Global climate models are often employed to investigate whether and how soil moisture anomalies affect weather and climate both in the current and in a potentially different future climate. A recent model intercomparison (Global Land-Atmosphere Coupling Experement, GLACE) demonstrated that the degree of land-atmosphere interaction varies widely between current state-of-the-art AGCMs (Koster et al 2002, 2004). Models with different inherent coupling strengths can lead to vastly different conclusions about climate sensitivity. Land-atmosphere coupling strength, or the extent to which a precipitation-induced soil moisture anomaly influences the overlying atmosphere and thereby the evolution of weather and the generation of precipitation, is reasonably strong in the Community Atmosphere Model (CAM2/CLM2) but is very weak in the Hadley Centre modeling system (HadAM3/MOSES2).

Through direct comparison of the coupling mechanism in these two models, we can evaluate what aspects of the model control the degree of land-atmosphere coupling. Key aspects of the indirect soil moisture-precipitation feedback are evaluated and compared. It is found that differences in the simulation of the diurnal cycle, and related differences in precipitation frequency and soil moisture variability, appear to be able to explain the large differences in coupling strength. In particular, it appears that the simulation of the diurnal evolution of boundary layer moist static energy and its relationship to moist convection are crucial factors that govern the strength of the soil moisture-precipitation feedback. In CAM2/CLM2, over wet soils boundary layer moist static energy grows steadily during the day, fed by strong evaporation into a only slowly deepening boundary layer, leading to heavy convective precipitation. In contrast in HadAM3/MOSES2, soil moisture exerts virtually no control on boundary layer moist static energy and henceforth convection is essentially unaffected by soil moisture and land-atmosphere coupling strength is low.