50
D.G. BARBER, A. THOMAS, AND T.N. PAPAKYRIAKOU
Results from the multivariate analyses over three different years of the SIMMS experiment (1992,1993 and 1994) show that a strong statistical relationship exists between the K*
variable and temperatures within the basal snow layer of both first -year and multiyear sea
ice types. In general we found that over both multiyear and first-year ice types K* explains
a statistically significant portion of the seasonal evolution in cr" when the multivariate model includes other components of the energy balance Eq. (2). During the seasonal transition
from winter to advanced melt period, as K* goes up so does 0'0. The shortwave flux (K*)
was found to have a statistically significant W in the presence of all pairwise and the single three-variable set with Q* and Ts. In the presence of L* the W for K* was not significantly different from zero in any of the pairwise or three-way combinations of variables.
Variability amongst the time steps (Table 1) indicates that the relationship between
0'0 and K* is not highly dependent on the time of day (i.e. the relationship is sufficiently
"seasonal" that all time steps produced significant W coefficients for K*). Multivariate
testing shows that K* also explains a significant portion of the observed variability in
multiyear ice forms. This relationship is, however, limited to the downward trend in the
multiyear ice scattering coefficient within the early melt and melt onset stages (Fig. 3),
and the relationship is inverse after the multiyear ice signature rebounds. The same statistical relationships for the basal layer of the snow volume, within a multivariate model, also explained a statistically significant portion of the observed variability in firstyear and multiyear microwave scattering. Further details on these statistical linkages
are available elsewhere (Barber et al. 1994, 1995; Thomas 1996).
In this section we have shown that there are strong statistical relationships between
various components of the energy balance and the time series evolution of the
microwave scattering coefficient. The univariate analyses were used to investigate individual components of the energy balance effects on the time series evolution of 0'0. We
also showed that many of the relationships can be explained using a physical scattering model and knowledge of the physical and electrical properties of the evolving
marine cryosphere. The multivariate analyses showed that K* and the basal layer snow
temperatures over first -year and multiyear sea ice explain a statistically significant portion of the observed variability in the temporal evolution of 0'0. Analysis of this relationship is consistent over 3 years of empirical evidence (SIMMS'92-94) and the physical causes of these relationships have been explained using the physical and electrical
properties of the snow-covered sea ice.
3.4
Generating Image Products
In this section we present research results which exploit the physical and statistical relationships described in the preceding section. We describe how synthetic aperture radar
can be used to detect the onset of melt within the marine cryosphere. We illustrate some
ongoing work in the field of image segmentation and classification using the time series
patterns of the scattering coefficient. We then extend the segmentation work into the
broad field of change detection and describe how change images may provide insights
into subtle changes in the energy balance of the sea ice surface throughout the annual
cycle. In the final section we describe how the temporal evolution of 0'0 can be used to
create climatological albedo classes of the marine cryosphere as the season progresses from winter through to the advanced melt period.
D.G. BARBER, A. THOMAS, AND T.N. PAPAKYRIAKOU
Results from the multivariate analyses over three different years of the SIMMS experiment (1992,1993 and 1994) show that a strong statistical relationship exists between the K*
variable and temperatures within the basal snow layer of both first -year and multiyear sea
ice types. In general we found that over both multiyear and first-year ice types K* explains
a statistically significant portion of the seasonal evolution in cr" when the multivariate model includes other components of the energy balance Eq. (2). During the seasonal transition
from winter to advanced melt period, as K* goes up so does 0'0. The shortwave flux (K*)
was found to have a statistically significant W in the presence of all pairwise and the single three-variable set with Q* and Ts. In the presence of L* the W for K* was not significantly different from zero in any of the pairwise or three-way combinations of variables.
Variability amongst the time steps (Table 1) indicates that the relationship between
0'0 and K* is not highly dependent on the time of day (i.e. the relationship is sufficiently
"seasonal" that all time steps produced significant W coefficients for K*). Multivariate
testing shows that K* also explains a significant portion of the observed variability in
multiyear ice forms. This relationship is, however, limited to the downward trend in the
multiyear ice scattering coefficient within the early melt and melt onset stages (Fig. 3),
and the relationship is inverse after the multiyear ice signature rebounds. The same statistical relationships for the basal layer of the snow volume, within a multivariate model, also explained a statistically significant portion of the observed variability in firstyear and multiyear microwave scattering. Further details on these statistical linkages
are available elsewhere (Barber et al. 1994, 1995; Thomas 1996).
In this section we have shown that there are strong statistical relationships between
various components of the energy balance and the time series evolution of the
microwave scattering coefficient. The univariate analyses were used to investigate individual components of the energy balance effects on the time series evolution of 0'0. We
also showed that many of the relationships can be explained using a physical scattering model and knowledge of the physical and electrical properties of the evolving
marine cryosphere. The multivariate analyses showed that K* and the basal layer snow
temperatures over first -year and multiyear sea ice explain a statistically significant portion of the observed variability in the temporal evolution of 0'0. Analysis of this relationship is consistent over 3 years of empirical evidence (SIMMS'92-94) and the physical causes of these relationships have been explained using the physical and electrical
properties of the snow-covered sea ice.
3.4
Generating Image Products
In this section we present research results which exploit the physical and statistical relationships described in the preceding section. We describe how synthetic aperture radar
can be used to detect the onset of melt within the marine cryosphere. We illustrate some
ongoing work in the field of image segmentation and classification using the time series
patterns of the scattering coefficient. We then extend the segmentation work into the
broad field of change detection and describe how change images may provide insights
into subtle changes in the energy balance of the sea ice surface throughout the annual
cycle. In the final section we describe how the temporal evolution of 0'0 can be used to
create climatological albedo classes of the marine cryosphere as the season progresses from winter through to the advanced melt period.