Forest Change Analysis Using Time-Series Landsat Observations ������������������������
Forest change is highly related to many surface processes of the Earth. Covering about 40% of the ice-free land surface, forests contain nearly 80% of the total carbon estimated to be in the terrestrial aboveground biosphere (Waring and Running 1998). Forest disturbance and recovery processes are major mechanisms that determine the carbon residence time in the terrestrial biosphere and the net carbon ux between the biosphere and the atmosphere (Law et al. 2004; Hirsch et al. 2004). By shaping the landscape pattern of forest age and structure (Peterken 2001), forest change can affect land hydrology, climate, and biogeochemical processes (Band 1993; Sahin and Hall 1996; Giambelluca et al. 2000), and it has complex but often adverse impacts on biological conservation by threatening the habitats of endangered species (DeFries et al. 2005; Zartman 2003; e.g., Kinnaird et al. 2003). Knowledge of forest disturbance history and recovery is therefore necessary in order to advance many earth science applications.