Lessons from the Field

How precision agriculture is pioneering analytics.

Farms are using analytics to solve unlikely problems. Intel is part of a broad network of researchers and developers working with tools such as sensors and drones to help them. These emerging agricultural applications may trigger some new thinking about one of your business problems.

When Nathan Stein hears the term “Internet of Things,” he thinks of corn and soybeans. These plants’ second-by-second adaptation to weather and soil conditions produces a nonstop stream of data that help him better run his family’s Iowa farm. Using analytics software developed for farmers, he can simulate the impact of water, fertilizer, and pesticide adjustments.

“I can basically virtualize the entire crop,” Stein says.

Stein is among a growing number of farmers using real-time data collection and computer-based analysis. Thanks to farmers like Stein—as well as researchers and companies developing technology for them—agriculture, the oldest of human industries, is becoming a prime testing ground for sensors, drones, and big-data analytics.

These methods are helping farmers increase yields, margins, and efficiencies on a massive scale—goals of every industry.

Something that works “in the context of large-scale farmscould allow for that application into other domains,” says Vin Sharma, director of strategy, product, and marketing for Big Data Solutions at Intel.

For example, a retailer could use a single-function foot traffic sensor to replace video analytics in measuring and improving the effectiveness of in-store displays. A fulfillment center manager could embed a sensor on a general-purpose drone to check inventory. And across many other industries, CIOs could implement sensor-derived data analytics to precisely control corporate resources ranging from raw materials to computing power. Targeted control promises efficiencies not only within the company, but potentially all along the supply chain.

I can basically virtualize the entire crop.

“The data collection, the model development and deployment, the analytic capabilities—we see that all of those activities are very similar across industries,” says Sharma. In response, Intel has built a Trusted Analytics Platform cloud service for building analytics applications, and also opensourced its components and the glue that makes them work together.

“Companies don’t have to reinvent the wheel every time; they can re-use common elements and then stitch their own domain expertise” to create industry-specific commercial or proprietary applications, he says.

Back on the farm, innovators of precision agriculture, as the field is known, are working with data analytics to resolve farming’s biggest challenges.

The Utah Water Research Laboratory at Utah State University, for example, is testing a drone called the AggieAir that director Mac McKee developed with funding from the U.S. Department of Agriculture.

A one-hour flight with a typical payload of three cameras—RGB, near-infrared, red-edge, or thermal-infrared—produces about 200 gigabytes of image files that, combined with proprietary software, could help a winemaker estimate the amount of water each grapevine on his vineyard needs, making watering more efficient and saving them money. McKee says flights of the next-generation AggieAir will produce about a terabyte of data.

And at the University of California at Davis, where Intel has funded research, Shrinivasa Upadhyaya is developing an in-field leaf monitor that uses a thermal-infrared sensor to detect a plant’s transpirational cooling.

The sensors, which account for environmental factors like ambient temperature, relative humidity, radiation, and wind speed, send data in real time to desktop computers and mobile devices, which in turn analyze the data using software Upadhaya and his students developed to determine which areas of a field need more or less water at a given time. Such site-specific data can help farmers use just the right amount of water needed, illustrating analytics’ ability to allocate resources and reduce waste.

“The availability of increasingly less expensive and “The availability of increasingly less expensive and more powerful sensor and analytics technologies is helping farmers to watch over their terrain more effectively,” says Chris Seifert, director of data science at San Francisco startup Granular, which makes cloud-based software for managing farms. “Farmers can have a much more thorough understanding of what’s happening on their land without having to go out and visit each acre each day.”

We anticipate that the data center and the edge devices are going to evolve together.

Granular uses APIs to upload sensor data from things like irrigation systems, tractors, and farm implements to Amazon Web Services. The biological systems of densely planted fields of corn, Seifert notes, “represent some of the ultimate things to connect and to be able to monitor remotely.”

Intel’s Sharma points out that this volume and variety of data types is “several orders of magnitude greater” than in traditional farming, when data collection meant a thumb in the soil. That’s why advanced analytics are necessary to put the ‘precision’ in precision agriculture.

“The goal is essentially yield management. You want to figure out which variables are most responsible for increasing the output of the farm,” he says. “Human beings aren’t going to sit and build a statistical model with a thousand dimensions. You’re going to have to use machine learning to do that effectively.”

Since 2010, Iowa farmer Stein has used aerial imagery from satellites and planes to detect information such as elevation, temperature, soil moisture, and chlorophyll levels. He exports images and data to mapping and analysis software from senseFly*—he works for the Swiss company as liaison between the company’s customers and engineers—to identify unhealthy areas of his crops.

One thing Stein has observed through the process of collecting and analyzing data is the extent to which conditions on his farm can change throughout a day. As the sun’s angle changes, and heat builds up in the ground, “You see a shift in thermal data, and you see the transpiration of the plants pick up and drop off,” he says.

The data Stein derived from the overhead imagery of his family’s farm “quickly showed us in the spring just how much damage not [installing] more drainage...was costing our corn field”—almost 40 bushels an acre.

“This fact alone triggered us to spend thousands of dollars to put in a new main and laterals, to adequately drain waterlogged soils,” says Stein.

Soon, he plans to soon use the senseFly* drones and software to optimize fertilizer distribution on his farm. Using senseFly’s software and a post-flight drone map, he could program a self-driving tractor to distribute a prescribed amount of fertilizer throughout the field.

Devices such as smart drones and autonomous tractors raise the question of where the intelligence lies and where data processing will happen: in equipment at the edge of the farm, or in a cloud-based data center? Sharma says the answer is both.

“There is a somewhat specious either/or argument in some parts of the industry,” he says. “We anticipate that the data center and the edge devices are going to evolve together.”

Sharma gives the human nervous system as an apt metaphor. You want enough reflexive intelligence at the edge to pull your hand off a hot stove without having to “think” about it. But central intelligence of the brain can help improve or override actions to create higher-level value. Future farms will pair smart semi-autonomous devices with cloud-based central command system that benefits from analyzing data across many locations.

Stein echoes that point. “Agriculture data’s very timely, and it has to be captured at a very precise moment, and it has to work every time,” he says. On today’s farm, he adds, a farmer is a “connoisseur of data.”

Download ‘From Data to Action’