Flood of floods: Here comes the rain

 by Stephen Battersby and Michael Le Page 

New Scientist Magazine issue 2804, 24 March 2011

Torrential downpours are on the increase, but is climate change to blame?

IT WAS a monstrous monsoon. Over just a few days in late July last year, more than 300 millimetres of rain fell on northern Pakistan. As the water swept down the river Indus, it killed close to 2000 people and affected 20 million more.

Pakistan was not the only place to suffer. Australia, China, Thailand, Brazil, the Balkans, Bangladesh, Indonesia, Colombia, the Philippines, Sri Lanka and Tennessee all experienced devastating floods in the past year. What's more, there were unusually heavy snowfalls in many regions, severely disrupting transport systems. Globally, 2010 was the wettest year ever recorded (see chart).

Inevitably after such freakish weather, people ask whether climate change is to blame. Is this flood of floods our fault? And is there worse to come?

There is no doubt that temperatures are rising, and basic physics suggests that warmer means wetter, because warmer air can hold more moisture. Observations confirm that the lower atmosphere holds about 5 per cent more water than a century ago, giving it that much more ammunition to unleash in a downpour or blizzard.

So does this explain the recent floods? "I don't think it is legitimate to assume that climate change played a role in these events until we've done the work," says Myles Allen of the University of Oxford. There are many types of floods, he says, and while climate change is making some extreme events more likely, others - such as floods caused by melting snow - may become less likely.

Others think we can say more. "The pervasive increase in water vapour changes the intensity of precipitation events with no doubt whatsoever," Kevin Trenberth of the US National Center for Atmospheric Research told a meeting in January. "Yes, all events. Even if temperatures or sea surface temperatures are below normal, they are still higher than they would have been, and so too is the atmospheric water vapour amount and thus the moisture available for storms."

For instance, El Niño's sister phenomenon, La Niña, which brings warmer surface waters - and thus moister air - to the western Pacific, has been blamed for some of the recent floods in Asia and Australia. Trenberth thinks high sea surface temperatures due to global warming also played a role. "The La Niña in place favours these sorts of things but the extra high sea surface temperatures make them record breaking," he says.

It is clear is that rainfall patterns are changing. In the US, for instance, Kenneth Kunkel of the National Climatic Data Center in Asheville, North Carolina, is analysing data from more than 1000 rain gauges across the country. Over the past century, the intensity of extreme rainstorms that occur once per year on average has risen 1.4 per cent per decade, he has found.

Put another way, in 2000, a "once a year" storm was dumping 14 per cent more rain on average than it would have done in 1900. "It would be wise for society to take this into account when building a dam to last 50 or 75 years, or a new housing development," says Kunkel. "That's not really done right now."

At the top end of the scale, the intensity of "once in 20 years" extreme rainstorms is increasing even faster, by about 3 per cent per decade, although for these rare events the statistics are more patchy. The overall increase also seems to be accelerating. Most of it happened in the last three decades, and Kunkel hasn't even included 2010 in his analysis yet. "I'm guessing that when the data come in, they are going to be high," he says.

Bucketing down

Similar studies show that heavy rainfall has been increasing in intensity in most other parts of the world too, just as climate models predict. These results can't be put down to changing methods or issues with scientific instruments, as almost all the figures come from the humble rain gauge, unchanged over the decades. "It is basically an 8-inch bucket - so simple that you almost can't call it technology," says Kunkel.

Insurers are also seeing a rise in claims for flooding. Much of that increase can be put down to demographic factors such as population growth, but not all, says Gerd Henghuber, a spokesman for insurance firm Munich Re. Claims for flood damage are rising faster than claims for other natural disasters such as earthquakes and eruptions. "The growing number of weather-related catastrophes most probably cannot be fully explained without climate change," he says.

So physics says that climate change should lead to more rain, buckets say that more rain is indeed falling and insurers say that that more rain is causing more destruction - a chorus suggesting that we are, indeed, bringing the skies down on our heads. But the case is not quite closed.

In theory, some long-term natural oscillation might be to blame for higher rainfall over the past few decades. "The climatic system has natural periodicities that operate over a whole range of timescales," says Stephen Gale of the University of Sydney in Australia, who studies past flooding. To pin the blame squarely on humans, we need to rerun the past century minus all the added greenhouse gases to see what happens. In the absence of spare Earths or a time machine, that means using computer models.

In 2007, a team including Francis Zwiers, now at the University of Victoria in British Columbia, Canada, compared the results of 14 different models with rainfall data for the 20th century. Simulations that included greenhouse gas emissions matched the overall global pattern seen so far, with increased rainfall in most regions (Nature, vol 448, p 461).

Still, that study looked only at total precipitation. When it comes to flooding, what really matters is whether climate change is making extreme events even more extreme. This is a harder question to tackle, partly because extreme events are rare and so our information on them is limited.

Extremes are also difficult to model, says Elizabeth Kendon of the Met Office Hadley Centre in Exeter, UK. One reason for this is that small-scale processes are important. The internal convection of storm clouds has to be approximated, for example, as it is on far too fine a scale to be modelled.

Despite these difficulties, Zwiers's team has now done a study comparing records of extreme rain and snowfall in the northern hemisphere between 1951 and 1999 with the simulations produced by eight models (Nature, vol 470, p 378). "Our research indicates that humans have contributed to the observed intensification of extreme precipitation, but we're not in a position to quantify the magnitude of that contribution," says Zwiers.

In another recent study, a team led by Allen looked at one specific event - the floods that occurred in the UK in autumn 2000, causing £1.3 billion worth of damage. With the help of computer time loaned by volunteers via climateprediction.net, they did thousands of model runs simulating the weather in 2000, with initial conditions both as they were and as they could have been without global warming. They found that climate change had nearly doubled the likelihood of this kind of flood (Nature, vol 470, p 382).

The limited predictive ability of the model does not undermine this finding, Allen says. The work is comparable to proving dice are loaded: you don't need to be able to predict the outcome of each throw, you just need to show sixes occur more often than expected.

Put it all together, and the case that global warming is leading not only to more rainfall but also to more extreme rainfall, and is thus making severe floods more likely, starts to look very convincing. The big question is, what is going to happen next?

For every 1 °C rise in temperature, the water content of the lower atmosphere is expected to rise by 7 per cent, and temperatures could rise by 4 °C as early as the 2060s. That is a lot more water in the air.

Climate models suggest that precipitation will become more intense pretty much everywhere: when it rains or snows, it will rain and snow harder. However, there could be longer dry intervals in between, and in some parts of the subtropics less rain than usual may fall overall. "You can have a situation where mean rainfall decreases but the extremes increase," says Kendon. In other words, droughts interspersed with floods.

Models produce a wide range of results when it comes to extreme events in the future. Take winter rainfall in the UK, for example. Based on a scenario in which global temperature rises just 3 °C, some models suggest that rainfall on the worst 5 per cent of rainy days will increase by 50 per cent, while others suggest there will be no change.

What's more, how much rain falls is only part of the story. Warmer weather will dry out soils faster, for instance, which reduces the risk of flooding. "The relationship between precipitation and flooding is immensely complex," says Gale. "Variation in a single factor, such as soil moisture, might mean that the same precipitation event could remain in channel or could generate a significant flood."

On top of this, we are changing the physical nature of river valleys enormously. "What we do is take all the agricultural land around cities, pave it, build houses and put in storm-water sewers, all of which stops water from infiltrating the ground where it can be stored and percolate slowly into rivers," says Gale. Instead, rainwater is funnelled rapidly into main rivers. So part of the problem with forecasting future flooding is not knowing how the landscape will change.

Flooded out

Another big problem is that we do not even know what has happened in the past. Records of flooding in most countries are surprisingly brief and full of holes. "We don't have the data to predict big events," says Gale.

Most government planners and engineers, for instance, rely on river gauging stations to record the height that floodwaters reach. These are a relatively recent tool. "If you have flood data that go back 50 years, you are doing well," says Gale.

What's more, gauging stations can be overwhelmed by a big flood. "Before 2010, the last big floods in Australia were back in 1990," says Gale. "More than 220,000 square kilometres was under water; the town of Nyngan was evacuated, entirely flooded out, and we have no idea how high the flood there was. Hydrologists went in after the event and estimated the discharge indirectly, but there is no official record, the flood was just too big."

So predicting the extent of a 100-year flood means extrapolating using statistical methods, and the results depend heavily on the assumptions used. "Predictions for the flow rate of a 100-year flood might vary by a factor of 4 or more," he says.

Zwiers's latest study suggests there could be yet another problem with flood forecasts. The team's "best guess" is that the observed increase in extreme precipitation is two to three times larger than global climate models predict, though the uncertainties are huge. It is possible that natural variability is to blame, but the most likely explanation is that the models are underestimating the extent of change. If so, extreme precipitation events may strengthen more quickly in the future than projected and have more severe impacts.

The upshot of all this is that while all the evidence points to an increased risk of flooding in large parts of the world as a result of climate change, no one can say for sure exactly which parts of the world or how bad it will be. This is not much help to planners.

Nevertheless, there are many measures we could be taking, from banning development in flood-prone areas to using permeable paving to reduce run-off. "There are lots of sensible things we could do," says Allen. He points out that these would benefit us whatever happens, come rain or shine.

Stephen Battersby is a consultant for New Scientist based in London. Michael Le Page is a features editor