Abstract:
Vegetation maps of a selected area within the boreal forest surrounding Fairbanks, Alaska, have been generated for the nominal years of 1985, 1995, 2005, and 2015 using Landsat 4 and 5 Thematic Mapper and Landsat 8 Operational Land Imager surface reflectance products at 30 meter spatial resolution using a decision tree classification. The maps include 9 U.S. Geological Survey (USGS) vegetation classes, as well as barren land, open water, and ice/snow classes that are consistent with the classes identified in the 2001 National Land Cover Database (NLCD) map of Alaska generated by the USGS. Classification steps are based on USGS methodology, with refinements for the boreal forest, to ensure further comparison to the 2001 USGS NLCD map for Alaska. The overall weighted accuracies of first order estimates of data quality using cross validation are 93.2%, 88.4%, 93.3%, and 86.9% for the nominal years of 1985, 1995, 2005, and 2015 maps, respectively, compared to 81.8% accuracy for the USGS NLCD 2001 product. This study demonstrates that the spatial and spectral resolution of Landsat data is the best available for mapping the vegetation of Alaska's boreal forest at 1:50,000 scale. It also shows that the boreal forests surrounding Fairbanks, Alaska have witnessed a decrease in the growth of evergreen forests, an expansion of shrub and an increase in wetland distribution, all of which have been reported as impacts of a warming climate in the Arctic and Sub-arctic.
