changes in roughness and albedo at the land surface.28 As noted above, the Earth's climate regulates the distribution of ecosystems, which in turn modifies land surface properties such as surface roughness and albedo, which then feeds back on the climate system. Elements of the water balance also regulate the carbon and nitrogen cycling on both continental and local scales.29 As such, soil moisture is a key component in the land surface schemes in GCMs, since it is closely related to evaporation and thus to the apportioning of sensible and latent heat fluxes. Accurate prediction of soil moisture is crucial for simulation of the hydrological cycle and of soil and vegetation biochemistry, including the cycling of carbon and nutrients at local, regional, continental, and global scales. It thus plays a significant role in atmospheric models, hydrological models, and ecological models.
Unfortunately, there exist large differences between models of soil moisture even for simulation runs with high-quality atmospheric forcing data in carefully chosen parameters.30 Therefore, the prediction of future soil moisture through coupled terrestrial-atmosphere models cannot be considered reliable, especially since the forcing data are necessarily inaccurate and the information required for specifying land surface parameters is crude. Moreover, current land surface schemes differ profoundly between models in terms of their structure and their treatment of various land surface processes such as evaporation, transpiration, and drainage; it appears that differences in scheme structure are of particular importance. 31
The differences among present land surface schemes32 used in models are manifested in a number of ways:
Different annual equilibrium when forced with the same atmospheric forcing data and the same land surface parameters.
Different descriptions of the seasonal cycle of soil moisture. The greatest dispersion occurs when vegetation contributes to the total evaporative flux, when there is a great atmospheric demand, and when the available soil moisture is limited.
Different partitioning of incoming precipitation among runoff-drainage, soil storage, and evaporation depending on timing and antecedent conditions.
Most schemes can be tuned to observations, but no single scheme predicts well all of the variables describing the land surface hydrology. Indeed, the consensus (single average) of all participating schemes generally outperforms any individual scheme. This suggests that individual schemes capture specific aspects of this complex system well but that no scheme yet captures the whole system satisfactorily and consistently. This issue is important and deserves attention.
As noted above, critical improvements in ecosystem modeling and its linkage to Earth system models will require the development of schemes for integrat-