Editorial for the Special Issue “Combining Different Data Sources for Environmental and Operational Satellite Monitoring of Sea Ice Conditions”

Satellite remote sensing is an important tool for continuous monitoring of sea ice covered ocean regions and spatial and temporal variations of their geophysical characteristics. Information on daily and weekly changes of the ice cover, provided by operational ice service, is essential for marine traffic and operations in ice-infested waters, and improves the understanding and forecasting of short-term interactions between atmosphere, ice, and ocean. When focusing on regional and local sea ice conditions, the synthetic aperture radar (SAR) is the most useful sensor due to its independence of daylight and cloud conditions and its high spatial resolution. However, the interpretation and analyses of SAR images are often prone to ambiguities. It is therefore beneficial to combine SAR images with data obtained from other types of satellite sensors (e.g., optical and thermal spectrometers, microwave radiometers, altimeters, scatterometers), and to compare them with results from airborne and ground-based field measurements when available. The retrieval of sea ice conditions and parameters does not only benefit from the combination of different data sources but also from linking such retrievals with results from modeling of sea ice thermodynamics and dynamics, or interpreting remote sensing data based on simulations of the interaction between electromagnetic radiation and sea ice. This Special Issue of Remote Sensing specifically addresses the potential of combining SAR with different complementary data sources in science studies and for operational applications, considering the most advanced technologies, for enhancing the sea ice monitoring capabilities and reducing ambiguities in data analysis. This also includes studies on improved methods for modeling of radar scattering and processing of data from complementary sensors.

Xu et al. [1] introduced a two-dimensional scattering model based on the finite element method (FEM) for simulating the microwave signatures from sea ice. In a first step, the comparison of results from the FEM, the small perturbation method and the method of moment reveal good agreement. Subsequently, the model is used to study the effect of subsurface scattering in sea ice. For newly formed sea ice with scattering contributions from both the air–ice and ice–water interfaces, the authors found that the backscattering coefficient (σ°) strongly oscillates as a function of sea ice thickness. This hampers the retrieval of the thickness of newly formed ice based on the magnitude of σ°. From simulating the scattering from first-year sea ice (FYI) with C-shaped salinity profiles the authors concluded that the salinity profile variations have very little influence on the backscattering at both C- and L-band.

Alekseeva et al. [2] compared sea ice concentration (SIC) obtained from the NASA Team (NT) algorithm, the ARTIST Sea Ice (ASI) algorithm, and the “Variation Arctic/Antarctic Sea Ice Algorithm 2 (VASIA2)” with ship-based observations carried out in the winter season throughout the years 1996–2005 in the Russian marginal seas in the Arctic. Here, the VASIA2 SIC algorithm has been developed by the authors of [2]. The observed discrepancies between ship data and the SIC charts were largest in summer and in open ice areas.

The automatic identification system (AIS) is used for broadcasting information on ship status and operations (e.g., speed and location). It is mandatory for most ship classes. Similä and Lensu [3] used the AIS database in combination with near-simultaneously acquired SAR imagery to create an estimated ship speed chart in the Gulf of Bothnia for ships that can navigate in ice without icebreaker assistance. The case study showed that by combining AIS and SAR data the maximum achievable ship speed can be well predicted for different ice conditions. The developed method can be applied to generate a chart of expected ship speeds based on a SAR image.

The increased role of altimeters in today’s cryospheric sciences is reflected by two articles that deal with measurements of the CryoSat-2 (CS2) altimeter [4,5]. Aldenhoff et al. [4] investigated the sensitivity of altimeter waveforms to variations of the sea ice surface on scales from a few hundred meters to about 10 km. They compared different CS2 waveform parameters with near-coincident SAR imagery acquired over the Beaufort Sea. The authors found that changes of the surface characteristics related to leads, large ridges or ice floes of different sea ice types were reflected in the changes of the waveform parameters. Shen et al. [5] focus on the retracking problem arising when processing altimeter data. Their usefulness for the retrieval of sea ice freeboard requires that the distance from the satellite to the reflecting surface is highly accurately determined. The proposed method relies on Bezier curve fitting (BCF) to the CS2 waveform. The results suggest that the sea ice freeboard data can be obtained with a higher accuracy when using the BCF method instead of the threshold first-maximum retracker algorithm which is commonly used in the freeboard estimation.

For estimating sea ice thickness (SIT) from radar or laser altimeter data, information on properties of the snow cover on sea ice is needed. Such information is also useful for seasonal sea ice forecast models, since the snow pack changes the sea-ice heat budget by increasing the albedo and acting as an insulating layer. Yackel et al. [6] investigated the retrieval of snow depth using data from C- and Ku-band spaceborne scatterometer acquired over landfast FYI in the Canadian Arctic Archipelago. To this end they investigated how daily time series of σ° are dependent on snow depth. They found that during the late winter season, a thinner snow cover displays a larger variance in daily σ° compared to a thicker snow cover on FYI. They demonstrated that it may be possible to estimate the relative thickness of snow on landfast FYI in the late winter season prior to melt onset.

Automatic detection of icebergs in satellite images supports their manual identification, which is necessary for safety of marine transport and operations. Soldal et al. [7] tested a new approach for iceberg detection which they applied to Sentinel-1 Extra Wide Swath (EWS) SAR images acquired over fast ice in the Barents Sea. Their approach is based on a filter making use of the cross-polarized (HV) channel for enhancing the contrast between icebergs and background (ocean or sea ice). Subsequent blob detection and a constant false alarm rate (CFAR) detector are applied. The authors compared their results with a database of 2000 icebergs visually identified in Sentinel-2 Multi Spectral Imager (MSI) data. The authors found that many icebergs of a size comparable to the spatial resolution of the EWS-mode are not detectable. Two reasons were given: (1) at C-band, not all icebergs are strong scatterers at HV-polarization; (2) icebergs and deformation structures present on fast sea ice can often not be distinguished since both may reveal equally strong responses at HV-polarization.

Methods for sea ice classification are useful for supporting analysts who operationally generate sea ice charts. Lohse et al. [8] introduce a fully automatic design of a decision-tree algorithm and demonstrate its application to ice type classification in SAR images. Each branch of the tree separates one single class and takes it out of the dataset. The order of classification steps and the respective optimal feature set (e.g., σ° at different polarizations and frequencies) are determined through a sequential forward feature selection (SFFS) by explicitly considering the achieved classification accuracy for a single class. The authors applied their algorithm on simulated data and SAR sea ice images. In the latter case they achieved improvements of the average per-class classification accuracies between 1% and 2.5% compared to traditional all-at-once classification. For single ice types, however, improvements up to 8% were obtained.

Pancake and frazil ice represent types that typically occur in the marginal ice zone. Aulicino et al. [9] developed a processing method for routinely retrieving ice thickness of frazil/pancake ice covers. To this end, they used the spectral changes in wave spectra that are derived from SAR imagery. They tested their method employing COSMO-SkyMed SAR images acquired over Antarctic frazil/pancake ice fields. The authors analyzed and validated their results through a comparison with co-located in situ observations collected during the 2017 PIPERS cruise in the Terra Nova Bay polynya and found good agreement between in-situ and retrieved thicknesses.

Idzelytė et al. [10] used images from different satellite SAR systems and Moderate Resolution Imaging Spectroradiometer (MODIS) acquired in a time interval from 2002 to 2017 to document ice cover conditions in the Curonian Lagoon (CL), Europe’s largest coastal lagoon located in the southeastern Baltic Sea. Using both satellite and in-situ data, the authors generate ice season duration (ISD) maps that reveal spatial details of ice growth and decay in the CL. An important observation is the pronounced shortening of the ice season duration in the CL at a rate of 1.6–2.3 days per year, which is much higher than reported for the nearby Baltic Sea regions. The authors found a correlation between the ISD, air temperature, and winter NAO index which is substantially higher when considering the lagoon-averaged ISD values derived from satellite observations compared to those derived from coastal records.