An improved method for the threedimensional variational assimilation of the radar seeable velocity and its numerical experiments

With the terminal velocity of raindrops, this paper improves a variational method that directly assimilates wind field information, including the radial velocity and the movement of radar echo (seeable velocity), as proposed in Wan et al (2005). According to the theoretical derivation and scale analysis, the terminal velocity of raindrops of a mesoscale system has a relatively large order of magnitude, so this term is an important role in the equation of radar echo conservation. The improved method - using the terminal velocity of raindrops - avoids the interference of local growth of cloud clusters, and also the advection of radar echo is to be more accurately expressed by the improved seeable velocity. The analysis of the observational radar dataset of ‘Fengshen’ demonstrates that the improved seeable velocity is regularly and reasonably distributed and distinctly shows the mesoscale convergence and divergence of the typhoon. With the application of the GRAPES3DVar system developed by the China Meteorological Administration (CMA), a group of observational system simulation experiments (OSSE) were operated. The results of improved method show that the increments of both the horizontal wind and the geopotential height assume a dipole flow pattern which is concord with the difference between simulated observations and free integrations, and the vertical variation of the meridional wind is also better. The structure of horizontal wind fields retrieved by the improved seeable velocity and radial velocity is more accurate. With the original method, the results of the OSSE show no these characteristics. With the observational radar data assimilated by the improved method, we conducted another group experiments. The initial field of sensitive experiment shows that the typhoon centre was adjusted and more contiguous with the observation and the horizontal wind fields had more obvious cyclonic rotation. Moreover, the distribution of 6hourly accumulative precipitation of the sensitive experiment was more similar to the observation. The maximum rainfall intensity was less, but it was superior to the control experiment.