The forward model in this system has the same architecture as the forecast model (viz. MSM) except that its horizontal resolution is reduced to 20km. The adjoint model has the same dynamical process as the forward model but its physical processes are simpler ones.
The background error covariance matrix is simplified by assuming homogeneity and Gaussian type. In order to greatly reduce the computational time of matrix calculation, trivial variance terms are skipped in the calculation without losing accuracy.
In order to properly assimilate precipitation data, the observational cost function is properly formulated, because the error distribution of precipitation amount cannot be considered as Gaussian. The performance of precipitation forecast is improved by assimilating precipitation amount with this system.
JMA is planning to install a 25 station operational wind-profiler network which covers all over Japan with approximately 100km intervals for the purpose of precise observation of lower atmosphere (approximately up to 5km). The impact of the assimilation of these wind-profiler data on prediction of heavy rainfall events will be demonstrated.