In the Bordeaux region of southwestern France, dozens of vineyards transform delicate grapes into bold red wine blends. Some bottles sell for thousands of dollars each. Prestigious chateaux boast the soil, microclimate and traditional methods that make their own wine superior, an inscrutable mix known as terroir.
“It's one of those terms that the wine industry likes to keep a little mysterious, part of the magic of wine,” said Alex Pouget, a computational neuroscientist at the University of Geneva.
Dr. Pouget is attempting to apply chemical precision to this je ne sais quoi. in a study Published Tuesday in the journal Communications Chemistry, he and his colleagues described a computer model that could identify which Bordeaux estate produced a wine based solely on its chemical composition. The model also predicted the year the wine was made, known as the vintage, with about 50 percent accuracy.
Although wine connoisseurs often claim they can distinguish between wines from the best estates, they rarely do blind tasting tests, he said. “People have been making these claims for decades, but we've never really had an objective measurement to prove this to be true,” he said.
Dr. Pouget grew up in Paris in a family that only drank Bordeaux (“You pretend Burgundy doesn't exist,” he said). As a young neuroscientist in the late 1980s, he studied the brain with machine learning, a type of artificial intelligence that identifies patterns in large data sets. He believed that these methods could be useful for the wine industry, but it took him 30 years to prove the idea.
He partnered with Stéphanie Marchand of the Institute of Vine and Wine Sciences in Bordeaux, which had created a database of 80 wines of different vintages from seven castles. The database contained the chemical signatures of each wine extracted from gas chromatography, an ancient and inexpensive method for breaking down substances into their molecular components.
The researchers trained an algorithm to look for common patterns in the chemical fingerprints of the wines. The results surprised them: the model grouped the wines into distinct groups that matched their geographic locations in the Bordeaux region. This demonstrated that the particularities of each estate had drastically influenced the chemistry of the wines produced there, just as winemakers have claimed for centuries.
The estates gave researchers permission to study their wines on the condition that they not be named. Dr. Pouget said that all the wines were part of the famous Bordeaux Classification of 1855a ranking instituted by Napoleon III to promote the best Bordeaux wines.
Dr. Pouget was surprised that the winemakers did not want to reveal their names, since the study's findings reinforced the notion that their wines were special. “I have scientific evidence that it makes sense to charge people money for this because they are producing something unique,” he said, laughing.
Independent researchers said the study was part of a wave of recent research using machine learning to decipher terroir. “This is where the field is going and where it needs to go to make sense of a lot of data,” said David Jeffery, a wine chemistry expert at the University of Adelaide in Australia.
For example, it has used machine learning to classify Shiraz wines from the Barossa Valley in Australia.
The focus, Dr. Jeffery said, “isn't just about what makes a wine great chemically.” The models could also help producers adjust their growing and winemaking practices to preserve the character of their product in years of unexpected weather. “This is especially important in a changing climate,” he said.
According to experts, another application of these models is to eradicate fraud, which is quite common among expensive wines. The producers have adjusted their bottles, labels and corks to make them more difficult to copy.
“If there are doubts about the origin of a wine, analyzing a wine from the estate as a reference point would probably allow us to know whether the wine is fake or not,” said Cornelis van Leeuwen, head of the viticulture and oenology department. in Bordeaux Sciences Agro.
The approach would likely work for any wine region, as long as the model has been trained on a wide variety of wines from different producers and vintages, said Dr. van Leeuwen, who was not involved in the new study. However, an open question is whether the model will retain its accuracy after a few years, he said.
Dr. Pouget, who has a large wine collection, hopes to repeat the study using some of his favorite types from the Châteauneuf-du-Pape region in southeastern France.
But among the best wines, he said, age is probably more important than provenance.
“I only drink old wine,” he said. “I think drinking things when they're under 15 is a bit criminal.”
