Automated iceberg detection around the Greenland Ice Sheet: bimonthly results from 2016 to 2018

Time-series of iceberg distribution around Greenland are important for parameterizing calving processes, quantifying freshwater flux, and incorporating into coupled ice-ocean models. However, the size and frequency of icebergs around Greenland are largely unknown. Here we present results from our automated iceberg tracking algorithm, which classifies icebergs around Greenland every 2 weeks from 2016 to 2018 . We use predominantly high-resolution IW Sentinel-1 SAR imagery and a combination of detection algorithms to characterize icebergs in open water, sea-ice and pro glacial mélange. Our algorithm first uses Random Forest (RF) supervised classification to mask out the various types of background (open water, sea ice, or proglacial mélange) , isolating potential iceberg pixels. A modified Constant False Alarm algorithm further identifies iceberg boundaries . The result is a bi-monthly dataset of iceberg distribution around Greenland fo r 3 complete years which allows us to assess regional and seasonal variations . We present the details of our technique, and show how regional trends in calving style impact iceberg distribution.