Pinpoint forecasting among the tools WSU brings to agriculture

This story is part of an AI series looking at how WSU is driving innovation in research and teaching through artificial intelligence. View the entire series as it becomes available.

We've come a long way from the Old Farmer's Almanac: These days, when farmers need to know about tomorrow's weather - or next month's weather - they can get a close-to-home forecast from Washington State University's AgWeatherNet.

That system does more than predict temperatures and precipitation. Drawing on data from 370 public-private stations across the state, it uses machine learning to power tools that help growers predict wheat yields, anticipate problems with pests, protect against frost and high heat, schedule irrigation, and more.

As the amount of location-specific data grows, it will fuel even more refined microclimate forecasting - down to the level of an acre.

"It is happening now and we are improving it further," said Lav Khot, the director of AgWeatherNet and a professor of precision agriculture in the Department of Biological Systems Engineering with a focus on agricultural automation. "That's where AI comes in: How can we mine the information from this data and make better management decisions? It's really changing the game."

WSU is playing a vital role in connecting big data with modern farming - marrying the strength of high-tech research with the public-service mandate of a land-grant institution to help farmers at every stage, from preparing for the growing season to harvest.

StockSmart, a tool developed at WSU in conjunction with the University of Arizona, uses GPS data to help ranchers and land agencies determine how much forage is available for livestock and where. A machine-learning guided metric developed in WSU Veterinary Extension uses data from dairy cattle to evaluate the disease burden among herds. WSU's smart orchard testbed in Mattawa is demonstrating how new smart agricultural technologies can be integrated and evaluated for meaningful adoption on the ground.

All helping to pave the way for the farm of the future.

WSU's role at the forefront of AI in agriculture is reflected in its leadership of AgAID, a $20 million institute established with funding from the U.S. Department of Agriculture as part of the National Science Foundation's National AI Research Institutes program. The Institute combines the efforts of multiple research institutions, industry, and government partners to address agricultural challenges in the Pacific Northwest.

The institute's mission is to build and sustain partnerships between AI and agricultural communities, driving research and the transfer of knowledge to meet some of agriculture's biggest challenges, including labor shortages, water availability, and climate change.

"Agriculture, globally speaking, is experiencing a revolution because of how much data is being generated and gathered," said Ananth Kalyanaraman, the director of AgAID and professor and director of the School of Electrical Engineering and Computer Science. "There's been a deluge of data. The immediate question when there is a lot of data is what you do with it - what is the value of data? The value is the information you can get to understand what is going on in the fields."

In addition to research, the institute has an educational goal - teaching students, obviously, but also helping introduce farmers and farm workers to new technologies. WSU and AgAID are in a prime position to do such outreach, based on the trust established over many years of the university's land-grant service.

"We need academia in the middle - academia is the way to connect big tech to a rural, societal challenge like agriculture," said Kalyanaraman, who also heads up the Advancing AI Research Working Group formed by the WSU Office of Research. "I think that education and workforce development very much have to go hand in hand with research when it comes to AI."

Forecasting streamflows

Most of the stream water that irrigates crops in the West comes from melting mountain snow. Being able to predict how much water the mountain snow will produce is crucial for farmers and resource managers.

One recent example is a new tool developed by Kalyanaraman's lab in collaboration with Kirti Rajagopalan, an assistant professor in the Department of Biological Systems Engineering, that could someday provide daily or weekly forecasts for water availability similar to weather forecasts, based on how much snow is in the mountains. Predicting how much water will be available is important for irrigation, flood prevention, fisheries, and hydropower.

In a recent publication, the WSU team demonstrated that their model was more accurate in predicting the snow-water equivalent - the amount of water available in the mountain snowpack - than current models in most cases across 500 locations, for both daily and weekly forecasts.

The plan is to eventually produce a public dashboard that agencies could use for both short-term and long-term planning.

That tool is just one of the ways that WSU researchers are working to build data to fuel AI applications to help farmers succeed.

"There are many different kinds of data we gather to understand different crop stresses and predict the crop stresses in advance, so the farmer can make better decisions," Kalyanaraman said.

WSU researchers are working to build data to fuel AI applications to help farmers succeed, including the ability to understand and predict different crop stresses.

Khot said that a key element of refining the precision of crop stressors forecasting is farm-level data - information on everything from soil conditions to tree canopies to local microclimates. The AI/ML driven decision support tools available through AgWeatherNet are already more precise than other forecasting models.

But the network is working with farmers to get them to contribute more localized information about conditions through AgWeatherNet Smart Farms and smartphone applications driven crowdsourcing.

"We are educating growers so they can collect better quality data on their farms that will contribute to realize AI-driven smart farms in the near future," Khot said.

As more and more such information is gathered, it will feed AI models to make ever more precise predictions.

"If you have the data sets that are collected for your block, you can synthesize everything and get the information in a quick second about your crop health, where the crop is doing well or poorly, and associated inputs management decision making" he said. "I'm just amazed at all the possibilities of what crowdsourced quality data and AI can bring into precision farming in the state of Washington."

Tina Hilding, the director of communications in the Voiland School of Engineering and Architecture, contributed to this report.