Calvin Coolidge, US president from 1923 to 1929, had a great dislike of ostentation. "Men in public office," he declared, "should substitute for the limelight that light that comes from the burning of the midnight oil." Consistent with this paradigm, he had a capacity for quiet, hard work, and a liking for trenchant phrases that were often humorous, but invariably terse. His character was such as to prompt Dorothy Parker, when told of Coolidge's death in 1933, unkindly to remark: "How can they tell?"
The same question could be asked about the Intergovernmental Panel on Climate Change, the IPCC, which a few days ago issued its latest report giving the direst warnings yet about the prospects for global climate. How does it know what the climate of the world may be more than 100 years from now? The answer lies in what are known in the trade as CGCMs; the acronym stands for Coupled General Circulation Models.
CGCMs, like the models used to produce the daily weather forecast, are based on a mathematical model of the atmosphere - a description of the behaviour of the atmosphere in terms of mathematical equations. The computer starts with a description of the current atmosphere, and applying the equations, works its way forward to some time in the future. But there are important differences between the "forecast" models, used to predict the weather a day or two ahead, and the "climate" models, which predict the general state of the Earth's climate decades hence. Forecast models have to predict in great detail the behaviour of relatively small-scale atmospheric phenomena, and are therefore sensitive to tiny errors in the accuracy of the data which define the state of the atmosphere at the beginning of the exercise.
Climate models, on the other hand, are concerned only with broad trends in the behaviour of the atmosphere, expressed in average figures for a large area. But then, the number of time-steps necessary to advance computationally over a period of 50 years or so, is vast. And climate models have their own complexities. In the case of, say, a five-day forecast, reasonable results can be obtained by assuming that there will be no change in sea-surface temperatures over that period. But if you have to think in terms of decades, the changes in the thermal structure of the oceans will be the very driving force behind the processes involved. The behaviour of the ocean itself must be predicted, and the atmosphere and ocean must be "coupled" - as the name CGCM implies.