An OSSE study for Doppler radar data assimilation strategies using four-dimension variational scheme

A series of the observation system simulation experiments (OSSEs) are conducted to investigate the influence of the assimilation strategies on forecast performances. The strategies include different background fields, radar coverage and assimilation windows. The simulated radar reflectivity and radial velocity observations of a squall line are assimilated into a cloud model using four-dimensional variational data assimilation (4DVAR) scheme. The forecast performances are evaluated by computing the three-dimensional spatial correlation coefficient of the rainwater mixing ratio between the control simulation and OSSEs. It is found that the 4DVAR assimilation system can reconstruct the environment wind field through assimilating the radial wind over the entire domain to make forecast performance comparable to the experiment with perfect background field. The clear echo can provide the positive influence on forecast performances even with poor background field. The short term forecast performance is dominated by the dynamic of the mesoscale system. However, the environment field from radar clear echo and the other observations affects the long term forecast. The cost function is hard to be minimized in longer assimilation window compared to the shorter assimilation window under the same numbers of minimization iterations. Therefore, the forecast performance of the experiment with long assimilation window is worse than cycling experiment and even single short assimilation window.