Automatic deforestation detection using time series Landsat images in a tropical forest of China

摘要

In this paper, we compared two versions of the Vegetation Change Tracker (VCT) algorithm, one based on the Integrated Forest Index (IFI) and the other on Disturbance index (DI) derived using the tasseled cap transformation. Time series Landsat ETM+ data from 1999 to 2010 were used in a tropical site of Xishuangbanna, Yunnan province of China. The results were evaluated using 376 field plots. The accuracies of IFI-VCT and DI-VCT are 82.7% and 79.9%, respectively. These two algorithms had different performances in different areas. The DI enhanced the separability between some tall agriculture and forest. Haze and seasonality affect DI result more than IFI.