Detecting Changes in Forest Vegetation using Multitemporal Landsat TM Data : A case study in the Shibata Forest, Niigata Prefecture

Abstract

This paper outlines three change detection methods, i.e., (1) direct multidate classif{cation (MDC), (2} 12-dimensional multitemporal principal component (MPC) and (3) 2-dimensional multitemperal principal component, to detect specific changes in forest cover due to new artificial regeneration, natural growth and forest cutting. Of thes.e 3 techniques, the MDC technique which simply classifies the multidate bands directly consistently providecl better delineation of forest change, In the selection of band dimensionality, the 12-dimensional MPC is the most effective, Not only this technique can reduce the dimensionality of the or{ginal data, but also effectively picks up the change of interest and provides nearly as the first technique. For all date-pairs (3-to 7-year intervals) the changes due to forest cutting and new artificial regeneration and tree height growth of young plantations were detected; however, the height growth of the larger trees, i.e., Sl-S2!3, Pl-P2 and P2-P3, could not be detected. As indicated in this study, as the time interva] increases, the ability of multidate TM to detect forest cover change inereases. IVithin the young forest plantation, differences in the density of under-story vegetation of Japanese cedar seedlings sometimes led to misclassification.