Professor Myles Allen , Head of Atmospheric, Oceanic and Planetary Physics at the University of Oxford, explains how researchers at Oxford are combining climate information with operational weather forecasting systems to understand the changing risks of extreme weather.
As temperatures climb towards the peak of our fifth heatwave of the summer, once again we are hearing the question 'is this the new normal ?' I recently heard an expert on the radio explaining 'what people need to understand is that climate change has happened.' If only.
The real problem is, of course, that the climate is continuing to change relentlessly - and predictably, at least as far as large-scale temperatures are concerned: after you account for random weather fluctuations, most regions and seasons have warmed in close to a straight line since the 1980s, which is precisely what we would expect from the steady build-up of greenhouse gases in the atmosphere.
Warming in Europe may have accelerated slightly because we have cleaned up the pollution that used to cause smogs and acid rain and which had also been masking the warming impact of greenhouse gases . Cue headlines such as ' Heatwaves caused by fall in pollution ' - conveniently focussing on less than half the story. But, as Stuart Jenkins demonstrated in his DPhil, any acceleration in temperature trends is barely detectable , meaning that the rate of warming still looks very close to a straight line. So why does it feel like the impacts of climate change are rising exponentially?
Part of the answer goes straight to the heart of 'extreme event attribution', one of the thorniest topics in climate science. The problem is there are two ways of framing the question of how climate change is affecting extreme weather, each suited to a different question. When I first wrote on this topic back in 2003, anxiously watching the flood waters off the Abingdon Road creep towards our kitchen door, I framed it in terms of greenhouse gases loading the weather dice, gradually increasing the odds on a six.
"It is dangerously complacent to talk about the 'new normal' as if we are already experiencing the climate of the future. We can stop climate change: not by banning fossil fuels, but by eliminating or replacing inefficient uses and safely disposing of the carbon dioxide generated by the remainder."
Whether the dice comes up six, or whether our kitchen would flood that year (it didn't), is a discrete event: it either happens or it doesn't. But another way of thinking about it is climate change painting an additional dot on each face of the dice. If the casino pays out for any roll of six or greater, the impact on your chances of winning is exactly the same as loading the dice to double the odds on rolling a six, but the message is subtly different. Rather than 'doubling the odds', climate change is simply adding one to the score, turning a five into a six and nudging an otherwise losing roll across the winning threshold. That sounds like a much smaller contribution, even though the change in risk is identical.
More controversially, thinking in terms of extra dots makes it tempting to claim that an observed event would have been 'impossible without climate change' - like rolling a seven. This is where the science becomes much less certain. We can observe only one climate: the one that actually happened. Any statement about what would have occurred in a world without past greenhouse gas emissions depends on models of a counterfactual climate that can never be tested directly.
So why does it feel that the impacts of climate change are suddenly coming thick and fast, when average summer temperatures and even the estimated contribution of climate change to peak temperatures have both been rising in close to a straight line? There are many factors in play but there is one simple explanation that's directly related to our work here in Oxford.
For most extreme weather indicators, like peak daily temperatures, the probability of crossing a threshold goes down approximately exponentially as the threshold increases: if the odds on temperatures peaking over 31°C in any given year is half the odds on them peaking over 30°C, then the odds on a peak over 32°C is half the odds on a peak over 31°C, and so on.
The rate of this exponential decline, sometimes known as the 'Gutenberg-Richter (G-R) slope', cannot be predicted reliably from basic physics. Nor can it be estimated directly from observations for the most extreme events, because our records are simply not long enough. The most accurate method of predicting it is by 'rolling the weather dice' many times with a model that reliably captures all the self-reinforcing processes that drive extreme weather.
"If you, or your children, or your business are sensitive to temperature thresholds, then the impacts of climate change on you really are increasing roughly exponentially. If you're sensitive to average temperatures, then impacts are increasing in a straight line."
Oxford has been pioneering this large-ensemble approach for many years - you may have even helped us by donating your computing time to the climateprediction.net project. The trade-off was that we had to use relatively simple climate models. Recently, however, Dr Nicholas Leach , part of Oxford's Predictability of Weather and Climate research group, has been addressing this challenge using the European Centre for Medium-Range Weather Forecasting (ECMWF) day-to-day weather forecasting system. Unlike most climate models, these forecasts are run at much higher resolution and are tested against real weather every day, with particular attention paid to extreme events.
Current DPhil student Olivia Vashti-Ayim , funded by the Man Group , has been comparing how different models represent the probabilities of very extreme weather. She found substantial differences in many parts of the world between the ECMWF forecasting system and the climate and statistical models widely used for extreme event attribution. In general, the ECMWF system produces 'fatter tails': the chances of the most extreme events fall away more slowly than in the other models.
Why does this matter, and how does it relate to your experience of summer heatwaves? If probabilities halve very quickly as temperatures rise, a given amount of warming of the entire distribution produces a large increase in the probability of exceeding any given threshold. A fatter tail means the probabilities decline more slowly, so the same amount of warming has a smaller impact on the odds of threshold exceedances, because they were more likely to have happened by chance.
Either way, if the whole distribution of daily temperatures shifts steadily upwards, then the chances of temperatures exceeding a particular threshold on any given day increase exponentially at the inverse of the G-R slope. So, if you, or your children, or your business are sensitive to temperature thresholds, then the impacts of climate change on you really are increasing roughly exponentially. If you're sensitive to average temperatures, then impacts are increasing in a straight line.
Olivia is currently funded by a Wellcome project on the impact of changing weather extremes on children's development, led by Professor Alan Stein . Children, like most of us, don't experience average temperatures: they experience the crossing of temperature thresholds, like schools being closed if daily temperatures look like exceeding 30°C. And the frequency of those threshold-crossings really is rising exponentially - and will continue to do so as long as we continue to warm the world.
Either way, it is dangerously complacent to talk about the 'new normal' as if we are already experiencing the climate of the future. We can stop climate change: not by banning fossil fuels (it is increasingly clear that will take too long), but by eliminating or replacing inefficient uses and safely disposing of the carbon dioxide generated by the remainder. This could be done , and surprisingly quickly. But as long as we continue just dumping carbon dioxide into the atmosphere, I'm afraid we're going to have to buckle up for more and faster changes to come.
