the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Assessing national exposure and impact to glacial lake outburst floods considering uncertainty under data sparsity
Abstract. Glacial Lake Outburst Floods (GLOFs) are widely recognized as one of the most devastating natural hazards in the Himalaya, which may catastrophic consequences including substantial loss of lives. To effectively mitigate these risks and enhance regional resilience, it is imperative to conduct an objective and holistic assessment of GLOFs hazards and their potential impacts of GLOFs over a large spatial scale. However, this is challenged by the limited availability of data and the inaccessibility to most of the glacial lakes in high-altitude areas. The data challenge is exacerbated when dealing with multiple lakes across an expansive spatial area. This study aims to exploit remote sensing techniques, well-established Bayesian regression models for estimating glacial lake conditions, cutting-edge flood modelling technology, and open data from various sources to innovate a framework for assessing the national exposure and impact of GLOFs. In the innovative framework, multi-temporal imagery is utilized with a Random Forest model to extract glacial lake water surfaces. Bayesian models, derived from previous research, are employed to estimate a plausible range of glacial lake water volumes and associated GLOF peak discharges, while accounting for the uncertainty stemming from the limited size of available data and outliers within the data. A significant amount of GLOF scenarios is subsequently generated based on this estimated plausible range of peak discharges. A GPU-based hydrodynamic model is then adopted to simulate the resulting flood hydrodynamics in different GLOF scenarios. Necessary socio-economic information is collected and processed from multiple sources including OpenStreetMap, Google Earth, local archives, and global data products to support exposure analysis. Established depth-damage curves are used to assess the GLOF damage extents to different exposures. The evaluation framework is applied to 21 glacial lakes identified potentially dangerous in the Nepal Himalaya. The results indicate that Tsho Rolpa Lake, Lower Barun Lake and Thulagi Lake bear the most serious impacts of GLOFs on buildings and roads, and influence existing hydropower facilities, while Lower Barun Lake, Tsho Rolpa Lake and Lumding Lake will experience the most impacts of GLOFs on agriculture areas. Four anonymous lakes (located at 85°37′51″ E, 28°09′44″ N; 87°44′59″ E, 27°48′57″ N; 87°56′05″ N, 27°47′26″ E; 86°55′41″ E, 27°51′00″ N) have the potential to impact more than 100 buildings, and the first three lakes may even submerge existing hydropower facilities.
- Preprint
(1915 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on hess-2023-260', Adam Emmer, 08 Feb 2024
This study aims at modelling potential future GLOFs from 21 Nepalese glacial lakes identified as potentially dangerous in previous study by Bajracharya et al. (2020). This study uses regression model to estimate plausible ranges of mean lake depths and so lake volumes and peak discharges. Further, exposed elements within the limits of modelled inundation areas are mapped and potential damage is assessed. Undoubtedly, such results are of value for disaster risk management authorities and I appreciate holistic approach going beyond GLOF modelling itself. While I’m very much in favor of GLOF hazard / risk assessment studies that consider a range of scenarios and I really appreciate the amount of work done, I have some thoughts for further improvements in the hazard assessment part.
Similar approach has been developed and employed by Veh et al. (2020) across the whole Himalaya. While thousands of lakes were considered in their study and the approach was suitable, I would expect bit more site-specific input data in case study of 21 lakes. For instance, moraine dam geometry (height and width) could be used to estimate max. breach depth and so max. flood volume that could differ substantially compared to the assumption of 100% lake volume release which is: (i) in general very unlikely for large lakes; and (ii) physically not even possible in many cases (when the height of a damming moraine is less than max. depth of the lake or the geometry of the dam is very flat).
Further, the procedure of random selection of 1000 scenarios and subsequent calculation of inundation frequencies and median of max. inundation depths for each lake is not appropriate because these scenarios are not equally probable. Reflecting on frequency-magnitude relationships of common GLOF triggers (various mass movements), low to moderate magnitude GLOFs are more frequent and more likely while extreme GLOFs are rare and less likely. Instead of selecting the scenarios randomly, my suggestion is to select them on purpose to cover the full range, with assigned weights (or ideally probabilities).
Since the modelling part lacks any validation, this is where frequency-magnitude relationship can come into play. I wonder whether employing your approach over past GLOFs can yield “typical extremity” of GLOFs (if you standardize the extremity of your scenarios for each lake on the dimensionless scale from 0-1)? While it is mentioned in Discussion section that the incompleteness of data about past GLOFs prevents the authors from attempting validation, I wonder whether any single GLOF characteristic (e.g., breach depth, flood volume, peak discharge, inundation area, etc.) could be used to validate the flood modeling results and estimate “typical extremity” of GLOFs in Nepal? Such an analysis could guide the weighting of your scenarios.
Overall, I’m in favor of recommending this study for further processing and subsequent publication after some modifications are considered. My suggestions to the authors are: (i) to consider dam geometry when estimating max. flood volume; (ii) to consider the validation of this approach with some of the past GLOFs in the country (and obtaining “typical extremity”); (iii) to consider avoiding the use of random selection of scenarios which may be misleading.
Citation: https://doi.org/10.5194/hess-2023-260-RC1 - AC3: 'Reply on RC1', Huili Chen, 16 May 2024
-
CC1: 'Comment on hess-2023-260', Taigang Zhang, 02 Mar 2024
Good day,Interesting paper. You might want to have a look at these papers recently published about GLOF risk assessment and lake bathymetry modeling. I hope it can be useful.
https://doi.org/10.1038/s41467-023-44123-z
https://doi.org/10.5194/tc-17-5137-2023
https://doi.org/10.1016/j.scitotenv.2021.150442Best regards
Taigang Zhang
Lanzhou UniversityCitation: https://doi.org/10.5194/hess-2023-260-CC1 -
AC1: 'Reply on CC1', Huili Chen, 15 May 2024
Thank you very much for sharing these excellent papers. They are very useful, especially the recent publications from 2023, which offer methods for estimating drainage volume and a framework for comprehensive GLOF risk evaluation. We have integrated these two recent publications into our introduction to enhance the discussion of current progress in the field of large-scale GLOF risk assessment.
Citation: https://doi.org/10.5194/hess-2023-260-AC1
-
AC1: 'Reply on CC1', Huili Chen, 15 May 2024
-
RC2: 'Comment on hess-2023-260', Anonymous Referee #2, 21 Mar 2024
This paper established the method for quantitative assessment of GLOF risk by combining Random Forest model to extract lake surface area, well-established Bayesian regression models to estimate glacial lake volume and peak discharge of outburst flood, hydrodynamic flood modelling, and damage analysis. Then the framework was applied to assess 21 glacial lakes in the Nepal Himalaya. I enjoy reading the methodology and believe it will contribute greatly to the quantitative assessment of the glacier lakes in the Himalaya with data sparsity. However, several major issues still need to be addressed carefully before further consideration of publication for this manuscript.
First, an isosceles triangle shape was assumed for the hydrograph of outburst floods (line 376, this information is better to be shown in the methods section by the way). I understand this assumption would simplify the calculation of hydrograph, which acts the key input for 2D hydrodynamic model. But this assumption needs to be justified before it can be used. If the hydrological monitoring data for the hydrograph of GLOF is too scarce, the authors can check the measured hydrographs of outburst floods for glacial lakes or barrier lakes in experimental research and see whether this assumption is close to the observations. The hydrograph shapes affect the interaction between morphology and hydraulics along the river significantly, so the assumption here needs to be made very carefully.
Second, I did not find the points in classifying the glacial lakes into three categories (lines 275-277). The classification standards were blurry and the glacier lakes were not analyzed by category (e.g., the volumes, peak discharges, or inundation areas of each class) in the results. I do not think it will make much difference to the clarity of the results if the classification is removed but will help reduce the length of the manuscript, which is already a bit too long.
Third, the manuscript is verbose in some sections and will benefit a lot if the irrelevant or repeating information is removed. For example, in lines 318 to 322, the lake areas from literature are listed, but these are not the results or findings of this study. So these lines can be shortened into one short sentence indicating the two glacier lakes are expending rapidly. Another example is the first paragraph in the discussion section. The paragraph adds very little information, mainly repeating what has been done in this work. It is fine to summarize the work in this study as the start of discussion but the summary needs to be concise. The second paragraph in the discussions has the same issue, with repeating information from the introduction and methodology section.
Apart from being verbose, the discussion section needs to be more focused. In lines 551 to 566, the authors introduced the backgrounds of hydropower projects in Nepal. This may help the readers to understand why the risk of hydropower stations was evaluated in this work, but too many details may become a deviation from discussing how the risk is distributed and varying in Nepal. Such information is more proper to be put into the supplementary materials rather than the maintext. Although the discussions include some comparisons with other studies to show the advantage of the methodology, I suggest the authors work on improving the depth of the discussions. For example, the assessment of inundation, exposure and damage has been presented in the results section, but the spatial distribution pattern, key influencing factors and the reasons or mechanism for the most severely affected glacier lakes can be further discussed. The discussion on the performance of the method used in this study is already enough but the interpretations of the outcomes of the method have not been dealt with in depth. But the interpretations will provide crucial insight to risk management of the glacier lakes for the study area.
Last, so many abbreviations were used in the manuscript but a list of abbreviations is missing. This creates extra difficulty for the readers to follow the manuscript. Also, figures need to be refined. Figures 4, 6, 7 and 8 do not show any ticks on the axes while the flow directions should be marked in figure 5.
Specific comments
Line 38: revise “… has observed…” to “is experiencing”.
Line 42: change “an objective and reproducible assessment” to “the requirement for reproducible assessment”.
Line 45: remove “typically focus on individual glacial lakes, which”.
Lines 54-57: the sentence can be more concise. Please rewrite.
Line 85: reference(s) are needed after “impact of GLOFs”.
Line 216: reference(s) are needed after “CPU-based counterpart”.
Line 231: the year seems to be 2022 from the reference list.
Line 239: are the values of Manning coefficients appropriate for Nepal? Please justify this setting.
Lines 345-357: most of the paragraph should be moved to the methods part. Please consider.
Figure 5: The locations of inset plots in the big map need to be marked.
Lines 402-405: The sentence should be moved to discussions.
Figure 8: It may be clearer if the results for the scenario when 100% lake water is released are presented together with the less severe scenarios.
Citation: https://doi.org/10.5194/hess-2023-260-RC2 - AC2: 'Reply on RC2', Huili Chen, 16 May 2024
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
437 | 145 | 43 | 625 | 31 | 31 |
- HTML: 437
- PDF: 145
- XML: 43
- Total: 625
- BibTeX: 31
- EndNote: 31
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1