Two techniques for automated sea-ice tracking, image pyramid area correlation (hierarchical correlation) and feature tracking, are described. Each technique is applied to a pair of Seasat SAR sea-ice images. The results compare well with each other and with manually tracked estimates of the ice velocity. The advantages and disadvantages of these automated methods are pointed out. Using these ice velocity field estimates it is possible to construct one sea-ice image from the other member of the pair. Comparing the reconstructed image with the observed image, errors in the estimated velocity field can be recognized and a useful probable error display created automatically to accompany ice velocity estimates. It is suggested that this error display may be useful in segmenting the sea ice observed into regions that move as rigid plates of significant ice velocity shear and distortion.< >
Sea ice ridges and keels (hummocks and bummocks) are important in sea ice research for both scientific and practical reasons. A long-term objective is to make quantitative measurements of sea ice ridges using synthetic aperture radar (SAR) images. The preliminary results of a scattering model for sea ice ridge are reported. The approach is through the ridge height variance spectrum Psi(K), where K is the spatial wavenumber, and the two-scale scattering model. The height spectrum model is constructed to mimic height statistics observed with an airborne optical laser. The spectrum model is used to drive a two-scale scattering model. Model results for ridges observed at C- and X-band yield normalized radar cross sections that are 10 to 15 dB larger than the observed cross sections of multiyear ice over the range of angles of incidence from 10 to 70 deg.
Ridges and keels (hummocks and bummocks) in sea ice flows are important in sea ice research for both scientific and practical reasons. Sea ice movement and deformation is driven by internal and external stresses on the ice. Ridges and keels play important roles in both cases because they determine the external wind and current stresses via drag coefficients. For example, the drag coefficient over sea ice can vary by a factor of several depending on the fluid mechanical roughness length of the surface. This roughness length is thought to be strongly dependent on the ridge structures present. Thus, variations in ridge and keel structure can cause gradients in external stresses which must be balanced by internal stresses and possibly fracture of the ice. Ridging in sea ice is also a sign of fracture. In a practical sense, large ridges form the biggest impediment to surface travel over the ice or penetration through sea ice by ice-strengthened ships. Ridges also play an important role in the damage caused by sea ice to off-shore structures. Hence, observation and measurement of sea ice ridges is an important component of sea ice remote sensing. The research reported here builds on previous work, estimating the characteristics of ridges and leads in sea ice from SAR images. Our objective is to develop methods for quantitative measurement of sea ice ridges from SAR images. To make further progress, in particular, to estimate ridge height, a scattering model for ridges is needed. Our research approach for a ridge scattering model begins with a survey of the geometrical properties of ridges and a comparison with the characteristics of the surrounding ice. For this purpose we have used airborne optical laser (AOL) data collected during the 1987 Greenland Sea Experiment. These data were used to generate a spatial wavenumber spectrum for height variance for a typical ridge - the typical ridge is the average over 10 large ridges. Our first-order model radar scattering includes both the quasi-specular and Bragg resonant scatter mechanisms. This model is extended to include contributions from volume scatter and scatter from discrete objects. Geometrical characteristics from the AOL survey and model calculations imply that for radar wavelengths and observation geometries that are dominated by the quasi-specular scattering mechanism radar backscatter from a ridge is a measure of peak ridge height. We present scattering model results and compare them with ridges observed during the LEADEX experiment of March-April 1992 when both X, C, and L-band aircraft SAR and the ERS-1 satellite SAR observed a region in the Beaufort Sea near 86 deg N, 10 deg W. Data were also collected documenting ridge characteristics on the surface. The surface data are used to generate a SAR signature via the scattering model described above. The predicted SAR signatures compare well with the SAR observations.
Abstract : The origins of HF surface wave radar for ocean wave and current measurements began with collaborative work at Stanford University and Scripps Institution of Oceanography in the late 1960's. Two of the participants in this project (Drs. Teague and Vesecky) have worked with HF radar observations of the ocean since these early experiments. Since that time, HF radar as an ocean sensing tool has progressed with increasing acceptance in the oceanography community over the last five years. During this grant a new HF radar design was completed, a prototype unit was constructed and is now being tested over Monterey Bay, California from a field site kindly provided at the Long Marine Laboratory of the University of California at Santa Cruz. This operation is in collaboration with the REINAS project at UC Santa Cruz that is also funded by ONR and the results of the radar measurements are being made available over the internet by the REINAS project. Initial results, including radial current field maps at four frequencies and variations of currents with time are to be presented at the International Geoscience and Remote Sensing Symposium in Singapore during August 1997. This report begins with a overview of the radar, its installation and some preliminary results. This is followed by two sections describing the operating characteristics of the radar and some further results, including measurements of vertical current shear in the top meters of the ocean. Section III contains the bulk of the system description with further information in volume II. The strong support and excellent suggestions from Drs. Dennis Trizna and Frank Herr at the Office of Naval Research are important in the present and future success of the project.
A new high-frequency (4-25 MHz) phased-array radar, constructed jointly by the University of Michigan, the Environmental Research Institute of Michigan and Stanford University, was installed at Santa Cruz, California in July, 1996. After initial equipment checkout and antenna calibration using a transponder carried on a small boat, regular data collection started in October, 1996. Installation of a second HF radar is planned at a site south of Santa Cruz to allow resolution of current vectors. Wind and wave data from moored buoys in the radar field of view, wind sensors at several coastal sites, and current measurements from several CODARs in the Monterey Bay area also are available.
Over-water wind speeds and directions derived from a ground-wave high frequency radar-the Multi-frequency Coastal Radar (MCR)-are compared to in-situ observations to examine the skill of the radar measurements. Conventional beam formation processing of radar data collected from two sites around Monterey Bay during summer 1997 is used to produce wind directions based on the relative strength of the positive and negative Bragg-resonant peaks, which correspond to wind-driven waves approaching and receding from the radar, respectively. The functional relationship that converts these signal levels, which primarily reflect surface wave conditions, to a wind direction are a central research issue. To address this issue, the remotely sensed estimates are compared against moored observations under variable wind conditions. Wind direction algorithms based on Bragg ratios applied to this or any HF radar cannot detect wind speed. However, the unique capability of the MCR to, simultaneously, measure at four Bragg-resonant frequencies is used to assess the sensitivity of near-surface current shear to the measured wind speed.
An unsupervised method that chooses and applies the most appropriate tracking algorithm from among different sea-ice tracking algorithms is reported. In contrast to current unsupervised methods, this method chooses and applies an algorithm by partially examining a sequential image pair to draw inferences about what was examined. Based on these inferences the reported method subsequently chooses which algorithm to apply to specific areas of the image pair where that algorithm should work best.
The reported work describes the results of using object-oriented image analysis techniques to study the areal distribution of first-year ice. The authors have examined the region roughly corresponding to 72/spl deg/N, 140/spl deg/W, which coincides with the area studied under the LEADEX campaign in the Beaufort Sea gyre. At least 32 ERS-1 images of this location have been analyzed so far, of which 12 are discussed in the paper. The findings suggest the likelihood in identifying a set of equations that can characterize the statistical areal distribution of first-year ice over a large portion of the Arctic. The findings further indicate that this relationship persists throughout the year. The results suggest that although the generation of first-year ice is driven by meteorological systems that result in a significant redistribution of multiyear ice, the statistics that describe the population distribution of first-year ice remain stable in the first order.< >
In July 1996 a new multifrequency (4-25 MHz) HF radar was installed at the Long Marine Laboratory (University of California at Santa Cruz) on the north coast of Monterey Bay. This radar is capable of observing near-surface currents at varying depths in the top two metres of the ocean. Observations were made over a ten-day period in March 1997 during which there was a strong land-sea breeze circulation over Monterey Bay. Radial current measurements corresponding to depths of about 0.3, 0.5, 1.0 and 1.4 m were made during this period using HF radar data from four operating frequencies. Acoustic Doppler Current Profiler (ADCP) measurements were made at the MBARI M1 buoy near the mouth of the bay. Time series and Fourier analyses of these data show that very near the surface the strongest periodic component is a diurnal one corresponding to the diurnally varying surface stress from the land-sea breeze. At deeper depths the diurnal component remains, but a semi-diurnal component grows in strength with increasing depth of the current measurement. Thus, multifrequency HF radar combined with current measurements from buoys and moorings can investigate the air-sea interaction from within about 30 cm of the surface to depths of 50 m and more.
