Recent significant subseasonal fluctuations of supraglacial lakes on Greenland monitored by passive optical satellites

Abstract

Temporarily impounded liquid water on the surface of the Greenland Ice Sheet (GrIS), prominently represented as supraglacial lakes (SGLs), may enhance ice flow and modulate surface meltwater runoff, serving as a dynamic indicator of the cryohydrologic cycle. Despite their importance in understanding glacier mass balance and regional climate change, a detailed description of SGLs and their intra-annual fluctuations across the entire GrIS remains understudied. Here, we present a deep learning-based approach to automatically map SGLs from passive optical satellite imagery across the entire GrIS during the melt seasons of 2017–2022. Approximately 150,000 Sentinel-2 and Landsat 8/9 images were utilized, each representing a 5-day average composite at a 10 km × 10 km grid resolution, with the Landsat images used as possible supplements. SGL predictions by the proposed method demonstrate high performance, achieving an F1-score of up to 0.959 compared to the independent test dataset. This high accuracy enables a detailed analysis of the key role SGLs play in enhancing surface ablation by absorbing solar radiation and delivering meltwater. The SGL-driven ablation effect was most pronounced in the South-West basin of the GrIS, where the peak lake area in July accounted for 44.9 % of the total GrIS-wide lake area. In contrast, the lowest magnitude (4.2 %) was observed in the South-East basin, despite similarly strong ablation in this region. Among all the generated SGL occurrence grids, peak SGL areas in certain grids (∼14 % of the total) were observed in May or September, rather than exclusively during the typical high-ablation months of June to August, reflecting regional and elevation-dependent variations. Grids further from the ice sheet margin generally showed peak SGL areas later in the melt season, which is evident in the western part of the GrIS. Monthly SGL peak areas shift dramatically from 253.18 ± 123.94 km² to 5084.90 ± 1043.26 km², with the lowest in May 2018 and the highest in August 2021. An extraordinary area spike occurred in September 2022 and was particularly monitored in the South-West basin, where abnormally intense rainfall and runoff simulated by the Modèle Atmosphérique Régional (MAR) model were recorded. Our study highlights the significance of examining SGL area changes at short temporal intervals to understand the dynamics of cryospheric hydrology under future climate scenarios.