For the purpose of examining the initial development of the atmospheric response to a warm SST anomaly placed at the equator, an ensemble switch-on experiment is conducted with an aqua planet GCM. An ensemble average of the size of 128 satisfactory reduces the transient noise caused by both small scale convective activity and large scale intraseasonal variability.

A convection center develops on the warm SST area in the first few days. As a barotropic response to the heating, pressure increases globally outside the low pressure region around the warm SST area. The response after this widespread pressure rise is consistent with Gill (1980); a warm Kelvin wave-like anomaly is emitted to the east of the convection center, and a warm Rossby wave-like anomaly is emitted to the west. The Kelvin wave-like signal propagates at a speed slower than that expected from its vertical wavelength, suggesting that the signal is a ``moist'' Kelvin wave. A transient decrease of precipitation occurs at the front of the signal, which is consistent with its slow propagation.

After several days, precipitation to the east of the warm SST area begins to recover because of the surface frictional convergence into the equatorial low pressure anomaly associated with the Kelvin wave-like structure. On the other hand, precipitation to the west of the warm SST area continues to decrease. It seems that the reduction of precipitation in the western area is caused by the surface frictional divergence out of the relatively high pressure anomaly associated with the Rossby wave structure.

The horizontal structure of the rainfall response settles down to that of the steady state response at around day 20. Afterward, the pressure in the tropical and subtropical latitudes rises slowly. This slow mass redistribution continues for several tens of days and finishes the whole development of the atmospheric response to the switch-on of the SST anomaly.