Snow line mapping using radar imagery, Place Glacier, B.C.
This thesis evaluates the effectiveness of ERS-1 synthetic aperture radar (SAR) imagery for mapping movement of the transient snow line in the Place Glacier basin during an ablation season. Hydrological models of mountainous regions require frequent and accurate measurement of snow and ice covered areas to produce timely forecasts. Studies of glacier mass balance also require this information to monitor glacier health and assess changes in water resources. These areas are extensive and often inaccessible. Snow and ice accumulation can be effectively mapped using imagery from optical sensors such as Landsat-TM. However, the presence of cloud cover often limits the use of optical image data at times when it is required. SAR data can be obtained independent of weather and time of day, thus providing a potentially timely source of hydrological and glaciological data. The two primary objectives of this study are to normalise the topographically induced distortions (radiometric and geometric) inherent in SAR imagery of rugged terrain and to delineate the snow line in the normalised imagery as the boundary between the wet snow and glacier ice facies.. SAR images acquired on June 19, 1992, August 28, 1992, and October 2, 1992 were chosen to represent different stages of the ablation period. A digital elevation model (DEM) with 60 m grid spacing is created from survey and digitised map data to provide the topographic information necessary for normalising the SAR imagery. The radiometric distortions are normalised with a cosine correction and the image texture is enhanced to take advantage of the spatial distribution of tonal variations within each image. To minimise geometric distortions and georeference the imagery each cosine corrected SAR image is ortho-rectified to an error of 40 m or better using the DEM and satellite orbital and ephemeris data. A supervised classification is performed on the ortho-rectified imagery to map the spatial distribution of snow and glacial ice within the basin. In June, when the glacier is still snow covered, the normalised imagery shows a significant difference between the return for wet snow at low incidence angles ( < 30Â°) and that for wet snow at large angles(> 30Â°), the former being several times greater. This is probably due to the difference between surface scattering at small incidence angles and volume scattering at large angles. The visual boundary between the wet snow and glacier ice surfaces on the ortho-rectified and classified August and October images is within 100 m horizontally and 50 m vertically of the snow line vectors obtained from field data. The glacier boundary is also discernible to within 50 m of the 1994 glacier outline. Several isolated bare ice areas that are marked with crevasses or runoff runnels give a low return similar to wet snow resulting in some confusion between glacier ice and wet snow. Despite the localised confusion between glacier ice and wet snow, examination of the methodology shows that SAR imagery is an effective and relatively inexpensive means of mapping glacier surface types.