The development of a three-dimensional variational data assimilation (3DVAR) system for the MM5 model has reached the optimization, tuning and forecast verification stage. This talk begin with an overview of the system, focussing on aspects which are unique to the MM5 implementation of 3DVAR. These aspects include the choice of control variables, the use of an extended analysis grid and the mix of observations used.

Results from tests of the system, which is to be released to the general user community in the next year or so, will be presented. As well as showing standard data assimilation diagnostics such as convergence rates and background and analysis fit to observations, forecast results from initial case studies will be shown. One such study is that of Typhoon Sam, a poorly forecast event which passed over Hong Kong on 22nd August 1999. Comparison of forecasts run from the 3DVAR and little_r (which 3DVAR is designed to replace) will be made. Finally, possible future developments to the MM5 3DVAR system will be briefly described.