Met Office: Statistical prediction of the winter NAO

	Statistical prediction of the winter NAO

Guidance from the statistical prediction method described here forms one input to our first assessment for Europe/UK winter, issued in July with updated forecasts issued from September. (Other inputs to the winter forecasts include potential influences from El Niño/La Niña, and changing expectations due to the climate warming trend.)

The North Atlantic Oscillation (NAO)

The North Atlantic Oscillation (NAO) is a phenomenon associated with winter fluctuations in temperatures, rainfall and storminess over much of Europe. When the NAO is in a positive phase westerly winds are stronger or more persistent, northern Europe tends to be warmer and wetter than average and southern Europe colder and drier. When the NAO is in a negative phase westerly winds are weaker or less persistent, northern Europe tends to be colder and drier and southern Europe warmer and wetter than average.

The phase and amplitude of the NAO is often described using an index. One of the simplest NAO indices is the winter difference in pressure at sea level between the Azores and Iceland. The index used here is somewhat different but it essentially captures the same phenomenon.

Winter 2008/9 NAO prediction

The figure below shows that the predicted winter NAO index for 2008/9 is weakly positive at +0.1 with a standard error of ±1.0. The small amplitude of the predicted index relative to the error bar means that the NAO prediction this year provides little signal for below- or above-normal European winter temperatures or precipitation. However, the prediction is consistent with a cooler, drier winter over northern Europe as a whole than experienced in winter 2007/8, when the observed index was +1.6.

Image: predicted winter NAO index for 2008/9 is weakly positive at +0.1 with a standard error of ±1.0

Details of the NAO statistical prediction technique

The aim of this statistical method is to use the slow variations and inertia of the Atlantic ocean to predict the atmospheric circulation. For this method to work, we require the ocean to have an impact on the overlying atmosphere. Scientific evidence for such a link from Atlantic sea-surface temperatures (SSTs) to the winter NAO was presented in Rodwell et al., Nature, 1999, 398, 320-323.

The prediction is based on a maximal covariance analysis of May North Atlantic SSTs and the following December to February 500 hPa geopotential heights. In essence, we look for a predictor pattern in May SSTs (left figure below) that, in the past, has tended to be associated with the following winter NAO. The NAO is defined here by the height pattern seen in the right figure below.

By taking the observed SST anomaly for May (figure below) and calculating how it projects onto the predictor pattern we can make a prediction for the winter NAO. If the projection is positive (i.e. the anomaly pattern looks similar to the predictor pattern) then the prediction is for a positive winter NAO. Conversely, if the observed May SST anomaly projects negatively onto the predictor pattern (i.e. it looks like the reverse of the predictor pattern) then we would predict a negative NAO. The NAO index used here is slightly different from the simple difference in surface pressures between the Azores and Iceland but the two time-series are quite similar with a correlation of 0.89.

We have estimated the skill of our predictions and it is reasonable but by no means high. The link from the Atlantic ocean to the NAO is not strong enough for any such prediction method to be very accurate. Hindcast tests, for the period 1948 to 1998, conducted in cross-validation mode (sometimes known as jack-knife mode) suggest that the correlation skill is about 0.45 with the correct sign of the NAO predicted for 66% of all winters. These figures are significant at the 99% and 98% confidence levels, respectively.

Further details of the prediction system are presented in Rodwell and Folland, Quarterly Journal of the Royal Meteorological Society, 2002, 128, 1413-1443.