37,438 24,404 29,909
0.93 0.806 0.994 0.947
21,694
(acres)
957 Conf. Int.
149,500 acres measured from U-2 photos taken in 1975.
1/
of irrigated lands than other films and their color representations are
close to those of LANDSAT color composites. On the Idaho study, we were
fortunate to get good CIR transparencies over our test site one day
after the LANDSAT pass. In the event that aerial photography cannot be
obtained, the PSU boundaries can be transferred to existing maps and the
irrigated lands be delineated completely on the ground to obtain actual
acreage (21) within each PSU.
If aerial photography is available, the block boundaries of the
sample PSU's are transferred from LANDSAT to the photographs by the use
of a transfer device, such as the Bausch and Lomb Zoom Transfer Scope.
Then, within the PSU all identifiable irrigated lands can be drawn on an
overlay of the air photo. Both the area of the irrigated lands and the
total area of the PSU are carefully planimetered and the percent that
the irrigated lands are of the total PSU area is calculated. This
percentage is then multiplied by the area of the PSU (acres or hectares)
to determine the actual acreage (a.). The computations are shown in
tabular form in Table 2. For more details on the derivation of the
probability sampling formula and the application to irrigation see
Langley (1975) and Johnson (1977), respectively.
Results
The results of four independent estimates are shown on the bottom
portion of Table 1. The actual acreage of irrigated land in the Idaho
test site was subsequently measured very accurately by planimeter on the
1:120,000 CIR transparencies and found to contain 149,500 acres (59,800
ha). All four estimates of total irrigated lands were close to the
actual irrigated acreage and well within one standard deviation of the
estimate. Interpreter 2 had the largest variance and sample error
(10.3%). His fifth observation (underlined on upper portion of Table 1)
was responsible for this increase in variance; he predicted 55 percent
irrigated from LANDSAT and the air photos measured 91.7 percent. Upon
examination of this PSU on the LANDSAT image, we found many wheat fields
which had lost their IR reflectance because they had been harvested
earlier in the season. However, these were fields with permanently
installed irrigation systems which had been irrigated earlier in the
season. One must realize that these kinds of errors can occur. When
they do, such errors will cause the variance and sampling error to be
inflated. Such errors can be minimized if the lower level aerial
photography or ground examination is made close to the time of the
LANDSAT overpass.
All computations can be made on a small desk or hand computer. The
method is surprisingly accurate and can be repeated with equally good
results. The photo interpretation, measurement of areas and computations
takes 12 to 15 hours to complete. If countries with developing irrigated
agriculture would be satisfied with sampling errors of 6 to 10 percent,
this technique should be considered.