Weather chaos

The early 1960s saw a boom in scientific studies, including meteorology, but in the background one particular sceptic was raising eyebrows. Edward Lorenz, a professor of meteorology at the Massachusetts Institute of Technology, discovered what still seems to be an outer limit to useful weather forecasts.The problem Lorenz found later became famous as 'chaos*. When a computer model begins to run, it has an incomplete view of the current weather, because it's starting with observations that may be separated by 160km/100 miles. Through a series of computer experiments, Lorenz found that small weather systems tucked out of view between the model points – as well as errors in the observations themselves – can grow with time as the model unfolds. Within a few days, the errors become large enough to jeopardize the entire forecast. To quote Lorenz's apt metaphor, which he used as the title of a talk,'Does a flap of a butterfly's wings in Brazil set off a tornado in Texas?* (His answer: perhaps. But so could all the wings of all the other butterflies – and, moreover, the flapping might actually prevent a tornado just as easily as it enables one to form.) One way forecasters get around the problem of chaos is to run a number of different models; the US runs several big ones at least four times a day. If all goes well, the errors will vary, while the bona fide weather features will show up in most or all of the models. Scientists may also introduce error on purpose. By running the same model twenty times or more, with tiny differences in the starting points each time,they can seewhich weather features a re robust enough to appear in all of the variations. This approach, called ensemble modelling, is quickly becoming more feasible as computer power increases, and it's also being used in climate modelling. Despite all of these tricks, the current thinking in the forecast world is that Lorenz was indeed right. Butterflies, and all the other things we can't specify, ensure that we will never exactly know what weather to expect at 4pm a month from today.