Frost & Freeze Date Explorer
How often Phoenix has hit freezing, by month and by station, from the climate normals and the station record, with what each threshold means for tender plants.
Threshold
The record, in three lines
- Winters the record covers
- 92
- Winters with a freeze
- 53 58% of them
- Last freeze at the airport
- 2019-01-02
A winter runs from September to the end of June, because a frost season does. Counted that way the record holds 92 winters.
The two dates to plan around
Earliest autumn freeze: 9 November. Latest spring freeze: 15 March.
Those are the outer ends of 92 winters at one station, not a rule. The median winter that froze at all froze on six nights.
January · below 32 °F
212 nights
212 of the 2,727 January nights in the record fell below 32 °F: about 2.3 in a typical January, and 7.8% of the nights in the file. The 1991–2020 normals count 0.4 nights below freezing in a normal January, and the record restricted to those same thirty years gives 9. Most of the 212 nights are in the winters before 1980.
Nights below the threshold, by month
Every night in the Sky Harbor record that fell below the threshold, added up by month across all 93 years. Pick a month to read it.
One scale for both thresholds, set by the tallest month at all: January below 36 °F, at 596 nights. The counts are printed on each column so the two thresholds can be read against each other rather than only looked at. Below 32 °F the record reaches back to 1933; a month with 0 means the airport has never had a night that cold in that month.
What each threshold does to a plant
The table is in two halves. The citrus rows carry the University of Arizona Cooperative Extension figures for cold damage. Everything below the citrus is this site describing how the plant looks after a frost: those rows have no publisher behind them and are marked that way.
| Plant | Below 36 °F | At 32 °F and below |
|---|---|---|
| Lemons and limes, fruit on the tree Sourced | No damage. The tree is not troubled by a night in the thirties. | Fruit on the upper and outer canopy takes damage below 28 °F. Lemons and limes are the most exposed of the citrus fruits. |
| Oranges, grapefruit, mandarins, fruit on the tree Sourced | No damage. | Hardier than lemons and limes, and the Extension guide rates most locally suitable grapefruit varieties as frost-hardy. Fruit is still the part most at risk. |
| Young citrus, first three winters Sourced | Nothing visible, and this is the stage worth protecting anyway: a young tree is the part of the planting most easily lost, and site choice and cover are the two things the Extension publication recommends. | The tree itself, not just the fruit, is at risk. Cover and a wrap on the trunk are the measures in the publication. |
| Bougainvillea Site description | The top growth is the first thing to mark, and a still night in the thirties can take the leaves. | The canopy can be killed back to the older wood, and an established plant usually regrows from the base through spring. |
| Lantana Site description | Leaves and flowers mark; the plant stops growing and looks finished for the winter. | Top growth can be killed to the ground, and the roots survive to regrow once the soil warms in spring. |
| Oleander and other old-planting shrubs Site description | Unaffected. | Usually unaffected at 32 °F. Damage here is a hard-freeze event rather than a frost, which is why a mature oleander hedge can be decades old and never have been covered. |
| Annuals and vegetables in the open Site description | This is the line the planting calendars work to: tender summer vegetables and flowering annuals are finished for the season or need a cover. | Tender annuals are killed outright, and the soil surface in the open can freeze. |
Rows marked Sourced carry a figure published by University of Arizona Cooperative Extension, at the link. Rows marked Site description are this site’s own account of how the plant performs in a Valley frost, written without a publisher behind the words: the same arrangement the monsoon checklist uses for advice no agency gives.
Freeze nights by decade, at one station
The same count, split by the decade each winter opened in. The airport’s freezes are almost all in the first half of the record.
- 1930s 74 5 of 6 winters
- 1940s 167 9 of 10 winters
- 1950s 80 10 of 10 winters
- 1960s 123 10 of 10 winters
- 1970s 47 9 of 10 winters
- 1980s 8 4 of 10 winters
- 1990s 1 1 of 10 winters
- 2000s 2 1 of 10 winters
- 2010s 9 4 of 10 winters
- 2020s 0 0 of 6 winters
Counted from the same station file. Two things are going on in this table and only one of them is measurable here: freezes have become rarer across the period, and the site itself has changed from an airport on the edge of a small city to an airport surrounded by a large one. A colder site would still be a colder site.
Two thresholds, because a frost and a freeze are different events
The freezing point is 32 °F, which makes it a strange number to organise a garden around. The damage happens before it. Frost forms on surfaces that are colder than the air a few feet above them, so on a still, clear night the roof of a car can carry ice while the thermometer reads 36 °F, and the top of a lantana can blacken without the air ever reaching freezing. The plants that suffer most here (bougainvillea, lantana, the summer annuals and vegetables) are the ones from somewhere warmer, and they respond to hours in the thirties, not to a single legal threshold.
That is why this page counts 36 °F and 32 °F and shows both on one scale. At Phoenix Sky Harbor the difference is not marginal: 596 January nights below 36 °F against 212 below 32. Roughly two thirds of the frost-risk nights in the record never reach freezing at all, which is a fact a gardener can use and a fact that a count of freeze days alone would hide.
Why the airport almost never freezes, and why that is not reassuring
Sky Harbor has not recorded a night below freezing since 2 January 2019, and it recorded eleven across the thirty years from 1991 to 2020. Read that against the 1940s, which produced 167 such nights across ten winters, and the shape of the record is unmistakable: freezes have become rare at this site, and almost all of them are in the first half of the record.
Two explanations sit behind that and the data on this page cannot separate them. One is the climate, which has warmed. The other is the site: the station sits where it always sat, but the land around it has become an airport and then a city, and asphalt, buildings, engines and air-conditioning exhaust all hold heat through the night. An urban minimum temperature is higher than a rural one on the same night, and that gap has widened here for ninety years. Whatever the split, the practical consequence is the same and is why the limits list leads with it: the airport’s warming is a fact about the airport, and a garden twenty miles east of it is a different site with a shorter, colder record or none at all.
The normals and the record disagree, and both are right
NOAA’s 1991–2020 normals give 0.4 nights below freezing for a normal January. The station record, counted here, gives 212 nights below freezing across 93 Januaries, about 2.3 a January. Those look like a contradiction and are not. A climate normal averages the most recent thirty-year window; the record is everything the station has ever measured, and most of what it measured in the way of freezes happened before 1980.
The check is easy to run and this page runs it: restrict the record to the normals’ own window and it gives 9 nights below freezing in 30 Januaries, or 0.3 a winter, which is the same quantity the normals are describing. Presenting either number alone would have been simpler and would have misled in one direction or the other. The normals answer "what is a typical January now"; the record answers "what has this station actually done", and a planting decision needs both, the second one especially, because a garden is built once and has to survive the tail of the distribution rather than the middle of it.
What a Valley freeze actually is
Almost every freeze in this record is a radiation frost: a clear, still night after a dry front has moved through, when the ground radiates its heat away to a black sky and the air nearest the surface cools fastest. That is why the wind matters more than the temperature forecast. On a windy night the air mixes and a garden stays warmer than the number suggests; on a still night the coldest air drains downhill and pools, and the lowest ground in a neighbourhood reads several degrees colder than the ground above it.
It follows that frost is a property of a place rather than of a night. The bottom of a slope, a walled courtyard, the low corner behind a block wall and the open lawn are four different climates on the same morning, and the airport (flat, paved, built-up, with the whole Valley’s warm air around it) is the mildest of them. The Extension publication this page draws on makes the same point from the other direction: the first and best protection for a citrus tree is where it was planted, long before anyone thinks about a cover.
How to use the numbers without over-reading them
There are three questions a reader usually has, and the tool answers the first two well and the third badly on purpose.
How often does it freeze here? That is the per-winter reading beside each month, and the honest version of it is the 1991–2020 window rather than the whole record, because that is the airport that exists now.
When is the risk over? The record puts the latest spring freeze at 15 March and the earliest autumn freeze at 9 November, which brackets the season at its extremes rather than at its middle. Those dates are the reason the planting calendars here treat February planting as a bet and mid-March planting as the safe side of the line.
Will it freeze in my garden next week? The tool will not answer that, and no history can. That is a forecast, the National Weather Service issues it, and a frost advisory or freeze warning from the Phoenix office supersedes everything on this page the moment it is issued.
- NOAA National Centers for Environmental Information: Global Historical Climatology Network, Daily (GHCN-Daily), station USW00023183, Phoenix Sky Harbor
Every count on this page. The nights below 36 °F and below 32 °F by month, the 511 nights below freezing the record holds, the freeze-season dates, the winter and decade counts, and the 1991–2020 sub-window counts were all counted from this file, using the same strict below-threshold convention NOAA uses. The file was downloaded and counted on 11 October 2026 and covers 1 June 1933 to 7 October 2026.
- NOAA National Centers for Environmental Information: U.S. Climate Normals 1991–2020, monthly summaries, Phoenix Sky Harbor (USW00023183)
The normal number of nights a month below 32 °F and below 40 °F, which is the figure the page compares the station record against. The comparison is the reason the page has a "why the two numbers differ" section rather than one number.
- NOAA National Centers for Environmental Information: U.S. Climate Normals product page
The definition of the 1991–2020 normals window and what a climate normal is, so the page can say plainly what it is comparing the station record against.
- University of Arizona Cooperative Extension: "Protecting a Citrus Tree from Cold" (az1222, Glenn Wright)
The guidance behind the young-citrus row of the plant table: that citrus lacks strong cold hardiness, that site selection is the long-term protection and cover the short-term one, and that the tree itself, not only the fruit, is what is at risk.
- University of Arizona Cooperative Extension: "December Monthly Gardening Guide for Pima County"
The one published temperature figure in the plant table: that exposed lemons and limes take damage on the upper and outer parts of the tree when temperatures drop to 28 °F, and that other citrus fruits handle cold better.
Every figure this tool uses was read from the source named above on 11 October 2026. Anything still waiting on a check is listed in the site's data notes rather than shown here as if it were settled.
What this explorer is, and is not
It is a count of what one thermometer at Phoenix Sky Harbor has done since 1933. It is not a forecast, it is not a probability for your garden, and it is not the same climate that a yard in the outer East Valley lives in. Every figure on the page is a night in the record rather than a winter, and the per-winter reading is printed beside it for that reason.
- This is the Sky Harbor record and nothing else. The airport is the warmest ground in the Valley on a still winter night (it sits in a built-up, paved, air-conditioned landscape that holds heat) and the older parts of the record were made when far less of that landscape existed. A garden in Queen Creek, Apache Junction, Anthem or the far East Valley is colder on the same night, and this page cannot tell you by how much. Mesa, Chandler and Queen Creek all publish their own climate summaries if you want a site closer to you.
- The counts are nights in the record, not winters. "596 nights below 36 °F in January" means 596 nights spread across 93 Januaries, and the page prints the per-winter reading beside it so the two are not confused. Dividing one by the other is not the same as a probability for next January.
- The record is not a stationary thing, and this page does not pretend it is. Freezes at Sky Harbor have fallen from 167 nights across the 1940s to none since 2 January 2019, and the likeliest cause is the city warming the site. Which half of that is climate and which half is the instrument is not a question a station record can answer, and this page does not try.
- Nothing here is a forecast. A count of past nights tells you what the odds were over the last ninety years, not what will happen in the next ten days. The National Weather Service issues the actual forecasts and warnings, and a frost advisory or a freeze warning from the Phoenix office is current information where this page is a history.
- Two thresholds only. 36 °F is where the tender plants mark and 32 °F is where the record calls it a freeze, and the gap between the two counts is the thing worth understanding. Nights below 28 °F, which is where citrus fruit is damaged, are in the record as well (108 of them across December and January) and the page names that figure without building a third scale around it.
- The plant table is mostly this site describing what a plant looks like after a frost. Only the two citrus rows carry a published temperature figure, from University of Arizona Cooperative Extension, and they are the only rows with a publisher behind the sentence. The rest are labelled in the table as descriptions rather than left to look like sourced facts. See the note under the table.
- A frost is a place, not a number. Cold air drains downhill and pools in the lowest ground, so the coldest spot in a garden is usually the bottom of a slope or a walled corner rather than the middle of the lawn, and the difference between a thermometer in the open and one against a west-facing wall is real. This tool has no way to see any of that.
- The two thresholds are this site’s choice. NOAA counts "below 32 °F" and publishes it; 36 °F is the number a gardener acts on, and it is counted here from the same record with the same strict convention: strictly below, so a night that stops exactly at 36 does not count. Change the threshold and every number moves.
- Covering a plant works, and it is not measured here. The Extension publications on the citrus rows are where to go for the method: what to cover, how to hold the cover off the leaves, and why a bare trunk wrapped in early winter matters more than a sheet over the canopy in a hard freeze.