classification attempts to actually determine land use and
not just land cover it naturally defines agricultural land
more broadly than would be possible from satellite imagery.
Consequently the USGS file has fewer areas that are left
in the category of barren land. Nevertheless, the two
files reveal good correspondence both in terms of magnitude
and spatial alignment of agricultural land.
Forests, the major use of land in South Carolina, are
difficult to classify into relevant subcategories utilizing
remote sensing procedures. In order to compare the two
processes in this study, electrostatic plots were generated
for cells classified as mixed, deciduous or evergreen
forests. The two plots for each category were then photo
graphically overlaid to display areas of agreement (Figs.
4-6). This process clearly reveals some major differences
in the two categorizing procedures and dramatically
indicates a need for a more systematic method for classifi
cation of forests from remotely sensed data.
According to the USGS classification, an area of mixed
forest must include a one-third intermixture of either
evergreen or deciduous trees, as well as have an overall
density of at least ten per cent (Anderson et al, 1976 p.
16). General agreement between the areas of mixed forest
can be found both in the northwestern portion and central
coastal sections of the State (Fig. 4). In other areas it
appears that the mixed category has been employed too
frequently. This probably is attributable to a combination
of insufficient ground truth information and a general
tendency to merge several spectral classes into this rather
nebulous category.
The most startling differences between the two classifi
cations are related to the categories of deciduous and
evergreen forests (Figs. 5 and 6). The most striking aspect
is the straight lines on the USGS plots for both deciduous
and evergreen forests. These lines correspond directly to
the map sheet boundaries and are particularly noticeable
between the Florence and Georgetown quadrangles (Fig. 2).
Obviously there was either something different in the raw
materials used for the classification or in the interpre
tation of the photography. Since the coastal map sheets
were compiled during the same time period, it is more
likely that the differences relate to the interpretation
process than to the imagery.
In addition to the interpretation problem with the USGS
maps, there is also a problem of inadequate spatial
correspondence with the Landsat derived map for both
deciduous and evergreen forests. A particular concern
relates to the classification of bottomland hardwoods that
are found along the major river systems. While they are
prominant features of the Landsat derived map of deciduous
forests, they are noticeably absent from the same USGS map.
Subsequent analysis revealed that most of the floodplains
for the coastal quadrangles had been classified as forested
wetlands by the USGS. Generally, the USGS has made a
purposeful attempt to delineate forested wetlands whenever
160