614.4 Branch Brown Bag Seminar

University of Maryland, Dept. of Geography
The series of Landsat instruments have produced a unique imagery record suitable for deriving forest change products at different spatial and temporal scales. Such products are needed in order to advance studies on many pressing environmental issues, including carbon accounting, ecosystem dynamics, sustainability, and the vulnerability of natural and human systems. Large quantities of Landsat images are often needed in order to develop forest change products with spatial and temporal characteristics suitable for such studies. In this seminar, the speaker will present two automated approaches for deriving forest change products using large quantities of Landsat images. The first approach is designed to reconstruct forest change history using Landsat time series stacks (LTSS) and a vegetation change tracker (VCT) algorithm. A LTSS typically consists of a sequence of images of the same location that have a nominal temporal interval (e.g., 1 image every two years). This approach will not only allow detection of forest changes, but also date the detected changes using the acquisition date of the LTSS images. The second approach is designed for forest change analysis in cases where available images are not adequate for assembling LTSS but are adequate for 2-date change analysis. This approach consists of a training data automation (TDA) method for delineating necessary training samples and uses an advanced support vector machines (SVM) algorithm to map changes. Both approaches have been tested extensively, and are being used to generate forest disturbance products through several NASA funded projects. Biographical Sketch: Chengquan Huang is a research scientist in the Department of Geography, University of Maryland. His research interests include land cover, ecosystem monitoring, and related applications in global change and earth system studies, biological conservation, and resources management. He collaborates closely with his colleagues in Geography and scientists at NASA, USGS, and US Forest Service on several land cover and forest change studies at national, continental, and global scales. Prior to his joining the University of Maryland in 2005, he spent about one year with Raytheon where he developed the science code for many NPOESS products. Prior to that, he worked at the USGS/EROS where he prototyped algorithms for developing the circa-2001 National Land Cover Database (NLCD 2001) and for developing many of the data products for the LANDFIRE project. He holds a Ph.D from the University of Maryland and an M.S. and a B.S. from Beijing University.