How season timing works
Growers want to know if a variety buds, flowers and starts ripening before or after another. This page explains where our answer comes from and how far to trust it.
What we measure
Three stages: budbreak, bloom and véraison (berries start to change). We compile 2,508 published dates, covering 173 varieties on the main scale, from 72 combinations of stage, place and year, at 7 places, from 1992 to 2021.
- USDA GRIN: Geneva, NY and Davis, CA collections, the largest source
- University of Vermont: yearly sheets, South Burlington, 2009 to 2014
- Iowa State University: Gilbert, Iowa, 2006 and 2011 to 2014
- CRAM Mirabel: Abbaye d'Oka vineyard, Quebec, 2010 and 2013 to 2015
- UMass Fruit Notes: Belchertown, Massachusetts, 2019 and 2021
- University of Wisconsin: West Madison, 2021
- Cornell, Willsboro trial: New York, average of 2010 to 2014
- Nova Scotia study: Annapolis Valley, average of 2017 to 2019
How seasons are lined up
A date alone says little: a warm spring makes everything early. So each season stays separate, and a variety's place is read against the others measured that same season.
Sources are connected through the varieties they share. If one vineyard only has X, E and F, and another has E and F with Marquette, we can place X among all of them. Two seasons join only if they share at least 3 varieties.
Some seasons also spread varieties further apart: at Geneva, budbreak spread about twice as wide as usual in 2012. The model allows for that, but only trusts it when a season has many varieties in common with the others.
What we set aside
Published data has errors. Everything we remove is listed in our files with the reason.
- Vidal's véraison at Geneva: up to October 7, far later than other measurements of Vidal. We use the Massachusetts values instead.
- Duplicate entries in one 2009 Davis study: 19 varieties listed twice, the second copy exactly 31 or 32 days later. We keep the copy that matches Davis's other years.
- Where a variety was measured comes from the study, not from where its plants are kept today.
- Values very far from what the rest of a season predicts are shown as hollow marks and don't count. Each one is judged against that season's own noise, so a messy year isn't punished.
- Multi-year averages are never treated as a season. They only place varieties with no other data, such as Sabrevois, and those are labelled.
How accurate is it?
We hide values the model has not seen and try to guess them. These are typical errors in days, half of the guesses being closer.
| Stage | Variety's average date | With seasons aligned | New source with only 3 shared varieties |
|---|---|---|---|
| Budbreak | 7.5 days | 2 days | 2.7 days |
| Bloom | 3 days | 1.3 days | 1.6 days |
| Véraison | 5 days | 3.2 days | 5 days |
In practice: don't read much into a gap of 2 or 3 days. Bloom is the most reliable stage. Véraison is the least, and one bad year can still move a variety that has only a few seasons.
In separate tests, predicting a new year at a known place worked best stage by stage: heat since March 1 for budbreak (about 3 days off), heat since budbreak for bloom (about 2 days) and simply days after bloom for véraison (about 4 days), when the earlier date is known. The chart doesn't use these predictions.
What it can't tell you
- Quebec isn't fully tested. Oka is our only Quebec vineyard, and its seasons use different stage definitions, so we can't properly check our predictions there.
- Davis is separate. Its varieties, mostly vinifera, share too few varieties with the rest, so they have their own scale and aren't comparable.
- Definitions differ. "Véraison" can mean first color at one place and soft berries at another. Lining seasons up absorbs part of that, not all of it.
- The chart doesn't use weather. It compares varieties with each other, season by season. It says nothing about ripening after véraison, when cool autumns slow everything down.
- Few seasons, less certainty. Each variety shows how many seasons and places are behind it. One season is a hint, not a verdict.
We're just trying to help: this is free, public data from other people's vineyards, and yours will have its own opinions. Please talk to local growers and your nursery before planting 2,000 vines because of a dot on a website; we can't take responsibility for them (or for what the neighbours say).