112
A. K. LIU AND C. Y. PENG
ing. In the Bering Sea, fish abundance is found to be highly correlated with yearly ice
extent. Fish select cold pools after ice retreat for survival as the food web descends
toward the ocean bottom. The cold pools depend on the yearly ice extent.
High-resolution SAR images are very useful for tracking the ice edge motion. A sequence
of seven SAR images of the MIl in the Chukchi Sea from September 27 to October 18,1991.
with 3-day intervals, have been investigated for ice edge advance/retreat by Liu et al.
(1994b). Simultaneous current measurements from the northeast Chukchi Sea, as well as
Barrow wind records, are used to interpret the MIl dynamics. On the basis of ice edge locations determined from the SAR images, the ice edge motion appears to be highly dynamic with hundreds of kilometers advancelretreat in 3 days. When the ice edges are relatively stationary, the formation of mesoscale eddy near the meandering ice edge is evident.
A new method for time-varying signal analysis, the wavelet transform, has been
developed for remote sensing applications at NASA/GSFC (Goddard Space Flight Center) during the past 3 years and provides spectral decompositions via the scale concept.
Basically, wavelet transforms are analogous to Fourier transforms, but are localized in
both frequency and time (e.g. Combes et al. 1989). Recent investigations in physical
oceanography (Ruskai et al.1992) verify the efficiency and abilities of these transforms
to analyze nonlinear dynamical ocean systems. Wavelet analysis of wind fluctuations
over wave groups has been reported by Liu et al. (1995) and Peng et al. (1995).
The two-dimensional wavelet transform is a highly efficient band-pass filter, which
can be used to separate various scale processes and show their relative phase/location
information,as done,e.g.,in SAR imagerybyLiu and Peng (1993b) and Liu et al. (1997).
The two-dimensional Gaussian wavelet (often referred to as Mexican-hat or Laplacian
of Gaussian) transform of a SAR image for small-scale features can be used with a
threshold as an edge detector (Canny 1986). The SAR scene can be wavelet transformed
with various scales to separate various texture or features. The Laplacian of Gaussian
(Mexican-hat) wavelet can also be used as a band-pass filter, and its first derivative as
a threshold based on the histogram of the transformed results. In the MIl study, the
ice edge location and ice floe motion in each SAR image can be delineated by using a
two-dimensional wavelet transform. The evolution of mesoscale features such as
fronts, ice edge meanders, leads, ice floes, surface film, and eddies can be easily tracked
by the wavelet analysis using multitemporal SAR images. The numerical simulation of
a two-dimensional ocean-ice interaction model can be used as a guide for the search
of ice floe, leads, and ice edge in the SAR images during the repeating cycles. This kind
of SAR processing by wavelet transform as described here can be used interactively as
needed and can provide a more cost-effective monitoring program that would keep
track of changes in important elements such as eddies, open leads, ice edge, and ice floe
in coastal polynyas and in the MIl.
The ASF Geophysical Processor System (GPS) has been quite successful in tracking
ice motion in the winter Arctic, but has some difficulties with summer ice, in polynyas,
and in the MIl. The wide-swath ScanSAR mode of RADARSAT will have the capability to image the Arctic sea ice cover every 3 days. The MIl is a critical boundary condition for large-scale climate change study. The assimilation of RADARS AT SAR data with
the ocean-ice interaction model can provide a more cost-effective monitoring program
that would keep track of changes in important processes (such as ice motion) of the
climate system. The advance or retreat of the ice edge, ice floe motion, and mesoscale
processes in polynyas and in the MIl can be tracked by the wavelet analysis using
A. K. LIU AND C. Y. PENG
ing. In the Bering Sea, fish abundance is found to be highly correlated with yearly ice
extent. Fish select cold pools after ice retreat for survival as the food web descends
toward the ocean bottom. The cold pools depend on the yearly ice extent.
High-resolution SAR images are very useful for tracking the ice edge motion. A sequence
of seven SAR images of the MIl in the Chukchi Sea from September 27 to October 18,1991.
with 3-day intervals, have been investigated for ice edge advance/retreat by Liu et al.
(1994b). Simultaneous current measurements from the northeast Chukchi Sea, as well as
Barrow wind records, are used to interpret the MIl dynamics. On the basis of ice edge locations determined from the SAR images, the ice edge motion appears to be highly dynamic with hundreds of kilometers advancelretreat in 3 days. When the ice edges are relatively stationary, the formation of mesoscale eddy near the meandering ice edge is evident.
A new method for time-varying signal analysis, the wavelet transform, has been
developed for remote sensing applications at NASA/GSFC (Goddard Space Flight Center) during the past 3 years and provides spectral decompositions via the scale concept.
Basically, wavelet transforms are analogous to Fourier transforms, but are localized in
both frequency and time (e.g. Combes et al. 1989). Recent investigations in physical
oceanography (Ruskai et al.1992) verify the efficiency and abilities of these transforms
to analyze nonlinear dynamical ocean systems. Wavelet analysis of wind fluctuations
over wave groups has been reported by Liu et al. (1995) and Peng et al. (1995).
The two-dimensional wavelet transform is a highly efficient band-pass filter, which
can be used to separate various scale processes and show their relative phase/location
information,as done,e.g.,in SAR imagerybyLiu and Peng (1993b) and Liu et al. (1997).
The two-dimensional Gaussian wavelet (often referred to as Mexican-hat or Laplacian
of Gaussian) transform of a SAR image for small-scale features can be used with a
threshold as an edge detector (Canny 1986). The SAR scene can be wavelet transformed
with various scales to separate various texture or features. The Laplacian of Gaussian
(Mexican-hat) wavelet can also be used as a band-pass filter, and its first derivative as
a threshold based on the histogram of the transformed results. In the MIl study, the
ice edge location and ice floe motion in each SAR image can be delineated by using a
two-dimensional wavelet transform. The evolution of mesoscale features such as
fronts, ice edge meanders, leads, ice floes, surface film, and eddies can be easily tracked
by the wavelet analysis using multitemporal SAR images. The numerical simulation of
a two-dimensional ocean-ice interaction model can be used as a guide for the search
of ice floe, leads, and ice edge in the SAR images during the repeating cycles. This kind
of SAR processing by wavelet transform as described here can be used interactively as
needed and can provide a more cost-effective monitoring program that would keep
track of changes in important elements such as eddies, open leads, ice edge, and ice floe
in coastal polynyas and in the MIl.
The ASF Geophysical Processor System (GPS) has been quite successful in tracking
ice motion in the winter Arctic, but has some difficulties with summer ice, in polynyas,
and in the MIl. The wide-swath ScanSAR mode of RADARSAT will have the capability to image the Arctic sea ice cover every 3 days. The MIl is a critical boundary condition for large-scale climate change study. The assimilation of RADARS AT SAR data with
the ocean-ice interaction model can provide a more cost-effective monitoring program
that would keep track of changes in important processes (such as ice motion) of the
climate system. The advance or retreat of the ice edge, ice floe motion, and mesoscale
processes in polynyas and in the MIl can be tracked by the wavelet analysis using