Weather: Supermodels keeping up with fashion

The global warming debate depends heavily on computer models of climate change. It may, however, take another

50 years before we know whether we can trust the predictions of current models.

As the Kyoto conference at the end of last year confirmed, the threat of global warming is one of the major concerns of our time. Anyone reading Sir John Houghton's excellent book Global Warming - The complete briefing (CUP, 1997) will find a convincing argument in favour of the major tenets of global warming: that the earth is getting hotter; that human activity (say "anthropogenic causes" if you want to impress people) is making a significant contribution to this warming; and that we really ought to do something about it before it is too late. Yet there is one aspect of the scientific study of the topic that I find not fully convincing.

Everything depends on computer models of climate change. These involve identifying all the components that can influence climate - including our orbit round the sun, the composition of our atmosphere, and the heat storage and transmission patterns of our oceans, among others - and relating them to observed changes in temperature. If we can produce a computer model that fits all the data from the past, we can be reasonably confident that it will also predict the future.

If we look at the graph of world temperatures, however, we see a wobbly, but overall fairly stable, pattern from 1860 to 1910, then a consistently rising curve until about 1940, relative stability again until around 1960, then a gradual decline until the mid-Seventies, since when we have been on a rising curve taking us to well above the previous peaks of the 1940- 60 era.

The best models nowadays are those that include the concentrations of both greenhouse gases and sulphate aerosols in their equations, leading to a remarkably close fit with the observed picture of the past 25 years.

For the previous century, there is broad agreement with observed figures, but, not nearly such a close fit as for recent times. And that is what is so disturbing. The present model looks as though it was designed to fit the rising temperatures of recent years - just as the computer models of the Seventies were designed to fit the fashionable worry of the time of falling temperatures.

If you design a computer model to explain a recent trend of rising temperatures, it is almost certain that it will predict a continuation of that trend - just as the models of the Seventies predicted an ice age around the corner.

There seems no doubt that the earth is still warming up after the Little Ice Age that ended around the middle of the 19th century. Within that pattern the last 25 years have seen steeper temperature rises than expected. We cannot yet be sure whether this is just one of those wobbles in the graph that cannot be explained, or whether it is all our fault. In another 50 years or so, we shall see how accurate today's models really are.