Climatically Relevant Singular Vectors for Coupled General Circulation Models

Singular vectors have proved a powerful tool in prediction and data assimilation studies of the atmosphere and climate systems. Such patterns are (linearly) optimal in exciting a response in a given dynamical system. In recent years the tool of choice in climate (ENSO) prediction and variability studies has become the coupled general circulation model (CGCM) which offers a relatively complete physical description of climate. Calculation of singular vectors for such models is a problematic undertaking however since the fastest growing modes are inevitably connected with weather rather than climate phenomenon. Here we present a novel method for overcoming this difficulty which allows us to calculate "climatically relevant" singular vectors. We test the method in a stochastically forced intermediate ENSO coupled model for which "climatic" singular vectors are known exactly. We then apply our methodology to the NASA CGCM and demonstrate convergence and robustness of our results.