The multilayer energy balance model of Peng et al. (1982) has been further developed to include a simple two-dimensional advective-diffusive deep ocean and to allow seasonal variation in order to study long-term transient climate response to a CO2 increase as well as its seasonal pattern. Comparisons between the model-simulated present climate and conventional and satellite data show that the model can simulate well the annual cycle of the surface temperature and the radiation budget of the atmosphere. In response to a hypothetical step function doubling of atmospheric CO2, the model reaches within 1/e of the equilibrium response of global mean surface temperature (2.6¿C) in 9-35 years for the probable range of vertical heat diffusivity in the ocean.

This large range of response time underlies the importance of properly using the heat diffusion coefficient in ocean models. In response to a projected CO2 trend based on estimates by the Carbon Dioxide Assessment Committee (National Research Council 1983), the model's transient response in annually and globally averaged surface temperature is 60--75% of the corresponding equilibrium response. The disequilibrium increases with increasing oceanic heat diffusivity. When the atmospheric CO2 level reaches twice the current level, in about a century, the global mean surface temperature increases by 1.5¿-2.0¿C, depending on the heat diffusivity of the ocean. Local warming at certain times of the year, however, may be 2--3 times greater than the annual and global average. In the northern high latitudes the response undergoes significant seasonal and latitudinal variations. A maximum occurs in the early winter, and a secondary maximum occurs in the spring.

In the southern hemisphere, large responses are confined to a narrow latitude zone bordering Antarctica and occur only in the cold months. The pattern of the seasonal and latitudinal distribution of the transient response remarkably resembles that of the equilibrium response to a doubling of CO2. This agrees with Bryan et al. (1982) and Spelman and Manabe (1984), implying that sensitivity studies of equilibrium climate can be used as a guide for predicting the latitudinal pattern of a warming trend. The abovementioned transient response is obtained by integrating the model from the equilibrium state corresponding to the year 1975. A comparison between this and another numerical integration of the model starting from the equilibrium state corresponding to the CO2 level in 1925 shows that the atmospheric CO2 increase before 1975 has only a very small effect on the model response after the year 2000. ¿American Geophysical Union 1987