FRUITA — When Perry Cabot looks at his 12-acre cornfield northeast of Fruita, he sees much more than a flat field among many other flat fields baking under a cloudless sky.
He sees a bountiful reserve of data and a way to fine-tune crop production in the drought-challenged Colorado River Basin using a new tool: artificial intelligence.
“This is the data-mining year. This is the year I’m looking for the nuggets,” said Cabot, an extension professor with Colorado State University, as he walked the field in April.
AI has long been in our cultural dialogue — e.g., robots waking up and taking over the world — and now it’s making its way into our hospitals, classrooms, Hollywood labor strikes and construction sites. Cabot and his research partners think AI has a place in agriculture, where the field of precision agriculture is already harnessing new technologies to boost ag practices.
The researchers are testing out their idea with a million-dollar, multiyear research project funded by the U.S. Department of Agriculture. If things go according to plan, they say their project could lay the foundation within two decades for a long-term shift in how farming is done in the Colorado River Basin.
“The thing I’m most excited about is we’re able to take something that’s extremely complex and then make it useful for farmers,” Cabot said.
What happens on farms and ranches in the Colorado River Basin has big implications for the rest of the basin, where the multibillion-dollar agriculture industry uses at least 75% of the basin’s water supply while growing food for the nation.
The basin provides water to millions of people. But its supply is dwindling, and water users are struggling to curb overuse in face of prolonged drought and climate change.
More efficient water technology would change how farmers operate their businesses, Cabot said. He emphasized that efficiency doesn’t mean conservation: Farmers could use their water savings to grow more crops, and support their livelihoods, instead of cutting down on water use entirely.
“I don’t like the word ‘minimize’ water. I hate that term. It’s optimizing,” Cabot said. “We’re trying to learn what nuggets of information exist to allow us to make decisions to optimize the water resources that we have.”
21st century irrigation
Optimizing resources on farms has a lot to do with weather, elevation changes, soil conditions, expensive equipment and irrigation systems.
A typical corn yield is about 200 bushels per acre, Cabot said. But a 10-acre field sees a ton of variety: Some acres produce 250 bushels, while others might only grow 150 bushels.
“If we’re applying the same amount of fertilizer and theoretically the same amount of water, then why are we not getting the same yield?” he said.
If a farmer puts 1 inch of water on an entire field, plants will have varying success because of the changing conditions in different sections. Water might run off higher areas of the field and gather in lower areas. Different soils have different absorption rates: Clay soils hold more water but absorb it more slowly, whereas water runs more quickly through sandy soils, Cabot said.
“We’re not following the trend in the natural environment. We’re not doing what the natural environment wants us to do,” he said.
But with modern technological improvements, farmers can design irrigation programs to apply varying amounts of water across their fields.
These varying rate sprinkler systems are about 80% efficient and are often programmed based on historical observations. Flood (furrow) irrigation systems, when farmers flow water down trenches running through their crops, are about 30% efficient, Cabot said.
But if farmers could adjust their irrigation practices according to a data-filled map of their cropland — or if an AI algorithm could do it for them — the whole process could improve. And that’s what Cabot and his research partners are trying to do.
At the CSU research station near Fruita, Cabot’s project is on schedule. They started collecting data this year after completing construction last year.
This is where Cabot’s nuggets come into play.
Cabot and his team took the seemingly flat 12-acre field and broke it into 15-by-40-foot zones, like a giant chessboard, he said. They’re harvesting every bit of data they can find: nitrogen levels, electrical conductivity, nutrient holding capacity, texture, water holding capacity and organic matter.
The team spent the spring and summer walking the field to gather measurements manually. They pulled data from satellite imagery, digital elevation models and spatial weather grids. “Salt sniffers,” devices that shoot electrical signals into the ground, measured the soil’s salt content in each zone.
In late spring, a system of pipes and ditches began delivering Colorado River water to the farm.
On a computer, Cabot’s team adjusted numbers on a spreadsheet tool to assign the amount of water for each zone. As the sprinkler system rolled over, hanging nozzles automatically changed how much water was released.
As the corn grew, the researchers monitored its growth stages, measuring plant leaf canopies and how quickly kernels appeared. Drones flying above the field mapped the vegetative index to determine how green plants are as an indicator of their health.
Similar study sites are set up elsewhere in the Colorado River Basin and Salinas River Valley in California. Researchers from the University of California Riverside, University of Arizona, Duke University and more are contributing their expertise and hope to wrap up the project in 2025.
The goal at this stage, Cabot said, is to figure out which variables impact crop yields the most. Next year, they’ll start feeding that data to AI algorithms to see if they can start getting better crops with less water and 10%-12% profit margins.
“We would never expect farmers to gather the resolution, the magnitude, the variety of data that we are gathering,” he said. “Our job is to grab everything and then grind it down.”
Could it work?
The farmers Cabot has talked to about the project aren’t really sure what to expect, he said.
“When was the last time anyone in the United States really liked anything with the word ‘artificial’ in front of it?” he said.
Even if the researchers can irrigate with more than 80% efficiency and keep profits at around 10%-12%, they’ll still face a critical challenge: making the technology easily accessible to agricultural producers.
Farmers would have to learn how to use the technology — one more task on already long to-do lists. The researchers plan to develop training programs and user-friendly web interfaces, including a smart phone app, to help ease the transition.
The program would also need to play well with the massively expensive equipment already being used on farms. Cabot’s research uses a Reinke Irrigation linear move system, the only variable-rate system of its kind in western Colorado, he said.
“The first 7 feet of this thing costs $30,000,” he said. The linear move itself cost about a quarter-million dollars, and the researchers might still install more improvements to their irrigation system to reduce water lost to wind.
The system also might be most effective on larger operations, like farms of 40 acres or more, Cabot said.
☀️ OUR RECOMMENDATIONS
Colorado’s largest operation, the Ute Mountain Ute Farm and Ranch in southwestern Colorado, is 7,600 acres. Simon Martinez, the general manager, was interested in learning more about Cabot’s research, hoping to farm more with less water, he said.
“I would want to look at it,” Martinez said. “I’m all for changing direction, but is that direction feasible for the long term?”
The Farm and Ranch is a multimillion-dollar operation that is already using many of the latest agricultural practices. Its staff is using variable-rate sprinkler systems, doing soil sampling and experimenting with more water-efficient crops. They even draw from hydropower and are considering incorporating solar technology.
Cabot said if the farm already uses variable rate irrigation, then his research would be able to help them operate based on up-to-date, changing conditions — like the health of plants in a particular zone at the time of irrigation.
“They’d be more nimble, they’d be more flexible, they’d be more real-time,” he said.