The Madden–Julian oscillation in ECHAM4 coupled and uncoupled general circulation models

Abstract

The Madden-Julian oscillation (MJO) dominates tropical variability on timescales of 30–70 days. During the boreal winter/spring, it is manifested as an eastward propagating disturbance, with a strong convective signature over the eastern hemisphere. The space–time structure of the MJO is analyzed using simulations with the ECHAM4 atmospheric general circulation model run with observed monthly mean sea-surface temperatures (SSTs), and coupled to three different ocean models. The coherence of the eastward propagation of MJO convection is sensitive to the ocean model to which ECHAM4 is coupled. For ECHAM4/OPYC and ECHO-G, models for which ~100 years of daily data is available, Monte Carlo sampling indicates that their metrics of eastward propagation are different at the 1% significance level. The flux-adjusted coupled simulations, ECHAM4/OPYC and ECHO-G, maintain a more realistic mean-state, and have a more realistic MJO simulation than the nonadjusted scale interaction experiment (SINTEX) coupled runs. The SINTEX model exhibits a cold bias in Indian Ocean and tropical West Pacific Ocean sea-surface temperature of ~0.5°C. This cold bias affects the distribution of time-mean convection over the tropical eastern hemisphere. Furthermore, the eastward propagation of MJO convection in this model is not as coherent as in the two models that used flux adjustment or when compared to an integration of ECHAM4 with prescribed observed SST. This result suggests that simulating a realistic basic state is at least as important as air–sea interaction for organizing the MJO. While all of the coupled models simulate the warm (cold) SST anomalies that precede (succeed) the MJO convection, the interaction of the components of the net surface heat flux that lead to these anomalies are different over the Indian Ocean. The ECHAM4/OPYC model in which the atmospheric model is run at a horizontal resolution of T42, has eastward propagating zonal wind anomalies and latent heat flux anomalies. However, the integrations with ECHO-G and SINTEX, which used T30 atmospheres, produce westward propagation of the latent heat flux anomalies, contrary to reanalysis. It is suggested that the differing ability of the models to represent the near-surface westerlies over the Indian Ocean is related to the different horizontal resolutions of the atmospheric model employed.