Within the same region, Landsat data
were acquired and analyzed to estimate
the proportion of wheat in 5 by 6 nautical
mile samples. This analysis was accom-
plished with procedures not requiring
ground truth. Figure 10 illustrates the
LACIE analysis procedure for area deter-
mination of wheat. Digital Landsat data
collected from worldwide receiving
stations were forwarded to Goddard Space
Flight Center (GSFC) where the LACIE
sample segments were extracted and trans-
mitted to JSC. There the digital data
were stored for use throughout the season
and transformed to imagery for analysis.
The LACIE analyst used ancillary
data (agromet data, crop calendars, and
Landsat full frames) to help him identify
the wheat fields in the imagery. The
analyst labeled a small sample of the
imagery as wheat and non-wheat to train
a pattern recognition algorithm on JSC
computers. The algorithm extended the
training to the complete segment and
produced a percentage of wheat. This
procedure was repeated several times
(multitemporally) during the crop season
LACIE RESULTS
The 1974-75 crop year, LACIE Phase I,
was devoted to developing the experi-
mental apparatus, assembling data bases
of historic agronomic and weather data,
developing sampling approaches, yield
models, and analytical procedures, and
training people. Preliminary testing and
evaluation was also done. During the 1975-
76 crop year, LACIE Phase II, the techno-
logy as modified in Phase I was evaluated
in the U.S. yardstick region, in the hard
red spring wheat of Canada, and in both
a spring wheat and a winter wheat region
in the U.S.S.R. Exploratory studies of
wheat identification and tests of yield
models were conducted in five other
important wheat regions.* The experi-
*India, China, Australia, Argentina,
and Brazil.
for each segment using accumulated Land-
sat images and updated ancillary data.
In the U.S. yardstick region, stratum
estimates for hectares, yield, and pro-
duction were made at the state level. In
Canada and the U.S.S.R., the strata were
considerably larger. The purpose of
analyzing estimates at the stratum level
was twofold:
1) To explore the geographic size at
which weather and hectarage (or acreage)
must be associated to attain a given
precision at the region or country level.
2) To investigate within a single crop
year how the technology performed in each
of the many strata with their different
agricultural and meteorological charac-
teristics. Thus, in a sense, testing the
technology in 10 subregions during a
single crop year equals testing it in a
single region for 10 crop years. This is
well illustrated in an analysis of the
different conditions that prevailed in
nine states in the yardstick region of
the U.S. during the three crop years.
mental results of Phases I and II, which
were most encouraging, permitted inves-
tigators to improve the algorithms and
procedures used to analyze the weather
and Landsat data.
An evaluation of the results of
Phase II led LACIE investigators to con-
clude that the approach worked well in
estimating the production of winter wheat
in both the U.S. and the U.S.S.R. Three
major problems encountered were the
difficulty in reliably differentiating
spring barley from spring wheat in Land-
sat data, the problem of developing yield
models in countries with little or no
reliable historic data, and the diffi-
culty in making accurate acreage esti-
mates in regions with small fields. For
example, because it was difficult for the
Landsat imagery to resolve the shape of
the strip faklow fields (long but narrow)
the wheat area in the hard red spring
NASA/JSC
DIGITAL
IMAGE DATA
LANDSAT
|
——— Cw
GSFC
"e