220
C. BERTOIA, J. FALKINGHAM, F. FETTERER
10.5.3
Synergistic Use with Other Data Types or Information
Ice type is more accurately determined when viewed at more than one wavelength in
the electromagnetic spectrum. More information is also gleaned from a time series than
from a single observation. For this reason, operational ice centers attempt to combine
data from various sources prior to labeling an ice map. There are two possible
approaches to data fusion:
1. Analyze the visible, infrared, in situ observations, passive microwave, and SAR data
separately, then mesh the geophysical labels.
2. View all sources concurrently, making decisions about ice characteristics with all
sources on screen.
Despite the digital workstation environment present in today's ice centers, some data
remain available only on paper (usually in situ observations from ships and aerial ice
reconnaissance), requiring the centers to implement a combination of these two
approaches.
Ice analysts are trained to start their analyses with the finest-resolution data available. Thus, in situ observations are considered first, followed by interpretation of SAR
imagery and then visible and infrared imagery. Finally, passive microwave is used for
all cloud covered regions not covered by SAR. The temperature and albedo of ice features derived from infrared and visible imagery is used to supplement the topography
and floe shape information provided by SAR, giving analysts a better decision base for
ice classification. For eX\imple, NIC ice analysts used a combination of DMSP OLS and
ERS-l SAR to characterize ice in the Beaufort Sea (Fig. 9).
The first challenge of this new digital environment at most ice centers is to map all
satellite data to the same projection. This is accomplished by the ingest functions of
the ice workstations. Rather than map all data to the same resolution, as would be
required for automated data fusion techniques, the ice centers have chosen to map
imagery to the optimal resolution for ice interpretation (in the case of SAR) or the
maximum sensor resolution (in the case of visible and infrared imagery). The next
step in the analysis process is to compensate for the temporal differences in images
which have been acquired since the last analysis period, from many different sensors.
This is accomplished using a combination of ice drift forecast/hindcast models and
surface wind and temperature analyses. Time series analysis has proven extremely
valuable in discriminating between clouds and ice (in the case of visible and infrared
imagery) and between ice and windy open water (in the case of SAR imagery). The
final step includes labeling ice features, compensating for resolution and geolocation
accuracies for each instrument. At all stages, of course, ancillary information such as
season, location, and oceanographic and meteorological conditions is incorporated
in the ice analysis.
C. BERTOIA, J. FALKINGHAM, F. FETTERER
10.5.3
Synergistic Use with Other Data Types or Information
Ice type is more accurately determined when viewed at more than one wavelength in
the electromagnetic spectrum. More information is also gleaned from a time series than
from a single observation. For this reason, operational ice centers attempt to combine
data from various sources prior to labeling an ice map. There are two possible
approaches to data fusion:
1. Analyze the visible, infrared, in situ observations, passive microwave, and SAR data
separately, then mesh the geophysical labels.
2. View all sources concurrently, making decisions about ice characteristics with all
sources on screen.
Despite the digital workstation environment present in today's ice centers, some data
remain available only on paper (usually in situ observations from ships and aerial ice
reconnaissance), requiring the centers to implement a combination of these two
approaches.
Ice analysts are trained to start their analyses with the finest-resolution data available. Thus, in situ observations are considered first, followed by interpretation of SAR
imagery and then visible and infrared imagery. Finally, passive microwave is used for
all cloud covered regions not covered by SAR. The temperature and albedo of ice features derived from infrared and visible imagery is used to supplement the topography
and floe shape information provided by SAR, giving analysts a better decision base for
ice classification. For eX\imple, NIC ice analysts used a combination of DMSP OLS and
ERS-l SAR to characterize ice in the Beaufort Sea (Fig. 9).
The first challenge of this new digital environment at most ice centers is to map all
satellite data to the same projection. This is accomplished by the ingest functions of
the ice workstations. Rather than map all data to the same resolution, as would be
required for automated data fusion techniques, the ice centers have chosen to map
imagery to the optimal resolution for ice interpretation (in the case of SAR) or the
maximum sensor resolution (in the case of visible and infrared imagery). The next
step in the analysis process is to compensate for the temporal differences in images
which have been acquired since the last analysis period, from many different sensors.
This is accomplished using a combination of ice drift forecast/hindcast models and
surface wind and temperature analyses. Time series analysis has proven extremely
valuable in discriminating between clouds and ice (in the case of visible and infrared
imagery) and between ice and windy open water (in the case of SAR imagery). The
final step includes labeling ice features, compensating for resolution and geolocation
accuracies for each instrument. At all stages, of course, ancillary information such as
season, location, and oceanographic and meteorological conditions is incorporated
in the ice analysis.