Improving SAFIS forest removal estimates through spectral change detection

摘要

Our overall goal is to improve removal estimates in the annual forest survey through the use of spectral change detection techniques. Estimates of removals represent one of the most significant weaknesses of traditional forest surveys because of the lack of serial correlation and because inventory plots are widely spaced and removals tend to be clustered. The change to annual forest surveys is expected to increase standard errors of volume estimates due to reduced sample size in any single year - with the greatest loss of precision expected to be in removal and mortality estimates due to greatly decreased plot density. Given that even over the course of a five year inventory cycle dramatic changes in forest resources are common in the southern United States, there is a pressing need to refine the precision of removal estimates. Our specific objectives are to determine the spectral change detection technique best suited for detection of complete harvests (clearcuts) in the southern United States, develop procedures by which implementation of a spectral change detection protocol can be automated and streamlined, and assess whether or not improved and timely knowledge of harvests improves forest removal estimates.