Global, Landsat-based forest-cover change from 1990 to 2000

摘要: Historical baselines of forest cover are needed to understand the causes and consequences recent changes assess effectiveness land-use policies. However, historical assessment global distribution change has been lacking due obstacles in image acquisition, computational demands, lack retrospective reference data for classification. As limitations access imagery power overcome, possibility is increased an automated classification cover. We used locally fit trees relate hind-cast observations “stable pixels” non-forest from circa-2000 Landsat spectral measurements taken circa-1990 epoch Global Land Survey collection images. Based on analysis nearly 30,000 images, forest-cover between 1990 2000 epochs was detected based joint probabilities two epochs. Assessed across a sample areas with coincident conterminous United States, resulting maps achieved 93% accuracy 84% change—comparable or even higher than many previous national efforts. likewise showed 88% change. The depict gross gains losses cover, as well their net initial strong effects extant land use temperate regions tropics over period, while wildfire dominated boreal zone. Regions high loss (e.g., Amazonia) were associated into agriculture, southeastern US, Sweden) intensive forestry. These results, including datasets, will be basis estimation efficacy policies analyzing correlation socio-economic factors.