The new science of extreme weather

How extreme event attribution transformed climate science’s take on weather

We have known for decades that climate change makes them more likely, but extreme weather events are nothing new. Now, thanks to recent developments in climate modelling, meteorologists can quantify how much more likely increased greenhouse gases have made a particular extreme weather event. It remains a rapidly developing field. One of its pioneers, Peter Stott, outlines where the new science of extreme weather is today.


Earth’s climate is changing. As global temperatures rise, the frequency and intensity of extreme weather events around the world also rise. The science of event attribution is being developed to assess the extent to which recent extreme weather events are linked to human-induced emissions of greenhouse gases. Already this science has shown its potential by demonstrating that many events have been affected significantly by climate change. Further research will make it possible to pin down in more detail how a wider range of weather extremes have been affected by human actions.

Global temperatures are now over 1.1 degrees Celsius warmer than they were in pre-industrial times. This rise is mainly due to human-induced emissions of greenhouse gases. With rising temperatures have come more intense and frequent heatwaves, floods and droughts around the world. This year the Canadian temperature record was smashed: the 49.6C recorded in June was followed by devastating forest fires that destroyed the town of Lytton in British Columbia. Extremely heavy rain in central Europe in July led to flash floods that destroyed many thousands of homes in Belgium and Germany. A severe drought in Madagascar has left more than one million people short of enough to eat. There can be no doubt that extreme weather brings with it disastrous consequences.

Nevertheless it does not follow that all extreme weather events can be blamed on human-induced climate change. Climate has always varied naturally. In the past, long before the industrial era, weather was occasionally sufficiently unusual or extreme to cause substantial harm to large numbers of people.  It follows therefore that some extreme weather events, even today, could be mainly associated with natural variations in climate. The natural El Nino phenomenon, for example, periodically brings warmer waters to the surface of the Eastern Pacific Ocean and this can lead to devastating flooding in Southern California and Northern Mexico.

Despite these natural variations in climate, global average temperatures have now risen well outside the range of variations that can be explained by natural factors. This is increasing the risk of damaging weather around the world. As temperatures rise globally, the changes of record-breaking temperatures being observed locally increases. A warmer atmosphere can contain more moisture – about 7% for each degree Celsius of warming – which can lead to heavier rainfall in storms. With warmer ocean temperatures, storms are also potentially more energetic with stronger wind speeds. And as sea levels rise, due to expansion of warming sea water and melting of ice on land, there is an increasing risk of storm surges leading to over-topping of sea defences.

Increasingly, National Meteorological Services, like the Met Office in the UK, are being asked to provide updates on the meteorology of extreme weather as it occurs. In the face of increasing threats to their lives and livelihoods from climate change, people are no longer satisfied with knowing whether a recent extreme measurement was a record-breaker or particularly unusual. They want to know about the role of human-induced emissions of greenhouse gases.

In the UK, thanks to an extensive network of measuring sites, the National Climate Information Centre at the Met Office provides comprehensive assessments of how temperature, precipitation and sunshine vary across the country and how recent extreme events of heat, rainfall or drought compare to past decades and centuries. Nevertheless, most of the media enquiries we face in the immediate aftermath of such events, which often cause major disruption and result in substantial economic losses, also ask us to comment on the role of climate change. To what extent was climate change responsible? we are asked. Is this a sign of things to come? To answer such questions, we can’t just rely on past records. We also need climate models. And we need the science of event attribution.


During the 2003 heatwave it was possible, I realised, to apply this concept for the first time thanks to recent developments in climate modelling. This meant that models were now able to simulate variations in regional temperatures with sufficient accuracy.

The science of event attribution has developed considerably since I published the first study to link an extreme weather event to climate change, the deadly European heatwave of 2003 that killed over 70,000 people. In a paper published in Nature in 2004, co-authored by Myles Allen and Daithi Stone of Oxford University, we applied the concept of probabilistic event attribution. This concept, first proposed by Myles Allen, was to attribute an individual weather event by calculating how much human-induced climate change had altered the likelihood of its occurrence. During the 2003 heatwave it was possible, I realised, to apply this concept for the first time thanks to recent developments in climate modelling. This meant that models were now able to simulate variations in regional temperatures with sufficient accuracy.

In our calculation, we compared the statistics of European extreme temperatures in climate models that included increasing greenhouse gas concentrations with models that only included natural factors. This allowed us to estimate how much human-induced emissions of greenhouse gases had changed the probability of the record-breaking temperatures seen that summer in Europe. Our results showed that human-induced climate change had very likely more than doubled the probability of such a devastating heatwave. It was a landmark study that prompted the development of more event attribution studies over the succeeding years.

Initially, much of the focus of such studies was on the impact of emissions of greenhouse gases on extreme heat. Extreme heat, as pointed out by a 2016 assessment of the science of event attribution made by a committee of the US National Academies of Sciences, Engineering and Medicine, is the easiest type of extreme weather to attribute. Temperatures vary over large areas, are well monitored and are well simulated by climate models, as for example was the case with the 2003 European heatwave. Thus many of the event attribution studies that have been made of events in recent years have been of heatwaves.

A striking result of these more recent studies is that some temperature extremes have become so much more likely due to climate change that they would have been effectively impossible in a non-industrial climate. This was the case for the extreme temperatures seen in the Arctic during the first 6 months of 2020 including a record-breaking temperature of 38C seen at Verkhoyansk in June, 2020. It was also the case for the extreme temperatures seen in June 2021 in Western North America that resulted in Canada smashing its all-time temperature record by almost 5 degrees to record 49.6C in British Columbia. While it is true that there have always been extremes of weather, the latest analyses show that recent heatwaves would not have been seen if humans had not altered climate, even over thousands or tens of thousands of years.

While most event attribution studies have been carried out on extreme heat, increasingly the focus is switching to other damaging weather events. Many studies have now been carried out on heavy rainfall events that have led to flooding. For example, a study carried out by the World Weather Attribution team into the devastating floods seen in Central Europe in June 2021 found that human-induced climate change increased the intensity of rainfall falling over a single day in this region by between 3 and 19%. This is an example of an attribution study that investigated the effect of human activities on the intensity of an event rather than its likelihood. Both approaches – attributing changes in likelihood and attributing changes in intensity – have been widely applied and both provide relevant, albeit complementary information.

Some types of weather event remain difficult to attribute. Wind storms are challenging to attribute as observations of very localised storms can be scarce or unreliable and models struggle to capture the very intense winds at the centre of extreme storm systems. Long-lasting droughts are also challenging as the processes involved, including the way the soils dry out and influence weather systems, are only approximately simulated by climate models. There is rapid progress being made however on improving models. Now that climate simulations are being made at weather forecasting resolution – kilometre scale grid-spacing – they are much better able to simulate the intense localised rainfall that leads to much of the flooding in the mid-latitudes. A recent study of Autumn rainfall made at the Met Office showed that the frequency of extreme daily precipitation in the UK during autumn has increased by about 60% due to human-induced climate change.

Crucially, the analysis used the same methods as in past peer-reviewed studies. This meant that it was not necessary to wait for a new peer-reviewed publication to appear before publicising the results.

Traditional event attribution studies have been carried out as research projects, requiring a lengthy period of time for the analysis to be done, written up, submitted to a scientific journal, reviewed by experts and then revised before it is ready for publication and the results released to the public. Increasingly however, people including the media, government officials and residents directly affected, do not want to wait many months or years to know how an extreme weather event has been affected by climate change. Nor should they have to, necessarily. We don’t wait for scientists to carry out a piece of research every time a weather forecast is made. Instead, weather forecasts are produced using methods based on previously published research. And as new research comes through, national meteorological services take advantage of it to improve their forecasts, thereby continually improving the service they can give to the users of their information.

The same principles apply to the rapidly developing field of operational attribution. This aims to exploit previously published methods to rapidly and reliably assess the extent to which human-induced climate change is affecting the risks of extreme weather. We know we won’t be able to do everything right away because the science isn’t yet sufficiently developed to rapidly assess the risks of some types of extreme, such as wind storms or droughts. But for some types of event, notably intense heatwaves, it is possible to do such rapid assessments based on peer-reviewed methodologies. Then we can inform people quickly about the altered odds of a recent heatwave and provide an outlook on the frequency and intensity of such heatwaves in future.

An example of such an operational attribution analysis was that carried out by the Met Office into the extreme temperatures seen in Europe during Summer 2021. This was the hottest summer on record with the average temperature over the continent close to 1 degree Celsius above the 1991-2020 average. An ensemble of climate model simulations of the climate of today was compared with a parallel ensemble of climate model simulations of the climate as it would have been without human influence. Crucially, the analysis used the same methods as in past peer-reviewed studies. This meant that it was not necessary to wait for a new peer-reviewed publication to appear before publicising the results.

The analysis showed that the record-breaking season in Europe would have been almost impossible without human influence on climate. The return time in the non-industrial world without greenhouse gas emissions would have been thousands of years. In the present climate, the extreme heat has an estimated return time of about three years. With continued emissions, such conditions could be seen every year by the end of this century.

Climate is changing and with it bringing a change in extreme weather and a rapid escalation in damaging impacts. The science of event attribution is now able to put many of these events, as they occur, into the context of past and future climate change. Of course, the future is not yet written. It depends heavily on whether the agreements made at climate negotiations including at Glasgow in November 2021 are successful in bringing greenhouse gas emissions down over the next few decades.

In this context of uncertainty, event attribution plays an important role in reducing the impacts of climate change. First, it helps to motivate climate action to reduce emissions by demonstrating the effects our past emissions are already having on making extreme weather more damaging. Second it helps people better adapt to what is coming their way through improving the understanding of the risks they face in future from extreme weather. As far as event attribution is concerned, understanding the present is key to a better future.

Latest Releases
Join the conversation