Clouds affect radiation both through their three-dimensional geometry and the amount, size and nature of the hydrometeors which they contain. In climate models these properties translate into cloud cover at different levels, cloud water content (for liquid water and ice) and cloud droplet (or crystal) equivalent radius. The interaction of clouds and radiation also involves other parameters (asymmetry factor of the Mie diffusion) which depend on cloud composition, and most notably on their phase. The subtle balance between cloud impact on the solar short-wave (SW) and terrestrial long-wave (LW) radiation may be altered by a change in any of those parameters. In response to any climate perturbation the response of cloudiness thereby introduces feedbacks whose sign and amplitude are largely unknown. While the SAR noted some convergence in the cloud radiative feedback simulated by different models between two successive intercomparisons (Cess et al., 1990, 1995, 1996), this convergence was not confirmed by a separate consideration of the SW and LW components.
Schemes predicting cloudiness as a function of relative humidity generally show an upward displacement of the higher troposphere cloud cover in response to a greenhouse warming, resulting in a positive feedback (Manabe and Wetherald, 1987). While this effect still appears in more sophisticated models, and even cloud resolving models (Wu and Moncrieff, 1999; Tompkins and Emanuel, 2000), the introduction of cloud water content as a prognostic variable, by decoupling cloud and water vapour, has added new features (Senior and Mitchell, 1993; Lee et al., 1997). As noted in the SAR, a negative feedback corresponding to an increase in cloud cover, and hence cloud albedo, at the transition between ice and liquid clouds occurs in some models, but is crucially dependent on the definition of the phase transition within models. The sign of the cloud cover feedback is still a matter of uncertainty and generally depends on other related cloud properties (Yao and Del Genio, 1999; Meleshko et al., 2000).
Most GCMs used for climate simulations now include interactive cloud optical properties. Cloud optical feedbacks produced by these GCMs, however, differ both in sign and strength. The transition between water and ice may be a source of error, but even for a given water phase, the sign of the variation of cloud optical properties with temperature can be a matter of controversy. Analysis from the ISCCP data set, for example, revealed a decrease of low cloud optical thickness with cloud temperature in the sub-tropical and tropical latitudes and an increase at middle latitudes in winter (Tselioudis et al., 1992; Tselioudis and Rossow, 1994). A similar relationship between cloud liquid water path and cloud temperature was found in an analysis of microwave satellite observations (Greenwald et al., 1995). This is opposite to the assumptions on adiabatic increase of cloud liquid water content with temperature, adopted in early studies, and still present in many models. Changes in cloud water path reflect different effects which may partially compensate, such as changes in cloud vertical extension or cloud water content. The role of low cloud optical thickness dependence on climate was tested in 2xCO2 experiments using the GISS GCM, in which simulations with fixed or simulated cloud optical properties were compared (Tselioudis et al., 1998). In spite of a low impact on global sensitivity, these results showed a strong cloud impact on the latitudinal distribution of the warming. High latitude warming decreased while low latitude warming increased, resulting in a large decrease in the latitudinal amplification of the warming (by 20% in the Northern Hemisphere and by 40% in the Southern Hemisphere).
Since the SAR, there has been much progress in the use of simplified tropical models to understand the impact of cloud feedbacks, and to address the issue of whether cloud feedbacks impose an upper limit on SST. This problem has been addressed in simplified models that maintain consistency with the whole-tropics energy budget (Pierrehumbert, 1995; Clement and Seager, 1999; Larson et al., 1999). Diagnostic studies have suggested both destabilising (Chou and Neelin, 1999) and stabilising (Lindzen et al., 2001) cloud feedbacks. None of this supports the existence of a strict limitation of maximum tropical SST of the sort proposed by Ramanathan and Collins (1991), which has been criticised on the grounds that it does not respect the whole-tropics energy budget, and that it employs an incorrect means of determining the threshold temperature for convection. It is beyond question that the increased cloudiness prevailing over the warmer portions of the Pacific has a strong effect on the surface energy budget, which is fully competitive with the importance of evaporation. Determination of SST, however, requires a consistent treatment of the top-of-atmosphere energy budget, and cannot be effected with reference to the surface budget alone. This does not preclude the possibility that other cloud feedback mechanisms could have a profound effect on tropical SST, and in no way implies that cloud representation is inconsequential in the tropics. Meehl et al. (2000) illustrate this point when they show how a change from a diagnostic prescription of clouds in the NCAR atmospheric GCM to a prognostic cloud liquid water formulation changes the sign of net cloud forcing in the eastern tropical Pacific and completely alters the nature of the coupled model response to increased greenhouse gases.
Other reports in this collection