual measurements, each spaced 10 years apart. It’d be more accurate to call it ‘Variation in …’.

——We thank the reviewer for pointing out this terminological inaccuracy. We fully agree that "interannual variation" typically refers to year‑to‑year changes, whereas our data consist of only four epochs at 10‑year intervals (1990, 2000, 2010, and 2020) rather than continuous annual observations; therefore, the use of this term was indeed imprecise. We have revised the figure titles and the corresponding text to accurately reflect the temporal characteristics of the data. The specific revisions are as follows:

“Variation in glacial lake number and area across the four epochs (1990, 2000, 2010, and 2020) (a) Variation in the number of glacial lakes (b) variation in the area of glacial lakes (c) ex-pansion rates of glacial lakes of different size classes in terms of number (d) expansion rates of glacial lakes of different size classes in terms of area”

Lines 284-286/Fig. 8b: The order of lake types in the x-axis of Fig. 8b seem off from the text in Lines 283-286. The text states that 2 lakes are supraglacial, but the figure indicates that there are over 10 supraglacial lakes and only 2 proglacial lakes.

——We thank the reviewer for carefully identifying this issue. We sincerely apologize for this oversight in our figure preparation—the labels for "Pro. lake" and "Sup. lake" on the x‑axis of Figure 8b were indeed reversed, causing the distribution of lake types shown in the figure to be inconsistent with the text description, and creating confusion for both the reviewer and readers. We have corrected this error in the revised manuscript. We once again apologize for the confusion caused by this figure preparation oversight, and we are grateful to the reviewer for the rigorous correction.

Line 358-360: Validation of the GLOF hazard metric in this study seems warranted (see General Comment 2).

——We thank the reviewer for this comment. As detailed in our response to General Comment 2, we have removed the statement that "validation is not necessary," supplemented back‑testing results using historically documented GLOF lakes, and acknowledged the limitations of validation in the Discussion.

Line 403-404: See General Comment 3 about statistical significance. To say there is a trend of increased precipitation, that trend must be statistically significant. From Fig. 12, I suspect that is not the case for precipitation. But if [proper] statistical tests indicate it is the case, then that should be noted in the text (e.g., … an increase of x mm per decade (p < 0.05).

——We thank the reviewer for this comment. This issue has been systematically addressed in our response to General Comment 3. In the revised manuscript, we have re‑analyzed the precipitation and temperature trends using the Mann–Kendall test and Sen's Slope estimator, and have explicitly reported the statistical significance of the trends. The original statement regarding "an increasing precipitation trend" has been removed from the text. The detailed statistical methods, results, and the revised discussion of climatic drivers have been fully presented in our response to General Comment 3, and are not repeated here. We thank the reviewer again for the careful correction.

Line 423/Fig. 12: In the figure, it looks like the data spans to 2022, but the figure caption says it stops at 2020. Also see General Comment 3 about the analysis for trends.

——We thank the reviewer for identifying this inconsistency. We apologize for this oversight in the figure caption. We have corrected the caption to read "from 1970 to 2022" in the revised manuscript, ensuring consistency with the actual data range shown in the figure. In addition, the statistical methods for trend analysis have been fully revised following the recommendations in General Comment 3, as detailed in our previous response. We thank the reviewer again for the professional comments.

Section 4.3/Line 231: A very cool analysis. Should ‘clustering analysis’ be written as a ‘cluster analysis’ if a cluster analysis was used? Or was an analysis used that clustered lakes, which is being called here a ‘clustering analysis’?

——We thank the reviewer for this careful terminological observation. Our original intention was to describe the analysis of lakes that are spatially clustered, rather than to imply the use of a specific clustering algorithm. Therefore, the original term "clustering analysis" was indeed imprecise and could easily be confused with the concept of clustering analysis in statistics. To avoid ambiguity, we have uniformly revised the relevant terminology throughout the text to "analysis of clustered lakes" or "analysis of lake clusters," to clarify that this refers to the analysis of spatially aggregated glacial lakes rather than the application of a clustering algorithm. The title of Section 4.3 remains unchanged (the use of "clustered" as an adjective in "Amplified GLOF risks from clustered glacier lakes" is appropriate); in the main text, "Clustering analysis" has been revised to "Analysis of clustered lakes."

Line 487-488: Has it been shown in the main text that proglacial lakes contributed the most to lake-area growth? If so, I missed that.

——We thank the reviewer for pointing out this issue. Upon checking, we find that the reviewer's judgment is entirely correct—we indeed did not provide sufficient data support for this conclusion in the Results section. Although our statistical results do indicate that proglacial lakes contributed the most to lake area growth, we did not present explicit figures or quantitative analysis to demonstrate this in the main text. Therefore, directly stating in the Conclusions that "proglacial lakes contributed the most to glacial lake area increase" lacks sufficient basis. To ensure that our conclusions do not exceed the scope of evidence, we have removed this statement from the Conclusions section in the revised manuscript, ensuring that all conclusions are strictly grounded in the results and analyses already presented in the preceding sections.

Line 509/Supplementary Materials: The satellite scenes used to build the inventory should be listed somewhere for reproducibility. They are typically located in the Supplementary Materials if not included as a table in the Methods section.

——We thank the reviewer for pointing this out. We agree that listing the satellite scenes used to build the inventory is essential for ensuring the reproducibility of the study. Accordingly, we have added a new table (Table S1) in the Supplementary Materials, which lists all satellite scenes used in this study, including the Landsat images used for lake mapping across the four epochs (1990, 2000, 2010, and 2020) and the recent high‑resolution base map images used in the figures. The table includes information on the sensor, Scene ID, acquisition date, Path/Row, cloud cover, and intended use for each image. Due to the addition of Table S1, the numbering of the original supplementary tables has been adjusted accordingly (the original Tables S2–S4 have been renumbered as Tables S2–S4, with the original Table S2 now becoming Table S3, and so on). The revised Supplementary Materials statement has been updated accordingly in the main text.

Supplementary Materials/Data Availability: Are there shapefiles (or geopackages) of the multi-temporal lake inventory from the Supplementary Materials that could be shared? It’s great to have the data in tables, but it would be cumbersome to use the inventory for a follow-up study. The impact of this study could increase with the availability of these data in geospatial format (as well as in tables).

——We thank the reviewer for this valuable suggestion. We fully agree that while providing the glacial lake inventory in tabular format facilitates quick reference, making the data available in geospatial formats such as Shapefile or GeoPackage would be more convenient for follow‑up studies and would enhance the reusability of our data. To this end, we have created a dataset in the Mendeley Data open repository and have generated a reserved DOI. The dataset is titled: "Multi‑temporal glacial lake inventory of the southeastern Tibetan engineering corridor (1990–2020)" (DOI: 10.17632/xt5t28zym8.1, currently in a pre‑release state and accessible only for review purposes). The dataset includes: glacial lake inventory Shapefiles for four epochs (1990, 2000, 2010, and 2020), containing attribute information such as lake ID, type (proglacial/supraglacial/detached), area, perimeter, centroid coordinates, and elevation for each lake; and the 2020 GLOF susceptibility assessment results Shapefile, including susceptibility levels and indicator scores for each lake. The temporary review link to the dataset is: https://data.mendeley.com/preview/xt5t28zym8?a=7d529408-94bd-41b2-888b-25cfc56073e0. The dataset will be made publicly available immediately upon formal acceptance of the paper and updated with the official DOI link. We have updated the Data Availability Statement in the revised manuscript to reflect the access information, enabling readers to directly access and use the data. We believe that open sharing of these data in geospatial formats will significantly enhance the reproducibility and academic impact of our study.

Author Response File: Author Response.pdf

Reviewer 4 Report

Comments and Suggestions for Authors

See report

Comments for author File: Comments.pdf

Author Response

GeoHazards

Re: revision on manuscript #geohazards-4452106

Dear Editors and Reviewers:

We would like to express our gratitude for the valuable comments and suggestions raised by the reviewer on our manuscript:

Evolution of Glacial Lakes and GLOF Risks to Transportation Routes in the Southeastern Tibetan Engineering Corridor

Authors’ response to reviewer’s comments

Reviewer 4:

General Assessment

The topic is squarely within scope for a geohazards journal with a practitioner-facing, infrastructure-oriented readership: a rapidly growing population of glacial lakes sits directly upstream of highways, a railway under construction, and a hydropower transmission corridor, in a setting of active tectonics and monsoon-influenced climate. The 30-year, four-epoch, dual-sensor dataset is a genuine asset, and explicitly tying susceptibility and expansion rate to specific pieces of infrastructure — rather than treating lakes in the abstract — is a welcome, applied framing. The cascading/clustering analysis is a valuable and, to my knowledge, under-explored angle in the regional literature. However, several methodological steps underpinning the manuscript's headline, infrastructure-relevant numbers are not sufficiently transparent or statistically supported for this level of venue.

——We thank the reviewer for the strong affirmation of the significance and applied value of our research topic. We fully accept the reviewer's criticism regarding the lack of transparency and statistical support in the original manuscript. In response, we have made systematic improvements in the revised manuscript, as detailed in our point‑by‑point replies below.

Major Comments

MC1.The four indicators (lake volume, distance from parent glacier, parent-glacier area, elevation) are combined with weights 0.4/0.25/0.25/0.1, but the manuscript does not explain how these specific values were obtained. Please state the statistical/analytical procedure used (correlation strength, PCA loadings, AHP/expert elicitation, regression-based weighting, etc.) and, ideally, show a sensitivity analysis of how the "very high"/"high" classification — which drives the infrastructure-risk conclusions — would shift under alternative, defensible weight sets.

——We thank the reviewer for raising this critical issue. We fully agree that the basis for weight assignment and sensitivity analysis are central to methodological transparency and must be clearly explained. In the revised manuscript, we have supplemented and improved the following three aspects:

(1) Basis for weight assignment. The weights of the four indicators were determined based on a comprehensive consideration of statistical discriminability, physical contribution, and literature conventions. In terms of statistical discriminability, we employed the Mann–Whitney U test to compare the differences between outburst and stable lakes for each indicator (Supplementary Table S2). The results show that lake volume has the highest discriminative power (U = 1568.0, Z = 4.702, p < 0.001), followed by distance to the parent glacier (U = 2537.0, Z = 2.248, p = 0.025) and parent glacier area (U = 2350.5, Z = 2.713, p = 0.007), while elevation shows the lowest discriminative power (p = 0.304). Accordingly, the weights were assigned in decreasing order. In terms of physical contribution, lake volume directly determines the potential flood magnitude; distance to the parent glacier and parent glacier area control the likelihood and magnitude of ice avalanches entering the lake; whereas elevation primarily influences the potential energy conversion process of outburst floods and plays a relatively indirect role. Meanwhile, the weight assignments also drew upon literature conventions in regional GLOF susceptibility assessments [30,54,59] and were locally adjusted based on the characteristics of the 12 historical outburst events in the study area.

(2) Approach to weight assignment. It should be noted that we did not employ quantitative methods such as PCA, AHP, or regression analysis to determine the weights. Instead, we adopted a comprehensive approach based on "statistical discriminability + physical contribution + literature conventions." This decision was made because: (i) the limited number of historical outburst events (n = 12) in this study does not satisfy the sample size requirements for PCA or regression analysis (which typically require a sample size at least 5–10 times the number of variables); (ii) AHP relies on expert judgment, whereas our weight assignment is primarily based on statistical test results, reducing the influence of subjectivity; and (iii) statistical discriminability directly reflects the ability of each indicator to indicate outburst events and provides the most objective quantitative basis for weight assignment. Therefore, while our weight assignment does contain a certain degree of subjective judgment, its core foundation remains the statistical test results.

(3) Sensitivity analysis. To test the robustness of the weight assignment scheme, we have added a sensitivity analysis in the revised manuscript (Supplementary Table S6). We designed six alternative weight scenarios, systematically varying each indicator weight within ±20% of the baseline values, and recalculated the susceptibility scores and the number of high and very high susceptibility lakes for each scenario. The results show that under all alternative weight scenarios, the number of high and very high susceptibility lakes remains stable between 77 and 81 (compared to 79 under the baseline scheme), with no significant changes in spatial patterns. This indicates that the assessment results are insensitive to reasonable variations in weight assignments, demonstrating the robustness of the weighting scheme, and confirming that the core conclusions of this study are not altered by minor adjustments in weights.

The above revisions have been made in Section 2.3 of the revised manuscript, and the sensitivity analysis results are presented in Supplementary Table S6.

MC2.The Discussion states that validation against independent outburst data "is not necessary… as the selection of GLOF sensitivity assessment indicators is based on the analysis of historical GLOF events." This does not substitute for validation: the same 12 events were used to select and calibrate the indicators, and the model's claimed value is precisely its transferability to currently-stable lakes. Given the small sample, at minimum a leave-one-out cross-validation is recommended.

——We thank the reviewer for raising this critical issue. We fully agree with the reviewer's view that validation was lacking in our study. The original statement that "validation is not necessary" was indeed inappropriate, and we have thoroughly removed it from the revised manuscript. To validate the effectiveness of our susceptibility assessment framework, we have supplemented the following validation efforts in the revised manuscript:

(1) Back‑testing. We applied the evaluation system to the four historically documented GLOF lakes within the study area for back‑testing. The results show that Guangxie Co, Lumu Co, and Anjue Zhuqulong glacier lake were all classified as having "very high" susceptibility, and Zhuxigou glacial Lake was classified as having "high" susceptibility. All historically documented GLOF lakes were correctly identified as either "high" or "very high" susceptibility, preliminarily confirming that our assessment framework can effectively distinguish lakes prone to outburst.

(2) Limitations of the validation. We have explicitly acknowledged in the Discussion that, given the limited number of historical GLOF events in our study area (only four cases), this validation remains a back‑testing exercise and cannot fully substitute for independent statistical validation. We agree with the reviewer's suggestion regarding leave‑one‑out cross‑validation—this method is indeed worth applying in a more complete framework; however, constrained by the current sample size, the results may not be sufficiently robust. As more GLOF records accumulate in the future, we will further improve the validation work.

The above revisions have been made in Section 4.1 (Discussion) of the revised manuscript, ensuring that the conclusions of the paper are strictly based on verifiable analytical results. The revised text is as follows:

“Validating GLOF susceptibility assessments remains challenging due to the com-plex and stochastic nature of GLOF triggers [30,43,59]. Nevertheless, to test the performance of our assessment framework, we back-tested it against historically documented GLOF lakes within the study area. The results show that Guangxie Co, Lumu Co, and Anjue Zhuqulong glacier lake were all classified as having "very high" susceptibility, while Zhuxigou glacial Lake was classified as "high" susceptibility. All documented historical outburst lakes were correctly rated as either "high" or "very high" susceptibility, supporting the capability of our framework to distinguish lakes prone to outburst. It should be noted, however, that given the limited number of historical GLOF events in the study area, this validation remains a back-testing exercise; fully independent statistical validation will require the accumulation of more GLOF records in the future. In addition, assessment indicators and methods inevitably vary across studies due to regional differences [32,43,54]. Nonetheless, two previous GLOF susceptibility assessments partially overlap with our study area, and comparing our results with these studies provides a more comprehensive regional perspective on GLOF risks.”

MC3. Group differences in Figure 2 (volume, elevation, distance, parent-glacier area) are called "significantly differed" without a test statistic; with n = 12 outburst lakes, a non-parametric test (e.g. Mann–Whitney U) with exact p-values is needed. Separately, the precipitation trend in Figure 12 (~5 mm/decade) corresponds to R² = 0.008 — essentially no linear explanatory power — yet is described as "an increasing trend" (line 403). A formal trend test (e.g. Mann–Kendall, which handles autocorrelated hydro-climatic series better than OLS R²) should be applied to both series before either is characterized as significant; the temperature series (R² = 0.411) is comparatively more defensible but would still benefit from this.

——We thank the reviewer for pointing out the deficiencies in the statistical testing. We fully accept the reviewer's criticism and have systematically revised the manuscript from the following two aspects:

(1) Group difference testing for Figure 2. We have performed the Mann–Whitney U non‑parametric test to evaluate the significance of differences between outburst lakes and stable lakes across all indicators. The results show that lake volume (U = 1568.0, Z = 4.702, p < 0.001), distance to the parent glacier (U = 2537.0, Z = 2.248, p = 0.025), and parent glacier area (U = 2350.5, Z = 2.713, p = 0.007) differ significantly between the two groups. In contrast, elevation (p = 0.304), glacier tongue slope (p = 0.344), slope near the lake (p = 0.084), air temperature (p = 0.612), and precipitation (p = 0.302) do not reach statistical significance. The complete statistical results (including sample sizes, means, standard deviations, U values, Z values, and exact p‑values) are summarized in Supplementary Table S2 and are described in detail in Section 2.3 of the revised manuscript. The caption of Figure 2 has also been updated accordingly. The revised text in the main body is as follows:

“To statistically evaluate the differences between historically outburst lakes and stable lakes across the candidate indicators, we employed the Mann–Whitney U non‑parametric test (Supplementary Table S2). The results show that lake volume (U = 1568.0, Z = 4.702, p < 0.001), distance to the parent glacier (U = 2537.0, Z = 2.248, p = 0.025), and parent glacier area (U = 2350.5, Z = 2.713, p = 0.007) differ significantly between the two groups. In contrast, no statistically significant differences were de-tected for elevation, glacier tongue slope, slope near the lake, air temperature, or pre-cipitation (p ≥ 0.05). Although elevation did not show a statistically significant differ-ence between outburst and stable lakes in our dataset, we nonetheless retained it as a fourth indicator in the susceptibility assessment. Elevation is widely used as an im-portant factor in GLOF susceptibility assessments [30,54,59]. Higher‑altitude lakes are typically situated in steeper terrain, where the substantial elevation drop would sig-nificantly amplify flood velocity and reach in the event of an outburst [43]. Meanwhile, high‑altitude regions are more sensitive to climate warming, with accelerated glacier retreat and more frequent ice avalanches indirectly increasing the likelihood of dam failure [44,45]. Given the well‑established physical importance of elevation and its common use in previous GLOF assessments, we assigned it a lower weight (0.1) in the susceptibility score to reflect its relatively weaker discriminative ability in this regional context, while still accounting for its potential contribution to failure susceptibility. In summary, four indicators, namely lake volume, distance to the parent glacier, parent glacier area, and elevation, were selected for GLOF susceptibility assessment in this study.”

Table S2. Comparison of indicators between stable and historically outburst glacial lakes

Indicator	All lakes (n = 257) Mean ± SD	Outburst lakes (n = 12) Mean ± SD	Mann-Whitney U	Z	p-value	Significance
Lake volume (×10⁶m³)	3.56 ± 13.72	4.71 ± 4.20	1568.0	4.702	< 0.001	***
Distance from the parent glacier (m)	1786.43 ± 2217.68	581.08 ± 828.92	2537.0	2.248	0.025	*
The area of the parent glacier (km²)	3.76 ± 9.97	5.68 ± 8.32	2350.5	2.713	0.007	**
Elevation (m)	4350.20 ± 441.69	4539.04 ± 412.57	2830.5	1.027	0.304	n.s.
Glacier tongue slope (°)	17.43 ± 9.30	14.15 ± 5.36	2839.0	0.947	0.344	n.s.
Slope gradient near the lake (°)	29.79 ± 14.75	31.62 ± 4.58	2616.0	1.729	0.084	n.s.
Temperature (°C)	-4.10 ± 3.09	-4.48 ± 3.82	2939.5	0.507	0.612	n.s.
Precipitation (mm)	714.89 ± 52.78	664.09 ± 103.68	2863.5	1.032	0.302	n.s.

(2) Climate trend analysis. Following the reviewer's suggestion, we have re‑analyzed the temperature and precipitation trends using the Mann–Kendall test and Sen's Slope estimator, instead of the previous OLS regression. The temperature trend is significant at the 0.001 level (Z = 6.65, p < 0.001, Sen's Slope = 0.045 °C yr⁻¹). The precipitation trend does not reach statistical significance (Z = 0.24, p = 0.81, Sen's Slope = 0.03 mm yr⁻¹). Accordingly, we have removed the statement that "precipitation also shows an increasing trend" from the revised manuscript, and explicitly state in the Discussion that precipitation change makes no statistically significant contribution to lake expansion. The caption of Figure 12 has also been updated accordingly. The revised text is as follows:

“To detect long‑term trends in the climatic time series, we employed the Mann–Kendall non‑parametric trend test to evaluate statistical significance and the Sen's Slope esti-mator to quantify trend magnitudes. The results reveal a highly significant warming trend in air temperature (Z=6.65, p<0.001), with a Sen's Slope of 0.045 °C/yr. In con-trast, the precipitation trend does not reach statistical significance (Z=0.24, p ≥ 0.05), indicating that interannual variability, rather than a long‑term directional change, dominates the precipitation regime. Climate change exerts direct influences on the hy-drological balance of glacial lakes. The observed lake expansion is primarily attributa-ble to regional warming, which accelerates glacier and snow melt and thereby supplies increased meltwater to glacial lakes [34,58,67]. No statistically significant contribution from precipitation changes to lake expansion was identified in this study.”

MC4.The headline cascading figures (e.g. a maximum combined volume of 452.36 × 10⁶ m³, a "+910.3%" increase) are obtained by summing individual lake volumes along a shared flow path. Stored volume is not equivalent to downstream hazard: GLOF destructiveness depends primarily on peak discharge, breach dynamics, and the fraction of stored water released (rarely 100%), not on total volume alone. Please state explicitly what release fraction is assumed, and either apply an established peak-discharge scaling relationship or frame the volume figures clearly as an upper-bound screening indicator rather than a hazard-magnitude estimate.

——We thank the reviewer for raising this critical issue. We fully agree with the reviewer's viewpoint that stored volume is not equivalent to downstream hazard intensity, and that GLOF destructiveness depends primarily on peak discharge, breach dynamics, and the fraction of released water, rather than total volume alone. We have carefully re‑examined the description of cascading volumes in the original text and have revised it following the reviewer's suggested second approach—namely, to clearly frame the volume figures as an upper‑bound screening indicator rather than a hazard‑magnitude estimate.

Specifically, we have revised the original statement that "the total water volume could reach a maximum of 452.36 × 10⁶ m³" to "the upper‑bound estimate of the total water volume could reach approximately 452 × 10⁶ m³," and have explicitly added the following clarification: "This value should be considered a conservative upper‑limit estimate for identifying potential high‑risk clusters, rather than an accurate prediction of the actual size of outburst floods." The purpose of this revision is twofold: first, to clarify that 452 × 10⁶ m³ is a static sum of lake storage volumes, without accounting for the fact that only a fraction of the stored water is typically released during an actual breach event; and second, to emphasize that this value serves to identify priority high‑risk clusters for further investigation, rather than to provide a quantitative prediction of actual flood magnitude or impact extent. As an upper‑bound screening indicator, it effectively helps to identify "which clusters merit priority for detailed hydrodynamic modeling," which is precisely the greatest value this study can offer for future research.

We are fully aware that accurate peak discharge estimation and breach dynamics simulation require detailed hydrodynamic modeling, including breach development processes, dam material parameters, and downstream channel cross‑sections and roughness data. These efforts fall beyond the scope of the current study but have been clearly identified as a priority direction for our future research. In the revised Discussion (Section 4.5), we have further emphasized this need, stating that future work should integrate high‑resolution topographic data, field channel cross‑section surveys, and breach parameter calibration to conduct systematic hydrodynamic modeling and quantitative hazard assessment of glacial lake outburst floods, thereby providing a more reliable scientific basis for early‑warning system design.

The above revisions have been made in Sections 4.3 and 4.5 of the revised manuscript.

MC5.The three cascading channels are identified only in the Discussion; no criterion (distance threshold, DEM-based flow routing, manual photo-interpretation) is defined in Section 2. For what is arguably the paper's most novel contribution, the detection method needs to be explicit and reproducible, ideally as its own Methods subsection.

——We thank the reviewer for raising this critical issue. We fully agree that the identification of cascading channels is one of the most novel contributions of this study, and the detection method must be clearly described and reproducible in the Methods section. To this end, we have added a new subsection (Section 2.4, "Identification of Cascading Lake Clusters") in the Methods section of the revised manuscript, in which we define in detail the three criteria for identifying cascading clusters: (1) Spatial connectivity: based on the flow direction extracted from the Copernicus DEM30 (30 m resolution), three or more glacial lakes with moderate or higher susceptibility were considered potentially connected if they are aligned along the same drainage line and if the downstream lakes lie within the potential outburst flood path of the upstream lake. (2) Distance threshold: the straight‑line distance between the outlet of the upstream lake and the shoreline of the downstream lake was required to be less than 5 km, ensuring that flood propagation remains sufficiently concentrated to trigger cascading failures. (3) Manual image interpretation: all candidate clusters identified through the above criteria were further validated by manual visual interpretation of high‑resolution satellite imagery to exclude topographic barriers (e.g., ridges, moraine mounds) that might block flood propagation, thereby ensuring that the connectivity assumptions are realistic. Based on the above criteria, we identified three cascading clusters within the study area, which are analyzed in detail in Section 4.3. The above revisions have been made in Section 2.4 of the revised manuscript, and the complete revised text is as follows:

“2.4. Identification of Cascading Lake Clusters

On the basis of individual lake susceptibility assessment, we further identified po-tential cascading lake clusters, defined as clusters in which an outburst from an up-stream lake may trigger successive failures of multiple downstream lakes along the same flow path. The identification was based on the following criteria. First, based on the flow direction extracted from the Copernicus DEM30 (30 m resolution), three or more glacial lakes with moderate or higher susceptibility were considered potentially connected if they are aligned along the same drainage line and if the downstream lakes lie within the potential outburst flood path of the upstream lake. On this basis, the straight‑line distance between the outlet (or dam toe) of the upstream lake and the shoreline of the downstream lake was required to be less than 5 km, ensuring that flood propagation remains sufficiently concentrated to trigger cascading failures. Finally, all candidate clusters identified through the above criteria were further validated by manual visual interpretation of satellite imagery to exclude topographic barriers (e.g., ridges, moraine mounds) that might block flood propagation, thereby ensuring that the connectivity assumptions are realistic.”

MC6.Table 2 does not state the units of A and V for any of the five formulas listed — a reproducibility gap in itself. More importantly, the single formula adopted (No. 5, calibrated on 47 Himalayan lakes) is applied without propagating its calibration uncertainty into the reported volumes, including the cascading-flood totals in Section 4.3, among the manuscript's most policy-relevant numbers. A sensitivity analysis using 2–3 of the Table 2 formulas (or a reported plausible range) would substantially strengthen confidence here.

——We thank the reviewer for pointing out the missing units in Table 2 and the insufficient uncertainty analysis in the volume estimation. We have revised Table 2 to clearly indicate the units of the variables (A: km², V: ×10⁶ m³, D_m: m), and have added in Section 2.3 an explanation of why Formula 5 was selected. In addition, we calculated the volume estimation results for all glacial lakes in the study area using each of the formulas, and found that the estimates do indeed show considerable variation across formulas. We attribute this variation primarily to the fact that empirical area–volume relationships are influenced by regional factors, and formulas developed in different regions differ substantially in both mathematical form and parameter values. In essence, there is no universally applicable formula. Therefore, when dealing with these empirical formulas, our task is not to reconcile their numerical differences, but rather to select the most appropriate formula based on the characteristics of the study area, while maintaining a clear awareness of the differences among them. The reasons for selecting Formula F5 are as follows:

First, the calibration dataset of F5 is the most similar to our study area. This formula was established using measured bathymetric data from 47 Himalayan lakes, representing the largest sample size (n = 47) among currently available area–volume relationships in the Himalaya. As an important part of the eastern Himalayan syntaxis, the southeastern Tibetan region shares strong similarities with the main Himalayan region in terms of glacier type (maritime glaciers), climatic conditions (monsoon influence), and lake genesis (moraine‑dammed lakes). Although F2 (Fujita et al., 2013) was also based on Himalayan data, its calibration sample size is relatively small and it assumes a fixed power‑law relationship between mean depth and area—an assumption that may introduce systematic bias for lakes with a wide range of areas. F4 (Sharma et al., 2018) is also based on Himalayan data, but its calibration samples are predominantly small to medium‑sized lakes, with limited coverage. Therefore, among the three Himalayan‑based formulas (F2, F4, and F5), F5 has clear advantages in both calibration sample size and coverage.

Second, F5 has been widely applied and validated in GLOF studies in the Himalaya. Since its publication, this formula has been used in multiple GLOF susceptibility assessments across the Third Pole region, and its consistency with measured data has been verified in several independent studies.

Third, F1 (Huggel et al., 2002) was based on data from the Swiss Alps, where the lakes are generally smaller in scale and the glacier type (continental glaciers) differs fundamentally from that of the southeastern Tibetan region (maritime glaciers), making it unsuitable for our study area. F3 (Patel et al., 2017) was based on data from the western Himalaya and adopts a linear assumption that may be applicable to small lakes, but for samples with a wide range of areas, the linear relationship may not accurately describe the actual depth–area relationship.

In summary, the numerical differences among the five formulas reflect the regional dependence of empirical relationships. F5 is the most regionally applicable formula for this study, and we therefore selected it as the volume estimation approach for this research.

The above revisions have been made in Section 2.3. We hope the revised manuscript meets the reviewer's expectations, and we welcome any further guidance.

MC7.The manuscript is framed around "GLOF Risks," but the modelled quantity is susceptibility (a hazard-side property of the lakes), combined only with proximity-to-road as a partial exposure proxy. No treatment of infrastructure vulnerability (bridge/culvert type and condition, embankment resilience, traffic exposure) is included — the standard third component of risk in the disaster-risk-reduction framing this topic invites. Recommend either adjusting terminology consistently to "susceptibility"/"hazard," or explicitly scoping what the risk framing does and does not cover.

——We thank the reviewer for pointing out the terminological inaccuracy. We fully agree that the quantity modeled in this study is essentially susceptibility rather than full risk. We have carefully reviewed the entire manuscript and systematically revised the terminology throughout.

(1) We have unified the terminology for the modeled results throughout the manuscript. We have revised "risk levels" to "susceptibility levels" and "high risk" to "high susceptibility" in the Methods, Results, and Discussion sections to accurately reflect what our assessment evaluates.

(2) We have retained the use of "risk" in the broader context. In the Abstract and Introduction, we keep the description of "GLOF risk" as research background, but we have made it clear that these descriptions refer to the research motivation rather than to the results of our modeling.

(3) We have explicitly defined the scope of this study. In the limitations section of the Discussion (Section 4.5), we have added a clarification: our assessment framework covers susceptibility and partial exposure (distance to roads as a proxy for exposure), but does not include infrastructure vulnerability (e.g., bridge/culvert types and conditions, embankment resilience, traffic exposure). Therefore, our identification of "lakes threatening roads" should be understood as an indication of potential hazard exposure rather than a quantitative risk estimate.

(4) Following the reviewer's suggestion, we have changed "Risks" to "Hazards" in the article title to more accurately reflect the scope of our research.

The above revisions have been made throughout the manuscript (title, Abstract, Methods, Results, Discussion, and Limitations).

MC8. Section 3.1.2 reports a "total change in area… of 7.787 km²" (Figure 5: 7.689 km² expansion + 0.098 km² shrinkage). The year-total areas given elsewhere (Figure 6b: 30.024 km² in 1990, 37.89 km² in 2020) imply a net change of 7.866 km², and even the net implied by Figure 5 alone (7.689 − 0.098 = 7.591 km²) does not match either figure. The discrepancy is small in relative terms (~3–4%) but should be reconciled, particularly since the Abstract's 26.2% area-growth figure appears to use the 7.866 km² value rather than the 7.787 km² value quoted in the body text.

——We thank the reviewer for identifying this data inconsistency. Upon careful checking, we found that the discrepancy arises from ambiguous definitions of the "total change in area": the 7.787 km² value represents the gross change including expansion, shrinkage, and newly formed lakes, whereas the 7.866 km² value is the net change derived directly from the difference between the 1990 and 2020 total areas. To ensure consistency and clarity, we have unified the net change to 7.866 km² (26.2%) throughout the manuscript and removed the ambiguous 7.787 km² figure. The Abstract's 26.2% remains unchanged, as it is calculated as 7.866/30.024, which is fully consistent with the other data. The revisions have been made in Section 3.1.2 of the manuscript.

Minor and Technical Comments
0.1 seems low relative to the narrative emphasis the indicator receives (lines 199–205, including a dedicated discussion of Figure 2c). A brief justification would help.

——We thank the reviewer for this careful observation. We agree that elevation receives relatively extensive discussion in the text despite its weight of only 0.1. It should be noted that the weight of 0.1 was determined based on statistical discriminability—the Mann–Whitney U test shows that elevation does not differ significantly between outburst and stable lakes (p = 0.304), making it the least discriminative among the four selected indicators and therefore assigned the lowest weight. The extensive discussion of elevation in the text does not arise from its statistical importance, but precisely because we need to explain to readers why a statistically non‑significant indicator was nonetheless retained in the assessment framework—which inherently requires more textual explanation. We justified the retention of this indicator from two perspectives: physical mechanisms (higher‑altitude lakes are typically situated in steeper terrain, where a large elevation drop would significantly amplify flood damage in the event of an outburst; high‑altitude regions are also more sensitive to climate warming, with more frequent glacier retreat and ice avalanche activity) and literature consensus, while assigning it a low weight to control its over‑influence on the assessment results. In addition, the sensitivity analysis (Supplementary Table S6) shows that when the elevation weight varies within ±20%, the number and spatial pattern of high and very high susceptibility lakes do not change significantly, further demonstrating that the low weight of 0.1 is reasonable and robust. We have added a brief clarification in Section 2.3 of the revised manuscript to address any potential confusion on the part of readers.

Lakes are mapped down to 0.005 km² (n = 257) but susceptibility is assessed only above 0.01 km² (n = 238); please state explicitly how many lakes this excludes.

——We thank the reviewer for pointing out this issue. In the revised manuscript, we have explicitly stated that the 0.01 km² area threshold excluded 19 lakes (accounting for 7.4% of the 257 mapped lakes), with areas ranging from 0.005 to 0.01 km². These lakes collectively account for less than 1.5% of the total glacial lake area and are considered to pose negligible GLOF threat due to their limited water storage capacity. The revised paragraph in the original text is as follows:

“Furthermore, to more effectively focus on vulnerable glacial lakes and mitigate the impact of redundant data, this study conducted GLOF susceptibility assessments on glacial lakes with an area greater than 0.01 km². This threshold excluded 19 lakes with areas between 0.005 km² and 0.01 km², which collectively account for less than 1.5% of the total glacial lake area and are considered to pose negligible GLOF hazard due to their limited water storage capacity [30,54].”

The text's 2.47 lakes/yr and 0.26 km²/yr read as simple endpoint-to-endpoint averages, while Figure 6a/6b report different OLS-fitted slopes (2.24 lakes/yr, R² = 0.757; 0.24 km²/yr, R² = 0.913). Please clarify which measure is used where.

——We thank the reviewer for pointing out this inconsistency in data reporting. Our original text indeed suffered from insufficient clarity on this point. The two sets of values were calculated using different methods, but we did not clearly distinguish between them in the text, which may have caused confusion for readers. We have now clarified and unified the presentation in the revised manuscript.

In this study, the growth rates of glacial lake number and area were calculated using two approaches:

The endpoint‑average rate: calculated directly from the difference between the 1990 and 2020 values divided by 30 years, reflecting the average rate of change over the entire study period. The values 2.47 lakes/yr and 0.26 km²/yr fall into this category and are used in the Abstract and main text to summarize the overall trend.

The OLS‑fitted slope: derived from linear regression using all four epochs (1990, 2000, 2010, and 2020) to fit the long‑term trend, with R² reported to assess the robustness of the trend. The values 2.24 lakes/yr (R² = 0.757) and 0.24 km²/yr (R² = 0.913) belong to this category and are used in Figure 6 to present the long‑term trends.

In response to this comment, the revised text in the original manuscript is as follows:

“From 1990 to 2020, both the number and area of glacial lakes in the study area showed an increasing trend (Figure 7). The number of glacial lakes increased from 183 to 257, with an average rate of 2.47 lakes per year over the 30‑year period (Figure 7a), while the total area increased from 30.024 km² to 37.89 km², with an average rate of 0.26 km² per year (Figure 7b). The linear‑fitted trends across the four epochs are 2.24 lakes/yr (R² = 0.757) and 0.24 km²/yr (R² = 0.913), as shown in Figure 7.”

"24 lakes corresponded with our evaluation results" would be more informative as a percentage agreement (≈80% for ref. [54]) or a confusion matrix across susceptibility classes.

——We thank the reviewer for this suggestion. In the revised manuscript, we have replaced the vague statement with a clear percentage agreement rate. The revised text is as follows:

“Additionally, the research by [54] examined glacial lake identification and susceptibil-ity assessment along the Sichuan–Tibet Railway. Within the area overlapping with our study, they identified a total of 30 glacial lakes with high or very high susceptibility. Of these, 24 were also classified as high or very high susceptibility in our study, indicating a high level of agreement (Consistency is 80%) between the two studies in the overlap-ping region, which further supports the reliability of our assessment results.”

This revision makes the presentation of the validation results more transparent and quantitative, and also helps readers to assess the reliability differences between this study and previous studies.

Please state which Landsat sensor (5/7/8/9) supplied each epoch, and how Landsat 7's SLC-off striping (post-2003) and the region's persistent cloud cover (noted as a defining regional trait, line 57) were handled — especially for the smallest lakes near the 0.005 km² threshold, which drive much of the reported number-based growth (Figure 6c).

——We thank the reviewer for pointing out the missing methodological details. We have supplemented the relevant information in the revised manuscript. (1) The specific Landsat sensor types used for each epoch are listed in the newly added Supplementary Table S1, and the data sources are reported in the main text (Section 2.2). (2) Regarding the SLC-off issue of Landsat 7 ETM+ which began on 31 May 2003, all Landsat 7 images used in the 2000 epoch were acquired in 2000, prior to the onset of the SLC-off problem, and thus are not affected by striping. For the 2010 epoch, we did not use Landsat 7 ETM+ images; instead, we used Landsat 5 TM images (Landsat 5 retired at the end of 2011, and data for 2010 were available), thereby completely avoiding the SLC-off issue. The 1990 epoch used Landsat 5 TM, and the 2020 epoch used Landsat 8 OLI; both are unaffected by the SLC-off problem. (3) Regarding cloud cover, the study area is located on the southeastern margin of the Tibetan Plateau, where cloud cover is common due to monsoon influence. To obtain high‑quality imagery, we adopted the following strategies: ① time‑window expansion: we applied a ±1‑year time window around each target year (e.g., the 1990 epoch covered 1989–1991) to increase the availability of low‑cloud imagery (Table S1); ② cloud‑cover threshold filtering: we preferentially selected images with cloud cover below 10%; when such images were insufficient, the threshold was relaxed to 30%, followed by visual inspection to exclude scenes with cloud cover over lake areas; ③ manual visual inspection: all candidate images were individually examined to ensure that the selected images were free of cloud cover over lake areas, or contained only thin clouds that did not affect lake boundary delineation. (4) Given that a relatively large number of small lakes near the 0.005 km² threshold (19 out of 257 mapped lakes are smaller than 0.01 km²) contribute significantly to the reported increase in lake number, we adopted the following measures to minimize the impact of image quality and cloud cover on their identification: ① all lake boundaries were manually delineated and inspected on a case‑by‑case basis, rather than relying solely on automated algorithms; ② for areas potentially affected by cloud cover or striping, we excluded those scenes or cross‑validated using adjacent epochs or alternative Path/Row scenes.

The above content has been supplemented in Section 2.2 of the revised manuscript, and the detailed image list and sensor information are provided in Supplementary Table S1.

Table S1. Satellite scenes used in this study

Epoch / Year	Sensor	Scene ID	Acquisition Date	Path/Row	Cloud Cover (%)	Primary Use
1990	Landsat 5 TM	LT05_134039_19891014	1989/10/14	134/39	13	Lake mapping
		LT05_134039_19891030	1989/10/30		3
		LT05_134039_19900627	1990/6/27		8
		LT05_134040_19891014	1989/10/14	134/40	18
		LT05_134040_19891030	1989/10/30		3
		LT05_135040_19891021	1989/10/21		22
2000	Landsat 7 ETM+	LE07_134039_20000716	2000/7/16	134/39	14	Lake mapping
		LE07_134039_20001020	2000/10/20		9
		LE07_134039_20010703	2001/7/3		12
		LE07_134039_20011023	2001/10/23		3
		LE07_134040_20010703	2001/7/3	134/40	19
		LE07_134040_20011023	2001/10/23		5
		LE07_135040_19991025	1999/10/25		8
2010	Landsat 5 TM	LT05_136039_20091019	2009/10/19	136/39	20	Lake mapping
		LT05_134039_20110808	2011/8/8	134/39	27
		LT05_134040_20111011	2011/10/11	134/40	14
		LT05_135040_20090926	2009/9/26	135/40	15
		LT05_135039_20110831	2011/8/31	135/39	24
2020	Landsat 8 OLI	LC08_134039_20191017	2019/10/17	134/39	8.03	Lake mapping
		LC08_134039_20200901	2020/9/1		15.06
		LC08_134039_20210920	2021/9/20		7.78
		LC08_134040_20210920	2021/9/20	134/40	10.44
		LC08_135039_20190602	2019/6/2	135/39	12.07
		LC08_135040_20211013	2021/10/13	135/40	10.73
		LC08_136039_20210918	2021/9/18	136/39	5.77

Note: The satellite scenes used in this study fall into two categories: Landsat imagery for glacial lake mapping across the four epochs (1990, 2000, 2010, and 2020), and recent high‑resolution Sentinel‑2 and Landsat 9 imagery (2024–2026) used as base maps for the figures. The Landsat images were sourced from the USGS Collection 2 Level‑2 Surface Reflectance products (https://earthexplorer.usgs.gov/), and the Sentinel-2 images from the ESA Copernicus programme (https://scihub.copernicus.eu/); all data are also accessible via the Google Earth Engine platform (https://earthengine.google.com/). Image selection prioritized summer–autumn months (June–October) to avoid misidentification caused by lake ice and snow cover, and applied a ±1‑year time window around each target year to acquire high‑quality imagery with cloud cover below 30%.

Copernicus DEM30 appears in the Figure 3 workflow but is not described in the Methods text; given several flood-pathway slopes in Section 3.2.3 are computed over short (5–18 km), steep distances, its resolution/vertical accuracy should be discussed.

——We thank the reviewer for pointing out this omission. We agree that the Copernicus DEM30 appears in the Figure 3 workflow but is not described in the Methods section, which does affect methodological transparency and reproducibility. We have added the following information at the end of Section 2.2 in the revised manuscript: (1) Data source and resolution. This study used the Copernicus DEM 30 m (COP‑DEM30) global elevation model (https://dataspace.copernicus.eu/), with a spatial resolution of approximately 30 m. According to the Copernicus DEM Product Handbook, the dataset has a vertical accuracy of less than 4 m (absolute, 90% linear error) and less than 2 m (relative, for slopes ≤ 20%) in flat to moderately sloping terrain. The revised paragraph in Section 2.2 is as follows:

“For topographic analyses, including lake elevation extraction and flood‑pathway slope calculations, we used the Copernicus DEM 30 m (COP‑DEM30) global elevation model, available at https://dataspace.copernicus.eu/. This dataset provides a spatial resolution of approximately 30 m and a vertical accuracy of < 4 m (absolute) and < 2 m (relative) according to the Copernicus DEM Product Handbook.”

Translating the distances and average slopes in Section 3.2.3/Figure 11 into approximate flood travel times to the nearest highway — feasible with the same DEM/flow-path data already used — would materially increase the paper's value for early-warning system design, which is the manuscript's own stated motivation.

——We thank the reviewer for this highly constructive suggestion. We fully agree that converting distances and slopes into flood arrival times would significantly enhance the practical value of the paper for early‑warning system design, which is indeed closely aligned with the original motivation of this study. However, after careful consideration, we have decided not to include flood travel time estimates in this paper, based on the following considerations:

Although a first‑order travel time estimation using the existing DEM and flow path data is technically feasible, such estimation relies heavily on parameter assumptions (e.g., Manning's roughness coefficient, hydraulic radius, flow regime, etc.), and we currently lack field survey data on river channel cross‑sections in the study area to constrain these parameters. Providing "arrival times" without validated parameters could create a false sense of precision. Regardless of how cautiously we word it, readers—especially engineering decision‑makers—are likely to interpret these numbers as actionable early‑warning time windows. Any deviation from reality could lead to misjudgment in early‑warning response strategies, and in extreme cases, could threaten human safety. Out of commitment to academic rigor and responsibility for hazard prevention applications, we believe that rough estimates without field‑based parameter calibration should not be presented as explicit early‑warning time information in this paper. Therefore, we have decided not to add flood arrival time estimates in this study, but instead to highlight hydrodynamic modeling and arrival time analysis as a clear direction for future research, as emphasized in the Discussion section (Section 4.4) of the revised manuscript. We plan to conduct systematic hydrodynamic modeling of glacial lake outburst floods in subsequent studies, integrating higher‑resolution topographic data, field channel cross‑section measurements, and hydrological parameter calibration, to provide a reliable quantitative basis for early‑warning system design.

We believe that this prudent approach is more consistent with the rigor required in hazard research and better avoids potential misinterpretation due to insufficient data. We hope the reviewer can understand this position.

8.Currently scattered only in the Conclusions (lines 504–507); a dedicated subsection, paired with concrete monitoring/early-warning recommendations for the priority lakes in Section 3.2.3, would strengthen the applied contribution.

——We thank the reviewer for this suggestion. We agree that the original early‑warning recommendations were scattered only in the Conclusions, which indeed did not highlight the applied value of the research. To address this, we have added a new section (Section 4.4, "Recommendations for Monitoring and Early Warning") in the revised manuscript, providing specific monitoring and early‑warning recommendations for the priority lakes. This section covers the selection of priority lakes for monitoring (the 14 glacial lakes with high expansion rates and very high susceptibility identified in Section 3.2.2, among which Guangxie Co and Lakes 8, 11, and 12 are of particular concern); the deployment of in situ water‑level sensors (including coordinated monitoring of upstream and downstream lakes in cascading clusters); a high‑resolution optical and SAR satellite remote sensing monitoring scheme; the construction of probabilistic prediction models based on historical GLOF records and lake inventory data; and recommendations for engineering protection and emergency response coordination for sections of National Highways G219 and G318 that lie within 10 km of potential outburst pathways. The above content has been integrated into Section 4.4 of the revised manuscript and is no longer scattered only in the Conclusions. The complete revised text is as follows:

“4.4. Recommendations for Monitoring and Early Warning

Based on the susceptibility assessment results and the identification of clustered lake channels, we propose the following recommendations for future monitoring and early warning. First, the 14 glacial lakes with high expansion rates and very high susceptibility identified in Section 3.2.2 (Figure 10) should be included in the priority monitoring list, among which Guangxie Co (Lake 4) and Lakes 8, 11, and 12 (Figure 11) are of particular concern due to their proximity to National Highways G318 and G219 and the active geological hazards (landslides, ice avalanches, and moraine dam erosion) in their surrounding areas [70]. For in situ observations, we recommend installing automated water-level sensors in the priority lakes to continuously monitor water-level fluctuations [71]. Water depth is one of the primary indicators of lake stability, and anomalous water-level changes may serve as important precursors to outburst events. For cascading lake clusters, sensors should be deployed in both upstream and downstream lakes simultaneously to capture the dynamic chain-reaction processes. Such systems are structurally simple, low-cost, and can be powered by solar energy, making them suitable for long-term operation in remote areas. For remote sensing monitoring, we recommend acquiring high-resolution optical and SAR satellite imagery at monthly or sub-monthly intervals to track changes in lake area, glacier retreat dynamics, and the development of hazard precursors such as ice avalanches, moraine dam erosion, and landslide activity [72]. Remote sensing data can complement in situ sensor data, providing important evidence for the early identification of outburst events. For predictive modeling, the historical GLOF records and lake inventory data compiled in this study can provide valuable prior information for constructing probabilistic prediction models. Field-observed water-level and area-change data can be dynamically assimilated into these models to enable updated assessments of outburst risk. For sections of National Highways G219 and G318 that lie within 10 km of potential outburst pathways (Table S3), we recommend reinforcing bridges and culverts along these routes to enhance their resilience against potential flood im-pacts. Where conditions permit, diversion channels or debris-flow barriers could be constructed to mitigate the effects of outburst floods. In addition, we recommend coordinating with local transportation authorities to develop targeted emergency response plans.”

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have made substantial improvements and have addressed most reviewer concerns adequately. The manuscript is significantly stronger than the original version.

Minor remaining issues to consider:

Add a dedicated "Limitations" subsection explicitly acknowledging:
- The 10-year temporal resolution and potential missed interannual variability
- Uncertainty in volume estimation
- The distinction between susceptibility and risk
- Potential seasonal misclassification
Consider adding the confusion matrix for historical events even if cross-validation is not performed
The "very high" amplification percentage (910.3%) might be better presented with a confidence range or uncertainty estimate

Author Response

GeoHazards

Re: revision on manuscript #geohazards-4452106

Dear Editors and Reviewers:

We would like to express our gratitude for the valuable comments and suggestions raised by the reviewer on our manuscript:

Evolution of Glacial Lakes and GLOF Risks to Transportation Routes in the Southeastern Tibetan Engineering Corridor

We are very grateful to the editor and the reviewers for their continued guidance. We appreciate your affirmation of our revisions in this round, and we also thank you for the remaining valuable comments that have helped us further improve the manuscript. In response to the three minor suggestions raised in this round of review, we have made revisions and provided responses point by point. The detailed point‑to‑point replies are as follows:

The authors have made substantial improvements and have addressed most reviewer concerns adequately. The manuscript is significantly stronger than the original version.

Minor remaining issues to consider:

Add a dedicated "Limitations" subsection explicitly acknowledging:

The 10-year temporal resolution and potential missed interannual variability

Uncertainty in volume estimation

The distinction between susceptibility and risk

Potential seasonal misclassification

——We thank the reviewer for this reminder. In the first round of revision, we have systematically addressed the four aspects mentioned above in Section 4.5 (Limitations) of the revised manuscript: temporal resolution (10‑year intervals), uncertainty in volume estimation, the distinction between susceptibility and risk, and potential lake misclassification caused by seasonal water‑level fluctuations. The specific content has been presented in Section 4.5 of the revised manuscript, as follows:

“4.5. Limitations

Consider adding the confusion matrix for historical events even if cross-validation is not performed

——We thank the reviewer for the further suggestion. We fully understand the intent behind visualizing the classification performance of historical events. However, given that there are only four historically documented GLOF lakes within the study area, the sample size is too small to justify a formal confusion matrix, which would appear somewhat redundant in length and might give readers the impression of "over‑presentation based on a small sample." Therefore, we have chosen to clearly describe the validation results of the four historical GLOF lakes through textual description in Section 4.1 of the revised manuscript:

The "very high" amplification percentage (910.3%) might be better presented with a confidence range or uncertainty estimate

——We thank the reviewer for this suggestion. We agree that the 910.3% amplification percentage, if presented without further explanation, could give readers a false impression of precision. In the revised manuscript, we have added clarifications regarding this percentage. First, we have revised “an increase of 910.3%” to “approximately 910%” and explicitly noted that it is calculated based on the volume estimation formula F5. Second, we clearly state that this percentage is derived from a static summation of lake volumes and represents an extreme upper‑bound scenario rather than a precise prediction of actual flood discharge. Regarding the uncertainty issue, we have explicitly defined this value as an upper‑bound estimate in the corresponding paragraph of the main text and noted that it should be understood as a conservative screening indicator rather than a precise prediction. We agree with the reviewer's view that further quantitative assessment of this uncertainty could be conducted through hydrodynamic modeling in future studies. The above revisions have been made in Section 4.3 of the revised manuscript, as follows:

“If the upstream Lake 14 (0.55 km²) were to experience an outburst, the total water volume could reach a maximum of 452.3⁶ × 106 m³ after converging with the down-stream Reduo Co (2.94 km²). This represents an increase of approximately 910% com-pared to a single outburst from Reduo Co, based on the volume estimation formula F5 [60]. It should be noted that this percentage is derived from a static summation of lake volumes and represents an extreme upper-bound scenario rather than a precise pre-diction of actual flood discharge.”

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

The text reads much better in this revision and the additions are well integrated with the original to improve the scientific quality.

However, I still have a problem with the geolocation as used (or not) in the paper and supplementary.

Table 1 does have a latitude and longitude and three decimal places is (just about) ok to identify the location buthaving look at the positions given I really thgink it should be 4 decimal.. However, in Supplementary there are only two decimal places, this is not sufficient and should be at least 3. In Table S4 however it is 8 decimals. This is far too much. please see: https://www.explainxkcd.com/wiki/index.php/2170:_Coordinate_Precision.

please also place them in [. ] brackets, as e.g. [29.754, 96.557] Table S3 gives the column row as an 'ID' . Keep the ID by all means but each row should have a [dLL] geolocation. This is what is the identifier (that anyone else can look at if necessary) is for, you ID only locates the feature in a diagram and could be in any order. Your data are important (and expensive!) so please make them count. Please also make them uniform by stating latitude then longitude and as one [. ] tuple in one column so they cvan be machine read if necessary and plotted as such. As above, please be consistent in usage and decimal pooint specification.

[dLL] geolocated points are still required for the figures. Evemn in Fig 4 with a values around a map you give these as degrees minutes. Please identify at least some of the points with a [dLL] and this should be in the figure so that it does not get separated from the image.

Author Response

GeoHazards

Re: revision on manuscript #geohazards-4452106

Dear Editors and Reviewers:

We would like to express our gratitude for the valuable comments and suggestions raised by the reviewer on our manuscript:

Evolution of Glacial Lakes and GLOF Risks to Transportation Routes in the Southeastern Tibetan Engineering Corridor

We are very grateful to the editor and the reviewers for their continued guidance. We appreciate your affirmation of our revisions in this round, and we also thank you for the remaining valuable comments that have helped us further improve the manuscript. In response to the suggestions raised in this round of review, we have made revisions and provided responses point by point. The detailed point‑to‑point replies are as follows:

The text reads much better in this revision and the additions are well integrated with the original to improve the scientific quality.

However, I still have a problem with the geolocation as used (or not) in the paper and supplementary.

——We thank the reviewer for once again raising detailed and professional comments on the geolocation issue. We fully agree that consistent coordinate formatting and reasonable precision are essential for data reusability. In response to the issues raised by the reviewer, we provide the following point‑by‑point replies:

(1) Regarding the inconsistency in coordinate precision. The reviewer noted that Table 1 uses three decimal places, the Supplementary Materials use two decimal places, and Table S4 uses eight decimal places—three different levels of precision. We fully accept this criticism. We have now unified the coordinates across all tables in the manuscript: all coordinates are now presented with four decimal places, which is sufficient to locate positions to approximately 10–20 meters—a reasonable precision for the purposes of glacial lake identification and localization in this study. The illustration on the website linked by the reviewer regarding coordinate precision is very instructive.

(2) Regarding the adoption of a unified [latitude, longitude] tuple format. We have standardized the coordinate format in all tables to [latitude, longitude] (e.g., [29.4641, 96.5023]), placed in separate columns to facilitate machine readability and direct import into GIS software. At the same time, we have retained the ID column in the supplementary tables for ease of reference in the text, but we have made the [dLL] coordinate column the primary geographic identifier.

(3) Regarding the correction of coordinate precision in the Supplementary Materials. We have revised all coordinates in the Supplementary Materials from two decimal places to four decimal places, ensuring consistency with the main tables.

(4) Regarding coordinate labeling in figures. The reviewer suggested that even in map‑based figures such as Figure 4, at least some points should be labeled with [dLL] coordinates to prevent the image from being separated from its spatial context. For map‑based figures containing over 200 glacial lakes (Figures 1, 3, 4, etc.), the figures already include complete latitude/longitude grids and scale bars, which serve as a standard spatial reference framework sufficient for regional spatial positioning. In addition, to further ensure data findability and reusability, we have uploaded the complete spatial vector data (Shapefile format) of all glacial lake inventories to the Mendeley Data open repository, and have provided detailed coordinate information for each lake in the Supplementary Materials. Therefore, we believe that labeling coordinate information for each lake on the map‑based figures would introduce redundancy and potentially compromise figure clarity and readability. We hope the reviewer can understand this approach.

(5) Regarding the placement of the coordinate column in tables. We have placed the [dLL] coordinate column in a prominent position (first or second column) in each table, making it the primary identifier for each row rather than hiding it at the end of the table.

The above revisions have been made in the corresponding sections of Table 1 and Supplementary Tables S3–S5 in the revised manuscript. We thank the reviewer for the strict requirements regarding data quality; these revisions have significantly enhanced the findability and reusability of our data, fully complying with the FAIR principles.

Author Response File: Author Response.docx

Reviewer 4 Report

Comments and Suggestions for Authors

See report

Comments for author File: Comments.pdf

Author Response

GeoHazards

Re: revision on manuscript #geohazards-4452106

Dear Editors and Reviewers:

We would like to express our gratitude for the valuable comments and suggestions raised by the reviewer on our manuscript:

Evolution of Glacial Lakes and GLOF Risks to Transportation Routes in the Southeastern Tibetan Engineering Corridor

We are very grateful to the editor and the reviewers for their continued guidance. We appreciate your affirmation of our revisions in this round, and we also thank you for the remaining valuable comments that have helped us further improve the manuscript. In response to the suggestions raised in this round of review, we have made revisions and provided responses point by point. The detailed point‑to‑point replies are as follows:

MC2: No independent validation; circularity dismissed as "not necessary"

Partially resolved

The dismissal is gone. A back-testing exercise now checks 4 historical lakes (Guangxie Co, Lumu Co, Anjue Zhuqulong, Zhuxigou) against the model (lines 435–450), and the authors candidly flag it as "a back-testing exercise" rather than independent validation. However, these 4 lakes are among the 12 events used to build the indicator set in the first place, so this remains largely in-sample. A leave-one-out cross-validation (excluding each event before scoring it) is still not performed but is now explicitly deferred to future work — an acceptable resolution for a minor-revision round, provided the text is not read as claiming independent validation.

——We thank the reviewer for this further clarification. We fully agree that since the four historically documented GLOF lakes are among the 12 events used to construct the indicator system, this validation is essentially a back‑testing exercise rather than an independent validation. To avoid misleading readers, we have adjusted the relevant wording in the main text accordingly.

Specifically, we have retained the description of the back‑testing results but have weakened the conclusive statement from “supporting the capability” to “indicating internal consistency of the assessment framework.” In addition, we have explicitly stated that this validation is a back‑testing exercise rather than an independent validation. Finally, we have identified a rigorous leave‑one‑out cross‑validation as a direction for future research, contingent on the accumulation of more GLOF records in the region.

These revisions have been made in Section 4.1 of the manuscript. We thank the reviewer again for the careful correction, which has made our description of the validation limitations more rigorous.

“Nevertheless, to test the performance of our assessment framework, we back-tested it against historically documented GLOF lakes within the study area. The results show that Guangxie Co, Lumu Co, and Anjue Zhuqulong glacier lake were all classified as having "very high" susceptibility, while Zhuxigou glacial Lake was classified as "high" susceptibility. All documented historical outburst lakes were correctly rated as either "high" or "very high" susceptibility, supporting the capability of our framework to distinguish lakes prone to outburst. Since these lakes are among the 12 historical events used to construct the indicator system, this validation remains a back‑testing exercise rather than an independent validation. The results merely indicate internal consistency of the assessment framework and are insufficient to robustly validate its predictive capability. A rigorous leave‑one‑out cross‑validation awaits the accumulation of more GLOF records in the region, which we identify as a direction for future research.”

MC6: Table 2 formulas lack units; no uncertainty propagation

Partially resolved

Units now stated (A in km², V in ×10⁶ m³; line 233–234) and the choice of formula [60] is now explicitly justified (largest calibration sample, closest regional match; lines 222–231). No sensitivity analysis across the five formulas in Table 2 was added, so the calibration uncertainty behind headline volumes (including the 452.36 × 10⁶ m³ figure) is still not quantified. Not blocking, but worth a sentence acknowledging this as a further limitation.

——We thank the reviewer for this further reminder. In the revised manuscript, we have supplemented a statement regarding the uncertainty in volume estimation in Section 4.5 (Limitations).

Specifically, in the second point of the Limitations section, we explicitly acknowledge that while we selected the formula with the strongest regional applicability and the largest calibration sample, we did not perform a systematic sensitivity analysis across the five formulas listed in Table 2 to quantify the calibration uncertainty behind the reported volumes, including the cascading‑flood totals in Section 4.3. We also note that the formula inevitably carries uncertainties when applied to lakes whose geometric characteristics differ from the calibration dataset, and that future bathymetric surveys of representative lakes in the study area would help to refine the volume estimates.

We believe this supplement honestly reflects the limitations of our study and also addresses the editor's concern regarding uncertainty quantification. The revision has been made in Section 4.5.

“4.5. Limitations

Several limitations of this study should be acknowledged when interpreting the results. First, the temporal resolution of our lake inventory (1990, 2000, 2010, and 2020) may not capture interannual variability or short-term rapid changes in lake area, par-ticularly for small lakes that respond quickly to climatic fluctuations. Future studies with higher temporal resolution would help to resolve this issue. Second, the lake volume estimation was based on an empirical area–volume relationship [60] calibrated on 47 Himalayan lakes. While this formula provides a reasonable approximation and was selected for its regional relevance and large calibration sample, we acknowledge that we did not perform a systematic sensitivity analysis across the five formulas listed in Table 2 to quantify the calibration uncertainty behind the reported volumes, including the cascading‑flood totals in Section 4.3. The formula inevitably carries uncertainties, especially when applied to lakes whose geometric characteristics differ from the calibration dataset. Future bathymetric surveys of representative lakes in the study area would help to refine volume estimates and reduce associated uncertainties. Third, this study focuses on GLOF susceptibility rather than comprehensive risk, the latter of which would require systematic assessments of exposure and vulnerability of downstream infrastructure, including bridge/culvert types and conditions, embankment resilience, and traffic exposure. Future studies integrating these components would contribute to a more complete risk assessment framework. Fourth, seasonal water-level fluctuations may lead to lake misclassification, particularly for small, shallow lakes whose water boundaries differ substantially between dry and wet sea-sons. To minimize this effect, we preferentially selected imagery from the summer–autumn months (June–October), when lake levels are relatively stable and lake boundaries are most discernible; nonetheless, residual seasonal variability cannot be entirely ruled out. Finally, while the cluster/cascade analysis provides a valuable qualitative assessment of flood amplification effects, quantitative simulation of cascading outburst dynamics would further strengthen hazard prediction capabilities.”

MC8: Internal inconsistency in total area-change figures (7.787 vs. 7.866 km²)

Partially resolved

The explicit "7.787 km²" claim was removed from the running text — a reasonable fix. However, the underlying figures were not reconciled: Figure 6's pie chart still shows 7.689 + 0.098 = 7.787 km², while Figure 7's bars (30.021 km² → 37.89 km²) imply a net change of 7.869 km². The same ~3–4% discrepancy persists between the two figures; it has become less visible rather than resolved. Recommend the authors verify the lake-matching datasets behind Figures 6 and 7 and reconcile the two values.

——We thank the reviewer for pointing out this issue. We have revised Figure 6 to reduce ambiguity and potential misinterpretation. Upon careful examination, we confirmed that the area change values in Figure 6 (the pie chart), namely 7.689 km² and 0.098 km², were derived from a 10 km × 10 km grid‑based statistical approach, which captures net changes at the grid‑cell level rather than tracking individual lakes over time. When a grid cell contains both expanding and shrinking lakes, or when lakes shift spatially between adjacent grid cells, the grid‑based results inevitably differ from the overall net change obtained by directly subtracting the total area in 1990 from that in 2020, with a systematic deviation of approximately 1%–5%.

To avoid confusion regarding the numerical discrepancy between “gross change (expansion plus shrinkage)” and “net change,” we have removed the specific area change values from Figure 6 and retained only the expansion rate (98.74%) and shrinkage rate (1.26%). These percentages are sufficient to clearly reflect the relative trends of glacial lake expansion and shrinkage in the study area. The original area values, being grid‑based coarse estimates, may mislead readers when compared with the net change value. Therefore, we have consistently adopted 7.866 km² as the standard net change in glacial lake area throughout the manuscript, which is derived from the direct subtraction of the total lake area in 1990 from that in 2020, and represents the most reliable indicator of the overall change trend.

After thorough verification and consideration, we would like to further clarify that the 10 km grid‑based statistics were designed to present the macroscopic spatial distribution pattern of glacial lake expansion and shrinkage across the study area, rather than to serve as a precise lake‑by‑lake quantitative change analysis. Therefore, it is normal and inevitable that the values obtained from this method deviate to some extent from the net change derived from the direct subtraction of total areas. Retaining the specific grid‑based numerical values in Figure 6 would not add informative value, but would instead risk causing confusion with the standard net change value (7.866 km²) adopted in this study. For this reason, we have removed the specific values and retained only the rates of expansion and shrinkage in the figure, so as to clearly convey the change patterns. We hope that this explanation meets with your approval. Should you have any further questions, we welcome your continued guidance at any time.

These revisions have been made in Figure 6 of the revised manuscript.

Round 1 stated imagery was filtered to "cloud cover below 10%"; Round 2 now states "cloud cover below 30%." Please clarify which threshold was actually used (or whether 10% was a Round-1), and confirm this does not affect the reported multi-temporal change-detection results, given the region's persistent cloud cover noted in the Introduction.

——We thank the reviewer for catching this detail; your observation is entirely correct. Upon verification, we confirm that our cloud‑cover filtering strategy actually employed a tiered threshold. In the initial screening stage, we used a 10% cloud‑cover threshold as the primary criterion. In practice, the 10% threshold was sufficient to identify the vast majority of glacial lakes in the study area, with only a few local areas affected by partial cloud cover. Therefore, in subsequent processing, we supplemented the dataset with additional imagery having overall cloud cover below 30%, which was used for local lake boundary identification, manual validation, and base map preparation for the figures. The sources, scene IDs, and cloud‑cover information for all images used are fully listed in Supplementary Table S1 for readers’ reference.

It should be noted that the reference to “10%” alone in the first‑round revision was a wording error in our description; we have now corrected it to reflect the actual threshold used, i.e., “30%.” Our actual cloud‑cover screening workflow consisted of three steps: first, we performed an initial filtering in GEE using the metadata cloud‑cover field with a threshold of 10% to acquire the majority of high‑quality images; second, for areas where the 10% threshold resulted in insufficient coverage, we supplemented the dataset with additional imagery having cloud cover below 30% to ensure complete spatial coverage of the study area; finally, all candidate images were visually inspected on a scene‑by‑scene basis to ensure that lake areas were cloud‑free. Any scene with cloud cover over critical lake areas was excluded, even if its overall cloud cover was below 30%, to guarantee the accuracy of lake boundary delineation and validation.

We confirm that this threshold adjustment does not affect the core conclusions of this study: on the one hand, all images used for area calculation were subject to rigorous manual quality control; on the other hand, the four epochs show consistent monotonic increases in both lake number and area, indicating that data quality remains stable across the time series.

Lake 11 is cited as "(Figure 11e)" in one sentence and "(Figure 12f)" in the next sentence of the same paragraph (Section 3.2.3); please verify against the correct current figure numbers.

——We thank the reviewer for identifying this figure citation error. We have verified and corrected the citation of Lake 11 in Section 3.2.3 to ensure consistency within the same paragraph. In addition, we have thoroughly checked all figure citation numbers throughout the manuscript and confirmed that, following the renumbering due to the addition or removal of figures, all related references have been updated accordingly, with no remaining citation errors or numbering inconsistencies. We appreciate your careful review.

the Data Availability Statement now points to an open Mendeley Data deposit for the lake inventory and susceptibility results — a good, unprompted improvement in transparency/FAIR compliance.

——We thank the reviewer for the recognition and affirmation of our efforts in data sharing. Indeed, in the revised manuscript, we have uploaded the complete glacial lake inventory data and susceptibility assessment results (in Shapefile format) to the Mendeley Data open repository to enhance the transparency and reproducibility of our study, and to ensure that the data are findable, accessible, interoperable, and reusable under the FAIR principles. The dataset is currently accessible via a private link for review purposes and will be made publicly available with a formal DOI upon acceptance of the paper. We once again appreciate the reviewer's encouragement, which further strengthens our commitment to data sharing and open science.

Author Response File: Author Response.docx

Review Reports

Re: revision on manuscript #geohazards-4452106

Reviewer 1:

Re: revision on manuscript #geohazards-4452106

Reviewer 2:

Re: revision on manuscript #geohazards-4452106

Reviewer 3:

Re: revision on manuscript #geohazards-4452106

Reviewer 4:

Re: revision on manuscript #geohazards-4452106

Re: revision on manuscript #geohazards-4452106

Re: revision on manuscript #geohazards-4452106