Biennial and Low-Frequency Components of El Niño/Southern Oscillation

Abstract

El Niño/Southern Oscillation (ENSO) is a coupled oscillation of sea surface temperatures (SSTs), winds, and air pressure in the eastern and central tropical Pacific, that repeats with quasi-regularity, every 2–7 years. Although the ENSO’s spectral peak is found at a 4–7-yr period, composite El Niño events, taken as the 84 months before and after the peak of each El Niño, show that the length of each event, and often the following La Niña if there is one, usually falls within a quasi-biennial (QB) range of around 18–42 months. We argue that the biennial range of ENSO events stems from the classical delayed oscillator dynamics, while the lower-frequency range is from interaction with the extratropics; these interactions also lead to much of ENSO’s irregularity. After applying an 18–42 month bandpass filter to historical monthly temperature record and comparing filtered temperature variance to that of the raw temperature anomalies, the tropical Pacific emerges as the major center of enhanced ratio of biennial-to-total variance. This suggests that ENSO might be primarily driven by processes in this frequency band, even if its spectral peak is at lower frequencies. Discriminating patterns that maximize the ratio of biennial-to-total variance of surface temperatures also point to ENSO as the primary and only significant mode, both when projected onto monthly and bandpass-filtered surface temperature and SST data. We also compare composites, power spectra, variance ratio maps and time series, and discriminating patterns from observations to some CMIP5 global climate models, many of which have ENSO be too regular, and/or attribute too much of ENSO’s variability to the QB timescale. Finally, to put these ideas in a dynamical perspective, we investigate a coupled model that includes biennial tropical dynamics augmented by extratropical feedbacks, which shows much more LF and decadal variability reminiscent of the observed ENSO behavior.