The role of the ocean boundary conditions for seasonal predictability

Dynamical seasonal predictions are becoming a useful tool for predicting climate variations. Despite the intrinsic predictability limit of atmospheric flow, these forecasts show often substantial skill. Now it becomes critical to better understand the controlling factors for short-term climate predictions. Slowly evolving sea surface temperatures (SST) are known to be key to seasonal climate predictions, but atmospheric initial conditions may play a role as well.

The goal of this study is to determine the importance of the state of the ocean surface in producing seasonal forecast skill. To this end, a suite of 10-member ensemble climate simulations has been performed with the NCEP reanalysis-2 atmospheric general circulation model. Over the 50-year period from 1950-1999, we have forecasted the state of the winter atmosphere from mid-December to the end of March of the following year. Each experiment has been forced with different combinations of initial and boundary conditions. This enabled us to find out how much predictability is related to the specification of SST boundary conditions, and how much is due to the specification of the initial state. We also derived the typical time scales that are associated with boundary and initial conditions. With this we can show for how long time the initial conditions influence the forecast result, and how much time it takes until the ocean surface exerts a substantial influence on the overlying atmosphere. Furthermore, we will discuss how precise SST boundary conditions must be specified for a good seasonal climate forecast, and what the differences are between climatological, persistent, and predicted SSTs.