For 31 days in a row in July 2023, Phoenix hit 110°F or more. The air was also very dry, and by the yardstick many climate studies use to decide whether heat can be survived, a wet-bulb temperature of 35°C (95°F), the city never came close to danger. A study published in March 2026 in Nature Communications argues that yardstick missed what was happening. Applying a model of how the human body sweats and stores heat, the authors estimate that a resting person over 65 in direct sun would have faced conditions past the threshold for fatal heatstroke in 24 percent of the six-hour periods that month.

That figure comes from a physiological model run on weather records. It is not a count of deaths, and it is one study, not settled consensus.

It is also a reading of the research, not medical advice.

Why a humidity-based limit underrates dry heat

Wet-bulb temperature folds heat and humidity into one number. It is roughly the temperature a wet cloth would reach as water evaporates from it in moving air, so at any given air temperature it falls as the air gets drier. In 2010, Steven Sherwood and Matthew Huber proposed in PNAS that six hours at a wet-bulb temperature of 35°C would be a hard upper limit for people, because skin could no longer lose heat to the air. Over the following decade, a long list of climate projections adopted that number as the line between survivable and unsurvivable heat.

The limit has a blind spot that matters in a place like Phoenix. It assumes sweat can keep pouring out and evaporating. In humid air, evaporation is the bottleneck. In dry air, evaporation is easy, and the bottleneck becomes the body itself: there is only so much sweat a person can produce in an hour, and older adults produce less of it and wet less of their skin.

Laboratory work had already pointed this way. In a 2022 study, a Penn State team including physiologist W. Larry Kenney put 24 people aged 18 to 34 in an environmental chamber, had them do light exercise such as slow walking, and raised the heat or humidity until their core temperature stopped stabilizing. That happened at wet-bulb temperatures of 25 to 28°C in hot, dry conditions and 30 to 31°C in warm, humid ones, according to the university’s release. That threshold marks the point where the body starts losing ground, not the point of death, but both numbers sat well below 35°C, and the dry-heat figure sat lowest.

What the 2026 model actually measured

The paper’s lead author is Sarah Perkins-Kirkpatrick of the Australian National University. Her co-authors include Jennifer Vanos and Gisel Guzman-Echavarria of Arizona State University and Ollie Jay of the University of Sydney. The model they used, called HEAT-Lim, was developed at those two universities and published by Vanos and colleagues in Nature Communications in 2023. The new paper takes it out of the abstract and asks whether real heat waves have already crossed its lines.

HEAT-Lim asks a narrow question. Given a temperature and humidity, would a person’s core temperature climb to 43°C (about 109°F), the point at which heatstroke is very likely to kill, within six hours? It assumes the person is resting and unclothed. It runs separately for an average younger adult and an average adult over 65, and for someone in shade or in direct sun, where it adds the sun’s radiant load as if the surroundings were 15°C hotter than the air.

The team ran that test on weather reanalysis data for six heat waves: Mecca in June 2024, Bangkok in April 2024, Phoenix in July 2023, Mount Isa in Australia in January 2019, Larkana in Pakistan in 2015 and Seville in August 2003. In every city, older people in direct sun crossed the fatal-heatstroke threshold at some point. In none did any six-hour period come close to a wet-bulb temperature of 35°C.

Why Phoenix came out on top

Phoenix had the highest frequency of deadly periods for older people in the sun: 24 percent of rolling six-hour windows across the month, or 28 percent when the team used weather-station records instead of the gridded data. Larkana was next at 18 percent, then Mecca at 12 percent.

Most of the Phoenix periods that crossed the line had relative humidity of 20 percent or less. That is the paper’s central point about the city, and the authors put it plainly: extremely hot and dry conditions can be just as deadly as hot and humid ones.

Shade changed the picture almost completely. For older people in shade, the model found a single six-hour period in the whole month that crossed the threshold. Younger adults never did, in sun or shade, although the authors say the heat very likely made ordinary activity difficult for them.

What Phoenix’s death records show, and where they do not fit

The human toll that summer is documented separately. The Maricopa County Department of Public Health’s 2023 heat deaths report counted 645 heat-related deaths, 52 percent more than in 2022 and the most ever recorded there. Sixty-four percent occurred in July, an average of 13 a day. Three quarters happened outdoors, and nearly two thirds of all victims were 50 or older.

Those figures fit the model’s warning about outdoor exposure, but they also show what it leaves out. Half of the outdoor deaths were people younger than 49, even though the model never placed its younger-adult profile in non-survivable conditions. Forty-five percent of victims were experiencing homelessness, and the county found substance use involved in 65 percent of deaths. Neither is part of a model that pictures a healthy person sitting still. Every indoor death occurred in a space without working cooling.

The paper itself cautions that its thresholds and real mortality figures are not straightforward to compare.

The limits of the estimate

The model counts only deaths from heatstroke. Many heat deaths are cardiovascular or respiratory, and they can happen at lower body temperatures, so the model misses part of the toll. It also starts every six-hour window with a normal body temperature, while a person exposed all day would start each window already warmer, which the authors say means the true danger thresholds are probably cooler and drier than the ones they used.

The weather data may also understate the danger. The reanalysis grid is about 31 kilometers across and leans on rural instruments, so it can run cooler than a city street. In Seville the team found it underestimated daily highs by up to 3°C.

Anyone worried about how heat affects them or an older relative can ask a doctor about medicines and conditions that make heat harder to tolerate, and local public health agencies publish heat guidance each summer.

The authors point to models that let heat build up over hours and days as the next step. Checking such models against the people who actually died, including the younger, unhoused and substance-using victims the county recorded, would show how much of the toll the thresholds can explain. Until that work is done, the Phoenix figure remains a model’s estimate of exposure, not a measure of how many people the heat killed.

Sources