A multivariate statistical model for forecasting anomalies of half‐monthly mean surface pressure

A multivariate statistical model for forecasting anomalies of surface pressure over Europe and the North Atlantic Ocean at medium to long range used routinely in the Synoptic Climatology Branch of the Meteorological Office is described. The model discriminates between broad classes of circulation type using a multivariate predictor space (including areal sea surface temperature anomalies). The forecasts may take categorical or probabilistic form, and the likelihood of the occurrence of categories of circulation index, temperature and rainfall over the U.K. can be produced using Bayesian methods. Emphasis is laid on the difficulties of applying classical statistical analysis to atmospheric fields, as this important aspect of long range forecasting by empirical methods tends to be neglected in the literature.