Conditional probability seasonal predictions of precipitation

Underlying, slowly evolving sea surface temperatures (SSTs) can influence atmospheric processes on seasonal time-scales raising hopes of producing successful seasonal forecasts using computer models. Unfortunately internal atmospheric instability and nonlinear dynamics of the atmosphere limit their skill even with perfect (observed) SSTs. An approach using analogues in an ensemble environment is shown to increase the skill of seasonal forecasts by using raw model ensemble forecasts to generate enlarged analogue ensembles consisting entirely of observations from other years. These new ensembles are a cheap and effective way of producing larger ensembles, which use readily available historical information about the real atmosphere which raw models do not. Conditional probability predictions from these new ensembles are shown to have higher skill than raw model forecasts in all seasons, in all parts of the world. They are also shown to be more skilful than climatology in certain seasons, also in all regions of the world. They can also be used to aid skill prediction.