David Williamson
National Center for Atmospheric Research

The products of climate models are temporal statistics based on a calculated evolution of an atmosphere model, when the model is in a forced-dissipative equilibrium, no longer under the influence of the initial conditions. Therefore additional evaluation techniques beyond deterministic test cases are needed. Convergence of deterministic test cases, while an important first step, is not sufficient. Because of uncertainties in the nonlinear, interactive forcing, absolute errors of the statistics attributable to the dynamical cores are difficult to determine. However comparative statements involving several cores can be made when they are coupled to the same parameterization suite. This addresses the critical question: from a set of schemes which produce the same quality statistical solution, which requires the least computer resources. Of course, such analysis requires resolution dependent signals, preferably monotonic ones. Examples of such analyses will be discussed which are based on integrations of an idealized, water covered earth: the aqua-planet. These integrations use the complete atmospheric model parameterization suite, but a simplified underlying surface. We will discuss various flow aspects that can provide the desired information, such as time means, variances, covariances, kinetic energy spectra, tropical wave propagation characteristics, frequency distributions of precipitation, etc. The same approach also is useful to study the resolution sensitivity of the parameterization suite coupled to a single dynamical core. This can be applied to models with uniform resolution as well as to scale-adaptive models. The idealized environment provides statistically significant results from shorter simulations than required with Earth-like simulations.

Presentation (PowerPoint File)

Back to Workshop II: Numerical Hierarchies for Climate Modeling