Using L-band SAR coherence to delineate glacier extent

Significant retreat has been seen in many glaciers worldwide. With increasing concern about the impact of global warming, there has been a concerted effort to monitor ongoing change. Projects such as Global Land Ice Measurements from Space (GLIMS) rely on the contributions of institutions worldwide to update a global database of digital glacier outlines. To a great extent, this effort relies on the manual extraction of glacier area from satellite optical instruments, such as the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). Unfortunately, dependence on optical imagery poses two significant challenges. First, clouds preclude the viewing of many glaciers, making periodic measurements impossible. Second, many alpine glaciers have significant debris cover, additionally complicating delineation of glacier extent from optical data. The sensitivity of synthetic aperture radar (SAR) coherence to motion offers a method for detecting glacial extent that is independent of weather and allows for the discrimination of active glaciers, with and without debris cover. In this paper, we describe an automated method for delineating the surface area of two glaciers in Alaska, one with and one without a debris-covered terminus. We use the periodic global observations of the Advanced Land Observing Satellite (ALOS) Phased Array L-band SAR (PALSAR) instrument to generate coherence images. Low coherence serves as our primary indicator of glacial extent. However, coherence can be degraded by dense vegetation and the topography of mountainous areas. Hence we develop a decision tree that also considers phase gradients and mountain slope to identify glacier boundaries. The result is a robust technique for generating digital glacier outlines. Coupled with the comprehensive global acquisitions of ALOS, it suggests a new method for measuring glacier extents around the world.