The hydrometeors have six categories: water vapor, cloud water, cloud ice, rain, snow, and graupel/hail. The bulk method is applied to parameterize 38 microphysical processes including autoconversion, nucleation, condensation, accretion, evaporation, freezing, melting, sublimation, deposition, etc. Except for cloud ice which is assumed to be hexagonal plates, hydrometeors are assumed to be spherical. The cloud water is supposed to be monodispersed with a number concentration; whereas cloud ice is assumed to be monodispersed as a function of temperature. The rain, snow, and graupel/hail are assumed to be inverse exponential size distributions (Marshall and Palmer, 1948; Gunn and Marshall, 1958; Federer and Waldvogel, 1975). The precipitation in all types of water species is assumed to fall with mass-weighted mean terminal speed. There are twelve prognostic variables containing velocity in x, y, z directions, pressure, potential temperature, turbulent kinetic energy, mixing ratio of water vapor, bulk cloud water, bulk cloud ice, bulk rain water, bulk snow aggregates, and bulk graupel/hail (Straka 1989). The Arakawa-C staggered grid (Arakawa and Lamb, 1981) is utilized in this model. The domain is 64 km × 64 km × 25 km with horizontal grid interval 500 m, and vertical interval 200 m. The time interval of integration is 2 seconds. The horizontal wind is subtracted, changing every 30 mins, to confine the convective activities within the computational domain. The upstream sixthorder, flux-conservative Crowley scheme is employed in advection terms (Tremback et al., 1987). Predictive variables are filtered every time step to handle nonlinear instability by a fourth-order numerical spatial diffusion operator (Klemp and Wilhelmson, 1978). A time filter is used in prognostic variables in the leapfrog scheme from odd and even time steps (Asselin, 1972). The radiation condition is utilized at the lateral boundaries to let patterns go out of domain without producing disturbances (Klemp and Wilhelmson, 1978). At the top boundary, all variables are given the values of the base state. A Rayleigh sponge layer is imposed to reduce gravity waves close to the top of the domain (Clark, 1977). All the hydrometeor concentrations are shown as mixing ratios. Negative numbers of them do not set to null nor participate in any calculation. Such a design is to avoid false accumulation of water species, which happens when negative hydrometeors mixing ratios are set to zero. The convection is initiated by a thermal bubble, whose center is 2 km above ground level, with the maximum 3.5 K at bubble center and the radius 10 km and depth 4 km in the surrounding of horizontal homogeneousness. The relative humidity of the bubble is set to be the same as outside the thermal bubble by varying mixing ratio. A sensitivity study of the thermal bubble size was conducted and the results show that, while the numbers change somewhat, the main conclusions remain the same (see Sec. 5).