Sea ice concentration estimation from Sentinel-1 Synthetic Aperture Radar images over the Fram Strait

Paper in proceedings, 2016

© 2016 IEEE.In this paper we present an algorithm for sea ice concentration estimation in the Arctic from C-band dual polarization Sentinel-1A SAR images. The algorithm is based on spatial autocorrelation and utilizes an artificial neural network to map the image information to sea ice concentration. The cross-polarization channel facilitates the improvement of concentration estimates of images with high backscatter over open water in the normally used co-polarization channel. Ice charts from the Norwegian meteorological institute are used for the training of the network and as a reference. A mean absolute error of 14.55 (ice concentration is given in the range from 0 to 100) of a test data set consisting of 20 images underlines the capabilities of the proposed algorithm.