US Grow Zones

Frost dates explained: what 10%, 30% and 50% really mean

Last updated · Sources: cited below

Every planting calendar starts from two dates: the last frost in spring and the first frost in fall. Those dates are often printed as if they were fixed points. They are not. A frost date is a probability, and choosing which probability to use is the single biggest decision in planning a vegetable garden. This guide explains how NOAA computes frost dates, what the 10%, 30% and 50% figures on our city pages mean, and which one to use for which crop.

Where frost dates come from

NOAA's National Centers for Environmental Information publishes the U.S. Climate Normals every ten years. The current set covers 1991–2020 and was first released in May 2021; it includes temperature normals for more than 7,300 stations.1 Among the hundreds of statistics in that release are freeze probabilities: for thresholds of 36, 32, 28, 24, 20 and 16°F, NOAA reports the dates by which the last spring freeze and the first fall freeze have occurred at probabilities from 10% to 90%.

This site uses the 32°F series as “frost” and the 28°F series as “hard freeze.” We take the curves straight from NOAA's annual/seasonal normals files for the station matched to each city (closest station within 150 m of the city's elevation, with at least 10 years of record — see the methodology).

Reading the probabilities

For spring, NOAA's figure is the chance that a 32°F night will still occur after a given date. Take Chicago, where we use the Northerly Island station 4.1 miles from the city center:

Chance of a later frostDateWhat it means
90%Mar 29Frost after this date is almost certain.
50% (median)Apr 10Half of springs have their last frost after this.
30%Apr 17About 3 springs in 10 still have a frost after this.
10%Apr 25Only about 1 spring in 10 has a frost after this.

Fall works the other way round: the 10% date is the early one. In Chicago there is a 10% chance that the first fall frost has already happened by Oct 24, a 50% chance by Nov 5, and a 90% chance by Nov 20.

Many published “average last frost dates” are the 50% date — the UC Master Gardeners of Sonoma County note that garden websites commonly cite NOAA's 50% probability.6 That is a coin flip — plant tomatoes on it and you should expect to lose them in about half of all years, unless you watch the forecast. Our city pages headline the 30% date as a planning date and show the 10% and 50% dates alongside, so you can choose your own risk.

How wide is the window?

The distance between the 90% and 10% dates tells you how erratic spring is. Across the 460 cities on this site with computable frost dates, the median gap is 36 days; the median gap between the 50% and 10% dates is 18 days. In other words, going from a coin-flip date to a 1-in-10 date typically means waiting two and a half more weeks.

The range is large. Anchorage, AK has one of the narrowest spring windows (20 days from 90% to 10%), while Daly City, CA, where frost is occasional rather than routine, has one of the widest (101 days). Wide windows are typical where frost is marginal: a few cold nights decide the statistics, so the dates scatter.

The frost probability timeline chart on each city page draws the whole curve, so you can see at a glance how quickly the risk falls. A steep curve means a predictable spring; a gentle slope means you will want frost cloth on hand for longer.

Frost versus hard freeze

Extension services usually separate a frost or light freeze (around 32°F) from a hard or killing freeze, typically defined as 28°F or lower, which can kill most non-dormant plants.2 The distinction matters for planting: cabbage, kale, spinach and peas shrug off a 32°F night, so the 28°F date is a better guide for them.

In our data, the median 28°F last-freeze date comes a median 13 days earlier than the median 32°F date. In Denver, for example, the 50% dates are Apr 22 for 28°F and May 4 for 32°F; in Atlanta they are Mar 9 and Mar 24.

Frost can also form when the official air temperature is above freezing. On clear, calm nights, leaves radiate heat to the sky and can drop to 32°F while the air a few feet up stays warmer.3 Weather stations measure air temperature in a shelter above the ground, so a forecast low of 34–36°F can still mean frost on low-growing plants in an open bed.

Which date should you use?

Crop typeExamplesSuggested spring date
Frost-tender, heat-lovingtomato, pepper, basil, eggplant10% date or later; they also need warm soil
Frost-tender, quickbeans, corn, cucumber, squash30% date, with a cover ready
Semi-hardylettuce, beets, potatoAround the 50% date, or a little before
Hardypeas, spinach, kale, onionWeeks before the 50% date; use the 28°F date as a guide

The calendars on this site follow this logic: weeks are counted from the 30% date, and tender crops are scheduled one to three weeks after it, which usually lands them near or past the 10% date.

Worked example: tomatoes in Denver

In Denver the spring 32°F dates are May 4 (50%), May 8 (30%) and May 16 (10%). Planting tomatoes on the 50% date means a frost after planting in about half of years. Waiting until the 10% date cuts that to about one year in ten, for a cost of 12 days. Because tomatoes also stall in cold soil, the later date usually costs little in harvest. Our calendar for Denver places transplanting one to three weeks after the 30% date — close to the 10% date — and the seed-starting window six to eight weeks before the 30% date.

For peas, which tolerate frost, the same logic runs the other way: they can be sown well before the 50% date, and the 28°F date (Apr 22) is a better guide to when a freeze could actually hurt them.

Why frost dates differ between websites

If two sources give different “last frost” dates for your town, three things usually explain it:

  1. Probability level. One site may show the 50% date, another the 30% or 10% date. The difference is typically one to three weeks.
  2. Station choice. The nearest station may be at an airport, a hilltop or a river valley. We pick stations by distance and elevation and show which one we used.
  3. Period. Older sources may still use the 1981–2010 normals. NOAA notes that most of the country was warmer in 1991–2020, while parts of the north-central U.S. cooled somewhat in spring.1

None of these is wrong — they answer slightly different questions. Knowing which one you are looking at lets you compare like with like.

Protecting plants when the forecast surprises you

Probability means some years will break the pattern. When frost is forecast after you have planted:

Where frost is rare

At some stations a 32°F night happens too seldom for NOAA to compute meaningful dates. On this site 52 of 512 cities fall in that category, mostly in coastal California, south Florida, Hawaii and Puerto Rico. For those cities we say “frost is rare” and show seasonal planting windows instead of inventing a date. We also report how often the station reached 32°F at all during 1991–2020.

Next: learn why your own garden may be colder or warmer than the nearest station in Microclimates, or use your fall frost date to plan succession and fall crops.

Sources

  1. NOAA National Centers for Environmental Information — U.S. Climate Normals (1991–2020)
  2. South Dakota State University Extension — “Frost and Freeze Impacts on Plants”
  3. University of California Statewide IPM Program — “Freezing and Frost”
  4. Iowa State University Extension and Outreach — “How to Protect Plants from Frost and Freeze”
  5. Michigan State University Extension — “Frost and freezing temperatures and ways to protect plants”
  6. UC Master Gardener Program of Sonoma County — “Sowing Planting Calculation”

Figures marked as “our data” or “on this site” were computed by US Grow Zones from the PRISM/OSU ZIP code listings (2012 and 2023) and NOAA 1991–2020 Normals; see the methodology.

More guides: How to read your hardiness zone · Why the 2023 hardiness zone map looks warmer · Microclimates · Succession and fall planting