How to read the dates
This page explains what the dates and numbers on the results page mean and how to use them in the garden. The examples are for Manole, a village near Plovdiv, using the data for 1996–2025.
The two pairs of dates
For every place you get four dates: a typical and a safe date for the last spring frost, and a typical and a safe date for the first autumn frost. They apply to the whole area of about 9 × 9 km that your place falls in. Spring and autumn are worked out separately, each from its own years — some areas have enough spring dates but not autumn ones, or the other way round.
The typical date is the middle one: the dates from all the years are put in order from earliest to latest, and the one in the middle is taken. So in about half the years the frost came by that date. For Manole, 16 of the 30 years had the last spring frost on or before the typical date, and 16 of the 30 had the first autumn frost on or before it. When two dates share the middle, we take the more cautious one: the later one in spring and the earlier one in autumn.
The safe date is the more cautious one. In spring it is the date after which there was no more frost in 9 of 10 years. In autumn it is the date before which there was no frost yet in 9 of 10 years. The two dates are calculated separately and promise nothing about the whole year at once: for Manole, 3 of the 30 years had frost after the safe spring date (April 11), 2 of the 30 had frost before the safe autumn date (October 30), and 25 of the 30 had neither.
In short, for Manole over 30 years of data: typical March 29 / November 25, safe April 11 / October 30.
Tip: for your first spring in a new place, go by the safe date. It is usually 7 to 14 days after the typical one, and the exact number for your place is on the results page. Before planting, also check the weather forecast and what the plant needs — no date promises there will be no more frost.
The chart
On the chart each year has two marks: a circle for the last spring frost and a diamond for the first autumn frost. The months are on the left.
Click or tap a year on the chart to see its dates. On a keyboard, use the left and right arrow keys.
The chart always shows all 30 years. Years outside the selected period stay, but their marks are hollow — so it doesn't look as if years are missing.
The 10, 20 and 30-year periods
The dates can be worked out from the last 10, 20 or all 30 years. Change the period and the dates are recalculated straight away.
There is no period shorter than 10 years. The site shows a safe date only if the selected period has at least 10 years with frost in that season — with fewer years we don't calculate it, because there are too few.
Under the dates it says by how many days the typical date for the last 10 years differs from the one for all 30. That shows which way recent years have gone, but a difference of a few days can also be chance.
The risk after a date
On the results page you can pick a date and see in how many of the years there was a spring frost after it.
Example for Manole: in 1 of 30 years there was a spring frost after April 20 — 3%.
This is how often it happened in the past, not a forecast for this year. Frost on the chosen date itself doesn't count — only after it. Spring frosts are those up to the end of June.
This works for spring only: in autumn more frosts follow the first one, so the question "will there be frost after this date" makes no sense there.
The safe date is the same calculation in reverse: instead of you picking a date, the site finds the date after which there was frost in only 1 of 10 years.
The frost-free season
For each year we work out the frost-free season length: how many days there are between the last spring frost and the first autumn frost, not counting the frost days themselves. The years in the selected period are used — 10, 20 or 30.
If a year has no recorded spring frost, the season is counted from January 1; if it has no autumn frost, through December 31; if it has neither, the season is the whole year — 365 days.
The typical length is the middle of those years. With an even number of years two values share the middle, and we take their average: for Manole, with 30 years selected, (236 + 238) / 2 = 237 days. (For the dates above we take the more cautious of the two middle dates; here, the average of the two day counts.) The shortest season was 194 days (1997), the longest 281 days (2023).
If you subtract the two typical dates yourself (March 29 and November 25), you get 240 days, not 237. It isn't a mistake: the typical length is worked out year by year and then the middle is taken, which isn't the same as the gap between the two typical dates.
The two elevations
The dates are calculated for the whole area of about 9 × 9 km, with one average elevation for it. For Manole, the area's average elevation is 99 m.
The place you pick is rarely right in the middle of the area and may be at a different elevation — Manole itself is at about 152 m.
The two elevations come from different sources. The page doesn't adjust the dates for them — it shows them so you can judge for yourself.
When the place is at least 50 m above the area's average elevation, the page warns that frost there probably ends later in spring and comes earlier in autumn. When it is at least 50 m below, it warns you not to count on earlier dates, because at night cold air flows downhill and settles in low ground. Manole is 53 m above the average elevation of its area, so the warning shows there.
The numbers are averages for the whole area, and in valleys they often show warmer nights than there really are — that is, less frost. There the real dates may be later in spring and earlier in autumn. But the difference in elevation alone can't tell you whether, or by how much.
Where the numbers come from
The numbers are calculated from ERA5-Land — a European weather archive from the Copernicus programme that holds temperatures day by day for recent decades. We use 30 full years: 1996–2025.
A frost day is any day when the lowest temperature 2 m above the ground is 0 °C or below. Right at ground level, frost can form even when it is slightly above zero at 2 m.
Each area is 0.1° × 0.1° (about 9 × 9 km). The dates are calculated in advance for all 2,080 such areas.
A year with fewer than 300 days of data isn't counted. The threshold of 10 years with recorded frost applies separately to spring and autumn and to each period — so an area can have spring dates but no autumn ones, or the other way round.
In the year-by-year table you'll see two different labels: "not enough data" — the data for that year falls short and it isn't counted; "no frost recorded" — there is data, but it shows no frost in that season.
Why it doesn't match what I see
The dates are calculated, not measured by a thermometer in your garden. They apply to an area of about 9 × 9 km (about 11 km north to south and 8 km east to west) — close to 90 km² with one number and one average elevation. An area like that has hills and hollows. In a low spot in your yard where cold air collects at night, frost can form even when there is none higher up. And on the ground, frost forms even when it is slightly above zero at 2 m.
Use the dates as a guide, not a hard rule — and write down when frost comes at your place. After a few years those notes will be the most accurate guide for your garden.