vegeta-
ics on
ropland
erage).
oscopic
in the
'ground-
grouped
| which
out 50
revealed
lied to
t pos-
ls were
lividual
»scribed
indicate
flight
proper
ssifica-
m 739 =
Individual pixels and pixel groups were actually located on the large-
scale CIR photos by transfer methods which are now briefly described.
The first step on the Goose Lake project was to locate the general
position of the low flight photos on the unsupervised classification of
the study area. These general areas were then transferred to USGS quad
sheets and to high elevation U-2 photos. The low flight color IR photos
were also located on the U-2 photos and USGS quad sheets. The individual
pixels and pixel groups were then located, described, and clustered into
management groups. A new classification of all of the plots was then
made and compared with the low flight CIR photos, and on-the-ground know-
ledge of the plots.
DISCUSSION
Inventory Efficiency
A large portion of the cost of an inventory project regardless of the
system is absorbed in the accumulation of "ground-truth." The Goose Lake
project originally involved 150 ground-truth plots on cropland and range-
land which varied from 10 to 160 acres. Each of these were visited two
or more times in the field. Each plot required from 2 to 4 hours per
trip.
Use of large-scale CIR photography would have reduced the number of plots
to about 50 with one trip to each. Thus, the time spent on each plot
with photos in hand would have been 1 to 2 hours.
The sequence followed on the Goose Lake project is not considered the
desired method for ground-truthing, nor vill it be followed on the five-
county project.
The revised inventory technique will follow these steps.
1. An unsupervised classification will be analysed and plots
selected from the 960,000 acres covered by low flight photography.
2. Individual pixels and pixel groups of the selected plots will be
located on the CIR photos and sampled on-the-ground.