Ultimately, importance of LA-I is expressed through its role in determining various regional climate systems. U.S. efforts over the period of review have addressed modeling of LA-I over the U.S. through the use of mesoscale models integrated for time periods of up to several years and with prescribed lateral boundary conditions from either NOAA or ECMWF forecast initialization archives or from global climate simulations. Giorgi et al. [1994a,b] have reported on a multi-year simulation with the Mesoscale Model 4 (MM4) driven by climate model output and described precipitation, soil moisture, evapotranspiration, and runoff. The mesoscale model evidently substantially corrects a large positive bias in the average U.S. precipitation, but maintains a relative excess over the western U.S. Whether the overall reduction in precipitation is a result of differences in model resolution or some other feature of the mesoscale model is not known. This question is important because many, if not most, GCMs appear to have a high bias in their overall land precipitation [e.g., as analyzed by Schultz et al., 1992]. Model evapotranspiration and soil moisture were found to be in substantial disagreement with observational data. However, the only observations used for the latter were observed monthly precipitation; furthermore, the comparison data are derived from a much cruder surface model than that used by Giorgi et al. [1994a,b].
Efforts have also continued to examine the LA-I consequent to a hypothetical complete deforestation of the Amazon. Nobre et al. [1991] obtained not only substantial reductions in evapotranspiration with deforestation but an even larger reduction in precipitation. Dickinson and Kennedy [1992] and Henderson-Sellers et al.[1993a] have addressed this climate response with the National Center for Atmospheric Research (NCAR) CCM1 model. Both studies also find precipitation decreases to be considerably larger than decreases in evapotranspiration (ET). Other models, however, may give less pronounced results. Further insights into the mechanisms underlying the numerical results can be provided by analytic studies with mechanistic models. Eltahir and Bras [1993b] developed a simple model interpreting earlier numerical simulations. Their model assumes that increased surface temperature increases precipitation but that reductions in precipitation produce a positive feedback that further reduces precipitation. This model neglects the effects of possible changes in boundary layer structure on precipitation.
Henderson-Sellers et al. [1993a] have argued the need for a considerable equilibration time for the model soil moisture changes and have carried out the integration for a substantially longer period (6 years) than did previous studies. They also argued for a substantially smaller change in surface roughness (i.e., degraded brush rather than degraded grassland) as the assumed end-state of forest conversion. An observational attempt to detect effects of Amazon deforestation [ Chu et al., 1994] gave a negative result. Scaling the effects inferred in complete deforestation studies to the relatively small amount of deforestation that has occurred up to now suggests an effect too small to be detectable above natural fluctuations. Moreover, understanding of what are the natural, long time-scale fluctuations within regions is weak, and the effect of patchy partial deforestation would not necessarily simply scale from results for complete deforestation.
One of the popular arguments for the importance of the Amazon forest for rainfall is its contribution of approximately 50% recycled rainfall. Brubaker et al. [1993] have analyzed the water budget of the Amazon and other regions and find only 30% of the water is recycled. This disagreement seems to lie, in part, in definitions. Their definition would give a recycling coefficient of 33% if the net water flux into the basin were equal to the basin evapotranspiration, as required for average precipitation to be double runoff (as approximately observed). Past authors have argued these conditions, and the smallness of transport out of the basin because of surrounding highlands implies a 50% recycling coefficient. Isotopic analyses have seemingly supported this result, but the correct answer depends on what fraction of land evapotranspiration is transported outside of the basin. Further analyses have been carried out by Eltahir and Bras [1994] who derive a recycling coefficient between 25% and 35% and explicitly argue that much of the evapotranspired water is carried out of the basin. They do not give details of their data analysis.
Further studies have addressed the question of the dependence of Sahelian rainfall on surface vegetation cover, soil moisture, and albedo. Xue and Shukla [1993] have carried out GCM sensitivity studies for this region that support the earlier inferences of positive feedbacks between albedo increase/vegetation decrease and reduced ET. Lare and Nicholson [1994] have also studied this question with observations and diagnostic modeling. They find an association between dry Augusts over the Sahel and a decrease in the strength of the African Easterly Jet, a system responsible for bringing rain-producing disturbances. Dirmeyer [1994] has studied, with numerical simulations and an idealized continent, the response of precipitation over the summer period to dormant vegetation or initial dry soil. He finds that either can lead to a drought.
Modeling of the summer Indian monsoon is another LA-I issue that has been addressed. Fennessy et al. [1994] have studied the sensitivity of precipitation in this system to various land boundary conditions. They find the greatest sensitivity to the prescription of orography, some dependence on initial soil moisture (for 90-day integrations started from 2 June), and little dependence on the details of the prescribed vegetation cover. Mechanisms of atmospheric heating and the boundary layer over the Tibetian Plateau have been addressed by Yanai and Li [1994].
Variations of precipitation amounts and distributions over a hypothetical tropical continent, in response to varying degrees of prescribed surface wetness, have been examined by Cook [1994]. She shows that with an idealized continent, longitudinal patterns of precipitation are obtained similar to those observed and that these patterns shift and precipitation is reduced as greater dryness is assumed.