|
Future decadal climate forecasts are likely to rely on ensembles initialised using small perturbations to oceanic and atmospheric conditions. In order to design efficient ensembles there is a need to identify those perturbations that grow most rapidly. Such perturbations may also be useful to identify where new ocean observations could improve forecast skill. We have employed two different methods to estimate such optimal perturbations for decadal forecasts of the Atlantic Ocean in the HadCM3 GCM.
Firstly, we use linear inverse modelling (LIM) to find the initial condition anomalies which grow most rapidly under a particular norm of interest. The regions identified as most sensitive to small perturbations are located in the far North Atlantic. Significant non-normal amplification is found, and the mechanisms of amplification involve a basin-wide overturning circulation response to the small perturbations. We also demonstrate multi-decadal predictability of the overturning strength, and of basin-wide temperature and salinity fields. Secondly, we are using an ensemble based technique which, unlike the LIM approach, enables optimal perturbations to be estimated for specific initial conditions, e.g. a high or low overturning strength. The latest results will be discussed. |
|