A Study On Assimilation Of AMSU And Its Application On Heavy Rainfall In Numerical Forecast

AMSU data is assimilated into the mesoscale model MM5 based on GRAPeS-3D-Var system, and its impact on rainfall numerical modeling is studied by experiments also. Conclusions drawn from the practice of a heavy rainfall process during 04-05, July, 2003 are: 1) results from the assimilation tests show that AMSU-A data contributes mainly to adjusting temperature, positive temperature increment companies with negative humidity increment, and verso also; AMSU-B data contributes mainly to adjusting humidity, positive humidity increment companies with negative temperature increment, and verso also. 2) Results from forecasts experiments show that both the rainfall location and intensity modeled are improved when AMSU-A/B data assimilated to initial condition fields. AMSU-A data has a positive effect on the intensity of rainfall modeling, and AMSU-B data improves the modeling of the location of rainfall. 3) This paper identifies the sensitivity location and variables of 6h forecast errors by the method of adjiont sensitivity analysis also, and they locate west and south west of the real rainfall field. The forecast skill may be improved if low-level observation data is of high quality. 4) The experiments of 4D-Var and forecasts show that, the forecast skill of rain and other model variables are improved when the retrieval fields of AMSU are assimilated into the initial condition by MM5 4D-Var system. The experiments in this paper show that AMSU data may have great latency in rainfall forecasts in summer in our country.