Want to know what computer models say about the climate? Roll a dice…
When it comes to predicting the future of our planet, how much faith would you put in a computer capable of a million-billion operations a second? It’s an awesome machine for an awesome task, but Britain’s climate experts believe the computing colossus, being readied for operation at the Met Office’s Hadley Centre for Climate Prediction and Research, will bring them closer to answering some of the most pressing questions facing the world today. Just how hot will our planet become over the coming decades? Where will the worst effects of global warming be felt? And what can be done to protect those most at risk?
The Hadley Centre supercomputer will start its assault on these issues next year using the most sophisticated simulation of the Earth’s climate ever devised. Its millions of lines of computer code make up a mathematical model of Earth’s atmosphere, land masses, oceans and ice-caps – even down to the level of dust in the air and the state of plants on the ground.
Fed with data about current climatic conditions, the supercomputer will then work out how today’s climate will change under various scenarios – such as projected future levels of man-made pollution. The results are eagerly awaited by climate scientists, who will make them the basis of their advice to the world’s governments on how to avert climatic disaster. But just how reliable are the results of these computer models? Can they be trusted by governments faced with making policy decisions that will affect all our lives?
Climate modellers admit that even the power of supercomputers like the Hadley Centre machine or Japan’s Earth Simulator is inadequate to do full justice to the complexity of the climate. Even so, they are confident that with ever more computing power, they will get closer to the goal of reliable climate prediction.
“This next generation of climate models will include detailed chemistry and interactions with the plants, soil and ocean biology,” says Dr Vicky Pope, Head of Climate Prediction at the Hadley Centre. “It is only with the introduction of new supercomputers that we will have enough computing power to run these models.”
But according to some experts, the drive for ever more computing power misses a far more basic problem with current climate models. They argue that the models are still too crude to be reliable, lacking subtle effects that can have a profound impact on the Earth’s climate.
And as evidence, these critics point to the recent discovery of worrying discrepancies between what climate models say should be happening and the reality. “The models are seen to disagree with the observations,” 3 3says Dr Fred Singer of the University of Virginia, adding that the models “should be viewed with caution.”
So who is right? With the future of the planet at stake, this is no academic squabble: everyone’s lives could be profoundly affected by the outcome of these supercomputer simulations.
While they may only now be attracting worldwide media interest, climate models have been around for over 50 years, with scientists working on them since the first commercial computers became available in the 1950s. But even then, the same two issues dogged the research: lack of computing power, and concern about what vital ingredients the models were missing.
The study of long-term weather patterns, usually over a number of years. It’s used to devise possible future projections by analysing current conditions and the causes for them
Computerised simulations based on real data that aim to show how the temperature and weather in a particular part of the planet will be affected by various atmospheric influences
Keep it simple, stupid
The dangers of using too simplistic a model became clear early on. In 1971 two NASA scientists carried out a pioneering study of an issue that remains central to climate research: the role of dust and smoke particles on global warming. Known technically as aerosols, these tiny particles have a host of sources: some natural, such as deserts and volcanoes, and some man-made, like coal power stations. The NASA study looked at how these aerosols might decrease global warming due to carbon dioxide by reflecting more of the Sun’s heat back into space.
Using estimates of man-made aerosol levels over the next 50 years, the researchers made a stunning discovery: the reflection effect had the power to halt global warming. They warned that temperatures could drop by 3.5°C, and might even plunge the planet into a new Ice Age.
While the researchers never claimed their scenario was inevitable, the idea of a new Ice Age attracted huge media coverage during the 1970s. The fact that climatologists came to believe the exact opposite is far more likely has, not surprisingly, prompted cynicism about climate models. Yet, in reality, the study was always seen by climatologists as just a crude prototype of the far more sophisticated models they have since gone on to build. For example, the NASA model failed to take into account the fact that aerosols tend to be concentrated in certain areas – for example, over industrialised Europe – while greenhouse gases are spread more evenly around the planet. The rates of aerosol production, and whether they all have a cooling effect, were also brought into question.
Today’s models take a much more sophisticated view of aerosols. They include not only their direct impact – such as their cooling effect, caused by reflecting sunlight back into space – but also their indirect effects. And chief among these is the way in which aerosols act as ‘seeds’ for the formation of clouds, whose effect on global warming is more complex.
According to Pope, clouds can produce both a cooling and a warming effect. “On a cloudy day, less heat from the Sun reaches the Earth’s surface, and we feel cool,” she explains. “On the other hand, on a cloudy night, the heat generated during the day is trapped, and the temperature near the surface remains relatively warm.”
Finding out how clouds help and hinder global warming is arguably the biggest challenge facing today’s climate models. Yet attempts to simulate the white, fluffy things that float in the sky still brings even today’s supercomputers to their knees. “Clouds have sizes of around several hundred metres, but the spatial resolution [or scale] of the best current climate models is 100km – which is bigger by a factor of 1000,” points out climate expert Professor Thomas Reichler of the University of Utah. “This means that individual clouds and their response to climate change are not well-simulated in current models.”
So it is important to model clouds in detail because their climatic impact depends on their height, thickness and thermal properties, which all depend on their individual size. But without the computing power to model clouds in detail, climatologists have been forced to capture just their basic features, like temperature and humidity. The hope is that this will give at least some guide to the impact of clouds on future climate.
“The biggest advances in recent years have been in our representation of how clouds, oceans, sea ice and climate interact with life on Earth”
According to new research, current computer models may overplay the link between clouds and global warming. Since 1980, the average temperature over Europe 3 has increased by 1°C – far faster than expected if only greenhouse gas increases were to blame. One possible explanation is that more sunlight is now reaching the ground because of clean air campaigns, which have greatly reduced the levels of cloud-forming aerosols over Europe.
To put this theory to the test, a team led by Christian Ruckstuhl of the Institute for Atmospheric and Climate Science in Switzerland measured aerosol levels at various sites over Europe. And the results seem to confirm that falling pollution levels are indeed responsible for the relatively rapid rise in temperature. But the team also found that the level of sunlight reaching the ground was barely affected by the presence of clouds. In other words, most of the observed warming appears to be directly due to the cut in aerosols – and not to their role in cloud formation, as assumed in many computer models.
Some researchers go further, arguing that the link between clouds and temperature may have been fundamentally misunderstood. In research published in June, researchers at the University of Alabama, Huntsville, claimed that clouds tend to respond (cloud cover increases) to increasing temperatures, rather than cause them. According to team member Dr Roy Spencer, their research suggests that clouds actually reduce global warming – contrary to what many models now predict – and may generally act as a brake on any type of climate change. “Our work has convinced me that cause and effect really do matter,” he says. “Unfortunately, in the process it also makes the whole global warming problem much more difficult to figure out.”
But clouds are just one feature of our planet whose subtleties pose a major challenge to climate prediction. The climate model being designed for the new Met Office supercomputer will include the effect of vegetation – and some researchers have warned that even such fine details as how moisture travels through plant roots can have a major effect on climate forecasts.
In the face of so much complexity, can faith be put in computer predictions about the climate of the future? “If our knowledge of global warming were entirely derived from climate models, I would probably also be more sceptical,” says Reichler. “But most of our knowledge is actually derived from relatively simple and well-established theories and physical laws.”
“Our work has convinced me that cause and effect really do matter. Unfortunately, in the process, it also makes the whole global warming problem much more difficult to figure out”
According to Reichler, the main role of computer models is to fill in the broad-brush conclusions – for example, by providing insight into the effect of climate change on specific regions. And the models do seem to be getting more reliable, he says. In research published earlier this year, Reichler and his colleague Dr Junsu Kim tested over 50 climate models used by research centres around the world. The challenge was simple: when fed with 20 years of data about climatic conditions up to the late 1990s, how well do the models predict today’s climate?
The results were impressive, with the best models producing predictions very close to what has actually been observed. “We can now place much more confidence in model-based projections of climate change than in the past,” says Reichler.
Not everyone agrees. A recent study of over 20 leading climate models suggest they fail to reproduce the basic features of global temperature changes observed since the late 1970s.
According to Dr John Christy of the University of Alabama, the models predict that the lower atmosphere (the so-called troposphere) should be warming more than the surface, with the biggest effects being seen in the tropics. “When we look at actual climate data, however, we do not see accelerated warming in the tropical troposphere,” says Dr Christy. “The warming trend we see in the tropics is less than half of what the models forecast.”
According to some climatologists, the culprit is not the climate models, but the observational data used for the comparison, which is itself subject to errors. The debate looks set to continue, but most experts insist that the big message emerging from climate models remains unchallenged: the planet is warming up, and human action is largely responsible for it. Says Reichler: “Based on the wide range of evidence, I have no doubt about the reality and seriousness of global warming – and the necessity to act.”
Robert Matthews is a visiting reader in Science at Aston University
A question of reflection
How measuring light on the Moon can improve climate models
The ghostly glow often seen lighting up the dark half of a crescent moon is giving climatologists a new way of testing the reliability of their computer models. Called ‘Earthshine’, the phenomenon is caused by sunlight reflecting off the Earth and striking the Moon. As such, its intensity depends on climate-related features, such as cloud cover, levels of sea-ice and snowfall over mountains. Measurements by Earth-based telescopes have shown a marked rise in the reflectivity of the Earth’s surface since 1999, which coincides with a slow-down in global warming. Climate scientists will run their computer models to check they can reproduce this link between reflectivity and recent levels of global warming – and it may help them explain it.
- After striking the Earth, the reflected sunlight is bounced off the Moon and returns to Earth where it can be observed by ground-based telescopes.
- Sunlight strikes the Earth and bounces off its surface. Different surfaces reflect different amounts of light – oceans reflect relatively little, clouds and ice a lot and land varies between the two.
FOR and AGAINST
Can we trust computer climate models?
Dr Vicky Pope, Met Office Hadley Centre
“Climate models solve complex mathematical equations based on well-established physical laws that define the behaviour of the weather and climate. The models are tried and tested in a number of ways. First, they’re used to reproduce the climate of the recent past. Second, they’re used to reproduce what we know about ancient climates, which is more limited. The Met Office Hadley Centre’s computer model is unique in that it is also used with more regional detail to produce the weather forecasts every day.
“Global climate models have become more complete over the years as computer power and our understanding of climate change has increased. The biggest advances in recent years – and where further advances are still needed – have been in our representation of how clouds, oceans, sea ice and climate interact with life on Earth. The resulting climate projections have demonstrably improved over recent years. Further improvements are necessary and achievable.”
Prof John Christy, University of Alabama
“There are so many complex and poorly understood processes in the climate system that long-term predictions should always be suspect. For example, there are large unknowns in our understanding of how the climate responds when it’s heated or cooled. This is directly related to clouds and rain – the two things climate models do a poor job of imitating. Recent observations indicate the climate is not as sensitive to increasing CO2 as is popularly believed, because clouds apparently react to keep the climate from going awry.
“What we need is more fundamental research focused on observational data. Our ignorance of the climate system is astounding. For example, we don’t have a global observation system that provides data from which we can determine the interplay of all of the climate factors. This is a fundamental requirement if we are going to improve climate models. Understanding what the climate does, and why, will lead to more reliable computer models.”
What the models predict for the future weather in the UK
2020 If greenhouse gas emissions remain relatively high, annual temperatures over the whole of the UK are likely to rise by approximately 0.5°C over the next 10 years.
2080 By 2080, marked regional differences are likely to emerge, with temperatures in the most densely populated South-East of England rising by as much as 3.5°C.
2020 Assuming relatively high greenhouse gas emissions, computer models predict that annual rainfall levels could drop by 10 per cent across a broad swathe of the UK by 2020, increasing drought risk.
2080 Most of the UK could face a higher risk of drought by 2080, through much drier summers in England, Wales and Ireland. However, eastern regions can also expect to see more rain
The supercomputers that recreate the real world to predict what the weather will do in the future
Predicting the future of the climate is one of the toughest problems in all of science – and that demands incredible computing power. The computer models work by mathematically slicing and dicing the entire surface of the Earth, complete with mountains, oceans and the atmosphere, into billions of ‘cells’, all of which interact with each other according to the laws of physics and chemistry.
Studying the effects requires computing behemoths like the machine now being installed at the Met Office’s headquarters in Exeter. Sitting in an air-conditioned building, the supercomputer is an array of interlinked processors whose combined power is equivalent to 100,000 home computers. When the machine reaches full power in 2011, it will be capable of a peak speed of a million billion operations per second. Yet even this isn’t enough to give the level of detail needed for reliable predictions about climatic change for individual countries. So earlier this year, climate scientists called for an international effort in order to create a ‘hypercomputer’ several thousand times faster even than the Met Office’s new machine in the Hadley Centre.
Your humble home PC can make climate predictions too
You don’t need a supercomputer to try climate modelling. Your PC is enough – once it’s teamed up with thousands of others across the world. Since 2003, around 150,000 PC owners have joined the Climateprediction.net project, set up by Dr Myles Allen of Oxford University to create the world’s most powerful climate model. It works by exploiting the computing power of PCs when they’re sitting around doing nothing. So far it has notched up the equivalent of 35 million years of run-time, allowing climate experts to investigate a host of ‘what if?’ scenarios. The project has shown how even slight changes in starting conditions can lead to dramatically different forecasts of future global temperature rises, ranging from less than 2°C to over 11°C.
Find Out More
Climate model that’s pooling the power of thousands of computers www.climateprediction.net
Reicher and Kim report into the effectiveness of climate models www.bit.ly/1q4sRI
A comparison study of climate models and tropical temperature trends www.bit.ly/3rkuoT
Computer models run by the Met Office www.bit.ly/Woj0f
Follow Science Focus on Twitter, Facebook, Instagram and Flipboard