Geoscience Reference
In-Depth Information
in the control of the ice edge and its location. Numerical models for ice edge
upwelling, in which the ice is allowed to be a dynamic medium (see Refs. 1,
4 and 5) show the effects from wave actions to be quite significant. In recent
years, there has been a massive increase in high-quality ocean/ice data for
research use. These data sources include SMMR, SAR, QuikSCAT, SSM/I,
and ASMR-E. Therefore, it is necessary to develop automated algorithms
to process the data for ocean and ice studies. Gloersen
et al.
6
reported
a frequency analysis of the Arctic sea ice concentration using the SMMR
data. Recently, Wu and Liu
7
developed an automated algorithm for ocean
feature detection, extraction and classification in SAR imagery, using two-
dimensional wavelet analysis. Remote sensing data can also be processed
to create ice motion products. Yu and Liu
8
carried out automated sea ice
texture classifications and motion analysis using SAR imagery. Zhao and
Liu
9
used wavelet analysis of QuikSCAT and SSM/I data to obtain daily
sea ice drift information for the Polar Regions. Here, we extend the study
of Yu and Liu
8
to include sea ice motion and deformation, as well as the
possible relationship to wind and wave effects in the MIZ. In the next sec-
tion, we describe our method and results for sea ice classification, motion
and deformation analyzes. In Sec. 3, we give some conclusions and outlook.
2. Methods and Results
We use SAR imagery to study sea ice motion and deformation in the MIZ.
We focus our study in the Bering Sea area. Figure 1 shows the raw SAR
imagery data taken on December 14, 2001. The spatial resolution is 100 m
×
100 m. The outline of the St. Lawrence Island is visible in the lower-right
quarter of the image, showing the geographic location of the SAR imagery.
To illustrate our method, we focus on the inset in Fig. 1, a small rectangular
area southeast of the St. Lawrence Island. We then choose two consecutive
SAR images of December 11 and 14, 2001 in order to study ice motion and
deformation. These are shown in Figs. 2 and 3, with the eastern tip of the
St. Lawrence Island visible in the upper right corner.
As was pointed out in Yu and Liu,
8
a segmentation technique with
DLT can be used to analyze and segment unstructured sea ice data. Since
a variety of motions and mesoscale activity exist in the MIZ, sea ice features
in SAR imagery vary at different scales for different times and locations in
the MIZ. Therefore, no direct global thresholding method can compensate
for all these variations. DLT is a method of generating global thresholds
through a dynamic local thresholding and we use this method in this study.
