A wide variety of land-surface models have now been developed that gives a response of land climate variables to atmospheric conditions. Some are general enough to match the wide range of conditions provided by a global model; others are intended for more regional or local applications. There is much similarity and overlap between models, and they all attempt to treat many of the same processes. Most such models have been constructed from elements of earlier models. However, there are also some substantial differences in objectives and details between models. New land-surface parameterizations have been developed by Xue et al. [1991] and Wood et al. [1992].
The Project for Intercomparison of Land-Surface Parameterization Schemes (PILPS) has been developed through the WCRP Global and Energy Water Experiment(GEWEX) program to provide systematic comparisons between all participating land-surface schemes [ Henderson-Sellers et al., 1993b; Pitman et al., 1993]. For prescribed atmospheric inputs over an annual cycle and presumably the same surface description, different schemes provide a wide range of portioning between sensible and latent fluxes. At least some of this variability can be ascribed to the wide variety of vegetation treatments and soil depths between the different models.
The Atmospheric Modeling Intercomparison Program (AMIP) project [ Gates, 1992] is providing a database of multi-year climate simulations from a wide variety of GCMs, allowing intercomparisons between land processes in a completely coupled framework. Randall et al. [1992] intercompare surface fluxes over the ocean for coupled models for prescribed increments of ocean surface temperature. A wide range of evaporative fluxes was found between models, even under these simple conditions, a conclusion that is being replicated for studies over land. Milly [1994] has studied the question as to what parameters are most important for determining the annual runoff to precipitation ratio. Key parameters are Budko's index of dryness, which is the ratio of annual net radiation determined potential evapotranspiration to precipitation, the soil water holding capacity, and the arrival rate of storms. Seasonal and stochastic variations of precipitation rates, potential evapotranspiration, and storm arrival rates are also significant.
Betts et al. [1993] have compared averaged fields and fluxes from
the First ISLSCP [International Satellite Land-Surface Climatology Project] Field Experiment
(FIFE) for summer and fall of 1987 (next section) with surface output from the European
Centre for Medium-Range Weather Forecasts (ECMWF) model. A number of substantial
defects in the model were thus revealed. The use of a 0.07 m slab soil layer
unshaded by vegetation gave soil heat fluxes that are much too large. The use
of 0.07 and 0.42 soil layers for water storage, with prescribed wetness values
below, gave zero evapotranspiration over a period when daytime fluxes greater
than 50 W/m
were observed. Model solar fluxes exceeded those observed by
approximately 10%. The treatment of land in this and other global Numerical
Weather Prediction (NWP) models has a substantial impact on the land surface
climatic data inferred from such models, which are widely used for diagnostic studies.
There is commonly a high bias in surface solar radiation in climate models
[ Dickinson and Kennedy, 1991; Garratt, 1994].
Milly and Dunne [1994] have carried out GCM sensitivity studies with soil water capacity varying from 0.01 to 1.2 m. They find approximately a 10% increase in global evapotranspiration for each doubling of the soil water capacity. About half of this atmospheric water increase is added to land precipitation and half to oceanic. Lakhtakia and Warner [1994] have made comparisons of simulations of a mesoscale model with a simple `bucket' model versus a soil-vegetation land model. Bonan [1994] has examined the dependence of land climatologies in Community Climate Model Version 2 (CCM2) on model resolution and concludes that improving resolution does not contribute noticeably to reducing discrepancies between modeled land-climate and observations.