Abstract
The sensitivity of simulations of shallow cumulus convection to their microphysical representation is explored, with a focus on the parameter space spanned by two common bulk schemes (Seifert and Beheng, SB, versus Khairoutdinov and Kogan, KK). Large-eddy simulation, simple models, and a priori analysis of the underlying microphysical equations are used as the basis for our study. The simulations are initialized using data derived from the Rain in Cumulus Over the Ocean (RICO) field study. Simulated clouds depths range between two and three kilometers. Microphysical sensitivities can largely be rationalized based on the behavior of simpler models. In particular a parcel model consisting of auto-conversion and accretion acting on a parcel condensing water at a fixed rate provides useful insight into the behavior of the microphysical schemes in the full simulation. To a first approximation the number concentration simply selects the cloud depth at which rain begins to develop, with different schemes predicting different cloud depths. Because of the interaction of auto-conversion and accretion the dependence of this cloud depth on cloud-droplet number concentration is considerably reduced from what would be deduced by an examination of auto conversion alone-suggesting a somewhat diminished role for the regulation of rain by the atmospheric aerosol. Dynamic feedbacks, such as the tendency for non-precipitating clouds to deepen more readily, can further dampen (and even reverse) the expected sensitivity of rain-rate on droplet number concentrations. Our analysis suggests that the commonly assumed exponential distribution for rain drops can strongly distort the sedimentation process in two moment microphysical schemes and that processes such as self-collection and drop break-up can not be neglected for shallow cumulus convection. [References: 26]