Influence of linearized physical processes on the GRAPES 4DVAR

The linearized physical processes can improve the convergence stability of the four-dimensional variational assimilation (4DVAR) minimization, and increase the significant description of the atmospheric physical processes and dynamics during the minimization. It is a very important component of the 4DVAR. In order to improve the analysis and forecast effect of GRAPES global 4DVAR, a set of linearized physical parameterizations have been developed for the GRAPES global model to improve the accuracy of the tangent linear model (TLM), especially the two moist linearized physical parameterizations. The development of the linearized physical parameterizations requires the simplification of those strong nonlinear terms in the non-linear physical parameterizations and the regulation of the linearized physical parameterizations, and reduces the abnormal growth of the tangent linear perturbation. At present, the following linearized physical processes are described in GRAEPS global model:Subgrid-scale orographic effect, vertical diffusion, deep cumulus convection and large scale condensation. The test method for the TLM accuracy with the linearized physical parameterizations is to compare the zonal mean errors between the perturbation evolution in the nonlinear model including full physics and the perturbation evolution in the TLM including the linearized physical parameterizations. It is shown that for finite size perturbations (analysis increments), the inclusion of physics improves the fit to the non-linear model. Then based on the adiabatic TLM, the effect of these linearized physical processes is examined for summer and winter cases for 12 h forecasts. The experimental results show that by adding two dry linearized physical processes (vertical diffusion and subgrid-scale orographic effects), the abnormal growth near the surface in the adiabatic TLM can be effectively suppressed, and the accuracy of the tangential linear mode can be greatly improved. By adding two moist linearization physics processes, i.e., deep cumulus convection and large-scale condensation, the accuracy of the moisture and temperature increments in the TLM can be improved in the tropics and middle to high latitudes, and thus the analysis and forecast effect of GRAPES global 4DVAR can be improved.