Results of the Ice Velocity Round Robin Algorithm Intercomparison Within the Esa Ice_Sheets_CCI Project
Merryman Boncori, John Peter1; Langer Andersen, Morten2; Kamstra, Martijn3; Dall, Jørgen4; Kusk, Anders4; Bech Andersen, Signe2; Bechor, Noa5; Bignami, Christian1; Gourmelen, Noel6; Joughin, Ian7; Jung, Hyung-Sup8; Luckman, Adrian9; Bevan, Suzanne9; Mouginot, Jeremie10; Neelmejer, Julia11; Scharrer, Kilian12; Scheuchl, Bernd10; Strozzi, Tazio13
1Istituto Nazionale di Geofisica e Vulcanologia (INGV), ITALY; 2Geological Survey of Denmark and Greenland (GEUS), DENMARK; 3Science [&] Technology AS, NORWAY; 4Technical University of Denmark, DENMARK; 5Massachusetts Institute of Technology, UNITED STATES; 6University of Strasbourg, FRANCE; 7University of Washington, UNITED STATES; 8University of Seoul, REPUBLIC OF KOREA; 9University of Swansea, UNITED KINGDOM; 10University of California, Irvine, UNITED STATES; 11GFZ Potsdam, GERMANY; 12ENVironmental Earth Observation (ENVEO), AUSTRIA; 13Gamma Remote Sensing and Consulting AG, SWITZERLAND

Within the ESA Ice_Sheets_cci project, a SAR-based processing chain is currently under development for the measurement of ice velocity time-series on continental scales, with particular focus on Greenland in the first project phase. As an input for algorithm design, the project team was asked to carry out a round robin algorithm intercomparison, by inviting members of the scientific community to apply their processing chains to a set of common datasets and subsequently identifying the algorithmic aspects with greatest influence on the quality of the generated products. Three families of SAR-based techniques were considered, namely Differential SAR Interferometry (DInSAR), Multiple Aperture Interferometry (MAI) and offset-tracking. Four datasets were chosen, covering Greenlandic glaciers and areas of the ice sheet margin. Participants were invited to carry out one or more tasks, chosen among the following four: (1) generate a slant-range velocity map from two 1-day ERS tandem pairs; (2) generate an azimuth velocity map from a 1-day ERS tandem pair; (3) generate slant-range and azimuth velocity maps from a 35-day ENVISAT ASAR pair; (4) generate slant-range and azimuth velocity maps from a 46-day ALOS PALSAR pair. For each task participants were also asked, if applicable, to provide an error characterization of their measurements and to provide detailed information on their processing algorithms. Responses were received from a total of 12 institutions worldwide, 9 of which external to the project team. Several participants carried out more than one task. For each of the latter, the delivered products were validated against available GPS measurements, and inter-compared to highlight the main differences. Throughout this work, as well during the validation phase and in all project documentation, the delivered results are treated anonymously. In this work we present the result of our analysis and, where sufficient information is available, we discuss the algorithmic steps, which have the greatest influence on the quality of the output velocity maps, particularly in terms of accuracy, coverage and error characterization.