| Home > Newsevents > Training > Rcourse_notes > PARAMETRIZATION > CLOUDS > |
Representation of clouds in large-scale
models
May 1992
By M. Tiedtke
| Table of contents | |||||||
|
Table of contents
Next Section
1 . Introduction
In part 1 of the lecture notes we discussed the parametrization of non-convective condensation processes. In this part we address the problem of representing cloud fields for the purpose of radiation calculations. Because clouds are represented poorly in large-scale models (e.g. lack of subgrid-scale cloudiness) clouds are normally diagnosed independently for radiation calculations. It is only recently that there have been attempts to treat clouds in a unified way. In this part we shall discuss the various cloud parametrizations developed for large-scale models, i.e. diagnostic schemes as well as prognostic schemes, including a new cloud scheme developed at ECMWF.
Representation of clouds and cloud-related processes is a central issue in large-scale modelling because clouds are among the most important regulators of the weather and climate of the Earth's atmosphere. Their importance for the atmospheric circulation, in particular at time-scales of climate change, has been recognized for many years but, because clouds interact in many ways with atmospheric processes such as turbulence, the larger-scale circulation and radiation, their role is still poorly understood. The lack of knowledge of cloud processes has also impaired the parametrization of clouds in climate and forecast models and, as a consequence, results from climate models are rather uncertain (e.g. Cess et al., 1989; Mitchell, 1989) and operational weather forecasts are affected in their quality. This is despite the considerable effort that has gone into improving the representation of clouds in models, which has progressed from very simple to rather complex schemes during the last two decades.
Initially their distribution was externally prescribed, often from zonal-mean climatological values ignoring their variation in space and time (e.g. Manabe and Holloway, 1971). Model-generated cloud fields became more realistic as clouds were tied to atmospheric properties such as relative humidity, vertical velocity and static stability. Diagnostic schemes of this type are successful in reproducing some of the gross features of global cloudiness but they lack a sound physical basis and, in particular, do not represent the interaction between clouds and the hydrological cycle of the model. However, because diagnostic schemes are simple and yet rather successful they are widely used in large-scale models (e.g. Slingo, 1987).
More recently, in searching for a unification of all cloud-related processes, schemes have been developed which use a liquid-water model variable. In these schemes the cloud evolution is linked directly to the model's processes-dynamics, radiative transfer, hydrology, convection and turbulence. Following Sundquist (1978), various prognostic cloud schemes have been developed for large-scale models (e.g. Hense and Heise, 1984; Le Treut and Li, 1989; Smith, 1990). The advantage of the unified approach is that it provides a realistic physical basis that is consistent with the rest of the model and accounts directly for the various feedbacks involving clouds (particularly important for climate studies). However, the accuracy of cloud parametrizations with this approach depends critically on the realistic treatment of advective transports of cloud variables, subgrid-scale processes, cloud microphysics and cloud optical properties. Considering the various processes we note that there exist large uncertainties in representing some of the processes.
The representation of advective transports poses severe problems for conventional numerical schemes as the presence of discontinuities typical for cloud fields causes large truncation errors which can lead to unrealistic values of cloud-water content. Recently, numerical methods have been developed which avoid this type of error (e.g. Williamson and Rasch, 1989), but these methods have not yet been applied regularly to the advection of cloud variables. Instead, advective transports of cloud water are often neglected; this may be justified at very low resolution and for warm clouds, but is less justifiable at higher resolution and for cold clouds, such as anvil clouds that can persist for a day or more while they drift over large distances (Ludlum, 1980).
Many clouds encountered on the globe depend strongly on processes which cannot be resolved in large scale models but are represented by means of parametrization which is still uncertain particularly for cumulus convection and boundary layer turbulence. There is also the additional difficulty to provide a realistic interface between the cloud scheme and the schemes for convection and boundary layer turbulence.
Cloud microphysical processes are represented crudely in current models. Because of limited computer resources detailed simulation of the microphysics of clouds (i.e. evolution of droplet spectra, nuclei activation, growth of precipitation drops etc.) is not practical. Instead, cloud processes are represented using a bulk water parametrization technique, where the liquid phase is subdivided into cloud water and precipitation water. Cloud water is assumed to form when the relative humidity exceeds a specified threshold value and precipitation processes are described by simple parametrizations for warm and cold clouds that are derived empirically. Uncertainty in the parametrization of precipitation is of particular importance as predicted cloud water content is highly sensitive to tunable parameters (Sundquist, 1978).
Cloud optical properties are currently specified in terms of liquid-water path on the basis of radiative-transfer theory for spherical cloud droplets. The optical properties of ice clouds are difficult to represent as they depend on the shape, size and orientation of ice crystals, which are not provided by numerical models. A particularly difficult problem is the calculation of grid-mean radiances in cases where various types of clouds (e.g. stratiform clouds and convective clouds at different stages of development) of different optical properties occur simultaneously within a grid area. As numerical models provide only cloud cover and grid averages of cloud water/ice content as input for radiation calculations, present radiation schemes consider only averaged conditions which may not provide sufficiently accurate grid-mean radiances.
The difficulties and uncertainties mentioned above impair cloud forecasts in large-scale models, but progress is expected to come from future cloud studies such as the GCSS (GEWEX cloud system study). The best framework to incorporate forthcoming results appears to be a prognostic cloud scheme where cloud related processes are treated in a unified way. The demands of such a scheme are high and are hardly met by present schemes, which do not treat all cloud processes in a fully consistent way and often contain empirical features which seem undesirable and unnecessary. In pursuing a more rigorous approach, a cloud scheme has been developed at ECMWF which goes beyond present schemes in important aspects. The scheme is described below together with diagnostic schemes and more conventional prognostic schemes.
Training Course Notes Front Page
Table of contents
Next Section
 |
11.06.2002 |
 |
© ECMWF |
