d USDA
vantify
1 also
these
explore
ccurate
"finish
graphy.
veloped
bed by
ırump£f,
f any
greatly
f the
T the
this
h £he
f.£he
vation
plots
= 133 -
"Ground-truth" plots were selected and located on gray scales of the
study area. The plots were then transferred to the U-2 photography and
the l5-minute quad sheets. Plots varied in size from 10-160 acres.
Plots were based upon the management criteria previously mentioned. Each
plot contained from one (1) to about ten (10) different pixel signatures
which were to be located and described.
Field personnel had little difficulty in locating the plots but had great
difficulty in locating and describing some of the individual pixels or
pixel groups. Vegetative descriptions tended to overlook some signifi-
cant differences such as an alfalfa field with a large wet spot in the
middle or an improved, tall wheatgrass field with a bare, shallow soil
spot in it. The area was finally classified based upon the best data
field personnel could obtain.
Statistics of the classification indicated an accuracy of from 50 to 90
percent on the original classification. We were never satisfied with the
results; however, funds never permitted extensive training of field per-
sonnel or money for the University personnel to travel to the field to
assist in gathering "ground-truth."
An opportunity arose to provide limited aerial photo coverage of a
limited number of plots for "ground-truth" analysis. This was done using
the system herein described. These plots were then further defined and a
new classification of those plots based upon field data and photo inter-
pretation of the low-flight CIR photos. A limited portion of the area
vas reclassified and compared with the photos and on-the-ground knowledge
of the plots. The results of the new classification were encouraging.
Based upon this limited trial a decision was made to use the low-flight
CIR photography to carry the brunt of the "ground-truth" for classifi-
cation of the Five-County Study. Timing of this Study is such that much
of the "ground-truth" will be unavailable by the time the classification
contract is awarded. A portion of the area is very dry, S-inch to 9-inch
precipitation, with vegetation die-off early in the spring. Therefore,
the photo strips were flown without the benefit of statistical analysis
or random plot selection. Bufficient strips were flown, however, to