The JMA has operated a mesoscale numerical weather prediction (Meso NWP) system since March 2001. The R/A has been assimilated in the mesoscale analysis from the beginning of the operation. The method was a physical initialization (PI) for the first one year and then replaced by the four-dimensional variational data assimilation system (Meso 4D-Var) in March 2002. Although the impact on the quantitative precipitation forecast (QPF) in the case of PI lasted only for the first few hours, that in the case of Meso 4D-Var lasted throughout all forecast time to 18 hours.
In addition, the Doppler radars have been installed at major 8 airports in Japan. Recently radars of the JMA radar network start to be replaced with the Doppler radars gradually. The radial velocity data were first assimilated by the Meso 4D-Var in March 2005. Currently, the data from the 9 sites are assimilated to contribute the improvement of the analysis.
We have been developing a new 4D-Var data assimilation system (JNoVA) for the mesoscale analysis. This new 4D-Var employs the JMA nonhydrostatic model while the Meso 4D-Var employs a hydrostatic spectral model. A preoperational version of the JNoVA has been compared to the operational Meso 4D-Var almost under the same configuration. Both R/A and the radial velocity data are assimilated in this experiment.
Although the preoperational version of the JNoVA considers only the water vapor, a research version of the JNoVA considers cloud water, cloud ice, rain and snow. So it becomes possible to assimilate the radar reflectivity data directly by the research version of the JNoVA. From the impact experiment of the radar reflectivity assimilation, it is found that the radar reflectivity data are useful to detect convective cells in the analysis and promotes the assimilation of surface precipitation data simultaneously.
In the presentation, we will talk about the history of radar data assimilation of operational NWP system at JMA and results of experiments by the JNoVA.