The data from a 9-day period of SOP-2 were analyzed at 6 h intervals using a fully automatic limited-area variational objective analysis system, with a primitive equations prognosis model providing guess fields for each analysis and thus maintaining time continuity between analyses. It is demonstrated that the scheme produced a set of analyses which showed a stable, consistent evolution of the synoptic systems throughout the period. This would not have been possible without manual intervention prior to FGGE.
Twenty-four hour prognoses based on these analyses showed, on average, equal or slightly greater skill than the equivalent operational prognoses which had access to a similar (although somewhat less timely) data base, as well as having manual input.
The entire experiment was repeated using updated rather than fixed boundary conditions during each 6 h prognosis, with the boundary tendencies being obtained from a hemispheric assimilation experiment. It is shown that this produced a small but significant improvement in the quality of the overall analysis system.