Figure 12. Video and radar data, best polynomial fits, and normalized difference between the
data and corresponding polynomial for the frazil ice run.
ference between the data and the polynomial was
generally less than 0.06. The mean and median
difference are both zero, with an RMS difference
of 0.022 m/s and a maximum difference of 0.065
m/s. The amplitudes of the velocity differences
are within the bounds of experimental error, and
their structure is random. We conclude that the
polynomial adequately describes the video data.
was a difference in velocity of 0.05 m/s, or a nor-
malized velocity difference of 0.09. Data repre-
senting the midpoint of the band obtained at 10-s
Figure 13. The part of the radar velocity record
intervals, the best fourth-order polynomial fit to
for breakup with dispersed ice floes in the river.
these data, and the normalized difference be-
Note the movement in time of strong returns from
tween the data and the polynomial are presented
higher to lower velocity, corresponding to the move-
in Figure 12. The mean radar velocity was 0.539
ment of floes through the antenna footprint.
m/s, and the normalized difference was general-
record during the period of dispersed ice motion.
ly less than 0.07. The mean and median differ-
Bands of velocity through time from floes produc-
ences are both zero, with an RMS difference of
ing strong backscatter are clearly visible.
0.023 m/s and a maximum difference of 0.067 m/s.
Small underestimates of α and β would cause a
The amplitudes of the velocity differences are
systematically high radar velocity, corresponding
somewhat larger than would be expected from
to most of the difference between the Doppler
experimental error, and their structure appears
and video mean velocities. In addition, strong in-
periodic. However, the oscillations in this record
are not supported by the video data. The oscillat-
footprint at constant speed spent almost 60% of
ing radar velocities occurred when the open wa-
this time on the high-velocity side of the bore-
ter area in the radar footprint increased and the
sight, potentially introducing a velocity bias.
number of targets decreased. When a single floe
However, for consistency and ease of comparison
traverses the footprint there is an apparent
we again assume that the systematic difference
change in velocity from high to low caused by the
between the radar and video velocities is due to
relationship between vertical angle and Doppler
video grid distortion. The video and radar veloc-
frequency. Figure 13 shows part of the breakup