Mathematical Models for Climate as a Link between Coupled Physical Processes and Computational Decoupling
Engineering Simulation: An International Journal of Electrical, Electronic and Other Physical Systems, 15(4), 509--544, 1998
Mathematical models for climate studies are treated as a coupling link between physical and computational models. These models are characterized by the fact that small-scale phenomena influence the large-scale properties of the modelling system, yet the former cannot be extracted from the latter using available hardware and computational procedures. Climate systems belong to the class of systems whose dynamics are only observable in transient states. As a result, the sensitivity of models to coupling procedures requires an examination of the schemes responsible for transporting data between components. It is proposed to perform such an examination, based on the connection between error growth and the degree of coupling of model components, using adaptive error control.