When Heat Becomes More Frequent, Longer, and More Widespread
The daytime evidence is the clearest test of whether human heat stress is changing beyond a few spectacular records. This chapter focuses on what the global study can establish from the hottest daily UTCI values and from the number, location and timing of threshold-exceeding days. The useful judgment is not simply that the world is warmer, but that intensity, recurrence, geographic reach and season length describe different ways a changing daytime hazard can affect preparedness.
The distinction is important because a trend in the hottest days is not a claim about every day. A count of days above a heat-stress threshold is not a count of people or illnesses. And a large percentage increase in a rare extreme does not necessarily mean that the extreme is common. The study’s value lies in keeping these signals separate and then showing how they point in the same broad direction.
What the hottest days can—and cannot—tell us
The global analysis uses the average of the ten highest daily maximum UTCI values in each year to describe the upper end of daytime heat stress. Since the 1970s, that indicator has risen by 0.27 ± 0.05 °C per decade. The choice of ten days matters. It makes the measure less dependent on one isolated daily anomaly than a single annual maximum would be, while keeping attention on the most intense conditions of the year. It does not describe the typical afternoon, and it does not mean that every day in the year warmed by the same amount.
An upper-tail indicator answers a focused question: how has the high end of the annual daytime heat-stress distribution moved? If it rises, the most severe conditions represented by that measure are becoming hotter on average. It cannot, by itself, tell us how many days crossed a level that matters for heat stress, how long those conditions lasted or how many people encountered them. Those questions require other measures.
The uncertainty attached to the trend is part of the result, not a decorative qualification. The estimate of 0.27 °C per decade is accompanied by an uncertainty of 0.05 °C per decade in the study’s presentation. That range describes uncertainty around the estimated trend; it is not a claim that each year or each location falls inside a narrow band. A global estimate can be precise enough to show direction while still leaving meaningful variation among regions and years.
The study therefore counts annual days when daily maximum UTCI reaches at least three thresholds: 32 °C for strong heat stress, 38 °C for very strong heat stress and 46 °C for extreme heat stress. Since the 1970s, the global averages have increased by about 2 ± 0.52 strong-heat-stress days per decade, 1 ± 0.18 very-strong days per decade and 0.2 ± 0.03 extreme days per decade. The numbers are counts of grid-cell conditions aggregated into a global land average, not a promise that each location gains exactly that many days.
That aggregation is easy to overlook. A global average of two additional strong-heat-stress days per decade is calculated across many land grid cells. It does not mean that a particular resident experienced precisely two extra days, or that every grid cell changed in the same direction. It summarizes a distributed pattern. The local question still requires a local series, and the personal question requires information about where someone was, what they were doing and how they could respond.
Together, the two kinds of measure show why “hotter” is an incomplete description. The hottest days can become more intense, while threshold counts show how often the environment enters a range associated with heat stress. A place could set a higher extreme without changing the number of ordinary hot days very much. It could also accumulate more strong-heat-stress days without producing a dramatic change in its single hottest day. A sound interpretation asks which part of the distribution has changed.
The increases are also not evenly distributed across categories. The study reports the largest relative frequency changes at the most severe thresholds. That pattern has a practical meaning: even when the most extreme category remains less frequent in absolute terms than moderate or strong heat stress, its proportional change can be large. Severe conditions can therefore become more common without becoming the everyday baseline everywhere.
There is a second reason to resist reading these figures as personal probabilities. The continental frequency distributions are built by summing occurrences across grid cells and days. Their absolute counts are not directly interpretable as the number of individual heat episodes or the number of people exposed. The result is evidence of changing environmental hazard conditions at the scale represented by the data. To estimate human exposure, the analysis must add population information; to estimate health outcomes, it would need outcome data and a different design.
This separation also protects against a common causal shortcut. A rising UTCI threshold count supports the conclusion that the modeled thermal environment is entering stressful ranges more often. It does not, on its own, establish a fixed increase in hospitalizations or mortality. Those outcomes can be related to the hazard, but they also depend on how long people are exposed, whether they can cool down, what they are doing and what health or social conditions they bring to the exposure.
A hazard that spreads across maps and calendars
The global average becomes more concrete when the study compares continents. Between the 1970s and 2015–2024, extreme heat stress occurred about 2.5 times as often in Europe and South America, twice as often in North America, and 1.8, 1.7 and 1.2 times as often in Africa, Oceania and Asia. These multiplication factors describe the relative change in aggregated exceedance frequency above the extreme threshold. They do not rank continents by total heat burden, and they do not imply that an extreme event carries the same practical meaning in every region.
Why can a rare category rise by a large factor and still occupy a small share of days? The denominator is the earlier frequency. Moving from a very small number of occurrences to a larger but still limited number produces a large ratio. Relative change is useful for showing how quickly a tail is shifting; absolute frequency is needed to understand how prevalent the condition is. A careful account uses both ideas and avoids presenting a multiplier as a direct measure of personal risk.
For example, a place that moved from almost no extreme-threshold exceedances to a few would show a striking relative increase, but that ratio would not tell us whether most residents regularly faced extreme conditions. Conversely, a region with a lower multiplier might still have a larger total burden if its baseline was already high. The study uses the multipliers to compare change between periods, not to create a league table of human suffering.
The study also finds a change in the map itself. Some locations that had recorded an average of zero heat-stress days in the 1970s recorded a non-zero annual average in 2015–2024. In parts of southern North America, southern Europe, northern and southern Africa and South America, the recent decade brought up to 50 additional days per year with at least strong heat stress. This is a different kind of signal from a hotter extreme at a location that has always experienced severe heat. It describes the arrival or regular appearance of a condition outside the earlier local pattern.
“Non-zero annual average” should not be read as “every place has a heat-stress day every year” or “every resident is exposed for 50 additional days.” It means that, in the gridded and averaged comparison, a condition absent on average in the earlier period appears in the later one. That is enough to identify a changing footprint, but not enough to describe the precise timing or intensity at a particular street, workplace or household.
That arrival can create an adaptation problem even when the new location is not the hottest place on the map. Buildings, work practices, public warnings and expectations may reflect historical experience. A region that has rarely crossed a strong-heat-stress threshold may have less institutional memory and less tested capacity for responding to it. The study does not measure each region’s preparedness, so it cannot quantify that capacity gap. It does show why a map of changing hazard footprint matters beyond a map of record temperatures.
The strongest local increases in the hottest-day measure are reported in Europe, northern Africa and the Arabian Peninsula, reaching up to 4 °C and locally 5 °C when the recent decade is compared with the 1970s. Elsewhere, increases of 2–3 °C dominate, while small areas in Greenland, coastal southwestern Australia and parts of India and Pakistan show decreases in the daytime indicator. The exceptions are not a refutation of the global pattern. They show that a global average is a summary of different regional pathways, not a claim of uniform change.
Season length adds a calendar dimension. In the northern extratropics, the study defines a heat-stress season as the interval between the first and last day in a year on which a selected UTCI threshold is exceeded. Compared with the 1970s, that interval is about 15 days longer for moderate heat stress, 12 days longer for strong heat stress, 9 days longer for very strong heat stress and 6 days longer for extreme heat stress on average. The change is not simply a higher peak; it is a longer period in which a threshold can matter.
The seasonal measure describes the span between boundary dates, not an uninterrupted heatwave. A season can lengthen because the first exceedance arrives earlier, the last one arrives later, or both. It also averages across extratropical grid cells and years, so it should not be used to predict the exact start or end date for a particular community. Its strength is comparative: it shows that the calendar window associated with heat stress has shifted.
Europe provides a concrete example. Moderate heat stress now begins on average in mid-May rather than early June, and strong heat stress begins in June rather than July. The result is a longer planning window for protecting people and infrastructure. It also changes how a community defines an exceptional period. If the season starts earlier and ends later, a plan built around a narrow midsummer episode may leave gaps at both edges.
The seasonal measure has a clear boundary. The analysis excludes the tropics because lower categories of heat stress are present much of the year and the seasonal cycle is less distinct. The measure therefore supports a meaningful comparison of seasonal timing in the northern extratropics, but it should not be generalized to every latitude. In tropical regions, the more relevant questions may concern persistence, severity shifts and nighttime relief rather than the first and last exceedance dates.
The four branches in the diagram should be read together but not collapsed. Intensity tells us how hot the upper tail became. Frequency tells us how often a threshold was crossed. Footprint tells us where the condition appeared. Season length tells us when it became relevant. Regional baselines affect the meaning of every comparison. None of these branches alone measures the health outcome, and none alone describes the full exposure of a population.
A useful reading habit follows from this structure. Before accepting a headline number, ask what is being averaged, which threshold is used, which periods are compared, and whether the denominator is days, grid cells or people. Then ask whether the result describes intensity, frequency, footprint or timing. Finally, ask what evidence would be needed to move from a changing hazard to a claim about exposure or health. Those questions do not weaken the finding; they prevent one valid finding from being stretched into several unsupported ones.
The study’s data design reinforces that point. The values are modeled, gridded indicators compared over two multiyear periods. That structure is strong for identifying broad, transboundary patterns and for reducing the influence of a single unusual year. It is not a fine-grained record of every street, building or individual. Statistical evidence that the global indicators are changing does not mean that every local trend is identical, nor that a local exception can be ignored when a decision is made there.
The daytime record therefore establishes a strong but bounded conclusion. The most intense daytime conditions have risen; days above heat-stress thresholds have increased; the hazard has expanded into locations that previously recorded none; and in much of the northern extratropics, the season has lengthened. These are different measurements of a related transition. To understand what the transition means for cumulative human exposure, one more temporal question is unavoidable: what happens after the sun goes down, and what if the body does not receive a meaningful break before the next hot day?