Methodology

How we calculate frost dates and planting windows

The calendars combine NOAA station data with crop timing rules drawn from university Extension publications. This page explains each step and its limits.

1. Climate data

All frost and temperature figures come from the NOAA NCEI U.S. Climate Normals for 1991–2020, read directly from the station files for one weather station per city:

We record NOAA's completeness flag for each station (“Standard”: 24+ years; “Representative”: 10+ years with gaps estimated from nearby stations) and show it on every city page. The project records the source files and their SHA-256 checksums. Automated tests compare selected frost dates with the NOAA station rows.

2. Which cities we cover

Each city uses one station, usually its main airport, named on the city page with its elevation. We only publish a city when the station has complete 32°F probability dates and NOAA's PRBOCC shows a spring 32°F freeze in at least 90% of years. Below that threshold a “last frost date” describes a rare event, and calendars built on it mislead. For that reason Phoenix, Tampa, Orlando, New Orleans, Fresno, San Francisco, Los Angeles are not covered yet.

A station is not your garden. Cold air drains into low spots, cities stay warmer at night, and elevation, slope and nearby water all shift frost dates. These differences can shift the dates at which your plot freezes. Use the probability tables to decide how much margin you want.

3. ZIP code search

ZIP codes are matched to the nearest covered station using the internal point of each ZIP Code Tabulation Area in the 2024 Census Gazetteer. We return station-based dates only when that point is within 30 miles of the station; this distance is our own product rule, not a scientific standard, and the distance is always shown. Hardiness zones come from the 2023 USDA Plant Hardiness Zone ZIP code listing (USDA-ARS and Oregon State University PRISM Group), shown unaltered. We do not draw zone maps.

4. Frost risk for a date

NOAA publishes nine probability dates (10%, 20% … 90%) per threshold. For a date between two of them we interpolate linearly. Example for Denver: the 50% date for the last 32°F freeze is May 5 and the 40% date is May 8, so a date in between shows a chance between 40% and 50%. Outside the published dates, we report the corresponding upper or lower bound. If multiple percentiles share the same date, we show their range.

5. Planting windows

For each crop we collected timing rules from university Cooperative Extension publications (listed on each crop page): weeks before or after the last frost to transplant or sow, indoor growing time, frost tolerance, soil temperatures and fall planting rules. When sources disagree we use a range covering them and say so on the crop page.

  1. Reference dates. “Typical” uses the 50% dates for 32°F. “Cautious” (planner option) uses the 10% dates: the last spring freeze comes later than that date in about 1 year in 10, and the first fall freeze comes earlier in about 1 year in 10.
  2. Spring windows = reference last-frost date + the crop's week offsets.
  3. Night-temperature rule for warm-season crops. UC Agriculture and Natural Resources and Ask Extension advise waiting for nights above 50°F before planting warm-season crops out. We therefore do not schedule outdoor planting of tender crops before the date when the station's normal daily low reaches 50°F. In cool-summer climates this matters a lot: in Seattle the median last frost is March 2, but normal night lows only reach 50°F around May 23.
  4. Indoor seed starting is counted back from the transplant date using the crop's indoor growing time (the method described by Rutgers and Iowa State Extension).
  5. Fall windows = reference first-frost date minus either the crop's published “weeks before first frost”, or the Virginia Cooperative Extension count-back rule: days to maturity + a 14-day Fall Factor (+14 days for frost-tender crops). Garlic follows University of Minnesota Extension: 1–2 weeks after the first killing frost (we use the median first 32°F freeze).

These rules come mostly from northern and central U.S. Extension services. In hot or mild-winter climates soil temperature and summer heat can dominate; city pages flag mild-winter stations and summer heat so you can compare with local Extension calendars.

6. What this site does not do

7. How we check the calculations

Automated tests compare the site's dates with values read directly from NOAA files, check that every probability series is ordered correctly, that indoor sowing always ends before transplanting, that fall windows never overlap spring windows and that warm-season crops are never scheduled before the night-temperature date.