The seasonal cycle of surface temperature is largely controlled by the land--sea distribution. Previous studies with a two-dimensional, seasonal energy balance model (EBM) suggested that large annual cycles on supercontinents could produce sufficiently high summer temperatures to melt summer snow, even when the continents were located in polar regions. The above calculations were done with a linear model. In this paper we test the sensitivity of these conclusions to seasonally varying snow albedo feedback by developing a new nonlinear two-dimensional, seasonal EBM. The model satisfactorily reproduces the present annual and semi-annual cycles, plus snow and sea ice margins; its senitivity is ≈25% greater than that of the linear model and comparable to a number of general circulation model (GCM) studies for altered CO2 levels. Experiments are described for a series of idealized supercontinent configurations; in addition we examine the effect of changing land--sea distributions on the climate of Greenland and Antarctica.

Supercontinent model simulations are similar to the previously published linear model results and provide further support for the hypothesis that ice-free states could occur and polar supercontinents. However, some of our results for smaller land masses differ from the earlier study. Although temperatures are systematically lower for the nonlinear run, we simulate>15 ¿C decrease in summer temperatures over Greenland during the last 100 million years in both the linear and nonlinear models. However, there is only a weak suggestion of a long-term trend in the southern hemisphere. Revised reconstructions of southern hemisphere geography further obscure any long-term trend in this region. Overall, model simulations provide further support for the propositon that changes in seasonality have played an important role in climate change over the past few hundred million years.