Impacts of initial conditions on cloud-resolving model simulations are investigated using a series ofsensitivity experiments.Five experiments with perturbed initial temperature,moisture,and cloud conditionsare conducted and compared to the control experiment.The model is forced by the large-scale verticalvelocity and zonal wind observed and derived from NCEP/Global Data Assimilation System(GDAS).Theresults indieate that model predictions of rainfall are much more sensitive to the initial conditions than thoseof temperature and moisture.Further analyses of the surface rainfall equation and the moisture and cloudhydrometeor budgets reveal that the calculations of vapor condensation and deposition rates in the modelaccount for the large sensitivities in rainfall simulations.