A quarter of US heat deaths happened below the temperature that triggers a heat alert

Heat alerts are built around thresholds — a temperature or heat index value that, once crossed, triggers a public warning. A study published this month in Weather, Climate, and Society asks a basic but underexamined question: do those thresholds actually line up with when people are dying from heat? The answer, drawn from nearly three decades of confirmed fatalities, is that they often don't [s1].

What the study did

Researchers used the National Oceanic and Atmospheric Administration's publicly available Storm Events Database to identify confirmed heat-related deaths across the contiguous United States, then matched those deaths against detailed weather reanalysis data (ERA5-Land) to examine the actual meteorological conditions present when each death occurred [s1]. The analysis covered 5,465 confirmed heat-related deaths between 1996 and 2024 [s1] and tested seven different heat-stress indicators against that mortality data: maximum and minimum dry-bulb temperature, the universal thermal comfort index, wet-bulb temperature, wet-bulb globe temperature, the standard heat index, and apparent temperature [s1].

The core finding

A quarter of the confirmed deaths in the dataset occurred when the heat index — the most common metric used to issue US heat alerts — was below 33°C (about 91°F), which the study describes as well beneath the benchmarks that commonly trigger a heat alert [s1]. In practical terms, a meaningful share of the people who died from heat in this dataset did so on days that would not have generated a warning under commonly used heat index thresholds.

Humidity metrics tell a more consistent story — outside the Southwest

The study also compared how consistently each of the seven metrics tracked with mortality across different climates. Outside the arid Southwest, humidity-sensitive metrics — wet-bulb temperature and wet-bulb globe temperature — showed the lowest variability among fatal cases, meaning they were the most consistent predictor of when a death occurred [s1]. In the Southwest specifically, that pattern flipped: plain maximum dry-bulb temperature was the most consistent metric there, reflecting the region's characteristically low humidity [s1]. The study frames this as an argument for region-specific metrics rather than a single national standard, since the dry-bulb-focused heat index that dominates US alerting was, by this analysis, not the most reliable indicator in most of the country.

Who is most at risk at lower temperatures

The analysis found that fatality thresholds for children under 10 were lower across every metric tested except wet-bulb temperature [s1] — meaning young children in the dataset tended to die at lower heat readings than the rest of the population, regardless of which indicator was used. The study also found that regions accustomed to extreme heat showed lower variability in the raw temperatures associated with fatal cases, while cooler regions showed more consistency in temperature anomalies — how far a reading departed from the local norm — or in rapid swings from cold to hot conditions [s1], a pattern consistent with the idea that acclimatization changes what "dangerous heat" means for a given population.

What the researchers recommend, and what they're careful not to claim

The study's authors argue their findings support refining heat alert thresholds by region and selecting the most stable metric available for each region, rather than applying one national standard uniformly [s1]. They are explicit, however, that these are described as preliminary findings, and that confirming them would require epidemiological studies built on complete, individual-level exposure data — something this analysis, based on confirmed-death records and reanalysis weather data rather than tracking individual people's actual exposure, does not provide [s1].

Why it matters for how alerts get issued

Heat alert systems exist to prompt behavior change — opening cooling centers, checking on vulnerable neighbors, canceling outdoor activities — and a threshold that misses a quarter of actual deaths represents a real gap between the warning system and the risk it's meant to flag. The study does not claim that fixing the threshold would have prevented those deaths; many of the people who died from heat below alert thresholds may have had other risk factors — age, chronic illness, lack of air conditioning, outdoor occupation — that a temperature-based alert alone would not have addressed regardless of the number attached to it.

Sources

• Optimizing U.S. Heat Alerts: A Multimetric Analysis of Heat-Related Mortality — Weather, Climate, and Society, 2026-05-05