Geostatistical characterization of ice surfaces from ERS-1 and ERS-2 SAR data, Jakobshavn Isbrae, Greenland

Abstract

Synthetic-aperture-radar (SAR) data are useful for the study of polar regions because of their independence of cloud cover and daylight, but their analysis is complicated by a lack of absolute reference in size and grey value. The geostatistical surface-classification approach provides a solution to this problem. Parameters extracted from variograms constitute a feature vector, which is characteristic of a surface morphological province. The goal is the classification of larger areas using the result of a characterization of prototype areas that are homogeneous with respect to ice-surface morphological types. This requires (1) a feasibility study and (2) comparison with ground-truth data. Both issues are investigated in this paper. Examples of 1995 ERS-1 and 1997 ERS-2 SAR data from Jakobshavn Isbrae serve to demonstrate that different ice-surface types can be distinguished automatically with geostatistical surface characterization. Ice-surface microtopography/roughness data collected during expedition MICROTOP 97 to the Jakobshavn Isbrae drainage basin provide the ground-truth data required to investigate whether areas that appear homogeneous on a SAR image are in fact homogeneous in ice-surface morphological properties. An analysis of contemporaneous ERS-2 SAR data using geostatistical surface characterization is presented.