资源环境科技发展态势分析平台

Recent studies of individual seasonal forecast systems have shown that the wintertime North Atlantic Oscillation (NAO) can be skillfully forecast. However, it has also been suggested that these skillful forecasts tend to be underconfident, meaning that there is too high a proportion of unpredictable noise in the forecasts. We assess the skill and overconfidence/underconfidence of the seasonal forecast systems contributing to the EUROpean Seasonal to Interannual Prediction (EUROSIP) multimodel ensemble system. Five of the seven systems studied have significant skill for forecasting the wintertime NAO at 2- to 4-month lead times. Four of these skillful systems are underconfident for forecasting the NAO. A multimodel ensemble (ensemble size 126 members) is both skillful and clearly underconfident. Underconfidence becomes more pronounced as the ensemble size increases. Certain years in the hindcast period are well forecast by all or most models. This implies that common teleconnections and drivers of the NAO are being captured by the EUROSIP seasonal forecasts.

Plain Language Summary In this paper we provide an intercomparison of seven seasonal forecast systems, with particular focus on the wintertime North Atlantic Oscillation (NAO). The wintertime NAO is the main driver of winter weather variability in the United Kingdom and Europe, and being able to forecast the NAO for the season ahead has potential benefits for many different sectors such as agriculture, energy, health, transport, and water resource management. We show that five of the seven systems studied can skillfully forecast the NAO, and a multimodel ensemble has even higher skill. Four of these skillful systems are found to be underconfident, which means that there is too high a proportion of unpredictable noise in the model. Being underconfident makes it more difficult to fully utilize the skill of a forecast. However, one system is skillful but not underconfident. We also find that there are common years in which the NAO is well forecast by all the skillful systems. This is an important result because it implies that common drivers of NAO predictability are being captured by these systems. These results are an important contribution to our understanding of seasonal forecasts systems and the predictability of the NAO.