The goal of the Climate System Model (CSM) project is to develop a comprehensive modeling system for climate, including both physical and biogeochemical aspects, and to apply appropriate configurations of the model to important scientific problems. The CSM project has always been committed to using comprehensive coupled models without flux adjustments and has successfully completed several simulations without significant trends in the sea surface temperatures. Major successes of the project over the last year have included: significant reduction in the rate of deep ocean drift; a simulation of the twentieth century climate; and the increasingly active participation of scientist from outside of NCAR in the development and application of the CSM.
The CSM working group structure is working reasonably well, with several active members both from inside NCAR and from the wider community in each group. A new biogeochemistry working group has been established, co-chaired by Scot Doney (NCAR) and Inez Fung (UC Berkeley), to foster modeling activities in marine and terrestrial biogeochemistry and ecosystems. The atmospheric biogeochemistry and anthropogenic climate change working group (S. Solomon and J. Kiehl, co-chairs) has been renamed the chemistry and climate change working group, consistent with its emphasis on atmospheric chemistry and on the climate of the past and next centuries.
A new version of the CSM (CSM1.2) was publicly released in July 1998. CSM1.2 is the third major configuration of the CSM. Both active and data models are available for each of the components (atmosphere, ocean, land, and sea ice). Standard resolutions are T31 & T42 for the atmosphere and land components, and a nominal 3x3 and 2x2 degree resolution for the sea-ice and ocean components. Other resolutions are also possible.
CSM1.2 implements the same numerical algorithms as the previous versions of the CSM, with many improvements for more ease of use and portability. The land model is now a separate executable, no longer contained within the atmospheric executable, which outputs independent netCDF files. The ocean model (NCOM1.4) also outputs netCDF files, leaving the atmospheric model as the only component still using a private data format. The latter is for efficiency reasons on machines with relatively small memories (such as the Cray C90). This distribution of the code is meant to be compiled and run on a Cray machine at NCAR. These restrictions are due to issues involving access to data files and linking to libraries. With some modest additional effort these issue can be resolved. For example, this code can be and has been adapted to run on Cray machines outside of NCAR and on both SGI and NEC machines. It is anticipated that the next release of the code will be considerably more portable.
A new feature included in CSM1.2 is a catchment basin runoff model within the land component, developed in collaboration with J. Famiglietti (University of Texas at Austin). Runoff from the land model is routed to 19 catchment basins. Each land cell can drain into at most 4 basins, on the basis of area weights determined from a 1/2 degree basin mask and wet surfaces (irrigated croplands, lakes, swamps, and glacial ice) are assumed to be mass balanced. At the moment, this term is diagnostic, since the runoff calculation is now performed in the land model and passed through the coupler to the ocean, but it is not yet used by the ocean model. Testing of the incorporation of runoff in the ocean model is nearly complete and coupled runs using the runoff calculation will begin soon.
Several new features have been developed for CSM over the past year and some are being incorporated in the climate of the twentieth and twenty first century simulations discussed below. The prognostic cloud water parameterization of Rasch and Kristjansson (1998) has been incorporated into the coupled simulations. In addition, the direct effect of sulphate aerosols have been included in the radiative transfer model. Several alternative formulations for the (indirect) effect of aerosols on the cloud drop size distribution have been tested by J Kiehl and T. Schneider and multi-decadal coupled simulations have been performed. However, the indirect effects are highly uncertain and are currently being omitted from transient climate scenarios. Sulphate aerosol distributions are obtained from the interactive sulphate chemistry model of Mary Barth (ACD) and Rasch. Three dimensional aerosol distributions can be either specified from previous simulations, or can be solved for interactively in the coupled model. Finally, the concentrations of CH4, N2O, CFC11, and CFC12, (the principal greenhouse gases except CO2) are solved in the model. The surface concentrations (not emissions) of these gases are specified and loss frequencies are applied, derived from the photochemical model of Solomon and R. Portmann of the NOAA Aeronomy Laboratory. The configuration and testing of CSM with all of these features was performed by B. Boville, L. Buja, M. Vertenstein, and B. Eaton (CMS).
The aerodynamic roughness length of sea ice used in the original CSM simulations (40 mm) was found to be unrealistically large by J. Weatherly (CCRS) and B. Briegleb (CMS) and the sensitivity to using a smaller value (0.5 mm), appropriate for relatively smooth first year ice, was tested. Note that the drag coefficient depends on the logarithm of the roughness length, so this change corresponds to a factor of 4 change in the surface stress over sea ice. This change has a modest impact on the ice distribution in either hemisphere. However, it significantly reduces the rate of Antarctic deep water formation, which remains too large but is now closer to that in the real ocean. In consequence, the drift in deep ocean salinity is reduced by an order of magnitude and the drift in deep ocean temperatures by a lesser amount. The smaller value is now used in all CSM simulations.
The tropical simulation of the ocean model has been improved significantly by reducing the vertical diffusion and incorporating an asymmetric horizontal diffusion tensor with greatly reduced cross flow diffusion (W. Large et al, OS). These changes result in a greatly improved simulation of the equatorial undercurrent in the Pacific and also of the surface countercurrents. The meridional resolution of the ocean model has also been increased to 0.6 degrees between 10S and 10N, but this has a much smaller impact on the simulation. The new ocean version has not yet been used in coupled simulations, but will be incorporated in the near future.
The third annual CSM workshop, held in Breckenridge Colorado in June 1998, was attended by over 160 scientists. The workshop included plenary sessions and meetings of the CSM working groups. Sessions were staggered so that participants could attend more than one working group. The workshop was followed by meetings of the CSM Scientific Steering Committee (SSC) and the CMAP Scientific Advisory Council. Results of a wide range of parameterization and other model experiments by NCAR, university, and other scientists were presented. Preliminary proposals for several improvements in the next generation of CSM component models were discussed in the working group meetings. Several of these proposed improvements have already been tested in coupled simulations, including the prognostic cloud water and sulphate aerosol mentioned above. Other proposals included: a new long wave radiation parameterization (M. Iacono, Atmospheric and Environmental Research Inc.); a generalized cloud overlap parameterization (J. Bergman, NOAA Climate Diagnostics Center); a common land model (R. Dickinson, University of Arizona and G. Bonan, CMS); plastic-viscous and elastic-plastic-viscous sea ice rheology (Briegleb, Weatherly, and Large); an ocean bottom boundary layer parameterization (Hecht,OS); and an asymmetric ocean diffusion tensor (Large et al, OS)
A major result of the workshop was an increased focus on improving the Tropical Pacific Ocean region simulation in the next generation coupled model. Some of the major tropical deficiencies in the ocean have already been addressed (see above). Unfortunately, the problems in the atmosphere appear to be intimately tied to the interaction of the physical parameterizations in the atmosphere (boundary layer, convection, clouds, and radiation). All proposed model improvements are being scrutinized for their impact on this region and some progress is expected, but no quick fix is available which is suitable for long term climate simulation.
The first CSM simulation of the 20th century climate was completed by the Chemistry and Climate Working Group. A new spinup and short (40 year) control simulation of the coupled system was performed for 1870 conditions. A transient forcing simulation was then performed using reconstructions of atmospheric concentrations for sulphate aerosol, CO2, O3, CH4, N2O, CFC11, and CFC12. The latter 4 gases are advected in the CSM, as described above, and CFC11 concentrations were scaled to account for the effects of other halocarbons. The globally averaged temperature increases by about 0.6 K between the late 19th century and the 1990s, with most of the increase occuring since 1970, in agreement with observations. The CSM simulation shows comparable levels of variability to the observed record prior to 1920, but does not capture the observed maximum in the 1940s, which is believed to be caused by increased solar irradiance. This hypothesis will be tested by including a reconstruction of the solar irradiance since 1870 in the next 20th century simulation. Indirect effects of sulphate aerosols and volcanic aerosols will also be incorporated in the near future.
The effect of the simulated global temperature increase after 1970 can be clearly seen in the evaporation and precipitation, which also increase after 1970. The effect on the runoff rates is less clear. Although an increase in global runoff is seen at the end of simulation, it is not outside the range of variability found earlier in the simulation, before the temperature and precipitation increase significantly. While there is some evidence of increasing snow accumulation on Antarctica and increased runoff in a few basins, others show no significant change.
The priority of 21st century climate scenarios was greatly increased late in FY98 to respond to requests for scenarios for the US national climate assessment. The 20th century simulation described above will serve as a realistic initial condition for 21st century simulations. Specifications for two forcing scenarios were completed by T. Wigley and S. Smith (CAS) and Solomon, including the same gases as used in the 20th century simulations. Consistent trace gas concentrations and sulfur dioxide emissions are given for a "business as usual" and for a plausible policy limited emissions scenario. Since the geographic distributions of anthropogenic sulfur dioxide emissions is expected to change with time, the aerosol model will be solved interactively in these scenarios. Testing of the coupled model was completed and the scenario simulations were started at the end of FY98.