← Back to the dashboard

How we spot heatwaves

What the dashboard shows, where the numbers come from, how current they are, and what they can’t tell you.

The short answer

A heatwave is heat that is unusual for the place and the time of year. 35 °C is an ordinary day in Riyadh in summer and an emergency in London. So the dashboard compares each of 323 cities, on every continent, with its own climate: the daily highs and night lows it had at the same time of year in 1991–2020. When a city is hotter than 9 in 10 of those days for at least three days in a row, we call it a heatwave. When the lowest temperature of the day is higher than on 9 in 10 of those days, we call it an unusually warm night. We count how many of both each city has had this year, against an average year and against every year since 1991.

How we know

Daily highs and lows. NOAA’s Climate Prediction Center collects daily reports from weather stations worldwide and turns them into maps of the daily maximum and the daily minimum temperature on a grid of half-degree squares, about 50 km wide. It adds each day within a day or two. We read the square of each city every night from NOAA’s public files.

The cities. The list follows a fixed rule, so that no one picks favourites: the capital and the largest city of every country with at least a million people, and every city of a million or more until a country has eight on the list. Cities within 60 km of a larger one on the list are left out, because they share its weather and often its grid square. The 48 cities the page started with in October 2026 all remain. Names, coordinates and populations come from GeoNames, a free geographical database.

The normal. From the same NOAA analysis for 1991–2020 we worked out, for each city and each day of the year, the average daily high and the 90th percentile: the temperature that only 1 in 10 days exceeded. We did the same for the daily lows. We use the 15 days around each date, so the normal changes smoothly through the year.

The heatwave rule. At least three days in a row above the 90th percentile. It is one of the standard definitions used by climate scientists (Perkins and Alexander 2013). Because it is relative to the time of year, an unusually warm spell in winter counts too: it is a heatwave in the sense of the record books, even if it is not dangerous.

Warm nights. A night counts as unusually warm when the day’s lowest temperature, which usually falls just before dawn, is above the 90th percentile of lows for that date. For the lows we use 1999–2020 as the normal, not 1991–2020. In NOAA’s analysis the lows before 1999 run colder than afterwards in parts of West Africa, the Middle East and South and South-East Asia, while the daily highs show no such step: Colombo had 0 to 4 warm nights a year from 1991 to 1997 and 84 in 1998. That is a change in the station reports, not in the climate. Nights matter because the body recovers from the day’s heat while it sleeps; a study of 24.7 million deaths in Japan found deaths 9 to 10% higher on hot nights than on other nights (Kim and others 2023).

This year against earlier years. We counted heatwave days in every year from 1991 to 2025, and warm nights from 1999, from NOAA’s yearly archive files, always with the same thresholds, and we count the current year from the nightly reads. The years 1991–2020 average out at “normal” by construction, because the thresholds were set from them.

The bigger picture. The world’s yearly average temperature comes from NOAA’s National Centers for Environmental Information. The statements on heatwaves and climate change are from the Intergovernmental Panel on Climate Change (IPCC) assessment of 2021, and the death tolls from peer-reviewed studies.

How current each part is

What the dashboard can’t tell you

It is not a heat warning. For warnings and advice in the coming days, follow your national weather service and health authorities.

A grid square is an average over about 50 by 50 km, not a thermometer in the city centre. City centres are often a few degrees hotter, and at the coast or near mountains the square can differ from the city’s weather station. For Los Angeles and Athens the nearest square is mostly mountains or sea, so we use the neighbouring square that matches the city’s station. For 34 coastal and island cities, among them Lima, Hong Kong, Cape Town and Barcelona, NOAA’s grid has no land square at the city itself; we use the nearest one, 30 to 80 km away, and the city panel says so. There the temperatures can differ from the city’s own, though unusual heat usually shows in both. About six in ten of the squares (202 of 323) contain no reporting station on most days; NOAA fills them from the nearest stations, so their values are smoother and less certain. The city panel marks these too.

The comparison between years is less exact than the daily one. NOAA’s analysis uses whichever stations reported on each day, and that set has changed since 1991. We checked the result two ways: for ten of the cities an independent weather reanalysis (ERA5) shows the same excess of heatwave days in 2026, and a larger one, and for a sample of the new cities its October 2026 temperatures are within about 2 °C of NOAA’s squares (up to 4 °C in mountain cities such as Quito).

Temperature is only part of the danger. Humidity matters as much for health, and it is not included here. The daily low is the lowest reading of the day whenever it occurred; on a few days a cold front in the evening sets it, not the night before.

The list holds large cities and capitals. Smaller places are not covered, and a country’s count of cities says nothing about how much heat it had.

The normal is 1991–2020, which was already warmer than the 20th century. A year with “normal” heatwave days still has more than the climate our grandparents knew.

Further reading

Credits

Daily temperatures from NOAA’s Climate Prediction Center, global temperature from NOAA’s National Centers for Environmental Information, both public US government data. City names, coordinates and populations from GeoNames, licensed CC BY 4.0. Heat-death estimates and IPCC statements as cited. We only download and present their public data.

← Back to the dashboard