An ENSO prediction approach based on ocean conditions and ocean–atmosphere coupling

A simple statistical modelfor the El Nino–Southern Oscillation(ENSO) prediction is derived based on the evolution of the ocean heat condition and the oceanic Kelvin wavepropagation associated with westerly wind events (WWEs) and easterly wind surges (EWSs) in the tropical Pacific. The multivariate linear regression model solely relies on the pentad thermoclinedepth anomaly evolution in 25 days along with the zonal surface wind modulation. It successfully hindcastsall ENSOs except for the 2000/01 La Nina, using the pentad (or monthly) mean tropical atmosphere ocean array data since 1994 with an averaged skill (measured by anomaly correlation) of 0.62 (or 0.67) with a 6-month lead. The exception is mainly due to the long-lasting cold sea surface temperature anomalies in the subtropics resulting from the strong 1998/99 La Nina, even though the tropical warm water volume (WWV) had rebounded and turned phases after 2000. We also note that the hindcastskill is comparable using pentad or monthly mean NCEP global ocean data assimilation system data for the same time period. The hindcastskill of the proposed statistical modelis better than that based on the WWV index in terms of the monthly correlation, normalized RMSEs and ENSO occurrences, which suggest that including the evolution of the subsurface ocean temperature anomaly and the WWEs/EWSs in the central tropical Pacific can enhance the ability to predict ENSO. The hindcastskill is also comparable to the predictions using other dynamical and statistical models, indicating that these processes are the keys to ENSO development. The dynamics behind the statistical modelare consistent with the physical processes of ENSO development as follows: the tropical WWV resulting from the interannually-varying meridional subtropical cell transport provides a sufficient heat source. When the seasonal phase lock of ocean–atmosphere coupling triggers the positive (negative) zonal wind anomaly in boreal summer and fall, an El Nino (a La Nina) will develop as evidenced by the Kelvin wavepropagation. The triggering dynamic may be suppressed or enhanced by the influence of extratropical Pacific sea surface temperature.