1 Sea ice concentration estimation during melt from dual-pol SAR scenes using deep convolutional neural networks: A case study Lei Wang, Student Member, IEEE, K. Andrea Scott, Member, IEEE, Linlin Xu, Member, IEEE, David A. Clausi, Senior Member, IEEE Abstract High resolution ice concentration maps are of great interest for ship navigation and ice hazard forecasting. In this case study, a convolutional neural network (CNN) has been used to estimate ice concentration using synthetic aperture radar (SAR) scenes captured during the melt season. These dual-pol RADARSAT-2 satellite images are used as input and the ice concentration is the direct output from the CNN. With no feature extraction or segmentation post processing, the absolute mean errors of the generated ice concentration maps are less than 10% on average when compared to manual interpretation of the ice state by ice experts. The CNN is demonstrated to produce ice concentration maps with more detail than produced operationally. Reasonable ice concentration estimations are made in melt regions and in regions of low ice concentration. Index Terms ice concentration, synthetic aperture radar, convolutional neural network I. I NTRODUCTION With recent reductions in Arctic ice extent, there has been growing economic interest in shipping and natural resource extraction in the Arctic [1]. To support safe Arctic operations and navigation in ice infested waters, timely high resolution information of the ice coverage is crucial [2]. In this study, we focus on ice concentration, which is deļ¬ned as the percentage of ice coverage over a given spatial area.