Parts | Days | Selection | Search | Updates | Downloads | Help
UP: Fachverband Umweltphysik
UP 3: Kryosphäre
UP 3.8: Talk
Tuesday, March 14, 2017, 14:30–14:45, GW2 B3009
Daily lead map of the European Arctic from Sentinel-1 SAR scenes — •Dmitrii Murashkin1, Gunnar Spreen1, and Marcus Huntemann2,1 — 1Institute of Environmental Physics, University of Bremen, Germany — 2Alfred Wegener Institute for Polar and Marine Research, Bremerhaven, Germany
Leads are linear-like areas with open water within sea ice cover. They are of interest for environmental science, weather forecasting and ship navigation in polar regions. Here, an algorithm that provides an automatic binary classification of leads is proposed. Synthetic Aperture Radar (SAR) satellites provide all-weather and season observations and the necessary high resolution to identify leads. Previously often a single co-polarized band was used for ice-water classification which can result in misclassification of leads under windy conditions. The presented algorithm benefits from the use of Sentinel-1 SAR dual channel products which include measurements in co- and cross-polarized modes. Exploiting information from both, the algorithm is capable to identify leads which can not be identified using single co-polarized measurements. The algorithm is based on Haralick texture features and a supervised classification algorithm. Its stability and high parallelization makes Random Forest Classifier a perfect tool for SAR image feature recognition. It allows per-pixel processing of images with speckle noise. Leads are identified in single SAR scenes, which are then compiled to maps covering a larger region from a set of individual products. Maps of lead distributions for the European Arctic with a resolution of 200 meters are presented.