A series of small-perturbation experiments has been conducted to demonstrate that an atmosphere-ocean coupled model and an atmosphere-only model produce significantly different intensities of intraseasonal oscillation (ISO) and phase relationships between convection and underlying SST associated with ISO. When exact the same initial conditions and SST from the coupled model are used to force the atmosphere-only model, both of them produce identical ISO solutions. Once a small amount of noise is introduced into the initial conditions and SST, the atmosphere-only model generates a considerable different ISO solution from the coupled model though the climatologies between them are almost the same.

The coupled model not only simulates a stronger ISO than the atmosphere-only model, but also generates a realistic phase relationship between intraseasonal convection and underlying SST. In the coupled model, positive (negative) SST fluctuations are highly correlated with more (less) precipitation with a time lead of 10 days as in the observations, suggesting that intraseasonal SST is a result of atmospheric convection, but at the same time, positively feeds back to increase the intensity of the convection. In the atmosphere-only model, however, SST is only a boundary forcing for the atmosphere. The intraseasonal convection in the atmosphere-only model actually is less correlated with underlying SST. The maximum correlation between convection and SST occurs when they are in phase with each other, which is in contrast to the observations. These results indicate that an atmosphere-ocean coupled model produces a more realistic ISO compared to an atmosphere-only model.