At each step forward the model "feels" both the specified incoming solar radiation, and the SST at its lower boundary. The incoming radiation is calculated according to the time of year - calculating the orientation of the earth to the sun. This, together with the evolving SST, will drive the model's "annual cycle" (e.g. the development of winter and summer in the model).

How does the model "feel" the SST at its lower boundary? Based on the prevailing SST, equations will specify the amount of heat and moisture that enter the lowest layer of the atmosphere at each time step. The received heat and moisture will then permeate through the model's atmosphere in the coming time steps, allowing the processes described in the first part of this lecture to influence the seasonal climate in the model. The seasonal climate simulated by the model (that is, the average weather over the 12,960 time steps) can be considered as a prediction of the seasonal climate given the prevailing SST forcing. Next, the model is run a second time, with a different SST pattern prescribed at its lower boundary. The new seasonal climate simulated by the model can be considered as a prediction of the climate given this new prevailing SST.

Even in the tropics, there will be a battle between SST controlling the seasonal mean climate, versus internal processes to the atmosphere (or other factors such as land- atmosphere interaction) that can also generate climate variability. One example is the Madden-Julian Oscillation. This is a 30-60 day oscillation that, once established, moves eastward around the tropics enhancing convection and precipitation as it passes through each location. In a given season, climate anomalies associated with such features may dominate, such that the impact of the prevailing SST anomalies is weak. In such situations, we say that the potential predictability of the seasonal climate anomaly, based on the SST is weak. We can make GCM experiments to study the evolution of this battle in past seasons. It allows us to study the impact of SST on seasonal atmospheric circulation anomalies, and to make some estimates of the predictability of seasonal climate anomalies from SST.

GCM experiments of the following type have now been widely made to study the predictability of the seasonal climate. For the example to be considered here, datasets of the observed SST in each month 1949-99 were created based on observations including those made by merchant ships and satellites. A GCM was started from arbitrary weather pattern typical of January, with the January 1949 SST field at the model's lower boundary. The model stepped forward, updating the SST as observed each month, until December 30 1999 was arrived at. With a ten minute time step for the model calculations, that makes a total of 2,643,840 time steps.