The question of how extreme weather events might
be linked to climate change is a key one.
It’s particularly important because in many
regions, extreme weather like heatwaves, floods and droughts cause
more damage than other, more predictable consequences of climate
change, such as sea-level rise.
Scientists know that an increase in average
temperature as the climate changes will lead to an increase
in the number or magnitude of some extreme events, while
others will get less likely.
But the chaotic nature of weather means it’s
generally impossible to say, for any particular event, that it only
happened because of climate change.
During the 2003 heatwave in Europe, summer
temperatures rose several degrees above those in 2000, 2001, 2002
and 2004. Image by Reto Stöckli, Robert Simmon and David Herring,
NASA Earth Observatory, based on data from the MODIS land
Probabilistic event attribution
On this basis, some people have concluded that
it’s effectively impossible to attribute extreme weather events to
greenhouse gas emissions.
But this isn’t quite right, and there are ways
we can explore the links.
My colleagues and I work on the emerging science
Probabilistic Event Attribution (PEA), which
assess how much human-induced climate change
affecting local weather events such
Our work focuses on assessing if and how past
emissions have contributed to the probability of any particular
extreme event changing – either becoming more or less
We use observations and statistical analysis to
assess weather events on a case-by-case basis. Because we don’t get
a lot of observations of the most extreme events,
almost by definition, assessments are based on
large numbers of climate model experiments, called
Such studies compare how often a particular
extreme weather event occurs in model experiments representing the
“world as it is” (with human influence on climate) with how often
it occurs in experiments representing the “world that might have
been”, where the estimated impact of human influence on climate is
Take the example of flooding.
The figure below shows a series of model runs for
peak flow in English and Welsh rivers during Autumn 2000. Each
circle represents a model run. The blue circles are the world as it
is – with climate change. The green circles are the world that
might have been – without climate change.
Return times of run-off (an indication of flooding) for the
“world as it is” (blue) and the “world that might have been”
without anthropogenic GHG emissions (green) for Autumn 2000 in
England and Wales. Black line shows the threshold exceeded in
observations. Adapted from Figure 10.18 IPCC AR5 WG1
In the current climate, shown by the blue dots, the chance of
exceeding the critical runoff threshold of 0.42 millimetres per day
observed in 2000 is one-in-ten in any given year. In the “world
that might have been” it would have been more like
Because we do not know exactly how the world without
anthropogenic climate change would look like (we do not have any
observations) we use several plausible ways to simulate the world
that might have been. That means there is more than one green
curve, indicating the large uncertainty around this kind of
But on average, this plot suggests human greenhouse
gas emissions has made flooding twice as
But not all damaging extreme weather events are
being made more frequent by human influence on climate. The plot
above, which looks at another river during spring, examines
flooding triggered by rapid melting of snow. Here we see that this
kind of event has been made less frequent by past GHG emissions –
the red arrow points to the right.
Public perception of extreme weather and
At the moment these studies require months to complete and are
published well after public attention to a weather event has
This can lead to situations where
extreme weather events
are listed as examples of the impacts of
anthropogenic climate change before
attribution studies have shown that it is
indeed an event made more likely by human greenhouse gas
In part, this is because immediately after an
extreme weather event the only available statement is that all
events are now happening in a changing climate.
While this is true, it doesn’t actually say
anything about the likelihood of the event occurring, and might
imply that all events are made more likely. Given that often we
find that climate change played no role in changing the likelihood
of an event, and sometimes we find climate change has made an
extreme weather event less likely, this is not ideal.
To avoid such issues, we are working on
developing ways to assess the links between a particular extreme
weather event and climate change that work a lot faster.
Real-time event attribution
Many uncertainties remain in attribution
studies. It’s also easier to work with some types of weather events
than others – there is generally more confidence in
heatwaves than those focusing on
precipitation. Investigation of hurricanes
and typhoons is currently limited by the ability of global climate
models to simulate these events.
Models also tend to produce more reliable
results in certain regions – for example, the mid-latitudes are
easier to model because their weather is much more random and does
not rely on getting large scale processes exactly right. Monsoon
regions, on the other hand, are trickier.
So far the events we’ve examined have been
picked rather randomly, leading to a very patchy picture of how
climate change is affecting us already. The next stage of our work
is to begin to fill the gaps and systematically investigate how
anthropogenic climate change is affecting us today.
We won’t be able to give a robust answer in
every case, due to limited data and limits to our understanding.
But we are working towards developing a more complete picture, that
can hopefully lead to better understanding of what climate change
means for extreme weather in different parts of the
Dr Friederike Otto is a research fellow in the ECI
Global Climate Science Programme at the Uiversity of Oxford and
scientific coordinator of climateprediction.net.