P5.2 EnKF analysis and forecast predictability of a tornadic supercell Storm

Tuesday, 28 October 2008
Madison Ballroom (Hilton DeSoto)
Edward R. Mansell, NOAA/NSSL, Norman, OK ; and L. J. Wicker

An Ensemble Kalman Filter (EnKF) system is used to assimilate radar velocity and reflectivity into a cloud-scale model for analysis of the 8 May 2003 Moore-Oklahoma City, OK, tornadic supercell storm. Data from the KOUN radar are assimilated from 20:46 UTC out through the approximate start time of the major tornado (22:10 UTC). Forecasts are carried out from analyses at lead times of 0 to 30 minutes prior to 22:10 to assess the skill of predicting a long-lasting low-level mesoscyclone. Assimilation tests are conducted at 1 and 2-km horizontal resolution with a variety of bulk microphysics parameterizations to begin to compare the performance (and computational cost) of schemes with increasing complexity. The Collaborative Model for Multiscale Atmospheric Simulation (COMMAS) has both single and double-moment microphysics schemes with and without the ice phase. (The double-moment schemes here predict hydrometeor number concentration in addition to mass.)

Convection is initiated by inserting randomly-perturbed warm bubbles in the boundary layer where cells are indicated by radar reflectivity. The ensemble members therefore require some spin up time, which allows for some comparison of precipitation initiation in the various microphysics schemes. Comparison of time-height reflectivity innovations (RMS and mean) reveal differences in microphysical behaviors, such as speed of precipitation generation and fall speed of precipitation hydrometeors. For example, the single-moment warm-rain (i.e., no ice phase) Kessler-type schemes tend to generate precipitation (and radar echo) more rapidly and thus tend to have lower initial reflectivity innovations. On the other hand, the addition of ice or number concentration provides greater diversity in fall speed and better mean innovations at middle to high altitudes at later times. Ice and number concentration also seem to help maintain higher ensemble spread, which is important for effective data assimilation.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner