Mapping forest canopy disturbance in the Upper Great Lakes, USA
This article is part of a larger document. View the larger document here.Abstract
A map of forest canopy disturbance was generated for Michigan, Wisconsin, and most of Minnesota using 42 Landsat time series stacks (LTSS) and a vegetation change tracker (VCTw) algorithm. Corresponding winter imagery was used to reduce commission errors of forest disturbance by identifying areas of persistent snow cover. The resulting disturbance age map was classed into 5-year age classes and then used to attribute age to forested pixels within the National Land Cover Database of 2011. Overall map classification accuracy was 84.9 percent when using Forest Inventory and Analysis data as reference. User's and producer's accuracies were high for persistent forest, nonforest, and water, but low for disturbed forest 5-year age classes, likely due to rarity of forest canopy disturbance on the landscape and confusion among 5-year age classes in both map and reference data sets.