According to NOAA (National Oceanic and Atmospheric Administration), as of mid-July, 2021, 89% of the U.S. West was in drought and 25% was in exceptional drought conditions.
An arrow pointing leftHome

How AI can help farmers fight drought in a hotter world

  • Mike Pearl

The agAID Institute, headquartered at WSU, is using AI to find hidden moisture in increasingly parched PNW soil.

As climate change makes water shortages increasingly frequent in the Pacific Northwest, droughts, like the one currently affecting over 70 percent of the region, will hit farms hardest. City-dwellers, tapping away at their keyboards in places like Portland and Seattle, could be forgiven for not noticing the drought at all, given the normal recent precipitation in those major cities. Even urbanites, however, will likely see the impact the water supply has on prices when they’re at the supermarket checkout counter.

But can AI help rescue farming in the Pacific Northwest? And moreover, might AI in the Pacific Northwest help save farming nationwide?

Apparently the federal government hopes so. Last year, the U.S. Department of Agriculture announced that it would carve off $4 million a year for five years to finance a collaborative research group headquartered at Washington State University, with Oregon State University and six other institutions, called the USDA-NIFA Institute for Agricultural AI for Transforming Workforce and Decision Support, or more concisely, the agAID Institute. IBM Research and a Walla Walla-based startup called are also partners on the project. Its mission, according to its website, is to “build and foster partnerships between AI and Ag communities, and create a transdisciplinary ecosystem for technology innovation and knowledge transfer.”

The agAID Institute hopes to apply AI to many areas of agriculture amid the anthropogenically changed climate. This will include a search for AI solutions to everything from the dangers of increasing temperatures on crops, to climate’s impact on farm labor. Two WSU researchers, computer scientist Kirti Rajagopalan and economist Michael Brady, are leading an effort to understand the impact of drought on what the institute’s director, Anantharaman Kalyanaraman, referred to in an email as “agricultural decisions.”

“Agricultural decisions” seems to be a key phrase at agAID. And one of those potential decisions, fallowing, aka “not planting,” seems to be on everyone’s mind. Artificial Intelligence can’t make it rain (yet), but according to Brady, AI has a role to play in “identifying patterns, and then using those to improve prediction.” The hope, he explained, is that farmers, armed with more and more granular information about the water supply, can make more and more granular decisions about water allocation, and thus, eventually become less reliant on fallowing to save resources, and improve crop yields even in drought conditions.

An agricultural economist in Washington, Brady, who grew up on a farm in Ohio, has identified two different troubling scenarios to focus on locally. If it’s a drought year, and you’re not one of the Yakima Valley’s most established farmers and thus entitled to the most senior water rights, you’re vulnerable to either water supply interruption — which is exactly what it sounds like — or water prorationing, in which your reduced supply will come in a constant, but less voluminous flow. Neither is pleasant, but they affect farms in different ways.

“Those are two very different worlds to be in,” Brady said.

An AI-based tool armed with information about a pinched water flow due to prorationing, or a period in which no water is going to be delivered at all, can, Brady hopes, be combined with information about what will grow best in what specific section in a specific field, under some specific set of drought conditions.

This may seem somewhat silly at first; we all know food grows where water flows. If there’s not going to be enough water, it’s a bad idea to plant anything, so it’s time for some fallowing. Conserve money and energy, and plant when the odds of a healthy crop are more favorable.

But, Brady pointed out, individual areas of any given field might be yellower than others year after year, and the system he’s hoping to build will know which areas struggle to produce healthy crops even in the best of times. “If you stand in some of these fields that are quite large, it can be hard to see that pattern,” Brady said. However, “The NASA Landsat satellites have been taking images for more or less every two weeks on every spot for 20 years.”

An AI might be trained on these millions of satellite photos to recognize stubborn blobs of brown during rainy years, or blobs of heroic green during drought years.

Better information means more advanced irrigation systems that, rather than just applying water uniformly — on a center pivot or a drip irrigation system — can process satellite images, find the spots that get drier, and, as Brady explained “combine those with soil moisture probes that automatically feed in real time data in addition to the historical record of all of these images and can adjust in an incredibly refined way exactly how much water is going in to each little part of the field.”

Longtime farmers might not be enthused to hear that AI is going to offer yet another complex and highly technical new solution to what is, essentially, the oldest and most primal problem in the history of agriculture: not enough water. Brady said he’s essentially describing “precision agriculture,” which is not an entirely new concept for farmers.

Companies have marketed and sold “precision” systems for seed and fertilizer application already, and while some have worked well, Brady said, in other cases “farmers have had them put on their combines and then just haven’t used them.”

A hypothetical future AI has to pull off a pretty fancy trick if it’s going to be useful: It has to produce a tool that instructs farmers in things like where to plant given a certain set of drought conditions, and that tool has to be easily accessible, navigable, and useful enough that farmers actually want to consult it.

To that end, there are two likely outcomes — the technology could be commercialized, and sold to farmers, or it could be made publicly available online, perhaps hosted on a university database.

If it’s commercialized, it would have to be worth it for farmers’ bottom lines. agAID is, after all, a federally funded research institute, not the research and development wing of some large agricultural company like Bayer-owned Monsanto, where researchers’ jobs are dependent on generating increased profits. Commercialization would, Brady said, involve “a significant additional step.”

Other areas of research under the agAID umbrella, such as labor, are finding it easier to show taxpayers what they’ve been up to, according to Brady. “It’s easy to see that when you’re selling a big machine that picks fruit,” he said. Commercial or not, agAID’s eventual AI-based drought technology is unlikely to be a big flashy gadget.

But for all Brady knows, agAID might produce nothing more tangible than a university-based resource for farmers like the ones that already exist on the WSU website. If AI produces unparalleled new insights that can be distilled down to a set of free instructions on, as Brady put it, “how to optimally refine your irrigation scheduling during a drought,” that would still spell success for agAID.