The white wave: when the acacia blooms in every Romanian commune

How fast does the white wave travel?

On the plain the date changes by about 1.7 days per degree of latitude, so the front advances roughly 65 km a day towards the north. From west to east it shifts by only 0.4 days per degree, so the front comes mostly from the south, with little west-to-east difference. On a slope the speed is of another kind: each 100 m of height adds about 3.3 days, so a hill 500 m above the plain blooms about 17 days after it.

Median typical full-bloom date (dot) and its middle half (line) for land in the acacia belt, by 100 m of height. Modelled.

How are these dates known?

No instrument records an acacia’s flowering date in every commune. The date is computed from temperature: degree-days above 2 °C added from 1 March reach a set total at full bloom. The temperature is the 0.1° E-OBS record corrected for height; one set of constants was fitted to 225 dated photographs of flowering acacia on iNaturalist (region 41–50.5 °N, 13–30 °E).

• Against the photos. Re-fitting with each year left out, the root-mean-square miss is 8.6 days (on the full fit, half of the photos are within 5 days), against 14.3 days for guessing one date for every photo. The model also follows the early and late springs: the correlation between the yearly medians of the photos and of the model is 0.95 (8 years, on the photos used for the fit). In Romania alone (22 photos) the miss is 12 days.
• Against beekeepers. Public posts by Romanian beekeepers, with a day, a stage and a town, were kept apart and never used in the fit. For 13 “in bloom” reports the median is 0.9 day earlier than the modelled date (range: 13 days earlier to 12 days later), and for 14 “starting” reports flowers were open a median of 9 days before the modelled date, as expected for the start of a flowering that peaks later.
• Against satellites. A satellite check found no flowering signal. In Sentinel-2 images, the visible brightness at 80 acacia records and at 80 oak records matched by place and height showed no extra brightening during the flowering window in 2021, 2024 or 2025: the difference in reflectance was +0.004, −0.003 and −0.008, against a bar set beforehand of +0.008 with an interval above zero. Each point is a 30 m patch of mixed trees, so a signal in pure stands is not ruled out. The Copernicus HR-VPP product, now on WEkEO, gives the start of the season, which is leaf-out, not flowering.

Each dot is one dated photo: modelled date at its place and year (across) against the date on the photo (up). The band is ±8 days; orange dots are in Romania.

How far the map can be trusted

• Dates are modelled. The photos and reports that check them are real and few; a commune’s date in a given year can be a week or more off.
• The temperature record has 10 km cells. The detail along valleys comes from the height correction, not from thermometers.
• Where acacia grows is not drawn from a stand map. The glow fades out between 500 m and 1,000 m, following where the recorded acacia stops. The 2019–22 survey shows where botanists walked, not how much acacia there is.
• The flowering shape is partly estimated. The opening about 9 days before full bloom comes from beekeepers’ reports; how fast the flowers fall is an estimate.
• Honey, frost and hive moves are not modelled.