Will it finish outdoors?

A season ends at the killing freeze rather than at the first frost. A mature plant takes a light touch of frost and growers carry on through it; what finishes a grow is a night at about −2°C, which these stations record nine to seventeen days after the first. Five maritime regions have no such night at 50%-probability at all and keep the first-frost date instead, which their pages say. Getting this wrong is what made an earlier version of this page tell a grower on Lake Superior that nothing photoperiod finishes there, while he harvests through late September and October.

Each region is named for a weather station, and the frost dates are that station’s: the median date of the last spring and first autumn freeze across 1991 to 2020, which is the 50%-probability date. US regions come from NOAA’s published Climate Normals. Australian ones are Bureau of Meteorology observations and the New Zealand ones come from that country’s national network; those and the European ones are worked out from the daily station records by the same definition, checked against NOAA’s own published figures first and agreeing with them inside a week. Your own plot will differ anyway: a sheltered, south-facing spot buys a week and a low frost pocket costs one.

The four New Zealand regions are normals over 1991 to 2004 rather than 1991 to 2020, because those stations’ daily records are only complete to 2004. Their pages say so. NIWA publishes normals for rainfall, temperature and sunshine but none for frost dates, so there is nothing more authoritative to use instead.

Somewhere missing? A region is only here if its station recorded a frost in most seasons. That rules out the Mediterranean, coastal southern California, Phoenix, Melbourne city and the Western Australian coast, where frost is not what ends a season, and it rules out anywhere nearer the equator than about 31 degrees, where the days never lengthen enough to hold a photoperiod in veg and the trigger this page turns on never fires. Japan, Korea and Chile are missing for a duller reason: their daily records reach this archive with roughly half the days absent, spread evenly through the year, and a frost date computed from half the days would put the last spring frost too early and the first autumn frost too late. Neither the Japan Meteorological Agency nor the South African Weather Service publishes a free frost-date normal to use instead. South Africa is represented by the one station whose record is complete.

Plan a single grow week by week → · Search every variety