In the paper, using a nonhydrostatic version of the Pennsylvania State University-National Center for Atmospheric Research fifth-generation Mesoscale Model (MM5) and its adjoint model, two sets of experiments, control experiment and assimilation experiment, for the Mei-Yu front rainstorm of 23～24 June 1999 were performed. The effects of rainstorm forecast and the initial fields of different experiments were compared. It is showed that, integrating all kinds of different type and different time data into numerical model, Four-Dimensional Variational(4DVAR) Data Assimilation method can produce an optional initial field and improve greatly the Mei-Yu front rainstorm forecast. 