Forecasting the weather: making sense of chaos

"Red sky at night, shepherd's delight. Red sky in the morning, shepherd's warning." For thousands of years, people have examined nature and its elements to predict the weather. Light, clouds and winds may indicate a likely change in the weather. Even today - in the absence of sophisticated measurements and computer simulations - people's livelihoods in many parts of the world rely on the handed-down knowledge of these signs.

At the heart of complex climatic systems lies the atmosphere. Changes in air and sea temperatures, wind, water vapour, even the presence of Saharan dust particles can impact on regional and global weather systems. Tropical weather systems in particular drive the global climate system, redistributing solar energy and determining the climate across the globe.

It is a chaotic, unpredictable system. Over half the world's population lives in the tropics, where failed rains or heavy flooding can be devastating, especially for farmers who rely on seasonal rainfall. Understanding the factors behind year-to-year variations in rainfall is therefore crucial. But how easy is it to determine seasonal weather patterns with any certainty? And will the science of forecasting the weather become more challenging as global warming makes the climate more unpredictable?

The probability of change

Despite the chaotic nature of climatic systems, the weather is partly predictable. Weather forecasters cannot know for certain that it will rain today or tomorrow, but they can estimate the probability of rain. With the help of supercomputers, weather forecasters can run a series, or ensemble, of forecasts with slightly different starting points to provide a range of possible weather scenarios. Modern weather forecasting usually involves running around 50 ensembles; a degree of similarity between the scenarios results in a higher probability of a particular weather event occurring.

The single largest cause of the planet's year-to-year climate variability is El Niño, which usually occurs every 3-7 years and generally lasts for about a year before conditions in the Pacific return to normal (see box). The strongest El Niño of recent times peaked in December 1998, resulting in floods in the Americas, storms in China, severe drought in Indonesia, and forest fires in Southeast Asia and Brazil. In Mexico, the town of Guadalajara saw snow for the first time since 1881. Models failed to predict the onset and intensity of these events, knowledge of which might have averted some of the 2,000-plus deaths and damage valued at over US$30 billion. However, advances in the long-range forecasting of variations in atmospheric pressure between the Pacific and Indian oceans, known as El Niño Southern Oscillation (ENSO), mean that scientists can now predict conditions up to 6-9 months ahead.

Whilst predicting El Niño events will undoubtedly bring social and economic benefits in the developing world, existing capacity to run simulations is insufficient for predicting the magnitude of global warming and the changes in the climatic system that may occur by the end of this century.

Extreme global warming: an unknown risk

The Intergovernmental Panel on Climate Change (IPCC) report of 2001 states that a doubling of carbon dioxide (CO₂) levels by 2080 will cause a global temperature rise of 2-5 degrees Celsius. However, new estimates published in Nature, based on computer simulations run in parallel on the computers of almost 100,000 people around the world, reveal a previously unknown risk of catastrophic warming of up to 10-12 degrees if CO₂ emissions go unchecked. This, according to Tim Palmer of the European Centre for Medium-range Weather Forecasts, is because the magnitude of global warming will depend critically on how CO₂ levels affect the primary natural greenhouse gas, water vapour.

Until recently, computer simulations could not accurately represent the deep convective clouds that form over the tropics and drive weather systems. Models have now been developed in which convective clouds are simulated properly, rather than parameterised. Century-long simulations would still take decades to run even on Japan's Earth Simulator, the world's most powerful computer, which can process climate models at 10¹² calculations per second. Further reducing uncertainty in climate change forecasts will require a computer working at 10¹⁵ calculations per second, or 1,000 times faster than the Earth Simulator. A collaborative group of institutes known as the European Network for Earth System Modelling (ENES, www.enes.org) is therefore lobbying the European Union to establish a European Virtual Earth Laboratory. This international supercomputing facility would provide cutting-edge climate modelling and a better understanding of the probability of extreme global warming and the measures required to mitigate and adapt to it.