190 
satellite images are integrated to 
obtain a representative six hour value 
for clouds. Rainfall coefficients are 
applied to a categorical cloud type 
(and/or infrared brightness) to make a 
first approximation of rainfall. A 
second rainfall estimate is obtained by 
using a procedure for inverse distance 
from ground stations within an area of 
influence of a grid cell. An algorithm 
produces an estimate of rainfall for a 
grid cell on a daily basis. 
The CROPCAST model simulates the crop 
environment, in part, through a 
soil-water budget used to estimate 
plant available moisture. Projected 
weather is included in the modeling 
system to obtain end - of-season 
estimates of yield at the grid cell 
level. Information from model output 
is available to clients with personal 
computer availability within 24 hours 
after data acquisition. The program is 
a subscription service and reports are 
routinely available prior to USDA 
forecasts. It should be added that the 
models have no direct spectral inputs. 
Producer Decision Models 
During the mid-70's, researchers began 
reporting crop growth models that 
attempted to mathematically describe 
the growth process. These models were 
primarily descriptive in that they 
attempted, via environmental inputs, to 
describe the various processes involved 
in plant growth, such as 
photosynthesis, water balance, etc.. 
Many of the models, over a period of 
time, became quite complex in structure 
because they took into account micro 
details of the growth process. 
Hundreds of lines of code were used to 
describe the various aspects of plant 
growth. 
One of these models is GOSSYM (Baker et 
al, 1983), a cotton production 
mechanistic model, is now being used 
operationally by cotton producers in 
the southern delta area of the United 
States. In 1988, about 200 farms were 
using the model to assist them in 
production management decisions. In 
order to make such a complex modelling 
structure "friendly" to a producer user 
community, an expert system called 
COMAX (Lemmon, 1986) was developed. To 
provide the daily inputs necessary to 
run the model, a typical GOSSYM user 
has an automated weather station. Data 
coming from the weather station 
includes temperature, rainfall, solar 
radiation, and wind run. These data 
are fed, via telephone modem, to the 
computer where the model is actually 
being run. Using both production input 
information and daily meterogical 
variables, the producer can determine 
management strategies and evaluate 
benefits and penalties, of production 
management decisions in a quantitative 
environment. 
Since the producer must invest a modest 
amount of money in software, pay for 
training and acquire hardware including 
the weather station, this is virtually 
a commercial model application. 
Although new, the system seems to be 
expanding fairly rapidly as a 
management tool. The Clemson 
University and the Mississippi 
Agriculture and Forestry Experiment 
Station have established the USDA-ES 
GOSSYM-COMAX Information Unit (GCIU) to 
support users. Training is provided 
and a staff can assist users in 
answering questions and provide update 
information. GCIU is producing a 
monthly newsletter to advise producers 
of particular aspects that come to 
light on system use, allow users to 
share information, announce training 
sessions and let producers know in 
advance of future enhancements in the 
system. GOSSYM-COMAX is the only known 
producer-level growth simulation system 
in operational use. Two other models 
(Acock, 1989) identified as FLEXCROP, 
to recommend wheat fertilization 
levels, and COFARM, to make fertilizer 
and tillage decisions, are being 
evaluated for operational use. 
SPECTRAL DATA AND YIELD FORECASTING 
During the mid 1980's, research started 
to determine what capability the NOAA 
polar orbiting satellites, NOAA-n 
series, have for agriculture. Two of 
the bands on these satellites were 
similar to those shown to be of 
agricultural importance from Landsat: 
specifically, channel 1, from .58 - .68 
um, and channel 2, from .72 to 1.10 um, 
respectively. Landsats provide the 
same coverage of the spectrum in the 
visiable and near infared. The 
resolution at nadir is approximately 
1.1 km, very gross. To counter balance 
this, at least potentially, is the 
frequency of coverage. Each of the 
NOAA-n polar orbiters provides one 
day-time pass. Normally there are two 
satellites in operation to provide 
morning and afternoon coverage. 
Landsat has either an 18 or 16 day 
repeat coverage with one satellite or 
when two satellites have been operable, 
twice that frequency. The advantage of 
the NOAA-n series satellites is the 
frequency of coverage. Because 
agricultural field sizes are generally 
much smaller than 1.1 km resolution, 
each resolution element is mixed for 
agriculture. However, production 
patterns tend to remain the same over 
time and stress, particularly moisture 
stress, in non-irrigated agriculture 
generally affects areas larger than 1.1 
kilometer. 
One approach that has been investigated 
is to examine spectral data from NOAA-n 
satellites to observe the temporal 
change during the growing season and, 
from this, empirically relate crop