Introduction

Climate warming induces extensive degradation of the cryosphere1,2,3. The widespread recession of mountain glaciers has resulted in two main influences: the rapid growth of glacial lakes4,5 and the frequent occurrence of ice avalanches and glacial lake outburst floods (GLOFs) in the cryosphere6,7,8,9,10. Globally, proglacial lakes (in contact with glacier terminus) are unfolding a narrative of lake expansion and ice retreat11. Other glacial lakes, including those developed on glacier surfaces or located downstream from glaciers, are expanding due to glacier downwasting and increased glacier meltwater12. Despite their relatively small area, generally less than 1 km2, glacial lakes are likely to dominate the number of lakes in alpine areas. For more than 680 million people residing in high mountain regions on Earth, the formation and expansion of glacial lakes contribute to regional water storage, hydropower development, and tourism13,14. In regions such as the European Alps, many alpine reservoirs have been constructed, storing glacier meltwater that would otherwise be lost15. However, these regions face the severe threat of abrupt and destructive GLOFs simultaneously originating from moraine- and ice-dammed lakes16. A recent global inventory of GLOFs has documented over 3000 events17, with the majority occurring in the Third Pole (encompassing the Tibetan Plateau and surrounding Himalayas, Karakoram, and Tianshan Mountains)18,19,20, Patagonian Andes21,22, and Alaska23,24, and these events have caused tens of thousands of deaths and substantial property losses25,26,27.

Scientifically, scholars have also delved into the processes of proglacial lake–glacier interactions28,29, the risk of glacial lake outbursts30,31,32, and the role of glacial lakes in the carbon balance in ice-marginal landscapes33,34. Analyzing these process mechanisms and assessing their impact potential will inform scientific recommendations for future economic development planning and local-level disaster prevention and mitigation efforts. A complete inventory of glacial lakes is fundamental for understanding these broader scientific questions related to glacial lakes. In recent years, numerous inventories of glacial lakes have been compiled across various glacierized regions, sometimes leading to redundant research efforts. Discrepancies in glacial lake delineation criteria have emerged, differing in the Third Pole, the Southern Andes, Alaska, and the Greenland Periphery. For example, questions arise regarding how far a water body within the buffer distance of a contemporary glacier can be considered a glacial lake and whether hydraulic connectivity with the glacier is a prerequisite. The inconsistent delineation criteria hinder regional comparative studies and comprehensive assessments. Shugar et al.4 pioneered the development of the first global inventory of glacial lakes. The reported count of 9414 glacial lakes worldwide in 2015–2018 is even less than some regional counts of glacial lakes. These differences are not actually discrepancies but are differences in measurement protocols. The primary reasons are that the choices of buffer distance and minimum extraction threshold. Shugar et al.4 are more concerned with finding ice-contact lakes and those which have very recently detached glaciers within a 1 km buffer around the global glaciers, with an area greater than 0.05 km2. In contrast, other studies often utilize larger buffer distances and smaller minimum extraction thresholds admitting lakes where the glaciers have detached decades or some centuries ago. The exhaustive mapping approaches have the potential to give a more holistic analysis of global glaciers and GLOF dynamics and meltwater storage, thereby enhancing the research scope of glacial lake changes and their association with climate-induced glacier retreat.

Given the abovementioned research deficiencies and the need to regularize glacial lake research, we evaluated glacial lake changes and their drivers on a global scale and discussed the research standards in glacial lake extraction and analysis. Using Landsat and Sentinel-2 images, we mapped an accurate inventory of glacial lakes to characterize their distribution and evolutionary features between 1990 and 2020. Through empirically resolved volume estimation, we quantified the glacial lakes’ interception function of sea-level rise by retaining glacier meltwater. In addition, we demonstrated the responses of different types of glacial lakes to climate warming and glacier thinning.

Results

Glacial lake distribution and changes worldwide

During the lake delineation process, a preliminary sample of lakes was acquired by extracting water bodies within a buffer distance of 10 km around global glaciers. Lakes with a hydraulic connection to the glaciers were identified as glacial lakes and were further categorized into glacier-fed lakes and ice-dammed lakes (see “Methods”). Furthermore, water bodies located within this 10 km buffer that were not fed by contemporary glacier meltwater were also delineated and classified as non-glacier-fed lakes. However, these non-glacier-fed lakes were not included in the glacial lake analysis.

For imagery acquired in 2020, a total of 71,508 glacial lakes (>0.002 km2) were systematically mapped on the foreland and surface of global glaciers using Sentinel-2 images (Fig. 1), covering an area of 21,770.9\(\pm\)544.2 km2. The collective volume of these lakes amounted to 1280.6\(\pm\)354.1 km3. At the regional scale, six glacierized regions hosted more than 5000 glacial lakes each: the Greenland Periphery (14,845 glacial lakes, constituting 21% of the total), the Third Pole (encompassing Central Asia, Southwest Asia, and Southeast Asia; 12,828, 18%), Western Canada and USA (8981, 13%), Alaska (8855, 12%), southern Arctic Canada (6532, 9%), and the Southern Andes (6021, 8%). Regarding the areal distribution, the glacial lakes in the Third Pole had a relatively lower average area, while the total glacial lake area in the other glacierized regions corresponded with their respective rankings in terms of the number of glacial lakes. To the best of our knowledge, we developed the first glacial lake inventory in Antarctica and the Subantarctic, including 605 glacial lakes with a total area of 159.05 ± 4.95 km2 and a volume of 10.6 ± 4.1 km3. The feature intervals showed that the small lakes (<0.1 km2) accounted for 79% of the total number, 6% of the area, and 11% of the volume of the global glacial lakes. The medium lakes (0.1–1 km2) accounted for 17%, 17%, and 7%. The large lakes accounted for 4%, 77%, and 92%. The altitude concentration intervals revealed that the glacial lakes in the Third Pole had the highest elevations, with an average value of 4515 m, while the elevations of a few glacial lakes in other glacierized regions exceeded 3000 m.

Fig. 1: Spatial distribution of global glacial lakes in 2020 and the intercepting effect on sea-level rise.
figure 1

The number and area of glacial lakes were aggregated for 1° × 1° tiles. The circles were scaled based on the number of glacial lakes in each tile and were colored according to the corresponding glacial lake area. The 19 Randolph Glacier Inventory (RGI6.0) regions were utilized to aggregate and display the glacial lake volume in 2020. The impeding effect of the global growth of glacial lake water on sea-level rise since the 1990s was also quantified and presented separately by region.

Full size image

For imagery acquired in the 1990s, we identified 46,520 glacial lakes worldwide using Landsat images. The total area and volume were 19647.9\(\pm\)1311.7 km2 and 1170.4\(\pm\)298.4 km3, respectively, demonstrating relative increases in the number, area, and volume of 54%, 11%, and 9%, respectively, over the past three decades (Supplementary Tables 1 and 2). The changes in the glacial lake area and volume were primarily contributed by the medium and large lakes. However, the capacity for climate and other forcings to affect relative percentage changes in area and volume of individual lakes is generally greater for numerous smaller lakes, which nevertheless could not exert a dominant influence on the total population’s relative changes in area and volume. In the Southern Andes, for example, a previous regional inventory of glacial lakes disclosed relative increases of 43% and 7% in their number and area during 1986–201621. At the regional scale, in our analysis, Alaska experienced the largest increase in the absolute glacial lake area, increasing by 648.7\(\pm\)62.6 km2 (25%), with a corresponding volume increase of 39.8\(\pm\)9.1 km3 (23%). The Greenland Periphery experienced the second largest increases (395.5\(\pm\)245.2 km2, 6%; 17.9\(\pm\)4.9 km3, 5%), followed by the Southern Andes (291.4\(\pm\)64.7 km2, 13%; 15.1\(\pm\)3.1 km3, 11%) and the Third Pole (190.9\(\pm\)65.6 km2, 27%; 7.2\(\pm\)2.6 km3, 33%). Overall, the glacial lakes in global glacierized regions were in an expansion state, except for Northern Arctic Canada, which was mainly affected by the diminished area of ice-dammed lakes.

Impacts on meltwater storage and delaying (intercepting) sea-level rise

By retaining the glacier meltwater flowing into the ocean, the volumetric increase in glacial lakes has the potential to intercept global sea-level rise. Previous assessments have indicated that in specific regions such as the western Greenland Periphery, the Northern Patagonia Icefield, and the Third Pole, this intercepting impact could contribute 1–10%35,36,37. Since the 1990s, there has been a net increase in the global glacial lake volume of 117.1\(\pm\)31.5 km3 (Supplementary Table 3), equivalent to four times that of the glacial lake total volume in the Third Pole in 2020. Assuming a mean ocean density of 1027 kg/m3 and a total area of 3.6\(\times\)108 km2, and considering that 370.8 km3 of water equates to a 1 mm rise in the global sea level38, the volumetric increase corresponds to a sea-level rise of ~0.3 mm, which is 0.16 mm greater than previous estimates4. Similarly, the collective volume of global glacial lakes in 2020 held about 3.45 mm of sea-level potential. During the same study period, we extrapolated and estimated a total ice loss of 7129.7 Gt for global glaciers39,40, contributing to a sea-level rise of 19.2 mm. This indicates that the intercepting impact of the global growth of glacial lakes is a modest 1.55% of the worldwide negative glacier mass balance. At the regional scale, Alaska, the Greenland Periphery, the Southern Andes, and the Third Pole accounted for 36%, 16%, 14%, and 7% of the intercepting impact, respectively (Fig. 1). Moreover, the relative contribution of glacial meltwater retained by glacial lakes varied considerably among regions. In New Zealand, the expansion of glacial lakes impeded approximately 9% of glacier meltwater during 1990–2020, followed by 6% in Scandinavia and 4% in the Southern Andes. This variation was determined mostly by regional glacial lake expansion and the relative glacial lake-glacier abundance. In addition, in the European Alps, Scandinavia, and Iceland, many alpine reservoirs have been constructed in past decades, such as the Kharanhjukar Hydroelectric Project (15°20′W, 64°55′N) in Iceland, with a storage area of ~56.7 km2. These reservoirs also play a crucial role in impeding runoff.

Ice-dammed lake changes and outbursts

We distinguished ice-dammed lakes from glacial lakes and analyzed their distribution and changes. A total of 4733 ice-dammed lakes were identified worldwide in 2020, with a combined area of 3310.8\(\pm\)72.7 km2 and a combined volume of 175.8\(\pm\)67.1 km3 (Supplementary Table 2). Ice-dammed lakes are generally impounded behind glacier dams and accumulate water from upstream glaciers, or they are formed along the downwasting margins of glaciers41. The stability of these lakes, situated at relatively high altitudes and reliant on effective ice dam blockage, has been remarkably affected by the consistent increases in atmospheric temperatures and ice dam thinning42,43. Since the 1990s, an increase in the number of ice-dammed lakes has been observed across all glacierized regions. However, this does not imply parallel increases in the relevant area and volume. While Southern Arctic Canada, the Greenland Periphery, the Russian Arctic, and the Third Pole experienced simultaneous relative increases in the number, area, and volume of ice-dammed lakes, Alaska experienced a relative increase of 64% in the number, yet the total area and volume decreased by −12% and –29%, respectively (Fig. 2). Similar trends were observed in Northern Arctic Canada (–8% reduction in relative volume), Iceland (–93%), Svalbard (–2%), and the Southern Andes (–9%). We noted that a large number of ice-dammed lakes experienced decreases in area and drainage due to the thinning and instability of ice dams. In particular, for larger pre-existing ice-dammed lakes, they were compelled to reduce their volume to balance the impact between ice dam thinning and static water pressure44,45, dominating the heterogeneous changes in the number, area, and volume of the ice-dammed lakes in these regions. The numerous newly formed ice-dammed lakes, located along the downwasting margins of glaciers with small sizes, were unable to compensate for the overall volume loss of the ice-dammed lakes.

Fig. 2: Global distribution of ice-dammed lakes in 2020.
figure 2

The larger ice-dammed lakes take precedence in the presentation. The circles were scaled according to the regional number of ice-dammed lakes, as well as the status of their volumetric changes.

Full size image

GLOFs originating from ice-dammed lakes are one of the most frequent and catastrophic cryospheric hazards. These events are characterized by a decreased trend in magnitude, earlier drainage occurrence, and manifestation at higher altitudes, and they are attributed to glacier thinning and destabilization of ice-dammed lakes over past decades46. A prior study revealed that the frequency of GLOFs originating from ice-dammed lakes has decreased since the 1970s, with an average annual occurrence of approximately 25 GLOFs47. This trend is contrary to simple expectations from global warming, partially due to the delayed response between climate warming, glacier thinning, and lake outbursts48. With certainly, it is also probably attributed to the incomplete nature of the global GLOF database. In Alaska, for example, Rick et al.49 documented 1150 GLOFs originating from ice-dammed lakes between 1985 and 2020 using a satellite approach, which is three times the previously recorded 339 GLOFs. We exported 2039 GLOFs originating from ice-dammed lakes from the global GLOF database17. The geographic distribution of these GLOFs revealed disparities with the primary distribution of the number and area of global ice-dammed lakes. Most regions, such as Northern and Southern Arctic Canada, the eastern and northern Greenland Periphery, the Russian Arctic, the interior of the Tibetan Plateau, and the Antarctic and subantarctic still have a remarkable potential to detect more drainage events from ice-dammed lakes in the past and will produce more in the future, which will contradict the observed decreasing trend in the frequency of GLOFs. For example, in the Northern Greenland Periphery, we found two large ice-dammed lakes that experienced a marked reduction in area and underwent at least three drainage events collectively during past decades (Supplementary Fig. 1). In addition, sporadic GLOFs reported in the interior of the Tibetan Plateau and the Antarctic region indicate further research potential50,51.

Drivers of spatial changes and differences

A distinct expansion of glacial lakes has been observed worldwide since the 1990s, with the exception of Northern Arctic Canada, the western and eastern Greenland Periphery, and a portion of Northern Asia (Fig. 3a). The regional patterns of glacial lake change closely aligned with the resolved large-scale, decadal changes in glacier mass (Spearman’s correlation coefficient r = 0.26). Pronounced dipoles, indicative of spatial concordance between glacial lake expansion and glacier mass loss, were notably observed in Alaska, Western Canada and the USA, the Southern Andes, and the Third Pole (Fig. 3b). Considering their definition, glacial lakes are mainly affected by glacier retreat and melting. In the Third Pole, an assessment revealed that glacier meltwater could contribute to more than 90% of the increased water volume of glacial lakes37, and the remainder was contributed by precipitation and permafrost degradation. Globally, changes in glacier mass were primarily caused by temperature changes and localized impacts from precipitation fluctuations39. Due to the evident time-lagged response between climate warming, glacier recession, and glacial lake expansion48,52, climate change only exerted an indirect influence on glacial lakes.

Fig. 3: Spatial patterns of global changes in glacial lakes, non-glacial lakes, glacier thinning, and precipitation over past decades.
figure 3

a Glacial lake expansion rate during 1990–2020. b Glacier mean elevation change rate during 2000–2019. c Non-glacier-fed lake expansion rate during 1990–2020. d Difference between 1990–1999 and 2010–2019 in the mean annual precipitation based on CRU TS V4.05. Each factor was aggregated for 1° × 1° tiles.

Full size image

In addition, we conducted a comparative analysis of the spatial changes in global glacial lakes and non-glacier-fed lakes during 1990–2020 (Fig. 3c). While non-glacier-fed lakes exhibited differences in their spatial evolution patterns compared with glacial lakes and glaciers (no correlation, r\(\to\)0), their spatial patterns were likely to be consistent with decadal-scale fluctuations in global precipitation (r = 0.15). Categorically, these non-glacier-fed lakes can be classified into two types: prevalent precipitation-fed lakes and permafrost-impacted thermokarst lakes. The dominance of precipitation-driven changes in non-glacier-fed lakes (Fig. 3d), as well as their distinctiveness from the observed spatial evolution patterns of glacial lakes, and a usual absence of some of ice-related hazard components (e.g., ice-cored moraine dams and glacier ice dams), underscores the necessity of excluding these extraneous lakes in glacial lake analysis. In 2020, a total of 79,614 non-glacier-fed lakes were identified worldwide in this study, covering an area of 5976.3\(\pm\)364 km2. Non-glacier-fed lakes formed substantial clusters in each glacierized region (Supplementary Fig. 2). Failing to exclude the impact of these lakes is likely to result in either overestimation or underestimation of the true distribution and change signals in the number, area, and volume of glacial lakes (Supplementary Table 4).

Discussion

We identified three major concerns: (i) global underestimation of the abundance of glacial lakes and their impacts on intercepting sea-level rise; (ii) heterogeneous changes in ice-dammed lakes in various glacierized regions that do not align well with the geographic distribution of their GLOFs; and (iii) the glacier dominant roles of melting and thinning in glacial lake changes, while precipitation likely had a greater impact on non-glacier-fed lakes. In this study, we employed uniform delineation criteria and high-resolution satellite images to map the global glacial lakes and worked on standardizing glacial lake analysis worldwide.

A glacial lake is defined as a water body affected by glaciers or located within a glacier zone53. However, there is currently no consensus on how to determine whether a lake has been affected by glaciation. Different inventories of glacial lakes, even within the same region, can differ in terms of image sources, digitization methods, minimum thresholds, spatiotemporal cover, and glacial lake definition (Supplementary Table 5). In the Third Pole, scholars seem to prefer setting a 10 km buffer around contemporary glaciers to extract glacial lakes5,12,54,55,56,57. In the Southern and Tropical Andes, the area of the Little Ice Age Maxima is frequently used in glacial lake censuses, estimated at 2–3 km around glaciers21,58,59. In Alaska and the Greenland Periphery, ice-marginal lakes or lakes within a 1 km buffer of contemporary glaciers are usually defined as glacial lakes36,43,60,61. These differences in delineation criteria are permissible. The maximum extent of the Little Ice Age varies in various glacierized regions, resulting in different buffer distances. Similarly, disparities in the spatial resolution of remote sensing imagery can lead to different minimum extraction thresholds. Conducting a meaningful comparative glacial lake analysis requires developing uniform delineation criteria. For example, How et al.61 identified 3345 glacial lakes (>0.05 km2) within a 1 km buffer of glaciers in the Greenland Periphery in 2017, whereas our inventory contains a comparable number of glacial lakes (3152) for the same buffer extent and minimum threshold in 2020. This slight difference is partially caused by regional diminution and drainage of ice-dammed lakes.

Importantly, the results of this study highlight the necessity of excluding non-glacier-fed lakes that confound the sample of glacial lakes. That is, only water bodies hydraulically connected to glaciers are considered glacial lakes. Few previous studies have considered this point. In terms of their formation and change characteristics, non-glacier-fed lakes differ markedly from glacial lakes41,62, and their existence can disrupt the changing characteristics of glacial lakes in the context of climate warming and glacier thinning. Furthermore, we merely provide a simplified classification of glacial lakes considering the objective of our study. The glacier-fed lakes exhibited rapid and extensive increases in number, area, and volume worldwide, yet ice-dammed lakes demonstrated a surge in number and heterogeneous changes in area and volume. The current systematic classification methods for glacial lakes can be approached in two ways: one based on the dam material (ice-, moraine-, bedrock-, and landslide-dammed lakes), and the other based on the topological position of the glacial lakes relative to their parent glaciers (supraglacial, proglacial, periglacial, and extraglacial lake)59,63. Each type of glacial lake is likely to exhibit distinct responses to climate warming and glacial retreat. For example, in the Third Pole, proglacial lakes at glacier terminus are presently the primary expanding glacial lakes, with accelerated expansion observed in recent years64. Glacier calving and frontal melting of glacier ice that exceeds the rate of ice flow toward the lake are important for these ice-connected proglacial lakes65,66. A glacier-fed lake may demonstrate various evolutionary features, starting as a supraglacial pond, developing into a large proglacial lake, and eventually detaching from the glacier62. They can generally be characterized by an area-volume power-law relationship that fits an idealized hemispheric three-dimensional body67. Ice-dammed lakes essentially act as weirs formed in valleys. Therefore, this study divides glacial lakes into two basic categories when calculating their volume to ensure the accuracy of the volumetric analysis. Investigating the evolution of the characteristics of different types of glacial lakes on a global scale is a compelling avenue for future research.

Glaciers are projected to substantially lose mass under future temperature increases68, indicating that glacial lakes will also continue to form and expand globally. Numerous glacier-bed overdeepenings have been detected in the Third Pole69, the Greenland Periphery70, the European Alpes71,72, the Peruvian Andes73, and New Zealand74. A rough estimation suggests a potential development of glacial lakes of 1–3 times the current level in the future, which will continue to impact global sea-level rise. While our assessment of the capacity of glacial lakes to retain meltwater for delaying sea-level rise is approximately twice as high as previous estimates4, it remains relatively low compared to the contemporary glacier mass loss worldwide. Especially, when considering factors such as endorheic basins and physical phenomena like evaporation and seepage during the water’s journey to the ocean, as well as the thermodynamic volume of mixing correction of freshwater in seawater, the intercepting impacts on a global scale may be considered negligible. However, regionally, in areas like New Zealand, Scandinavia, and the Southern Andes, the intercepting impacts of glacial lakes are pronounced and exhibit strong regional variations. The role and impact of future glacial lakes as water resources, as well as potential flood hazards, will inevitably continue to be highlighted with ongoing climate warming and glacier retreat.

The threats to downstream human communities and infrastructure from GLOFs will increase. Compared with GLOFs originating from ice-dammed lakes, those originating from moraine-dammed lakes possess a higher destructive potential and are more widely distributed in space17. Glacial lakes progressively expand closer towards steep headwalls, increasing the outburst susceptibility, while the continuous upstream shift of people and infrastructure leads to a surge in exposure as socio-economic development progresses75. Hazard and risk assessments, whether conducted on a basin or regional basis, indicate that GLOF threats under future scenarios will be multiples of current levels32,76. The most recent devastating GLOF occurred in Indian Sikkim on October 4, 2023. The South Lhonak moraine-dammed lake (88°11′E, 27°54′N) drained, leaving more than 200 people missing or dead and destroying downstream roads, bridges, and reservoirs. Analyzing historical glacial lake changes and GLOF characteristics, assessing current and future glacial lake hazards and risks, and assisting decision-makers and stakeholders in preventing and mitigating GLOF risks have been the focus of various research teams77,78. These extraordinary undertakings rely on a well-developed global inventory of glacial lakes and standardized regional mapping of the future, warranting a more effective response to the escalating challenges posed by glacial lake dynamics in the context of climate warming.

Methods

Glacial lake mapping

In the past decade, many glacial lake inventories have been established worldwide using satellite or aerial imagery. Extraction methods, such as object-based image analysis, water indices-based classification, and machine learning algorithms, enable cost-effective and rapid segmentation of water bodies from optical remote sensing imagery. However, these automatic or semi-automatic extraction methods require extensive manual post-processing to address the missed glacial lakes and misplaced non-water bodies, as well as correcting impacts from cloud cover, hill shadows, and snow. Currently, the most widely used method of delineating glacial lakes is visual interpretation, which is adaptable to different glacierized regions where the terrain is complex and the spectral characteristics of glacial lakes are highly variable35. This method remains the most accurate and cross-comparable mapping approach in alpine regions58,59,79. In this study, glacial lakes were defined as water bodies primarily fed by contemporary glacier meltwater within a 10 km glacier buffer zone. We manually delineated the glacial lakes globally in two time windows, 1990 and 2020, using a total of 1255 Sentinel-2A/B (10 m) and 765 Landsat (30 m) images. These images were selected with acquisition times in the late period of regional glacial ablation and with a cloud coverage of less than 10%. The minimum extraction threshold was set at 0.002 km2. Northern Arctic Canada, the South American Patagonia, and the Antarctic and subantarctic are covered insufficiently by Landsat4/5 thematic mapper (TM) products around 1990, and thus, Landsat7 enhanced thematic mapper (ETM) images acquired in 2000 were utilized to fill the data gaps. Water bodies within the 10 km-buffer zone hydraulically connected to glaciers were classified as glacier-fed and ice-dammed lakes, while others were labeled as non-glacier-fed lakes. Ice-dammed lakes utilize glaciers as dams to accumulate water in upstream valleys, essentially functioning as a type of weir. Visually, the glacial lakes generally appeared gray or bluish on real-color composited images due to the glacier impact, while the non-glacier-fed lakes primarily affected by precipitation or permafrost tended to appear dark black. We critically classified the various lakes into categories based on expert judgment and geographical recharging by glaciers. Non-glacier-fed lakes were excluded from the glacial lake analyses throughout.

Leveraging statistical data from global glacial lake bathymetry67, we established empirical relationships between the lake volume and area, focusing on glacier-fed lakes and ice-dammed lakes (Supplementary Fig. 3). The equations exhibited a power–law relationship for glacier-fed lake areas smaller than 5 km2. As the area gradually increased, the relationship with the volume exhibited a linear growth80. Due to the limited number of data points available for fitting the scaling relationship between ice-dammed lakes and glacier-fed lakes (>5 km²), we were unable to utilize the Bootstrap resampling method to calculate volumetric uncertainty accurately. Therefore, the 95% uncertainty of the volume estimates was uniformly utilized instead of their uncertainty calculation. It is important to note that this approach may result in an underestimation of the volumetric uncertainty.

$$V=43.32{A}^{1.55}({R}^{2}=0.93,\,P < 0.01,{{{{{\rm{glacier}}}}}}-{{{{{\rm{fed\; lakes}}}}}},\, < 5{{{{{{\rm{km}}}}}}}^{2}),$$
(1)
$$V=16.9{A}^{1.83}({R}^{2}=0.97,{P} < 0.01,{{{{{\rm{ice}}}}}}-{{{{{\rm{dammed\; lakes}}}}}},\, < 5{{{{{{\rm{km}}}}}}}^{2}),$$
(2)
$$V=53.61A+362.02({R}^{2}=0.96,P < 0.01,{{{{{\rm{glacial}}}}}}-{{{{{\rm{fed}}}}}}\, {{{{{\rm{lakes}}}}}},\,\ge 5{{{{{{\rm{km}}}}}}}^{2}),$$
(3)

where V and A denote the volume and area of the glacial lake in 106 m3 and km2, respectively.

Quality control

Two primary sources of uncertainty in the glacial lake mapping were omissions and delineation. Initially, we digitized the glacial lakes in 2020 using high-resolution Sentinel-2A/B images. Then, we mapped the predetermined locations of the glacial lakes in 1990 using Landsat images, minimizing omissions resulting from the varying spatial resolutions of the satellite imagery. In addition, to reduce the workload and ensure mapping accuracy, we utilized parts of previous regional inventories as the initial locations of glacial lakes. A small number of lake boundaries from the same period, including those reported by Wang et al.5 in the Third Pole, by Andreassen et al.81 in Norway, by How et al.61 in the Greenland Periphery, and by Wieczorek et al.82 in Svalbard, were incorporated, and each additional glacial lake boundary was labeled and further corrected. Prior comparative analysis in the Greenland Periphery showed that the difference in the number of glacial lakes identified using automated and manual extraction methods was within \(\pm\)8%. In this study, each glacial lake underwent at least three reviews, reducing this uncertainty to approximately less than \(\pm\)3%. The identification, delineation, and classification of the glacial lakes were meticulously carried out by a single expert, ensuring rigorous quality control and continuous inspection55,83. Ultimately, we manually created 336,731 km of glacial lake boundaries globally, equivalent to 8.4 times the Earth’s circumference. The uncertainty in the glacial lake delineation was estimated by multiplying 0.5 pixels by the lake perimeter.