Forecasting perfect storms


Science is giving us advance warning of dangerous weather combinations known as ‘compound events’. Michael Lucy reports.


John Finney Photography / Getty Images

The Russian spring of 2010 was unseasonably dry. Then the summer arrived with a blistering heatwave. Hundreds of fires sprang up in forests across the country’s west. Vast tracts of the country were clouded with smoke; in Moscow the air became almost unbreathable.

All told, the summer left 55,000 dead and sent aftershocks across the planet. Devastated farms meant wheat exports were shut down, which drove up world food prices and – in some analyses – spurred on the Arab Spring uprisings the following year.

Nothing like it had been seen before. A year later the same words were used to describe the catastrophic Brisbane floods that killed 35 and forced a whole town to be rebuilt on higher ground. Six years later we heard the same refrain in the wake of Hurricane Harvey, which devastated Houston.

These ‘never before seen events’ now have a technical name: ‘compound events’.

They are perfect storms blown up by cascading events that amplify one another’s effects.

The problem is we’re doing a poor job of predicting them. In the past, long-term weather records provided a fair guide to future weather. We could predict a location’s potential for floods, droughts, storms and heatwaves.

As the climate changes, all that is going out the window. “Historical records of natural hazards can no longer be our sole source of information on how future hazards will play out,” says Seth Westra, an engineer at the University of Adelaide who studies climate risk.

Westra is one of a small but growing group of researchers who are using the tools of risk analysis and physical weather modelling to predict and prepare for the compound events of the future.

The key insights are that global warming means climatic events will cascade in new and unprecedented ways.

In the Russian heatwave, for example, the lack of rainfall combined with the heat and the fires and the smog to deadly effect. Hurricane Harvey, for another example, killed more than 100 people and inflicted US $125 billion in damage, not because it was a particularly big storm but because it hit a high-pressure system that stopped it moving inland and dissipating. Instead, it looped around along the coast, spending more time over Houston and delivering more rain.

If you imagine natural catastrophes as being doled out by the roll of invisible dice, studying history has allowed scientists to build tables of how often each number is likely to come up. As human carbon emissions warm the planet, however, the weighting of the dice themselves is shifting.

“We can’t just look at the biggest flood we’ve ever had any more and use that as the basis for planning,” says Westra.

Instead, we will need to understand not only the physical processes that drive catastrophes – the rainfall that causes flooding, the heatwaves that set the scene for fire – but also how they are connected. Till now, long-range climate projections usually focussed on single variables, perhaps predicting heatwaves or rainfall without studying the interactions between them and the physical processes that tie them together.

These connections – how the roll of one dice will affect the next – are also changing. Take extreme heat: it is more likely to occur than in the past, but it is also more likely to occur at the same time as extremely high or low rainfall.

Global warming means climatic events will cascade in new and unprecedented ways.

Climate scientists have begun coming to grips with the problem of compound events in the past six years or so. As recently as 2012, a 500-page report by the Intergovernmental Panel on Climate Change (IPCC) contained barely half a page on the topic.

In 2014, however, Westra, along with his University of Adelaide colleague Michael Leonard and others, put together a framework for thinking about compound events and the risks they pose.

Earlier thinking about future catastrophes would often take a top-down approach, Leonard says, beginning with a large-scale climate projection and then trying to work out its effects on the ground. A better angle of attack, he says, is to think in terms of networks: start with a catastrophic result like fires or floods and map the events that could lead to it.

“You work back from there,” he says, using climate models to find the odds of the specific combinations of conditions that could trigger disaster.

Andy Pitman, the head of the Centre of Excellence for Climate Extremes at the University of New South Wales, says it’s tricky work.

“It’s not the majority of extreme events that catastrophically affect people,” says Pitman. He cites Australia’s east coast low-pressure systems as an example. If one hits Sydney, it fills up the dams. If another hits a few days later, there will be some flooding because the ground is already wet and the dams are full.

“If you have three in succession,” says Pitman, “you get catastrophic flooding and the dams might fail.” Water released from overfull dams was responsible for the flooding of Brisbane in 2011, and a similar prospect might befall Sydney. “So we’re moving our attention away from the probability of one east coast low to the probability of three east-coast lows hitting the coast in quick succession.”

The next steps, he says, are to improve forecasting – using better dynamic models of the Earth’s atmosphere to see what’s coming next. The impact of Hurricane Harvey was predicted by weather forecasts in the days before it hit. The Russian heatwave, on the other hand, was too big to be seen in its entirety much in advance.

As computer power increases, so does the scale of forecasts in time and space. In future, Pitman says, “there’s a good chance they could capture these kind of things”.

Changes in how extreme weather events combine are also a concern for the insurance industry. The traditional gold standard for insurers has been to calculate risk from past records, but this is beginning to change.

Most climate models don’t zoom in to the level of individual properties that might be insured against flood or fire, or even that of large events like cyclones, says Kate Simmonds, a catastrophe analyst at reinsurance broker Willis Re in Sydney. Timescale is the other issue: insurance contracts typically only last 12 months, so the longer view may not be relevant for current insurance premiums.

“There is a competitive advantage in thinking further into the future,” Simmonds says, “but you don’t want to incorporate changes that might happen after 30 years because then you’re going to be offering much more expensive policies.”

David Bresch, a former head of the Atmospheric Perils team at reinsurance giant Swiss Re who is now at ETH Zurich, says insurers should increase prices where appropriate. This would give customers an incentive to prepare for the future, to keep their properties insurable. Insurers, he says, “would be better informed by taking a long-term view into account.”

Even in purely scientific terms, the study of compound events still has a long way to go. Jakob Zscheischler, a climate scientist at ETH Zurich, is working on an international project to bring together climate scientists, impact modellers and statisticians.

“We need to establish methods to study different types of compound events. We also don’t really know how to evaluate how well climate models simulate compound events.”

Compound events may play an increased role in the IPCC’s Sixth Assessment Report – the official roundup of the state of climate science – due to be published in 2022. Sonia Seneviratne, a coordinating lead author on the report’s chapter on extreme events, is prohibited from discussing its content but regards compound events as “a very promising new area of research”.

Science is gearing up to give us better predictions of the ‘never before events’ that are coming our way. The challenge will be to prepare for them.

  1. https://www.reuters.com/article/us-russia-heat-deaths-idUSTRE69O4LB20101025
  2. https://muse.jhu.edu/article/582525
  3. https://www.nature.com/articles/s41558-018-0156-3
  4. http://iopscience.iop.org/article/10.1088/1748-9326/8/3/034014/meta
  5. http://www.pnas.org/content/112/37/11484?ijkey=fc1996caff13c39b10d7048bd0be233757e07bf6&keytype2=tf_ipsecsha
  6. https://wg1.ipcc.ch/srex/srexdownload.html
  7. https://onlinelibrary.wiley.com/doi/abs/10.1002/wcc.252
  8. http://www.cost.eu/COST_Actions/ca/CA17109
Latest Stories
MoreMore Articles