This paper derives an approach to systematic correction of errors in the mean dynamic topography (MDT) used for altimeter data assimilation in ocean circulation models. An a priori error covariance is used to help identify bias in the MDT. The method is then demonstrated in the operational Forecasting Ocean Assimilation Model (FOAM) from the U.K. Met Office. Two independent MDTs are compared: the original MDT currently used in the model and a new product of the European project Gravity and Ocean Circulation in the North Atlantic (GOCINA) which has an a priori error covariance field associated with it. A third MDT is then derived by estimating and correcting the systematic error in the GOCINA MDT using off-line bias calculations. The results are compared using diagrams that split the traditional root-mean-square error (RMSE) in the model sea level into two independent measures, a systematic error and a variable or dynamical error. The dynamical error is a measure of the capacity of the model to represent the temporal variations of the observations, in this case the time-varying altimetry. It is shown that the dynamic sea level error can be improved by the choice of MDT despite the fact that it is formally independent. In this way it is demonstrated that the bias correction method derived for the MDT also leads to improvement in the forecasting of altimetric sea level anomalies in the FOAM model.