NAO Forecasts

If the NAO is not completely random (white noise), then it may be possible to forecast it with some measurable skill. The NAO has marked variability on the quasi-biennial time scales and this might enable some seasonal forecasts to be made. However, it should not be forgotten that the mid-latitudes contain a large amount of high frequency weather disturbances and these may cause skills to deteriorate. The NAO has recently shown some strong decadal trend behaviour and it may be possible that coupled ocean-atmosphere models could emulate these trends. Here are some online forecasts of the North Atlantic region and the NAO:

Because of its dominant impact on the weather and climate of Europe, there is a growing interest in quantifying the possible limits of seasonal and interannual predictability of the North Atlantic oscillation. Unfortunately, however, the NAO is a noisy mid-latitude phenomenon with an almost flat power spectrum close to that of white noise and so even the best possible linear predictions are not able to explain more than about 10\% of the total variance \footnote{\, the variance of the hindcast residuals is 90\% of the variance of the original NAO series.} of wintertime means from 1864-1996 (Wunsch 1999). For example, a simple fractionally integrated noise model can be used to to make one-year ahead forecasts that have a small yet significant correlation ($r=0.17$) with the observed NAO index for the period 1864-1998 (Stephenson et al. 2000). Rodwell et al. (1999) obtained a higher correlation of $r=0.41$ for the ensemble mean NAO simulated by an ensemble of 6 atmosphere-only runs over the period 1947-97. However, it should be noted that these simulations are not useful forecasts since they rely upon knowing a priori the sea surface temperatures. Furthermore, much of the apparent skill in these simulations is primarily due to longer term decadal variations that can be duplicated with a simple 2-variable linear coupled model (Bretherton and Battisti 2000). Since more than 75\% of the variance of the NAO resides in shorter than decadal time scales, the real challenge in NAO forecasting is to be able to predict the shorter term variations.

UKMO ensemble forecasts made as part of the PROVOST project

• Hurrell, J.W., 1995: Decadal trends in the North Atlantic Oscillation and relationships to regional temperature and precipitation. Science 269, 676-679.