Glacial lake outburst floods threaten millions globally

The study addresses the global danger from glacial lake outburst floods (GLOFs) by integrating physical lake conditions with social exposure and vulnerability, components that have not been jointly quantified at global scale. The authors situate the research within accelerating glacier mass loss and expansion of glacial lakes due to climate warming, highlighting complex and location-specific GLOF triggers (e.g., impulse waves, overfilling, dam degradation) that make probabilistic hazard assessments difficult without local data. Historical impacts vary widely by region (e.g., thousands of deaths in Peru versus hundreds in the Alps over centuries), underscoring the importance of local exposure and vulnerability. While glacial lakes and downstream populations have increased since 1990, global socio-economic vulnerability is thought to have decreased heterogeneously. Prior global efforts focused largely on physical hazard; this work seeks to quantify contemporary (circa 2020) GLOF danger by combining lake conditions (as a proxy for intensity), exposure (population near modeled runoff tracks), and vulnerability (national/subnational indices), to identify hotspots and guide mitigation and research priorities.

The paper synthesizes prior work on glacier sensitivity to climate change, global glacier mass loss, and the rapid growth of glacial lakes since 1990. It reviews GLOF trigger mechanisms (mass-movement-induced waves, runoff overfilling, moraine/ice dam degradation) and past regional hazard assessments in High Mountain Asia and the Andes. The authors note that previous global-scale assessments largely emphasized hazard (lake number/area/volume, geomorphological predisposition) without fully integrating exposure and social vulnerability. They also summarize regional disparities in research focus, with earlier hotspots in Iceland, North American Cordillera, and Hindu Kush–Karakoram, and a recent shift to the Himalayas, leaving the Andes comparatively under-studied despite growing lake numbers and deglaciation.

The authors develop a semi-quantitative GLOF Danger Index (GDI) combining three normalized components: lake conditions (proxy for potential intensity), exposure (population along modeled outburst paths), and vulnerability (social and institutional susceptibility). Key steps:

• Study units: 1,089 level-4 Global Water Resource Zones (WRZ) containing glacial lakes, grouped into European Alps, Andes, High Mountain Asia (HMA), Pacific Northwest (PNW), and a residual group termed High Arctic and Outlying Countries. The authors acknowledge WRZs are not true catchments but use them for global consistency.
• Lake conditions (intensity proxy): Number and total area of mapped glacial lakes per basin/country/region were normalized (y = X/Max) and multiplied to produce a 0–1 score. Lake area is used in preference to volume due to global inconsistency in area–volume scaling and DEM artifacts in high mountains. Probability of failure is treated as unknown; the metric is consequences-oriented, flagging basins where a failure could be more intense.
• Exposure: Potential GLOF runoff paths were approximated by delineating river networks downstream of each glacial lake up to 50 km, with exposed populations defined as those residing within 1 km of these river courses. Population exposure per basin/country/region was normalized by the maximum exposed population to yield a 0–1 exposure score. Populations were also binned by distance from source lakes (5 km increments up to 50 km; some analyses referenced 15 km intervals) to profile proximity distributions.
• Vulnerability: Combined three indicators with equal weighting after normalization to a 0–1 range: Corruption Perceptions Index (CPI, national), Human Development Index (HDI, subnational at first administrative level aggregated), and a national Social Vulnerability Index (SVI) adapted to avoid double-counting and missing data. SVI excluded four variables (safe drinking water, good sanitation, literacy, unemployment) due to redundancy/missingness. The final proxy for vulnerability ranges from 0 (least) to 1 (most vulnerable); equation presented as Vulnerability = 1 – [HDI × 1 – CPI × SVI] after normalization and equal weighting. The approach captures both local population fragility and national response capacity, while acknowledging hidden finer-scale heterogeneity.
• GLOF danger computation: GDI = Lake Conditions × Exposure × Vulnerability. Results were ranked (10 highest to 1 lowest) to identify hotspots.
• Data sources: Glacial lake inventories (1900–2018), global population datasets, national CPI (Transparency International), HDI (UNDP), national SVI, and WRZ basin delineations. The framework is designed to be repeatable for temporal updates as data improve.

• Globally, an estimated 15 million people are exposed to potential impacts from GLOFs.
• Exposure concentration: Over 50% of exposed populations are in four countries—India, Pakistan, Peru, and China.
• Regional exposure: High Mountain Asia (HMA) has the highest exposure and populations living closest to glacial lakes; approximately 1 million people live within 10 km downstream of a glacial lake. About 2% (~300,000) of the globally exposed population reside within 5 km of a glacial lake, two-thirds of whom (≈66%; ~198,000) are in HMA. Populations in the PNW and High Arctic/Outlying Countries tend to be >35 km downstream.
• Proximity distribution: Nearly half (48%) of exposed populations globally are located 20–35 km downstream of lakes.
• National exposure ranking: India scores highest on exposure (1.000), Pakistan second (0.701); Sweden lowest (0.001). Regionally, HMA scores highest on exposure (1.000), High Arctic/Outlying Countries lowest (0.109).
• Lake-condition (hazard proxy) patterns: Highest individual lake-condition scores occur in Greenland (1.000) and Canada (0.685), lowest in Ecuador (0.001). Within HMA, most countries have scores <0.100 except China (0.319). Greenland, despite the highest hazard proxy, has no population along modeled runoff tracks, yielding zero danger.
• Vulnerability: HMA is most vulnerable (0.768), PNW least (0.336). Afghanistan (0.919) and Pakistan (0.837) are the most vulnerable nations, while Switzerland (0.194) and New Zealand (0.156) are least vulnerable.
• Research gap: Despite high danger potential, the Andes are under-studied; the region’s number of glacial lakes increased by 93% over two decades (versus 37% in HMA), and the region hosts the second- and third-most dangerous basins. Peru ranks third globally in national danger.
• Drivers of danger: Exposed population and vulnerability strongly modulate danger; areas with many/larger lakes do not necessarily have high danger if exposure is low (e.g., Greenland).

The study demonstrates that contemporary GLOF danger is governed not only by lake conditions but critically by where people live relative to potential flood paths and their capacity to cope. HMA stands out for high exposure and proximity—particularly in Pakistan—implying low warning times and high uncertainty in event magnitudes for many communities. The Andes display comparable danger to HMA due to rapidly growing lake numbers, rising downstream populations, and persistent social vulnerability, yet the region remains under-researched. The findings argue for a holistic risk paradigm integrating hazard, exposure, and vulnerability rather than focusing solely on physical metrics like lake area/volume. Anticipated expansion of agriculture, hydropower, and tourism at higher elevations is likely to increase exposure, especially in HMA and the Andes. The authors emphasize urgent needs for early warning systems, evacuation planning, zoning, and community outreach tailored to social and cultural contexts, particularly where resources are limited and awareness is low. They highlight that similar outburst events can have vastly different consequences depending on socio-political and economic contexts, reinforcing the centrality of vulnerability in determining impacts. Although the framework provides a global baseline for current danger, temporal dynamics remain uncertain; future danger will evolve with changing lake conditions, population distributions, and socio-economic development.

By combining normalized proxies for lake conditions, exposure, and vulnerability into a GLOF Danger Index, the study quantifies and maps contemporary (circa 2020) global danger, identifying High Mountain Asia and the Andes as principal hotspots. Over 15 million people are exposed globally, with a majority concentrated in India, Pakistan, Peru, and China, and about one million people living within 10 km of glacial lakes in HMA. The analysis shows that exposure and vulnerability often outweigh physical lake metrics in driving danger, explaining why regions with extensive lakes (e.g., Greenland) can have negligible danger due to minimal exposure. The authors provide a globally consistent, updateable framework to guide prioritization of mitigation and research. They call for targeted national- and basin-scale risk assessments, expanded early warning and preparedness, and focused research in under-studied high-danger regions—especially the Andes—and for future work to evaluate temporal changes in lake conditions, exposure, and vulnerability to understand evolving danger trajectories.

• Probability not modeled: The framework omits explicit probability of lake failure due to data limitations; lake area and number serve solely as intensity proxies. Thus, high potential impact areas may still have low risk if failure probability is low.
• Basin delineation: Use of Global Water Resource Zones (not true hydrological catchments) can cause spatial artifacts, especially in large plains/coastal regions (e.g., Chilean Patagonia).
• Simplified exposure modeling: Runout approximated to 50 km along river networks with a 1 km buffer may underestimate far-field impacts (some GLOFs exceed 120 km) or overgeneralize pathways; discharge attenuation with distance not explicitly modeled.
• Vulnerability proxies and scale: National/subnational indices (CPI, HDI, SVI) mask finer-scale heterogeneity; equal weighting and variable omissions in SVI (due to double-counting or missingness) introduce uncertainty.
• Global data constraints: DEM artifacts in high mountains hinder globally consistent geomorphic predictors; area–volume relationships are not applied due to inconsistency, potentially missing volume-related nuances.
• Potential textual inconsistencies: Some reported exposure counts and percentages in results sections appear internally inconsistent, reflecting data/reporting limitations in the source text; main headline figures are emphasized.