How a satellite-robot-AI trio is helping to improve vineyards one grape leaf at a time

While consumers worry about the taste of corked wine, growers fret over disease. It’s the most limiting factor to wine production, managing it is expensive, and climate change is making it worse. But technology is driving the next agricultural revolution. And a group of growers and researchers are harnessing new tech’s power to build a global disease surveillance program that can help reduce soil-borne plant disease’s harmful impacts.

Dr. Katie Gold’s Grape Sensing, Pathology, and Extension Lab at Cornell University (Gold Lab) is leading the charge on plant disease sensing, a modern field of research that combines plant pathology, machine learning, and remote sensing to detect diseases at its early stages. The sooner you catch the first signs of plant disease, the more you can save. But you can’t manage disease without seeing it. And like someone wearing a monocle, an optical upgrade is long overdue.

The majority of disease detection methods require in-person observation and analysis. But agriculture today has grown far too large for this kind of detailed treatment. To modernize the process, the Gold Lab uses autonomous robots from the Cyber-Agricultural Intelligence and Robotics Laboratory (CAIR Lab) and spacecraft to monitor grapevines through hyperspectral imaging. Under the supervision of humans, roving robots on the ground and satellites orbiting above are expanding what’s possible for the field and revealing new insights formerly unseen.

On four acres of vineyards in the center of Finger Lakes wine country, the Gold Lab studies plant pathology with the enthusiasm of a master sommelier tasting a vintage. One recent study investigated how effectively high resolution satellite imagery could detect downy mildew. This harmful disease splotches the vine’s normally green leaves with yellow and brown hues. The less green the leaf, the more severe the disease. Satellites can detect plant health by observing wavelengths of light reflected by vegetation. This reflected light can then be used to calculate vegetation indices, including NDVI and EVI, the latter of which is often more accurate for monitoring agricultural systems. In other words, the satellites spot disease as disease spots the leaves.

To see how well the satellites stacked up to the measurements made by the robotic Phytopatholobot and humans, the three departed for the vineyards (humans by foot, robot by wheels, and satellites by orbit), collected data, and compared findings afterwards. Results showed that severity ratings by humans correlated to both robot and satellite EVI, but that the robots’ and satellites’ ratings weren’t on the same page.

This problem, however, is actually an opportunity. The Gold Lab and the CAIR Lab are now exploring an AI approach that helps sync all these data sources. The ultimate vision is a feedback loop wherein the satellites spot downy mildew in the vineyards, lead the robot to the affected locations to take further measurements of the grapevines, and then tell the satellites how accurate they were. The cycle then repeats, but with improved results.

The team is also incorporating PlanetScope 8-band imagery at 3m resolution to add more spectral channels to the mix. They anticipate that all these information layers will help enhance their ability to detect, manage, and prevent plant disease.

Wine is a product of its grapes. And its grapes are a product of their environment. Meaning a wine’s flavor profile can drastically change when grape growing conditions are altered. This presents a challenge for vineyards working in the era of climate abnormality. How do you keep your vineyards healthy when the very air around them is changing, or prevent them from accruing the undesirable smoke-tainted flavor as wildfires rage nearby?

While there isn’t one panacean technology to address these questions yet, places like the Gold Lab are pioneering use-cases like the AI-robot-satellite trio to improve global food security one grape, vine, and vineyard at a time.