Implementation of Terrain Resolving Component by Immersed Boundary Method for Variational Doppler Radar Analysis System (VDRAS)

The Variational Doppler Radar Analysis System (VDRAS) developed by National Center for Atmospheric Research (NCAR), has been widely used for nowcasting and also quantitative precipitation forecast nowcasting (QPN). Although many successful applications of VDRAS were shown in the recent years, the lack of ability to resolve terrain effect limits its performance when complex topography is within the analyzing domain.

For the purpose of improving the analysis system, Ghost Cell Immersed Boundary Method (GCIBM) is applied to this 4D-VAR radar data assimilation system includes both cloud model and its adjoint model. The GCIBM-implemented model can simulate reasonable terrain effects by enforcing terrain boundary conditions implicitly through ghost cell grids and is suitable for any Cartesian grid code. The GCIBM is chosen for our needs since VDRAS cloud model constructed in Cartesian coordinate.

In this study, the validation experiment is accomplished by a two-dimensional linear mountain wave simulation first. The numerical simulation shows compatible result with proposed analytical solution. An extended three-dimensional leeside vortex simulation also conducted. A pair of symmetric lee vortices exists and becomes elongated as time evolves, represents good agreement with previous studies. For the moist case, a parallel forecast experiment is set to examine the modified VDRAS cloud model by qualitative comparison with WRF model forecast under the same atmospheric environment. Results show our Cartesian-grid model can simulate very similar convection evolutions with WRF model constructed in follow-terrain coordinate. Furthermore, in order to prove the completeness for the modification of corresponding adjoint model, an OSSE for radar data assimilation is also presented. Virtual radial velocity and rainwater observations obtained from WRF simulation are assimilated in full model domain coverage for five minutes assimilation window. The retrieval from dynamic and microphysics variables show great agreement with the truth simulation. In addition, the following cold pool propagation forecast also demonstrates its promising usage for QPN.