Abstract. This paper presents a statistical diagnostic to interpret the dynamics of nonlinear variability in models. It consists in applying the Multichannel Singular Spectrum Analysis to the variability of a suitable equation of a dynamical system. This method enables us to evaluate the sensitivity of the oscillatory features of a physical model to tendencies in its mathematical formulation. Hence, responses in amplitude and frequency are determined in the phase space of the model. The example of interannual variability of the tropical Pacific climate through the analysis the Sea Surface Temperature equation of a hybrid coupled model is used to illustrate and validate the methodology.