Forest species classification based on optical properties using hyperspectral data

Possibly the most important factor for management and monitoring of forests is the knowledge of the type of species that covers the area of interest. The advent of hyperspectral remote sensing with its narrow bandwidth provides new potential to significantly improve forest species classification. The large data redundancy, common to hyperspectral datasets, makes species classification challenging. The overall goal of this thesis is to develop a methodology for identifying various forest species based on reflectance features sensitive to the vegetation parameters that are systematically different by species.;In this study, the state parameters that are systematically different among forest species were first identified then the optical indices were developed and calculated from remote sensing data to reflect these state parameters. The developed indices were validated using the high-spatial resolution hyperspectral remote sensing data Compact Airborne Spectrographic Imager (casi). The results demonstrate improvements in the accuracy and efficiency of tree species identification using the optical indices that are sensitive to the state parameters of the forest canopies.;This study undertakes the systematic analysis using a three-tiered approach: a pigment biochemical analysis, a tree cover simulation, and finally a species classification, making this study unique in that all 3 analysis steps are performed together. There has, of course, been research completed on these individual aspects before, but it is believed this is the first study to include all these steps together.