In this paper an algorithm for ice/water classification of C- and L-band dual polarization synthetic aperture radar data is presented. A comparison of the two different frequencies is made in order to investigate the potential to improve classification results with multi-frequency data. The algorithm is based on backscatter intensities in co- and cross-polarization and autocorrelation as a texture feature. The mapping between image features and ice/water classification is made with a neural network. Accurate ice/water maps for both frequencies are produced by the algorithm and the results of two frequencies generally agree very well. Differences are found in the marginal ice zone, where the time difference between acquisitions causes motion of the ice pack. C-band reliably reproduces the outline of the ice edge, while L-band has its strengths for thin ice/calm water areas within the icepack. The classification shows good agreement with ice/water maps derived from met.no ice-charts and radiometer data from AMSR-2. Variations are found in the marginal ice zone where the generalization of the ice charts and lower accuracy of ice concentration from radiometer data introduce deviations. Usage of high-resolution dual frequency data could be beneficial for improving ice cover information for navigation and modelling.

Automatic and visual sea ice classification of SAR imagery is impeded by the incidence angle dependence of backscatter intensities. Knowledge of the angular dependence of different ice types is therefore necessary to account for this effect. While consistent estimates exist for HH polarization for different ice types, they are lacking for HV polarization data, especially for multiyear sea ice. Here we investigate the incidence angle dependence of smooth and rough/deformed first-year and multiyear ice of different ages for wintertime dual-polarization Sentinel-1 C-band SAR imagery in the Beaufort Sea. Assuming a linear relationship, this dependence is determined using the difference in incidence angle and backscatter intensities from ascending and descending images of the same area. At cross-polarization rough/deformed first-year sea ice shows the strongest angular dependence with -0.11 dB/1° followed by multiyear sea ice, -0.07 dB/1°, and old multiyear ice (older than 3 years) with -0.04 dB/1°. The noise floor is found to have a strong impact on smooth first-year ice and estimated slopes are therefore not fully reliable. At co-polarization we obtained slope values of -0.24 dB/1°, -0.20 dB/1°, -0.15 dB/1° and -0.10 dB/1° for smooth first-year, rough/deformed first-year, multiyear and old multiyear sea ice respectively. Furthermore, we show that imperfect noise correction of the first subswath influences the obtained slopes for multiyear sea ice. We demonstrate that incidence angle normalization should not only be applied to co-polarization but should also be considered for cross-polarization images to minimize intra ice type variation in backscatter intensity throughout the entire image swath.

Knowing the location and type of sea ice is essential for safe navigation and route optimization in ice-covered areas. In this study, we developed a deep neural network (DNN) for pixel-based ice Stage of Development classification for the Baltic Sea using Landsat-8 optical satellite imagery to provide up-to-date ice information for Near-Real-Time maritime applications. In order to train the network, we labeled the ice regions shown in the Landsat-8 imagery with classes from the German Federal Maritime and Hydrographic Agency (BSH) ice charts. These charts are routinely produced and distributed by the BSH Sea Ice Department. The compiled data set for the Baltic Sea region consists of 164 ice charts from 2014 to 2021 and contains ice types classified by the Stage of Development. Landsat-8 level 1 (L1b) images that could be overlaid with the available BSH ice charts based on the time of acquisition were downloaded from the United States Geological Survey (USGS) global archive and indexed in a data cube for better handling. The input variables of the DNN are the individual spectral bands: aerosol coastal, blue, green, red and near-infrared (NIR) out of the Operational Land Imager (OLI) sensor. The bands were selected based on the reflectance and emission properties of sea ice. The output values are 4 ice classes of Stage of Development and Ice Free. The results obtained show significant improvements compared to the available BSH ice charts when moving from polygons to pixels, preserving the original classes. The classification model has an accuracy of 87.5% based on the test data set excluded from the training and validation process. Using optical imagery can therefore add value to maritime safety and navigation in ice- infested waters by high resolution and real-time availability. Furthermore, the obtained results can be extended to other optical satellite imagery such as Sentinel-2. Our approach is promising for automated Near-Real-Time (NRT) services, which can be deployed and integrated at a later stage at the German Aerospace Center (DLR) ground station in Neustrelitz.

Radar altimetry in the context of sea ice has mostly been exploited to retrieve basin-scale information about sea ice thickness. In this paper, we investigate the sensitivity of altimetric waveforms to small-scale changes (a few hundred meters to about 10 km) of the sea ice surface. Near-coincidental synthetic aperture radar (SAR) imagery and CryoSat-2 altimetric data in the Beaufort Sea are used to identify and study the spatial evolution of altimeter waveforms over these features. Open water and thin ice features are easily identified because of their high peak power waveforms. Thicker ice features such as ridges and multiyear ice floes of a few hundred meters cause a response in the waveform. However, these changes are not reflected in freeboard estimates. Retrieval of robust freeboard estimates requires homogeneous floes in the order of 10 km along-track and a few kilometers to both sides across-track. We conclude that the combination of SAR imagery and altimeter data could improve the local sea ice picture by extending spatially scarce freeboard estimates to regions of similar SAR signature.