Yield Monitoring

Fred Hapgood




All this Spring cold rain fell on the fields of the Corn Belt, leaving them chilled and sodden. On the few days when no rain fell, thick clouds prevented the ground from drying out.

Farmers cannot set seed in water-saturated soils (planting machinery compacts wet ground, making a poor incubator for baby roots), but neither can they wait forever for the soil to dry. Past a certain date every day of delay costs the fall harvest a few percent of yield. For many farmers that window was narrowing quickly. They sat inside, paced, called each other, told farmer stories, and recalled the glory years in the 70's, when it seemed like the price of corn would rise forever.

Some, like Doug Harford of Grundy County in Illinois, were a bit better prepared. Harford, a blonde column of a man in his mid-40's, had laid drainage pipes under his fields, and these had carried off enough water to allow him to squeeze some seed into the ground. Still, he too had a few hours to sit inside, watch his fields soak, and sum up the roller coaster ride he's been living since he took over his father's farm two decades back.

Every business has its uncertainities, Harford says, but farming is extreme. People in the cities read about the major droughts and floods, but the unpredictability runs into every corner of the enterprise. A crop farmer plants seed into a combination of biology (soil flora and fauna), geology (soil type), and meteorology, all of which fluctuate drastically over space or time or both. The same inputs will not produce the same output from one year to the next, nor do any two acres on a farm grow the same yield on the same year, let alone two acres from different farms. No two situations are ever equal. "People have a tendency to confuse experiences with experiments," he says. "Agriculture is mostly experiences."

This unpredictability defines farming culture, making farmers conservative, skeptical, and careful. It also makes farming management more of an art than a science. Other businesses can calculate the tradeoffs between marginal returns and incremental resource costs quantitatively, objectively. A farmer considering whether an extra hour of field inspection makes sense on a given acre must account for the effect of changes in soil type, seed variety, field and planting geometry, tillage practices, yield histories, drainage, pest populations, density of organic material, and so on, all of which can fluctuate wildly across a single field.

Rationalizing these variables into hard numbers would require massive amounts of measurement. In the absence of this data farmers make do by weighing these questions "intuitively". When they do write down numbers, they are estimates; what Harford 'WAGs', for wild-assed guesses. WAGs are not arbitrary: a skilled farmer can manage a host of complexities through "feel". But every farmer's "feel" has its natural capacity. Managing through feel sets an upper bound on the number of acres that can be managed effectively, and this limits the efficiencies of scale farmers can bring to bear.
Harford believes that if the country is going to continue to enjoy the social and economic benefits of low food prices, sooner or later farm management will have to become more quantitative; less of an experience, more of an experiment.

Harford has been thinking about this problem for some time and he knew that in theory there was a solution to this problem: measuring the flow of harvested material every few seconds as it passes through the combine, addressing each measurement with the location of the field segment that "unit" of material came from, and then using these addresses to plot the yield measurements on a map. The farmer could then control the variables by overlaying the yield data with maps representing the various factors, for instance, comparing the difference in yield that comes from laying 10 units of fertilizer Y on soil type Z, when Z is planted with seed "alpha", with laying 12 units of fertilizer on exactly the same combination of factors.

The theory behind this process, known as yield monitoring, had been discussed in the academic literature since at least the 1950's and actually used as an instrument in agricultural research since the 60's. Commercial farmers had not touched it, since the labor required -- stopping to weigh the units of harvested material every few feet, determining the coordinates of the area that had grown that unit, and handling the tremendous volumes of data required to make the overlays, was prodigious. Agricultural research stations had the staff and students to do that kind of scut work; businesses that had to make a profit did not.

In the late 80's, as often happens, someone too far outside the industry to have heard what was and was not practical looked over and saw what seemed like a huge opportunity. Al Myers was a power transmission engineer at the time, working for what was then Sundstrand, Inc. Though he was no farmer, he did know computers, and he saw that the digital engines were growing cheap and powerful enough that many of the most onerous information handling tasks involved with yield monitoring could be automated.

He also knew there was a new idea in the location business: in 1983 Ronald Reagan (in the wake of the KAL007 disaster) had announced that a network of navigation satellites built by the Defense Department would be made available for civilian use. These satellites broadcasted their orbital position and a timestamp continuously, allowing receivers moving around on the ground to use simple triangulation to find their location. This whole system was called GPS, for Global Positioning System.

Myers saw he could use these satellites to find his locations. The single remaining difficulty was breaking the harvest into measurable, standardized, units. For several years Myers worked evenings and weekends trying to find a way to automate this last function without forcing farmers to rebuild their harvesting equipment. Finally he emerged with the idea of a plate that stood in the flow of harvested material. A transducer behind the plate measured the force striking the plate face continuously, then handed the values to a processor, which corrected for complicating factors like the speed of the combine and the water content of the crop. Depending on the combine, the entire system might break an acre up into about 500 measurement units or elements, the agricultural equivalent of the pixels on a display screen or the dots in a newspaper photo. (A farm like Harford's, with 1500 acres, would be represented with about 750,000 elements.)

The 'Ag Leader 2000' went on the market in 1992 for $3300. That year and the next not many sold. "When farmers are harvesting they go flat out," Myers explained. "They're a little reluctant to get involved with something that looks like it might force them to stop and read a manual." One of his few customers was Doug Harford, who ran the machine during his 1993 and 1994 harvests. "The big surprise for me was the amount of variation in a single field," he says, opening up a loose-leaf folder full of maps. (Myers says that his customers often report seeing a 50% variation from one section of a field to another.)

Each map was speckled with a dozen different colors; each color represented a different level of soybean or corn production, in the case of beans, for instance, ranging from 30 bushels/acre to 70. Harford turned to a map of a rectangular 160-acre field with a notch cut out of one end. The notched end, about 20 acres worth, was a solid high-yielding green, while the rest of the field flickered through the shades of yellow and red that coded for lower productivity levels. "In 1994 this end grew 20 bushel/acre more than the rest of the field," he says. "This caught my eye. With beans at $5 a bushel, bringing the rest of the field to that level that would make a $14,000 difference." He had no idea what was taking that fourteen grand out of his pocket.

"In 1994 we were growing all beans in that field. When I checked the records from 1993, when we were growing corn, I didn't see the same variation. I knew that the year before that, in 92, we had planted beans in most of the field, but oats on that end," pointing to the region that was glowing emerald in the 1994 data set. "So it looked like we had a parasite that ate beans but not corn or oats, and that could survive one year without food but not two."

The farmer hired a soil testing company to probe for microscopic wildlife. (When they arrived he used the data from his yield monitor to plan the pattern of their sampling sites.) He found he had an infestation of a soybean-loving nematode that up till then had been encountered mostly in the South, but was known to be moving North. "This was the first time we'd seen it on that scale, distributed through a whole field like that," he says. Alerted by the monitor, Harford looked for and found the parasite on several other fields of his. He is considering his next step. He might use nematode-resistant seeds or try starving it out by growing corn two years in a row, skipping the rotation into beans. Both have their short-term disadvantages, but the evidence from the yield monitor is that the fight is well worth making.

On the other hand, sometimes the monitor makes problems go away. "Some of my acreage is leased," he says. "Recently I was driving through a field with its owner. We reached a section where the drainage is deteriorated a bit. The plants looked short and ugly and partly on that basis we decided to upgrade the system. "This would have meant replacing 8" with 15" diameter pipe and running the new size through fields belonging to three other people. It would have been complicated and expensive." So before Harford went ahead he checked the field on the monitor. "I found that we were only losing about $1500 of yield in that region. $1500 wouldn't even have paid for the interest on the investment needed to do all that work." The ugliness of the plants had been deceptive. "We used to fix things like this."

Some of the messages he sees in the data he is still thinking about. "What ought to be our best soils actually turn out to have the lowest yields. Maybe that's because the most crop has been taken out of the best soils, but the fertilizer has been applied evenly." If he settles on this hypothesis, Harford might add more fertility to the very soils that were supposed to need it least. On the other hand, he says he isn't seeing any direct relation between productivity and fertilizer distribution. "That's interesting," he says shaking his head. "That sure isn't what the fertilizer companies tell you."

Harford expects yield monitoring to play a big role in his technology management. Every year the seeds, machinery, and chemicals companies pump a small river of studies through Harford's fax machine, all showing the handsome increases in productivity awaiting any farmer canny enough to buy their products. "They always seem to claim to give you about six bushels/acre difference (of corn)," he says. "Does this mean that if I bought everything I'd grow 300 bu/acre?" (200 bushels would be an excellent year.) "I doubt it."

He knows that just because a product did something heroic 150 miles to the south last year doesn't mean it would do anything on his farm this year. It might not grow the six extra bushels in his particular region to begin with, and even if it did unrelated factors might depress production by six bushels, preventing the farmer from noticing the improvement. The only dependable way to evaluate a technology is to "amplify" the signal over time by using it on the same fields for many years running, but that necessarily retards the rate of technological innovation. Harford tries to stay up-to-date with new seeds -- he has twelve varieties working in his fields this year, the oldest being all of four years old -- but he sometimes suspects that his decisions are really being driven by the vendor. "We have our complaint of the month," he says; "the salesmen all sell to that. It's possible the movement is all lateral."

A farmer armed with a yield monitor is not such an easy target. "We've always had to take other people's data grown on other people's fields," Harford days. "Now we have our data, gathered from our fields. If a seed salesman promises me six bushels, and his seed doesn't deliver, I'll know in a year. And so will he. To a farmer, validation is power."

Only 200 farmers used the full Yield Monitor with GPS package in the 1994 harvest, but Harford is sure the device will spread quickly; he says the effect on revenues is just too great. (Myers says his sales have quadrupled this year, and John Deere has announced plans for a combine that will come with Yield Monitoring built in.) Further, the technology can only get more powerful, as on-the-fly testers are developed to measure different soil properties (nitrogen, pH, phosphates) continuously and equipment appears that can handle each field unit separately, matching the right seed to the right soil, applying quantities of fertilizer or chemicals that are tuned to each specific unit.

Perhaps in the near future, most farmers will be growing two crops, one measured in bushels and one in bits. The implications for agriculture are profound. Myers expects leading edge farmers to become far more experimental, since they will be able to prove out theories and put them into practice very quickly.

Harford suspects farm size might grow significantly, since yield monitoring will make it possible to handle far more variables than in the past. Farms have already been growing, from a national average of about 350 acres in 1967 to 500 acres in 1994, but Harford says he wouldn't be surprised to see consolidation proceed much faster over the next twenty years.

On a still deeper level, Harford suspects yield monitoring might help change the basic paradigm of farming to one where the farmer adapts to the soil instead of forcing the soil to adapt to him. He makes his point by pulling out a map. "We have these drainage pipes running under the fields. Sometimes the soil settles down between them, making parallel rills. They're annoying to run over, so last year I trimmed them down. And yield went down. Right there."

This was not a big surprise to Harford; he likes to point out that corn did not evolve in a world in which the ground was being plowed up every year. He thinks soil is more usefully seen as a biological tissue, organized around common systems and networks of exchange and circulation, than an inorganic commodity that can be torn up and ripped apart. But the issue has divided farmers; it's been an argument. Now he has the numbers. "The implications of yield monitoring for the way farmers do soil management are profound indeed," he says quietly.

Every technology has its dark side and yield monitoring also raises the possibility that crop farming might end up where poultry farming is now: so rationalized and controlled that it could be done by 9-to-5 corporate employees or even by robots. That won't necessarily happen, but it's possible that Harford has picked up a tool that will end by leaving no room in agriculture for a person of his skills and intelligence. It's an unknown; but of course if Harford was afraid of unknowns he would never have been a farmer in the first place.