An Improved Strategy for the Evaluation of Cloud Parameterizations in Gcms

OCTOBER 2003 AMERICAN METEOROLOGICAL SOCIETY | G reat effort is continually being made to improve the parameterization of clouds in global circulation models (GCMs), and as a result the physical realism of cloud parameterizations has been improved considerably. Almost all GCMs now use a prognostic equation to represent the evolution of cloud condensate (e.g., Sundqvist 1978; del Genio et al. 1996; Fowler et al. 1996). The treatment of cloud cover is more diverse, ranging from simple diagnostic relations (e.g., Sundqvist 1978; Slingo 1987; Smith 1990) to a fully prognostic treatment (e.g., Tiedtke 1993). Through evaluation studies, model developers want to be able to expose flaws in these parameterizations, and, if possible, reveal reasons for those flaws. Evaluating the “model clouds” against their real-life counterparts, however, is becoming increasingly difficult because the models themselves are becoming increasingly complex. Current GCMs are among the most complex of physical models, not only because they describe a large number of processes but also because they include nonlinear interactions. Many studies have evaluated the representation of clouds and their radiative effects in GCMs. Their approaches vary widely, but most are evaluations of either the model climate or case studies. As we shall see, there is an unfortunate gap between these two approaches. In the numerous studies using one or more of the techniques we outline below, a lack of coherence in the application of these techniques to the same model is clearly visible. Most likely this is because each of the techniques as such requires substantial resources. However, this lack of coherence-a lack of strategy-when evaluating cloud parameterizations, has led to a considerable dilution of efforts. Results of a number of recent studies will be used in an illustrative fashion to highlight some of the evaluation techniques in use today and to propose how a technique of compositing by dynamical regime might bridge the gap between model climate and case study and thereby provide new insight into cloud parameterization. Most of the studies used here have been carried out with various versions of the European Centre for Medium-Range Weather Forecasts (ECMWF) global forecast model, which applies the cloud parameterization of Tiedtke (1993) with recent modifications described in Jakob (2001). After assessing the current evaluation techniques we will propose AN IMPROVED STRATEGY FOR THE EVALUATION OF CLOUD PARAMETERIZATIONS IN GCMS