Estimating Weather and Climate Predictability on Attractors

Abstract

Predictability is deduced from phase space trajectories of weather and climate variables which evolve on attractors (local surface pressure and a δ 18O-record). Predictability can be defined by the divergence of initially close pieces of trajectories and estimated by the cumulative distance distributions of expanding pairs of points on the single variable trajectory. The e-folding expansion rates characterize predictability tune scales. As a first estimate one obtains a predictability time scale of about two weeks for the weather variable and 10?15 thousand years for the climate variable.