54
D.G. BARBER, A. THOMAS, AND T.N. PAPAKYRIAKOU
Fig.11. Results of
the Kolmogorov•
Smirnov non-para.5
metric difference in
.25
•
.,
distribution tests for
•
~;g
all pairwise compar. 12
"'.c .. '"
u·.,
•
'0 :::>
isons of the six ice
.06
., ..
. ",
"3 '~
sites discussed in the
-.; .03
>
c;)~
text. The P-values
d...
•
..!:
indicate the level of
.02
•
significance in a
.01
u
•
•
»:0
hypothesis test con.0039
•
•
=a~
trasting the seasonal
•
u .,
.0019
•
• •
:B ·~
evolution of 0°
.00097
is·§
between the two ice
•
(I) . ~
sites denoted on the
.00048
Cl
" " .... 3: 3: -< 3: 3: 3: -< ~ ~
V>
!Q !Q
x axis ("contrast
-< -< :T -< -;.
3: ~
-< -< 3:
~
~ I< "
T
~
;; ;; ~ ' " ' " ~ ;!
comparisons").
~ ~
V>
-< U,
:i;l :i;l
...
~
~
3:
~ ~
"
~
~
';
(Adapted from Bar!o
~
~
~
Q- ,,'
~ ~ ~
~
.."
~
ber et al.1995)
~
~ ~
~
~
Contrast Comparisons
ly distinguishable, indicating excellent separation of these ice sites based on the seasonal evolution of their microwave scattering coefficient cr". The remainder of the contrast comparisons indicate that the MYI and the YMYI site were all statistically distinguishable from each of the three first-year ice types (with the exception of"YMYI-thin
FYI", which we consider to be marginal with a P-value = 0.023). The second-year ice
(SYI) was not distinguishable from any of the first-year ice types.
These results illustrate the utility of using the time series evolution as additional information in the classification of sea ice types. Saline ice types (first-year) are easily separated from non-saline types (multiyear) because of the effect of atmospheric forcing
on the dielectrics of the saline surface (Barber and Thomas 1997). Separation within
multiyear ice sites is possible because of the stability of the seasonal pattern and magnitude shifts due to the age of the multiyear ice. Further details are available elsewhere
(Barber et al. 1995).
3.4.2.2
Segmentation
The time series relationships illustrated in Sect. 3.2.3 (Fig. 3) can be used to separate
first year from multiyear forms of sea ice. The approach we are investigating is still in
development (Thomas 1996), but the physical mechanisms which allow for this segmentation are well understood (Barber and Thomas 1997). Physical principles, and as
a result, segmentation approaches, are different within the winter season and the period spanning the early melt to advanced melt stages of the transitional period.
Within the winter season segmentation is based on the observation that multiyear
ice signatures are stable and first-year smooth ice signatures are variable. Ice containing brine in the surface layer and/or within the basal layer of the snow cover can be
D.G. BARBER, A. THOMAS, AND T.N. PAPAKYRIAKOU
Fig.11. Results of
the Kolmogorov•
Smirnov non-para.5
metric difference in
.25
•
.,
distribution tests for
•
~;g
all pairwise compar. 12
"'.c .. '"
u·.,
•
'0 :::>
isons of the six ice
.06
., ..
. ",
"3 '~
sites discussed in the
-.; .03
>
c;)~
text. The P-values
d...
•
..!:
indicate the level of
.02
•
significance in a
.01
u
•
•
»:0
hypothesis test con.0039
•
•
=a~
trasting the seasonal
•
u .,
.0019
•
• •
:B ·~
evolution of 0°
.00097
is·§
between the two ice
•
(I) . ~
sites denoted on the
.00048
Cl
" " .... 3: 3: -< 3: 3: 3: -< ~ ~
V>
!Q !Q
x axis ("contrast
-< -< :T -< -;.
3: ~
-< -< 3:
~
~ I< "
T
~
;; ;; ~ ' " ' " ~ ;!
comparisons").
~ ~
V>
-< U,
:i;l :i;l
...
~
~
3:
~ ~
"
~
~
';
(Adapted from Bar!o
~
~
~
Q- ,,'
~ ~ ~
~
.."
~
ber et al.1995)
~
~ ~
~
~
Contrast Comparisons
ly distinguishable, indicating excellent separation of these ice sites based on the seasonal evolution of their microwave scattering coefficient cr". The remainder of the contrast comparisons indicate that the MYI and the YMYI site were all statistically distinguishable from each of the three first-year ice types (with the exception of"YMYI-thin
FYI", which we consider to be marginal with a P-value = 0.023). The second-year ice
(SYI) was not distinguishable from any of the first-year ice types.
These results illustrate the utility of using the time series evolution as additional information in the classification of sea ice types. Saline ice types (first-year) are easily separated from non-saline types (multiyear) because of the effect of atmospheric forcing
on the dielectrics of the saline surface (Barber and Thomas 1997). Separation within
multiyear ice sites is possible because of the stability of the seasonal pattern and magnitude shifts due to the age of the multiyear ice. Further details are available elsewhere
(Barber et al. 1995).
3.4.2.2
Segmentation
The time series relationships illustrated in Sect. 3.2.3 (Fig. 3) can be used to separate
first year from multiyear forms of sea ice. The approach we are investigating is still in
development (Thomas 1996), but the physical mechanisms which allow for this segmentation are well understood (Barber and Thomas 1997). Physical principles, and as
a result, segmentation approaches, are different within the winter season and the period spanning the early melt to advanced melt stages of the transitional period.
Within the winter season segmentation is based on the observation that multiyear
ice signatures are stable and first-year smooth ice signatures are variable. Ice containing brine in the surface layer and/or within the basal layer of the snow cover can be