Characterization of the long-time and short-time predictability of low-order models of the atmosphere

Abstract

Methods to quantify predictability properties of atmospheric flows are proposed. The ``Extended Self Similarity'' (ESS) technique, recently employed in turbulence data analysis, is used to characterize predictability properties at short and long times. We apply our methods to the low-order atmospheric model of Lorenz (1980). We also investigate how initialization procedures that eliminate gravity waves from the model dynamics influence predictability properties.