Analysis Results for Nor.th Central Illinois
LANDSAT Scene CIA; August 3, 1975
The major goal of this analysis was to make crop acreage estimates for
corn and soybeans in CIA.
These estimates were to be based on LANDSAT data
as an auxi.liary variable using a regression type estimator.
The area to be
considered was seven counties in North Central Illinois:
Dekalb, Lee, McHenry, Ogle, and Winnebago.
The area originally to be considered (scene Cl) had to be reduced to the
seven counties (CIA) because of cloud cover problems in the south and east
sections of Cl.
The part with cloud cover was analyzed on another scene with
a different date.
The estimates for the seven individual counties were computed
along with their relative sampling errors and also for the seven county aggregate
Four of the county estimates were used along with county estimates from
other scenes to make an estimate for the Northwest Crop Reporting district.
estimate can be directly compared to the JES estimate for that area and year.
Although the precision of the individual estimates cannot be compared directly
to JES estimates; the RE2 relative efficiency in the tables is a measure of the
gain, in terms of lower variance, of the regression estimate over the JES direct
expansion type estimate.
II. DESIGNING AND EVALUATING CLASSIFIERS
The classification categories were determined from the llNot Background"
Any crop was included with more than (or close to) 200 pixels.
These ten crop types (or covers) were then clustered and fourteen categories were
determined, with waste, oats, water, and cropland pasture having two categories
each and with wheat and oat stubble combined to one category.
Two levels of prior probabilities were studied; priors proportional and
expanded reported acres (PER) and Equal Priors (EP).
The PER priors were obtainec
from the '0 pooling of the seven counties with the Direct Expansion estimator.
Two methods were considered to allocate the da