| Summary: | This dataset contains synthetic aperture radar (SAR) images of supraglacial lakes on the Antarctic ice shelves, acquired by the Sentinel-1 satellite. Supraglacial lakes are surface meltwater features that can affect the stability and dynamics of ice shelves by hydrofracturing, basal sliding, or surface thinning. Monitoring the spatio-temporal distribution and evolution of these lakes is important for understanding the impact of surface meltwater on ice shelf collapse and sea level rise. The dataset covers four key regions: the Amery, Roi Baudouin, Nivlisen, and Riiser-Larsen ice shelves, which are characterized by extensive surface meltwater features. The images are single-polarized (VV or VH) and have a spatial resolution of 10 m. The images are processed using a modified U-Net algorithm for semantic segmentation of supraglacial lakes, based on deep learning and multiscale feature extraction. The algorithm is trained and validated on Landsat 8 and Sentinel-2 optical imagery, which provide high-resolution and multispectral information on supraglacial lake detection. The dataset provides binary masks of supraglacial lake extent for each Sentinel-1 image, as well as metadata on acquisition date, polarization, orbit number, and region name. The dataset also includes a decision-level fused product of Sentinel-1 and Sentinel-2 maximum lake extent for January 2020, which reveals a more complete supraglacial lake coverage than the individual single-sensor products. The dataset covers the austral summer seasons from 2015/2016 to 2019/2020 and contains over 1000 images. The dataset is intended to provide a first-ever continental record of supraglacial lake extent and volume in Antarctica using SAR imagery, which can complement optical datasets and overcome limitations of cloud cover and polar darkness. The dataset can be used for studying the intra-annual and interannual variability of supraglacial lake occurrence, depth, and drainage events, as well as their relationship with surface air temperature, ice shelf geometry, ...
|
|---|