“…Because of the advantage of high temporal frequency, many quite subtle disturbance events of forest, such as defoliation, diseases, insect pests and regeneration, can be captured based on the change of vegetation spectral attribution (Goodwin et al, 2008;Hermosilla, Wulder, White, Coops, & Hobart, 2015;Zhu, Woodcock, & Olofsson, 2012). In addition to its wide applications in forest ecosystems, such method has also been applied to quantify changes of impervious surfaces in urban environments (Powell, Cohen, Yang, Pierce, & Alberti, 2008;Schneider, 2012), coral reef health (Palandro et al, 2008) and fire events (Röder, Hill, Duguy, Alloza, & Vallejo, 2008). Pre-classification change detection techniques, whether using dense time series image data or not, generally only generates "change" vs. "no-change" maps, but do not specify the type of change (Berberoglu & Akin, 2009;Lu et al, 2004;Singh, 1989).…”