157
were determined by an algorithm that utilizes a moving
window of 3 by 3 pixels to select distinct training fields
in an unsupervised manner. Each pixel was then assigned to
respective spectral class with the ELAS maximum likelihood
program (MAXL31). The pixels were geographically refer
enced to the UTM grid using a least squares fit to a series
of common control points taken from USGS's 7% minute
topographic quadrangels.
Extensive site surveys were then used to determine the
land cover associations for the different spectral classes.
This ground truthing effort was directed toward training
sites that were selected from aerial photography, ortho
photo quads and similar imagery. The resultant eight
land cover classes generally followed the USGS level I
categories (Table 1). No attempt was made to classify any
subareas within urban areas, however, the forests were
subdivided into the level II classes of deciduous, ever
green and mixed.
The classified data for the nine scenes were then mathe
matically mosaicked into a single file. This data set was
resampled into square cells, 63.5 meters on a side. Color
separations for a general land cover map of the State at a
scale of 1:500,000 were generated from the data set. The
file comprises a major component of the state-wide
geographical information system that has combined several
other variables. Although more rigorous ground truthing
procedures and multi-temporal data sets have been used for
more detailed analysis in selected sites, the file has met
the basic needs for state-wide coverage and has created a
great deal of interest in Landsat processing within the
State.
USGS LAND USE AND LAND COVER MAPS
The USGS land use and land cover mapping program represents
the first attempt to systematically inventory the use of
land in the United States (Place 1977). The program
utilizes high altitude photography to determine level II
land use and land cover polygons for the USGS 1:250,000
scale basemaps. A major component of the program has been
the generation of digital files for these maps. Potentially
these files represent a major data base for computer
mapping and geographical analysis.
South Carolina became actively involved in the USGS program
in 1978 when its Coastal Zone Management Plan mandated the
inventory and mapping of land use. At that time the USGS
provided the State with its Geographical Information
Retrieval and Analysis System (GIRAS) and digitized polygon
files for the James Island and Savannah quadrangles (Fig. 2).
The Graphics Division at the University of South Carolina
digitized parts of three other maps and then generated the
combined polygon files required for the production of the
Coastal Zone Land Use and Land Cover Map that covered an
eight county area (Cowen et al 1979).
Over the next couple of years the USGS furnished the