The development of biosphere models for future global climate models will require the successful prediction of the natural vegetation “congruent” to the climate. A highly generalized (nine class) grouping of Holdridge life zones has been used here as a means of investigating three-step “coupling” of a land-surface scheme into a global climate model. Another requirement for future modeling is the development of a method of spatial aggregation of land-surface data to the required global model resolution while retaining the necessary degree of variability in the data for subsequent climate simulations. It is shown that the aggregation of data rather than the selection of a dominant ecotype is likely to improve future simulations. A third vital component of future modeling is the representation of important sub-grid scale processes, especially those associated with hydrological exchanges between the land surface and the atmosphere. It is shown that the incorporation of a scheme for representing sub-grid scale precipitation and lakes improves the level of realism in the land-surface models and should improve the overall climate simulations.