Abstract: Two comparison experiments, i.e.NOAMSUA and AMSUA, were performed for simulating a very heavy rainstorm occurring in Wuhan City and the region to its east in 21-22 July 1998.In Exp.NOAMSUA, radiosonde observations were assimilated while in Exp.AMSU-A radiosonde observations and brightness temperatures of NOAA-15 AMSU-A channels 4-14 were assimilated simult aneously.Then the assimilation analyses of the both comparison experiments were taken respectively as initial conditions of numerical prediction model for simulating that process.In this paper, an incremental three-dimensional variational(3D-Var) method was used as data assimilation method.The MM5 mesoscale model(version 3) act ed as assimilat ion model and predict ion model as well.The analysis increments of Exp.A MSU A w ere compared with that of Exp.NOAMSUA, so did the forecast results.The comparisons in terms of the analysis increments showed that direct assimilating the AMSU-A bright ness temperatures adjusted the temperature backg round of the mid and upper troposphere more distinctly.The result is attributed to the temperature information of the mid and upper layers in Exp.AMSUA being enriched by the addition of t he AMSU-A channels 4-14 measurements, which could only detect temperature above 700 hPa except channe l4 detecting surface at mospheric temperature.The simulation results showed that the forecasts of temperature and mixing ratio of water vapor were generally better in Exp.AMSUA than in Exp.NOAMSUA.It was also found that in each experiment, the simulation of the major influencing weather systems in the heavy rainstorm process, the low-level south west vortex and the cold shear was preferable, and the predicted areas and grades of heavy rains somew hat fitted to the real time rainf alls.But all the precipitation forecasts of Exp.AMSUA seem not superior to that of Exp.NOAMSUA, which may also result from few information on humidity being contained in the AMSU-A measurements.