This research project will explore the utility of adjoint-based methods of ensemble prediction for El Nino Southern Oscillation (ENSO), with the view to support the development of a new operational ensemble prediction scheme for ENSO. The PIs will use a hierarchy of coupled ENSO models that represent different levels of approximation of coupled ocean-atmosphere General Circulation Models (GCM). In addition to exploring singular vector and stochastic optimal based ensemble prediction techniques, they will compare these techniques with the method of bred modes that is currently used for ensemble weather prediction at the National Centers for Environmental Prediction. The ensemble prediction experiments will be analyzed using conventional statistical techniques and using a new theoretical framework based on information theory. The research builds on existing methods and ideas that have been applied successfully to Numerical Weather Prediction on timescales of 5-10 days using complex state-of-the-art atmospheric GCMs. The techniques have been applied successfully to ENSO prediction using simple coupled models, however, the application of existing ideas to complex ENSO prediction presents some significant new challenges that need to be addressed. The work is important because it has the potential to improve ENSO forecasts, which has important societal benefits.
