3. Analytical Technique

3.1 Position Determination Method and Positional Accuracy of the Grounding Line

An InSAR image shows a strip of dense fringe between the ice sheet and ice shelves. This strip is sometimes referred to as the "grounding zone"⁶⁾ . When the InSAR image explicitly shows the continental-side boundary of the fringe as in Fig. 2, the grounding line can be identified. Since an InSAR image does not itself have geometric position information, the grounding line position is determined by referring to the position information of the Radarsat Antarctic Mapping Project (RAMP; Jezek and RAMP Product Team, 2002)⁷⁾. In Fig. 2(a), the grounding line is colored pink. The part of the line in the blue frame is enlarged in Fig. 2(b). Fig. 2(c) shows a diagrammatic sketch of phase profile between points A and B in Fig. 2(b). Thus, a grounding line appears as a dense fringe pattern. The line closest to the continent is digitized as the grounding line as shown in Figs. 2(b)-(c). This definition is similar to those of Rignot (1996)⁸⁾ and Ozawa et al. (2002)⁹⁾. When extracting a grounding line, three types of errors should be considered: 1) digitizing error, 2) geometric correction error between the RAMP image and the InSAR image, and 3) positional error of the reference RAMP image. A digitizing accuracy of about ±2 pixels is empirically possible if the magnification factor is sufficient. The digitizing error can thus be regarded as ±100 m since the spatial resolution of InSAR image is 50 m. The geometric correction error is below four pixels, hence the accuracy is ±200 m. For the RAMP reference image, Jezek and RAMP Product Team (2002) have reported the accuracy as ±200 m. Here, a RAMP image refers to a pan-Antarctic mosaic SAR intensity image produced from SAR images shot by the RADARSAT-1 satellite. This image has orthogonal-coordinate position data with a polar stereographic projection developed on a WGS84 Ellipsoid. Therefore, the spatial positional error σ of an extracted grounding line is calculated by a root-some-square (RSS) according to the following formula:

Therefore, the spatial positional accuracy of an acquired grounding line is theoretically ±300 m.

Fig.2 Definition of grounding line by InSAR analysis

3.2 Extraction of the Grounding Line in an Ice Sheet Area

Fig. 3 shows the grounding line extraction flowchart. Firstly, the SAR raw data of ERS-1 or ERS-2 is reproduced (1). The reproduced image is a complex image called a Single Look Complex (SLC) that contains phase information. In addition, a microwave backscatter intensity image (hereinafter, referred to as an SAR intensity image) (1) is also obtained.
Interference processing using an ERS-1 and an ERS-2 SLC image is executed in order to acquire an interferogram (2). This interferogram is a complex-number image, containing phase difference information, with the same pixel-line position (the spatial positions of pixels) as the corresponding SAR intensity image. An interferogram, however, cannot be used directly for geometric correction because no ground control points (GCPs) information from terrain or bare rock can be acquired.
A geographical coordinate system is necessary to compare an SAR interferogram with an ADD or to give a spatially correct grounding line position. Therefore, corresponding Ground Control Points (GCPs) between the RAMP and SAR intensity images are acquired for geometric correction. From these, a geometric correction factor (3) is calculated from them and superimposed on a RAMP image.
The interferogram is geometrically corrected (4) using the calculated geometric correction factor. This geometric correction produces an interferogram with coordinate information, with a pixel size of 50 m.
A grounding line is extracted manually by visual interpretation using Geographic Information System (GIS) software (5) using the final interferogram. The acquired grounding line information is saved as a Shape file, which is a vector format. The grounding line is then superimposed on a RAMP image with a grounding line from an ADD image and an interferogram on a single layer for comparative interpretation (6).
SAR image reconstruction and interferometric SAR processing were executed using Swiss Gamma's SAR analytical software “Gamma SAR Processor” (Gamma Remote Sensing, 2000)¹⁰⁾.
We used the software “Erdas Imagine 8.7 Professional” of Leica Geosystems GIS & Mapping (LGGM) for geometric correction processing between images and the GIS software “ArcGIS version 9.1” of ESRI for image and vector data superimposition and GIS analysis.

Fig.3 Grounding line extraction processing flow