3 Role of SAR in Surface Energy Flux Measurements Over Sea Ice
53
Fig.l0. Seasonal
evolution ofthe mul-5.0
tiyear (MYI), young
ERS-Io·
SlMMS'93 Period
multiyear (YMYI)
---2000LAT 0
and the second-year
0
..!?,
-7.5
1200 LAT .
ice (SYI) total relac
tive scattering coeffi0
'.c
dent aO for the periu
..,
CI:l
od day 50-190. The
'" -10.0
'"
SIMMS'93 field perie
od and the approxiU
c./l
mate overpass times
.§
for the ERS-1 SAR
.., -12.5
are indicated. Solid
~
lines are a spline fit.
CI:l
..,
(Adapted from Barc./l - 15.0
ber et a1.1995)
E ..,
>
«
0;
~ -17.5
-20.0
50
70
90
110
130
150
170
190
Julian Day
had the thickest hummocks with the largest pore spaces of the three forms of multiyear ice. This increases the volume scattering component of 0"0, thereby causing a higher average 0"0 for the MYI relative to the YMYI and SYI sites. The reduction in scattering after day 146 corresponds to the increase in water in liquid phase to detectable levels (i.e. 1-3%). The 0"0 continued to decrease throughout both the pendular and funicular regimes (i.e. up until day 167). Once the snow cover was completely ablated to a
melt pond cover the scattering increased (days 167-176). Draining of the surface melt
ponds after day 176 corresponded to a decrease in 0"0.
To evaluate the utility of the time series pattern in the classification of these six sea
ice types we subjected the seasonal evolution patterns to a Kolmogorov-Smirnov test.
This test ascertains the probability that the two distributions being compared arise from
the same parent distribution (i.e. they are statistically indistinguishable). The test
assumes a continuous distribution, sampled in an independent, unbiased fashion. The
test is sensitive to both magnitude and shape of the distribution (Sokal and Rohlf 1981),
making it a good metric for our stated objectives.
Results of this analysis indicate that considerable variability exists between the pairwise contrasts of the seasonal evolution of 0"0 for the six ice sites tested (Fig. 11). Pairwise comparisons of the first-year ice sites amongst themselves (i.e. "FYI-thick FYI",
"FYI-thin FYI", "thin FYI-thick FYI" contrast) is statistically distinguishable (P-value
= 0.0039). The contrast "FYI-thin FYI" is marginally separable (P-value = 0.07) and the
"FYI-thin FYI" contrast is statistically indistinguishable (P-value = 0.35). Pairwise
comparisons of the multiyear ice sites between themselves (i.e. "MYI-SYI",
"MYI-YMYI", "YMYI-SYI") shows all pairwise contrast comparisons to be statistical-
53
Fig.l0. Seasonal
evolution ofthe mul-5.0
tiyear (MYI), young
ERS-Io·
SlMMS'93 Period
multiyear (YMYI)
---2000LAT 0
and the second-year
0
..!?,
-7.5
1200 LAT .
ice (SYI) total relac
tive scattering coeffi0
'.c
dent aO for the periu
..,
CI:l
od day 50-190. The
'" -10.0
'"
SIMMS'93 field perie
od and the approxiU
c./l
mate overpass times
.§
for the ERS-1 SAR
.., -12.5
are indicated. Solid
~
lines are a spline fit.
CI:l
..,
(Adapted from Barc./l - 15.0
ber et a1.1995)
E ..,
>
«
0;
~ -17.5
-20.0
50
70
90
110
130
150
170
190
Julian Day
had the thickest hummocks with the largest pore spaces of the three forms of multiyear ice. This increases the volume scattering component of 0"0, thereby causing a higher average 0"0 for the MYI relative to the YMYI and SYI sites. The reduction in scattering after day 146 corresponds to the increase in water in liquid phase to detectable levels (i.e. 1-3%). The 0"0 continued to decrease throughout both the pendular and funicular regimes (i.e. up until day 167). Once the snow cover was completely ablated to a
melt pond cover the scattering increased (days 167-176). Draining of the surface melt
ponds after day 176 corresponded to a decrease in 0"0.
To evaluate the utility of the time series pattern in the classification of these six sea
ice types we subjected the seasonal evolution patterns to a Kolmogorov-Smirnov test.
This test ascertains the probability that the two distributions being compared arise from
the same parent distribution (i.e. they are statistically indistinguishable). The test
assumes a continuous distribution, sampled in an independent, unbiased fashion. The
test is sensitive to both magnitude and shape of the distribution (Sokal and Rohlf 1981),
making it a good metric for our stated objectives.
Results of this analysis indicate that considerable variability exists between the pairwise contrasts of the seasonal evolution of 0"0 for the six ice sites tested (Fig. 11). Pairwise comparisons of the first-year ice sites amongst themselves (i.e. "FYI-thick FYI",
"FYI-thin FYI", "thin FYI-thick FYI" contrast) is statistically distinguishable (P-value
= 0.0039). The contrast "FYI-thin FYI" is marginally separable (P-value = 0.07) and the
"FYI-thin FYI" contrast is statistically indistinguishable (P-value = 0.35). Pairwise
comparisons of the multiyear ice sites between themselves (i.e. "MYI-SYI",
"MYI-YMYI", "YMYI-SYI") shows all pairwise contrast comparisons to be statistical-