Variability of Landsat Thematic Mapper data in boreal deciduous and mixed-wood stands with conifer understory

Publisher Website
Google Scholar
Abstract
In this study, we examine Landsat TM satellite multispectral imagery and several image processing strategies to determine the most accurate method to detect and map white spruce understories in deciduous and mixed-wood stands in Alberta. These stands may be considered as part of the conifer land base that is defined as stands which contain or are projected to contain a minimum conifer volume at rotation. Images acquired in late April (leaf-off) and late July (leaf-on) were used to generate signatures for three levels of understory (heavy, light, nil) in five overstory classes. Separability statistics indicate that a reasonable degree of success can be obtained in mapping some of the understory classes with conventional classification tools. Linear discriminant functions using different classification schema and discriminating variables are presented to indicate the level of accuracy that may be obtained in a supervised classification mapping exercise.