An automated algorithm to retrieve the location and depth of supraglacial lakes from ICESat-2 ATL03 data

Abstract

Supraglacial lakes are widespread on the surface of the Antarctic Ice Sheet and Greenland Ice Sheet due to global warming. Estimating the location and depth of supraglacial lakes is critical for evaluating the impact of surface meltwater on ice dynamics. The accuracy of traditional image-based methods is limited, especially when supraglacial lakes are deep. Recently, Ice, Clouds, and land Elevation Satellite-2 data have been used to develop algorithms to measure the depth of supraglacial lakes. The ICESat-2-based methods are currently either semiautomated or sensitive to noise, which makes them less efficient and less accurate in detecting supraglacial lakes. In this study, we develop an automated algorithm for the retrieval of the location and depth of supraglacial lakes using ICESat-2 ATL03 data. We investigated 10 supraglacial lakes as case studies over Antarctica and Greenland (4 lakes on the Amery Ice Shelf, Antarctica and 6 lakes in southwestern Greenland). The results show that our algorithm can detect supraglacial lakes at depths up to 8.25 m. The RMSE for the lake depth retrievals in Antarctica is approximately 0.30 m and that of the lakes in Greenland is approximately 0.32 m. Our algorithm reveals high robustness and accuracy, especially in shallow areas. Additionally, the fully automated algorithm does not require prior knowledge and guarantees efficiency when there is an extensive study area. This result implies that our algorithm has the potential to address large amounts of data and may even be useful to estimate meltwater volumes across the surface of the entire ice sheet.