| description abstract | he variability of the North Atlantic Oscillation is a key aspect of Northern Hemisphere atmospheric circulation and has a profound impact upon the weather of the surrounding land masses. Recent success with dynamical forecasts predicting the winter NAO at lead times of a few months has the potential to deliver great socio-economic impacts. Here we find that a linear regression model can provide skillful predictions of the winter NAO based on a limited number of statistical predictors. Identified predictors include El-Niño, Arctic sea ice, Atlantic SSTs and tropical rainfall. These statistical models can show significant skill when used to make out-of-sample forecasts and we extend the method to produce probabilistic predictions of the winter NAO. The statistical hindcasts can achieve similar levels of skill to state-of the art dynamical forecast models, although out-of-sample predictions are less skillful, albeit over a small period. Forecasts over a longer out-of-sample period suggest there is true skill in the statistical models, comparable with that of dynamical forecasting models. They can be used both to help evaluate, and to offer insight into sources of predictability and limitations of, dynamical models. | |
