11 The RADARSAT Geophysical Processor System
253
surements of the 'ice particles' that are located on the initial grid which covers a region
or the entire Arctic Ocean. The sampling interval is determined by repeat coverage of
these points by SAR imagery. With an initial grid spacing of 5 km, there are approximately 500000 points in this array. Over a winter season of approximately 8 months,
there are close to 32 observations of each trajectory. The gridded 50-km Eulerian product represents weekly ice motion which is spatially and temporally interpolated to the
grid locations from these trajectories. We also produce routine Eulerian observations
of ice motion from image pairs for locations like the Greenland Sea and Bering Strait,
where the character of the ice motion is such that the deformation does not allow us to
maintain the density of points to define a Lagrangian grid. In these special regions, we
select a high-repeat coverage frequency.
Ice Age Histogram. The ice age distribution of sea ice specifies the fractional area
covered by ice in different age classes, within a given region, as a function of time. This
ice age distribution is computed from the field of Lagrangian trajectories described
above. The algorithm for determining ice age works only in the winter: the assumption
is that there is ice growth in all new leads. The process for ice age determination is initialized shortly after fall freeze- up and is operated till the onset of melt. For each observation of a cell, we update the ice age histogram. The resolution of age is dependent on
the sampling interval of the area of interest. As an example, for a series of images with,
say, 7 days between successive images, the age classes would be: 0-7 days, 7-14 days,
14-21 days, 21-28 days, ... , first-year ice, multiyear ice, and ridged first-year ice. We note
that the ice age categories for a given age histogram do not necessarily have uniform
age ranges because the sampling intervals can be nonuniform. The multiyear ice fraction is obtained using an ice classification algorithm. At each step, the accumulated
freezing-degree days associated with each age class are also recorded. The surface air
temperatures for computing the freezing -degree days are extracted from analyzed temperature fields (described below). There are three ice age histogram products. The first
ice age histogram product contains records oflocal ice age distributions and the accumulated freezing-degree days of each age class within each Lagrangian cell. The interpolated ice age histogram product provides uniform age categories (3-day) rather than
the nonuniform age categories in the previous product. The last product is a gridded
(50-km) product which summarizes the ice age histograms on an Earth fixed grid.
Ice Thickness Histogram. The ice thickness distribution specifies the fractional area
covered by ice in given thickness ranges within a region as a function of time. The ice
thickness distribution is estimated from the ice age distribution using an empirical relationship between freezing-degree days and ice thickness as described above. The highest resolution products provide the ice thickness histograms of the Lagrangian cells.
Again, a 50-km gridded product of the ice thickness histograms is produced from the
cell observations.
Open Water Fraction. Summer ice conditions are characterized by an open water fraction product. The summer open water fraction in a Lagrangian cell is estimated from
kinematics and backscatter data. Kinematics gives the expected area change relative to
a previous observation and the backscatter distribution provides an estimate of the areal coverage of open water in a cell. We plan to use both measurements to obtain a better estimate of the open water fraction.
Backscatter Histogram of Lagrangian Cells. The backscatter histograms of the
Lagrangian cells are recorded in this product. The backscatter histogram for each cell
253
surements of the 'ice particles' that are located on the initial grid which covers a region
or the entire Arctic Ocean. The sampling interval is determined by repeat coverage of
these points by SAR imagery. With an initial grid spacing of 5 km, there are approximately 500000 points in this array. Over a winter season of approximately 8 months,
there are close to 32 observations of each trajectory. The gridded 50-km Eulerian product represents weekly ice motion which is spatially and temporally interpolated to the
grid locations from these trajectories. We also produce routine Eulerian observations
of ice motion from image pairs for locations like the Greenland Sea and Bering Strait,
where the character of the ice motion is such that the deformation does not allow us to
maintain the density of points to define a Lagrangian grid. In these special regions, we
select a high-repeat coverage frequency.
Ice Age Histogram. The ice age distribution of sea ice specifies the fractional area
covered by ice in different age classes, within a given region, as a function of time. This
ice age distribution is computed from the field of Lagrangian trajectories described
above. The algorithm for determining ice age works only in the winter: the assumption
is that there is ice growth in all new leads. The process for ice age determination is initialized shortly after fall freeze- up and is operated till the onset of melt. For each observation of a cell, we update the ice age histogram. The resolution of age is dependent on
the sampling interval of the area of interest. As an example, for a series of images with,
say, 7 days between successive images, the age classes would be: 0-7 days, 7-14 days,
14-21 days, 21-28 days, ... , first-year ice, multiyear ice, and ridged first-year ice. We note
that the ice age categories for a given age histogram do not necessarily have uniform
age ranges because the sampling intervals can be nonuniform. The multiyear ice fraction is obtained using an ice classification algorithm. At each step, the accumulated
freezing-degree days associated with each age class are also recorded. The surface air
temperatures for computing the freezing -degree days are extracted from analyzed temperature fields (described below). There are three ice age histogram products. The first
ice age histogram product contains records oflocal ice age distributions and the accumulated freezing-degree days of each age class within each Lagrangian cell. The interpolated ice age histogram product provides uniform age categories (3-day) rather than
the nonuniform age categories in the previous product. The last product is a gridded
(50-km) product which summarizes the ice age histograms on an Earth fixed grid.
Ice Thickness Histogram. The ice thickness distribution specifies the fractional area
covered by ice in given thickness ranges within a region as a function of time. The ice
thickness distribution is estimated from the ice age distribution using an empirical relationship between freezing-degree days and ice thickness as described above. The highest resolution products provide the ice thickness histograms of the Lagrangian cells.
Again, a 50-km gridded product of the ice thickness histograms is produced from the
cell observations.
Open Water Fraction. Summer ice conditions are characterized by an open water fraction product. The summer open water fraction in a Lagrangian cell is estimated from
kinematics and backscatter data. Kinematics gives the expected area change relative to
a previous observation and the backscatter distribution provides an estimate of the areal coverage of open water in a cell. We plan to use both measurements to obtain a better estimate of the open water fraction.
Backscatter Histogram of Lagrangian Cells. The backscatter histograms of the
Lagrangian cells are recorded in this product. The backscatter histogram for each cell