Fig.13. Phenomenological evolution of multiyear ice aO and
aO over the four SAR scattering
seasons proposed by Livingstone et al. 1987. (Adapted from
Thomas 1996)
Albedo
D.G. BARBER, A. THOMAS, AND T.N. PAPAKYRIAKOU
Winter
Earl Melt Advanced
Melt Onset Melt
................. . .
Multi-Year Ice 0"
o"-a model period
ice. Change detection methods capture this seasonal change in ice type scattering in a
consistent and definable fashion (Sect. 3.4.2). It is also evident that the stability of multiyear sea ice scattering in the winter season, followed by a consistent reduction over
the early melt season, provides a vehicle by which climatological albedo may be estimated using microwave scattering of the surface. The conceptual relationship between
surface albedo and the time series evolution of microwave scattering over multiyear
sea ice (Fig. 13) illustrates the general relationship which will be exploited here.
3.4.3.1
Methods
Statistical relationships between shortwave and microwave interactions were explored
using regression analysis between multiyear ice 0'0 and albedo. Results from detailed
analyses (Thomas 1996) indicated that the time period spanning the end of the winter
season into the early melt and melt onset periods would be optimal for estimating surface albedo. Due to the limited number of paired data points (multiyear ice cro with coincident albedo measurement) for each individual SIMMS year, ERS-1 and multiyear ice
albedo data from SIMMS'92-'95 were combined, not only to increase sample size, but
also to make the model representative of inter-annual variations in multiyear ice types.
Both multiyear ice 0'0 and ~ao were used as independent variables. ~ao was calculated
relative to the winter mean multiyear ice 0'0 for each SIMMS year. This resulted in a
series of ~ao values which become more negative the more temporally removed they
are from the winter mean. A statistical model was constructed to relate ~ao to albedo
over multiyear ice surfaces. Confidence intervals were computed at the 90% level. Computation of the confidence intervals [Eq. (4) 1 was integral to computing statistically distinguishable albedo classes based on the modelled relationships:
aO over the four SAR scattering
seasons proposed by Livingstone et al. 1987. (Adapted from
Thomas 1996)
Albedo
D.G. BARBER, A. THOMAS, AND T.N. PAPAKYRIAKOU
Winter
Earl Melt Advanced
Melt Onset Melt
................. . .
Multi-Year Ice 0"
o"-a model period
ice. Change detection methods capture this seasonal change in ice type scattering in a
consistent and definable fashion (Sect. 3.4.2). It is also evident that the stability of multiyear sea ice scattering in the winter season, followed by a consistent reduction over
the early melt season, provides a vehicle by which climatological albedo may be estimated using microwave scattering of the surface. The conceptual relationship between
surface albedo and the time series evolution of microwave scattering over multiyear
sea ice (Fig. 13) illustrates the general relationship which will be exploited here.
3.4.3.1
Methods
Statistical relationships between shortwave and microwave interactions were explored
using regression analysis between multiyear ice 0'0 and albedo. Results from detailed
analyses (Thomas 1996) indicated that the time period spanning the end of the winter
season into the early melt and melt onset periods would be optimal for estimating surface albedo. Due to the limited number of paired data points (multiyear ice cro with coincident albedo measurement) for each individual SIMMS year, ERS-1 and multiyear ice
albedo data from SIMMS'92-'95 were combined, not only to increase sample size, but
also to make the model representative of inter-annual variations in multiyear ice types.
Both multiyear ice 0'0 and ~ao were used as independent variables. ~ao was calculated
relative to the winter mean multiyear ice 0'0 for each SIMMS year. This resulted in a
series of ~ao values which become more negative the more temporally removed they
are from the winter mean. A statistical model was constructed to relate ~ao to albedo
over multiyear ice surfaces. Confidence intervals were computed at the 90% level. Computation of the confidence intervals [Eq. (4) 1 was integral to computing statistically distinguishable albedo classes based on the modelled relationships: