Validation and Comparison of Ice Velocity Fields from SAR Data - Results of the ESA Glaciers_cci Round Robin Experiment
Scharrer, Kilian1; Strozzi, Tazio2; Nagler, Thomas1; Wiesmann, Andreas2; Nuth, Christopher3; RR, contributors4
1ENVEO IT, AUSTRIA; 2GAMMA, SWITZERLAND; 3University Oslo, NORWAY; 4various institutions, UNITED KINGDOM

Ice velocity is an important parameter to evaluate the dynamic response of glaciers to climate change and to calculate the mass fluxes. Repeat pass SAR data enable the mapping of ice motion at high accuracy by means of differential processing techniques, including SAR interferometry and offset tracking. Offset tracking, coherent and incoherent, has been proven to be an attractive alternative to SAR interferometry in observing ice flow, as it provides two components of the velocity vector (in line of sight and along track) with a higher robustness. Feature tracking does not require the preservation of coherence. Therefore this technique has been widely applied for mapping the ice flow of large glaciers, ice caps and ice sheets using SAR systems. Although the implementation of offset tracking is straight forward the details of data processing may affect the quality of the retrieved ice flow products. In order to validate and inter-compare the ice velocity maps from various institutions, Round Robin Exercises were initiated in the frame of the ESA Glacier_CCI project. Round Robin experiments are intended as benchmark tests in order to support the selection of the algorithm, which is most suitable for automatic production of ice velocity fields. Three study sites with different characteristics and different SAR data sets were chosen for the Round Robin experiments. The first test site is Breiamerkurjokull, an outlet glacier of Vatnajokull ice cap in Iceland draining towards the Atlantic Ocean, which is covered by repeat pass TerraSAR-X scenes separated by 11 days. In-situ GPS measurements coincidently acquired with the TerraSAR-X acquisitions and kindly provided by the University of Iceland, enable independent validation of the derived products. The Karakorum test area represents alpine type of glaciers and is covered by repeat pass ENVISAT ASAR Image Mode data separated by one year. A pair of Landsat images acquired with a similar time interval is used for validation. The third test area is the Vestfonna ice cap in the high Arctic region of Svalbard, with ALOS-PALSAR L-band SAR data separated by 46 days. Several in-situ velocity measurements on Vestfonna are available to validate the ice motion products. Participants of the Round Robin Experiments were invited to select a dataset, apply their software for ice flow mapping, and return the results together with a feedback form describing the processing steps and selected parameters. Overall -seven groups provided results and three groups returned results for two tasks. Together with participants of the ESA Glacier_CCI project the effort of our experiment lead to a total number of 15 datasets (5 for Breiamerkurjokull, 6 for Karakorum, and 4 for Vestfonna) for the RR inter-comparisons. In this paper we summarize the results of the intercomparison of ice velocity products for the three data sets and on the absolute validation using GPS data.