How we calculate frost dates

Every number on this site comes from public climate data and a small set of documented rules. This page explains each step, so you can judge how far to trust the dates for your garden.

1. Frost dates come from NOAA climate normals

NOAA's National Centers for Environmental Information publish U.S. Climate Normals: 30-year statistics for thousands of weather stations, currently for 1991–2020. For each station NOAA estimates the dates of the last spring freeze and first fall freeze at several probability levels. We use the 32°F (0°C) threshold, the usual definition of a frost for gardeners.

We use 7,245 stations with usable records. Stations whose probability dates are out of order are left out rather than published.

2. What the 10%, 50% and 90% dates mean

• Median (50%) last frost: in half of years the last frost comes earlier, and in half it comes later. This is the date most sources call the "average last frost date".
• 10% date: only 1 year in 10 has a frost after this date. It's the safer date for planting out tomatoes, peppers and other frost-tender crops.
• 90% date: in 9 years out of 10 frost continues past this date. Planting before it is a gamble.

Fall works the same way in reverse: the 10% first-frost date is the early date, when frost arrives in only 1 year in 10. The growing season is the number of days between the median last spring frost and the median first fall frost.

3. Four types of climate

A single "last frost date" doesn't make sense everywhere, so every station is put into one of four groups:

• Regular frost season (6,580 stations): NOAA defines spring and fall dates. These places get the full calendar.
• Occasional light frost (367 stations): dates exist, but the frost-free season is 300 days or longer. Frost comes on a few winter nights, as in Jacksonville or the San Francisco Bay Area.
• Frost is rare (118 stations): NOAA can't define a date because most years never reach 32°F, and the station averages fewer than 30 freezing nights a year.
• Frost possible in any month (180 stations): no frost-free season can be defined, and the station has 30 or more freezing nights a year. These are high mountains and the Arctic. They get advice on short-season crops and season extension instead of a calendar.

Dates that cross New Year (a spring frost in December, or a fall frost in January) are kept in order instead of wrapping around, so the probability table stays correct in mild climates.

4. Matching a ZIP code to a weather station

ZIP code locations come from GeoNames postal code data. Each ZIP code uses its nearest station by straight-line distance, up to 150 km. Half of ZIP codes are within 10 km of their station, 96% are within 25 km and 99.6% are within 40 km. The farthest is 148 km. Each page names the station and gives its distance and elevation, and pages more than 40 km from their station say so.

City pages group the ZIP codes that share a city name. If two towns in the same state share a name but are more than 25 km apart, they get separate pages, with the county added to the name. A city page uses the station nearest the center of its ZIP codes. A ZIP code whose own nearest station is different gets its own page.

5. Hardiness zones

Zones come from the PRISM Climate Group data behind the 2023 USDA Plant Hardiness Zone Map, via the open-source frostline project. A ZIP code missing from that data borrows the zone of the nearest ZIP code within 15 km. City pages show the most common zone among their ZIP codes. The "typical" dates on zone pages are medians over every ZIP code in the zone with a regular frost season. This is not the official USDA map, so check the USDA site for the authoritative zone.

Zones measure the coldest winter night, not season length. Two places in the same zone can have last frost dates weeks apart, which is why every location page compares its dates with its zone.

6. Planting windows

Each crop has windows measured in weeks before or after the median frost dates, following common guidance from university extension services. For example, tomatoes are started indoors 6–8 weeks before the last frost and transplanted 1–3 weeks after it. We apply those offsets to your dates. Two extra rules:

• A fall planting window is shown only if it starts at least two weeks after the spring planting ends. Otherwise the season is too short for a real second crop. Garlic is the exception, since it's always planted in fall.
• Warm-season crops are flagged ⚠ when their days to harvest plus two weeks is longer than the frost-free season.

Places with occasional light frost and a median last frost before late January don't get a frost-based calendar, because heat and rainfall matter more there than frost.

Limitations

• These are 30-year statistics, not forecasts. Always check the forecast before planting tender crops.
• Frost varies over short distances. Valleys and low spots collect cold air, while cities, slopes and nearby water stay warmer. Elevation differences between you and the station matter most in the mountains.
• NOAA's freeze dates measure air temperature about 1.5 m (5 ft) above the ground. Frost can form on plants at ground level on clear, calm nights even when the air stays above 32°F.

Spotted a date that looks wrong for your area? Tell us. Local reports are how we find stations that don't represent their surroundings. Data built 2026-11-11 from the 1991-2020 normals.