Satellite change detection of forest harvest patterns on an industrial forest landscape

Abstract

Multiple dates of Landsat Thematic Mapper (TM) imagery were analyzed to compare a decade of forest harvest activity in northern Maine's industrial forest. Unsupervised clustering on three-date sequences of Normalized Difference Vegetation Index classified the harvest type and time period when change occurred. Problematic clusters were reclassified using the Normalized Difference Moisture Index, which improved the detection of light partial harvests. The procedure classified clearcuts and partial cuts in all time periods with at least 80% agreement with 250 reference sample points. Patch characteristics of the harvest areas revealed a shift from large clearcuts in the late 1980s and early 1990s to fewer and smaller clearcuts in the middle to late 1990s. By the late 1990s, large clearcuts (>40 ha) were eliminated, and partial cuts increased in size. Partial cuts affected more land area than did clearcuts in all time periods. Comparison of harvest patches between the two largest landowners in the study area revealed significant differences in the number and distance between clearcut harvest areas. The trends in decreasing clearcut harvesting practices over the past decade appear to have been influenced by forestry legislation, according to the results of this study and an independent report by the Maine Forest Service. The changing harvest patterns on the landscape were detectable on time-series satellite imagery.