Much as we might mock our weather forecasts, meteorologists are now pretty darn accurate. Today’s three-day weather forecasts are on average more reliable than a one-day forecast was in the 1980s, and even a seven-day forecast is respectable. So will we be tuning in to the 17-day forecast by 2050?
To try to answer this question, Falko Judt, from the National Center for Atmospheric Research in Boulder, Colorado, used a supercomputer to run two nearly identical weather simulation models over a 20-day period. The first began with the actual weather observations from 19 October 2012; the second model tweaked the starting conditions by a tiny fraction – about one-thousandth of a degree.
The findings, which were published in the Journal of Atmospheric Sciences, found that the two simulations produced very similar forecasts for the first six days, but thereafter they began to diverge until after 17 days they had nothing in common. Multiple small scale events, such as thunderstorms, drove regional atmospheric patterns off in random directions after a week or so. This suggests that with perfect weather observations forecasters can produce near-perfect six- to seven-day forecasts, but forecasting accurately beyond this isn’t likely to be realistic any time soon.
