How it works
Every page shows a modelled start of spring and compares it with that city's 1991–2020 average. This page explains why it is modelled, how the average is calculated, and why it covers those years.
Why a model, not observations
Nobody records the first bloom in every city every spring, and the records that exist come in late:
- Germany's weather service collects its lilac observations once a year. On 18 October 2026 its open data held 283 for this spring, against about 910 for each of 2024 and 2025.
- In Sweden, open lilac records start in 2008. Near the cities on this site there is a median of 3 to 6 a year, in some years only one.
The model needs only daily temperatures. They arrive about five days late and go back to 1940, so every city gets a date for every spring, worked out the same way.
The observations are what the model is checked against: from 1951 in Germany and from 2008 in Sweden.
The calculation
- For each of the 30 years in 1991–2020, the model gives the city a start-of-spring day from that year's temperatures. Days are counted from 1 January, which is day 1.
- The average is the plain mean of those 30 days.
- A year's "days from average" is its own day minus that average. A minus sign means earlier, a plus sign later.
The average is of the model's own results, not of observed bloom. Both sides of the comparison come from the same model and the same temperature data, so an error that is the same every year cancels out.
| City | Average day of the year | Shown as |
|---|---|---|
| Cologne | 111.7 | 22 April |
| Frankfurt | 113.1 | 23 April |
| Munich | 119.6 | 30 April |
| Berlin | 121.4 | 1 May |
| Hamburg | 122.7 | 3 May |
| Malmö | 134.8 | 15 May |
| Gothenburg | 139.7 | 20 May |
| Stockholm | 148.1 | 28 May |
Why the date says "around"
The average is rarely a whole day, so pages show the nearest whole day, as a date in a 365-day year. In a leap year the same day of the year falls one calendar date earlier.
Differences are calculated from the unrounded days, not from the dates shown. A date and its difference can therefore look one day apart.
Why 1991–2020
- It is the World Meteorological Organization's current climatological standard normal: the most recent 30 years ending in a year that ends in 0.
- The USA National Phenology Network compares against the same years in its maps of spring in the United States. Its averages come from different temperature data.
It is a recent period, and spring already came earlier in it than before. In the model, the 1961–1990 average falls 4 to 11 days later than the 1991–2020 one in these cities. "The same as average" means the same as a recent spring, not an earlier one. The World Meteorological Organization keeps 1961–1990 as its reference period for long-term climate change.
Observed lilac bloom in Sweden starts in 2008, so the Swedish averages for 1961–1990 are untested.
Those observations are based on data from Deutscher Wetterdienst (CC BY 4.0) and from SLU Naturens Kalender (CC0).