In situ validation of segmented SAR satellite scenes of young Arctic thin landfast sea ice

The use of satellite remote sensing techniques for the observation and monitoring of the polar regions has increased in recent years due to the ability to cover larger areas than can be covered by ground measurements, However, in situ data remain mandatory for the validation of such data. In April 2016 an Arctic fieldwork campaign was conducted at Kongsfjorden, Svalbard. Ground measurements from this campaign are used together with satellite data acquisitions to improve identification of young sea ice types from satellite data. This work was carried out in combination with Norwegian Polar Institute's long-term monitoring of Svalbard fast ice, and with partner institutes in the Center for Integrated Remote Sensing and Forecasting for Arctic operations (CIRFA). Thin ice types are generally more difficult to investigate than thicker ice, because ice of only a few centimetres thickness does not allow scientists to stand and work on it. Identifying it on radar scenes will make it easier to study and monitor. Four high resolution 25 km x 25 km Radarsat-2 quad-pol scenes were obtained, coincident in space and time with the in situ measurements. The field teams used a variety of methods, including ice thickness transects, ice salinity measurements, ground-based radar imaging from the coast and UAV-based photography, to identify the different thin ice types, their location and evolution in time. Sampling of the thinnest ice types was managed from a small boat. In addition, iceberg positions were recorded with GPS and photographed to enable us to quantify their contribution to the radar response. Thin ice from 0.02 to 0.18 m thickness was sampled on in a total nine ice stations. The ice had no or only a thin snow layer. The GPS positions and tracks and ice characteristics are then compared to the Radarsat-2 scenes, and the radar responses of the different thin ice types in the quad-pol scenes are identified. The first segmentation results of the scenes present a good consistency with the ground data and the recorded thin ice edge. We are able to distinguish between several of the thin ice types.