A method for estimating and assessing modes of interannual variability in coupled climate models

Abstract

The seasonal mean of a climate variable consists of: slow-external; slow-internal; and intraseasonal components. Using an analysis of variance-based method, the interannual variability of the seasonal mean from an ensemble of coupled atmosphere-ocean general circulation model (cgcm) realisations is separable into these three components. Eigenvalue decomposition is applied to the covariance matrices to obtain, for each component, the dominant modes of variability (eigenvectors) and their associated variance (eigenvalues) for the climate variable. Here, a method is described that assesses the modes of interannual variability in cgcms against those obtained from reanalysis data based on observations. A metric is defined based on the pattern correlation