Seasonal glacier ice velocities are important for understanding controlling mechanisms of ice flow. For many Greenlandic glaciers, however, these measurements are limited by low temporal resolution. We present seasonal ice velocity changes, melt season onset and extent, and ice front positions for 45 Greenlandic glaciers using 2015–2017 Sentinel‐1 synthetic aperture radar data. Seasonal velocity fluctuations of roughly half of the glaciers appear to be primarily controlled by surface melt‐induced changes in the subglacial hydrology. This includes (1) glaciers that speed up with the onset of surface melt and (2) glaciers with comparable late winter and early melt season velocities that show significant slowdown during most of the melt season and speedup during winter. In contrast, less than a quarter of the study glaciers show strong correspondence between seasonal ice speed and terminus changes. Our results pinpoint seasonal variations across Greenland, highlighting the variable influence of meltwater on year‐round ice velocities.

Abstract-Data driven classification algorithms have proven to do well for Automatic Target Recognition (ATR) in Synthetic Aperture Radar (SAR) data. Collecting datasets suitable for these algorithms is a challenge in itself as it is difficult and expensive. Due to the lack of labelled datasets with real SAR images of sufficient size, simulated data plays a big role in SAR ATR development, but the transferability of knowledge learned on simulated data to real data remains to be studied further.In this paper we show the first study of Transfer Learning between a simulated dataset and a set of real SAR images. The simulated dataset is obtained by adding a simulated object radar reflectivity to a terrain model of individual point scatters, prior to focusing. Our results show that a Convolutional Neural Network (Convnet) pre-trained on simulated data has a great advantage over a Convnet trained only on real data, especially when real data is sparse. The advantages of pre-training the models on simulated data show both in terms of faster convergence during the training phase and on the end accuracy when benchmarked on the MSTAR dataset. These results encourage SAR ATR development to continue the improvement of simulated datasets of greater size and complex scenarios in order to build robust algorithms for real life SAR ATR applications.

Abstract. We present the Programme for Monitoring of the Greenland Ice Sheet (PROMICE) ice velocity product (https://doi.org/10.22008/promice/data/sentinel1icevelocity/greenlandicesheet) (Solgaard and Kusk, 2021)) which is a September 2016 through present time series of Greenland Ice Sheet ice-velocity mosaics. The product is based on Sentinel-1 synthetic aperture radar data and has a 500 m spatial resolution. A new mosaic is available every 12 days and span two consecutive Sentinel-1 cycles (24 days). The product is made available within ~10 days of the last acquisition and includes all possible 6 and 12 day pairs within the two Sentinel-1A cycles. We describe our operational processing chain in high detail from data selection, mosaicking and error estimation to final outlier removal. The product is validated against in-situ GPS measurements. We find that the standard deviation of the difference between satellite and GPS derived velocities is 20 m/yr and 27 m/yr for the vx and vy components, respectively. This is within the expected bounds, however, we expect that the GPS measurements carry a considerable part of this uncertainty. We investigate variations in coverage from both a temporal and spatial perspective. Best spatial coverage is achieved in winter due to excellent data coverage and high coherence, while summer mosaics have the lowest coverage due to widespread melt. The southeast Greenland Ice Sheet margin, along with other areas of high accumulation and melt, often have gaps in the ice velocity mosaics. The spatial comprehensiveness and temporal consistency make the product ideal for monitoring and studying ice-sheet wide ice discharge and dynamics of glaciers.

The seasonal response to surface melting of the Northeast Greenland Ice Stream outlets, Zachariae and 79N, is investigated using new highly temporally resolved surface velocity maps for 2016 combined with numerical modeling. The seasonal speedup at 79N of 0.15 km/yr is suggested to be driven by a decrease in effective basal pressure induced by surface melting, whereas for Zachariae its 0.11 km/yr seasonal speedup correlates equally well with the breakup of its large ice mélange. We investigate the influence 76 km long floating tongue at 79N, finding it provides little resistance and that most of it could be lost without impacting the dynamics of the area. Furthermore, we show that reducing the slipperiness along the tongue‐wall interfaces produces a velocity change spatially inconsistent with the observed seasonal speedup. Finally, we find that subglacial sticky spots such as bedrock bumps play a negligible role in the large‐scale response to a seasonally enhanced basal slipperiness of the region.

Abstract. The Northeast Greenland Ice Stream (NEGIS) extends around 600 km upstream from the coast to its onset near the ice divide in interior Greenland. Several maps of surface velocity and topography of interior Greenland exist, but their accuracy is not well constrained by in situ observations. Here we present the results from a GPS mapping of surface velocity in an area located approximately 150 km from the ice divide near the East Greenland Ice-core Project (EastGRIP) deep-drilling site. A GPS strain net consisting of 63 poles was established and observed over the years 2015–2019. The strain net covers an area of 35 km by 40 km, including both shear margins. The ice flows with a uniform surface speed of approximately 55 m a−1 within a central flow band with longitudinal and transverse strain rates on the order of 10−4 a−1 and increasing by an order of magnitude in the shear margins. We compare the GPS results to the Arctic Digital Elevation Model and a list of satellite-derived surface velocity products in order to evaluate these products. For each velocity product, we determine the bias in and precision of the velocity compared to the GPS observations, as well as the smoothing of the velocity products needed to obtain optimal precision. The best products have a bias and a precision of ∼0.5 m a−1. We combine the GPS results with satellite-derived products and show that organized patterns in flow and topography emerge in NEGIS when the surface velocity exceeds approximately 55 m a−1 and are related to bedrock topography.