Sensitivity of Precipitation Modeling to Uncertainty of Initial Conditions

Abstract

The meaningful precipitation simulation and estimate require sophisticate models with accurate cloud microphysical and radiative parameterization schemes. One of such models is cloud-resolving model. The precipitation processes are highly nonlinearly associated with the dynamic, thermodynamic, cloud microphysical and radiative processes, which makes precipitation simulations very sensitive to temperature, water vapor, and parameterization schemes of these physical processes (Grabowski et al. 1998; Donner et al. 1999; Guichard et al. 2000; Xu et al. 2002; Petch et al. 2002, 2008; Petch 2004, 2006; Phillips and Donner 2006; Keil et al. 2008). Petch and Gray (2001) found that the model domain size, horizontal resolution, use of a third dimension, and cloud microphysical parameterization have impacts on the model simulations. Petch et al. (2002) revealed that the finer horizontal resolution (<250 m) is required to realistically reproduce the development of boundary-layer clouds. Cheng and Xu (2006) studied the effects of turbulence closures on cloud simulations and found that fully prognostic quasi-Gaussian based third-order closures produce more and deeper shallow cumuli but smaller and narrower convective clouds than intermediately prognostic double-Gaussian based third-order closures do. Khairoutdinov and Randall (2003) showed that the uncertainties of precipitation simulations are much more sensitive to the uncertainties of initial conditions than the uncertainties of cloud microphysical parameterization schemes. Li et al. (2006) and Gao and Li (2008) conducted 2D cloud-resolving model simulations with the perturbed initial PW and temperature conditions and compared these perturbed experiments with the control experiment. They found significant differences in cloud and precipitation simulations with the given precipitation water perturbations. The further budget analysis indicates that the errors of initial conditions affect cloud and precipitation ­simulations through a biased condensation process.