2014 年 10 巻 p. 145-149
A new 4-dimensional variational data assimilation system with 0.5-km grid spacing (NHM-4DVAR.v3) was developed by integrating the nonhydrostatic storm-scale 4D-Var (NHM-4DVAR.v2) and the Japan Meteorological Agency (JMA) nonhydrostatic model (NHM) based Variational Data Assimilation System (JNoVA). Both systems are based on the JMANHM, but horizontal resolutions, their formulations, adjoint models of physical processes, and observation operators are different. NHM-4DVAR.v3 comprises advantages of both systems: a penalty term, optimization of lateral boundary conditions, and observation operators for advanced observations. This development aimed at improving the forecast accuracy of hazardous weather at meso-γ-scales (5∼20 km). In this paper, the characteristics of NHM-4DVAR.v3 and some results, including the integrated formulations, are presented. An assimilation experiment of actual observations using NHM-4DVAR.v3 with 2-km grid spacing was found to show improvement over NHM-4DVAR.v2 at the same resolution. As a final goal, NHM-4DVAR.v3 was applied with a 0.5-km resolution. The comparison between assimilation results by NHM-4DVAR.v3 with 0.5- and 2-km horizontal resolutions indicates that analyses with super high resolutions can reproduce more detailed atmospheric features such as convective clouds.