Antarctic ice shelf front dynamics between 2015 and 2023

Abstract

The Antarctic ice sheet is fringed by ice shelves regulating the ice flow into the ocean. The fronts of these ice shelves are constantly moving and are sensitive indicators of glaciological and environmental change. Previously, Antarctic ice shelf front change was not observed regularly due to limited availability of satellite imagery and time-consuming manual front delineations. The era of freely available SAR satellite data and recent developments in image processing with artificial intelligence created new opportunities for monitoring ice shelf front dynamics on a regular basis. Here, we present the IceLines dataset providing continuous time series of calving front dynamics for 36 major Antarctic ice shelves since 2015. The dataset consists of over 19,000 front positions extracted from Sentinel-1 satellite data by using a convolutional neural network called HED-Unet. The automatically extracted front positions can deviate from manual delineated fronts due to fast ice, mélange and icebergs close to the front by 209±12 m (5.2 pixel) on dual polarized imagery and 432±21 m (8.8 pixel) for single-polarized imagery whereas the frontal movement can be determined with higher accuracies of 63±68 m (1.6 pixel) for dual and 107±126 m (2.7 pixel) for single polarized imagery. To minimize errors and enhance quick usability, automatic separation of unreliable front positions (still accessible) is applied for an easy analysis of the dataset. This contribution features the analysis of the IceLines dataset providing new insights into Antarctic calving front dynamics by investigating intra-annual calving front dynamics, changing advance rates of ice shelf fronts, recent calving events and overall calving front change.