Parameterization of sub-grid scale processes in climate models

Three methods of increasing sub-grid scale heterogeneity in the representation of the land surface in AGCMs are described: a method of spatial aggregation of parameter data to AGCM resolution which retains the original variability in the data in subsequent climate simulations; a parameterization of sub-grid scale precipitation, interception and runoff; and an attempt at a model of sub-grid scale lakes for climate models. The relative sensitivity of the land surface to these modifications is discussed. It is shown that the aggregation of data is likely to improve simulations while incorporating the precipitation or lake parameterizations will improve the level of realism in the land surface models, but may not improve the climate simulations.