Humans must eat to live. One essential part of global monitoring is our 
ability to watch our food supply. But more importantly, we must be able to 
react to changes in our food supply. The monitoring capability must be at a 
temporal and spatial scale adequate to detect these changes. The purpose of 
hemispherical crop monitoring is to record change in the crops as some 
measure of average yield. Our paper describes a new way of presenting the 
information so yields in kg ha’ can be provided. 
The first important step in monitoring for change is the production of a 
cloud free image. In 1986, Manitoba Remote Sensing Centre received a 
contract from the Canada Centre for Remote Sensing to build the Canada Crop 
Information System. DIPIX Technologies developed the system. This system 
enabled daily NOAA AVHRR (Advanced Very High Resolution Radiometer) imagery 
to be composited and pieced together (mosaiced) into a single cloud free 
image to cover the whole Northern Hemisphere. The maximum value for each 
pixel was also preserved for the week. The test data set was NDVI 
(Normalized Difference Vegetation Index) for the Canadian Prairies. 
The Northern Hemisphere was calculated from NESDIC (NASA) GVI (Global 
Vegetation Index) data. The resolution of the GVI data is about 18 km (15 km 
at the equator to 30 km near the pole) . These parameters set the temporal 
and spatial bounds (ie. daily data is used to produce a weekly composite and 
1 km data is used to produce an 18 km average pixel). 
The second step was the realization that satellite imagery must be absolutely 
calibrated to monitor change. We must^ be able to produce more than an index. 
We need crop information in kg ha . The Saskatchewan Research Council 
developed the results through a contract from the Canada/Saskatchewan Crop 
Insurance Corporation. 
Calibration must be made for different satellites even in the same series, 
haze, and off-nadir stretch. All of these difficulties have been overcome in 
this scheme. But acceptance of the results would not be complete without a 
way to produce crop information in real values. This is difficult without 
knowing the crop planted. We found that by dividing this week's NDVI by last 
year's maximum NDVI gave us a percent yield map. The resulting image is 
actually a percent of last year's yield. Normally, farmers have a planting 
rotation derived from tradition. The year-to-year rotation is uniform for a 
region. But even if the farmers change their cropping practice due to market 
or climate conditions, the changes will show on the resulting image. The 
image will quickly point out where these changes are occurring. With the use 
of GIS (Geographic Information Systems), last year's crop yield can be 
multiplied by the weekly percent yield map to produce an estimate of this 
year's yield on a weekly basis. 
BACKGROUND 
This paper was made possible because 
of a previous contract to provide an 
operational crop and forage yield 
estimate for Saskatchewan (Figure 
1) . Yield estimates were provided 
within ten days after the end of 
every week, starting June 15 and 
ending August 7, 1988 and 1989. The 
breakthrough was the production of 
cloud-free imagery by the Canada 
Centre for Remote Sensing and 
Manitoba Remote Sensing Centre. 
This breakthrough enabled the 
Saskatchewan Research Council to 
produce the crop and forage esti 
mates for the Canada/Saskatchewan 
Crop Insurance Corporation. 
The previous study provides 
calibration using 1985, 1986, 1987 
and 1988 data. NOAA [USA National 
Oceanographic Atmospheric 
264