Asynchronous coupling schemes between ice sheet and atmospheric forcing models are evaluated for use in long-term ice-age simulations. In these schemes the ice sheet and atmospheric forcing are run together for short synchronous periods (T
s), alternating with longer asynchronous periods (T
A) during which the ice sheet is run with atmospheric information extrapolated from the previous synchronous period(s). Two simple ice-sheet models are used that predict ice thickness as a function of latitude, and the atmosphere is represented by a prescribed pattern of net annual accumulation minus ablation. The pattern is shifted vertically to represent long-term orbital variations, stochastic inter-annual weather variability and ice-sheet albedo feedback.
Several asynchronous schemes are evaluated by comparing results with those from fully synchronous runs. The best overall results are obtained using a scheme in which the forcing during each asynchronous period is linearly extrapolated from its means in the previous two synchronous periods. Differences from the synchronous results are caused primarily by poor sampling of the stochastic forcing component, which exaggerates the stochastic ice-sheet fluctuations. We examine how these errors depend on T
s and T
A, and outline implications for GCM ice-age simulations.