Forest structure classification using airborne multispectral image texture and kriging analysis

A study in southwestern Alberta conifer stands was designed based on the premise that obtaining image texture from a kriging surface could improve the accuracy of forest structure classification because the resulting textures would be less sensitive to random variations in spectral response and more related to structural features, such as crown size and shape. We extracted image spectral data, textural derivatives and a kriging surface from multispectral CASI imagery (2 m spatial resolution); 83% classification accuracy (K_hat=0.79) was obtained in nine species composition classes.