A frost date isn't a forecast — it's a bet

v1 — 2026-07-14.

You look up "last frost" for your area, a date comes back, and your brain files it as a fact: May 14th. From the 14th, you're clear.

No. That date is not a prediction of anything. It is a statistic from decades of past years, and planting on the day it names is not following an instruction — it is accepting a probability.

What a "frost date" actually is

Frost dates come from climate normals: decades of weather-station records, summarised into probabilities. The source that publishes them says so plainly: they are climatological probabilities based on historical data, not weather forecasts — the dates in the tables are best guesses of what might occur in a given year, based on trends observed in previous years.

Nobody looked at your year. Nobody looked at your garden. They looked at the past.

That's why the tables don't give one date: they give several, one per probability level. And picking a level is not a technical detail — it is picking how much risk you accept. An early date is an aggressive bet; a late one is conservative. The date doesn't change the weather; it changes your exposure.

The percentile IS the risk

This is the idea worth internalising: a percentile always leaves a remainder.

If your start date is the one by which 90 % of years have already had their last frost, that means exactly what it looks like: in 1 year out of 10, it frosts afterwards. That is not the table failing. That's the table working: it is telling you the price.

Choosing "the average date" (the 50 %) sounds reasonable until you say it out loud: it is flipping a coin with your season.

Your window is squeezed from both ends

Outdoors you don't have a deadline, you have a sandwich:

• In front, the spring frost: planting too early exposes the plant to a late frost that takes it out.
• Behind, the autumn frost: the plant has to finish before the first frost of the end. That is your real deadline, and it is not negotiable.
• In the middle, the sky: outdoors you do not choose when flowering starts. The photoperiod decides, when days get short enough.

That third line is the one most people ignore, and it's covered in it isn't the light — it's the dark: flowering arrives when the night gets long enough, whether it suits you or not. You choose the planting date. Latitude supplies the rest.

Sometimes the honest answer is "it doesn't fit"

Put the three together — spring frost, the cycle the plant needs, autumn frost — and the answer may be that there is no gap. That in your climate, that plant profile does not finish in time.

That isn't an error: it's the result. This site's outdoor start-window calendar says so when it happens, instead of handing you a tidy date that would lead to a frozen harvest. It would rather refuse than lie to you.

What these dates don't know about you

And here is the warning no table comes printed with:

• Normals come from a weather station, not from your plot. Elevation, which way the slope faces, and hollows where cold air pools can move your real date by days or weeks.
• They are history, not the future. A strange year is a strange year, and the tables don't see it coming.
• This site's calendar coverage is limited, and we don't hide it: today it only includes three US regions, because the European agencies we reviewed don't publish the early-autumn frost percentile the model needs. We would rather not have your region than invent its data.

Use them as what they are: the starting point of a decision that you make, knowing the risk, and adjusting with what you know about your own ground.

The summary

A frost date is a probability from decades past, not a forecast for your year. The percentile you pick is the risk you accept: 90 % leaves out 1 year in 10. Your window is squeezed by two frosts at the ends and by photoperiod in the middle — the flowering date is not yours to choose. And if it doesn't fit, the honest answer is that it doesn't fit.

Sources

• Nelson, N., El-Khoury, L. & Boyette, M. (2020, rev. 2024), "Interpreting Freeze / Frost Probabilities from the National Centers for Environmental Information", NC State Extension (retrieved 2026-07-14)