Atmospheric General Circulation Model

In 1982, a copy of the UCLA atmospheric general circulation model (AGCM) was generously provided by A. Arakawa for use at the Goddard Space Flight Center. One of the key features of this model was (and is) the use of an explicit planetary boundary layer (PBL) depth, embedded into the vertical structure of the model through the use of a modified sigma coordinate (Suarez et al., 1983; Randall et al., 1985).

From the beginning, a key application envisioned for the transplanted model was studying the processes by which clouds interact with the other components of the climate system. In this context, a major limitation of the model provided by UCLA was that its radiation parameterization was outdated. For this reason, during the mid-1980s, a new radiation parameterization was developed by Harshvardhan et al. (1987). Subsequently the AGCM has in fact been used for many studies of the role of clouds in climate (e.g., Randall et al., 1989; Harshvardhan et al., 1989; Randall et al., 1990; Fowler and Randall, 1994; Fowler et al., 1996). At present, we are testing a new and more accurate radiation parameterization developed by Graeme Stephens and colleagues at CSU. In the late 1980s and early 1990s, the Harshvardhan radiation scheme was adopted by many modeling centers around the world. The parameterization was made freely available by its developers.

A second limitation of the UCLA AGCM of the early 1980s was that it did not include a realistic parameterization of land-surface processes. To remedy this, a project was undertaken with Piers Sellers and colleagues, which ultimately led to the development of SiB, the Simple Biosphere Model, which has been very influential in the land-surface modeling arena (see referenced papers by Sellers and colleagues). Numerous studies have subsequently been carried out in this area (e.g., Randall et al., 1996; Denning et al., 1996). Further work is ongoing in collaboration with A. S. Denning, I. Fung, and others. SiB and its variants are in use at many modeling centers around the world. The parameterization was made freely available by its developers.

In 1988, the AGCM was transplanted to CSU, where it is affectionately known as BUGS. Subsequent work, briefly reviewed below, has involved extensive model development in a variety of areas, and also applications of the model to scientific issues related to climate processes. The work has been carried out by a team led by David Randall and involving central and essential contributions by numerous students and research staff. The fruits of this research have been shared with the community through publications, give-aways of model components, and interactions with the Community Climate System Model project.

Following the upgrading of the model's radiation and land-surface components, attention was turned to the parameterization of both convective and stratiform clouds. The Arakawa-Schubert parameterization was modified to use a prognostic closure (Randall and Pan, 1993; Pan and Randall, 1998), and to allow multiple simultaneous cloud base levels (Ding and Randall, 1998). The prognostic closure is now being used operationally at the Japan Meteorological Agency, and also at UCLA. The parameterization was made freely available by its developers. Further work on convection is ongoing now, involving refinements of the prognostic closure (Lin et al., 2000) and a radical reformulation of the cumulus cloud model used in the parameterization.

A new stratiform cloud parameterization called Eauliq was developed by Fowler et al. (1996); it includes prognostic variables for cloud water, cloud ice, rain, and snow, in addition to water vapor. Eauliq also features direct coupling with the model's cumulus parameterization. Current work is extending Eauliq by incorporating a prognostic cloud amount and a simple parameterization of the effects of mesoscale vertical motions (Randall and Fowler, 1999).

During the early 1990s, graduate student Ross Heikes developed a shallow water model based on an icosahedral grid (Heikes and Randall, 1995 a, b), and using the vorticity, divergence, and mass as the primary prognostic variables on an unstaggered grid (Randall, 1994). Encouraged by the success of this work, we began development of a version of BUGS based on the geodesic grid. This effort reached fruition at the end of the 1990s (Ringler et al., 2000). We have developed a version of BUGS which uses semi-implicit time differencing to allow a long time step. Recently we have completed a version which, through the use of MPI, can run efficiently on computers with many processors. Current work involves the development of a new horizontal differencing scheme, with emphasis on the nonlinear properties of the scheme.

As mentioned earlier, a unique feature of BUGS, inherited from the UCLA AGCM, is the incorporation of an explicit PBL depth with a modified sigma coordinate. In the early 1990s, the PBL parameterization was modified to use a prognostic turbulence kinetic energy (TKE). This is a first step towards the use of multiple prognostic variables that represent various measures of subgrid variability. Our entrainment parameterization makes use of the prognostic TKE. We have also developed a very new approach to the parameterization of the surface fluxes (Zhang et al., 1996), which makes use of the prognostic TKE.

We are currently collaborating with UCLA in generalizing the embedded-PBL paradigm to allow arbitrarily many layers inside the PBL sub-domain of the model. In order to make good use of these layers, we need a realistic parameterization of the turbulent processes which couple the layers together. Our approach to this problem is to combine mass-flux ideas with ideas borrowed from the higher-order-closure modeling literature (Lappen and Randall, 2001 a, b, c).

In addition, we have been following the lead of the UCLA group by testing their isentropic and hybrid theta-sigma vertical coordinates. We currently have working versions of BUGS which use these vertical coordinates with the geodesic grid, but without realistic physical parameterizations. We are currently working to construct a new full-physics version of BUGS which will use the hybrid sigma-theta coordinate in combination with the multi-layer embedded PBL mentioned above.

Finally, we have been collaborating for several years with the ocean modeling team led by Robert Malone at the Los Alamos National Laboratory (LANL), and also with oceanographer Tommy Jensen who was at CSU and is now at the International Pacific Research Center on the campus of the University of Hawaii. These collaborations have resulted in the coupling of our AGCM with POP, the Parallel Ocean Program (a full ocean GCM) developed at LANL, and with TOMS, the upper-ocean model developed by Tommy Jensen. We are using the coupled model to investigate the role of clouds in atmosphere-ocean interactions.

Further information about our AGCM and related work can be found at http://kiwi.atmos.colostate.edu/BUGS/.

In summary, much of what we do is model development. We feel that several components of our model are at the leading edge of the field. These include:

• The parallelized geodesic dynamical core.
• The embedded PBL with a prognostic TKE and the surface flux parameterization of Zhang et al. (1996).
• The prognostic cumulus parameterization.
• The prognostic stratiform cloud parameterization.
• The land-surface parameterization.
• The new radiation parameterization developed by Graeme Stephens and colleagues.