Global heat stress intensification and its expanding footprint on the human population

Rebecca Emerton, Julien Nicolas, Anna Lombardi & Claudia Di Napoli  Nature Climate Change (2026) || Climate warming is driving more frequent and intense heat extremes. Yet, changes in heat stress, the leading cause of weather-related mortality, remain poorly quantified at the global scale. Here, using the Universal Thermal Climate Index, we assess heat stress globally since 1950, examining daytime extremes, nocturnal heat and compound daytime–nighttime events, revealing a pronounced, multidimensional intensification. Extreme ‘feels-like’ temperatures have become more frequent on every continent, and the spatial footprint of hazardous heat has expanded, exposing previously unaffected regions. Heat stress days and tropical nights have increased, with some regions experiencing up to 50 additional heat stress days annually and an extended heat stress season. The hottest nights of the year are warming faster (0.32 °C per decade) than the hottest days (0.27 °C), and compound events are more frequent, severe and prolonged. Population exposure to dangerous heat has increased markedly, driven by intensifying heat stress in addition to population growth.

Emerton, R., Nicolas, J., Lombardi, A. et al. Global heat stress intensification and its expanding footprint on the human population. Nat. Clim. Chang. (2026). https://doi.org/10.1038/s41558-026-02670-5

---

---

Heat is the leading cause of weather-related mortality worldwide and exacerbates underlying illnesses such as cardiovascular, respiratory and mental health conditions¹. As global warming intensifies, heatwaves are becoming more frequent, longer and more severe^2,3,4, further increasing the risk of heat-related illness and death⁵. One quarter of heatwaves between 2000 and 2019 were virtually impossible without climate change⁶. Yet, while trends in heatwaves are well documented, many people worldwide are exposed to chronic heat⁷, and less is known about how these changes translate into trends of heat as experienced by the human population, particularly at the global scale.

Thermal stress indices quantify heat stress, defined as the net heat load on an individual, accounting for environmental and personal parameters⁸. Due to the nonlinearity of these indices, whereby equal warming at higher temperatures leads to a greater increase in thermal indices, perceived temperatures are rising faster than air temperature⁹. Here, we use the Universal Thermal Climate Index (UTCI), a widely adopted biometeorological thermal stress index¹⁰ accounting for temperature, humidity, wind and radiation, alongside modelling how the human body responds to the environment. Expressed as a ‘feels-like’ temperature (°C), the UTCI classifies ten categories of cold to extreme heat stress, based on impacts on the human body¹¹.

The ERA5-HEAT reanalysis^12,13 provides globally consistent UTCI data from 1940 to the present, enabling long-term, transboundary assessment of trends. Recent analysis¹⁴ revealed mean UTCI increases of up to 2.5 °C across North America, Europe and Asia in 1991–2020 compared with 1941–1970^14,15. These patterns highlight the need for a comprehensive global assessment of evolving heat stress extremes, including their intensity, frequency and duration, which is addressed in this work.

Using ERA5-HEAT, we examine changes in heat stress across the globe from 1950 to 2024 and contrast recent conditions (2015–2024) with the 1970s, the point from which heat stress indicators have risen markedly at the global scale. We assess trends in maximum and minimum feels-like temperatures on the hottest days of each year and quantify the occurrence of strong (UTCI ≥32 °C), very strong (UTCI ≥38 °C) and extreme (UTCI ≥46 °C) heat stress. Because high nighttime temperatures amplify health risks by limiting recovery and increasing mortality¹⁶, we also analyse the evolution of tropical nights (minimum temperature ≥20 °C). We characterize tropical nights by severity of nocturnal heat stress, and investigate compound events^9,17 that couple daytime and nighttime heat.

To link physical trends to societal impacts, we combine UTCI data with global demographic data¹⁸ to quantify population exposure to heat stress and how the number of people experiencing hazardous thermal conditions has evolved. We further attribute changes in exposure to the relative influences of population growth versus changes in heat stress.

Our analysis reveals a multidimensional intensification of heat stress worldwide, with implications for human exposure.

Heat stress trends at the global scale

Global average indicators of feels-like temperatures, heat stress days and tropical nights have all risen markedly since the 1970s (Fig. 1). The maximum feels-like temperature on the ten warmest days of the year has increased by 0.27 ± 0.05 °C per decade since the 1970s, while the ten warmest nighttime minima have risen even faster, at 0.32 ± 0.05 °C per decade. These trends translate to an additional two days per decade with at least strong heat stress, one with at least very strong heat stress and two additional tropical nights.

Building on these trends, we map changes across the globe, comparing the most recent decade (2015–2024) with the 1970s to capture the magnitude of change (Fig. 1).

Daily maximum feels-like temperatures on the ten warmest days of the year have increased across most regions, with the strongest warming in Europe, northern Africa and the Arabian Peninsula—up to 4 °C, and locally 5 °C—warmer in the last 10 years than in the 1970s. Elsewhere, increases of 2–3 °C dominate, although small areas in Greenland, coastal southwestern Australia and parts of India and Pakistan show decreases of 1–3 °C. Research supports the decreasing trend in the UTCI in parts of India^19,20,21,22. The warmest nighttime minima have risen almost everywhere, including regions with slight daytime cooling. The largest increases occur in the northern extratropics: up to 6 °C in northwestern North America and northern Canada, 5 °C in northern Africa, and 4 °C in Europe and the Arabian Peninsula.

Heat stress days and tropical nights have proliferated globally. The spatial footprint of heat stress has also expanded, meaning that regions previously untouched by extreme heat are now affected (Extended Data Fig. 1). Subtropical regions, including in southern North America, southern Europe, northern and southern Africa, and South America, now experience up to 50 more days per year with at least strong heat stress. Many of these areas also see more days with very strong heat stress, and some places, such as in northern Africa, the Arabian Peninsula, Australia and parts of western North America have also seen an increase in extreme heat stress days. In the tropics, strong heat stress typically remains nearly year-round, but there are shifts towards the more severe categories. Exceptions include parts of India, Pakistan and coastal southwestern Australia, where heat stress days have declined, although India and Pakistan are experiencing more tropical nights.

Changes in the frequency and duration of heat stress

On every continent, feels-like temperatures are more frequently exceeding thresholds for very strong and extreme heat stress in our current climate (2015–2024) compared with the 1970s (Fig. 2). Extreme heat stress, at which point urgent action is required to prevent severe health impacts¹¹, now occurs 2.5 times more often in Europe and South America, twice as often in North America and 1.8, 1.7 and 1.2 times more often in Africa, Oceania and Asia, respectively. Extended Data Table 1 provides the equivalent frequency changes for the Intergovernmental Panel on Climate Change Sixth Assessment Report reference regions.

Although extreme heat stress remains rare in Europe, affecting 0.2% of days and locations in the last 10 years, this represents a 138% relative increase from 0.08% in the 1970s. In Africa, the frequency has risen from 4.1% to 7.4%, the highest of any continent. Africa also records the greatest overall heat stress burden, with a strong heat stress frequency of 70% in the past decade, compared with 63% in South America, 55% in Oceania, 24% in Asia, 15% in North America and 9% in Europe.

Across all continents, the largest relative increases in frequency occur in the most severe heat stress categories.

In the Northern Hemisphere, the season during which heat stress is experienced has increased in duration (Fig. 3). Averaged across the northern extratropics, the first and last days with moderate heat stress now span a season 15 days longer than in the 1970s. For strong heat stress, the season is an average of 12 days longer; for very strong, 9 days; and for extreme, 6 days.

Europe and Africa (Northern Hemisphere) show the longest season extensions. In Europe, moderate heat stress now begins on average in mid-May rather than early June and persists until nearly October, while strong heat stress starts in June instead of July. For these levels of heat stress, the shortest heat stress season in the last 10 years was longer than the longest season in the 1970s (Extended Data Fig. 3). Similar, although less pronounced, trends occur in Africa, North America and Asia. In Africa, the extreme heat stress season has lengthened by 28 days—the largest increase observed, with the shortest extreme heat stress season in the last 10 years lasting 11 days longer than the longest season in the 1970s Elsewhere, the extreme heat stress season remains similar in duration but shifts later, most notably in North America, where it begins 10 days later and ends 7 days later than in the 1970s.

Compound heat stress days and tropical nights

Rising nighttime feels-like temperatures underscore the need to assess not only the occurrence of tropical nights but also the intensity of heat stress they impose. Here, we classify tropical nights by the heat stress category of their minimum overnight UTCI (Fig. 4).

There is a decreasing trend in the number of tropical nights with no heat stress (UTCI <26 °C) since the 1970s, accompanied by an increasing trend in the number of tropical nights with moderate or strong heat stress. The proportion of tropical nights with moderate heat stress peaked at 11.4% in 2024, the warmest year on record for the globe, up from a minimum of 2.2% in 1965. Before 1998, a strong El Niño year^23,24, the share never exceeded 5.8%.

Strong heat stress on tropical nights was virtually absent before 1998 (≤0.02%) and remained at or below 0.04% until 2023, but rose to 0.08% in 2024. No tropical nights have yet recorded very strong or extreme heat stress.

More frequent heat stress days and tropical nights are driving an increase in compound heat events (Fig. 5), which globally have become more frequent and prolonged, especially in Europe, Africa and North America.

In Europe, the occurrence of single-day compound events (one heat stress day followed by one tropical night) has increased by 73% since the 1970s. Events lasting 15–30 days are now 3.4 times more common, and the longest observed durations—up to 120 consecutive days—have almost doubled in frequency.

In Africa (Northern Hemisphere), shorter-duration compound events (1–90 days) have declined as sequences of up to a full year (271–365 days) have increased by a factor of 2.8. In the Southern Hemisphere, Africa has seen increases across all durations, with the longest events (181–270 days) now occurring 2.8 times more frequently. Asia and Oceania exhibit smaller changes, with modest increases for most durations and larger increases for the longest events.

Population exposure to heat stress

Population exposure to heat stress has increased in recent decades, driven by rising frequency, intensity and duration of heat stress alongside population growth and redistribution (Fig. 6). Comparing exposure by heat stress severity and duration in the last 10 years (2015–2024) with the 1970s reveals that increases are steepest for the most severe categories and longest durations, meaning more people are exposed, for more days, and to more intense heat stress.

In the 1970s, 55% of the global population experienced at least 90 days of strong heat stress annually; this has risen to 70% in our current climate. Exposure to at least one day of extreme heat stress has grown from 16% to 22%; accounting also for the increase in population, this represents an additional one billion people. For most categories, the increase in exposure caused by changes in heat stress is equal to or larger than the increase caused by population growth and redistribution. This may be further amplified by the spatial variability of population growth, with some of the largest increases occurring in regions with hot climates²⁵.

Mapping annual person-days of exposure to very strong heat stress in the last 10 years (Extended Data Fig. 4) identifies regions of greatest population exposure, including sub-Saharan Africa, South and Southeast Asia, the Arabian Peninsula and the Mediterranean.

Given ERA5-HEAT limitations in capturing extremes and microclimates such as urban heat islands²⁶, these estimates are likely to be conservative, particularly for cities. They also draw on a hybrid gridded population dataset that, while providing long-term coverage, carries uncertainties in early-period and regional population distributions.

Discussion and conclusions

This study provides insights into changes in the frequency, severity and duration of heat stress with climate change, and a comprehensive global-scale analysis of not only daytime extremes but also nocturnal heat and compound daytime–nighttime events. Our analysis is based on the UTCI; by incorporating the combined effects of temperature, humidity, wind and radiation, and how the human body responds to the thermal environment, the UTCI offers a physiologically relevant measure of thermal stress. We demonstrate a clear and accelerating climate-driven intensification of heat stress since the 1970s, with implications for the health of billions of people worldwide.

Feels-like temperatures have risen markedly, with nighttime minima increasing faster (global average of 0.32 ± 0.05 °C per decade since the 1970s) than daytime maxima (0.27 ± 0.05 °C per decade). These changes translate into more frequent and severe heat stress days and, in the Northern Hemisphere, a prolonged heat stress season. Importantly, the most dramatic shifts often occur at the highest thresholds of heat stress, where health risks are greatest. The spatial footprint of hazardous heat stress has expanded, exposing previously unaffected regions and posing adaptation challenges not only in areas where heat stress is becoming more extreme, but also in those where it has historically not been experienced.

Population exposure has increased across all categories and durations of heat stress, compounded by population growth. In the 1970s, 55% of the global population experienced at least 90 days of strong heat stress annually; this has risen to 70% in our current climate. Exposure to at least one day of extreme heat stress has increased from 16% to 22%, representing an additional one billion people who now experience extreme heat stress annually compared with the 1970s. In the future, exposure to heat stress will continue to increase, with the largest absolute increases in low-latitude regions, and the largest relative increases in high-latitude regions²⁷. A recent United Nations Children’s Fund report²⁸ highlights that around 559 million children are already exposed to high heatwave frequency, and that virtually every child on Earth is forecast to face more frequent heatwaves by 2050, even if global warming is kept below 2 °C above the pre-industrial level, as agreed in the 2015 Paris Agreement. Extreme heat can be particularly life-threatening to young children, as they are less able to regulate their body temperature²⁸.

Our findings also reveal previously unquantified aspects of heat stress at the global scale, focussing on changes in compound heat stress days and tropical nights. Periods of heat stress during the day with little relief overnight can have severe implications for health and mortality, and the evidence indicates that the duration of exposure to heat stress is an important factor for human health²⁹. We demonstrate that the number of tropical nights with no heat stress is declining, due to an increasing proportion of nights in which feels-like temperatures still exceed heat stress thresholds.

We show that compound events combining consecutive heat stress days and tropical nights have become more frequent and prolonged on every continent, with the most dramatic changes in Europe and Africa. These patterns underscore the growing importance of heat stress dynamics beyond just daytime extremes.

The observed rise in compound events aligns with emerging evidence that concurrent daytime and nighttime heat poses disproportionate health risks due to the absence of nocturnal cooling^30,31. Our classification of tropical nights by heat stress severity adds further insight to this understanding, revealing that nocturnal heat is not only more frequent but also more intense.

Our findings complement previous work on heatwave trends under anthropogenic warming^5,6,32,33,34 and extend our understanding by using the UTCI and considering heat stress year-round rather than confined to specific events, given the exposure of many people worldwide, predominantly in the tropics, to chronic heat⁷. While most global assessments focus on air temperature, our use of the UTCI reveals additional dimensions driven not only by changes in temperature but also by changes in humidity, wind and radiation. This distinction is critical: recent studies^9,30 show that the nonlinearity of thermal indices means they change at a different rate than air temperature alone and, in some cases, may diverge from temperature trends—for example, in parts of India, where daytime feels-like temperatures have decreased since the 1970s while nighttime minima have increased.

Increasing heat stress and exposure has profound implications for climate adaptation and public health. Heat stress is already the leading cause of weather-related mortality^1,35,36, and our results show that the dangers continue to increase. Adaptation strategies must address both daytime and nighttime heat. Heat-health action plans, early warnings systems and urban cooling interventions are essential to reduce exposure. Integrating heat stress metrics into climate risk assessments and adaptation strategies is vital, as temperature-based thresholds may underestimate risk in humid or low-wind environments. At present, adaptive capacity varies across regions and populations, but such adaptation strategies may help to reduce vulnerability.

While this study focuses on historical trends to understand how our current climate differs from that of recent decades, projections under different emissions scenarios are urgently needed to inform long-term adaptation planning. At present, consistent global climate projections for the UTCI do not exist; future work should endeavour to produce and assess such projections. Recent research into exposure to wet-bulb temperature found that humanity is more vulnerable to moist heat stress than had previously been proposed, and that in the future, moist heat extremes will lie outside the bounds of human experience and beyond current mitigation strategies for billions of people²⁹. High-resolution datasets will also be critical for quantifying urban heat stress for example. While current state-of-the-art reanalyses such as ERA5-HEAT have been found to be suitable and useful for climate and health studies²⁶, they cannot fully resolve microclimates such as urban heat islands. Our estimates may therefore underrepresent extremes and population exposure, particularly for cities. In addition, the UTCI models thermal stress for a standardized reference individual and does not account for individual variability, including differences in age, body composition or acclimatization.

Overall, our findings reveal a multidimensional intensification of heat stress, and an increase in exposure that exceeds the effect of population growth alone. Exposure is increasing across all severity thresholds, as the magnitude, frequency and duration of heat stress all increase, both during the day and at night. These changes pose escalating threats to health, livelihoods and economic productivity worldwide. Without urgent mitigation and adaptation, billions more people may face dangerous heat stress conditions in the coming decades. These comprehensive results represent important considerations for climate policy and planning that are essential to safeguard human health in a warming world.

Methods

Data

Reanalyses of temperature and thermal stress

Reanalysis datasets combine observational data with state-of-the-art models to ‘fill in the gaps’ where observations are scarce or absent, providing globally consistent datasets of past weather and climate. The ERA5^37,38,39 and ERA5-HEAT^12,13 reanalyses are used here to allow long-term and transboundary assessment of thermal stress trends. They provide global (excluding Antarctica, for ERA5-HEAT) hourly data from 1940 to near-real time at a 0.25° grid resolution. The variables used are 2-m air temperature from ERA5, and the UTCI from ERA5-HEAT. ERA5-HEAT combines air temperature, wind speed, radiation and dewpoint temperature (which, together with the air temperature, represents the humidity) from ERA5, and for any combination of these, the UTCI is defined as the air temperature of a reference outdoor environment that would elicit in the human body the same physiological model’s response (sweat production, shivering, skin wettedness, skin blood flow and rectal, mean skin and face temperatures) as the actual environment¹². It therefore represents a feels-like temperature in degrees Celsius.

The UTCI categorizes thermal stress on the basis of the feels-like temperature and corresponding physiological responses. Throughout this study, we use these established thresholds for at least moderate (26 °C), strong (32 °C), very strong (38 °C) and extreme (46 °C) heat stress. These physiologically based thresholds are designed to be globally applicable, enabling a consistent and interpretable assessment of heat stress and trends globally, with direct relevance for heat-health applications and planning. The UTCI models thermal stress for a standardized reference individual, however, and does not account for individual variability, including differences in age, body composition or acclimatization.

Population data

The dataset used for the population exposure analysis presented here is the ‘Hybrid gridded demographic data for the world, 1950–2020’ dataset¹⁸, which provides 5-year population bands at a 0.25° grid resolution. This dataset is chosen due to its consistent temporal and spatial coverage with the ERA5 dataset; the 0.25° resolution release of this dataset was designed explicitly for the purpose of matching ERA5. It combines the National Aeronautics and Space Administration (NASA) Socioeconomic Data and Applications Center (SEDAC) Gridded Population of the World version 4 dataset for 2000–2020 with the ISIMIP Histsoc gridded population data for 1950–1999 and the United Nations World Population Program demographic modelling data.

Methodology

Heat stress trends at the global scale

To conduct a comprehensive analysis of global heat stress trends, a range of heat stress indicators were defined: (1) the average daily maximum UTCI on the 10 days with the highest daily maximum UTCI per year, (2) the average daily minimum UTCI on the 10 days with the highest daily minimum UTCI per year, (3) counts of heat stress days for different categories of heat stress, and (4) counts of tropical nights.

For (1) and (2), the daily maximum and minimum UTCI were extracted from the ERA5-HEAT dataset for each grid cell. For each year from 1950 to 2024, the ten highest daily maximum values and the ten highest daily minimum values were identified, and their means were calculated to produce annual metrics representing the hottest days and nights. This approach enables a more robust characterization of changes in extreme UTCI conditions by increasing the sample size and reducing sensitivity to single-day anomalies, compared with using only the single hottest day or night of each year.

For (3), the daily maximum UTCI for each day of the year was compared with the thresholds for at least strong (≥32 °C), very strong (≥38 °C) and extreme (≥46 °C) heat stress in order to count the number of days per year exceeding these thresholds at each grid cell. This process was repeated for each year from 1950 to 2024. The methodology for (4) is similar. The daily minimum air temperature was extracted from ERA5 for each day and compared with the 20 °C threshold used to define a tropical night. The number of days per year on which the daily minimum temperature did not fall below 20 °C was counted.

All indicators are based on the maximum or minimum of the data during each 24-h period from 00:00 to 23:59 UTC; adjustments are not made for local time zones.

To produce a global annual time series for each indicator, the mean over all land grid cells was computed. The global trends indicated in Fig. 1 were estimated using least-squares linear regression of the annual values from 1970 to 2024, applied to each indicator, and multiplied by ten to express changes per decade.

Epochal differences

Based on the indicators described above, global maps of the differences between the 1970s and the most recent decade were created. The 1970s were chosen as the reference period because this marks the onset of a clear and sustained global trend in all indicators. The comparison between these two decades is therefore continued throughout the study.

To quantify the change in key heat stress indicators between the 1970s and the most recent decade, the decadal mean for the 1970s (1970–1979) and for the last 10 years (2015–2024) was calculated for each grid cell, and the former subtracted from the latter to obtain the spatial pattern of change.

This epochal differences approach provides a direct and temporally transparent measure of change. By contrasting two multiyear periods, it avoids the assumptions of linearity and temporal homogeneity inherent in long-term trend fitting, while suppressing the influence of interannual variability associated with the El Niño Southern Oscillation and other short-lived perturbations. In addition, because the method produces spatial fields representing concrete, absolute changes between two well-defined periods, it offers an intuitive and accessible way to convey changes in heat stress indicators that is both useful for scientific analysis and for communicating findings to a broad audience.

Changes in the frequency of heat stress

To examine changes in the frequency of occurrence of heat stress conditions for each continent, frequency distributions of daily maximum feels-like temperatures were computed using the UTCI. For all land grid cells within each continental domain, daily maximum UTCI values were extracted for two periods: the 1970s and the most recent decade (2015–2024). For each period, the total number of occurrences within successive 1 °C UTCI bins were calculated by summing counts across all grid cells and all days within each decade. This yields frequency distributions characterizing the frequency of the full range of feels-like temperatures experienced on each continent for each of the two decades considered.

To quantify changes in extreme heat, we computed the frequency with which the UTCI exceeds established thresholds for at least strong (32 °C), very strong (38 °C) and extreme (46 °C) heat stress. For each threshold and continent, the multiplication factor is defined as the ratio of exceedance frequencies in the last 10 years relative to the 1970s.

The continental domains used in this study are as follows: Europe 35–70° N, 25° W–45° E; North America 15–70° N, 170–50° W; Africa 35° S–38° N, 20° W–55° E; South America 60° S–15° N, 90–30° W; Asia 5–80° N, 45–180° E; Oceania 50° S–10° N, 110–180° E.

Changes in the duration of heat stress

To expand on the analysis of the changing frequency of heat stress, we assessed how both the timing and duration of heat stress have changed by analysing the length of the ‘heat stress season’ for each continent in the 1970s and in the most recent decade. The heat stress season is defined as the interval between the first and last days in a year on which UTCI exceeds thresholds for heat stress. For each year and for four UTCI thresholds representing at least moderate (≥26 °C), strong (≥32 °C), very strong (≥38 °C) and extreme (≥46 °C) heat stress, we identified the earliest and latest exceedance dates for each grid cell.

Because seasonal cycles differ between hemispheres, exceedances were evaluated from January to December in the Northern Hemisphere and from July to June in the Southern Hemisphere. The tropics (20° N–20° S) were excluded because heat stress occurs throughout much of the year in tropical regions, meaning that a distinct season cannot be reliably defined.

For each continent, extratropical land grid cells were averaged to obtain the decadal mean onset and end dates for the heat stress season in the 1970s and in the last 10 years (2015–2024). Season duration was then calculated as the difference between these mean onset and end dates, allowing comparison of how the period of the year affected by different levels of heat stress has evolved over time.

Compound heat stress days and tropical nights

Intensity of heat stress on tropical nights

A key objective of this study was to assess changes not only in daytime heat extremes, but also in nocturnal heat, given that high nighttime temperatures amplify health risks by limiting recovery and increasing mortality¹⁶. To investigate this, we first classify tropical nights according to the severity of heat stress experienced.

For each day in the period from 1950 to 2024, ERA5 daily minimum air temperature was extracted and used to generate a binary mask identifying whether each land each grid cell experienced a tropical night (minimum temperature ≥20 °C).

Daily minimum UTCI from ERA5-HEAT was then extracted for the same period. For each day, the tropical night mask was applied to the UTCI field, retaining only those grid points that experienced a tropical night. For each year, the number of grid-cell-days (that is, all land grid cells across all days) falling within each UTCI heat stress category (no heat stress <26 °C, moderate 26–32 °C, strong 32–38 °C, very strong 38–46 °C and extreme ≥46 °C) was counted. These counts were then normalized by the total number of grid-cell-days with tropical nights in that year, producing the percentage of tropical nights during which minimum UTCI still exceeded thresholds for heat stress.

This approach allows characterization of the severity of nighttime heat conditional on tropical night occurrence, rather than conflating changes in overall tropical night frequency with changes in nocturnal heat stress intensity, and provides a clear measure of how often nights that are already warm also impose physiologically meaningful heat stress.

While the UTCI provides a physiologically based assessment of heat stress, the categories were originally developed for daytime conditions under typical solar and metabolic loads. As a result, applying these categories to nocturnal minimum UTCI introduces some limitations. At night, the absence of solar radiation, reduced activity levels and altered thermoregulatory dynamics mean that human heat strain may not correspond directly to the same UTCI threshold ranges used during the day. Consequently, the categorization applied here should be interpreted as an indicative measure of relative nocturnal heat stress severity for assessing changes in the intensity of warm nights. Future work should endeavour to evaluate the most appropriate physiologically relevant thresholds for nocturnal heat stress, an emerging focus in recent studies⁴⁰.

Frequency and duration of compound events

Having considered both daytime and nighttime heat stress thus far, the next component of the analysis focuses on changes in the frequency and duration of compound events, defined here as sequences of consecutive heat stress days and tropical nights.

To identify such events, we used the tropical night masks described above for the 1970s and the most recent decade (2015–2024). In parallel, we constructed daily masks for strong heat stress days, retaining only land grid cells where the daily maximum UTCI reached at least 32 °C. For each day, these two masks were combined to isolate grid cells experiencing both strong daytime heat stress and a tropical night within the same 24-h period.

For each decade (the 1970s and 2015–2024), and for each land grid cell, sequences of consecutive days meeting this compound event criterion were identified, and the duration (in days) of each sequence was recorded. Compound events were then aggregated into duration bins ranging from 1 day (that is, 1 day and 1 night) to up to a full year (271–365 days), yielding the total frequency of events of different durations across all days and all land grid cells within each continent, and for the globe as a whole. To account for hemispheric differences in seasonality, results for Africa were calculated separately for the northern and southern hemispheres. For each region and compound event duration bin, a multiplication factor was derived as the ratio of event counts in the most recent decade to those in the 1970s.

Population exposure to heat stress

To quantify changes in population exposure to heat stress over time, we combined ERA5-HEAT UTCI data with the ‘Hybrid gridded demographic data for the world’ dataset¹⁸, again comparing the most recent decade (2015–2024) with the 1970s. As the population dataset is available in 5-year intervals, data from 1975 were used to represent the population in the 1970s, and data from 2020 to represent the most recent decade. The total population across all age bands was used. The total global population statistics are noted in Fig. 6 for 1975 and 2020, and agree well with other sources and datasets.

Population exposure was quantified through two complementary metrics. First, the total person-days of heat exposure were calculated for each grid cell by multiplying the decadal average annual number of heat stress days by the population in the corresponding grid cell, and summing globally. This was done for both the 1970s and the last 10 years (2015–2024), and for three heat stress thresholds; at least strong (≥32 °C), very strong (≥38 °C) and extreme (≥46 °C). The results for the most recent decade (2015–2024) are mapped in Extended Data Fig. 4.

Second, to assess the scale of exposure to sustained heat, we evaluated the number of people living in locations that exceeded a minimum number of heat stress days per year, in the 1970s compared with the last 10 years (2015–2024). For each threshold of ≥1, ≥10, ≥14, ≥30 and ≥90 days, a binary mask was applied to identify grid cells where the decadal average annual number of heat stress days exceeded the specified count of heat stress days, again for the same three heat stress thresholds. The exposed population was then computed by summing the population across those grid cells and normalizing by the total global population for the corresponding decade.

As part of this analysis, the increase in exposure to heat stress between the 1970s and the last 10 years (2015–2024) was decomposed into the increase caused by population growth (and redistribution) and the increase caused by changes in heat stress. This was done using counterfactual decomposition. First, the exposure in the 1970s was computed as outlined above, combining the population and heat stress days data for the 1970s. Second, a counterfactual scenario was computed, combining the population in 2020 with the heat stress days data for the 1970s, to assess the exposure if the climate had remained the same but population growth and redistribution had occurred as observed. Comparing these gives the increase in exposure due to population growth. Finally, the exposure in the last 10 years was computed, combining the population and heat stress days data for 2015–2024. Comparing this with the counterfactual scenario gives the increase in population exposed due to changes in heat stress.

Expressing exposure as a proportion of the total population at the time, rather than as an absolute number, reduces the influence of population growth over time, as absolute exposed population inevitably increases with global population growth; percentage exposure, by contrast, reflects the share of the population experiencing hazardous heat and is therefore more indicative of climate-driven changes in heat stress exposure.

Several limitations should be acknowledged when interpreting the exposure estimates produced by this analysis. The method relies on gridded population snapshots taken at a single reference year for each decade, which cannot capture intradecadal demographic changes, such as migration or urbanization. The exposure metrics are also sensitive to the spatial and temporal resolution of the underlying datasets. Given ERA5-HEAT limitations in capturing extremes and microclimates such as urban heat islands²⁶, the exposure estimates are likely to be conservative. Historical population reconstructions, particularly in earlier decades, may depend on sparse or inconsistent census inputs, introducing additional uncertainty when comparing population exposure across decades. Finally, while expressing exposure as a percentage of the global population helps to separate population growth from climate-driven changes, it cannot completely disentangle the combined effects of spatial population redistribution and growth, and changing patterns of heat stress.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

All datasets used in this study are freely and openly available. ERA5 reanalysis data can be downloaded from the Copernicus Climate Data Store (https://cds.climate.copernicus.eu). The ERA5 daily minimum 2-m air temperature data are available at https://cds.climate.copernicus.eu/datasets/derived-era5-single-levels-daily-statistics, and the ERA5-HEAT UTCI hourly data alongside postprocessed statistics such as daily/monthly/seasonal/annual maximum and minimum values and counts of heat/cold stress days are available at https://cds.climate.copernicus.eu/datasets/derived-utci-historical. These data can also be explored interactively using Thermal Trace (thermaltrace.climate.copernicus.eu). The ‘Hybrid gridded demographic data for the world, 1950–2020’(ref. ¹⁸) dataset (0.25° resolution) is available via Zenodo at https://zenodo.org/records/6011021 (ref. ¹⁸).

Code availability

The Python and bash scripts (which make use of Climate Data Operators (CDO)) used in the data analysis are available from the corresponding author upon reasonable request.

References

1. Heat and Health (World Health Organization, 2024); https://www.who.int/news-room/fact-sheets/detail/climate-change-heat-and-health
2. Domeisen, D. I. V. et al. Prediction and projection of heatwaves. Nat. Rev. Earth Environ. https://doi.org/10.1038/s43017-022-00371-z (2023).
3. Martinez-Villalobos, C. et al. Accelerating increase in the duration of heatwaves under global warming. Nat. Geosci. https://doi.org/10.1038/s41561-025-01737-w (2025).
4. Perkins, S. E., Alexander, L. V. & Nairn, J. R. Increasing frequency, intensity and duration of observed global heatwaves and warm spells. Geophys. Res. Lett. https://doi.org/10.1029/2012GL053361 (2012).
5. Callahan, C. W. et al. Increasing risk of mass human heat mortality if historical weather patterns recur. Nat. Clim. Change 16, 26–32 (2026).Article  Google Scholar 
6. Quilcaille, Y. et al. Systematic attribution of heatwaves to the emissions of carbon majors. Nature 645, 392–398 (2025).Article  CAS  Google Scholar 
7. Cruz, M. et al. Where heat does not come in waves: a framework for understanding and managing chronic heat. Environ. Res. Clim. 4, 023002 (2025).Article  Google Scholar 
8. McGregor, G. R. & Vanos, J. K. Heat: a primer for public health researchers. Public Health 161, 138–146 (2018).Article  Google Scholar 
9. Zhu, J., Wang, S. & Fischer, E. M. Increased occurrence of day–night hot extremes in a warming climate. Clim. Dyn. 59, 1297–1307 (2022).Article  Google Scholar 
10. Jendritzky, G., de Dear, R. & Havenith, G. UTCI—why another thermal index? Int. J. Biometeorol. 56, 421–428 (2012).Article  Google Scholar 
11. Di Napoli, C., Pappenberger, F. & Cloke, H. L. Verification of heat stress thresholds for a health-based heatwave definition. J. Appl. Meteorol. Climatol. 58, 1177–1194 (2019).Article  Google Scholar 
12. Di Napoli, C. et al. Thermal comfort indices derived from ERA5 reanalysis. Copernicus Climate Change Service (C3S) Climate Data Store https://doi.org/10.24381/cds.553b7518 (2020).
13. Di Napoli, C. et al. ERA5-HEAT: a global gridded historical dataset of human thermal comfort indices from climate reanalysis. Geosci. Data J. 8, 2–10 (2020).Article  Google Scholar 
14. Hamed, M. M., Alasow, A. A. & Shahid, S. Global trends in human thermal stress: a spatiotemporal analysis from 1940 to 2020. Earth Syst. Environ. 9, 2529–2541 (2024).Article  Google Scholar 
15. Mǎrmureanu, L. et al. Changes in the thermal stress across Europe between 1940–2023. Int. J. Biometeorol. 69, 3569–3586 (2025).Article  Google Scholar 
16. Roye, D. et al. Short-term association between hot nights and mortality: a multicountry analysis in 178 locations considering hourly ambient temperature. Environ. Int. 203, 109719 (2025).Article  Google Scholar 
17. Chen, Y. & Zhai, P. Revisiting summertime hot extremes in China during 1961–2015: overlooked compound extremes and significant changes. Geophys. Res. Lett. 44, 10 (2017).Article  Google Scholar 
18. Chambers, J. Hybrid gridded demographic data for the world, 1950–2020. Zenodo https://doi.org/10.5281/zenodo.3768003 (2020).
19. Naskar, P. R. et al. Spatiotemporal variations of UTCI based discomfort over India. J. Earth Syst. Sci. 133, 47 (2024).Article  Google Scholar 
20. Park, J. et al. Aerosol-induced surface cooling elevates relative humidity in the Indo-Gangetic Plain. Commun. Earth Environ. https://doi.org/10.1038/s43247-025-02513-9 (2025).
21. Liu, J. et al. CO₂ radiative forcing induces summer cooling over India. Nat. Commun. 17, 2724 (2026).Article  CAS  Google Scholar 
22. Sarin, T. S. & Vinoj, V. Assessing the relationship between atmospheric aerosols and maximum surface air temperature over the Indian region. Sci. Rep. https://doi.org/10.1038/s41598-026-40641-0 (2026).
23. Copernicus Climate Change Service. Global climate highlights 2024. ECMWF https://climate.copernicus.eu/global-climate-highlights-2024 (2025).
24. McPhaden, M. J. Genesis and evolution of the 1997–98 El Niño. Science 283, 950–954 (1999).Article  CAS  Google Scholar 
25. Gu, D., Andreev, K. & Dupre, M. E. Major trends in population growth around the world. China CDC Wkly https://doi.org/10.46234/ccdcw2021.160 (2021).
26. Urban, A. et al. Evaluation of the ERA5 reanalysis-based Universal Thermal Climate Index on mortality data in Europe. Environ. Res. 198, 111227 (2021).Article  CAS  Google Scholar 
27. Xu, X. et al. Increasing meridional disparity of population exposure to heat stress. Geogr. Sustain. 7, 100391 (2025). Google Scholar 
28. United Nations Children’s Fund. The State of the World’s Children 2024: The Future of Childhood in a Changing World (UN, 2024); https://doi.org/10.18356/9789211069631
29. Vecellio, D. J. et al. Greatly enhanced risk to humans as a consequence of empirically determined lower moist heat stress tolerance. Proc. Natl Acad. Sci. USA 120, e2305427120 (2023).Article  CAS  Google Scholar 
30. Liu, J. et al. Nonlinear exposure–response associations of daytime, nighttime, and day–night compound heatwaves with mortality amid climate change. Nat. Commun. 16, 624 (2025). Google Scholar 
31. Yang, J. et al. Drivers of associations between daytime–nighttime compound temperature extremes and mortality in China. Nat. Commun. Med. 4, 125 (2024). Google Scholar 
32. Perkins-Kirkpatrick, S. E. & Gibson, P. B. Changes in regional heatwave characteristics as a function of increasing global temperature. Sci. Rep. 7, 12256 (2017).Article  CAS  Google Scholar 
33. Rogers, C. D. W. et al. Sixfold increase in historical northern hemisphere concurrent large heatwaves driven by warming and changing atmospheric circulation. J. Clim. 35, 1063–1078 (2022).Article  Google Scholar 
34. Seneviratne, S. I. et al. in Climate Change 2021—The Physical Science Basis (eds Masson-Delmotte, V. P. et al.) 1513–1766 (IPCC, Cambridge Univ. Press, 2021); https://doi.org/10.1017/9781009157896.013
35. Romanello, M. et al. The 2023 report of the Lancet Countdown on health and climate change: the imperative for a health-centred response in a world facing irreversible harms. Lancet 402, 2346–2394 (2023).Article  Google Scholar 
36. European State of the Climate 2023 (Copernicus Climate Change Service & World Meteorological Organization, 2024); https://doi.org/10.24381/bs9v-8c66
37. Hersbach, H. et al. The ERA5 global reanalysis. Q. J. R. Meteorol. Soc. 146, 1999–2049 (2020).Article  Google Scholar 
38. Soci, C. et al. The ERA5 global reanalysis from 1940 to 2022. Q. J. R. Meteorol. Soc. https://doi.org/10.1002/qj.4803 (2024).
39. Lopes, F. M., Dutra, E. & Boussetta, S. Evaluation of daily temperature extremes in the ECMWF operational weather forecasts and ERA5 reanalysis. Atmosphere 15, 1 (2024).Article  CAS  Google Scholar 
40. Gong, P. et al. Action on hot nights: a scoping review of nighttime heat thresholds and their critical role in urban climate resilience. Sustain. Cities Soc. https://doi.org/10.1016/j.scs.2026.107303 (2026).