LONG TERM CHANGE DETECTION USING CONTINUOUS FIELDS OF TREE COVER FROM 8KM AVHRR DATA FOR THE YEARS 1982-2000

摘要

This paper describes the use of the long term 8km Pathfinder Advanced Very High Resolution Radiometer (AVHRR) dataset in detecting changes in global tree cover. The procedure involves the use of a regression tree in predicting percent tree cover for the years 1982-2000. Training data are created from high-resolution imagery and used with phenological metrics derived from the annual AVHRR time series. Using the 18 years of predicted tree cover, change can be measured using a simple regression. Results are qualitatively promising for tropical regions. Countries exhibiting extensive reductions in tree cover include Brazil, Paraguay, Bolivia and Indonesia. Venezuela exhibits the most extensive gains in tree cover due to plantations in the eastern part of the country. Confidence maps have also been generated and allow for filtering the change pixels for more probable areas of change.