next up previous
Next: PARALLEL ALGORITHMS Up: Parallel Community Climate Previous: INTRODUCTION

History of the NCAR Community Climate Model

Over the last decade, the NCAR Climate and Global Dynamics (CGD) Division has provided a comprehensive, three-dimensional global atmospheric model to university and NCAR scientists for use in the analysis and understanding of global climate. Because of its widespread use, the model was designated a Community Climate Model (CCM). The original version of the CCM (CCM0A) was based on the Australian spectral model developed by W. Bourke, B. McAvaney, K. Puri, and R. Thurling [6] [26] and was described in [37]. An important broadening of the concept of the NCAR community model occurred in late 1981 with NCAR's decision to utilize the same basic code for global forecast studies (both medium- and long-range) and for climate simulation. Economy and increased efficiency could then be achieved by documenting and maintaining only one set of modular codes. The use of one basic model for both forecasting and climate studies was also seen to have great potential scientific value since a major part of medium-range (one- to two-week) forecast error is due to the drift toward a model climate which differs from that of the atmosphere. Thus, improvements in the climate aspects of the model should lead to improvements in forecasts. Similarly, many physical parameterizations are deterministic rather than statistical in the sense that they are based on the details of the current model state rather than on some past statistical properties. Thus, performance aspects of parameterized physics can be studied, improved, and verified by examining them in a forecast mode.

Because of the extension of the role of the CCM to include forecast studies as well as climate studies, and because of the expected widespread use for both purposes by university as well as NCAR scientists, a versatile, modular, and well-documented code became essential. The initial version designated CCM0B was developed to meet these requirements. This code grew out of an adiabatic, inviscid version of the spectral model developed at the European Centre for Medium Range Weather Forecasts (ECMWF) by A.P.M. Baede, M. Jarraud, and U. Cubasch [2] to which physical parameterizations and numerical approximations matching those of CCM0A were added. The physical parameterizations included the radiation and cloud routines developed at NCAR [28] and convective adjustment, stable condensation, vertical diffusion, surface fluxes, and surface-energy-balance prescription developed at the Geophysical Fluid Dynamics Laboratory (GFDL) [32] [24] [33] [19]. The vertical and temporal finite differences matched those of the Australian spectral model [6]. The resulting model code, designated CCM0B, was described in a series of technical notes which included a User's Guide [31], a description of model subroutines [39], a detailed description of the continuous algorithms [38], and circulation statistics from long January and July simulations [42].

The advantages of the community model concept, in which many scientists use the same basic model for a variety of scientific studies, were demonstrated in workshops held at NCAR in July 1985 [1], July 1987 [44], and July 1990 [45]. Fundamental strengths and weaknesses of the model have been identified at these workshops through the presentation of a diverse number of applications of the CCM. Much constructive dialogue has taken place between experts in several disciplines at these meetings leading to continued improvements in the CCM with each release.

CCM0B was followed with CCM1 in July of 1987 and included a similar set of detailed technical documentation [40] [5] [4] [43] [16]. Substantial changes were incorporated in the radiation scheme, including a new solar albedo parameterization accounting for the solar zenith-angle dependence of albedo on various surface types, improvements to the absorption of solar radiation by HO and O, improvements to the long wave absorptance algorithms for HO, CO and O, changes to account for the liquid water content of stratiform clouds in determining their emissivity, and incorporation of a new finite-difference scheme in the long wave part of the radiation model (see [22]). The vertical finite-difference approximations were modified to conserve energy without adversely affecting the model simulations, and frictional heating was included so that the momentum diffusion produced a corresponding heating term in the thermodynamic equation. The latter two improvements resulted in the energy in the model being conserved to the order of one W m and moisture to one-hundredth W m energy equivalent over 90-day periods. The horizontal diffusion was modified to a form in the troposphere and included a partial correction for evaluating the operator on pressure surfaces rather than sigma surfaces. The local moisture adjustment was generalized to provide for a global horizontal borrowing [30] in a conserving manner. The vertical diffusion was converted to a nonlinear form for which the eddy-mixing coefficient depended on local shear and stability. The diffusion was applied throughout the atmosphere rather than only below 500 mb as done in CCM0B, which eliminated the need for a dry convective adjustment in the troposphere. The surface drag coefficient was made a function of stability following Deardorff [8] and the equation of state was modified to formally account for moisture in the atmosphere (i.e., virtual temperature was used where appropriate and the variation with moisture of the specific heat at constant pressure was accounted for). In addition to the above changes to the physics, CCM1 included new capabilities such as a seasonal mode in which the specified surface conditions vary with time, and an optional interactive surface hydrology [7] which followed the formulation presented by Manabe [24]. Since the CCM1 could also be used as a global forecast model, codes to prepare initial data in the CCM history tape format from analyzed observed atmospheric data, such as FGGE Level IIIb analyses [25], and codes to perform nonlinear normal mode initialization [13] [12] were made available.

As a result of the biennial CCM workshops mentioned earlier, the underlying philosophy of the CCM was modified. The original intent was to provide a stable, robust model applicable to a variety of problems. Thus the most recent developments in model physics were deliberately not included in the physical parameterizations in order to provide stable, well known algorithms. This approach leads to more straightforward interpretation of experimental results. The discussions in the workshops highlighted the strengths of this approach, but also pointed out the need for a state-of-the-science model to address many of the very important climate questions being raised today. The decision was made that the next version of the CCM should be brought up to date in all its aspects. Thus the most recent version of the CCM, CCM2, which is expected to be released during the summer of 1991, incorporates the most ambitious set of changes to date.

The bulk of the effort in the NCAR Climate Modeling Section over the last several years has been to improve the physical representation of a wide range of key climate processes in the CCM, including clouds and radiation, moist convection, the planetary boundary layer, and transport. The resulting changes to the model have resulted in a significantly improved simulation and fundamentally better climate model. On the parameterized physics side, changes include the incorporation of a diurnal cycle, along with the required multilayer heat capacity soil model, and major improvements to the radiation scheme, including a -Eddington solar scheme (18 spectral bands), a new cloud albedo parameterization, a new cloud emissivity formulation using liquid water path length, a new cloud fraction parameterization, and a Voigt correction to infrared radiative cooling in stratosphere. The moist adiabatic adjustment procedure has been replaced with a stability-dependent mass flux representation of moist convection, and an explicit planetary boundary layer parameterization is now included, along with a modified gravity-wave drag parameterization which introduces changes in the generation and vertical distribution of momentum drag as well as providing the framework for a longer-term non-isotropic formalism.

On the dynamics side, a semi-Lagrangian transport scheme is now the default for water vapor as well as an arbitrary number of other scalar fields (e.g., cloud water variables, chemical constituents, etc.) and the vertical coordinate makes use of a hybrid (or generalized ) formulation. The model has been developed for a standard horizontal spectral resolution of T42 (2.8 degrees by 2.8 degrees transform grid) with 18 vertical levels and a top at approximately 2.9 mb. The entire model code is also being entirely rewritten with three major objectives: much greater ease of use and modification; conformation to a plug-compatible physics interface; and the incorporation of single-job multitasking capabilities.

CCM2 provides the basis for a large body of experimental and developmental efforts by a large community of university and NCAR climate investigators, many of whom may not be directly involved in the CHAMMP initiative. Because of the community nature of the enterprise, new methods and process modules are continually emerging. The new methods will be incorporated in future releases and versions of the model as seems appropriate for computer efficiency and the requirement for increased capabilities.

next up previous
Next: PARALLEL ALGORITHMS Up: Parallel Community Climate Previous: INTRODUCTION

John B. Drake
Wed May 15 09:51:22 EDT 1996