The Compound Hazard Century

Across the places where 6.5 billion people live (those with good station coverage), 11% of warm-season months were extremely hot in 1951–80 by that period's own standard, which is what the definition implies. By 1995–2024, 30% were. In 2024 the share reached 59%, a record, and every year from 2022 to 2024 was above 45%. Drought months, by contrast, stayed at about one in ten: 10.0% in the baseline, 10.3% in 1995–2024 and 10.1% in 2021–24, with large swings, peaking in the El Niño droughts of 1983 and 1992.

Where the two meet, the change is dramatic. In 1951–80, 3.1% of warm-season months were both extremely hot and in drought; in 1995–2024, 5.3%; in 2021–24, 6.9%. For a typical place, that means about one hot drought month per decade in the baseline, two in 2021–24. Because drought frequency is flat, almost all of the rise comes from droughts becoming hotter: in 1951–80, 31% of drought months were extremely hot; by 1995–2024, 52%; in 2021–24, 69%. In a log decomposition, 95% of the increase in hot droughts comes from the rising heat of droughts and 5% from their frequency.

The most extreme single years still belong to El Niño. In 1983, 11.6% of warm-season months were hot and dry, as drought struck Australia, Indonesia, southern Africa and the Sahel at once; 2022 came second at 10.2%, with record heat and drought in Europe, China's Yangtze basin and the Horn of Africa. The difference now is that heat no longer needs a rare drought year to arrive. Any drought is likely to be a hot one, which is what makes the overlap a hazard in its own right: hot droughts dry soils and vegetation faster, stress crops at flowering, raise water demand just as supply falls, and prime landscapes for fire.

In most climates, the warm season's temperature and rainfall move in opposite directions: a dry month is a sunny, hot month. The mechanism runs both ways. When soils are moist, much of the sun's energy goes into evaporating water, which cools the surface. When they dry out, that energy heats the air instead, so drought amplifies heat. And heat, by raising the atmosphere's demand for water, dries soils and vegetation faster, so heat deepens drought. This land–atmosphere feedback is why the two hazards were already 2.8 times more likely to coincide than chance before warming took off.

Measured as the correlation between each place's warm-season temperature and three-month rainfall (both detrended), the coupling is negative in 138 of 143 countries, home to 7.4 billion people, with a population-weighted value of −0.36. It is strongest in monsoon Asia and southern Africa, Thailand (−0.61), Bangladesh (−0.58), Lesotho, Laos and the Philippines (−0.55), and weakest in the hyper-arid Gulf, where there is almost no warm-season rain to couple with. The literature finds the same: concurrent hot and dry summers become more frequent in a warmer climate partly because the two become more interdependent (Zscheischler and Seneviratne, 2017).

What warming has added is a shift in the whole temperature distribution. In the baseline, a drought month was 3.5 times as likely to be extremely hot as any other month (31% against 9%). Today the gap is smaller (52% against 27%), not because the coupling weakened but because extreme heat now also happens outside drought. The odds of a hot drought therefore depend on both: the coupling sets how much a drought tilts a month towards heat, and warming sets how far the dial has already moved. The simulator in section 9 lets readers vary each.

The map shows the merger place by place. Hot droughts are now common across the Mediterranean, from Spain and Algeria through Italy and the Balkans, and in Iran, in the Gulf of Guinea from Côte d'Ivoire to Nigeria, across much of Brazil and Venezuela, and in pockets of East Africa (Rwanda, Burundi, Tanzania), Mozambique and central Europe. In these places, 10–20% of warm-season months in 1995–2024 were both extremely hot and in drought, against about 3% in the baseline.

The "droughts that are hot" view shows how general the change is. In most of Europe, Africa, Latin America and Southeast Asia, more than half of recent drought months were also extremely hot; in Italy, 84%; in Mexico, 78%; in Spain, 78%. The coupling view shows that the underlying link between dry and hot is almost everywhere, strongest in monsoon Asia, where the warm season is the pre-monsoon dry season and rain brings cooling.

South Asia is the clear exception to the trend. Despite strong coupling, India's and Bangladesh's hot-dry share fell slightly (India 4.3% to 3.7%), and extreme warm-season heat rose far less than elsewhere (India from 11% to 13% of months). Studies link part of the muted warming in the Indo-Gangetic plain to irrigation and aerosol pollution, and CRU station coverage over India thins after 2000, so this result is less certain. The whiplash view shows rising dry-to-wet swings in the southern Sahel and central Africa, the Andes, the Caribbean, East Africa, Myanmar, Pakistan and northern India.

By region, hot droughts rose most in sub-Saharan Africa (2.8% to 8.2% of warm-season months, in the parts where drought can be measured), Europe and Central Asia (3.3% to 7.7%), the Middle East and North Africa (2.1% to 7.2%) and Latin America (3.1% to 6.2%). East Asia (2.8% to 4.1%) and North America (2.6% to 3.5%) rose less, and South Asia fell (4.3% to 3.7%). In 54 countries, home to 1.4 billion people, the hot-dry share at least doubled.

Spain is the most extreme large case: 18.5% of warm-season months were hot and dry in 1995–2024, against 2.5% in 1951–80, because both heat (59% of months) and drought (24%) rose. Iran (1.7% to 15.2%), Algeria (2.1% to 14.9%) and Nigeria (2.2% to 11.8%) follow, with Ghana, Benin and Togo above 16%. Brazil's share more than tripled to 9.1%, and Germany's and Italy's reached 10–12%.

By income group, the jump in heat itself was largest in low-income countries: 44% of warm-season months were extremely hot in 1995–2024 against 23% in lower-middle-income and 37% in high-income countries, reflecting the small natural variability of tropical climates, where a modest warming pushes many months past the old extremes. In low-income countries, 52% of drought months are now hot, up from 28%. These are the economies most dependent on rain-fed farming and least able to buffer a hot-drought harvest.

Flooding interacts with drought through the land. Soils baked hard by drought, or stripped of vegetation by fire, absorb less rain, so the first heavy rains after a drought run off faster, carrying soil and ash into rivers. Reservoirs drawn down in a drought can be overwhelmed when the rain returns, and water managers who must hold water against the next drought have less room to absorb a flood. Swain and colleagues (2025) call this hydroclimate volatility and find that three-month whiplash has increased by 31–66% since the mid-20th century, mainly because a warmer atmosphere both holds more water and pulls more from the land.

Monthly station data show the same direction. Very wet months (top 10% of the 1951–80 climate, at least 50 mm) made up 10.0% of months in the baseline and 12.1% in 1995–2024, a 21% rise. Whiplash, a very wet month within three months after a drought month, rose from 1.8% to 2.3% of months, or 30%, at the low end of Swain's range. The rise mostly reflects wetter extremes overall: the chance of a very wet month after a drought (10.8%) is close to the chance of one at any time (12.1%), so whiplash is rising because both ends of the distribution are stretching, not because droughts now trigger floods more directly.

Whiplash rose in 85 of 140 countries. Among countries of 10 million people or more, the largest increases are in Chad and Cameroon (1.3% to 4.9% of months), Peru and Haiti (to 4.3%) and Myanmar (1.5% to 3.9%), followed by Cuba, North Korea, Uganda and Bolivia; among the most populous, Pakistan rose from 1.9% to 3.2% and India from 1.9% to 3.1%. The pattern is uneven by income: very wet months became more common in high-income countries (10.0% to 12.9%) and less common in low-income ones (10.0% to 9.2%), where drying dominates. Monthly totals miss the flash floods that come from hours of rain, so these figures understate the short-duration side of the problem.

Pollution interacts with heat and drought in three ways. Drought and heat dry out vegetation and soil, so fires and dust storms put more particles into the air. Heat and sunshine speed the chemistry that turns traffic and industrial emissions into ozone, and the stagnant high-pressure weather of heatwaves traps pollution near the ground. And heat and air pollution strain the same organs, the heart, lungs and kidneys, so each makes the other more deadly. Across 620 cities, the mortality effect of the hottest days was 5.3% where PM10 was 10 µg/m³ and 12.8% where it was 90; for ozone, 2.9% at 40 µg/m³ and 12.5% at 160 (Stafoggia et al., 2023).

To test the first channel at scale, this report links satellite-based annual PM2.5 for 136 countries (1998–2024) to each country's warm-season hazards, comparing each country with itself in other years and removing global year effects. The signal is modest on annual averages: a warm season entirely in drought raises PM2.5 by about 5% (t = 3.4); heat alone does not raise it, and there is no measurable extra effect from heat and drought together. Annual national averages dilute what are short, local episodes, so the effect shows up most clearly in the extremes.

The largest pollution surprises in the panel are hot-dry fire and haze years: Australia in 2019 (+50% above expectation), Guatemala in 1998 (+40%), Malaysia in 2019 (+35%, smoke from Indonesian fires), Portugal in 2005 (+34%), Switzerland and Hungary in the 2003 European heatwave (+27%, +25%) and Russia in 2010 (+27%). Globally, landscape fires added 2.5 µg/m³ to average PM2.5 exposure in 2010–19, about four times more in low-income than high-income countries (Xu et al., 2023). At daily scale, this platform's city study found ozone higher on hot days in 131 of 149 cities, and 85% of hot days polluted against 75% of other days.

Climate models cannot yet simulate monthly compound events place by place with confidence, but their climatologies show where the conditions for each pair of hazards will overlap. Three pairs matter: humid heat with persistent drying, humid heat with heavier downpours, and drying with heavier downpours, the climate of whiplash. Today, 1.9 billion people live with frequent humid heat. Under SSP2-4.5, 0.23 billion live with both heat and drying by the 2050s and 0.72 billion by the 2090s; 0.71 billion with heat and heavier downpours; 1.35 billion with at least two of the three.

The pathway matters enormously after mid-century. Under SSP1-2.6, the number facing two or more compound hazards peaks at 0.43 billion in the 2070s and falls to 0.34 billion by the 2090s. Under SSP3-7.0, with high emissions and fast population growth, it reaches 5.2 billion: 1.9 billion with heat and drying and 3.5 billion with heat and heavier downpours, as a hotter atmosphere loads tropical storms with more water. Holding population at 2020 levels, the SSP3-7.0 figure is 2.6 billion, so roughly half of the high-end exposure comes from climate and half from where population grows.

The countries most exposed by the 2090s under SSP2-4.5 are in the Middle East and the Sahel: Iraq (90% of people), Saudi Arabia, Egypt, Niger, Mali and Sudan, followed by Mozambique, Venezuela, Syria and Pakistan. Under SSP3-7.0, nearly everyone in Niger, Egypt, Mali, Sudan, Iraq and Burkina Faso, and more than three quarters of Thailand, Pakistan, Mozambique and Nigeria, would live with two or more. These projections use humid heat, so they understate the Mediterranean's dry-heat risk; and climatologies cannot capture month-to-month whiplash, which Swain and colleagues project to more than double (+113%) at 3 °C of warming.

The atlas summarises the grid by country, weighting each cell by its population. The first two layers show the share of warm-season months that were both extremely hot and in drought in 1995–2024 and the change since 1951–80. The third shows the share of drought months that were also extremely hot, the clearest single measure of the merger.

The coupling layer shows how strongly warm-season heat and rainfall move together in each country; the more negative, the more a dry month tilts towards heat. The whiplash layer shows the change in dry-to-wet swings, and the outlook layer the share of the population projected to live with two or more compound hazards in the 2090s under SSP2-4.5.

Countries are blank where station coverage is too thin (for example, Somalia) or where the warm season is normally rainless, so a three-month drought cannot be measured against it (most of the Sahara, the Gulf and the central Sahel). Hover over or tap a country for its card; clicking loads it into the simulator. The sortable table lists all 143 countries.

The simulator treats each country's warm-season temperature and three-month rainfall as a pair of linked bell curves, using the country's own measured coupling. Warming shifts the temperature curve to the right; a drier or wetter climate changes how often rainfall falls below the drought threshold. The share of months that are both hot and dry then follows from the overlap of the two tails.

It is calibrated so that, at each country's observed warming since 1951–80, it reproduces the observed share of extremely hot months; the warming slider then scales that shift. Across 114 countries, its predicted hot-dry shares for 1995–2024 match the observed ones closely (r = 0.98, mean 6.9% against 6.9%). The comparison with "independent" shows how much the coupling adds.

The chart traces the hot-dry share as warming rises from 0 to 4 °C at the chosen drought frequency. Outputs are given as months per decade (each decade has 30 warm-season months) and as the chance that a drought month is extremely hot. The curves flatten at the drought frequency: once almost every warm-season month is extremely hot by the old standard, every drought is a hot drought, and drought frequency becomes the only lever left. It is a stylised model: real distributions are not perfectly normal, and coupling may itself strengthen with warming.

Warn for combinations, not single hazards. Heat-health warnings, drought monitors, fire-danger ratings and air-quality alerts are usually run by different agencies on different scales. A hot drought is simultaneously a heat, crop, water, fire and smoke emergency. Joint triggers, such as heat plus soil-moisture deficit for crops and fire, or heat plus pollution forecasts for health, give earlier and better-targeted warnings, and the data to build them already exist in most national weather services.

Keep water in the landscape. Soil moisture is the switch that turns a dry spell into a heatwave and a downpour into a flood. Measures that hold water in soils and storage, such as conservation farming, agroforestry, restored wetlands and floodplains, managed aquifer recharge and well-run reservoirs, damp both ends of the whiplash. Water managers increasingly co-manage drought and flood, holding storage flexibly instead of reserving it for one or the other, as Swain and colleagues recommend.

Cut the fuel for pollution events, and cut emissions. Managing fuel loads and fire use, enforcing bans on agricultural and peatland burning in dry seasons, and controlling dust at its sources reduce the smoke and dust that hot droughts release. Cutting combustion emissions lowers the baseline that heat makes deadlier, with immediate health returns. And cutting greenhouse gases is what separates a 2090s world of 0.34 billion people living with compound hazards from one of 5.2 billion.

Observations: CRU TS 4.09 monthly mean daily maximum temperature (tmx) and precipitation, 0.5°, 1951–2024 (Harris et al., 2020). Cells count if both variables are station-informed in at least 75% of months; these hold 6.5 of 7.6 billion people. Per cell and calendar month, against 1951–1980: hot = tmx ≥ 90th percentile; drought = 3-month rainfall ≤ 10th percentile, with the 10th percentile above zero and a 3-month mean of at least 30 mm; very wet = monthly rainfall ≥ 90th percentile and ≥ 50 mm; whiplash = very wet with a drought month in the previous 1–3 months. Hot-dry is measured in each cell's three warmest calendar months. Weights: SSP2 2020 population (CCKP), or land area.

Country results are population-weighted over covered cells; hot-dry figures are shown only where at least half of the warm season is measurable (114 countries). Coupling: correlation of detrended warm-season tmx and 3-month rainfall anomalies. Pollution: SatPM PM2.5 (van Donkelaar et al., via Our World in Data), 136 countries, 1998–2024; log PM2.5 on hot, drought and hot-dry warm-season shares with country and year fixed effects, standard errors clustered by country. Projections: CCKP CMIP6 multi-model medians, 0.25°, for heat index above 35 °C on ≥ 30 days, 12-month SPEI ≤ −0.5 and maximum 5-day rainfall ≥ 1.2 × 1995–2014, with SSP population for 2030–2090. Simulator: bivariate normal with each country's coupling, calibrated to its observed hot-month share.

Caveats: monthly data miss heatwaves and floods lasting days or hours; percentile thresholds measure change against each place's own past, not absolute danger; CRU station coverage is thin in parts of Africa, the Middle East and India after 2000; arid warm seasons cannot be assessed for drought; annual PM2.5 dilutes short episodes and includes dust; the panel shows association, not causation. Projected pairs are climatology overlaps, not co-occurring events, and humid-heat thresholds understate dry heat. Exposure is not impact.