Climate change accelerates the extensive retreat of glaciers, leading to the widespread development of glacial lakes. A holistic picture of the spatial distribution of glacial lakes worldwide is a critical base for tracking the outburst hazards. By employing a semi-automated mapping approach and rigorous quality control, this study inventories 117,352 glacial lakes (≥0.01 km2) worldwide (the ice cap/sheet of Antarctic and Greenland excluded), with a net area of 24,755.84 ± 2,971.33 km2. The evaluation result shows this global inventory of glacial lakes (GIGLak) has an overall accuracy of 89.37% and 91.42% in number and area, respectively. These glacial lakes are widely distributed in different altitudes, ranging from the Earth’s third pole to the coasts. Most glacial lakes are distributed in the Greenland periphery, High-Mountain Asia, Alaska, Canada, and the Cordilleras. The number of glacial lakes between 0.01–0.1 km2 accounts for 77.24% of the total count but only 11.82% in area. The classification of glacial lakes as four types indicates that the ice-uncontacted proglacial lakes dominate the number (67.07%) and area (53.04%) worldwide.
Two open-source datasets, GSW (1984-2020, taking the maximum water area range) and GLAD (1999-2020, aggregating annual layers to extract the maximum range), were used to extract glacial lake water areas. GLAD supplemented the narrow and elongated glacial lakes not identified by GSW; At the same time, MERIT DEM was used to distinguish terrain shadows that were mistakenly identified as glacial lakes (generating>10 ° slope maps to assist in non water body judgment), and RGI v6.0 glacier cataloging was used to determine the location and range of glaciers.
(1) Using GSW (1984-2020, taking the maximum water area range) and GLAD (1999-2020, aggregating layers to take the maximum range) to extract glacial lake water areas, GLAD supplements the narrow glacial lakes not identified by GSW;
(2) Using MERIT DEM to distinguish terrain shadows (generating>10 ° slope maps to assist in identifying non water bodies), and using RGI v6.0 to determine glacier ranges.
(3) Identify glacial lakes using a 3-kilometer buffer zone at the end of the glacier, and overlay high-resolution images to extract boundaries.
(4) Manually verifying surface water vectors and ESRI images, removing fake glacial lakes, and filling in omissions, two people spent 2600 hours processing approximately 6 million polygons from 2020 to 2022. According to the meltwater supply, GFL and NGFL are classified, and GFL is further divided into three categories based on glacier topology and manually corrected. Assuming that the artificial error follows a Gaussian distribution, only the influence of mixed pixels is considered for accuracy evaluation.
The data quality is good.
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
10.6084/m9.figshare.26310967
