A case study and batch verification on high resolution numerical simulations of severe convective events using an analysis system based on rapid-refresh 4-D variational radar data assimilation

The high-resolution numerical simulations of 18 convective events occurred over the Beijing-Tianjin-Hebei region have been conducted using a numerical analysis system based on the rapid-refresh 4-D variational assimilation (RR4DVar) technique of multi-radar observations and a 3-D cloud-scale numerical model. Observations of both reflectivity and radial velocity at 6-min interval from six CINRAD radars and the integration of observations updated at 5-min interval from auto weather stations (AWS) in the study region are assimilated in the meso-scale numerical model simulations. Results from observations and the simulation of a selected convective event are analyzed first. Simulations of all the 18 events are verified later. The case study indicates that the simulated low-level 3-D dynamical and thermodynamical fields and water vapor can be used to explicitly interpret the local initiation and organization of convective storms and the formation of meso-scale linear convective systems over complex terrain area. Using the high-resolution simulation results, mechanisms for the consecutive initiation (regeneration) and upstream back building of convective cells in the linear MCS can be revealed; effects of topography forcing on the formation and evolution of convective storms can be demonstrated clearly. Simulations of the 18 convective events are verified against high-frequency observations from one wind profiler, two ground-based microwave radiometers, one second-level radiosonde, and one boundary layer tower. The verification results indicate that the biases and root-mean-square errors (RMSE) of simulated wind speed, wind direction, and temperature at 0-3 km levels are relatively low. The bias and RMSE of wind speed are smaller than -0.5 and 0.9 m/s at the lowest model level of 187.5 m, and smaller than -0.9 and 1.6 m/s at the highest verification level of 2.8125 km, respectively. Within 0-3 km levels, the error of wind speed increases with height. The bias of wind direction is between 14° and 22°, and the RMSE is less than 38°. The bias and RMSE of temperature are less than -1℃ and 1.8℃, respectively. The bias and RMSE of simulated low-level wind speed, wind direction, and temperature inside the convective system are slightly larger than that outside the convective storm. The study implies that the modeling system has significant advantages in nowcasting and warning initiation. It can well simulate the evolution, dissipation, and life cycle of convective storms.