The statewide forest/nonforest classification of Pennsylvania using Landsat MSS dataA procedure is described for processing the large volume of data needed to generate a Landsat-derived forest resource map (forest/nonforest mask) for the state of Pennsylvania, for use in a defoliation assessment program. Landsat coverage of the 28 million acres encompassing Pennsylvania requires portions of ten Landsat frames, totalling approximately 76 million pixels. A prime effort of this project was to efficiently and accurately classify this data into forest or nonforest categories. A specialized approach to Bayesian classification was developed which involved the use of MSS5 and MSS7; the identification of a single class, forest; and the use of the confidence map to generate the forest/nonforest mask. Statewide, for 7-acre contiguous forest areas, the overall agreement achieved between the mask and reference data was 90 percent.
Document ID
19840050542
Acquisition Source
Legacy CDMS
Document Type
Conference Paper
Authors
Russo, S. A. (Computer Sciences Corp. Silver Spring, MD, United States)
Stauffer, M. L. (Computer Sciences Corp. Silver Spring, MD, United States)