(Photo illustration by TJ Mullinax/Good Fruit Grower)
(Photo illustration by TJ Mullinax/Good Fruit Grower)

Washington State University’s efforts to foster collaboration between its engineers and agriculture scientists recently received a $20 million federal investment. The funding will support development of a new digital agriculture institute, bringing together over 50 research and extension staff, including from 10 partnering universities, to bring artificial intelligence expertise to agriculture challenges. 

“This is not a project, it’s an institute,” said Ananth Kalyanaraman, a WSU computer science professor and director of the new AgAID Institute. “We are trying to build an ecosystem here.”

The ecosystem includes horticulturists and agricultural engineers, as well as experts in many aspects of computer science, AI, human-computer interaction and water policy. The mission is broad, encompassing specialty crops down the West Coast, but Washington’s tree fruit and grape crops will be a major focus. Unlike a grant project, with clearly defined objectives from the outset, the U.S. Department of Agriculture and National Science Foundation-funded AI institute plans work with several thrusts: farm intelligence, labor intelligence and water intelligence. 

“It’s not a static proposal; I think it will deliver products that are not as tangible today,” said WSU extension specialist Bernardita Sallato. “That’s why it’s called an institute. It’s a platform for evolving and creating new expertise.”

Within each of those arenas there are some clearly defined projects, Kalyanaraman said, including some building off of existing tree fruit industry-funded research into orchard frost prediction, smart irrigation and automated pruning, among others. Further areas of need and opportunity will emerge as the scientists collaborate across their own disciplines and with growers through organizations such as the Washington Tree Fruit Research Commission. 

“The innovation really needs to be on what exactly the problem is you need to solve, and that comes from working closely with end users, be they farmworkers or orchard managers or water districts,” said Alan Fern, an Oregon State University computer scientist and leader of OSU’s AI team for this project. 

Artificial Intelligence 101

This is far from the first attempt to use AI to solve agricultural problems. But thanks to head starts from teams at WSU and OSU, a foundation has been laid for the new institute to hit the ground running.

“This effort provides a much wider expertise and infrastructure that we will benefit from, and a chance to work with computer science and machine learning experts that with smaller projects we regularly conduct, we don’t have the resources to include,” said Manoj Karkee, an associate professor for biological systems engineering at WSU’s Center for Precision Agriculture and Automated Systems, who leads the institute’s labor intelligence team working on automation of farming tasks, such as orchard pruning. “We have a lot of knowledge and foundational work done. With this collaboration, we want to bring it to the next level.”

The timing is right, Kalyanaraman agreed. 

“With data being our primary currency in agriculture now, I think it’s a ripe area for AI and machine learning techniques to help us understand complex problems,” Kalyanaraman said.

AI may be starting to feel like a buzzword. The best way to think about it: an umbrella term for a suite of computational approaches that can make systems smarter. 

“One way to get that intelligence is through machine learning. The key word there is learning, and we need data to learn. We look at past observations of what has happened in the field and use that data to learn what the system was doing,” Kalyanaraman said. “Essentially, the data presents an incomplete view of how a farm is evolving over a season. How best to leverage that data is going to come through machine learning.”

Fern encouraged the agriculture scientists and stakeholders to view AI as a suite of tools for doing more with data, not a magic wand. 

“I don’t like to say ‘AI,’ I just like to talk about technology that can do things better than we’ve been able to do before,” Fern said. “AI paints a picture of magic in some cases, and I don’t like that.”

On the orchard automation front, Karkee said AI can help in several ways. One project will be looking at orchard tasks that are already mechanized, such as tree shaking nut harvest, and automate aspects of it to improve worker safety and productivity. Another project aims to advance the work his team is already doing around automated pruning — using advanced algorithms to make more nimble and adaptable the existing prototypes that analyze tree architecture and make pruning decisions.

Research examples

Many ongoing areas of research for WSU’s orchard engineers and horticulturists will be looped into the institute’s efforts. On the farm intelligence thrust, better frost prediction and sensor-driven orchard irrigation are among the first projects the team will tackle. 

Existing approaches to predict frost risk and cold hardiness just can’t capture all the variability that new, AI approaches can use to reduce uncertainty, Fern said. 

“They use handcrafted, scientific models to make these predictions, but they are not very sensitive to specific locations and not able to leverage other sensor data you might want to put on your farm to improve predictions,” Fern said. “We can start with their models, that people are using right now, and using the AI techniques to learn the error of those models — using the data more aggressively than we already are — we can basically correct those models.” 

With this approach, the more data, the better the model. So, the team also plans to set up a system for growers to input frost data for their specific locations — knowing that, as a result, the model will deliver more accurate predictions for them in the future, Fern said. 

When it comes to sensor-driven irrigation, WSU physiologist Lee Kalcsits collected data on several different technologies at several orchard sites last year. Working with computer scientists will offer new ways to analyze that data and start to develop smart irrigation models, he said.

In vineyards, there’s even more data to work with, as Karkee and WSU viticulture professor Marcus Keller have been collaborating for several years on remote-sensing technology for irrigation. 

All that data makes it a very promising arena for the AI experts to collaborate, Fern said. The challenge will be organizing and understanding the sensor data available so that end users, growers or irrigators, will want to use such a system.

“What I’d like to envision is the manager of an orchard or vineyard would be able to say, ‘Hey, I’d like the water stress of the plant to have this profile and I want the system to tell us what the water schedule should be,’” Fern said, “and it should be adaptive based on what the weather is and optimized based on previous years and the current year.” 

The technology approach to building that smart system could then be transferred to other crops, he added. “Once you can nail it for one type of crop, you just change the data inputs.”

In the arena of water management, the focus will be more regional. The effort will zero in on modeling water needs for agriculture in both the Columbia River Basin and in California, with collaborating water experts at University of California, Merced, and complex modeling experts at University of Virginia.

“Water is all about supply and demand and demand impacts downstream. It’s a pretty complex question to think about with climate change, crop change,” and all the instream and out-of-stream needs for water, said Georgine Yorgey, associate director of WSU’s Center for Sustaining Agriculture and Natural Resources, who works on water and climate change issues. 

For example, if climate change extends growing seasons, some growers might double-crop and use more water while others will remove crops that no longer thrive in the local climate. Water modeling needs to account for both the complex physical systems and the human ones, she said.

“We’ve pushed the tools we already have to their limits,” Yorgey said. “Bringing in people working on advanced data analysis and AI will give our team new capabilities.” 

—by Kate Prengaman