World Just Saw Its Hottest Day Ever, Here's How It Was Analysed

On Sunday, July 21, the world had its hottest day on record. Just 24 hours later, that record broke again making last Monday very likely the hottest day in thousands of years.

It may seem improbable for scientists to gauge the world's hottest day given that they don't have temperature monitors in every corner of the world and less than a century of relatively widespread observations. But they've developed a technique that's increasingly useful as the planet heats up.

This month's shocking heat findings, announced by the EU's Copernicus Climate Change Service, are based on "reanalysis," a technique that mixes temperature data and models to provide a global view of the climate. The center creates a nearly real-time picture of the Earth's climate, including temperature, wind and precipitation, for (roughly) every 30-square-kilometer chunk of the planet's surface.

This reanalysis goes back to 1940, and it allows researchers to say with confidence when a record is broken, whether for a day, month or year. Beyond the new daily heat record, the data also shows that 2023 was the hottest year ever recorded and that every calendar month for the past 13 months has been the hottest on record.

Though there aren't thermometers in every corner of the world, Copernicus receives a large amount of weather data that it uses to underpin its reanalysis. "We have this constant flow of information coming into the center," says Carlo Buontempo, director of the Climate Change Service, which is part of the European Centre for Medium-Range Weather Forecasts (ECMWF).

Scientists at the center receive 100 million readings per day about weather conditions from around the world. Observations come from airplanes, satellites, ships, radar and surface-level weather stations - all feeding real-time information about temperature, wind, rain and snow information, as well as other factors like air pollution. This information is fed into a model, known as ERA5, which is already equipped with historic information about the global climate.

There are gaps in these observations because the data sources don't cover every part of the world. Weather conditions like cloudy skies may also reduce the amount of data coming from sources like satellites. To fill these gaps, the scientists take the predictions they have already made, based on the long-term ERA5 model, and test them against the observations. That means a forecast that predicts a particular temperature in a particular place will be tested against all the data researchers receive about the weather in that place and nearby, as well as broader forces like ocean currents and air circulation.

This is done repeatedly while assessing how compatible the prediction is with what's actually been recorded. The model also accounts for any errors in the recorded data, and relies on the laws of physics, including the weather patterns, currents and airflow that govern how the global climate works.

In this way, it's possible to create a complete picture that is as accurate as possible. That's what allows scientists to confidently declare records like when the world experiences the hottest day in human history.

Globally, five weather services - the US's National Oceanic and Atmospheric Administration and NASA, the ECMWF, the China Meteorological Administration, and the Japan Meteorological Agency - carry out continuous appraisals of global temperature using this technique. While their models differ slightly, the five groups have come to similar conclusions about record heat in recent months and years.

Historical data is trickier to come by. The longest-running temperature series, the Central England Temperature in the UK, started in the 17th century. Data from before humans were systematically monitoring temperatures comes from sources like bubbles of gas trapped in glacial ice, or tree rings. These sources aren't as specific as a thermometer reading, but it's possible to say with confidence that recent temperatures are likely the highest in around 100,000 years, Copernicus says.

Meteorologists also have a good idea when a particularly significant day, like the hottest day on record, is on its way. This is partly because global mean temperatures usually peak between early July and early August. Last year's hottest day - which was the previous record for the hottest ever - occurred in early July amid a historic oceanic heat wave. An intensifying El Nino - a natural global climate phenomenon that usually means hotter temperatures globally - provided yet another clue that record heat was brewing.

Until this July, it looked for a while like the world wouldn't set a new daily record, says Buontempo. "The global mean temperature for the oceans started rising again," he says. "Some of the people who systematically monitor our predictions started to sound alarm bells." By the start of last week, they were paying extra attention to the reanalysis and getting ready to make an announcement.

This technique isn't just useful for making "hottest day ever" announcements: It's being used to train AI forecasting models, especially for "ensemble" weather forecasts, which represent multiple possible future scenarios. It's also used by solar energy companies to help homeowners work out how much energy their panels might generate, and by wind energy companies to plan where to put wind farms.

Copernicus is currently working on a new model, known as ERA6, which will be more precise - dividing the world into 14 kilometer squares - and incorporate many more historic data sources, including early satellite readings from the 1970s.

For Buontempo, more important than any one day is the recent extraordinary streak of record-breaking months, given that's a better indicator of how rapidly the world is warming. But pinpointing a specific day does make a changing climate feel much more immediate.

"I think we have to make it more tangible, more direct, more visible," he says. "It is important that people are informed."

(Except for the headline, this story has not been edited by NDTV staff and is published from a syndicated feed.)