A multi-year survey of 95 Swiss vineyard sites — 680 samples in total — paired microbiome profiling with untargeted metabolomics and found that berry-associated fungi showed the strongest site-specific signatures of any biological variable measured. Machine learning classifiers trained on fungal community composition could predict the microclimatic subregion of a vineyard with accuracy that exceeded predictions based on soil chemistry alone. The microbes on the grape, not the minerals in the soil, were the most reliable fingerprint of place.
This complicates the classical concept of terroir. For centuries, winemakers have attributed regional character to geology — limestone versus slate, clay versus sand, the depth of the water table, the mineral content of the bedrock. These factors do matter, but the Swiss data suggest they matter partly because they select for different microbial communities, which in turn produce different volatile metabolites during fermentation. The geological signal is real but indirect. It arrives in the glass not as dissolved minerals but as the metabolic byproducts of organisms that preferred those minerals' habitat.
The study also revealed an uncomfortable temporal dimension. Fungal communities on berries shifted measurably across the three years of sampling, tracking vintage-to-vintage climate variation. A warm, dry summer favored fermentative yeasts associated with distinct aroma profiles; a cool, wet one shifted the community toward molds that compete for substrate. Terroir, in this light, is not a fixed property of a place but a probability distribution — a tendency shaped by geography but modulated by weather, management, and microbial succession.
The implication extends to any domain that attributes outcomes to fixed causes. When a system's character emerges from the interaction between a stable substrate and a variable biological community, the “signature” of the place is never the place alone. It is the place as interpreted by whatever is living in it at the time.