Mapping and spatial distribution characteristics of glacial lakes in Xinjiang based on Sentinel-2 imagery
Glaciers in Xinjiang,northwestern China,have experienced drastic retreat phenomena due to climate warming in recent decades.Glacier retreats have led to an increase in glacial meltwater runoff and accelerated the formation and expansion of glacial lakes.Especially in the core area of the Silk Road Economic Belt and cli-mate change-sensitive areas,ice-marginal compound disasters,such as glacial surges and glacial lake outburst floods(GLOFs),have entered a period of high incidence and acceleration.The resulting economic losses and potential risks will also become increasingly severe and prominent.Accurate mapping and monitoring of these glacial lakes is therefore critical to understanding the response of GLOF hazards to climate change.In this study,we conducted a detailed mapping of glacial lakes in Xinjiang from April to October 2022 using the DUNet se-mantic segmentation model and Sentinel-2 imagery,extracted the maximum boundaries of glacial lakes in Xinji-ang from April to October 2022,and manually inspected to exclude non-glacial lakes and supplement glacial lakes missed due to shading.We used three-level basins to divide distribution area of glacial lakes in Xinjiang in-to 11 subregions.We assigned detailed attribute information to each glacial lake,such as center coordinates,pe-rimeter,area,elevation of the center point of the glacial lake,subregion name,subregion code,absolute error,relative error,and so on.In addition,we screened out glacial lakes with an area greater than 0.06 hm².We ana-lyzed spatial distribution of these lakes and trend of change over the past decade in conjunction with the existing inventory of glacial lakes.Results showed that the glacier mapping based on Sentinel-2 got more lake numbers and achieved lower relative errors than the existing Landsat-based inventories,with average relative errors of large(>10 hm²),medium(>1～10 hm²),and small(≤1 hm²)lakes are 2.29%,10.02%,and 27.71%,respec-tively.The relative error of this glacial lake dataset with an area larger than 0.81 hm² is 18.36%,which meets the mapping accuracy of monitoring glacial lake changes in Xinjiang for more than 20 years.About 6 854 glacial lakes with a total area of 200.36 km² were mapped throughout Xinjiang.Specifically,small glacial lakes ac-counted for 70.32%of the total number of glacial lakes but only 7.51%of the total glacial lake area;medium glacial lakes and large glacial lakes accounted for 29.68%of the total number of glacial lakes,but the area ac-counted for 92.49%of the total area.The patterns of lake distribution are heterogeneous in different mountain-ous areas,and the glacial lakes are numerous in the Altai Mountains(1 474),western Tianshan(Ili River ba-sin)(1 170),and southwest Tianshan(Tarim inflow area)(915).The number of glacial lakes peaks in the alti-tude range of 3 300～3 600 meters.In the last 30 years,glacial lakes with areas of less than 10 hm² increased sig-nificantly in number and area,especially in the Altai and the Western Tianshan Mountains.Therefore,the re-gional differences in the distribution of glacial lakes in Xinjiang are significant,with a large number of small gla-cial lakes,which tend to be more sensitive to climate change and change more drastically and can effectively re-flect the changes in regional climate and glaciers.This study provides an important dataset for monitoring and di-saster assessment of glacial lakes in Xinjiang,which is essential for a deeper understanding of the response of glacial lakes to climate change and the development of effective coping strategies.Future studies will improve the construction of a multi-period glacial lake catalog database,further analyze the changing characteristics of glacial lakes in Xinjiang,and explore GLOF disaster and its coupling relationship with climate,glaciers,and other elements.
glacial lakeXinjiangremote sensingspatial distribution characteristicsdeep learningSentinel-2 imagery
NETL
NSTL
万方数据
