Land cover maps have been produced using satellite imagery to monitor forest resources since the launch of Landsat 1. Research has shown that stacking leaf-on and leaf-off imagery (combining two separate images into one image for processing) may improve classification accuracy. It is assumed that the combination of data will aid in differentiation between forest types. In this study we explored potential benefits of using multidate imagery versus single-date imagery for operational forest cover classification as part of an annual remote sensing forest inventory system. Landsat Thematic Mapper (TM) imagery was used to classify land cover into four classes. Six band combinations were tested to determine differences in classification accuracy and if any were significant enough to justify the extra cost and increased difficulty of image acquisition. The effects of inclusion/exclusion of the moisture band (TM band 5) also were examined. Results show overall accuracy ranged from 72 to 79% with no significant difference between single and multidate classifications. We feel the minimal increase (3.06%) in overall accuracy, coupled with the operational difficulties of obtaining multiple (two), useable images per year, does not support the use of multidate stacked imagery. Additional research should focus on fully utilizing data from a single scene by improving classification methodologies.
Each regional journal of applied forestry focuses on research, practice, and techniques targeted to foresters and allied professionals in specific regions of the United States and Canada. The Southern Journal of Applied Forestry covers an area from Virginia and Kentucky south to as far west as Oklahoma and east Texas.