The Improvement Of Observation Operator Of Doppler Radar Radial Velocity Assimilation And The Application In GRAPES-3Dvar

With the development of radar radial velocity observations applied in meso-scale models, researches focus on Doppler radar radial velocity assimilation are increasing rapidly. The mainstream methods mainly include two types-direct assimilation and indirect assimilation. However, the former adopting traditional emulator canjust ingest the radial component of wind fields while the later would produce retrieval errors because of the wind-retrieving process.Based on GRAPES-3Dvar system, the pre-processed module for radar radial velocity assimilation was established and the assimilation module was improved(including traditional and advanced emulators) in this research.After conducting the accuracy tests and single analysis unit assimilation tests, the single and multiple Doppler radar radial velocity assimilation and simulation tests for several rainfall events in North China were conducted to analysis the differences on the adjustment for initial conditions and prediction adopted these two emulators. The conclusions are as followed:1 Compared with traditional emulator just ingesting radial component of wind fields, the advanced one can ingest the radial velocity and those spatial variations into system, guaranteeing the complete observations. Compared with the indirect emulator, the advanced emulator calculate the optimal wind field under the restraints of background error covariance and physical constraints, omitted the retrieval process under the local strong constraints, reducing the calculation errors to great extent and achieve the goal that organism combination of retrieve and assimilation.2 Adopting area as the analysis unit, the advanced operator can utilize more observation data and achieve the data thinning result without introducing interpolating errors compared with the traditional grid point analysis unit. Moreover, the area analysis unit can depress the influenced weight for assimilation parameters of the single observation data, which can improve the reliability of observed information to some extent.3 Through the results of single analysis unit assimilation tests, all the emulators can ingest radial information effectively while there exits great differences in the tangential velocity. The tangential velocity of traditional emulator is close to that of background, at the same time, because the uniform wind assumption is perfectly satisfied, the tangential velocity of indirect operator is close to the observation value. The results of advanced emulator varied between the results of tradition emulator and indirect one, which mean that the compatibility of advanced emulator through the variation of analysis unit.4 The results of Doppler radars radial velocities assimilation tests show that the initial conditions produce more meso-scale dynamical and thermodynamical features after assimilating radar data, effectively ingesting the local features coming from radar radial velocities, and the initial conditions of advanced emulator are superior to those of traditional emulator. Moreover, compared the accumulated rainfall simulations of all tests and calculating the precipitation scores, the rainfall distributions of advance tests approach to observations most, reflected through the accurate simulating of rainfall band and restraining the virtual rainfall centers, the precipitation scores also illustrate this conclusion.