Inventory Design
Because of constraints imposed by funds, manpower and time, a
sampling design, rather than a complete enumeration, is usually relied
upon to provide inventory information needed by a land managing agency.
This was the situation facing the Idaho Department of Water Resources.
The sampling design used in Idaho is described below. However,
many kinds of sampling designs are available for consideration: simple
random sampling, stratified random sampling, ratio estimation, cluster
sampling, double sampling with regression, systematic sampling and
probability sampling. Multistage variable probability sampling was
chosen as being most appropriate to the kinds of remote sensing data
available--LANDSAT color composites and small scale (1:120,000) color
infrared (CIR) transparencies. Based upon successful remote sensing
surveys conducted earlier (Wert and Roettgering 1968, Langley et al.
1969, Heller and Wear 1969, Langley 1975), the use of coarse resolution
data such as afforded by LANDSAT, to select areas for more extensive
examination seemed to apply for irrigated lands.
A sampling grid which corresponds to a map base in common use by
the land manager should be selected. The size of each sampling block
should not be too large or too small; we chose a block size of two by
two miles (3.6 by 3.6 km). A five by five km would also be a convenient
size. Each block is our primary sampling unit (PSU). Directions for
preparing a sampling grid follow.
Preparation of sampling grid
Draw grid in ink at some convenient drafting scale--1" = 5 mi or 1
em = 5 km. Number rows and columns for subsequent photo interpretation
and block relocation. Copy drafted grid onto orthochromatic film (Kodalith
or equivalent) with a copy camera. Enlarge negative to exactly 1:1,000,000
scale onto stable base film; this procedure will produce a positive
transparency at the same scale as the LANDSAT color composite. Orient
grid in a north-south or east-west direction to correspond with normal
map orientation and tape the positive transparency to the LANDSAT color
composite over the area selected for sampling (Fig. 2).
Photo Interpretation
The photo interpreter should take care to select a LANDSAT image
which was taken at a time when the bulk of the irrigated crops are near
maturity. In southern Idaho, this period occurs near August 1 each
year. If possible, underflight photography is most useful if taken
during the same periods as those when the LANDSAT data were collected.
In Idaho, the LANDSAT data were imaged on August 5, 1975, while the high
altitude CIR and the large scale (1:8,000) CIR 70 mm photos were both
photographed one day later on August 6, 1975. This was a most fortunate
coincidence of timing.