We will review ideas of weather predictability, with the Lorenz (1963) discovery of chaos (a word coined by Jim Yorke in 1975). Lorenz concluded that even if we had perfect models of the atmosphere and perfect knowledge of the initial conditions ("except for a butterfly"), we could not predict the weather beyond two weeks. At that time, weather forecasts were useless beyond a day, so this limit on predictability was only of academic interest.

A widely used example of a chaotic system is the Lorenz (1963) 3- variable model with a solution that switches chaotically between two regimes. Four undergraduate women interns were able to use a simple method known as "breeding fast growing perturbations" to develop accurate rules forecasting when the current regime would end and how long the next regime would last. These same ideas, showing that even in complex systems chaos has simple structures, are applied to models used for weather forecasting. Chaos is "attacked" using ensemble forecasts, extending the skill of weather forecasts to 10 days, and approaching the theoretical limit of predictability. The El Niño phenomenon, being due to instabilities in the coupled ocean- atmosphere, makes possible predictions of mean weather conditions several months in advance.

--------------------------------------------------------------------------------------
For more information, contact:

Adil Hassam
hassam@umd.edu
301-405- 1417

Arthur LaPorta
alaporta@umd.edu
301-405-3291