Glacial Lake Outburst Floods in High Mountain Asia: Historical Evidence, Future Changes, and Risk-Reduction Strategies from a Remote-Sensing Perspective

Highlights

What are the main findings?

• GLOFs in HMA are strongly clustered by sub-region and dam type.
• Robust prediction of GLOF event timing, magnitude, and frequency remains constrained by uncertainties.

What are the implications of the main findings?

• Synthesize historical evidence, future changes, and risk-reduction strategies of GLOF across HMA.
• Research priorities are proposed to support durable climate adaptation and disaster risk reduction across HMA.

Abstract

Glacial lake outburst floods (GLOFs) are a major cryosphere-related hazard in High Mountain Asia (HMA), where glacier mass loss and changing hydroclimatic conditions are reshaping glacial-lake systems and increasing the prevalence of potentially unstable lake–dam configurations. However, current knowledge remains fragmented across HMA. Therefore, this review synthesizes historical evidence, future changes, and risk-reduction strategies of GLOFs across HMA from a remote-sensing perspective. Historical evidence derived from satellite archives, multi-temporal lake inventories, geomorphological analyses, and documented event records indicate that reported GLOFs in HMA are strongly clustered by sub-region and dam type, with moraine-dammed lakes representing the dominant source of documented events, while ice-dammed lakes remain important in several mountain belts. The compiled record also shows that GLOFs have caused severe human, economic, geomorphic, and ecological losses. Future projections based on glacier evolution, glacial-lake expansion, and climate-sensitive hazard assessments indicate continued glacial-lake growth under global warming. However, reliable prediction of future GLOF event timing, magnitude, and frequency remains constrained by uncertainties in glacier evolution, dam stability, and triggering processes. This review further shows that effective GLOF risk reduction in HMA requires integrated systems that combine hazard and risk mapping, early warning, structural interventions, and non-structural measures. It also highlights the need to better link remote sensing with monitoring, assessment, and implementation frameworks, and proposes an integrated management cycle to support practical risk reduction. It concludes that the most urgent research priorities are harmonized multi-temporal lake inventories, targeted field observations, explicit consideration of heatwaves and compound extremes, transparent uncertainty propagation, and stronger operationalization of monitoring and warning systems to support durable climate adaptation and disaster risk reduction across HMA.

1. Introduction

High Mountain Asia (HMA) is one of the world’s leading hotspots for glacial lake outburst flood (GLOF) hazards due to rapid deglaciation, widespread growth of ice- and moraine-dammed lakes, and the frequent occurrence of extreme rainfall events and heatwaves [1,2]. Global warming accelerates the loss of glacier mass in HMA, resulting in substantial changes in cryospheric hazards, water resources, and associated environmental problems [3,4]. As a result, accelerated glacier retreat and thinning promote the formation and expansion of glacial lakes and may increase the potential for GLOFs [5,6]. Because much of this evidence has been derived from satellite observations, multi-temporal glacial-lake inventories, and related geospatial analyses, remote sensing has become central to understanding the spatio-temporal evolution, hazard characteristics, and downstream implications of GLOFs across HMA. Therefore, it is important to improve understanding of glacial lake and GLOF dynamics, spatio-temporal distribution, and potential consequences to support sustainable development, climate adaptation, and disaster risk management in HMA.

GLOFs occur when a glacial lake drains abruptly after the stability of its natural dam is compromised or overtopping occurs, producing sudden downstream flooding [7,8]. In HMA, GLOFs most commonly originate from moraine- and ice-dammed lakes, while outbursts from bedrock-dammed lakes are comparatively rare and are typically associated with overtopping [9,10,11]. The likelihood of outburst is conditioned by hydroclimatic variability, which regulates meltwater supply, lake filling, and dam stability over time [12,13]. Outburst initiation is also frequently associated with discrete external disturbances that destabilize the dam or promote overtopping, including earthquakes, cloudbursts, ice or rock avalanches, landslides, and related slope failures, which may breach the dam directly or induce sudden lake-level rise and overflow [1,14]. Figure 1 illustrates the GLOF hazard chain, including erosion–deposition processes and downstream damage pathways. In this review, GLOF hazard refers to the physical outburst process and associated flood or debris-flow intensity, including outburst susceptibility, flow depth, velocity, and inundation extent. GLOF risk is used in the formal disaster-risk sense as the potential for adverse consequences arising from the interaction of hazard, exposure, and vulnerability.

Long-term glacial-lake inventories for the HMA region provide the essential data foundation for GLOF research. Multi-temporal Landsat-based inventories have documented substantial changes in glacial-lake number, area, and spatial distribution, including HMA-wide inventories for 1990 and 2018 [6], Third Pole inventories for approximately 1990, 2000, and 2010 [15], and HMA lake-area change assessments extending to 2020 [16]. Regional studies have further provided important long-term evidence for specific mountain belts, including Himalayan lake-change assessments for 1990–2015 [17], Nepal Himalaya inventories for 1977–2017 [18], and northwestern Indian Himalaya inventories for 1984–2016 [19]. More recent annual products, such as the 30 m Hi-MAG glacial-lake dataset for 2008–2017 [20], have improved the temporal resolution of glacial-lake monitoring across HMA. Together, these inventories show that glacial lakes have generally expanded in many glacierized parts of HMA, although the magnitude and timing of change vary strongly among sub-regions. They also provide the baseline information needed for lake-volume estimation, susceptibility screening, historical GLOF interpretation, and future hotspot assessment. However, inventory comparisons must consider differences in minimum lake-area threshold, glacier-proximity rule, lake-type definition, seasonal image selection, and manual quality-control procedures, because these methodological choices strongly affect lake counts, mapped area, and the identification of potentially dangerous lakes.

Many studies suggest that GLOF hazard is likely to increase across HMA, particularly in basins with debris-covered glaciers and unstable moraines [5,17,21,22]. Model-based assessments further indicate substantial growth in future glacial-lake area and associated hazard potential during the 21st century in HMA [21,23]. However, robust projection of the timing, magnitude, and frequency of future GLOFs in HMA remains difficult because key controls are still poorly constrained, including incomplete event inventories, limited field measurements of lake and dam properties, and uncertainty in future glacier mass loss and climate forcing [4,24,25].

In response, GLOF risk reduction in HMA increasingly combines lake-level interventions, early warning, preparedness planning, land-use regulation, and governance coordination across lake, catchment, and community scales [24,26,27]. However, implementation remains uneven in remote and transboundary mountain regions because monitoring, maintenance capacity, institutional roles, and cross-border coordination are often limited. Therefore, it is necessary to evaluate how remote-sensing evidence can better support practical risk-reduction planning and implementation across HMA.

Despite the growing body of work, recent review articles have addressed GLOF hazards from different spatial and thematic perspectives. Zhang et al. [3] provided a broad global critical review of glacial-lake and GLOF characteristics, changes, future hazards, and risk frameworks. Ahmed et al. [24] reviewed glacial-lake expansion and associated GLOFs in the Himalayan region, while Ahmed [27] provided a global overview of GLOF hazard and risk-management strategies. Nurakynov et al. [28] synthesized advances in remote sensing for GLOF monitoring and risk assessment, whereas Shah and Ishtiaq [29] focused on adaptation strategies in the Hindu Kush Himalaya. Bazai et al. [30] examined surge-glacier-related GLOF hazards in the Karakoram, emphasizing ice-dammed lake formation and outburst processes. These studies provide important foundations, but their spatial focus, thematic scope, methodological depth, and management emphasis differ considerably. Moreover, existing reviews have not yet systematically integrated HMA-specific historical GLOF evidence, glacial-lake inventory development, Earth-observation methods, future hazard changes, uncertainty sources, and operational risk-reduction strategies within a single remote-sensing-oriented analytical framework.

Therefore, this review provides an HMA-focused, remote-sensing-oriented critical synthesis of GLOFs by integrating historical evidence, glacial-lake inventory development, assessment methods, future hazard changes, uncertainty sources, and risk-reduction strategies. Specifically, this review aims to: (i) synthesize glacial-lake evolution and documented GLOF characteristics, including dam type, failure mode, and triggering conditions; (ii) compare remote-sensing and modelling approaches for lake mapping, volume estimation, susceptibility screening, and flood simulation; (iii) identify where and why GLOF hazard may intensify across HMA sub-regions; and (iv) evaluate how Earth-observation evidence can support early warning, structural and non-structural measures, governance coordination, and implementation priorities.

2. Review Design, Literature Search, and Evidence Synthesis

This review was designed as a structured critical narrative synthesis to integrate evidence on glacial-lake evolution, historical GLOF occurrence, remote-sensing and modelling approaches, future hazard changes, and risk-reduction strategies across HMA. The literature was identified from major scientific databases and technical sources, including Web of Science, Scopus, Google Scholar, ScienceDirect, SpringerLink, MDPI, Copernicus publications, ICIMOD reports, and relevant NASA, ESA, and national/international technical documents. The search combined terms related to hazard processes, regional context, remote-sensing datasets, methodological approaches, and risk management, including “GLOF/glacial lake outburst flood”, “glacial lake hazard”, “High Mountain Asia”, “Himalaya”, “Karakoram”, “Hindu Kush”, “Pamir”, “Tien Shan”, “Tibetan Plateau”, “glacial lake inventory”, “remote sensing”, “Landsat”, “Sentinel”, “SAR”, “InSAR”, “ICESat-2”, “SWOT”, “UAV”, “DEM”, “lake-volume estimation”, “susceptibility assessment”, “hydrodynamic modelling”, “hazard mapping”, “early warning”, and “risk reduction”. The overall search scope and evidence-synthesis framework are summarized in Figure 2.

Studies were prioritized when they met at least one of the following criteria: (i) they reported historical GLOF events or version-controlled GLOF inventories relevant to HMA; (ii) they developed or applied glacial-lake inventories, lake-area change analysis, lake-volume estimation, susceptibility screening, breach modelling, flood routing, or downstream exposure assessment; (iii) they used satellite observations, geospatial analysis, field validation, UAV data, hydrodynamic modelling, or machine-learning approaches relevant to GLOF assessment; or (iv) they discussed risk-reduction, early-warning, policy, governance, or adaptation strategies for high-mountain GLOF-prone regions. Studies outside HMA were considered only when they provided transferable methodological insights for glacial-lake mapping, GLOF modelling, uncertainty assessment, or risk management. Purely local glacier studies with no direct relevance to glacial lakes, GLOFs, remote sensing, modelling, or risk reduction were excluded. Non-peer-reviewed sources were used selectively for recent events, technical inventories, project information, and operational risk-reduction examples where the peer-reviewed literature was not yet available.

For historical event synthesis, version-controlled GLOF databases and published inventories were used as the primary evidence base. Recent events that occurred after the coverage period of major inventories were treated as case-study additions rather than automatically merged into the statistical database. To reduce duplication, reported events were cross-checked using available information on location, date, lake name, dam type, trigger, downstream impact, and original source. Where information on dam type, sub-region, or process attribution was incomplete, events were retained but clearly identified as unknown or unclassified. The evidence was then organized by sub-region, dam type, remote-sensing dataset, modelling approach, uncertainty source, and management relevance. This approach allowed to evaluate not only what has been reported across HMA, but also how differences in datasets, mapping thresholds, modelling assumptions, and reporting quality influence the interpretation of historical evidence, future hazard assessment, and practical risk-reduction strategies.

3. Remote-Sensing Methods Used for the Assessment of Glacial Lakes and GLOFs in HMA

3.1. Geographic Scope of High Mountain Asia

High Mountain Asia (Figure 3) is a mountain–plateau system comprising the Hindu Kush, Karakoram, Himalaya, Kunlun, Pamir, Tian Shan, and the Tibetan Plateau, extending across Afghanistan, Pakistan, India, Nepal, Bhutan, China, Tajikistan, Kyrgyzstan, Kazakhstan, and Uzbekistan [31,32]. HMA spans strong gradients in elevation and relief, from deeply incised mountain terrain to high-elevation plateaus, which creates sharp spatial contrasts in local climate and cryospheric conditions [33,34]. These contrasts also make regionally consistent hazard assessment difficult using field observations alone, thereby increasing the importance of remote sensing and geospatial analysis for documenting glacial-lake distribution and GLOF-related conditions across HMA. These contrasts shape glacier regimes and meltwater pathways and contribute to pronounced heterogeneity in where glacial lakes form and how they evolve among sub-regions. The geographic breadth and physiographic heterogeneity of HMA provide the essential spatial context for region-scale synthesis of glacial-lake and GLOF-related evidence.

HMA contains the largest concentration of glaciers outside the polar regions, sustaining major transboundary river systems such as the Indus, Ganges, Brahmaputra, and Mekong [36]. The HMA has over 95,000 glaciers and around 30,000 glacial lakes that combined cover about 2000 km² [6,37]. These headwaters support downstream water and food security and underpin livelihoods in both mountain communities and densely populated lowlands [32,38]. Because river networks cross national boundaries, upstream cryospheric change can influence water availability, hydropower reliability, and resource planning across multiple countries. This strong downstream dependence makes HMA strategically important for understanding climate-sensitive hydrological change and related hazards.

The climate of HMA reflects the combined influence of the South and East Asian monsoons and midlatitude westerlies, which shape seasonal moisture transport and circulation over the Tibetan Plateau and surrounding ranges [39,40]. This circulation produces strong spatial gradients in precipitation amount and seasonality, with a pronounced summer peak on monsoon-influenced southern and eastern slopes and greater westerly influence toward the northwest [39,40]. The plateau environment is characterized by strong solar radiation, low mean temperatures, and a large diurnal temperature range, all of which influence snow persistence and melt timing [31,41]. In addition, warm extremes such as heatwaves can temporarily intensify meltwater production and lake-level rise during the melt season.

3.2. Glacial-Lake Mapping and Inventory Generation

Comprehending glacial-lake attributes, distribution, and evolution is essential for evaluating and monitoring GLOF hazard in HMA because lake outlines and morphometric descriptors are the starting inputs for hazard screening and modelling workflows (Figure 4). In HMA, most region-scale glacial-lake assessments rely primarily on remote sensing because field-based mapping is logistically difficult, spatially uneven, and often impossible across many high-elevation and transboundary settings. Mapping has progressed from expert-led manual delineation of high-resolution optical and radar imagery, which can be highly accurate in complex terrain but is labour-intensive and may be sensitive to observer bias and data gaps [42,43]. Regional-scale efforts increasingly rely on automated or semi-automated classification using spectral water indices, most commonly the Normalized Difference Water Index (NDWI), to delineate lake boundaries consistently over large areas [6,44]. The expansion of satellite archives, together with cloud-computing platforms, has enabled reproducible multi-temporal processing for systematic lake inventories and change detection [45]. More recently, artificial intelligence and machine-learning segmentation have further improved the efficiency of large-area mapping in remote parts of HMA, especially when integrated with robust quality-control steps [46,47,48].

Furthermore, Landsat imagery remains widely used for NDWI-based lake mapping because its long, consistent archive supports repeated observations and multi-decadal inventories [5,49,50]. However, optical classifications can be degraded by seasonal snow and ice cover on lakes, mountain shadows, turbid water, and persistent cloud cover, all of which can increase omission and commission [50]. Additional uncertainty may arise from differences in image acquisition season, minimum mapping thresholds, and the treatment of supraglacial ponds, ice-contact lakes, and partially debris-covered margins, which can reduce comparability among inventories. Spaceborne SAR reduces cloud and illumination constraints, but its operational time series is often shorter than Landsat records and lake-surface backscatter can vary with roughness conditions, so SAR is typically used to complement optical mapping or support validation and targeted case studies [42,51]. Even with cloud computing and advanced classifiers, expert visual inspection and targeted manual correction remain important to remove persistent misclassifications and maintain temporal consistency in long-term monitoring products [52]. Therefore, differences in sensors, classification workflows, mapping thresholds, seasonal image selection, and validation practice can produce non-trivial inconsistencies in lake counts, area estimates, and the identification of potentially dangerous lakes across HMA inventories. These issues become especially relevant when comparing and integrating region-scale datasets, as summarized in the key inventories of glacial lakes and GLOFs applied across HMA (

Table 1. Key glacial-lake inventories, GLOF event databases, and glacier-outline products applied or discussed across HMA.

Dataset/Study Region	Spatial Coverage	Time Period of Data Used	Inventory Type	Minimum Lake-Area Threshold/Mapping Unit	Reference/Use Note
Glacial lake inventory of High Mountain Asia	HMA	1990; 2018	Glacial-lake inventory	≥0.0054 km2 (reported mapped size range: 0.0054–6.46 km2)	HMA-wide Landsat inventory used for long-term lake distribution and area-change comparison [6].
Regional-scale assessment of Himalayan glacial-lake changes	Himalaya	1990–2015	Glacial-lake inventory/change analysis	≥0.0081 km2 (nine Landsat pixels)	Regional Himalayan lake-change assessment using satellite observations [17].
Third Pole glacial-lake inventory	Third Pole region; subset overlaps HMA	1990; 2000; 2010	Glacial-lake inventory	>0.003 km2	Third Pole-scale historical glacial-lake inventory for multi-decadal comparison [15].
Glacial lakes in the Nepal Himalaya	Nepal Himalaya	1977–2017	National/regional glacial-lake inventory	≥0.0036 km2	Decadal inventory of glacial lakes in the Nepal Himalaya [18].
Glacial-lake inventory of the northwestern Indian Himalaya	Northwestern Indian Himalaya	1984–2016	Regional glacial-lake inventory	Minimum mapping unit reported in the original source	Used for regional Indian Himalayan comparison; included because threshold and mapping rules affect inter-inventory comparability [19].
Hi-MAG annual glacial-lake dataset	HMA	2008–2017	Annual glacial-lake inventory	≥0.0081 km2 (nine Landsat pixels)	Annual 30 m Landsat-based HMA glacial-lake inventory [20].
HMA lake area changes	HMA and sub-regions	1990–2020	HMA glacial-lake area-change product/synthesis	≥0.0081 km2 where reported for Landsat-based HMA comparison	Used for sub-regional lake-area trend comparison [16].
ICIMOD HKH glacial-lake inventory/status report	Five major HKH river basins: Amu Darya, Indus, Ganges, Brahmaputra, Irrawaddy, and Mansarovar Interior Basin	2005 ± 2 years	Glacial-lake inventory and potentially dangerous lake assessment	≥0.003 km2	ICIMOD HKH lake inventory and status assessment [53].
HMA_GLI/NSIDC High Mountain Asia Near-Global Multi-Decadal Glacial Lake Inventory	Near-global, including HMA	1990–2018; five multi-year periods	Glacial-lake extent product	Product-specific; 30 m Landsat-derived polygon product	Near-global/HMA-relevant lake extent product; useful for GLIMS/RGI-linked comparison [54].
Global Landsat-derived glacial-lake inventory	Global; HMA subset	1990–2018	Global glacial-lake inventory	0.05–200 km2	Larger-threshold global comparison product; useful for understanding threshold effects [5].
HMAGLOFDB v1.0	HMA	1833–2022	GLOF event database	Not applicable (event database, not a lake-boundary inventory)	Event-level HMA GLOF database used for historical event synthesis [55].
ICIMOD HMA GLOF database/Regional Database System	HMA	Access-date dependent/updated records	GLOF event database	Not applicable (event database, not a lake-boundary inventory)	GLOF event metadata and regional database [56].
Global historic GLOF database	Global; HMA subset extractable	850–2022; HMA subset depending on use	GLOF event database	Not applicable (event database, not a lake-boundary inventory)	Global event database; not a glacial-lake inventory [2].
GLIMS/RGI glacier-outline products, if used	Global/HMA glacier outlines	Version-dependent	Glacier-outline product	Not applicable (glacier outlines, not lake inventory)	Included only as glacier-outline reference products for glacier context, glacier proximity, and glacier identifiers; not lake-boundary inventories [57,58].

). Therefore, lake inventories should be compared cautiously because methodological differences can materially affect lake counts, area estimates, and the identification of potentially dangerous lakes.

3.3. Emerging Remote-Sensing Capabilities for GLOF Monitoring and Process Interpretation

Conventional optical lake mapping based on Landsat, Sentinel-2, and water indices remains essential for multi-decadal glacial-lake inventories in HMA, but it represents only one component of remote-sensing-based GLOF assessment. Landsat-based inventories have enabled region-wide and multi-decadal mapping of lake area and shoreline change, while annual 30 m products such as Hi-MAG provide consistent HMA glacial-lake coverage for recent decades [5,20]. However, lake outlines alone are insufficient for GLOF assessment because hazard interpretation also depends on lake level, volume change, dam geometry, glacier–lake contact, slope instability, ice-dam deformation, glacier-surge dynamics, flood and debris-flow pathways, and downstream exposure. Therefore, a more complete Earth-observation framework requires linking each sensor or method to specific GLOF-relevant variables and uncertainty sources (Table 2).

Satellite altimetry is becoming increasingly important for monitoring lake-surface elevation and estimating water-storage change in remote mountain basins. ICESat-2 ATLAS photon-counting laser altimetry provides elevation measurements that can be used to retrieve lake-surface height and, in optically clear shallow-water settings, support proxy bathymetry or lake-depth estimation when combined with optical imagery [59,60]. Studies on Tibetan Plateau lakes have demonstrated that ICESat-2 photon data can help reconstruct lakebed topography and water depth where photons penetrate the water column and where adequate denoising and track coverage are available [59,60]. SWOT further expands this capability by providing wide-swath measurements of surface-water elevation and extent for lakes, reservoirs, rivers, and wetlands, offering new opportunities to monitor lake-level change and downstream hydrological response in remote high-mountain catchments [61,62]. However, both ICESat-2 and SWOT have limitations in steep terrain, including track spacing, lake-size constraints, water-surface roughness, snow and ice cover, shoreline complexity, and data gaps in narrow valleys.

Radar-based approaches provide complementary information where optical imagery is limited by cloud, seasonal snow, shadow, or short revisit windows. Sentinel-1 and ALOS-2 SAR can support all-weather lake and flood mapping, while InSAR can detect deformation of ice dams, glacier termini, moraine slopes, and potentially unstable slopes adjacent to proglacial lakes [63,64,65,66,67]. Optical and SAR pixel-offset tracking are also important for identifying slope motion, glacier surging, and rapid ice-flow changes that may precede lake impoundment, dam failure, or impulse-wave generation. These methods are particularly relevant in the Karakoram, western Kunlun, Pamir, and related surge-glacier environments, where ice-dammed lake formation and repeated drainage events are closely linked to glacier dynamics [63,67]. For example, studies of Kyagar and Shisper/Shispare glaciers show that glacier surface velocity, terminus advance, ice-dam formation, and lake evolution can be reconstructed from multi-source optical, SAR, DEM, and feature-tracking datasets [63,64,65,66,67].

High-resolution commercial satellite imagery from Planet, Maxar, Pléiades, and similar constellations improves sub-event monitoring and event reconstruction by resolving narrow breach channels, shoreline changes, landslide sources, debris-flow deposits, damaged infrastructure, and post-event lake-level changes [68,69,70]. Such imagery is especially valuable for recent events, including the 2023 South Lhonak GLOF, where optical imagery, InSAR deformation analysis, and multi-source remote-sensing observations supported pre- and post-event interpretation of lake drainage, slope instability, dam breaching, and downstream impacts [68,69,70]. Thermal-infrared and land-surface-temperature products can complement optical and radar data by identifying thermal anomalies associated with glacier or moraine thermal stress, wet debris, exposed or buried ice, and possible englacial drainage pathways where spatial resolution and atmospheric conditions are suitable [4,70]. Nevertheless, thermal data remain underused in regional HMA GLOF studies because of coarse spatial resolution, mixed-pixel effects, complex terrain illumination, and limited field validation.

Overall, no single sensor provides a complete GLOF-monitoring solution. Optical imagery is best suited for long-term lake-area inventories, SAR improves all-weather mapping, altimetry provides elevation and storage-change information, InSAR and offset tracking reveal deformation and movement precursors, high-resolution imagery supports event-scale interpretation, and DEM products provide the topographic basis for dam geometry and downstream flood routing [4,5,20,59,60,61,62,63,64,65,66,67,68,69,70]. The most robust future GLOF assessment frameworks in HMA will therefore require multi-sensor fusion, uncertainty reporting, and targeted field validation rather than reliance on a single mapping product.

Table 2. Critical comparison of remote-sensing capabilities for GLOF monitoring and assessment in High Mountain Asia (HMA).

Sensor/Platform	Typical Spatial/Temporal Resolution	GLOF-Relevant Retrieved Variable	Main Strengths for HMA GLOF Research	Key Limitations	Reference
Landsat Series	30 m; 16-day revisit; archive from the 1970s/1980s depending on mission	Lake area, shoreline migration, lake expansion, glacier–lake contact	Best long-term optical archive for multi-decadal inventories and historical change detection.	Cloud, seasonal snow/ice, terrain shadow, mixed pixels, and limited sub-event detail.	[6,20]
Sentinel-2 MSI	10–20 m; approximately 5-day revisit with two satellites	Lake boundaries, small lake changes, glacier–lake contact, post-event shoreline change	Higher spatial and temporal detail than Landsat; useful for annual to seasonal monitoring and recent event interpretation.	Cloud, snow, mountain shadow, turbidity, and strong sensitivity to seasonal image selection.	[5,20]
Sentinel-1 SAR	10 m class; 6–12-day repeat depending on orbit availability	Open-water extent, flood traces, wet surfaces, lake-surface roughness	Cloud-independent observations during monsoon/cloudy periods; complements optical lake mapping.	Layover and radar shadow in steep terrain; wind, rough water, and frozen surfaces may complicate interpretation.	[68,70]
Sentinel-1/ALOS-2 InSAR	10–30 m class; repeat-pass interferometry	Ice-dam deformation, moraine/slope instability, glacier-terminus displacement, precursor deformation	Detects deformation not visible in optical imagery; valuable for unstable slopes and dams around glacial lakes.	Temporal decorrelation over snow, ice, water, and debris; geometric distortion in high relief; requires careful processing.	[68,70]
Optical/SAR Pixel-Offset Tracking	Sensor-dependent; days to months depending on image pairs	Glacier velocity, surge propagation, slope displacement, landslide motion, pre-failure movement	Captures rapid movement and surge dynamics that can generate ice-dammed lakes or impulse-wave triggers.	Requires suitable image pairs and texture; affected by snow cover, shadow, decorrelation, and geolocation error.	[70,71,72]
ICESat-2 ATLAS	Along-track photon-counting laser altimetry; approximately 17 m footprint; repeat-track sampling	Lake-surface elevation, water-level change, shallow-water/proxy bathymetry, storage-change constraints	Provides elevation information for remote lakes; photon data can support water-depth/bathymetry estimation in suitable clear-water conditions.	Track spacing, small/narrow lake coverage, water turbidity, ice/snow cover, shoreline complexity, and slope effects.	[59,73]
SWOT	Wide-swath Ka-band radar interferometry; ~21-day repeat; water bodies meeting mission size thresholds	Water-surface elevation, surface-water extent, storage change, river/lake hydrodynamic indicators	Systematic surface-water elevation and extent measurements for lakes, rivers, reservoirs, and wetlands; new opportunity for remote basin monitoring.	Mission data are recent; complex terrain, lake size, valley geometry, and algorithm maturity constrain high-mountain applications.	[74,75]
Planet/Maxar/Pléiades and Other Very-High-Resolution Optical Imagery	Sub-metre to ~3 m; daily to tasking-based depending on provider	Breach channels, landslides, debris-flow deposits, damaged infrastructure, lake drainage, shoreline changes	Resolves narrow breach channels and local geomorphic impacts; strong for post-event mapping and detailed case-study reconstruction.	Cost, access restrictions, cloud dependence, limited historical continuity, and tasking constraints.	[68,70,76]
UAV/Structure-from-Motion Photogrammetry	cm–dm scale; campaign-based	Dam geometry, outlet elevation, breach morphology, shoreline position, deposit thickness, local validation data	Very high spatial detail; useful for validating satellite products and planning site-level mitigation.	Limited spatial coverage; requires field access, permissions, weather windows, and operational safety.	[76]
Thermal Infrared/Land-Surface Temperature Products	30–1000 m depending on sensor; revisit from daily to 16-day depending on product	Thermal anomalies, surface wetness/ice indicators, glacier/moraine thermal stress, possible seepage or buried-ice indicators	Can complement optical/SAR indicators where thermal stress, buried ice, or englacial drainage pathways are suspected.	Coarse resolution, mixed pixels, atmospheric effects, terrain illumination, and limited direct validation for GLOF precursors.	[70,76]
DEMs: SRTM, ASTER, TanDEM-X, Copernicus DEM, UAV DEMs	10–90 m for regional DEMs; cm–dm for UAV DEMs; static or repeat-based	Dam height, freeboard proxy, slope, lake catchment geometry, flow path, inundation model input, exposure corridor	Essential topographic basis for susceptibility screening, breach modelling, flood routing, and downstream exposure assessment.	Vertical error, voids, radar penetration into snow/ice, temporal mismatch, glacier-surface change, and scale mismatch.	[22,76]

Note: The table emphasizes GLOF-relevant variables and limitations rather than presenting one sensor as sufficient for all tasks. Resolution and revisit values are approximate and depend on mission mode, orbit, latitude, cloud/snow conditions, and data availability. Abbreviations: DEM, digital elevation model; InSAR, interferometric synthetic aperture radar; SAR, synthetic aperture radar; SWOT, Surface Water and Ocean Topography; UAV, uncrewed aerial vehicle.

Table 2. Critical comparison of remote-sensing capabilities for GLOF monitoring and assessment in High Mountain Asia (HMA).

Sensor/Platform	Typical Spatial/Temporal Resolution	GLOF-Relevant Retrieved Variable	Main Strengths for HMA GLOF Research	Key Limitations	Reference
Landsat Series	30 m; 16-day revisit; archive from the 1970s/1980s depending on mission	Lake area, shoreline migration, lake expansion, glacier–lake contact	Best long-term optical archive for multi-decadal inventories and historical change detection.	Cloud, seasonal snow/ice, terrain shadow, mixed pixels, and limited sub-event detail.	[6,20]
Sentinel-2 MSI	10–20 m; approximately 5-day revisit with two satellites	Lake boundaries, small lake changes, glacier–lake contact, post-event shoreline change	Higher spatial and temporal detail than Landsat; useful for annual to seasonal monitoring and recent event interpretation.	Cloud, snow, mountain shadow, turbidity, and strong sensitivity to seasonal image selection.	[5,20]
Sentinel-1 SAR	10 m class; 6–12-day repeat depending on orbit availability	Open-water extent, flood traces, wet surfaces, lake-surface roughness	Cloud-independent observations during monsoon/cloudy periods; complements optical lake mapping.	Layover and radar shadow in steep terrain; wind, rough water, and frozen surfaces may complicate interpretation.	[68,70]
Sentinel-1/ALOS-2 InSAR	10–30 m class; repeat-pass interferometry	Ice-dam deformation, moraine/slope instability, glacier-terminus displacement, precursor deformation	Detects deformation not visible in optical imagery; valuable for unstable slopes and dams around glacial lakes.	Temporal decorrelation over snow, ice, water, and debris; geometric distortion in high relief; requires careful processing.	[68,70]
Optical/SAR Pixel-Offset Tracking	Sensor-dependent; days to months depending on image pairs	Glacier velocity, surge propagation, slope displacement, landslide motion, pre-failure movement	Captures rapid movement and surge dynamics that can generate ice-dammed lakes or impulse-wave triggers.	Requires suitable image pairs and texture; affected by snow cover, shadow, decorrelation, and geolocation error.	[70,71,72]
ICESat-2 ATLAS	Along-track photon-counting laser altimetry; approximately 17 m footprint; repeat-track sampling	Lake-surface elevation, water-level change, shallow-water/proxy bathymetry, storage-change constraints	Provides elevation information for remote lakes; photon data can support water-depth/bathymetry estimation in suitable clear-water conditions.	Track spacing, small/narrow lake coverage, water turbidity, ice/snow cover, shoreline complexity, and slope effects.	[59,73]
SWOT	Wide-swath Ka-band radar interferometry; ~21-day repeat; water bodies meeting mission size thresholds	Water-surface elevation, surface-water extent, storage change, river/lake hydrodynamic indicators	Systematic surface-water elevation and extent measurements for lakes, rivers, reservoirs, and wetlands; new opportunity for remote basin monitoring.	Mission data are recent; complex terrain, lake size, valley geometry, and algorithm maturity constrain high-mountain applications.	[74,75]
Planet/Maxar/Pléiades and Other Very-High-Resolution Optical Imagery	Sub-metre to ~3 m; daily to tasking-based depending on provider	Breach channels, landslides, debris-flow deposits, damaged infrastructure, lake drainage, shoreline changes	Resolves narrow breach channels and local geomorphic impacts; strong for post-event mapping and detailed case-study reconstruction.	Cost, access restrictions, cloud dependence, limited historical continuity, and tasking constraints.	[68,70,76]
UAV/Structure-from-Motion Photogrammetry	cm–dm scale; campaign-based	Dam geometry, outlet elevation, breach morphology, shoreline position, deposit thickness, local validation data	Very high spatial detail; useful for validating satellite products and planning site-level mitigation.	Limited spatial coverage; requires field access, permissions, weather windows, and operational safety.	[76]
Thermal Infrared/Land-Surface Temperature Products	30–1000 m depending on sensor; revisit from daily to 16-day depending on product	Thermal anomalies, surface wetness/ice indicators, glacier/moraine thermal stress, possible seepage or buried-ice indicators	Can complement optical/SAR indicators where thermal stress, buried ice, or englacial drainage pathways are suspected.	Coarse resolution, mixed pixels, atmospheric effects, terrain illumination, and limited direct validation for GLOF precursors.	[70,76]
DEMs: SRTM, ASTER, TanDEM-X, Copernicus DEM, UAV DEMs	10–90 m for regional DEMs; cm–dm for UAV DEMs; static or repeat-based	Dam height, freeboard proxy, slope, lake catchment geometry, flow path, inundation model input, exposure corridor	Essential topographic basis for susceptibility screening, breach modelling, flood routing, and downstream exposure assessment.	Vertical error, voids, radar penetration into snow/ice, temporal mismatch, glacier-surface change, and scale mismatch.	[22,76]

Note: The table emphasizes GLOF-relevant variables and limitations rather than presenting one sensor as sufficient for all tasks. Resolution and revisit values are approximate and depend on mission mode, orbit, latitude, cloud/snow conditions, and data availability. Abbreviations: DEM, digital elevation model; InSAR, interferometric synthetic aperture radar; SAR, synthetic aperture radar; SWOT, Surface Water and Ocean Topography; UAV, uncrewed aerial vehicle.

3.4. Lake Volume Estimation and Empirical Equations Applied in HMA

Lake-outline mapping alone is not sufficient for hazard quantification because lake volume provides the physical upper bound on water available for release and strongly conditions plausible outburst magnitude (Figure 4) [77]. Because direct bathymetric surveys are scarce in high-elevation environments, many regional assessments in HMA estimate lake volume indirectly from remotely sensed lake area, topographic context, and empirical scaling relationships [5,78,79]. The empirical relationships compiled for HMA include both global and regional calibrations and show substantial variation in sample size, explained variance, and domain of applicability (Table 3), reflecting differences in lake type, geomorphic setting, and calibration datasets [5,78,80]. For example, Table 3 includes relationships developed for lakes above specific size thresholds (e.g., >0.5 km²), for moraine-dammed lakes, for proglacial lakes, and for different HMA sub-regions, implying that selection should match lake type and mapping threshold [5,81]. Table 3 also shows that some relationships incorporate additional shape terms beyond area alone, which can improve performance where suitable calibration data exist [80]. Accordingly, uncertainty in remotely sensed lake extent, seasonal image selection, and inventory definition can propagate directly into volume estimates and influence subsequent hazard interpretation.

Once volume is estimated, studies commonly use empirical formulations and simplified breach approaches to generate plausible peak discharge and hydrographs for downstream simulation (Figure 4) [82,83,84]. The reliability of volume and discharge estimation therefore depends on (i) the quality and temporal consistency of lake outlines, (ii) the selection of an appropriate area–volume relationship (Table 3), and (iii) the transparency with which uncertainty is propagated into subsequent modelling steps. Where feasible, targeted field measurements and UAV/photogrammetry-derived geometry can strengthen calibration datasets and reduce uncertainty, but regional screening still largely depends on empirical relationships because bathymetry remains limited for most HMA lakes [85,86,87]. As a result, it is good practice to report the chosen relationship, its calibration setting (global vs. HMA sub-region), and the relevant lake-size/domain constraints (Table 3).

Table 3. Empirical area–volume relationships for glacial-lake volume estimation.

Formula	Studied Region	N	R2	Reference	Applicability/Note
V = 0.1217A1.4129	Global	15	0.95	[74]	Bedrock-, ice-, and moraine-dammed lakes
V = 0.05057A1.2884	Global	42	0.38	[78]	Bedrock-, ice-, and moraine-dammed lakes
V = 0.1746A1.3725	Global	30	0.60	[78]	Bedrock-, ice-, and moraine-dammed lakes
V = 0.3211A1.324	Global	57	0.57	[78]	Bedrock-, ice-, and moraine-dammed lakes
V = 0.1697A1.3778	Global	45	0.75	[78]	Bedrock-, ice-, and moraine-dammed lakes
V = 1.0 × 10−17A1.76	Global	122	0.99	[5]	Lakes > 0.5 km2
V = 0.0354A1.3724	HMA (Himalayas)	20	0.50	[88]	Moraine-dammed lakes
V = 0.087A1.434	HMA (Himalayas)	20	0.50	[88]	Moraine-dammed lakes
V = 4.067 × 10−20A1.184 − 3.218 × 10−15(RmaxW/maxL)	HMA (Central Himalayas)	27	0.96	[80]	Lakes ≥ 0.1 km2; includes shape-ratio term
V = 5.574 × 10−27.73A2.455 − 2.005 × 10−16(RmaxW/maxL)	HMA (Central Himalayas)	70	0.80	[80]	Lakes < 0.1 km2; includes shape-ratio term
V = 1.26 × 10−26A2 + 5.6 × 10−24A + 1.32 × 10−17	HMA (Himalayas)	35	0.98	[80]	Lakes ≥ 0.5 km2
V = 2.35 × 10−25A1.4083	HMA (Himalayas)	227	0.90	[80]	Lakes < 0.5 km2
V = 5.5 × 10−0.5A1.25	HMA (Himalayas)	16	N/A	[89]	Moraine-dammed lakes; South Lhonak case set
V = 0.0578A1.4683	HMA (Himalayas)	33	0.93	[90]	Moraine-dammed lakes
V = 4.3244 × 10−14A1.5307	HMA (Himalayas)	15	N/A	[91]	Moraine-dammed lakes
V = 5.22 × 10−18A1.1766	HMA (Himalayas)	6	0.99	[89]	Moraine-dammed lakes
V = 4.93 × 10−16A0.9304	HMA (North Himalayas)	15	0.99	[31]	Moraine-dammed lakes
V = 0.096A1.426	HMA (Central Himalayas)	15	N/A	[92]	Ice-, moraine-, and supraglacial-dammed lakes
V = 4 × 10−5A2 + 5.0564A	HMA (Western Himalayas)	N/A	0.96	[81]	Proglacial lakes

Note: V denotes glacial-lake volume and A denotes glacial-lake surface area. Unit conventions follow the original references listed in the table and should be checked before direct application, because empirical area–volume relationships may be reported using different combinations of area and volume units. Where equations are applied after unit harmonization, V is expressed in m³ and A in m². N is the number of bathymetric measurements used to develop the relationship; R² denotes the coefficient of determination. RmaxW/maxL is the ratio of maximum lake width to maximum lake length. N/A = not available or not applicable. Each equation should be applied only within the lake-type, regional, and area-range domain reported by the original source.

Table 3. Empirical area–volume relationships for glacial-lake volume estimation.

Formula	Studied Region	N	R2	Reference	Applicability/Note
V = 0.1217A1.4129	Global	15	0.95	[74]	Bedrock-, ice-, and moraine-dammed lakes
V = 0.05057A1.2884	Global	42	0.38	[78]	Bedrock-, ice-, and moraine-dammed lakes
V = 0.1746A1.3725	Global	30	0.60	[78]	Bedrock-, ice-, and moraine-dammed lakes
V = 0.3211A1.324	Global	57	0.57	[78]	Bedrock-, ice-, and moraine-dammed lakes
V = 0.1697A1.3778	Global	45	0.75	[78]	Bedrock-, ice-, and moraine-dammed lakes
V = 1.0 × 10−17A1.76	Global	122	0.99	[5]	Lakes > 0.5 km2
V = 0.0354A1.3724	HMA (Himalayas)	20	0.50	[88]	Moraine-dammed lakes
V = 0.087A1.434	HMA (Himalayas)	20	0.50	[88]	Moraine-dammed lakes
V = 4.067 × 10−20A1.184 − 3.218 × 10−15(RmaxW/maxL)	HMA (Central Himalayas)	27	0.96	[80]	Lakes ≥ 0.1 km2; includes shape-ratio term
V = 5.574 × 10−27.73A2.455 − 2.005 × 10−16(RmaxW/maxL)	HMA (Central Himalayas)	70	0.80	[80]	Lakes < 0.1 km2; includes shape-ratio term
V = 1.26 × 10−26A2 + 5.6 × 10−24A + 1.32 × 10−17	HMA (Himalayas)	35	0.98	[80]	Lakes ≥ 0.5 km2
V = 2.35 × 10−25A1.4083	HMA (Himalayas)	227	0.90	[80]	Lakes < 0.5 km2
V = 5.5 × 10−0.5A1.25	HMA (Himalayas)	16	N/A	[89]	Moraine-dammed lakes; South Lhonak case set
V = 0.0578A1.4683	HMA (Himalayas)	33	0.93	[90]	Moraine-dammed lakes
V = 4.3244 × 10−14A1.5307	HMA (Himalayas)	15	N/A	[91]	Moraine-dammed lakes
V = 5.22 × 10−18A1.1766	HMA (Himalayas)	6	0.99	[89]	Moraine-dammed lakes
V = 4.93 × 10−16A0.9304	HMA (North Himalayas)	15	0.99	[31]	Moraine-dammed lakes
V = 0.096A1.426	HMA (Central Himalayas)	15	N/A	[92]	Ice-, moraine-, and supraglacial-dammed lakes
V = 4 × 10−5A2 + 5.0564A	HMA (Western Himalayas)	N/A	0.96	[81]	Proglacial lakes

Note: V denotes glacial-lake volume and A denotes glacial-lake surface area. Unit conventions follow the original references listed in the table and should be checked before direct application, because empirical area–volume relationships may be reported using different combinations of area and volume units. Where equations are applied after unit harmonization, V is expressed in m³ and A in m². N is the number of bathymetric measurements used to develop the relationship; R² denotes the coefficient of determination. RmaxW/maxL is the ratio of maximum lake width to maximum lake length. N/A = not available or not applicable. Each equation should be applied only within the lake-type, regional, and area-range domain reported by the original source.

3.5. Susceptibility Screening and Hazard Classification

Lake identification and volume estimation are necessary but insufficient; hazard evaluation also requires assessing whether a lake–dam system is susceptible to failure under plausible initiating processes and internal weakening (Figure 4). Susceptibility screening typically integrates geomorphological and hydrological indicators such as dam freeboard, lake and catchment geometry, glacier–lake contact, and surrounding slope conditions, alongside evidence of past instability where available [43,93,94]. Process understanding distinguishes external disturbances that can destabilize dams (e.g., avalanches, landslides, extreme precipitation, seismic shaking) from intrinsic degradation mechanisms (e.g., seepage, piping, melting of buried ice, progressive erosion) that reduce dam integrity over time [94]. These distinctions matter because moraine-dammed lakes often fail by overtopping erosion or internal piping, whereas ice-dammed lakes may drain abruptly through flotation or subglacial conduit development, implying different indicator sets and scenario assumptions [95,96]. However, susceptibility classification remains highly sensitive to the quality of remotely sensed inputs, the completeness of historical event inventories, and the transferability of screening criteria across contrasting HMA sub-regions.

Operational screening is implemented using multi-criteria decision analysis or statistical and machine-learning models that integrate physical predictors with available event inventories to classify lakes into hazard categories [43,97]. Remote-sensing predictors (e.g., glacier retreat metrics or thermal anomalies) increasingly support screening in data-scarce areas by extending spatial coverage across HMA [43]. At the same time, machine-learning applications remain constrained by limited GLOF cases, imbalanced datasets, and uneven regional reporting, which can reduce model robustness and generalizability if validation is weak or geographically narrow. However, model transferability remains constrained by incomplete event documentation and reporting bias, so susceptibility outputs are most defensible when presented as prioritization tools rather than deterministic predictions [79]. Consistency in baseline datasets is also important, since different lake and GLOF inventories differ in spatial coverage, time period, and mapping thresholds, as summarized in Table 1. Overall, susceptibility screening is most useful as a regional prioritization tool when field data are limited, whereas detailed process-based assessment should be reserved for priority lakes where lake geometry, dam condition, slope instability, and downstream exposure require closer evaluation.

3.6. Breach Modelling and Hydrodynamic Simulation of GLOFs

Hydrodynamic simulation typically links (i) breach initiation and growth to generate a source hydrograph and (ii) downstream flood routing to estimate inundation extents, flow depths, and velocities for hazard mapping and engineering design (Figure 4) [77,98]. In most HMA applications, these simulations depend heavily on remote-sensing-derived inputs, including lake extent, dam geometry, topography/DEMs, channel alignment, land cover, and exposure layers for downstream impact assessment. Breach development is sensitive to dam composition and structure, breach geometry and growth rate, and lake level, and uncertainty in these inputs often dominates uncertainty in peak discharge and inundation [84,99]. Scenario-based modelling is therefore widely used to represent alternative failure mechanisms such as overtopping erosion, internal piping, and impulse-wave overtopping from mass movements into lakes [99]. These hydrographs provide consistent boundary conditions for routing models and allow comparison across lakes under standardized assumptions.

A range of hydrodynamic tools is applied in HMA depending on terrain complexity and the processes that must be represented. As summarized in

Table 4. Hydrodynamic and process-based models used for GLOF modelling, with representative versions, and inputs.

Model/Program	Representative Software Version/Release	Dimension/Governing Equations	Key Represented Processes	Standardized Typical Inputs	Representative HMA Event or Case	References
HEC-RAS	HEC-RAS 5.x–6.x where 2D modelling is used; exact version not always reported	1D/2D Saint-Venant and shallow-water equations	Clear-water flood routing, inundation extent, flow depth and velocity; sediment transport in selected configurations	DEM; lake volume or breach hydrograph; dam/breach geometry; channel/floodplain roughness; upstream/downstream boundary conditions; calibration water levels or observed flood extent where available	South Lhonak/Teesta basin; Central and Eastern Himalaya; SE Tibet; Indian Himalaya	[101,102,103]
BASEMENT	BASEMENT 2.x/3.x; exact version varies among cited studies	1D/2D Saint-Venant equations with morphodynamic and breach modules	Flood routing, sediment transport, dam-breach evolution, morphodynamic response	DEM; lake volume; breach parameters; inflow/outflow hydrograph; bed-material/granulometry; roughness; boundary conditions; observed deposits or inundation for validation	Indian Himalaya; Nepal Himalaya; Karakoram process simulations	[104,105,106]
RAMMS	RAMMS: Debris Flow 1.x; exact version not consistently reported	2D depth-averaged Voellmy-type mass-flow model	Debris-flow routing, flow transformation, runout and deposition	High-resolution DEM; release volume or hydrograph; Voellmy friction/rheology parameters; entrainment assumptions; calibration deposits/runout	Nepal Himalaya and Pamir debris-flow or GLOF-related mass-flow cases	[105,107]
r.avaflow	r.avaflow 2.x; exact version varies among cited studies	2D multi-phase mass-flow model using NOC-TVD numerical scheme	Water–debris mixture, impulse waves, process chains, erosion–deposition and entrainment	DEM; release mass/volume; water–solid fractions; rheology/friction parameters; entrainment settings; observed deposits or runout for validation	Jinwuco, Tibet; process-chain GLOF reconstructions	[98,108]
NWS-FLDWAV	NWS-FLDWAV; legacy 1D flood-routing model, version not consistently reported	1D Saint-Venant equations	Unsteady clear-water flood routing and breach-hydrograph propagation	Channel cross-sections; breach hydrograph; downstream boundary condition; roughness; lateral inflow where available	Sagarmatha/Nepal Himalaya GLOF routing applications	[109,110]
HR BREACH	HR BREACH/HR Wallingford breach model; version not consistently reported	1D physically based breach-growth model	Dam-breach erosion, breach widening/deepening and source hydrograph generation	Dam height and geometry; dam material; erodibility; lake level/volume; breach-initiation assumptions	Dig Tsho reconstruction and moraine-dam breach applications	[111]
DL Breach	DL Breach/dam-break breach model; version not consistently reported	1D Saint-Venant equations with breach formulation	Dam breach, water/debris routing and hydrograph generation	Dam geometry; breach parameters; lake volume; DEM/channel geometry; roughness; downstream boundary conditions	Eastern Himalaya GLOF modelling cases	[112]
MIKE 11/MIKE 21	MIKE 11/21 releases used in cited studies; exact version varies among cited studies	1D/2D Saint-Venant and shallow-water equations	Flood routing, inundation mapping and sediment transport in selected modules	DEM or cross-sections; breach/discharge hydrograph; roughness; boundary conditions; sediment parameters where used; observed flood extent/depth if available	Indian Himalayan and Tibetan basin GLOF simulations	[113,114,115,116]
SMPDBK	SMPDBK; version not consistently reported	1D Saint-Venant routing	Dam-break flood routing and simplified debris-flow routing	Dam parameters; breach hydrograph; channel geometry/cross-sections; roughness; downstream boundary condition	SE Tibetan Plateau and Indian Himalaya applications	[117,118,119]
FLO-2D	FLO-2D Pro/FLO-2D releases used in cited studies; exact version varies among cited studies	2D grid-based shallow-water equations with mud/debris-flow options	Flood and debris-flow routing, inundation depth and velocity, sediment/debris effects	DEM; computational grid; inflow/breach hydrograph; roughness; rheology/sediment concentration; observed inundation/deposits	Imja Tsho and SE Tibet GLOF simulations	[116,120]
TELEMAC-2D	TELEMAC-2D v7/v8 family; exact version varies among cited studies	2D shallow-water equations	Hydrodynamics, flood routing, inundation and sediment transport where coupled	DEM/bathymetry; mesh; boundary conditions; breach/source hydrograph; roughness; calibration flood levels or extent	South Lhonak, Sikkim/Teesta basin	[121]

Note: DEM denotes digital elevation model. Software versions are representative versions or model generations reported in the literature and should not be interpreted as the latest available release. Where exact versions are not consistently reported in HMA applications, this is stated explicitly. The standardized input column reports comparable input groups across models: terrain data, lake/dam geometry, breach or release hydrograph, roughness/rheology parameters, boundary conditions, and validation data where available. Many inputs are remote-sensing-derived, including DEMs, lake extent, dam geometry, channel alignment, land cover, post-event deposits, and exposure layers.

, widely used models include HEC-RAS (commonly implemented in 1D/2D forms based on Saint-Venant/shallow-water equations) and other platforms that can represent inundation mapping and, in some configurations, sediment transport, requiring inputs such as DEMs, breach hydrographs or breach parameters, and channel/floodplain roughness. Where process-chain behaviour and debris-laden flows are important, mass-flow and two-phase tools (e.g., RAMMS, r.avaflow) are increasingly used to represent flow transformation and sediment entrainment under steep mountain conditions [98,100]. Model performance therefore depends not only on hydraulic formulation, but also on the spatial resolution, vertical accuracy, and temporal representativeness of the remote-sensing-derived terrain and boundary-condition data used to initialize simulations. The models, dimensionality, typical inputs, and representative HMA applications should be reported consistently to support comparability, as structured in Table 4. In practice, model selection should follow the dominant process and decision need: 1D/2D hydraulic models are most suitable for flood routing and inundation mapping, while mass-flow or two-phase models are more appropriate where debris entrainment, landslide-generated impulse waves, or coupled process chains are expected.

4. Historical GLOFs and Documented Losses in HMA

4.1. Historical Records and Sources of GLOFs

Historical GLOFs in HMA represent rapid lake drainage following failure or overtopping of natural dams, most commonly moraine or ice, releasing impounded water and often debris in short-duration floods [33,94]. Here, the historical GLOFs record is reconstructed from multiple evidence streams, including satellite imagery, geomorphic and sedimentary traces, river-gauge information, published reports, and documented accounts [33]. However, triggers and failure modes are frequently unknown or inferred because observations at the time of failure are sparse, and diagnostic evidence may be short-lived or never recorded in remote terrain [3]. These limitations imply that catalogues are informative but incomplete, and apparent spatial–temporal patterns should be interpreted with explicit consideration of reporting bias and uneven observation effort through time [122,123].

Based on available compilations, approximately 711 reported GLOF records are identified across HMA for the historical synthesis [2]. These records include both dated events and records with incomplete temporal attribution; therefore, decadal analyses were based only on records with available event-year information, whereas sub-regional and lake–dam type summaries were calculated from the full compiled dataset. The records were organized using a single event-level dataset following the HMAGLOFDB structure, and the regional and lake–dam type percentages were calculated relative to the full compiled total. The HiMAP-based sub-regional grouping shows that reported GLOFs are concentrated mainly in the Tien Shan (n = 255; 35.9%), Karakoram/Western Kunlun (n = 160; 22.5%), Himalaya East/Hengduan Shan (n = 137; 19.3%), and Himalaya West/Central (n = 115; 16.2%). Smaller proportions are reported from the Hindu Kush (n = 19; 2.7%), Pamir and Alay (n = 13; 1.8%), and Tibet and Plateau/Fringe regions (n = 4; 0.6%), while records with insufficient information for sub-regional attribution are retained as unknown/unclassified (n = 8; 1.1%). These statistics should be interpreted as reported database records rather than a complete census of all historical outbursts, because observation intensity, source availability, and event documentation differ among sub-regions and across time [122,123].

For statistical reporting, GLOF records were grouped using a HiMAP/HMAGLOFDB-based regional aggregation. The detailed physiographic sub-regions shown in Figure 3 were aggregated into the broader categories used in

Table 5. GLOF event counts by HiMAP-based sub-region and lake–dam type in HMA.

Category	Class	Number of GLOF Events	Percentage of Total Records (%)
Sub-region	Tien Shan	255	35.9
	Karakoram/Western Kunlun	160	22.5
	Himalaya East/Hengduan Shan	137	19.3
	Himalaya West/Central	115	16.2
	Hindu Kush	19	2.7
	Pamir and Alay	13	1.8
	Tibet and Plateau/Fringe regions	4	0.6
	Unknown/unclassified by sub-region	8	1.1
	Total	711	100.0
Lake–dam type	Moraine-dammed	378	53.2
	Ice-dammed	207	29.1
	Supraglacial	73	10.3
	Water-pocket	22	3.1
	Unknown/unclassified by lake–dam type	22	3.1
	Bedrock-dammed	6	0.8
	Landslide-dammed	3	0.4
	Total	711	100.0

as follows: Dzhungarsky Alatau, Northern/Western Tien Shan, Central Tien Shan, and Eastern Tien Shan were grouped as “Tien Shan”; Pamir Alay, Western Pamir, and Eastern Pamir as “Pamir and Alay”; Karakoram and Western Kunlun Shan as “Karakoram/Western Kunlun”; Western Himalaya and Central Himalaya as “Himalaya West/Central”; Eastern Himalaya, Gangdise Shan, Nyainqentanglha, and Hengduan Shan as “Himalaya East/Hengduan Shan”; and the remaining Tibetan Plateau and fringe sub-regions as “Tibet and Plateau/Fringe regions.” Records with insufficient spatial information were retained as “unknown/unclassified.”

Decadal variability also differs by mountain-range sub-region and lake–dam type, highlighting that historical occurrence is not stationary in the reported record (Figure 5). The Tien Shan shows a strong mid-20th-century increase followed by a decline in the 1990s, whereas the Karakoram record exhibits multiple peaks beginning early in the 20th century, consistent with the occurrence of temporary lake formation associated with dynamic glacier behaviour in some basins. In the Himalaya and Hindu Kush, reported events show a more gradual rise, although totals remain lower than in the Tien Shan and Karakoram. These decadal patterns likely reflect both physical variability and major shifts in observation and reporting capacity, including the transition to satellite-era detection [122,123]. Accordingly, the historical record is best used to characterize event types, plausible process pathways, and regional contrasts, rather than to infer precise long-term frequency trends without careful bias treatment.

4.2. Historical GLOFs by Lake–Dam Type

4.2.1. Moraine-Dammed GLOFs

GLOFs in HMA arise from multiple lake–dam configurations, and the reported GLOFs are dominated mostly by moraine-dammed lakes. Moraine-dammed outbursts can be initiated by overtopping driven by mass movements or calving-related displacement waves, or by internal weakening processes such as seepage, piping, and melt-out of buried ice within unconsolidated dam material [94,96,124]. In the compiled record, moraine-dammed GLOFs account for ~53.2% (378/711) of reported GLOFs between 1900 and 2024, although their occurrence varies among sub-regions and lake settings [49,53]. Glacier-contact lakes can be especially sensitive to nearby disturbances, whereas detached proglacial lakes may become less exposed to some triggers as glacier–lake distance increases [125]. Overall, moraine-dammed events remain the largest contributor to reported historical GLOFs in HMA, consistent with the prevalence of moraine-impounded lakes in several mountain belts (Figure 3).

4.2.2. Ice-Dammed GLOFs

GLOFs from ice-dammed lakes are the second largest reported group, accounting for 207 records (29.1%) in the compiled dataset. Ice-dammed GLOFs are strongly controlled by glacier dynamics and drainage through or beneath the ice dam, including flotation and enlargement of subglacial conduits that can accelerate discharge rapidly once drainage initiates [85,95,126,127]. A defining characteristic is that ice dams can re-establish after drainage, enabling repeated outburst cycles in some settings [95,126]. In the compiled record, ice-dammed events are most common in Central Asia within HMA and less frequent in the Southeast Asia region (Figure 3). This type-specific behaviour is important because it affects both how events are documented historically and how recurrence is interpreted.

4.2.3. Subglacial and Bedrock-Dammed GLOFs

Subglacial/englacial and bedrock-dammed GLOFs are reported infrequently, and their representation in historical inventories is strongly shaped by detectability constraints. Subglacial outbursts can leave limited preserved geomorphic evidence and can be difficult to detect with optical remote sensing, contributing to under-reporting [128]. In contrast, bedrock dams are generally more stable than ice or unconsolidated moraine dams, and reported outbursts are rare and typically associated with overtopping, including displacement waves generated by mass movements entering the lake [9,10,11]. In HMA, bedrock-dammed events represent 0.8% of the compiled record (n = 6), while bedrock-dammed lakes can still be common in some Himalayan settings [49,53]. These contrasts emphasize that rarity in the event catalogue does not necessarily imply rarity of the lake type.

4.2.4. Water-Pocket and Landside-Dammed GLOFs

Water-pocket and landslide-dammed lakes contribute only a small fraction of reported events but remain relevant because their effects can be large despite low frequency. Water-pocket outbursts arise from small, isolated water bodies and account for 22 reported records (3.1%) in the compiled dataset [55]. GLOFs from landslide-dammed lakes are rare, accounting for three records (0.4%), and typically involve failure of loose valley-blocking material after overtopping or progressive erosion [10,11]. In addition, 22 records (3.1%) remain unknown or unclassified by lake–dam type because the available records do not provide sufficient information for confident attribution. Because geomorphic traces for small or rare lake types can be short-lived, these shares may be underestimated relative to moraine-dammed failures [129].

The historical record indicates strong lake–dam type dependence and pronounced regional clustering of reported GLOFs in HMA (Figure 3 and Figure 5), while remaining subject to substantial uncertainty due to incomplete documentation and reporting bias [33,122,123]. This historical evidence provides an observational basis for interpreting projected future hazards and for evaluating assessment and risk-reduction approaches in the following sections.

4.3. Documented Losses Associated with GLOFs in HMA

4.3.1. Human and Economic Losses

GLOFs in HMA have caused substantial loss of life and severe disruption to mountain communities, demonstrating that rare outburst events can produce extreme downstream consequences [3,55]. In the compiled database (Table S1) used for this review, reported fatalities total 7008 when compound and multi-hazard events are included. However, this aggregate value is strongly influenced by the 2013 Kedarnath/Chorabari disaster in Uttarakhand, India, where the lake-breach component was embedded within a broader compound event involving extreme monsoon rainfall, debris flows, and slope failures [55,93,123]. Excluding the 2013 Kedarnath/Chorabari fatalities from the primary GLOF-attributed mortality estimate reduces the fatality total to 1008. Therefore, the database-wide total and the more conservative primary GLOF-attributed estimate should be interpreted in parallel rather than as a single undifferentiated mortality statistic.

This distinction is important because global and regional GLOF databases differ in how they retain and describe compound-event losses. Lützow et al. [2] compiled historical GLOF records and associated consequences, while emphasizing that event documentation and attribution vary substantially across regions and time periods. Similarly, Shrestha et al. [55] preserved HMA event-level information on lake type, triggering conditions, drainage process, and downstream consequences, but many historical records still contain incomplete or uncertain attribution. Consequently, fatalities recorded in GLOF databases may include consequences from process chains in which lake drainage interacts with extreme precipitation, landslides, debris flows, or pre-existing slope instability. For this reason, loss statistics are best interpreted as reported database impacts with explicit attention to compound-event attribution, rather than as uniformly direct consequences of lake outburst alone.

Beyond fatalities, reported GLOFs have caused extensive damage to infrastructure, settlements, agricultural land, and hydropower systems across HMA. Documented consequences include at least 122 destroyed bridges, more than 2200 destroyed buildings, approximately 71 km² of affected agricultural land, severe damage to hydropower facilities with a combined capacity of about 164 MW, and reported economic losses of approximately USD 5.3 billion [55]. GLOFs commonly damage roads, bridges, hydropower facilities, irrigation infrastructure, and farmland, generating both direct repair costs and prolonged service disruption [47,123]. The 1985 Dig Tsho outburst in Nepal illustrates this pattern, as it destroyed a hydropower facility and damaged bridges and agricultural land, with long-lasting consequences for local development [33]. Because transport and energy assets are often concentrated along narrow mountain river corridors, disruption can propagate far beyond the immediate flood footprint by interrupting access, markets, and emergency response [122]. However, reported economic losses often exclude indirect and long-term costs and remain uncertain due to reporting gaps and limited post-event field verification [55].

4.3.2. Geomorphic and Ecological Disturbance

GLOFs can drive major geomorphic change by mobilizing large sediment volumes, scouring channel beds and banks, and depositing debris across floodplains, thereby altering river morphology and sediment connectivity [94]. Channel widening, braiding, and avulsion can modify conveyance and roughness and leave legacy conditions that persist for years to decades, influencing subsequent flood behaviour and infrastructure exposure [77]. Aquatic systems can experience abrupt increases in turbidity and fine sediment, degrading habitat quality and affecting freshwater biodiversity [130]. In some basins, sustained erosion and deposition have modified river courses and local base-level controls, contributing to longer-term shifts in hydrological behaviour downstream [33].

4.3.3. Cascading Processes, Transboundary Consequences, and Uncertainty

Loss estimation for individual events is subject to substantial uncertainty, especially when GLOFs occur during monsoon floods or other extremes that generate overlapping damage pathways [93]. Early post-event assessments may miss delayed and cascading damage, such as secondary landslides from undercut slopes that occur hours to weeks after peak flow, complicating attribution and accounting [77,94]. Comparability across sub-regions is further limited by incomplete reporting and shifts in observation capacity through time, including changes in gauge networks and satellite-era detectability [122,123]. These uncertainties mean that synthesized loss totals are best treated as conservative and context-dependent indicators rather than complete accounting.

GLOFs in HMA also commonly occur within cascading process chains, where avalanches, landslides, seismicity, and extreme precipitation can destabilize dams or generate displacement waves that amplify downstream destruction [94,122]. Because major rivers originating in HMA cross national boundaries, outburst floods can propagate far downstream and create cross-border consequences, increasing the importance of coordinated monitoring and response [131]. The 2016 Gongbatongsha event exemplifies this, where a lake outburst in Tibet evolved into a debris-laden flood that crossed into Nepal and damaged hydropower and transport infrastructure [99,132]. Together, cascading behaviour and transboundary propagation underscore why effects cannot be evaluated purely at the lake scale and motivate integrated risk-reduction planning, which is addressed in the next section [55].

5. Future Changes and Risk Management Strategies for HMA

5.1. Future Changes in Glacial Lakes and GLOF Hazard in HMA Under Global Warming

5.1.1. Future Glacial-Lake Development and Hotspots

Future changes in GLOFs across HMA are commonly inferred from how glacial lakes are expected to form, expand, and redistribute under global warming, because lake storage and lake–dam configuration provide the physical preconditions for outburst events [3,23,133]. However, lake growth alone does not uniquely determine outburst likelihood or event magnitude, because dam type and stability, local glacier dynamics (including surging), and initiating environments evolve differently among sub-regions [23]. Accordingly, scenario-based results should be interpreted as guidance on where future lake development and associated hazard hotspots may intensify, rather than as deterministic predictions of event timing. Model-based assessments consistently indicate substantial increases in glacial-lake area and volume across HMA by the end of the twenty-first century, but the interpretation of these results depends on the climate-scenario framework used. Furian et al. [23] used CMIP6 climate projections under Shared Socioeconomic Pathway (SSP) scenarios and projected substantial glacial-lake growth by 2100. Under SSP1-2.6, they estimated an additional 474 ± 121 km² of lake area and 22.8 ± 6.7 km³ of lake volume, whereas under SSP5-8.5 they estimated an additional 833 ± 148 km² of lake area and 39.7 ± 7.7 km³ of lake volume by 2100 [23]. These results show that future glacial-lake expansion is strongly scenario-dependent and spatially heterogeneous across HMA.

A clear distinction is required when interpreting projections derived from SSP- and RCP-based scenario frameworks. SSP5-8.5 and RCP8.5 both represent high-end pathways associated with approximately 8.5 W m⁻² radiative forcing by the end of the twenty-first century, but they are not identical. RCP8.5 is a CMIP5 concentration pathway, whereas SSP5-8.5 is a CMIP6 scenario that combines a high-forcing trajectory with a fossil-fuel-intensive socioeconomic storyline. Therefore, results derived from SSP5-8.5 and RCP8.5 can be interpreted qualitatively as evidence of high-emission or high-forcing futures, but they should not be merged quantitatively without acknowledging differences in scenario design, socioeconomic assumptions, emissions trajectories, climate-model generation, and forcing implementation. At the synthesis level, available projections indicate that future glacial-lake development in HMA should be interpreted as a spatially heterogeneous redistribution of lake storage, not simply as a uniform increase in lake area and volume [23]. High-forcing scenario assessments suggest that future lake-storage hotspots are likely to concentrate in particular mountain belts, but these spatial patterns should be interpreted qualitatively when comparing RCP- and SSP-based studies because the scenario frameworks differ in design, assumptions, and model generation [3,23]. Therefore, the most robust implication is that future hotspot assessment should focus on where expanding or newly forming lakes coincide with unstable glacier termini, steep surrounding slopes, moraine- or ice-dam conditions, and downstream exposure, rather than treating projected increases in lake area or volume as direct forecasts of GLOF occurrence.

A complementary perspective comes from inventories of potential future lakes based on over deepened basins exposed during retreat, which highlight where new lakes could form and where surrounding slopes may predispose lakes to mass-movement impacts [23,134]. Furian et al. [135] identified 25,285 potential future lake basins (>10⁴ m²) across HMA with total area 2683 ± 773.8 km² and volume 99.1 ± 28.6 km³, and they evaluated the predisposition of larger potential lakes to effects from adjacent slopes. Importantly, they reported a redistribution of future lake area/volume and associated hazard potential away from parts of the southwestern Himalaya toward the Karakoram, indicating that hotspot patterns may reorganize spatially as glaciers retreat. These findings reinforce that future lake change in HMA is best viewed as both an increase in total lake storage and a spatial reconfiguration of where the most consequential lake-slope settings may emerge.

5.1.2. Future GLOF Hazard and Risk Patterns

Syntheses combining historical evidence with future lake trajectories suggest that GLOF hazard in HMA is likely to intensify under ongoing global warming, although the magnitude and timing of change remain region-dependent [3,23]. Zhang et al. [3] highlighted the scenario-based evidence indicating substantial hazard increases by 2100 in HMA under strong global warming. Future GLOF hazard and risk patterns should therefore be interpreted through the combined effects of projected lake development, dam stability, trigger potential, downstream exposure, and vulnerability. Future assessments should not infer risk from lake expansion alone; instead, they should integrate lake storage, lake–dam configuration, slope-instability potential, hydrodynamic routing, settlement and infrastructure exposure, vulnerability, and local response capacity [3,23,129].

Future hazard change is also expected to be shaped by the emergence of new lake–slope configurations, not only by growth of existing lakes [23,133,135]. Where expanding or newly formed lakes coincide with steep, unstable terrain, mass-movement impacts into lakes can generate larger displacement waves and increase the likelihood of overtopping-driven failure, thereby amplifying hazard locally [23]. Consequently, future hotspot identification is most defensible when it jointly considers projected lake storage and the surrounding topographic predisposition to impacts, consistent with the hotspot-shift and slope-predisposition evidence reported for HMA [135].

5.1.3. Uncertainty and Interpretation Limits

Despite consistent signals of increasing lake area and volume, translating future lake estimates into robust forecasts of GLOF occurrence and magnitude remains uncertain [136]. A key limitation is that region-scale lake projections cannot reliably determine future dam type, outlet geometry, or outlet elevation, which strongly influence outburst susceptibility and breach behaviour, and these uncertainties propagate into any hazard inference [3,94,104]. Projection outcomes also vary because of differences in glacier-evolution assumptions, ice-thickness estimates, and how lake formation is represented, leading to scenario sensitivity and methodological dependence in both totals and hotspot location [23,135].

5.2. GLOF Hazards and Risk Management Strategies for HMA

Effective GLOF risk reduction in HMA relies on combining hazard and risk mapping, early warning systems, and complementary structural and non-structural measures. Hazard-focused tools identify plausible outburst pathways, flow depths, velocities, and inundation extents, whereas risk-focused tools combine hazard information with exposure and vulnerability indicators to prioritize downstream communities, infrastructure, and assets where losses would be most likely or severe [34]. Because GLOFs are typically rapid-onset events with cascading process chains, successful strategies must balance targeted engineering at selected lakes with preparedness and governance across downstream corridors [98,104]. Figure 6 conceptualizes GLOF hazard and risk management as an iterative cycle linking risk analysis with disaster risk reduction, response, and recovery, and shows how structural and non-structural measures fit within this cycle.

5.2.1. Hazard and Risk Mapping

Hazard and risk mapping is a cornerstone of GLOF management because it links lake conditions to downstream flow pathways and identifies where damaging depths, velocities, and inundation footprints are plausible under defined scenarios [90,137,138]. In practice, mapping integrates remote sensing and DEM-based terrain analysis with empirical lake-volume/peak-discharge estimation and scenario-based hydrodynamic modelling to generate hazard layers representing plausible flood extents and intensities [139,140]. When hazard layers are combined with exposure and vulnerability information (population, critical infrastructure, access constraints), the result is a ranked basis for prioritizing lakes, corridors, and communities for monitoring, early warning systems deployment, or targeted interventions [122]. Because lakes and downstream assets change over time, mapping is most useful when it is updated iteratively and benchmarked against observations where possible [141].

5.2.2. Early Warning Systems

Early warning systems are a key strategy for reducing loss of life because they translate monitoring into timely alerts and actionable responses [104]. Effective systems integrate four interdependent components: risk knowledge, monitoring and forecasting, communication, and response capability [27,142]. Monitoring typically combines in situ sensors (e.g., lake level, outlet discharge, precipitation, ground motion) with repeat satellite observations to detect rapid lake and slope change where access is limited [5,43]. Communication and response depend on redundancy (sirens, radio, mobile networks, community messengers) and standard operating procedures that define alert levels and evacuation routes and are tested through drills [93,104,122]. The most common operational limitations are false alarms, power or telemetry outages, and maintenance constraints, which reinforces the need to embed early warning systems within local governance and sustained training rather than treating it as a standalone technical installation [143].

5.2.3. Structural and Non-Structural Measures

Structural interventions reduce hazard by modifying lake storage, stabilizing outlets, or strengthening vulnerable dam sections, and they are most effective when designed using site-specific field investigation and scenario modelling [27,95]. Common measures include controlled lake lowering (e.g., syphons, outlet excavation), engineered and armoured spillways, outlet-channel stabilization, and selective dam reinforcement where feasible [34]. Performance depends on dam material and ice content, lake geometry and inflow regime, valley confinement, site accessibility, and long-term operation and maintenance capacity [34,104]. Because no engineering solution eliminates residual hazard, structural works should be implemented alongside monitoring and early warning systems to ensure that reduced peak outflow translates into reduced losses [24].

Non-structural measures reduce losses by lowering exposure and strengthening response capacity without modifying the lake itself. Core elements include land-use planning and hazard zoning, risk communication and education, and preparedness planning that co-develops evacuation routes, shelters, and standard operating procedures with local communities [79]. These measures are most effective when institutionalized, regularly updated, and linked to drills so that alerts lead to timely action [104]. Persistent constraints include limited institutional capacity, communication gaps, and financial limitations that restrict long-term maintenance and equitable access to warnings and safe relocation [22]. Because many HMA rivers are transboundary, interoperable warning protocols, coordinated data sharing, and joint response planning are essential for managing downstream cascading consequences across borders [132]. The relevance of these issues is illustrated by two recent policy-relevant cases: the 2023 South Lhonak Lake GLOF in Sikkim and the 2025 Gyirong/Rasuwagadhi transboundary flood on the China–Nepal border (Box 1). These cases demonstrate that operational GLOF risk reduction in HMA requires a shift from static lake inventories toward multi-sensor monitoring of lake evolution, slope instability, dam conditions, downstream exposure, hydropower vulnerability, and transboundary communication pathways.

Box 1. Recent policy-relevant GLOF cases in HMA: South Lhonak 2023 and Gyirong/Rasuwagadhi 2025.

South Lhonak Lake, Sikkim, India, 3–4 October 2023: The South Lhonak Lake GLOF is one of the most consequential Himalayan GLOF disasters in recent years and provides a clear example of a cascading high-mountain hazard process. Recent event reconstructions indicate that the disaster involved slope or moraine instability above South Lhonak Lake, displacement-wave generation, rapid lake drainage, downstream flood and debris-flow propagation, sediment entrainment, and interaction with hydropower infrastructure [68,69,70]. The event caused major downstream destruction, including failure of the 1200 MW Teesta-III hydropower dam, and demonstrates that lake-growth monitoring alone is insufficient without slope-instability assessment, hydropower exposure mapping, process-chain modelling, real-time warning, and emergency-response planning.

Gyirong/Rasuwagadhi, China–Nepal border, July 2025: The July 2025 Gyirong/Rasuwagadhi transboundary flood illustrates a different but equally important risk pathway: rapid drainage of a supraglacial lake in a transboundary basin. Preliminary satellite-based assessments and media reports indicated that the flood originated from drainage of a supraglacial lake in Tibet, China, and propagated into Nepal’s Bhote Koshi/Rasuwagadhi corridor, causing fatalities, missing persons, damage to cross-border infrastructure, disruption of the Nepal–China trade corridor, and impacts to hydropower and transport systems [144,145,146]. Because some details of this event remain preliminary, it is treated here as a recent policy-relevant case rather than as a fully finalized database entry. Nevertheless, it highlights the urgent need for cross-border glacial-lake monitoring, rapid satellite-based event attribution, shared warning protocols, and coordinated disaster response in HMA transboundary river corridors.

Synthesis: Together, these two cases show that recent GLOF risk in HMA is increasingly shaped by compound and cascading processes rather than lake breaching alone. South Lhonak emphasizes the need to integrate slope-instability monitoring, hydropower exposure, and process-chain modelling into pre-event hazard assessment, while the Gyirong/Rasuwagadhi event highlights transboundary data-sharing and early-warning challenges. Both cases strengthen the argument that remote sensing should be used not only for periodic lake inventories but also for near-real-time monitoring, event reconstruction, exposure assessment, and operational risk reduction.

5.2.4. Integrated Implementation Framework

A coordinated approach is required to translate GLOF hazard assessment into durable risk reduction in HMA, combining mapping, monitoring, early warning system and structural interventions at prioritized lakes, and governance mechanisms that support sustained maintenance and preparedness [98]. Figure 6 conceptualizes GLOF hazard and risk management as an iterative cycle linking risk analysis with disaster risk reduction, response, and recovery, and it shows how structural and non-structural measures fit within this cycle. Figure 7 presents a five-phase pathway for building GLOF-resilient communities, moving from diagnosing vulnerabilities to strengthening monitoring and early warning, building awareness and preparedness, developing resilient infrastructure and planning, and sustaining coordinated action. Together, these frameworks highlight that effectiveness depends as much on sustained capacity and governance as on technical modelling and engineering design. The distinctive value of the proposed frameworks is that they translate the review evidence into an HMA-focused implementation pathway by linking remote-sensing-based monitoring, historical GLOF records, future lake-change assessment, uncertainty propagation, early warning, mitigation measures, governance coordination, and community resilience within a single operational cycle.

The phases shown in Figure 7 require different but coordinated responsibilities. Scientists, remote-sensing specialists, glaciologists, hydrologists, geomorphologists, and modelling experts are primarily responsible for generating and updating glacial-lake inventories, detecting lake and slope changes, modelling breach and flood scenarios, assessing uncertainty, and translating technical results into interpretable hazard and exposure products. Hydrometeorological agencies and disaster-management authorities are responsible for converting these products into warning thresholds, alert levels, standard operating procedures, and operational monitoring protocols. Local governments, infrastructure operators, NGOs, and community organizations are responsible for evacuation planning, preparedness drills, public communication, maintenance of warning infrastructure, and last-mile dissemination. Therefore, the transition from scientific knowledge to actionable risk reduction depends on sustained communication between researchers and authorities, including shared databases, co-produced hazard maps, agreed warning thresholds, periodic training, and feedback after drills or events.

Existing early-warning and mitigation examples show that technical systems alone are insufficient unless they are embedded in institutional and community response structures. In Nepal, Tsho Rolpa and Imja Tsho illustrate the role of lake lowering, monitoring, warning systems, and community preparedness in reducing potential downstream losses [63,64]. In Bhutan, the Punakha-Wangdue GLOF early-warning system demonstrates the importance of formal standard operating procedures, agency roles, sirens, communication chains, and preparedness planning [65]. Monitoring systems such as Cirenmaco in the Tibetan Plateau further illustrate how real-time sensors and defined warning thresholds can support downstream warning where maintenance capacity and communication pathways are reliable [66]. These examples indicate that effective GLOF early warning requires risk knowledge, monitoring and detection, communication of warnings, and response capacity [67].

6. Synthesis-Level Lessons and Future Research Priorities

6.1. Synthesis-Level Lessons from GLOF Research in HMA

Across HMA, the evidence reviewed in this paper shows that GLOF hazard is controlled by the coupled evolution of glacial lakes, dam stability, glacier dynamics, slope processes, hydroclimatic extremes, and downstream exposure. Five synthesis-level lessons emerge. First, historical GLOF records remain incomplete and unevenly documented across sub-regions, which limits robust inference about long-term frequency trends and trigger attribution. Second, methodological differences among lake inventories, minimum mapping thresholds, lake-volume equations, susceptibility indicators, DEM products, and hydrodynamic models can strongly influence hazard interpretation. Third, uncertainty propagates through the full assessment chain, from lake mapping and volume estimation to breach modelling, flood routing, exposure assessment, and risk prioritization. Fourth, recent events demonstrate that GLOFs often occur as part of compound and cascading process chains involving rainfall, slope failure, glacier instability, sediment entrainment, and infrastructure exposure. Fifth, remote-sensing products are most useful for risk reduction when they are linked to field validation, early warning, institutional coordination, and operational decision-making. These lessons provide the basis for the future research priorities outlined below.

6.2. Concrete Future Research Priorities for GLOF Assessment and Risk Reduction in HMA

6.2.1. Build Harmonized Multi-Temporal Lake Inventories for GLOF Applications

A first priority is to build consistent multi-temporal glacial-lake inventories for HMA that are explicitly designed for GLOF applications, using consistent mapping thresholds, glacier–lake proximity rules, and temporal coverage. These inventories should report lake type and glacier-contact state where possible, because these attributes directly affect GLOF susceptibility interpretation and scenario selection. Explicit quality control is needed for seasonal ice, shadows, debris influence, and cloud contamination, along with transparent version control so GLOF-relevant trends are comparable through time. Open reference datasets for training and validation are essential so automated mapping can scale without losing accuracy or temporal consistency. A unified inventory baseline would immediately improve the consistency of GLOF lake screening, hotspot identification, and multi-temporal monitoring across HMA.

6.2.2. Expand Targeted Field Observations of GLOF-Critical Parameters

A second priority is targeted field measurement of the parameters that most strongly control GLOF magnitude and model uncertainty, particularly lake bathymetry, outlet geometry and elevation, and dam composition including ice content and evidence of seepage or piping. Because field resources are limited, campaigns should focus on high-leverage lakes identified through screening and hotspot analyses, where improved constraints would most reduce uncertainty in breach development and peak outflow. Field datasets should be collected in a way that directly supports model calibration and benchmarking, including repeat measurements where feasible. Expanding calibration datasets for area–volume relations and dam properties will reduce systematic bias in regional GLOF scenario modelling and improve transferability across sub-regions.

6.2.3. Integrate Heatwaves and Compound Extremes into GLOF Hazard Assessment

A third priority is to treat warming-driven extremes explicitly, especially heatwaves and compound warm and wet periods, as both conditioning factors and potential initiators of GLOFs. Future studies should link extreme-temperature and intense-precipitation diagnostics to lake-level changes, snow and ice conditions, and slope-instability indicators, because these factors can elevate overtopping likelihood and breach potential over short timescales. This is necessary to stress-test GLOF early-warning thresholds, scenario design, and engineering assumptions under non-stationary extremes. Heatwave-aware monitoring and modelling will also help identify where GLOF hazard may increase disproportionately compared with background lake growth alone.

6.2.4. Standardize Uncertainty Propagation and Benchmarking of GLOF Models

A fourth priority is to propagate uncertainty consistently from lake mapping through volume estimation, breach generation, and routing, so GLOF hazard outputs are delivered as credible ranges rather than single-point estimates. In parallel, benchmarking datasets and minimum reporting standards are needed for comparing GLOF breach and routing models, including documentation of DEM resolution, roughness assumptions, breach parameterization, calibration evidence, and performance metrics. Where relevant, modelling should represent cascading GLOF process chains such as mass movement into lakes producing impulse waves and sediment-rich floods, because simplified representations can underestimate hazard in steep terrain. Consistent benchmarking will improve comparability across studies and increase confidence in decision-ready outputs.

6.2.5. Operationalize Early Warning, Risk Reduction, and Transboundary Coordination

A final priority is operationalization, ensuring that improved assessment products can be sustained through interoperable monitoring, warning, and response systems in remote and transboundary basins. Early warning must be designed around redundancy, maintenance feasibility, clear alert protocols, and regular drills that reliably convert warnings into action. Because many HMA rivers cross borders, compatible data standards, coordinated alerting, and joint response planning are essential to reduce downstream consequences beyond the source catchment. Strengthening institutional capacity and long-term funding models is therefore as important as technical advances for durable risk reduction. Together, these priorities translate the synthesis-level lessons into actionable research and operational needs for improving GLOF assessment, monitoring, and risk reduction across HMA.

7. Conclusions

GLOFs remain a major cryosphere-driven hazard in HMA, where glacier mass loss, changing hydroclimatic extremes, and evolving glacial-lake systems are reshaping high-mountain risk landscapes. This review provides an HMA-focused, remote-sensing-oriented synthesis of historical GLOF evidence, glacial-lake evolution, assessment methods, future hazard changes, uncertainty sources, and risk-reduction strategies. The reviewed evidence shows that reported GLOFs are strongly clustered by sub-region and lake–dam type, but historical records remain incomplete and affected by uneven documentation and attribution uncertainty.

The review further demonstrates that remote sensing has become central to GLOF assessment in HMA, supporting lake inventory generation, lake-level and storage monitoring, susceptibility screening, breach and hydrodynamic modelling, exposure mapping, and event reconstruction. However, the reliability of GLOF hazard and risk assessment depends on how uncertainties propagate through lake mapping, volume estimation, dam and slope characterization, model parameterization, and exposure analysis. Future GLOF hazard is expected to be shaped not only by glacial-lake expansion but also by lake–dam stability, triggering processes, compound extremes, slope instability, and downstream exposure.

Effective GLOF risk reduction in HMA therefore requires integrated systems that combine harmonized lake inventories, targeted field observations, multi-sensor monitoring, transparent uncertainty assessment, early warning, structural and non-structural measures, and transboundary coordination. By linking Earth-observation evidence with process understanding and operational risk-management needs, this review highlights a pathway toward more credible and actionable GLOF assessment under continued warming.