68 Estimating Convective Storm Turbulence with Radar to Evaluate the Paramerisation of Turbulence in NWP Models

Monday, 28 August 2017
Zurich DEFG (Swissotel Chicago)
Matthew Marcus Feist, Univ. of Reading, Reading, U.K.; and C. D. Westbrook, P. Clark, T. Stein, H. W. Lean, and A. Stirling

The effects of turbulence on the structure and evolution of convective clouds remain unclear in observations and numerical weather prediction (NWP) models. NWP models often employ simple sub-grid turbulence schemes to parameterise the effects of turbulence on spatial scales lower than the model resolution. At high resolution (less than 1 km grid length), the characteristics of simulated convection can be highly sensitive to the application of turbulence parameterisations. In order to evaluate the performance of turbulence schemes, we have developed an algorithm to retrieve eddy dissipation rates ε, from radar observations of convective storms. By applying this retrieval throughout large datasets of radar observations collected during the DYMECS project, we have investigated the characteristics of convective storm turbulence statistically for model evaluation. Dissipation rates in our observations generally range from 0.001 to 0.1 m²/s³, with the largest values typically found within, around and above convective updrafts. In convective clouds, the 95th percentile of ε has been found to generally increase with height, while distributions of ε in individual clouds are approximately log-normal. In convective updraft regions, mean values of ε are found to be positively correlated with the maximum updraft velocity (r = 0.63, p < 0.001), and the mean horizontal shear in the vertical velocity (r = 0.55, p < 0.001). A weaker positive correlation was found with updraft height (r = 0.26, p < 0.01). We have begun to compare these observations to ε produced at 100-m resolution in the Met Office Unified Model (Met UM), which employs a Smagorinsky sub-grid turbulence scheme. Generally, the strength of ε in convective clouds simulated in the Met UM is smaller than those observed. Going forward, direct statistical comparisons will be made between observations and ε found in convective clouds simulated on DYMECS case days.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner