2.0
peak, and rawpeak records, respectively. Given
the various measurement errors, these compari-
1.6
sons suggest fundamental agreement of the
methods. However, additional development of
1.2
the peak return data reduction method is needed
because of the significant underprediction of the
0.8
mean velocity.
0.4
Ice velocities exceeding 5 m/s during breakup
Video Data
have been reported on some rivers. The maxi-
0
with a given system is limited by the Nyquist
0.04
sampling rate, fd ≤ N/2. With our system, veloci-
ties of up to 6 m/s can be obtained with Doppler
frequencies less than 200 Hz, indicating that the
0
Nyquist rate does not restrict velocity measure-
ment at the high end of the observed range. Un-
0.04
like optical methods, the large changes in river
stage that can occur during breakup do not affect
Doppler radar velocity measurement.
0.08
0
10
20
30
Time (min)
Frazil ice run--December 1993
Figure 10. Ice velocity data, best polynomial fit,
Doppler radar ice velocity and simultaneous
and normalized difference between the data and
video ice velocity records were obtained for the
polynomial of the video ice breakup record.
Connecticut River at the CornishWindsor
bridge on 28 December 1993. Frazil pans and
floes were moving downstream during the
2.5
measurement period at a river flow of about 170
2.0
m3/s, just prior to ice cover formation. Over sev-
eral hours of data collection, the river flow and
1.5
ice velocity conditions were relatively steady,
and we selected a typical 1000-s record for analy-
1.0
sis. The video setup and grid were the same as in
Video Data
0.5
Peak Data
the previous event. The radar antenna was
Raw Data
mounted at the same height and location as be-
0
fore, about 11 m above the lower water surface.
0
10
20
30
Time (min)
To reduce the data processing requirements of
the method, an antenna with a much narrower
Figure 11. Comparison of ice velocity results
for breakup after multiplying the video and
beam width was used. With slower ice motion
peak radar results by a constant that equates
the mean velocity of each record.
creased to improve the velocity resolution. Using
Almost identical correlation coefficients were
solve velocity differences of 0.05 m/s is only 1.5
obtained for all pairs of the raw radar, peak radar,
and polynomial video velocity records, ranging
sponding frequency shift increases to 9 Hz with a
between 0.933 and 0.950. The ice velocitytime
30-GHz millimeter-wave source. A millimeter-
traces obtained with each method are superposed
wave source provides sufficient resolution for ice
velocity measurement at speeds above about 0.1
return results by 1.191 and the video polynomial
m/s. The modified radar system specifications
results by 1.067 to correct for the mean offset from
are given in Table 1.
the raw radar results. The root-mean-square
The video velocity data, best fifth-order poly-
(RMS) difference obtained between any pair of
nomial fit, and difference between the data and
records was less than 7.7 cm/s, corresponding to
the polynomial normalized by the polynomial
dimensionless RMS velocity differences of 0.033,
velocity are given in Figure 12. The mean video
0.037, and 0.048 between the videoraw, video
velocity was 0.518 m/s, and the normalized dif-
9