Identification of the Madden–Julian Oscillation With Data‐Driven Koopman Spectral Analysis

Abstract The Madden‐Julian Oscillation (MJO), the dominant mode of tropical intraseasonal variability, is commonly identified using realtime multivariate MJO (RMM) index based on joint empirical orthogonal function (EOF) analysis near‐equatorial upper and lower level zonal winds outgoing longwave radiation. Here, in place conventional EOFs, we apply an operator‐theoretic formalism dynamic systems theory (the Koopman operator) to extract analog RMM that exhibits certain features refine characterization predictability MJO. In particular, spectrum operator eigenfunctions, with eigenvalues corresponding periods, contains a leading period ∼50 days. Moreover, amplitude this eigenfunction seasonal modulation clearly peaked boreal winter. Finally, phase space formed by complex smoother temporal evolution higher degree autocorrelation than RMM, which may contribute enhanced predictive skill.