Some Problems in the Numerical Simulation of Climate Variability Using High-Resolution Coupled Models

Abstract

The development and application of high-resolution coupled climate models poses a number of problems. The large timescale separation between different climate subsystems requires special numerical coupling techniques and imposes difficult tradeoffs between the spatial and temporal resolution and the length of integration. The lack of any external boundary condition restraints, apart from the prescribed solar insolation, gives rise to model drift problems. In extensive simulations on the timescale of the slow system (e.g., 10³ years for a coupled ocean-atmosphere model), the more rapid fluctuations of the fast system (e.g., atmospheric weather variability) can no longer be explicitly resolved. They can nevertheless give rise to significant low-frequency variability and must therefore be parameterized as stochastic forcing terms. Finally, the analysis and dynamical interpretation of the complex space-time variability of a high-resolution coupled model cannot be readily carried out with standard statistical analysis methods and requires the development of alternative techniques. Most of the problems are interrelated. Various methods for resolving these problems are discussed and illustrated by specific examples.