Results will be presented showing large variations in precipitation produced by cloud-scale simulations of deep and organized convective storms owing to variations in the ice phase microphysical parameterization. Utilized within the Straka non-hydrostatic cloud model is the Lin-Farley-Orville (LFO) microphysics scheme, which includes prognostic equations for water vapor, two categories of liquid water (cloud water and rain) and three categories of frozen water (cloud ice, snow, and "large ice"). Large ice is the graupel/hail category.

Storms are simulated in a thermodynamically unstable environment with a half-circle wind hodograph capable of supporting rotating supercell thunderstorms. Eight inverse-exponential size distributions, based upon in situ observations reported in the literature, are tested by varying the intercept parameters and particle densities for the large ice category. On one end of the parameter space, the large ice distribution is more representative of smaller, low-density graupel, while on the other end the large ice distribution is more representative of larger, high-density hail.

Large ice distributions with a smaller intercept parameter and/or a larger particle density result in storms associated with a greater precipitation accumulation on the ground (and less aloft). For example, the ground precipitation accumulation after 2 hours for storms with a "hail-weighted" large ice distribution was about 3 times that of those with a "graupel-weighted" distribution (78 versus 27 teragrams). Smaller sensitivities were found by varying the intercept parameters for the rain and snow categories, within their observed ranges.

These results have important implications for storm scale prediction.