Global crop forecasting, still very much in the experimental stage, is a massive undertaking that calls on the research facilities and technologies of such organizations as the National Aeronautics and Space Administration (NASA), the United States Department of Agriculture (USDA), and the National Oceanic and Atmospheric Administration (NOAA).
There is no magical system by which satellites can report a simple and infallible prediction, as crop yields are subject to vagaries of policy, unforeseen weather changes, and other variables.
Nevertheless, the study of data from weather satellites, of multispectral and return beam vidicon (television) images from Landsat satellites, combined with analysis of regression models, historical records of particular crops, yield models, and crop calendars, allows scientists to assess current crop status and to estimate harvest yields with a reasonable degree of accuracy.
The Apollo 9 flight in March 1969 provided the first significant test of acquiring data about the earth’s resources by remote sensing from space.
Today, a Landsat satellite (Landsat 3), launched on March 5, 1978, orbits the earth at an altitude of about 570 miles in a circular, near polar, sun synchronous orbit. This means that the same point on the earth’s surface is viewed every 18 days. The satellite collects data by means of multispectral scanners (optical mechanical scanners), vidicon TV cameras, and radio systems.
The ability to photograph in each part of the electromagnetic spectrum has proved particularly helpful in studying vegetation and its condition. Color infrared photography, which was developed for detection of camouflaged installations during World War II, records information that the human eye cannot detect.
A multispectral scanner is an instrument, which may be mounted on a satellite or aircraft, that senses reflected and/or emitted radiation in any number of bands of the electromagnetic spectrum, depending on the design of the scanner. Landsat 3’s multispectral scanner senses radiation in five bands: green, red, thermal infrared, and two in near infrared. Multispectral scanners contain a device that splits a beam of light into spectral components. The various wavelengths are, in turn, converted to electrical signals, which can be transmitted to the earth by telemetry and then recorded on magnetic tape.
An examination of images taken in, say, three different portions of the spectrum can show that corn, alfalfa, stubble, and bare soil stimulate different responses and thus can be distinguished from one another.
Each substance has its own unique “signature,” or tone, ranging from black to numerous shades of gray to white, produced by its reflectance within a particular band. The “signature” of each substance over time thus depends on which band is examined, what stage the crop is in, soil moisture conditions, and so on. Within the green band, for instance, mature green alfalfa produces a “medium” amount of reflectance. Bare soil also produces an image of medium brightness. In the near infrared band, however, alfalfa’s reflectance is high, but that of bare soil is medium. It is thus necessary to compare and contrast multiple images, in different bands, in order to differentiate types of crops and soil.
Once various crops are identified, multispectral analysis is used to detect crop disease as well. Infrared photographs show that vigorous plants have very high reflectance, and unhealthy ones reflect less.
It is fortunate that the human eye cannot see infrared light, for we might be blinded by the reflectance of these healthy plants. When translated into color images, the reflectance shows healthy plants as reddish, sickly ones as blue or green. Other uses for multispectral imagery that relate to crop yields include determining the mineral and moisture content of different soils and recording stages of growth and the area covered by a particular crop.
There is, however, much disagreement among scientists about just how much can be gleaned from such data and to what extent the findings in one area can be applied to another. There is also considerable debate about which factors are the most significant in forecasting crop yields.
Everyone has a bias; the National Environmental Satellite Service (NESS), for example, endorses the usefulness of meteorological measurements. NESS has a geostationary operational environmental satellite (GOES) system made up of two spacecraft located on the 75 degree and 135 degree west meridians. (The satellites orbit the earth at the same speed as its rotation, thus retaining constant positions in relation to the planet, 22,000 miles above the surface.) They provide early warning of climatic conditions affecting crops, such as drought, freezing temperatures, precipitation and snow cover estimates, and measurements of solar radiation at the earth’s surface. Polar orbiting satellites, circling the earth twelve to fourteen times a day at altitudes of 500 to 900 miles, also gather meteorological data.
Still, whether one studies Landsat images or meteorological data, or both, the information is essentially useless for crop prediction without a corresponding study of yield models and crop calendars.
A yield model is a statistical model based on a long history of crop yield and weather conditions in a particular area. (It is also necessary to take into account a trend line yield, which assumes that yield increases with time because of technological advances.) A crop calendar is the normal schedule of crop development, from seedbed preparation and planting to flowering, maturity, and eventual harvesting. A crop calendar varies not only from crop to crop (and from one crop subclass to another), but also from one geographical area to another. Permanent conditions in an area may determine, to some extent, the impact of weather variations. For example, in Indiana, soil moisture is always high, so a dry year might not seriously affect crop growth; whereas in Iowa, where the soil contains less moisture, a dry year could be harmful.
Using all known technology, NASA, USDA, and NOAA in the mid 1970’s collaborated on one of the most successful experimental harvest forecast projects to date. The Large Area Crop Inventory Experiment (LACIE) had three phases, beginning in November 1974 with the study of the nine state Central Plains region of the United States. Phase II brought the Soviet Union and Canada into the picture.
Information collected in those two phases was used in Phase III, which led to a forecast of the size of the 1977 Soviet wheat crop that was accurate within 6 percent of official Soviet figures released six months later.
