Alexander S. Komarov

Sea Ice Motion Tracking From Sequential Dual-Polarization RADARSAT-2 Images

A new sea ice motion tracking algorithm that operates with two sequential synthetic aperture radar (SAR) RADARSAT-2 ScanSAR images is presented. The feature tracking approach is based on the combination of the phase-correlation and cross-correlation methods. An algorithm for selecting control points, a matching technique, an approach for filtering out error vectors, and a confidence levels setting for output drift vectors were specifically developed in order to increase the systems robustness and accuracy. We evaluated ice motion tracking results derived from HH and HV channels of RADARSAT-2 ScanSAR imagery and formulated a condition where the HV channel is more reliable than the HH channel for ice tracking. Furthermore, we found that the ice motion tracking from the HV channel is not affected by noise floor stripes, which are prominent in the cross-polarization RADARSAT-2 ScanSAR images. The developed sea ice tracking technology was implemented at the Canadian Ice Service, Environment Canada for operational use. The system was successfully run to provide operational support of field work in the Arctic Ocean in compliance with the United Nations Convention on the Law of the Sea in the spring of 2010.

Ocean Surface Wind Speed Retrieval From C-Band SAR Images Without Wind Direction Input

Two new models for wind speed retrieval from C-band synthetic aperture radar (SAR) data have been developed, based on a large body of statistics on buoy observations collocated and coinciding with RADARSAT-1 and -2 ScanSAR images. The first models independent variables are co-polarization (HH) normalized radar cross-section (NRCS), and antenna beam incidence angle. The second models predictors are HH NRCS, cross-polarization (HV) NRCS, instrument noise floor, and incidence angle. The latter model has better accuracy than the first because of using an additional HV variable. Furthermore, we found that the proposed models without wind direction input demonstrated a better accuracy than CMOD_IFR2 and CMOD5.N models in combination with the SAD HH co-polarization ratio (VV/HH), which require wind direction input. These results were confirmed on a large independent subset of collected data. The developed wind speed retrieval models, in conjunction with our previously developed ice motion tracking algorithm, can be a useful tool for studying sea ice dynamics processes in the marginal ice zone. The developed models have been integrated into a quasi-operational system at the Meteorological Service of Canada.

A Study on the C-Band Polarimetric Scattering and Physical Characteristics of Frost Flowers on Experimental Sea Ice

A focused study on the C-band polarimetric scattering and physical characteristics of frost-flower-covered sea ice was conducted at the Sea-Ice Environmental Research Facility over a three day period. Sea ice was grown in an outdoor pool outfitted with automated sensors to monitor environmental conditions. C-band polarimetric scattering measurements were conducted continuously at a range of incidence angles, and surface roughness statistics were obtained at discrete times using a laser scanner system LiDAR. Four stages of development were identified that exhibited notably different physical and scattering characteristics: 1) initial formation; 2) surface brine expulsion; 3) frost flower growth; and 4) decimation. An optimal polarization and incidence angle is not readily apparent for the purposes of identifying the frost flower development Stages I-III; however, the lower incidence angles (25° and 35°) appear to be most sensitive to the surface brine expulsion. Only the dual-polarization measurements at low incidence angles (e.g., 25°) could be used to identify the onset of the decimation stage. Backscatter increased rapidly during the initial formation, with a local maximum corresponding to ~ 80% areal coverage of frost flowers, followed by a local minimum when the surface was covered by a brine-rich surface layer, connoting that surface brine expulsion may be identified using polarimetric scatterometry.

Parameterization of Centimeter-Scale Sea Ice Surface Roughness Using Terrestrial LiDAR

Microwave scattering from sea ice is partially controlled by the ice surface roughness. In this paper, we propose a technique for calculating 2-D centimeter-scale surface roughness parameters, including the rms height, correlation length, and form of autocorrelation function, from 3-D terrestrial light detection and ranging data. We demonstrate that a single scale of roughness can be extracted from complex sea ice surfaces, incorporating multiple scales of topography, after sophisticated 2-D detrending, and calculate roughness parameters for a wide range of artificial and natural sea ice surface types. The 2-D technique is shown to be considerably more precise than standard 1-D profiling techniques and can therefore characterize surface roughness as a stationary single-scale process, which a 1-D technique typically cannot do. Sea ice surfaces are generally found to have strongly anisotropic correlation lengths, indicating that microwave scattering models for sea ice should include surface spectra that vary as a function of the azimuthal angle of incident radiation. However, our results demonstrate that there is no fundamental relationship between the rms height and correlation length for sea ice surfaces if the sampling area is above a threshold minimum size.

Numerical and Experimental Evaluation of Terrestrial LiDAR for Parameterizing Centimeter-Scale Sea Ice Surface Roughness

Terrestrial light detection and ranging (LiDAR) offers significant advantages over conventional techniques for measuring the centimeter-scale surface roughness of natural surfaces, such as sea ice. However, the laser scanning technique is inherently limited, principally by the following: 1) the high inclination scanning angle of the sensor with respect to nadir; 2) the precision of the laser ranging estimate; and 3) the beam divergence of the laser. In this paper, we introduce a numerical model that has been designed to simulate the acquisition of LiDAR data over a regular rough surface. Results from the model compare well (r2 = 0.97) with LiDAR observations collected over two experimental surfaces of known roughness that were constructed from medium-density fibreboard using a computer numerical control three-axis router. The model demonstrates that surface roughness parameters are not sensitive to minor variations in the LiDAR sensors range and laser beam divergence, but are slightly sensitive to the precision of the ranging estimate. The model also demonstrates that surface roughness parameters are particularly sensitive to the inclination angle of the LiDAR sensor. The surface RMS height is underestimated, and the correlation length is overestimated as either the inclination angle of the sensor or the true roughness of the surface increases. An isotropic surface is also increasingly observed as an anisotropic surface as either the inclination angle or the true surface roughness increases. Based on the model results, we propose a set of calibration functions that can be used to correct in situ LiDAR measurements of surface roughness.

Electromagnetic Wave Scattering from Rough Boundaries Interfacing Inhomogeneous Media and Application to Snow-Covered Sea Ice

In this study a new analytical formulation for electromagnetic wave scattering from rough boundaries interfacing inhomogeneous media is presented based on the first-order approximation of the small perturbation method. First, we considered a scattering problem for a single rough boundary embedded in a piecewise continuously layered medium. As a key step, we introduced auxiliary wave propagation problems that are aimed to link reflection and transmission coefficients in the layered media with particular solutions of one-dimensional wave equations at the mean level of the rough interface. This approach enabled us to express the final solution in a closed form avoiding a prior discretization of the inhomogeneous medium. Second, we naturally extended the obtained solution to an arbitrary number of rough interfaces separating continuously layered media. As a validation step, we demonstrated that available solutions in the literature represent special cases of our general solution. Furthermore, we showed that our numerical results agree well with published data. Finally, as a particular special case, we presented a formulation for scattering from inhomogeneous snow-covered sea ice when the dominant scattering occurs at the snow-ice and air-snow interfaces.

Modeling and Measurement of C-Band Radar Backscatter From Snow-Covered First-Year Sea Ice

In this paper, we present model and measurement results for C-band HH and VV normalized radar cross-sections (NRCS) from winter snow-covered first-year sea ice with average snow thicknesses of 16, 4, and 3 cm. The brine content in snow pack was low in all three case studies, which is typical for cold winter conditions. We used the first-order approximation of the small perturbation theory accounting for surface scattering from the air-snow and snow-ice rough interfaces and continuously layered snow and sea ice. The experimental data were collected during the Circumpolar Flaw Lead system study in the winter of 2008 in the southern Beaufort Sea from the research icebreaker Amundsen. Good agreement between the model and experimental data were observed for all three case studies. The model results revealed that the scattering at the snow-ice rough interface is usually stronger than that at the air-snow interface. Furthermore, both model and experimental NRCS values (at VV and HH polarizations) were considerably higher for thin-snow cover compared with the thick-snow-cover case. We associate this effect with the lower attenuation of the propagated wave within the thin-snow pack in comparison to the thick-snow pack. We also demonstrated that different brine volume contents in snow with close thicknesses of 4 and 3 cm did not affect the backscattering coefficients at certain incidence angles and polarization. Our findings provide the physical basis for winter snow thickness retrieval and suggest that such retrievals may be possible from radar observations under particular scattering conditions.

Inversion-Based Sensitivity Analysis of Snow-Covered Sea Ice Electromagnetic Profiles

For the microwave remote sensing of snow-covered sea ice dielectric profiles, the sensitivity of the normalized radar cross-section data with respect to the complex permittivity and thickness values is investigated. Our results show that the data collected closer to the nadir in monostatic setups, and the data collected closer to the specular angle in bistatic setups represent higher sensitivity values. Using both synthetically and experimentally collected data sets, we demonstrate that the inversion of data sets having higher sensitivity values results in enhanced reconstruction accuracy.

Marine wind speed retrieval from RADARSAT-2 dual-polarization imagery

New regression models for the retrieval of marine wind speeds from RADARSAT-2 dual-polarization images and results of the comparative analysis of these models with the C-band geophysical model function CMOD_IFR2 are presented. In the two new models, synthetic aperture radar (SAR) co-polarization (VV) and cross-polarization (VH) radar cross-sections, instrument noise floor, and antenna beam incidence angle were used as independent variables. In addition, one of the models also included wind direction input. To build and test the empirical relationships between wind speed and RADARSAT-2 parameters, we created a database containing 570 samples of ocean buoy wind speed observations collocated and coincident with the information obtained from SAR images. The models were tested on independent wind speed data reaching up to 40 knots. The CMOD_IFR2 was also tested on the same data. Analysis of the test results proved a higher accuracy of the new regression models, including the model without wind direction, as compared with CMOD_IFR2, which uses wind direction information. This means that the degraded accuracy of the SAR wind retrieval model without input of the wind direction can be compensated for by using cross-polarization backscatter information. The models were validated through current buoy and image data, provided by the quasi-operational Wind Information Processing System (WIPS). The developed models have been integrated into and are currently functioning within WIPS (at the Meteorological Service of Canada).

Retrieval of Young Snow-Covered Sea-Ice Temperature and Salinity Evolution Through Radar Cross-Section Inversion

This paper utilizes an electromagnetic inverse-scattering algorithm to quantitatively reconstruct the vertical temperature and salinity profiles of snow-covered sea ice from time-series monostatic polarimetric normalized radar cross-section (NRCS) data. The reconstructed profile at a given time step is utilized to provide a priori information for the profile reconstruction at the subsequent time step. This successive use of a priori information in the inversion algorithm results in achieving high reconstruction accuracy over the time period of interest. This inversion scheme is evaluated against the experimental data collected from snow-covered sea ice grown in an Arctic ocean mesocosm facility. It will be shown that the time evolution of the temperature, salinity, and density profiles of an artificially grown snow-covered sea ice can be quantitatively reconstructed using single-frequency time-series radar cross-section data assuming that these profiles are initially known with sufficient accuracy.