EnKF Data Assimilation and Forecasts of the Goshen County, Wyoming, Supercell during VORTEX2
Timothy A. Supinie, Youngsun Jung, and Ming Xue
Data collected by rawinsondes, profilers, surface stations, and three Weather Surveillance Radar-1988 Doppler (WSR-88D) radars, in addition to other field observations collected by VORTEX2, are assimilated using the Ensemble Square Root Filter (EnSRF) algorithm to examine the Goshen County, Wyoming, supercell from 5 June 2009. Because accurate prediction at both the mesoscale and storm-scale are important for the evolution of a supercell storm, storm-scale ensemble analyses at 1 km grid spacing are nested inside mesoscale ensemble analyses at a grid spacing of 3 km. The 40-member initial mesoscale ensemble is initialized by adding mesoscale perturbations to a 6-hour forecast on the 3 km grid from the 1200 UTC 5 June 2009 run of the NCEP NAM. Data are assimilated every 30 minutes from 1800 to 2100 UTC in the 3-km ensemble and every 5 minutes from 2100 to 2200 UTC in the 1-km ensemble. 1-hour ensemble forecasts using the ARPS model are then launched at 2200 UTC.
The impact of the dense VORTEX2 observations along with other model configuration and data assimilation tuning parameters, such as computational mixing (CM) parameter and horizontal correlation length scale for initial condition perturbations, on forecasts of the Goshen County supercell are assessed in this study. The forecasts are compared against observations, and reflectivity and vortex probability swaths generated from the forecasts are examined. A model and data assimilation configuration that gives a subjectively good forecast has been found using a CM parameter (normalized to the grid spacing) of 8×10-4 s-1 and a horizontal correlation length scale for initial condition perturbations of 16 km.