Gloria Ferrer vineyards a Sonoma County pioneer in using AI to monitor crops

As climate change brings about a more unpredictable climate, with heat waves, late frosts and drought, a Sonoma Valley vineyard has said it’s the first in Sonoma County to use artificial intelligence to more accurately monitor its crops.

Gloria Ferrer Caves and Vineyards uses AI to track growth, regulate water use and predict yields across its 330 acres of grapevines in the Carneros wine region in Sonoma County, Gloria Ferrer vineyard manager Brad Kurtz said.

“I just get so excited by the possibilities of the things that they are developing,” Kurtz said about Gloria Ferrer’s AI partnerships. “You wouldn’t be able to get this level of data doing it by hand. There’s just no way to get this sort of granularity.”

AI combines computer learning with large data sets that can be used for any number of disciplines, from language processing and speech recognition to image creation. Gloria Ferrer employs several different AI technologies, one of which creates an algorithm based on thousands of images of individual vines captured by a toaster-sized camera that routinely maneuvers through the vineyard. That algorithm will help the winemaker determine future grape yield.

Tom Shapland, CEO of Tule Technology, an AI company focused on irrigation planning, said AI can be simply described as “a computer system that can do something that we normally think of humans doing.”

An unyielding past meets the future

Vineyard managers have traditionally relied on historical references and human observations to monitor the health of vines and their potential yield.

Even the most thorough of vineyard managers cannot get a fully representative sample of what their ultimate yield will be. Kurtz said the “best crop estimators in the world,” who perform their estimates manually, have a margin of error of about 20% before crops have bloomed.

“We’re signing and authorizing contracts right now for what our yields will be during harvest,” Kurtz said while walking through rows of chardonnay grapes. “If you have a 10-acre vineyard, if you count 1% of the vines in that block, you're talking about maybe 100 vines. That's a huge undertaking.”

The yield of a vineyard is important in the winemaking process that comes after. Like a recipe for cake, the ratios of grapes need to be just right to create the intended flavor of a wine.

“If I think we need 1,000 tons of grapes to produce the wines that we can sell in a given vintage and all of a sudden, I do crop estimations, and there's 900 tons out there — we’re 100 tons of grapes short,” Kurtz said. “So I need to go out there and go buy some fruit or vice versa.”

But this year, Kurtz is getting data from a toaster-sized camera from Bloomfield AI, which creates and stores a digital copy of each plant.

The camera is attached to the front of a tractor and takes thousands of images of every plant. And in the same way that an AI-generated image is created using a trove of composite images, Bloomfield AI’s technology creates an algorithm based on the captured images to predict yields more accurately than manual approaches ever could, said Mark DeSantis, CEO of Pittsburgh-based Bloomfield Robotics, which has an office in St. Helena.

“It's taking a very, very detailed image and it's taking it at a pixel level,” DeSantis said. “It's looking for the same patterns and features that a human would when they're looking at your vine … They're looking at their size or color, how many clusters, leaves; the tendrils, the corridors and all those features.”

While Kurtz walked through chardonnay vines at Gloria Ferrer on Tuesday, he discovered a vine where he suspected a mouse had eaten the emerging buds. Nearby, a tractor with Bloomfield’s camera was making its first pass through the vineyard this season and captured the missing buds, updating the predicted yield that Kurtz could access with a finger swipe on an app.

“Yield monitoring is just the very tip of the spear of what this technology is going to bring to agriculture,” Kurtz said.

Drought and adaptation

A pressure chamber, also known as a “pressure bomb,” has traditionally been used to measure how tightly a vine is holding onto water, which offers an indication of how stressed a vine is — a critical measure to ensure the quality of a grape.

It takes a leaf from a vine, wraps it in a plastic bag, and applies pressure to it in a tank using nitrogen. When enough air pressure is applied, the leaf releases water and provides a reading that dictates how stressed the plant is.

“It's a pretty high-level skill to teach someone,” Kurtz said. “You have to have a very good technician to be able to go out and take readings. And like I said, it takes a lot of time.”

Yet, it answers two vital questions to vineyard managers: How thirsty are the plants compared to how thirsty a farmer wants them to be? And given that, how much does a plant need to be irrigated?

Shapland, Tule Technology’s CEO, said the use of AI will elicit a more accurate prediction of how plants will react to different levels of water using predictive models. Tule, one of Gloria Ferrer’s AI partners, uses an evapotranspiration sensor that measures the amount of evaporation out of leaves over an acre and provides pressure readings attached to them.

By utilizing these readings, an algorithm can indicate how stressed vines are.

The ability to predict the amount of water needed for a crop, Shapland said, will become even more important as California goes under a “water revolution.”

“There's the way that things were always done in the past: People turn on the water on May 15 and they put it on for eight hours in every field, every week until they harvest,” he said. “There's still a lot of people in California that still irrigate like that.”

The sustainable stewardship of water, and the increasing regulation surrounding the resource, will require farmers to turn over a new leaf. Some, like Kurtz, have already turned toward the future.

“You can walk down a vineyard row and just scan 500 vines down at vineyard row, and it'll take the average reading of each vine,” Kurtz said. “So you get a measurement that's both accurate and representative of your entire vineyard.”

Since its founding in Davis in 2014, Tule Technology has worked with hundreds of farmers throughout California to change farm irrigation practices with improved monitoring and predictive models.

As Sonoma County wrestles with less predictable weather systems that may threaten its marquee industry, Kurtz is hedging his bets across his vineyard through the use of AI.

“These things are the future and we’re excited to partner with these companies to be some early adopters with them,” Kurtz said. “It’s going to change the way we do everything.”

Contact Chase Hunter at chase.hunter@sonomanews.com and follow @Chase_HunterB on Twitter.