In about half of springs the last frost comes after the median last frost date, and across the lower 48 the date by which 9 springs in 10 are clear of frost sits a median of 16 days later.1 In zone 6a, the zone with the most weather stations behind it, the median last frost is April 30 and the 9 in 10 date is May 16.
A last frost date as it is usually printed is the median: even odds that the last 32F freeze of spring lands on or after it. NOAA computes the same station's date at several probability levels, from the chance that the last spring freeze falls on or after a given day, and the 10 percent date is the one with only 1 spring in 10 freezing later.1 Reading it beside the median shows how much cushion the median leaves. The table gives both for every zone with enough stations to support a median.
| Zone | Median last spring frost | Clear in 9 springs of 10 by | Days between | Stations | Example city |
|---|---|---|---|---|---|
| 3b | May 17 | May 31 | 14 | 89 | Hibbing, Minnesota |
| 4a | May 14 | May 28 | 14 | 301 | Fargo, North Dakota |
| 4b | May 11 | May 25 | 14 | 487 | Rochester, Minnesota |
| 5a | May 9 | May 24 | 14 | 721 | Minneapolis, Minnesota |
| 5b | May 4 | May 18 | 15 | 791 | Lincoln, Nebraska |
| 6a | April 30 | May 16 | 16 | 1044 | Chicago, Illinois |
| 6b | April 24 | May 10 | 16 | 1033 | Columbus, Ohio |
| 7a | April 18 | May 5 | 17 | 954 | Boston, Massachusetts |
| 7b | April 7 | April 24 | 16 | 695 | New York, New York |
| 8a | March 30 | April 16 | 17 | 707 | Charlotte, North Carolina |
| 8b | March 19 | April 8 | 19 | 713 | Dallas, Texas |
| 9a | March 4 | March 27 | 23 | 482 | San Antonio, Texas |
| 9b | February 5 | March 8 | 30 | 384 | Houston, Texas |
| 10a | January 17 | February 20 | 30 | 178 | Phoenix, Arizona |
| 10b | January 15 | February 8 | 24 | 45 | Los Angeles, California |
Where the margin is widest
The gap runs from 14 days in zone 3b to 30 days in zone 9b. In the warmest zones a freeze is not a yearly event, and NOAA computes these dates only over the springs that reach the threshold at all, so there they describe how late a frost can come in a year that has one, not the odds that one comes.1
How these numbers were built
Each weather station in NOAA's 1991 to 2020 climate normals2 carries the date of its last 32F spring freeze at probability levels from 10 to 90 percent, and this table uses the 50 and 10 percent dates.1 NOAA derives them from 10,000 simulated years of daily low temperatures rather than from the 30 observed springs alone, which keeps a single odd year from swinging the result.1 Every ZIP code in the lower 48 is assigned to its nearest station and to its USDA zone, and the zone comes from the Plant Hardiness Zone Map, which sorts places by average annual extreme minimum winter temperature rather than by frost dates.3 A station counts toward each zone its ZIPs fall in. A zone row is the median of its stations, column by column, and the days column is the median of each station's own gap, so it will not always equal the distance between the two date columns. Zones with fewer than 20 stations are left out rather than shown on a thin sample: 3a (5), 11a (10), 11b (2).
What the table cannot tell you
These are air temperatures at a NOAA station, so they describe the station, not the bed. A zone row is also a median across many stations that disagree with each other. For one place, the per-ZIP frost lookup gives the figures for the station that serves it, and the zone table and state table give the medians alongside their typical range.