Near Real Time Delivery of Sea Ice Information Retrieved from Spaceborne Synthetic Aperture Radar

Abstract

Sea ice is subject to constant change. Within just a few hours the wind can turn, shoving sea ice together over kilometres and causing pressure ridges to form – obstacles that are difficult or impossible even for icebreakers to overcome. Synthetic Aperture Radar (SAR) satellites are able to observe small- and large-scale structures in sea ice – in any weather, through clouds and darkness. In this contribution, we show new algorithms that aim at retrieving sea ice information from SAR data in particular information on high resolution sea ice drift and sea ice classes. The sea ice drift retrieval is based on the well-known phase correlation technique executed on subsequent, co-located SAR acquisitions. The sea ice classification utilizes a convolutional neural network and - by exploiting the different polarizations - differentiates multiyear ice, first-year ice, new ice, open leads, and rough ice. In a new study, both algorithms are interlinked to perform multi-temporal sea ice classification using a stack of drift-compensated SAR acquisitions (instead of just a single acquisition). The multi-temporal approach is intended to reduce misclassifications and improve reliability of the output. The developed algorithms are foreseen to be integrated into the operational data processing chain at DLR ground station network sites in order to provide sea ice information to maritime users in near real-time (NRT).