7 Mapping the Progression of Melt Onset and Freeze-Up
131
onset signature change is clear. For freeze-up, Winebrenner et al. (1996) combined a
wide area SAR study in the Beaufort and Chukchi Seas with a buoy temperature/SAR
study in the Beaufort and concluded that temporally erratic but generally low summer
backscattering cross sections stabilize at winter multiyear ice values (roughly -9 dB)
within 7 days of continuously subfreezing air temperatures. Schwartz et al. (1994)
observed spring and autumn transitions in the Beaufort and Chukchi Seas, but compared with analyzed temperature fields based primarily on shore-based data, and
reached similar conclusions.
In an effort to reap the advantages of SAR while mitigating the costs, Winebrenner
et al. (1994; 1996) proposed and demonstrated machine-automatable algorithms for retrospective estimation of melt onset and freeze-up dates within geographical cells using
ERS-1 SAR imagery. This approach utilizes many low-resolution (100-m) SAR images
but immediately reduces the information volume to time series of back scattering cross
section histograms, one for each geographic cell. This avoids any need to monitor
backscattering cross section histories for individual ice floes.
More or less simultaneously with these developments, several investigators found that
characteristic seasonal backscattering changes could also be identified for several terrestrial targets in addition to sea ice. An illustrative, rather than exhaustive, list of examples includes the following. Smith et al. (1997) found that ERS-1 SAR backscatter observations of alpine glaciers show clearly the onset and progression of spring melt. On the
Greenland ice sheet, Drinkwater and Long (1994) found that 14.6 GHz backscattering
observed using the Seasat wind scatterometer not only delineated snow zones in winter but showed clearly the progression and extent of the melt season at all altitudes.
Early et al. (1994) found evidence of temporal variation in snow facie boundaries
between 1978 and 1991 by comparing scatterometer observations from ERS-1 with those
from Seasat. Wismann et al. (1996) found continuous temporal variation from 1991 to
1995 in 5.3-GHz ERS-1 scatterometer observations of Greenland, and found valuable seasonal information in scatterometer observations of Siberian tundra, the Sahel, and
tropical and boreal forests as well. Thus, the observations of seasonal transitions on sea
ice may be seen within a wider geophysical and remote sensing context, and methods
developed in connection with sea ice may prove useful in developing automated processing methods for other geophysical targets and variables.
In this chapter, we begin with a review of the methods and results of Winebrenner et
al. (1994,1996) for automatable mapping of seasonal transition dates and melt season
length on Arctic sea ice. This review includes neither new data nor results, but we focus
particularly on the problems and essential features of machine automation that contribute to a useful geophysical data product. In the following section, we present new
results based on Seasat wind scatterometer observations and the resolution enhancement algorithm of Long et al. (1993) for the freeze-up period in the Beaufort and
Chukchi Seas during 1978. We briefly compare these results with those obtained by
Drinkwater et al. (1994) for seasonal transitions on Antarctic sea ice. The observations
clearly show the observability of freeze-up and its temporal progression in 14.6-GHz
observations, as well as the applicability of Long's resolution enhancement method to
moving targets such as sea ice during a period of temporal change. Thus it appears that
scatterometry data may usefully serve to set the context for, augment, or in some cases substitute for SAR observations even for targets such as sea ice that vary strongly on
small horizontal scales. Insight gained in the development of methods for SAR and scat-
131
onset signature change is clear. For freeze-up, Winebrenner et al. (1996) combined a
wide area SAR study in the Beaufort and Chukchi Seas with a buoy temperature/SAR
study in the Beaufort and concluded that temporally erratic but generally low summer
backscattering cross sections stabilize at winter multiyear ice values (roughly -9 dB)
within 7 days of continuously subfreezing air temperatures. Schwartz et al. (1994)
observed spring and autumn transitions in the Beaufort and Chukchi Seas, but compared with analyzed temperature fields based primarily on shore-based data, and
reached similar conclusions.
In an effort to reap the advantages of SAR while mitigating the costs, Winebrenner
et al. (1994; 1996) proposed and demonstrated machine-automatable algorithms for retrospective estimation of melt onset and freeze-up dates within geographical cells using
ERS-1 SAR imagery. This approach utilizes many low-resolution (100-m) SAR images
but immediately reduces the information volume to time series of back scattering cross
section histograms, one for each geographic cell. This avoids any need to monitor
backscattering cross section histories for individual ice floes.
More or less simultaneously with these developments, several investigators found that
characteristic seasonal backscattering changes could also be identified for several terrestrial targets in addition to sea ice. An illustrative, rather than exhaustive, list of examples includes the following. Smith et al. (1997) found that ERS-1 SAR backscatter observations of alpine glaciers show clearly the onset and progression of spring melt. On the
Greenland ice sheet, Drinkwater and Long (1994) found that 14.6 GHz backscattering
observed using the Seasat wind scatterometer not only delineated snow zones in winter but showed clearly the progression and extent of the melt season at all altitudes.
Early et al. (1994) found evidence of temporal variation in snow facie boundaries
between 1978 and 1991 by comparing scatterometer observations from ERS-1 with those
from Seasat. Wismann et al. (1996) found continuous temporal variation from 1991 to
1995 in 5.3-GHz ERS-1 scatterometer observations of Greenland, and found valuable seasonal information in scatterometer observations of Siberian tundra, the Sahel, and
tropical and boreal forests as well. Thus, the observations of seasonal transitions on sea
ice may be seen within a wider geophysical and remote sensing context, and methods
developed in connection with sea ice may prove useful in developing automated processing methods for other geophysical targets and variables.
In this chapter, we begin with a review of the methods and results of Winebrenner et
al. (1994,1996) for automatable mapping of seasonal transition dates and melt season
length on Arctic sea ice. This review includes neither new data nor results, but we focus
particularly on the problems and essential features of machine automation that contribute to a useful geophysical data product. In the following section, we present new
results based on Seasat wind scatterometer observations and the resolution enhancement algorithm of Long et al. (1993) for the freeze-up period in the Beaufort and
Chukchi Seas during 1978. We briefly compare these results with those obtained by
Drinkwater et al. (1994) for seasonal transitions on Antarctic sea ice. The observations
clearly show the observability of freeze-up and its temporal progression in 14.6-GHz
observations, as well as the applicability of Long's resolution enhancement method to
moving targets such as sea ice during a period of temporal change. Thus it appears that
scatterometry data may usefully serve to set the context for, augment, or in some cases substitute for SAR observations even for targets such as sea ice that vary strongly on
small horizontal scales. Insight gained in the development of methods for SAR and scat-