Abstract
LANDSAT-1 and -2 multispectral scanner (MSS) data from six overpass dates (April 2, May 17, June 4, July 10, October 17, and December 10, 1975) showed that MSS digital data for bare soil, cloud tops, and cloud shadows followed a highly predictable linear relation (soil background line) for MSS bands 5 and 7 (r2 = 0.974). Increasing vegetation development, documented by leaf area index measurements, for 1973 grain sorghum fields, was associated with displacement of sorghum MSS digital counts away from the soil background line. The LANDSAT data space surrounding the soil background line for MSS5 and MSS7 was divided into 10 decision regions corresponding to water; cloud shadow; low, medium, and high reflecting soil; cloud tops; low, medium, and dense plant cover; and a threshold region into which no LANDSAT data are expected to fall. We demonstrated that, using a table look-up procedure, based on these 10 decision regions, LANDSAT scenes could be classified into meaningful vegetation density levels, soil brightness levels, and water from the raw satellite data without prior knowledge of local crop and soil conditions. These procedures used deviate from current pattern recognition and remote sensing practice but should lead to faster and more automated machine processing of satellite MSS data for monitoring crop development, and for associating vegetation vigor and yield in large area crop yield prediction efforts. For these purposes, the procedures can be used with ancillary meteorological and crop calendar information as part of a decision tree analysis for rapid classification of vegetation, soil, or water conditions.
Date of this Version
1977