Search

The paper presents the potential of combining satellite radar data and neural networks for quasi-automatic detection of glacier grounding lines. The conducted research covered five years and was carried out in the area of the Amery Ice Shelf. It has a very complex shoreline, so its grounding-line location is uncertain. Thus, it has always been the subject of much research. The main objective of our work was to find out if Synthetic Aperture Radar data combined with a deep learning implementation would enable rapid detection of ice shelf grounding lines over large areas. For this purpose, 290 radar images from the Sentinel-1 satellite covering 46 000 km² were used. Processed by the Differential Interferometry of Synthetic Aperture Radar four-pass method, the images formed a timeconsistent series between 2017 and 2021. As a result of performed calculations, a total length of 1280 km of grounding line was determined. They were validated by comparing with other independent data sources based on manual measurements. It has been demonstrated that the combination of satellite radar data and automated data processing allows for obtaining high-precision results continuously in a very short time. Such an approach allows monitoring of grounding line position in the long term with intervals of less than one week. It enables analysis of the dynamics changes with unprecedented frequency and the identification of patterns. [ABSTRACT FROM AUTHOR]