The skill of dynamical long‐range forecasts, including the effect of sea surface temperature anomalies

The Meteorological Office 5‐level general circulation model has been used to investigate the skill of wintertime, long‐range (50‐day) forecasts. In spite of its low vertical resolution and relatively simple parametrization (by present day standards), its winter climatology is shown to be realistic not only in terms of the time‐mean flow but also in the behaviour of transients. In order to eliminate the less predictable synoptic scales, 15‐day means have been used to assess the skill at longer ranges. Even with climatological sea surface temperatures, the largest‐scale features of the flow are found on average to have a small but significant correlation with the real atmospheric evolution out to about 20 days. However, the skill is far from uniformly distributed between the 17 forecasts studied. Three forecasts exhibited some skill throughout the 50 days, while in five forecasts the anomaly correlation fell to zero within 10 days. The use of observed tropical sea surface temperatures improved the forecasts in four out of five years. Examination of the model response to tropical sea surface temperature anomalies shows that it is particularly sensitive to anomalies in the west Pacific and to a lesser extent the west Atlantic. In this region anomalies as small as 1 K or less, similar to the uncertainty in their measurement, can have marked mid‐latitude effects.