Stochastic modeling of climate variability

In this talk stochastic models of the climate system and its various component subsystems will be reviewed. Stochastic modeling has come a long way since the early studies of Leith (1975) and Hasselmann (1976). In particular, stochastically driven linear models are now acknowledged to capture a large part of ENSO dynamics, tropical-extratropical interactions, extratropical storm-track dynamics, and extratropical air-sea interaction. As such they provide valuable insight into the dynamics of these phenomena. Stochastically driven linear models have also been shown to be competitive with much more comprehensive nonlinear models (up to and including global coupled general circulation models) in predicting these phenomena. The talk will highlight some important theoretical and observational developments leading to this progress. It will also be shown how the dynamic of the earth’s nonlinear high-order chaotic climate system asymptotically approaches that of a stochastically driven linear system on longer than synoptic time scales. This general tendency not only justifies the linear approximation but also implies fundamental upper bounds on climate predictability. An important emerging area of research, that of implementing stochastic representations of unparameterized processes in GCMs and gauging their impact on climate variability and drifts, will also be discussed briefly.