The next time you visit one of the many vineyards around the Finger Lakes region, take a closer look at the grape vines. While leaves may appear perfectly lush, healthy and green, there may be disease-causing viruses or fungi unseen to the eye.
Cornell AgriTech has a long history of collaborating with grape growers to improve vineyard productivity — in terms of grape breeding, development of climate-adapted grape varieties and pest and disease management. One professor researching the future of viticulture is Prof. Kaitlin Gold, grape pathology, who leads the Grape Sensing, Pathology and Extension Lab.
GrapeSPEC is one of the primary extensions responsible for grape disease management in New York state, specializing in digital viticulture technologies that could improve disease management decision-making.
Grapes are a crop susceptible to a number of diseases that, if left untreated, result in significant decreases in crop yields — which farmers have long tried to avoid. These include viral diseases like grapevine leafroll disease, or fungal diseases like Botrytis bunch rot and powdery mildew.
To minimize crop damage, it is crucial to detect diseases early and deploy management strategies, whether they be applying fungicides or eliminating diseased grapevines to stop the spread. However, this poses a significant challenge to grape growers, as symptoms can vary and take as long as a year to show up after infection — if they do so at all.
Gold and her lab work on grapevine disease sensing using imaging spectroscopy, or sensors capable of detecting grapevine diseases even when grape crops are asymptomatic. Although there may be no difference to the eye, the physiological processes of infected grapevines are affected at a cellular level, since their defense mechanism against the disease is activated.
According to Gold, these sensors are able to detect a range of light of 2100 nanometres, compared to the 300 nm range that human eyes are capable of. By capturing these subtle interactions of light with the plant, researchers are able to measure aspects of plant physiology that are otherwise invisible to farmers growing these crops — such as water content and starch concentration — in a non-destructive way.
“This technology was developed to tell NASA what Mars is made up of, so [it can detect] diverse aspects of environmental chemistry,” Gold said. “And we can leverage that together to accomplish early disease detection [in grapevines].”
Imaging spectroscopy at GrapeSPEC is utilized across a range of spatial scales, including proximal and remote sensing. Autonomous rovers are used in proximal sensing, through which they can scan the foliage in each row of the vineyard. In contrast, satellites are used in remote sensing, which can capture spectral data of the whole vineyard at once.
Gold noted that while remote sensing offers unparalleled coverage of vineyards, proximal sensing can provide specific biological information and detailed coverage of the foliage on the grapevines. Both complement each other in detecting diseases.
“They are like yin and yang,” Gold said. “They fix each other’s weak spots.”
Machine learning can be used to process large amounts of data that the sensors generate. Fernando Romero Galvan grad, a Ph.D. student under Gold, works on a remote sensing project in collaboration with NASA to detect grapevine leafroll disease in Californian vineyards.
Using known parts of healthy, symptomatic and asymptomatic grapevines, Galvan trains a machine learning model to distinguish between healthy and diseased grapevines — asymptomatic or not — based on spectral data that the sensors capture.
However, while imaging spectroscopy is a powerful technology, there are a number of caveats that come with it.
“One of the curses with spectroscopic imagery is that this dataset is so sensitive that lots of things affect it — the time of day data was collected, the weather conditions,” Galvan said.
Galvan added that even the location of data collection can heavily influence disease diagnosis. More data would need to be collected across different conditions to determine if the signal for disease can be detected despite data variation due to confounding factors, he said.
Galvan said seeing the rate of infection inspired him to help the grape grower community, but he also noted the gravity of plant disease.
“It’s also hard to be excited because this is people’s lives [and] livelihoods that are going up in flames due to a virus,” Galvan said.
For Gold, conversations with growers and other stakeholders in California drove her research for detecting asymptomatic grapevine leafroll virus.
“It is very much like a two-way street,” she said. “[T]heir goals have really defined the research that we do.”
While imaging spectroscopy in early disease detection is a huge technological advancement, it is not meant to replace current disease management efforts, but rather to complement it, according to David Combs, a research support specialist at GrapeSPEC. Having such a way to detect disease at growers’ disposal will help inform ground-based solutions and allow them to target diseased grapevines precisely and efficiently.
“Having this technology to advance towards disease control and having fewer toxic materials in the environment — it’s great,” Combs said. “These things were unheard of 20 or 30 years ago, and now, the idea is becoming reality.”
Gold expressed confidence that this technology will be essential to ensuring environmental and financial sustainability of the agricultural sector in the face of climate change.
“Precision management technologies, digital agriculture, [non-destructive sensing],” Gold said. “This is the future.”
Dan Hong Loh can be reached at [email protected].