Global time series and temporal mosaics of glacier surface velocities, derived from Sentinel-1 data

Abstract

Abstract. Consistent and continuous data on glacier surface velocity are important inputs to time series analyses, numerical ice dynamic modelling and glacier mass flux computations. Since 2014, repeat-pass Synthetic Aperture Radar (SAR) data is acquired by the Sentinel-1 satellite constellation as part of ESA’s (European Space Agency) Copernicus program. It enables global, near real time-like and fully automatic processing of glacier surface velocity fields at up to 6-day temporal resolution, independent of weather conditions, season and daylight. We present a new near global data set of glacier surface velocities that comprises continuously updated scene-pair velocity fields, as well as monthly and annually averaged velocity mosaics at 200 m spatial resolution. The velocity information is derived from archived and new Sentinel-1 SAR acquisitions by applying feature and speckle tracking. The data set covers 12 major glaciated regions outside the polar ice sheets and is generated in an HPC (High Performance Computing) environment at the University of Erlangen-Nuremberg. The velocity products are freely accessible via an interactive web portal that provides capabilities for download and simple online analyses: http://retreat.geographie.uni-erlangen.de. In this paper we give information on the data processing and how to access the data. For the example region of Svalbard, we demonstrate the potential of our products for velocity time series analyses at very high temporal resolution and assess the quality of our velocity products by comparing them to those generated from very high resolution TerraSAR-X SAR (Synthetic Aperture Radar) and Landsat-8 optical (ITS_LIVE, GoLIVE) data. We find that Landsat-8 and Sentinel-1 annual velocity mosaics are in an overall good agreement, but speckle tracking on Sentinel-1 6-day repeat acquisitions derives more reliable velocity measurements over featureless and slow moving areas than the optical data. Additionally, uncertainties of 12-day repeat Sentinel-1 mid-glacier scene-pair velocities are less than half (< 0.08 m d−1) of the uncertainties derived for 16-day repeat Landsat-8 data (0.17–0.18 m d−1).