Designment Of The MM5 Genetic Algorithm Assimilation System And Its Application

The technique of variational data assimilation has been considered as aneffective method in the aspect of improving the quality of the initial fields ofnumerical weather predication (NWP). Genetic algorithm (GA), as a rising theory,which has many characteristic, just solved the problem of adjoint variational dataassimilation that it is easy to fall into the trap of local optimization.MM5 GA assimilation system is established in this paper, of which base oftheoretic and the steps of achievement are introduced. We take a heavy rainfallprocess during 09-10, July, 2005 as the experimental to compare the assimilationefficiency and simulative with GA assimilation system and adjoint modelassimilation system. The results suggested that: both GA assimilation system andadjoint model assimilation system can effectively improve the initial field ofnumerical prediction and somewhat enhanced the prediction effect of physical fieldsand rainfall. The experiment showed it slightly surpasses follows adjoint modelassimilation system. Parallel algorithm showed GA assimilation system has higherefficiency than adjoint model assimilation system. Then based on the results of thelast experience, the system was adjusted and added the cloud-drift wind into it, witha case of heavy rainfall taking place from 00h July 2 to 00h July 3, 2005, severalnumerical simulation experiments of different schemes are performed. The resultsare as follows. The GA assimilation system similarly has the assimilation ability tothe non-conventional data. The cloud-drift wind can obviously improved the initialwind field and temperature field to obtain a more superior initial filed and enhancesthe forecast of the model. Compared the GA assimilation system added thecloud-drift wind with adjoint model assimilation system in this experience, theformer has slightly better prediction in the precipitation field, physics field andanalytic fields of rainstorm's formation.