| Summary: | Ice shelf calving is an important mass loss component of the Antarctic Ice Sheet (AIS). The discharge of the AIS is sensitive to changes in ice shelf extent and thickness due to ice shelf buttressing effects. Up to now, it has been challenging to continuously monitor calving front change as the manual delineation of ice shelf fronts is very time-consuming and suitable satellite data was missing. Since the launch of Sentinel-1, plenty of imagery over the Antarctic coastline is freely available. To overcome the tedious manual work, we present a novel ice shelf monitoring service called “IceLines” which enables the near-real-time tracking of front positions based on Sentinel-1 SAR data. IceLines is based on a Deep Learning (DL) algorithm to automatically extract the border between ocean and ice sheet from high volumes of Sentinel-1 data. Further processing creates a line shapefile for each front position based on suitable Sentinel-1 acquisitions to create a time series of calving front movement. A truncation algorithm secures that erroneously extracted fronts due to surface melt, wind roughening sea or dry snow are excluded from the time series. For validation, manually delineated fronts for randomly selected dates (outside the training data time period) are created. The accuracy is measured as the mean and median distance between DL-extracted and manually delineated fronts. Finally, the continuously extracted calving fronts are made available to the public via the DLR GeoService (geoservice.dlr.de). The webservice will provide researchers with up-to-date front positions on major Antarctic ice shelf extents. This will allow to easily include ice shelf front fluctuations in future analysis in order to provide more accurate estimates on the mass balance of the AIS.
|
|---|