Hey all — I’ve been working on a wine project for a while and wanted to share it here to get some feedback from people who actually care about this stuff.
It’s a web app that tries to map and score terroir at the site level using climate and land data (heat accumulation, stress factors, soil, etc.), and then connects that to which varietals and vintages should perform best at a given location.
You can explore a map of vineyard sites and toggle different components (overall terroir score, heat, stress, smoke exposure, etc.), and each site has a breakdown with:
a short description of the growing conditions
which varietals seem best suited there based on the model
vintage-level scoring based on weather patterns
The goal isn’t to replace tasting or producer knowledge, but to give a more structured way to reason about why a place produces what it does.
It’s free to use, no login required:
I’d genuinely appreciate any feedback — especially from folks who know regions deeply or think this kind of modeling is flawed. Happy to talk through how it works too if anyone’s interested.
Very cool, and thanks for sharing. Would you mind explaining the source of the 6 inputs to your score. Where do they come from and what is their level of resolution? Also, if you could explain some of the methodology behind how the score is produced, that would be appreciated.
Edit: seems like there are a lot of data quality issues when you click in to some of the regions. There’s a “SECRET GARDENS WINERY & FARM, FLORIDA” near Orinda, CA. An “ALASKA LEGENDS, ALASKA” on Maui. Seghesio is in the middle of the San Francisco Bay.
On the data side, it’s a mix of publicly available climate datasets, gridded weather, and soil databases, all normalized to a consistent grid. Resolution varies a bit by source, but generally lands somewhere in the ~1–5km range for climate, with soils being a bit coarser.
Methodology-wise, I’m not trying to predict quality directly, but rather how well a site’s conditions align with what different varietals tend to prefer. The score is essentially a composite of those factors, calibrated against known regions and benchmark sites to keep things grounded.
Still very much a work in progress — a lot of the effort has gone into making sure the outputs behave sensibly across regions rather than optimizing any one metric. And heard on the data quality - looking into ways to more reliably geolocate.
Reviving this a few months on — the tool’s come a long way since April.
The biggest change is it’s not US-only anymore: there are vineyard sites mapped
across ~23 countries now, with appellation boundaries, so Europe gets the same treatment the US did.
There’s also region and varietal pages where you can look in aggregate at the most favorable growing regions to really drill down where within an Appellation to try or new places for your favorite varietal.