Why Phoenix nights stay hot

A normal January day in Phoenix swings 19.0 °F between its coolest and its hottest hour. A normal July day swings 19.2 °F. Drawn as two lines on the same axes, they are very nearly the same shape: a shallow dip overnight, a steady climb through the morning, a flat top across the afternoon, a fall that picks up speed after dark.

What changes between the two months is not the shape. It is the height, and it changes by about 38 °F, evenly, at both ends. In January the arc runs from 47.6 °F at seven in the morning to 66.6 °F at four in the afternoon. In July it runs from 85.6 °F at six in the morning to 104.8 °F at four in the afternoon.

Read those two ranges against each other and the striking number is not the 104.8. It is that the coolest hour of a normal July day is 19 °F warmer than the hottest hour of a normal January day. By the time July arrives there is no hour left that a person could describe as cool.

Two normal days, and the shape they share

The number that does not fall

The chart is the argument, and the table is the count. Take the normal July day and ask how many of its twenty-four hours sit at or above a plain 90 °F, which is warm by any standard and is the point at which the National Weather Service’s heat products start to matter here.

No month outside that list reaches 90 °F at any hour.

July’s seventeen hours leave exactly seven below the line, and they run from two in the morning to eight. That is the honest version of the advice to do things early: the window is real, but it is narrow and it is dark, and it closes before most people start. September is the month that catches people out: it holds eleven hours at or above 90 °F while the calendar has already turned, which is why the month-by-month article keeps a September section of its own.

The 100 °F row is the other one worth reading twice. July holds eight hours at or above 100 °F and August seven, and those hours are not a spike around noon. They stretch across the afternoon and into the early evening, so a normal Phoenix summer evening is still a hundred-degree evening.

What the tenth percentile says

Averages hide the nights that are worse than average, so it is worth asking what the edges of the distribution look like rather than only the middle.

At 6 a.m. in July (the coolest hour of the coolest part of the day) one morning in ten is cooler than 79.7 °F, and one in ten is warmer than 91.2 °F. The good tail of July nights does exist, and it bottoms out near 80 °F, not near 70 °F.

At 4 p.m. the same month, one afternoon in ten is cooler than 98.1 °F and one in ten is warmer than 110.8 °F. Ten of July’s twenty-four hours have their tenth percentile at or above 90 °F, which means that even on the cool end of thirty years of July days, those ten hours hold at 90 °F or above. Thirteen of them have their ninetieth percentile at or above 100 °F, so one July day in ten passes 100 °F in thirteen separate hours.

The averages are the useful planning number. The percentiles are the reminder that a bad July night is worse than the average rather than a variation on it.

Why the heat has nowhere to go

Open desert loses its heat quickly after sunset, because there is very little there to have stored any. A city does not, and the Arizona State Climate Office puts the reason plainly: paving, roofs and walls take in sunlight through the day and release it slowly after dark, so the air above a built-up area cools far more slowly than the air above open ground. The office reports a difference of 10–14 °F between Phoenix Sky Harbor and the nearby community of Wickenburg, and describes the effect as strongest on calm, dry summer nights, which is to say, on the nights a Phoenix summer is made of.

That is the mechanism behind the flat bottom of the July line. The city is not holding last week’s heat; it is holding the afternoon’s, and giving it back over the hours when a desert would have shed it. It also explains why the two lines on the chart stay so far apart at night and come closest in the late afternoon: the urban difference is a nighttime phenomenon more than a daytime one.

A yard is a small version of the same thing, which is what the Sun Path & Shade Finder is for. Where a wall, a tree or a paved area stands decides how much of the afternoon’s sun a given patch of ground takes in, and therefore how much it has to give back at midnight.

What a warm night does to a person

The CDC names people without air conditioning as a group at higher risk from heat, and lists it alongside age and chronic illness rather than as a footnote. In a climate where the coolest hour of a July day averages 85.6 °F, that is not a general statement.

The reason a warm night matters is that it removes the recovery. A body that spends an afternoon above 100 °F can shed its built-up heat overnight in most climates; here the night itself is above 85 °F, so there is no such hour to use. That is why this site treats the overnight hours as part of the planning problem rather than as the safe half of the day, and why the Heat-Safe Hours Planner runs a full twenty-four hours rather than a daytime window.

It is also why housing is the other half of the answer, and why the article on keeping a house cool at 115 degrees belongs next to this one. On a night that never drops below 85 °F, the difference between a house that vents its stored heat after dark and one that does not is the difference between sleeping and not.

This page describes what the normals do. It does not diagnose anyone and it does not tell anyone what to do about a medical problem. If a person is confused, cannot be woken, or has had a seizure in the heat, the guidance is to call 911.

Where this page stops

Everything above is an average over 1991–2020, and an average is the wrong tool for tomorrow. A specific night can be well under the line or well over it, and the percentiles in the middle of this page are the more honest guide to how far.

There is one more thing to be straight about, because it is a place where two NCEI products measure slightly different things. This page uses the hourly normals throughout, and the coolest hour of a normal July day in that product is 85.6 °F. NCEI’s monthly product, which reports a single daily low rather than an hour-by-hour curve, puts the normal July low at 84.5 °F. The two figures are a degree apart and neither is wrong; they are counting different things, and this page stays with the hourly set so that the chart, the table and the text all come from one dataset.

For the reason the nights matter as much as the days, the dew-point and heat-index article is the companion piece: it covers why the humidity correction points the other way at Sky Harbor, and why the local heat alerts are built on a product that accounts for duration rather than on the heat index.