Reconstructing Holocene Glacier Changes in West Greenland From Multispectral ASTER Imagery

To understand the mass balance of the Greenland Ice Sheet and to identify mechanisms controlling that balance and Greenland's contribution to future changes in global sea level, it is crucial to construct longer temporal records, reaching back to the Little Ice Age (LIA) or beyond. The primary objectives of this project are to develop procedures for mapping glacial trimlines, marking maximum glacier extent during the LIA, and terminal moraines indicating earlier advanced terminus positions, in central west Greenland using multispectral ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) images. The motivation for using satellite imagery for mapping glacial-geological features is the greater spatial coverage that can be achieved, as opposed to the traditional method of field mapping in restricted areas. ASTER imagery provides spectral bands spanning from the visible to the thermal infrared bands, including two stereo bands, enabling us to map the spectral properties of the Earth's surface as well as to obtain surface topography. This poster presents examples of mapping the 3D shapes of glacial geomorphological features using supervised classification, visual interpretation and advanced pattern recognition methods, and results of the volume change computation and interpretation, focusing on the Jakobshavn drainage basin. For trimline mapping, a Digital Elevation Model (DEM) was generated from the stereo bands of the same data set, followed by orthorectification using Ground Control Points (GCPs) and checkpoints extracted from stereo aerial photographs and digital maps. Surface reflectance was estimated from the raw DN values by applying the Empirical Line Correction model for atmospheric effects. Maximum likelihood classification, in supervised mode, was applied to distinguish different land cover types. Classification of the ASTER image with nine non-thermal bands provides a good discrimination between the exposed fresh rock surfaces, moraines of the trimzone, and the rocky patches within the vegetated area. However, debris-covered ice is often misclassified as trimzone and snow-covered vegetation pixels are classified as trimzone. Because inclusion of topographic information usually improves the classification accuracy of debris-covered ice and snow-covered moraines, an object-based classification was applied to delineate different landscape units from a merged data set created from the ASTER spectral bands and the surface slope. A simple ice sheet model was used to calculate ice volume changes in the Jakobshavn Drainage Basin since the LIA from the trimline and contemporary ice sheet margin elevations. We also estimated the total contribution of the Greenland Ice Sheet to global sea level change by extrapolating our results.