Ice Clouds in Numerical Weather Prediction Models: Progress, Problems, and Prospects

Abstract

The properties of cirrus, as well as the role ice clouds play in the atmosphere, have been extensively described in the previous chapters. To represent the effects of cirrus in atmospheric models, several intimately linked processes need to be described. These processes include the generation and dissipation of ice clouds as well as their interaction with the radiative fluxes throughout the atmosphere. In this chapter the cloud parameterization aspects of this problem (i.e., the treatment of the generation and dissipation of ice clouds), are discussed in the context of global numerical weather prediction (NWP) models. Aspects of the radiative transfer in ice clouds can be found in chapter 13. The main focus of the current chapter is on the cloud parameterization used in the global forecast model of the European Centre for Medium-Range Weather Forecasts (ECMWF). This parameterization will serve as an example in highlighting the progress made, the problems encountered, and the prospects for improving the representation of ice clouds in atmospheric models. The principles of representing clouds in global NWP models are identical to those in general circulation models (GCMs) used for climate research (see chapter 15). Although ice clouds are the focus of this book, a substantial part of this chapter will be concerned with the overall treatment of clouds in numerical models of the atmosphere. In fact, many models used in NWP today distinguish ice clouds from mixed-phase and water clouds only as a function of temperature. Cloud parameterizations in GCMs have evolved rapidly over the last few years. Section 16.2 is a general overview of the progress made. Section 16.3 will describe the cloud parameterization that is currently used in the ECMWF forecast model as a specific example for a state-of-the-art cloud parameterization in NWP. General aspects of the simulation of ice clouds with this model will be presented. GCM simulations of the atmosphere are very sensitive to the treatment of clouds in general (Senior and Mitchell 1993; Rasch and Kristjansson 1998) and to assumptions about cloud ice in particular (Fowler et al. 1996; Jakob and Morcrette 1995). Section 16.4 gives an example of those sensitivities and the model design problems that can arise when model sensitivities exist in combination with a lack of observations, as noted for cloud ice by Stephens et al. (1998).