Figure 2: Schematic chart of the procedure of ensemble experiment
The ensemble experiment is carried out as follows. Starting from an isothermal state at rest, the model is integrated for 10700 days with the zonally uniform SST distribution where the warm SST area is removed ( Figure 1). The initial conditions for the ensemble runs are the set of 200 data taken every 50 days from day 700 of the above time series. Each run is integrated for 50 days with the warm SST area as the surface condition.
The reason for choosing 50 days as the sampling interval of the initial data is that the memory of individual convective activity is considered to remain for the period much shorter than 50 days. It is also important that the period of intraseasonal variability in the model without the warm SST area is about 36 days, which does not have a large common divisor with the sampling interval. The phase of the intraseasonal variability is regarded to be distributed randomly among the initial conditions.