Eighteen days: how far the Burgundy harvest has moved

The Burgundy harvest has been dated for a very long time. The series Barrelheads holds, compiled by Daux and colleagues and archived by NOAA, gives a composite harvest date for the region in 605 of the 653 years from 1354 to 2006, counted in days after 31 August. It stops in 2006. Everything below therefore describes a shift that was already on the record twenty years ago, and says nothing about the seasons since.

Take the decade averages first. In the 1950s the harvest fell 28.5 days after 31 August; in the 1960s, 31.1 days, which is 1 October. The 1970s averaged 26.6 days and the 1980s 25.8, both in the last week of September. Then the date moved: 18.4 days in the 1990s, and 12.6 days in the seven years from 2000 to 2006, which is 12 or 13 September. Between the 1960s and 2000–2006 the average harvest came forward by 18.4 days. Nearly three-quarters of that, 13.2 days, happened after the 1980s.

The starting point matters, and it flatters the number. The 1960s were the latest decade of the twentieth century in this series. Measured instead against the average of the five centuries of data before 1900, which is 26.0 days after 31 August, the harvests of 2000–2006 were 13.4 days early rather than eighteen. Either way the recent years sit well outside the long run of the record.

The extremes make the same point. The latest harvest in the whole series is 1816, at 51.5 days after 31 August, around 21 or 22 October. The earliest is 2003, at ten days before 31 August, which is 21 August. No other year in the series comes within a week of it: the next earliest is 1420, at 28 August. Since 1950 the latest harvest was 1956, on 12 October, and before 2003 the earliest was 1976, on 9 September. In the fifty years from 1950 to 1999, three harvests fell before 15 September. In the seven years from 2000 to 2006, three did.

For the second half of the series we can set the dates against our own weather record: ERA5-Land daily temperatures for the Côte de Beaune and the Côte de Nuits, from 1950. For each of the 57 years to 2006 we took the mean temperature from 1 April to 31 August and compared it with that year's harvest date. The relation is close and steady. On the Côte de Beaune the harvest came 8.7 days earlier for each degree Celsius of April-to-August warmth; on the Côte de Nuits, 8.8 days. Temperature over those five months accounts for about 85 per cent of the year-to-year variation in the date. The result does not hang on 2003: with that year removed, the figures are 8.4 and 8.5 days per degree.

The individual years line up as the figures suggest. The warmest April-to-August of 1950–2006 across the two Côtes was 2003, at 18.6 °C, and it gave the earliest harvest. The coolest was 1956, at 13.4 °C, and it gave the latest harvest since 1950. The third warmest, 1976, gave the second earliest.

The decades can be put through the same arithmetic. April to August averaged 14.7 °C across the two Côtes in the 1960s and 16.2 °C in 2000–2006, a difference of 1.5 degrees. At 8.75 days per degree, that difference corresponds to about thirteen days. The harvest moved eighteen. The year-to-year relation with temperature therefore accounts for roughly thirteen of the eighteen days, and the remaining five are not explained by this measure. We do not know from these data what they are. A harvest date is a decision as well as a consequence of the weather, and a series of dates cannot separate the two.

Three limits should be stated. The harvest series is a regional composite built from many local records, not the date at any one estate or village. The temperatures are a gridded reanalysis, not a station reading. And the series ends in 2006: this year's picking is outside it, the series we use has no dates after that year, and we make no comparison with 2026.

What the two records show together is narrow and firm. Over more than half a century, a warmer April-to-August in Burgundy meant an earlier harvest, at a rate of close to nine days per degree, and by the first years of this century that had already moved the average date from the start of October to the middle of September.

Harvest dates: Daux, V. et al. (2012), Climate of the Past, 8, 1403–1418; data held by NOAA/WDS Paleoclimatology, doi 10.25921/2xg2-mt70. Weather: Contains modified Copernicus Climate Change Service information 2026.