Ensembles of boreal summer GCM integrations for 1987 and 1988 are conducted with and without interactive soil moisture to evaluate the degree of climate drift in the coupled land-atmosphere model system, and to gauge the quality of the specified soil moisture data set — from the Global Soil Wetness Project. It has been found that positive feedbacks between evaporation and rainfall can lead to a drift to either anomalously dry or wet conditions, depending on the systematic errors in precipitation (probably via moisture flux convergence errors). Specified soil moisture can offset some of this error by removing the feedback, and thus the drift, at the expense of violating the surface water balance. Use of specified GSWP soil moisture leads to improved simulations of rainfall patterns, independent of the errors in magnitude, indicating that the GSWP product is useful and can also be used to supply initial conditions to fully coupled climate integrations. Integrations using specified soil moisture from the opposite year suggest that the interannual variability in the GSWP data set is significant, and contributes to the quality of the simulation of precipitation above what would be possible with only a mean annual cycle climatology of soil moisture.
Ensembles with different data sets of global vegetation and soil distributions and properties are also compared to evaluate how climate simulations are affected by errors and uncertainty in the specification of land surfaces characteristics