Relationships between Optimal Precursors Triggering NAO Onset and Optimally Growing Initial Errors during NAO Prediction

Abstract

Abstract Based on a viewpoint that the North Atlantic Oscillation (NAO) is a nonlinear initial-value problem, the predictability of NAO event onset is studied through investigation of the relationship between the optimal precursor (OPR) to its onset and the optimally growing initial error (OGE) in onset prediction. The problem is explored by the method of conditional nonlinear optimal perturbation with a triangular T21, three-level, quasigeostrophic global spectral model. For the NAO onset, there are two types of OGEs. Numerical results show that, with the optimization time of 3 days, a type-1 OGE bears a great resemblance to OPR, and the similarity coefficient between them is 0.98 for both positive (NAO+) and negative NAO (NAO−). A type-2 OGE is also characterized by a similar pattern to OPR, but with an opposite sign. With the extension of the optimization time to 7 days, the similarity coefficient between OPR and type-1 (type 2) OGE gradually decreases to 0.82 (−0.81) for NAO− and 0.87 (−0.57) for NAO+. However, in the linear regime, such high similarity between OPR and OGE can only be found with an optimization time of 3 days. Further analysis reveals that a type-1 (type 2) OGE has a similar growth behavior to that of the corresponding OPR of the same-phase (opposite phase) NAO event, both of which develop into a dipole NAO anomaly pattern. This similarity between OPR and OGE suggests that the nonlinear process plays an important role in the NAO event, which simultaneously provides a theoretical foundation for its targeted observations.