Forest cover classification using Landsat Thematic Mapper data for areal expansion of line LAI estimate generated through airborne laser profiler

摘要

A simple cover classification of Canadian boreal forest was conducted using Landsat Thematic Mapper (TM) imagery to expand a line estimate of leaf area index (LAI) into a two-dimensional regional one. The line estimate had been made through a 600km long continuous vegetation profile obtained by airborne laser altimetry. The present study area of 170×30km straddles the central portion of the laser profiling transect, from Wandering River north to Fort McMurray, Alberta, Canada. A total of eight land cover types were identified first in the field, and then some 83 training points and another 74 reference points were chosen and recorded for a supervised classification and its accuracy assessment respectively. By applying a supervised procedure to Landsat TM data in two different seasons, these eight cover types, consisting of six vegetated covers, i. e. closed and open conifer forests, conifer woodland, closed and open broad-leaved forests and marsh thicket, and two nonvegetated covers, i. e. bare ground and water surface, were classified. The classification was basically successful with an overall accuracy of 76%. Finally, using an overlay of this land cover map and the airborne laser profiling flight track, the mean LAI for each type of vegetation cover was obtained, and subsequently fed back to the land cover map to form a false-color map showing the two-dimensional distribution of LAI over the entire study area.