A Study Of Application Of Cloud Drift Wind Data Assimilation In MM5 Adjoint-model Assimilation System

The accuracy of modern numerical prediction mainly depends on the exact model that can simulate the synoptic course and the initial field that can describe the initial time. Recently the development of numerical models has run to perfection, so whether the initial field stands to exactness counts. The optimal initial field not only meets the dynamical and thermal expectations but also reflects the true atmospheric state. To obtain the optimal initial field, a variety of conventional and non-conventional observations of multi-time levels as well as the method of utilizing effectively are needed. In this context conventional and non-conventional observations (CDW) are assimilated by MM5 adjoint-model assimilation system so as to do research on the quality of the assimilation system and the role of CDW in improving the prediction.In this text the principal of constructing MM5 adjoint-model assimilation system is introduced amply. Furthermore data assimilation experiments are conducted with MM5 Adjoint-model Assimilation System made by adjoint codes. With a case of heavy rainfall taking place from 00h July 23 to 00h July 24,2002, several numerical simulation experiments of different schemes are performed. The results are as follows. MM5 Adjoint-model Assimilation System not only can improve the initial field effectively and promote the coordination with the model but also can enhance the forecast on the precipitation and other elements. The assimilation of CDW has an improvement on quality of upper wind. The effect of direct numerical simulation with utilizing the CDW to amend the initial field gains the advantage over the one not. When using the adjoint-model assimilation system, the experiment adding CDW can directly improve the wind forecast, but comparing the one not adding CDW it has little advantage over other elements forecast. The result of sensitivity experiment is that adding the proportion of the non-conventional observation properly plays a certain role in improving the numerical prediction. But further efforts are needed to realize it.