HR: 1340h
AN: C13B-0283    [Abstracts]
TI: Preliminary Hyperspectral Land Cover Classification of the Alaskan Arctic Coastal Plain- Application to Landscape Chronologies and the Thaw-Lake Geomorphic Cycle
AU: Hinkel, K M
EM: kenneth.hinkel@uc.edu
AF: University of Cincinnati, Department of Geography, Cincinnati, OH 45221-0131 United States
AU: * Beck, R A
EM: richard.beck@uc.edu
AF: Central State University, International Center for Water Resources Management P.O. Box 1004, Wilberforce, OH 45384 United States
AU: Eisner, W R
EM: wendy.eisner@uc.edu
AF: University of Cincinnati, Department of Geography, Cincinnati, OH 45221-0131 United States
AU: Tweedie, C E
EM: tweedie@msu.edu
AF: Michigan State University, Department of Plant Biology, 100 North Kedzie Hall, East Lansing, MI 48824-1031 United States
AB: Georeferenced ground- and helicopter-acquired spectral libraries have been assembled for a variety of vegetation and other land cover types related to drained thaw lake basins on the Arctic Coastal Plain of Alaska between Barrow and Atqasuk. Challenging illumination conditions and resulting low signal-to-noise ratios in the short wave infrared (SWIR) portion of the spectrum have persuaded us to focus our efforts on the visible and near infrared (VNIR) parts of the spectrum. Preliminary supervised classifications of atmospherically corrected VNIR hyperspectral Hyperion imagery from the EO-1 satellite with ground- and helicopter-acquired training spectra show generally good agreement with ground truth observations of vegetation with some exceptions. The supervised classifications of Hyperion imagery are particularly useful in the search for recently drained thaw lake basins, and for some key vegetation types usually associated with specific age-related stages in basin vegetation succession. Comparison of ground- and helicopter-acquired training spectra illustrate significant differences associated with spatial scaling issues and the need for hyperspectral spectral resolution. For example, some of the oldest geomorphic surfaces have some of the highest helicopter-acquired Normalized Difference Vegetation Index (NDVI) values due to absorption of red wavelengths by water in well-developed ice-wedge troughs despite relatively modest near infrared (NIR) reflectance values. Helicopter-borne NDVI measurements of these older surfaces are similar to those of recently drained thaw-lake basins. However, map view outlines of the older and irregular high NDVI surfaces are quite different from those of the younger high NDVI surfaces, which are usually similar to recently drained thaw lakes. These observations suggest that the most effective methods for tundra land cover classification with regard to the thaw lake geomorphic cycle will probably be hybrid methods that combine hyperspectral and textural classifications at relatively high (sub-5-meter) spatial resolutions.
DE: 9315 Arctic region
DE: 1823 Frozen ground
DE: 1890 Wetlands
DE: 1894 Instruments and techniques
DE: 1640 Remote sensing
SC: Cryosphere [C]
MN: 2004 AGU Fall Meeting