Impact of multi-moment microphysics and model resolution on predicted cold pool and reflectivity intensity and structures in the Oklahoma tornadic supercell storms of 3 May 1999

We report on the results of numerical experiments designed to test the impact of single and multi-moment bulk microphysical parameterizations (BMP) on high-resolution convection-permitting simulations of the 3 May 1999 Oklahoma tornado outbreak using the ARPS model. This outbreak was characterized by severe, long-lived supercells that often tracked in close proximity (50 km or less) to each other without significant destructive interference, and displayed relatively weak and localized cold pools. We perform real data simulations starting from analyzed initial conditions that include radar observations, as well as more idealized simulations with a horizontally homogeneous environment defined by a sounding extracted from the inflow region of the real-data simulations, in order to understand the impact of microphysics on the cold pool and reflectivity structures of the simulated storms.

A strong resolution dependence of the microphysics impact is found, with higher-resolution (500 m grid spacing or smaller) simulations (Fig. 1) showing much better agreement in the cold pool and reflectivity structures with the observations when using a double or triple-moment BMP developed by Milbrandt and Yau (MY, 2005). The double-moment scheme predicts mixing ratio and number concentration for either an exponential or gamma distribution for 4 classes of ice species and 2 classes of liquid. However, single-moment BMP schemes with default values of intercept parameters, including the single-moment version of the MY scheme, a Lin-type scheme default in ARPS, and the WRF single-moment 6-class scheme (WSM6) all produced cold pools that were much too large, strong, and dry, even at high resolutions. Although reducing the value of the intercept parameter for rain and hail in single-moment schemes has been shown to reduce cold pool strength in previous studies, the multi-moment schemes effectively predict such parameters with much less arbitrariness. However, the performance of a single-moment scheme (WSM6) based on a diagnostic rain intercept parameter will be also be examined in this study. The triple-moment scheme adds radar reflectivity to the predictive equations, allowing the shape parameter in the gamma distribution to vary independently, and preliminary results suggest that the improvement in going from 2 to 3 moments is significantly less than from going from 1 to 2, at least inasmuch as the reflectivity and cold pool structures are concerned. At coarser resolutions (1 km grid spacing and larger), all BMP's featured overly-intense cold pools, including the double and triple-moment MY schemes, suggesting that 1 km or coarser grid spacings are inadequate for reproducing even the qualitative behavior and evolution of the convective cold pools in this case. Attempts are being made to determine the underlying causes of these differences.

Figure 1 shows simulated surface equivalent potential temperature, reflectivity, and wind vectors at 1 hr for 4 simulations with 500 m horizontal grid spacing: the ARPS Lin scheme, and the 1, 2, and 3 moment versions of the MY scheme