the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Bayesian-Informed Hybrid Deep Learning for GLOF Susceptibility and Hazard Escalation in the HKKH Region (2010–2020)
Abstract. Rapid glacier retreat, complicated geography, and unpredictable weather make the Hindu Kush–Karakoram–Himalaya (HKKH) region extremely susceptible to glacial lake outburst floods (GLOFs). Current GLOF susceptibility assessments seldom take temporal dynamics or uncertainty into account, instead concentrating on either upstream lake conditions or downstream repercussions. An integrated, uncertainty-aware GLOF susceptibility framework that combines hybrid deep learning models with Bayesian probabilistic classification is presented in this paper. Spatiotemporal variations in glacial lakes and downstream terrain are captured using multi-temporal Landsat (2010–2016) and Sentinel-2 (2016–2020) imagery, SRTM DEM, Randolph Glacier & ICIMOD Inventory, morphological, hydrological, spatial, and topographical variables, and recorded GLOF events. CNN-LSTM, CNN-RNN, and Transformer-CNN models are trained using probabilistic labels produced by Bayesian inference. With AUC values between 0.90 and 0.92, the models demonstrate high predictive performance. High-altitude northern and central HKKH regions are becoming more vulnerable due to increased glacier melt, according to hazard escalation maps from 2010 to 2020. For regional GLOF risk assessment and disaster risk management, this framework offers a scalable tool.
- Preprint
(1408 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-2333', Jhon Sarria, 30 Sep 2026
-
AC1: 'Response to reviewer's comments', farkhanda abbas, 04 Oct 2026
The manuscript addresses an important topic related to GLOF risk management in a highly vulnerable region. It employs advanced methodologies to obtain relevant results that could contribute to improving plans and actions aimed at mitigating the impacts of these events. Overall, the methodology is robust, and the data are consistent with the objectives of the study. I recommend publication after minor revisions.
The figure title appears both above the figure and in the figure caption below it. Please consider removing the duplicated title above the figures to avoid redundancy and ensure consistency with the journal's figure-formatting requirements. Additionally, several figures appear to present similar or overlapping results. Please consider whether some of these figures could be combined or removed to improve the clarity and conciseness of the manuscript.
Response: the duplicate titles from all figures are removed. The susceptibility maps look similar but they have minor difference but not exactly the same.
Line 39: The word “triggers” should be corrected.
Response: typo error corrected (tigger to triggers)
Line 85: The legend and information presented in Figure 1 are difficult to read clearly. Consider increasing the font size and/or improving the figure resolution.
Response: figure improved accordingly front size increased for readability.
Line 95: “A minimal area threshold of 0.002 km² (about three Landsat pixels) was used to rule out small or transient ponds that are unlikely to cause GLOFs.” Does the threshold of 0.002 km² itself provide sufficient evidence that a lake is unlikely to cause a GLOF? I was unable to find information in the cited reference (Xu, 2006) that clearly explains or supports the use of this specific threshold. Please provide further justification or an appropriate reference supporting this criterion.
Response: Thank you for this important observation. We agree that a lake area of 0.002 km² alone cannot provide sufficient evidence that a lake is unlikely to generate a GLOF. We have therefore revised the wording to clarify that the 0.002 km² threshold was applied as a remote-sensing screening and inventory-quality criterion rather than as a physically based GLOF-hazard threshold. The use of 0.002 km² is supported by previous satellite-based lake inventories and glacial-lake studies that have employed this magnitude of minimum lake area to reduce uncertainty associated with the detection and delineation of very small water bodies. For example, Verpoorter et al. (2014) used approximately 0.002 km² as a minimum size in the GLOWABO lake inventory because smaller water bodies were difficult to reliably verify relative to image noise. Similarly, glacial-lake inventories in the Himalaya have used 0.002 km² as a minimum mapping threshold. We have consequently removed the implication that lakes below this threshold are inherently incapable of causing GLOFs. The revised manuscript now makes clear that the threshold was used for quality control, whereas GLOF susceptibility was evaluated using multiple physical and environmental factors. Referred to line 98 to 102.
Line 100, Table 5, and Figure 3: It is not clear how the Bayesian prior weights and rankings for the key GLOF susceptibility parameters were determined. Provide further details on the methodology used to establish these prior weights and rankings.
Response: Thank you very much for identifying the lack of clarity regarding the prior weights used in our Bayesian framework. In an effort to keep the original description concise, some essential methodological details regarding the derivation and interpretation of the prior weights were inadvertently omitted. We have therefore revised and rearranged this section (Lines 180–200) to provide a clearer explanation of both the parameter-specific prior weights and the class-level prior weights. The revised text now explicitly describes how the parameter rankings were established, how ordinal-ranking-informed normalized Bayesian prior weights rather than conventional ROC weights are formed and used and how these weights differ from the class-level prior probability used in the Bayesian classification. This revision provides a more transparent description of the prior-weighting procedure and clarifies the role of each type of prior information within the Bayesian framework.
Lines 106–113: The criteria used to identify non-GLOF lakes are presented, but the criteria used to identify GLOF-prone lakes are not clearly described. Clarify how the GLOF-prone lakes were identified.
Response : line 120-125 : Lakes that did not satisfy the predefined non-GLOF criteria were not assumed to have experienced a GLOF; instead, they were treated as referenced GLOF-prone samples. A random subset 225 of these samples were subsequently selected to construct the GLOF-prone class used in the Bayesian analysis.
Line 107: It is not clear how the threshold of 0.7 was established for classifying lakes as GLOF-prone or non-GLOF. Please provide the rationale or methodological basis for selecting this threshold.
Response: 0.7 is a conservative Bayesian decision threshold, mathematically corresponding to 70% posterior probability and 2.33:1 posterior odds, and it is broadly consistent with the use of elevated susceptibility cutoffs in HKKH GLOF studies.
Lines 207–208: This sentence appears to describe the methodology rather than the results. Please consider moving it to the appropriate section of the manuscript.
Response: Thank you for the comment. We have retained the sentence within Section 3.1 because it provides the context necessary for interpreting the model-performance results presented in Table 6. However, we revised its wording to integrate the evaluation measures directly with the reported results, rather than presenting it as a separate methodological statement. The subsection now focuses on the performance metrics and their corresponding results.
Line 210: The abbreviation “HNN” has already been defined. It is not necessary to define the abbreviation again.
Response: corrected thanks
Line 234: Hindu Kush–Karakoram–Himalaya (HKKH) has already been defined earlier in the manuscript, it is not necessary to define the abbreviation again.
Response: corrected thanks
Line 241: The abbreviation “GLOF” has already been defined.
Response: corrected thanks
Lines 264–265: “...were used to determine the susceptibility classes; 0.0–0.2 denotes extremely low susceptibility, 0.2–0.4 low, 0.4–0.6 medium, 0.6–0.8 high, and 0.8–1.0 very high.” Please clarify how boundary values between two classes are treated. For example, if the susceptibility value is exactly 0.4, is the lake classified as having low or medium susceptibility? The classification intervals should be defined unambiguously, including the treatment of boundary values.
Response: The posterior-probability threshold of 0.7 was used solely as the decision criterion for distinguishing GLOF-prone from non-GLOF lakes and should not be interpreted as a boundary between the five susceptibility classes. The susceptibility classes were independently defined from the normalized susceptibility index using fixed intervals of 0.2, 0.4, 0.6, and 0.8. Thus, a posterior probability of 0.7 indicates a GLOF-prone classification under the Bayesian decision rule, whereas a normalized susceptibility value of 0.7 corresponds to the High susceptibility class. The normalized susceptibility index was divided into five predefined classes using equal-width intervals of 0.2. The class boundaries were defined as follows: Very Low (0.0 ≤ S < 0.2), Low (0.2 ≤ S < 0.4), Medium (0.4 ≤ S < 0.6), High (0.6 ≤ S < 0.8), and Very High (0.8 ≤ S ≤ 1.0). These intervals were adopted as a consistent classification scheme for interpreting the normalized 0–1 susceptibility index. Boundary values were assigned to the higher susceptibility class; therefore, S = 0.4 is classified as Medium, S = 0.6 as High, and S = 0.8 as Very High.
-
AC1: 'Response to reviewer's comments', farkhanda abbas, 04 Oct 2026
-
RC2: 'Comment on egusphere-2026-2333', Anonymous Referee #2, 13 Oct 2026
The manuscript trains four hybrid deep learning models with naïve Bayes labels to map GLOF susceptibility in the HKKH region for 2010–2020. The topic is relevant, and the effort to account for label uncertainty over such a large region is appreciated. However, I have some concerns about the overlap between the labelling indicators and the model inputs, the training/testing design, the consistency of several numbers, and the comparison with the 2025 events. In my view these issues need substantial work, so I would suggest major revision and resubmission.
Major comments:
- It seems that the labelling indicators (Lines 98–103, Table 2) largely overlap with the model inputs (Section 2.3), and that the non-GLOF lakes are selected by low values of the same indicators (Lines 113–117), whereas GLOF-prone lakes are defined by high values (Table 2). If Table 6 is computed against the Bayesian labels (Lines 213–214), it may mainly reflect how well the networks reproduce the labelling rule; if it is computed on the GLOF inventory (Lines 108–112), the negative samples are still lakes selected for low indicator values. It would also help to clarify whether this inventory was used to set the priors (Lines 172–175). I would suggest listing the input features, evaluating against observed GLOF occurrence with representative non-GLOF lakes, and comparing with simple baselines (e.g., naïve Bayes, logistic regression, random forest).
- I found the design of the training and test sets in Lines 104–112 difficult to follow: 675 + 225 lakes are used for training, and "900 lakes with confirmed past GLOFs" for testing. If the test set contains only positive samples, AUC, PR-AUC and calibration could not be computed on it. Could the authors give the size and class composition of each set and fold, the data used for Table 6 and Figures 3–4, and the source of the 900 GLOF lakes?
- More detail on Section 2.4 would help readers reproduce the method, in particular how the class-conditional parameters were estimated, how the Beta prior (Line 189) and the weights in Table 5 enter the posterior of each lake, and whether the networks used soft labels or labels binarized at 0.7 (Lines 106–107, 192). Definitions of lake volume, breach volume, excess volume and IPP, and the sensitivity analysis mentioned in Lines 335–337, would also be useful.
- Some numbers appear inconsistent and may need checking: Figure 2 and the text differ in the Sentinel-2 period (2016–2022 vs. 2016–2020) and the minimum lake area (0.1 vs. 0.002 km²); the breach volume threshold in Line 115 (0.5 million m³) is ten times the lake volume threshold in Lines 116–117; the colour bar of Figure 7 ends at about 5.5 million m³, below the 20 million m³ threshold in Table 2; Table 4 implies mean depths of only 0.5–5 m; Lines 210–212 do not match Table 6; and Figure 8 totals about 250,000, versus 23,000 lakes (perhaps lake-years?).
- It would be helpful to describe how the hazard escalation and bivariate scores (Figures 10–11) were calculated and how Landsat and Sentinel-2 were harmonized. Sections 3.2 (Lines 243–247) and 3.3 (Lines 280–289) seem to describe different regional patterns, which could be reconciled. Section 3.4 is qualitative, and some events (Uttarkashi; Gilgit-Baltistan melt floods; the ice-dammed Shishper lake) may not be representative of the processes modelled; some cited references also predate the events. Reporting the predicted susceptibility of the actual source lakes would make this comparison more convincing.
Minor comments:
- The manuscript would benefit from language editing; Lines 287–299 are somewhat repetitive. Typos: "tiggers" (Line 39), "take use of" (Line 62), "according on" (Line 301).
- Susceptibility, hazard and risk are used interchangeably; consistent use would help.
- Some references seem unrelated to the statements. Please check.
- Table 3: please consider giving the final hyperparameters, the CNN input, and the definition of the HNN.
- Line 118: "Table 3" should probably be Table 4. Equations are not numbered; superscripts are missing in Lines 95 and 114–117.
- Most figures have low resolution and small fonts (Figures 1, 2, 5–15). A basemap would help in Figures 5–11; "HKHH" should be "HKKH".
- Figure 4: please give the number of bins and samples per bin.
- Section 4: separating the Discussion from the Conclusions and comparing with previous studies would be useful.
Citation: https://doi.org/10.5194/egusphere-2026-2333-RC2 -
AC2: 'response to second reviewer', farkhanda abbas, 23 Oct 2026
The manuscript trains four hybrid deep learning models with naïve Bayes labels to map GLOF susceptibility in the HKKH region for 2010–2020. The topic is relevant, and the effort to account for label uncertainty over such a large region is appreciated. However, I have some concerns about the overlap between the labelling indicators and the model inputs, the training/testing design, the consistency of several numbers, and the comparison with the 2025 events. In my view these issues need substantial work, so I would suggest major revision and resubmission.
Major comments:
- It seems that the labelling indicators (Lines 98–103, Table 2) largely overlap with the model inputs (Section 2.3), and that the non-GLOF lakes are selected by low values of the same indicators (Lines 113–117), whereas GLOF-prone lakes are defined by high values (Table 2). If Table 6 is computed against the Bayesian labels (Lines 213–214), it may mainly reflect how well the networks reproduce the labelling rule; if it is computed on the GLOF inventory (Lines 108–112), the negative samples are still lakes selected for low indicator values. It would also help to clarify whether this inventory was used to set the priors (Lines 172–175). I would suggest listing the input features, evaluating against observed GLOF occurrence with representative non-GLOF lakes, and comparing with simple baselines (e.g., naïve Bayes, logistic regression, random forest).
Response : We thank the reviewer for this valuable suggestion. In the revised manuscript, we have clarified the predictor variables used by the four hybrid deep-learning models and their relationship to the Bayesian labelling framework. We have also clarified the composition of the independent test dataset, comprising 900 documented historical GLOF lakes and 900 stratified pseudo-absence lakes, which were reserved for model evaluation referred line 152 to line 165. We acknowledge that the overlap between the labelling indicators and model inputs may influence the interpretation of model performance, and we have addressed this limitation in the revised discussion. Although comparisons with conventional classifiers, including naïve Bayes, logistic regression, and random forest, would provide additional insight, the present study focuses on evaluating the performance of the four selected hybrid architectures. We have acknowledged the absence of these baseline comparisons as a limitation and identified them as a priority for future research.
- I found the design of the training and test sets in Lines 104–112 difficult to follow: 675 + 225 lakes are used for training, and "900 lakes with confirmed past GLOFs" for testing. If the test set contains only positive samples, AUC, PR-AUC and calibration could not be computed on it. Could the authors give the size and class composition of each set and fold, the data used for Table 6 and Figures 3–4, and the source of the 900 GLOF lakes?
- Response : to reduce the length of paper some of important details had been missing we have added the necessary details again it’s 625 + 625 lakes are used for training it’s a typo error for training and second it secondly 900 lakes with confirmed past GLOFs Plus 900 pseudo-absence samples were selected from eligible glacier lakes without documented historical GLOF occurrences using stratified pseudo-absence sampling strategy base on which AUC and other performance matrix has been evaluate we have mentioned details from line – line . the 900 GLOF identified from Glacier Lake points ICIMOD is available at ICIMOD and Global Glacial Lake Inventory: 2015. Available at Zenodo.
- More detail on Section 2.4 would help readers reproduce the method, in particular how the class-conditional parameters were estimated, how the Beta prior (Line 189) and the weights in Table 5 enter the posterior of each lake, and whether the networks used soft labels or labels binarized at 0.7 (Lines 106–107, 192). Definitions of lake volume, breach volume, excess volume and IPP, and the sensitivity analysis mentioned in Lines 335–337, would also be useful.
Response: We thank the reviewer for this constructive comment. We agree that additional methodological detail is necessary to improve the reproducibility and transparency of the Bayesian classification framework. Accordingly, we have expanded Section 2.4 to clarify the estimation of class-conditional likelihoods, the roles of the parameter-specific prior weights and the class-level Beta prior, and the procedure used to derive the posterior probabilities for individual lakes.
We have also clarified the distinction between continuous posterior probabilities and binary classifications based on the 0.7 threshold, specifying how the resulting labels were used to train the hybrid deep-learning models. In addition, we have expanded the definitions of lake volume, breach volume, excess volume, and IPP, including their calculation procedures, units, and supporting references where applicable. Finally, the description of the sensitivity analysis has been revised to identify the parameters and thresholds examined and explain how changes in these settings affected the resulting susceptibility classifications. These revisions improve the methodological clarity and reproducibility of the proposed framework. Line 194 to 250
- Some numbers appear inconsistent and may need checking: Figure 2 and the text differ in the Sentinel-2 period (2016–2022 vs. 2016–2020) and the minimum lake area (0.1 vs. 0.002 km²); the breach volume threshold in Line 115 (0.5 million m³) is ten times the lake volume threshold in Lines 116–117; the colour bar of Figure 7 ends at about 5.5 million m³, below the 20 million m³ threshold in Table 2; Table 4 implies mean depths of only 0.5–5 m; Lines 210–212 do not match Table 6; and Figure 8 totals about 250,000, versus 23,000 lakes (perhaps lake-years?).
Response: thanks it’s 2016–2020 we had corrected the typo error in image. 0.1 km² is used as threshold value with Bayesian framework to select the glof and non glof samples 625+625 and 0.002 km2 for screening of water bodies extracted using NDWI. The two thresholds refer to different parameters: the lake-volume threshold (0.05 million m³) represents the estimated volume of water stored in a lake, whereas the breach-volume threshold (0.5 million m³) refers to the breach-related volume criterion used in our framework. the colour bar of Figure 7 ends at about 5.5 million m³ we crop the image because of it size now the full range of volume is visible and it’s also justify the use of 20 million m³ threshold for volume which not a universal threshold value but many studies in HKKH used this thread hold we had also provided literature references (Table 4 implies mean depths of only 0.5–5 m ?) I couldn’t exactly understand what you ask but table 4 contain almost 250,000 lakes over the years 2010 to 2020 and we randomly presented some of them but if required we can provide the whole table as supplementary files. (Lines 210–212 do not match Table 6) corrected. Figure 8 totals about 250,000, versus 23,000 lakes (perhaps lake-years?) exactly the figure trace all 23000 lake over the year of 2010 – 2020.
- It would be helpful to describe how the hazard escalation and bivariate scores (Figures 10–11) were calculated and how Landsat and Sentinel-2 were harmonized. Sections 3.2 (Lines 243–247) and 3.3 (Lines 280–289) seem to describe different regional patterns, which could be reconciled. Section 3.4 is qualitative, and some events (Uttarkashi; Gilgit-Baltistan melt floods; the ice-dammed Shishper lake) may not be representative of the processes modelled; some cited references also predate the events. Reporting the predicted susceptibility of the actual source lakes would make this comparison more convincing.
Response: the description about hazard escalation and bivariate scores in (Figures 10–11) were calculated and how Landsat and Sentinel-2 were harmonized are mentioned accordingly reference line 335 – line 350. We thank the reviewer for identifying the apparent inconsistency between Sections 3.2 and 3.3. We have revised both sections to distinguish the spatial distribution of lake area and volume, model-derived susceptibility, the bivariate hazard score, and temporal hazard escalation. Section 3.2 now clarifies that the concentration of larger lakes in parts of the northern and central HKKH does not necessarily imply the highest susceptibility across all models. Section 3.3 identifies the south-eastern Himalaya as the most consistent hotspot of very high susceptibility across the four deep-learning models and acknowledges the greater model-to-model variation in the Hindu Kush and Karakoram. We have also clarified that the bivariate hazard score combines lake volume and susceptibility, whereas the escalation score represents the change in that indicator between 2010 and 2020. These revisions provide a consistent interpretation of the regional patterns while preserving the differences represented by the individual figures. We thank the reviewer for highlighting the limitations of the original Section 3.4. We have rewritten the section to clarify that the post-2020 event review is a qualitative comparison of subsequent event locations with the spatial susceptibility patterns identified during 2010–2020, rather than a representative modelling exercise or formal statistical validation. The revised discussion examines events in the eastern and central Himalayan arc alongside localized events in the northern HKKH, while distinguishing conventional GLOFs from other glacier-related or high-mountain flood processes. The revised interpretation identifies broad geographical consistency with the study findings while explicitly acknowledging that individual source-lake susceptibility classes and changes in GLOF frequency cannot be established from this qualitative comparison alone. Citations also provided and cross checked.
Minor comments:
- The manuscript would benefit from language editing; Lines 287–299 are somewhat repetitive. Typos: "tiggers" (Line 39), "take use of" (Line 62), "according on" (Line 301).
Response: typos corrected
- Susceptibility, hazard and risk are used interchangeably; consistent use would help.
Response: We thank the reviewer for highlighting the inconsistent use of susceptibility, hazard, and risk. We have standardized these terms throughout the manuscript according to the quantities assessed in our framework. Susceptibility now refers to the relative propensity of individual lakes to experience a GLOF, as represented by the four deep-learning models. The combined indicator is consistently termed the bivariate hazard score, calculated from normalized lake volume and model-derived susceptibility, while hazard escalation refers to its change between 2010 and 2020. We have also clarified that these indicators are not calibrated probabilities of GLOF occurrence and do not constitute a comprehensive risk assessment because exposure and vulnerability are not explicitly quantified. References to “risk” have been revised where they previously described susceptibility scores or the bivariate hazard indicator.
- Some references seem unrelated to the statements. Please check.
Response : references rechecked
- Table 3: please consider giving the final hyperparameters, the CNN input, and the definition of the HNN.
Response: Hybrid Neural Network (HNN), hyperparameters table updated accordingly.
- Line 118: "Table 3" should probably be Table 4. Equations are not numbered; superscripts are missing in Lines 95 and 114–117.
Response : yes it’s table 4 thanks reference corrected. Equation numbering and missing superscripts checked .
- Most figures have low resolution and small fonts (Figures 1, 2, 5–15). A basemap would help in Figures 5–11; "HKHH" should be "HKKH".\
Response: resolution ,fonts size improved and HKHH corrected
- Figure 4: please give the number of bins and samples per bin.
Response: suggestion noted for figure 4.
- Section 4: separating the Discussion from the Conclusions and comparing with previous studies would be useful.
Response: discussion and conclusion section improved accordingly thanks.
Data sets
DATASET F. Abbas https://www.usgs.gov
Model code and software
GITHUB REPOSITORY F. Abbas https://github.com/shaminkhan/GLOF/tree/main
Interactive computing environment
Jupyter Notebooks F. Abbas https://github.com/shaminkhan/GLOF/tree/main
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 315 | 198 | 46 | 559 | 51 | 45 |
- HTML: 315
- PDF: 198
- XML: 46
- Total: 559
- BibTeX: 51
- EndNote: 45
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript addresses an important topic related to GLOF risk management in a highly vulnerable region. It employs advanced methodologies to obtain relevant results that could contribute to improving plans and actions aimed at mitigating the impacts of these events. Overall, the methodology is robust, and the data are consistent with the objectives of the study. I recommend publication after minor revisions.
The figure title appears both above the figure and in the figure caption below it. Please consider removing the duplicated title above the figures to avoid redundancy and ensure consistency with the journal's figure-formatting requirements. Additionally, several figures appear to present similar or overlapping results. Please consider whether some of these figures could be combined or removed to improve the clarity and conciseness of the manuscript.
Line 39: The word “triggers” should be corrected.
Line 85: The legend and information presented in Figure 1 are difficult to read clearly. Consider increasing the font size and/or improving the figure resolution.
Line 95: “A minimal area threshold of 0.002 km² (about three Landsat pixels) was used to rule out small or transient ponds that are unlikely to cause GLOFs.”
Does the threshold of 0.002 km² itself provide sufficient evidence that a lake is unlikely to cause a GLOF? I was unable to find information in the cited reference (Xu, 2006) that clearly explains or supports the use of this specific threshold. Please provide further justification or an appropriate reference supporting this criterion.
Line 100, Table 5, and Figure 3: It is not clear how the Bayesian prior weights and rankings for the key GLOF susceptibility parameters were determined. Provide further details on the methodology used to establish these prior weights and rankings.
Lines 106–113: The criteria used to identify non-GLOF lakes are presented, but the criteria used to identify GLOF-prone lakes are not clearly described. Clarify how the GLOF-prone lakes were identified.
Line 107: It is not clear how the threshold of 0.7 was established for classifying lakes as GLOF-prone or non-GLOF. Please provide the rationale or methodological basis for selecting this threshold.
Lines 207–208: This sentence appears to describe the methodology rather than the results. Please consider moving it to the appropriate section of the manuscript.
Line 210: The abbreviation “HNN” has already been defined. It is not necessary to define the abbreviation again.
Line 234: Hindu Kush–Karakoram–Himalaya (HKKH) has already been defined earlier in the manuscript, it is not necessary to define the abbreviation again.
Line 241: The abbreviation “GLOF” has already been defined.
Lines 264–265: “...were used to determine the susceptibility classes; 0.0–0.2 denotes extremely low susceptibility, 0.2–0.4 low, 0.4–0.6 medium, 0.6–0.8 high, and 0.8–1.0 very high.” Please clarify how boundary values between two classes are treated. For example, if the susceptibility value is exactly 0.4, is the lake classified as having low or medium susceptibility? The classification intervals should be defined unambiguously, including the treatment of boundary values.