Sensitivity of Single-Column Model Solutions to Convective Parameterizations and Initial Conditions

Two sets of single-column model (SCM) simulations are performed to determine whether the SCM solutions are more sensitive to model parameterization schemes than to initial perturbations in temperature and moisture profiles. The first set of simulations (S3) used the Zhang and McFarlane scheme for the deep convection and the Hack scheme for the shallow convection, while the second set (S2) used the Hack scheme for all types of convection. The same random perturbation used by Hack and Pedretti is applied in S2 and S3. The observed total (horizontal and vertical) advections of temperature and moisture during the Tropical Ocean Global Atmosphere Coupled Ocean–Atmosphere Response Experiment are used to force all simulations. A major difference in temperature and moisture biases occurs between the ensemble means of the two sets of simulations, and is much larger than the standard deviation of each set. Differences are also evident in cloud and radiative properties. This demonstrates that SCM solutions can be more sensitive to the model physics than to the initial perturbations. In other words, the deterministic aspects of SCM solutions dominate the nondeterministic aspects, which is important for their continued use in developing parameterization schemes of convection and clouds in large-scale models. This point is also supported by the SCM simulations using several available longer observational datasets over different regions.