The Multi-Scale/Block Batch-wise Variational Data Assimilation And Application In Forming Tropical Cyclone Initial Circulation

In the data assimilation, the most critical factor of controlling propagation characteristics about the observation information is the background error covariance (BEC), which is closely related with the flow pattern of weather system. The flow-dependent BEC greatly influences the quality of analysis. At present, homogeneous and isotropic BEC in the three-dimensional variational data assimilation(3DVar) is not consistent with the actual situation of tropical cyclone(TC). The four-dimensional variational data assimilation (4DVar) only implicitly develops BEC, but its inherent defects, such as model linearization and physical process parameterization, could result in unreasonable development of the BEC. The flow-dependent BEC is used in Ensemble Kalman filter (EnKF), but there are some issues, such as underestimate representation of ensemble members and filtering divergence, and that the more serious problems arises when collection members of TC forecast have a systematic deviation. Actually, the weakness of these data assimilation techniques used in the initial formation of tropical cyclones seriously impact on the ability of the TC analysis and prediction. In allusion to the needs of assimilation techniques, it is necessary to obtain a flow-dependent BECconformed to TC circulation characteristics, and establish the data assimilation method,which can use the flow-dependent BEC. Subsequently, it is necessary to optimize BEC’s estimation scheme and improve observation operator design and applications. Consequently, the scattered spatial and temporal distribution and indirect observation data, such as TC warning and satellite observations, are efficiently mixed together into an objective analysis data that is consistent with the laws of physics and representatives of TC structure. Thus, it would be helpful to improve analysis, research and forecasts for TC.Based on the analysis of advantages and disadvantages of current data assimilation methods, this paper proposes the ObservationBatch-wise Variational Data Assimilation and designs the Multi-Scale/Block Variational Data Assimilation accordingly, thus it is more suitable for data assimilation of TC. It is also discussed here that the estimation scheme of the flow-dependent BEC in tropical cyclone data assimilation and its impacts on TC data assimilation. What’s more, two typical data, TC warning and meteorological satellite FY2E infrared observations, are chosen for the assimilation experiment. The results show that the ObservationBatch-wise Variational Data Assimilation and the Multi-Scale/Block Variational Data Assimilation proposed in this paper have a positive effect, which can make good use of the flow-dependent BEC. The BEC’s estimation scheme, which is aimed at the TC circulation, is also reflected the real characteristics of TC circulation and correctly reflects the observation information transmission. The application technologies of two typical observation is effective; it can play a role in the formation of TC initial circulation and improve the forecast of TC path and intensity.The features work and innovations in this paper include three points. First, the Observation Batch-wise Variational Data Assimilation and the Multi-Scale/Block Variational Data Assimilation are proposed. Second, the estimation scheme of flow-dependent BEC is established, which can reflect the characteristic of the TC circulation. Finally, by means of the multi-scale/block variational data assimilation, the technology of forming TC’s initial circulationand the application scheme of typical observation are established.