The interplay of buoyancy, moisture and condensation on scales ranging from millimetres to tens of kilometres is the defining physical feature of atmospheric convection, and is the source of much of the challenge in representing convection in climate models. Deep convection is in large measure responsible for the very existence of the troposphere. Air typically receives its buoyancy through being heated by contact with a warm, solar-heated underlying surface, and convection redistributes the energy received by the surface upwards throughout the troposphere. Shallow convection also figures importantly in the structure of the atmospheric boundary layer and will be addressed in Section 18.104.22.168.
Latent heat release in convection drives many of the important atmospheric circulations, and is a key link in the cycle of atmosphere-ocean feedbacks leading to the ENSO phenomenon. Convection is a principal means of transporting moisture vertically, which implies a role of convection in the radiative feedback due to both water vapour and clouds. Convection also in large measure determines the vertical temperature lapse rate of the atmosphere, and particularly so in the tropics. A strong decrease of temperature with height enhances the greenhouse effect, whereas a weaker temperature decrease ameliorates it. The effect of lapse rate changes on clear-sky water vapour feedback has been studied by Zhang et al. (1994), but the significance of the lapse rate contribution (cf. item (5) of the SAR, Technical Summary) has been somewhat exaggerated through a misinterpretation of the paper. In fact the variation in lapse rate effects among the models studied alters the water vapour feedback factor by only 0.25 W/(m2K), or about 10% of the total (Table 1 of Zhang et al., 1994).
There is ample theoretical and observational evidence that deep moist convection locally establishes a "moist adiabatic" temperature profile that, loosely speaking, is neutrally buoyant with respect to ascending, condensing parcels (Betts, 1982; Xu and Emanuel, 1989). This adjustment happens directly at the scale of individual convective clouds, but dynamical processes plausibly extend the radius of influence of the adjustment to the scale of a typical GCM grid box, and probably much further in the tropics, where the lapse rate adjusts close to the moist adiabat almost everywhere. All convective schemes, from the most simple Moist Adiabatic Adjustment to those which attempt a representation of cloud-scale motions (Arakawa and Schubert, 1974; Emanuel, 1991), therefore agree in that they maintain the temperature at a nearly moist adiabatic profile. Moist Adiabatic Adjustment explicitly resets the temperature to the desired profile, whereas mass flux schemes achieve the adjustment to a near-adiabat as a consequence of equations governing the parametrized convective heating field. The constraint on temperature, however, places only a limited constraint on the moisture profile remaining after adjustment, and the performance in terms of moisture, clouds and precipitation may be very variable.
Since the SAR, a variety of simulations of response to CO2 doubling accounting for combinations of different parametrizations have been realised with different models (Colman and McAvaney, 1995; Yao and Del Genio, 1999; Meleshko et al., 2000). The general effects of the convection parametrization on climate sensitivity are difficult to assess because the way a model responds to changes in convection depends on a range of other parametrizations, so results are somewhat inconsistent between models (Colman and McAvaney, 1995; Thompson and Pollard, 1995; Zhang and McFarlane, 1995). There is some indication that the climate sensitivity in models with strong negative cloud feedback is insensitive to convective parametrization whereas models with strong positive cloud feedback show more sensitivity (Meleshko et al., 2000).
Other reports in this collection