QBO influence on extratropical predictive skill

Abstract

The quasi-biennial oscillation (QBO) in the zonal wind in the tropical stratosphere is one of the most predictable aspects of the circulation anywhere in the atmosphere and can be accurately forecast for many months in advance. If the stratospheric QBO systematically (and significantly) affects the tropospheric circulation, it potentially provides a predictable signal useful for seasonal forecasting. The stratospheric QBO itself is generally not well represented in current numerical models, however, including those used for seasonal prediction and this potential may not be exploited by current numerical-model based forecast systems. The purpose of the present study is to ascertain if a knowledge of the state of the QBO can contribute to extratropical boreal winter seasonal forecast skill and, if so, to motivate further research in this area. The investigation is in the context of the second Historical Forecasting Project (HFP2), a state-of-the-art multimodel two-tier ensemble seasonal forecasting system. The first tier, consisting of a prediction of sea surface temperature anomalies (SSTAs), is followed by the second tier which is a prediction of the state of the atmosphere and surface using an AGCM initialized from atmospheric analyses and using the predicted SSTs as boundary conditions. The HFP2 forecasts are successful in capturing the extratropical effects of sea surface temperature anomalies in the equatorial Pacific to the extent that a linear statistical correction based on the NINO3.4 index does not provide additional extratropical skill. By contrast, knowledge of the state of the stratospheric QBO can be used statistically to add extratropical skill centred in the region of the North Atlantic Oscillation. Although the additional skill is modest, the result supports the contention that taking account of the QBO could improve extratropical seasonal forecasting skill. This might be done statistically after the fact, by forcing the QBO state into the forecast model as it runs or, preferably, by using models which correctly represent the physical processes and behaviour of the QBO.