Sea ice surface roughness is an important control on heat and momentum exchanges between the atmosphere and ocean, as well as albedo through the formation and distribution of surface melt ponds during melting conditions. Roughness and melt pond distributions are also spatially heterogeneous, making them difficult to assess within the context of resolving the mechanisms contributing to sea ice melt. In this study we address the utility backscatter and polarimetric parameters from C- and L-band frequency, polarimetric synthetic aperture radar (polSAR), systems for deriving sea ice surface roughness and melt pond fraction statistics at high spatial resolutions (metre-scale). Airborne laser scanner measured surface topography data from first-year and multiyear sea ice targets during the late winter period (April 2016 and April 2017) in the Canadian Arctic Archipelago (CAA), collected coincident to RADARSAT-2 and ALOS-2/PALSAR-2 polSAR datasets, are used in an assessment of optimal SAR imaging parameters for deriving roughness statistics. Shallow incidence angle measured linear HH-polarization backscatter at C-band frequency shows the strongest correlation with level and deformed first-year ice surface roughness (R=0.86). The inclusion of polarimetric information does not substantially improve the relationship. A method for predicting sea ice melt pond fraction from winter measured HH-polarization backscatter from the Sentinel-1 SAR is detailed, along with application of the technique to a 9-year record of RADARSAT-2 imagery of the CAA (2009-2017). Relationships between predicted melt pond fraction on summer sea ice area in the CAA are assessed in the context of seasonal sea ice prediction. Removing an outlier year, that being 2013 when surface air temperature was lower than normal throughout the melt season and buffered sea ice ablation, results in statistically significantly de-trended correlations between melt pond fraction and August/September sea ice area at the weekly scale up to R=-0.73. This result points to potential for C-band SAR predicted melt pond fraction to be used in a suite of sea ice forecasting tools accounting for dynamic/thermodynamic processes contributing to sea ice ablation.

The increasingly ice-free Arctic Ocean is transitioning from a quiescent ice-covered basin to a more energetic ocean, dominated by free-surface phenomena. A key aspect of this transition is the emerging presence of surface waves. In the Southern Ocean, sea ice extent increased from the beginning of the satellite record until 2014, at which time it showed a dramatic decrease. This behaviour has been poorly captured by climate models, and wave forcing has been proposed as a potentially contributing mechanism. Waves are known to break up ice over large areas, but they also have important effects on ice formation and the turbulence structure of the upper ocean. These processes have the potential to create feedback mechanisms that could either increase the rate of ice melt, or help stabilize the ice pack.

Our current understanding of the interactions between waves and ice is largely limited by the availability of observations. Until very recently, models and theoretical results could only be compared against a few experiments, limited in geographical extent and parameter space. That dearth of data was due to the high costs and technical difficulties of making traditional buoy measurements in polar latitudes. This presentation will discuss results from airborne and satellite remote sensing techniques that are being developed to overcome those challenges.

Findings from two airborne campaigns will be discussed. The measurements captured wave propagation in Arctic sea ice over a varied parameter space of wave, wind, and ice conditions. By capturing wave evolution over flight paths of order 100km, these measurements provide insight into the physical mechanisms of wave attenuation in the marginal ice zone. Key results include constraints on the validity of wave scattering models, and modified wave fetch relations in ice-forming conditions.

At basin scales, Sentinel-1 SAR imagery has been used to develop the largest-ever dataset of waves in sea ice. The dataset contains thousands of unique spectral wave measurements, and continues to grow with regular new acquisitions. The measurements surround the Antarctic continent with all-season, multi-year, coverage. This dataset shows the spatiotemporal evolution of the wave field in the Antarctic marginal ice zone, allowing assessment of the large-scale significance of the dynamical effects of the wave field. Wave forcing was found to be comparable to wind forcing over the outer 50 km of the Antarctic marginal ice zone. Although significant spatial variability was observed, waves were found to provide a mean on-ice forcing, potentially providing a stabilising effect on the ice edge.

The global-scale implications of the combined airborne and satellite-based results will be considered in the context of observed long-term changes in polar ice cover. Finally, planned future work leveraging the remote-sensing techniques and results presented here will be discussed.

The European Space Agency’s Envisat and CryoSat-2 satellite radar altimeter missions provide a continuous 17-year (2002-present) dataset with the potential to estimate Arctic sea ice thickness up to 81.5°N. However, geometric sampling differences cause discrepancies in the sea ice floe size distributions measured by the missions. We find that Envisat sampling is biased towards wider, thicker sea ice floes. This leads to a mean sea ice thickness difference between Envisat and CryoSat-2 of 77 cm over first year ice and 22 cm over multiyear ice. By applying an Envisat thickness correction that is dependent on floe size distribution we provide a novel, physically-based approach to addressing the bias between the satellites. The correction enables us to produce a reconciled 17-year time series of Arctic sea ice thickness in the seasonal ice zone. Here we present 17 years of sea ice thickness changes, ice thickness error budget, and evaluation with a suite of in situ data (airborne, moored buoy, and submarine). There is the potential to extend this methodology to our current estimates of Antarctic sea ice freeboard.

Snow thickness retrieval methods on sea ice using various high frequency (e.g. Ka- and Ku-band) active microwave channels is an active area of research in the sea ice research community given the importance of snow as a physical variable impacting energy, physical and chemical processes operating across the ocean-sea-ice-atmosphere interface. Through its thermodynamic control on atmospheric and oceanic heat fluxes, snow cover dictates sea ice growth and decay rates, which in turn control air-sea mass and energy transfers in the polar regions. Despite the importance of snow on sea ice as an integral component of the marine cryosphere, much of the Arctic is severely lacking regionally specific and accurate snow on sea ice information with respect to its thickness distribution over space and time. Previous work has highlighted the role of a saline snow basal layer and its influence on brine volume and dielectric properties towards increasing microwave backscatter and shifting the main scattering horizon of Ku-band altimeter data. In this study, we assess daily backscatter (sigma naught) data from the Advanced SCATerometer (ASCAT) C-band enhanced resolution (4.45 km) product to investigate the variance in winter backscatter of FYI within the Canadian Arctic Archipelago (CAA). This higher resolution data product can better resolve small-scale geographical features, such as coastlines, narrow channels, and leads and polynyas. We utilize the all passes product in which sigma naught (σ°) is normalized to a 40° incidence angle.


In this work, our objective is to describe and assess a simple empirically-derived methodology for estimating relative snow thickness on smooth, land-fast first-year sea ice (FYI) based on the variance in microwave backscatter during the late winter season prior to melt onset. We investigate the FYI ice type because we are interested in exploiting the thermodynamic-brine-dielectric effect (Barber et al., 1998; Barber and Nghiem, 1999) within the snow cover and its effect on microwave backscatter. We specifically select land-fast FYI because we are interested in assessing how a locations’ snow thickness distribution modulates the thermodynamic and dielectric properties of the snow cover over the course of variable atmospheric forcing during the winter season. We also specifically assess smooth FYI because we are interested in measuring the thermodynamically induced effect on basal brine layer snow dielectrics and subsequent microwave scattering, and not scattering resulting from macroscale surface roughness features or deformed ice that is often heterogeneously distributed across an icescape within an ASCAT 4.45 km resolution cell. Our hypothesis stemming from previous research is that thinner snow covers on smooth, thick (> 1.5 m) FYI should manifest a larger variance (at the daily time scale) in spaceborne microwave backscatter for a given air temperature change and forcing, at the diurnal to daily time scale, in late winter prior to melt onset. This is because the basal snow layer brine volume will respond to this air temperature forcing more rapidly and with greater change for a thin snow cover compared to a thicker, greater insulating, snow cover, owing to the lower (higher) snow thermal conductivity or higher (lower) snow thermal diffusivity, respectively. As a result, the dielectric properties will also manifest a correspondingly smaller or larger change and directly affect the amount of surface and/or volume scattering from the snow/ice system. Results suggest that spatially significant patterns of relatively thin and thick snow on sea ice can be identified across the CAA. Our assessment of these patterns with Environment and Climate Change Canada (ECCC) Regional Ice-Ocean Prediction System (RIOPS) estimates are in good agreement. The results suggest potential for exploiting this simple, yet robust, approach for quantitative estimates of snow thickness on sea ice validated with in situ observations.

In our study, we present a method to extrapolate sea ice thickness (SIT) derived from CryoSat-2 (CS-2) radar altimeter elevation measurements, both temporally and spatially, using Sentinel-1 (S-1) SAR data. The resulting sea ice thickness product, thus combining S-1 SAR and CS-2 altimetry, is a significant step towards the operational use of radar altimetry for supporting sea ice navigation.

We have studied the combination of SIT data derived from CS-2 and S-1 C-band SAR data for estimating SIT in the spatial resolution of 500 m. Our study was conducted for the winter 2017 (January-April 2017) over a study area in the Kara and Barents seas. Daily SAR mosaics generated from S-1 EW Ground Range Detected Medium resolution (GRDM) dual-polarized (HH/HV) SAR data were used. The mosaics are downsampled to a resolution of 500 m from the original SAR imagery and the final SIT estimates are also provided in this resolution. A SAR segmentation, based on the ICM algorithm [1] utilizing both of the SAR polarization channels is performed and multiple texture features are computed for each SAR segment. The texture features include local autocorrelation, entropy, coefficient of variation, kurtosis, HH/HV cross-correlation as well as segment-wise edge and corner point densities. A feature vector consisting of the mentioned features is constructed for all the SAR segments and assigned to the center point (2-D center of mass) of the segment. The CS-2 SIT values processed with the FMI implementation of PySIRAL (a python software package for altimeter SIT processing) are used to assign an initial SIT to those SAR segment which have at least N SIT measurements available within the segment. Different values of the parameter N have been studied. The temporal requirement is that the SIT values are acquired within M days preceding the SAR mosaic date (M=3 and M=7 examined). Segment-wise median of the CS-2 SIT values within each segment is assigned to these SAR segments. For the remaining SAR segments, still without a SIT value assigned to them, SIT values are determined by utilizing the segment-wise SIT values already existing and the similarity of the segments. To assign the missing SIT values we minimize a target function among all the segments with a SIT value assigned based on the CS-2 data. The target function to be minimized over all the segments, paired with the segment for which the SIT value will be assigned, with a CS-2 thickness is a function of the pairwise Euclidean distance between the segment locations, defined by their center points , and the Euclidean distances of the segment feature vectors with a linear weighting of the two terms. The linear weights are estimated using a set of SIT segments as the training data. Based on this extrapolation SIT values for all the SAR mosaic segments are computed. Because CS-2 SIT values typically tend to overestimate SIT for thin ice, we additionally perform a remapping of the SIT values as a post-processing step. This remapping is determined using a matching between SAR/CS-2 based SIT distributions and SIT distributions extracted from TOPAZ-4 ice model [2] nowcasts over the study area for a representative data set. The matching of the distributions is performed by minimizing the Kolmogorov-Smirnov distance of the two (cumulative) distributions. This approach assumes that the ice model gives a good estimation of the overall SIT distribution over the study area but does not always give good local SIT estimates as we have seen in our earlier studies.

The results have been compared to the SIT values derived from Russian Arctic-Antarctic Research Institute (AARI) ice charts and those of the FMI SIT retrieval algorithm [3] based on S-1 SAR, microwave radiometer (AMSR2) and the TOPAZ-4 ice model over the study area. The results show that it is feasible to use CS-2 ice thickness instead of using TOPAZ-4 data as background data for SIT estimation, thus estimating SIT based on EO data alone. Additionally, microwave radiometer data can be included to improve detection of thin ice areas.

References:
1. Besag, J. E., On the Statistical Analysis of Dirty Pictures, Journal of the Royal Statistical Society, Series B, v. 48, n. 3, pp. 259–302, 1986.

2. Sakov, P, F. Counillon, L. Bertino, K. A. Lisæter, P. R. Oke, A. Korablev, TOPAZ4: an ocean-sea ice data assimilation system for the North Atlantic and Arctic, Ocean Science, v. 8, pp. 633-656, 2012.

3. Simila, M, M. Makynen, J. Karvonen, A. Gegiue, A. Gierisch, Modeled sea ice thickness enhanced by remote sensing data, Proc. European Space Agency Living Planet Symposium, Prague, Czech Republic, 9-13 May 2016, ESA SP-740, 2016.

Variability of the Arctic sea-ice mass balance is determined by sea ice production and melt on the one hand, and sea ice export on the other hand. Sea ice exported through the Fram Strait also represents a significant fresh water input to the North Atlantic, which could in turn modulate the intensity of the thermohaline circulation. We present estimates of winter sea ice volume export through the Fram Strait using CryoSat- 2 sea ice thickness retrievals and three different drift products for the years 2010 to 2017. The export rates vary between -21 and -540 km³/month. We find that ice drift variability is the main driver of annual and interannual ice volume export variability, and that the interannual variations of the ice drift are driven by large scale variability of the atmospheric circulation captured by the Arctic Oscillation and North Atlantic Oscillation indices. On shorter timescale, however, the seasonal cycle is also driven by the mean thickness of exported sea ice, typically peaking in March. Considering Arctic winter multiyear ice volume changes, 54 % of the variability can be explained by the variations of ice volume export through the Fram Strait. Here, we also show first results from extending the ice volume export time series back to 2002, including the ESA Climate Change Initiative sea-ice thickness record using data from CryoSat-2 and its radar altimeter predecessor Envisat.