Cloud resolving simulations of tropical cloud systems: Using field program observations to evaluate ice phase microphysics parameterizations

This study analyzes two sets of cloud model simulations of deep, precipitating, convective systems. One is a 29-day summertime, continental case from which two convective systems are selected. It is taken from the ARM Program summer 1997 SCM intensive observation period (IOP), at the SGP site. The second is a 52-day simulation of tropical maritime conditions in the central Pacific Ocean, from which intensive study of the MCS of 11-12 August 1999 during the TRMM Kwajalein Experiment is used for a large set of simulations and sensitivity testing.; The relationship of simulated microwave brightness temperatures and radar reflectivities to the simulated convective vertical velocities shows that these radiative quantities are strongly dependent on vertical velocities. Statistically, the TRMM measurables are good proxies for convective intensity.; Comparison of simulation results with the TRMM observation statistics indicates that the model tends to underestimate microwave brightness temperatures at ice scattering frequencies and overestimate radar reflectivities. The differences between the statistics of KWAJEX simulation and available ground-based precipitation radar observations increase with height and reach a maximum near 9 km. The extreme graupel mixing ratios are the most likely reason for the unrealistically high simulated radar reflectivity.; A series of experiments are conducted using different intercept parameters of the particle size distribution of graupel (N0g) and different graupel density (rho0g). Changes of graupel intercept parameter and density have relatively small influence on simulated convective velocity. The simulated snow and graupel amounts are most sensitive to the graupel intercept parameter. Change of the graupel intercept parameter has a significant influence on the microphysical terms associated with graupel production. The bulk microphysical parameterizations are the major reason leading to the overestimated graupel in numerical simulation results.; The sensitivity to horizontal grid resolution and domain size has been tested. The simulated statistical convective properties change considerably as horizontal grid resolution changes from 1000 to 500 m, but change less as the grid size goes from 500 m to 250 m. As the horizontal domain sizes are varied from 64 km x 64 km to 128 km x 128 km, little change of the simulated statistical convective properties is found.