Information theory and dynamical system predictability

Predicting the future state of a turbulent dynamical system such as the atmosphere has been recognized for several decades to be an essentially statistical undertaking. Uncertainties from a variety of sources are magnified by dynamical mechanisms and given sufficient time, compromise any prediction. In the last decade or so this process of uncertainty evolution has been studied using a variety of tools from information theory. These provide both a conceptually general view of the problem as well as a way of probing its non-linearity. Here we review these advances from both a theoretical and practical perspective. Connections with other theoretical areas such as statistical mechanics are emphasized. The importance of obtaining practical results for prediction also guides the development presented.