FIGURE C.1 Wavelength regions and bandwidths for two remote sensing satellites, Landsat 7 and ASTER. SOURCE: NASA Goddard Space Flight Center and U.S. Geological Survey.
2008). Thus, newer sensors with pixel sizes on the order of 3 to 30 m provide more accurate representations (and changes over time) of such parameters as glacial area, location of the glacier terminus, and location of the ELA.
Snow and ice are distinct from surrounding terrain in clear weather. Thick cloud cover can complicate such distinction. However, measurements of the 1.6 to 1.7 μm wavelengths (wavelength band 4 of ASTER) can overcome this issue. At these wavelengths, clouds are reflective but snow and ice are absorbing (Racoviteanu et al., 2008). Techniques including single-band ratios and the normalized difference snow index (NDSI) use the high brightness values of snow and ice in the visible wavelengths to distinguish them from darker rock, soil, or vegetation (Racoviteanu et al., 2008).8 The single-band ratio and NDSI techniques are also fast and robust, making them relatively easy to automate over extensive areas. Challenges to automatic mapping of glaciers using band ratios include (1) the presence of proglacial and supraglacial lakes, (2) the presence of fresh snow on surfaces other than a glacier, and (3) debris cover on glaciers (Racoviteanu et al., 2008). Lakes are misidentified as glacial ice because liquid water and ice have similar bulk optical properties in the visible and near-infrared wavelength. Glaciers always have snow above the ELA. However, surrounding bedrock, moraines, tundra, and other surfaces may also have snow. These snow-covered surfaces can easily be misclassified as part of a glacier. Care must be taken to discriminate snow on surrounding areas from snow on glaciers. Often multiple years of imagery are needed to do so.
Debris cover is one of the greatest challenges in remote sensing of glaciers. In the visible and near-infrared wavelengths, debris cover has optical properties very similar to the surrounding moraines. Automated methods of analyzing spectral information are ineffective in mapping ice covered by debris. However, manual digitization is time-consuming and subject to human error. Thus, debris on glaciers may result in underrepresenting glacial area and overcalculating rates of glacial retreat.
8 NDSI is calculated as (VIS - SWIR)/(VIS + SWIR), where VIS is band 1 of ASTER (0.52-0.6 µm) at 15 m and SWIR is band 4 of ASTER (1.6-1.7 µm) at 30 m.