To gain a better understanding of atmospheric conditions, we simulate global-scale climatic processes using global circulation models (GCMs), which are computer simulations of mathematical models of how the atmosphere operates and interacts with the hydrological cycle (Wright et ah, 1993). GCM experiments have provided pertinent information about relationships between atmospheric circulation and mechanisms of climate forcing such as land uplift (Ruddiman and Kutzbach, 1990), vegetation cover (Foley et ah, 1994), volcanic activity (Bryson, 1989), desert formation (Joussaume, 1989), ocean circulation (Lautenschlager et ah, 1992) and ice cover (Lautenschlager and Herterich, 1990).
GCMs aim to simulate 3D structures and flows of the atmosphere, but even the most powerful computers are unable to model the scale and complexity of the atmospheric circulation system. The surface of the Earth is represented by a grid (the size of the grid cells varies commonly between 4x5° and 11.5 x 11.25°. In each grid cell, values are plotted representing selected Earth surface parameters. The parameters normally fall into two categories. Boundary conditions are surface values of, for example, sea-surface temperatures (SSTs), albedo, radiation, atmospheric transparency, sea-ice cover and topography (e.g. Kutzbach and Ruddiman, 1993). When modelling modern climate, direct measurements of physical parameters comprise the input data, while palaeoclimatic modelling is based on estimates obtained from proxy data. Dynamic conditions include flow and can be obtained by parameterization of surface processes (heat and moisture exchange, convection, Coriolis force, shear constants and atmospheric pressure) (Street-Perrott, 1991; Wright et ah, 1993).
Since GCMs involve enormous amounts of data and long computer runs by supercomputers, GCMs are only undertaken in a few highly specialist centres such as the National Center for Atmospheric Research (NCAR) in the USA; the Goddard Institute for Space Studies in USA; the UK Meteorological Office; and the Deutsche Klima Rechen Zenter (DKRZ) in Hamburg, Germany.
Several GCM efforts have concentrated on modelling climatic conditions at selected time intervals with appropriate proxy data, for example, the late glacial maximum around 18,000 yr bp. Figure 7.3 shows schematically how input boundary conditions, acting as the basis for GCM experiments, have varied since 18,000 yr bp.
GCM experiments are of two main types: analogue experiments and sensitivity experiments (Street-Perrott, 1991). Analogue experiments, or so-called 'realistic' experiments, attempt to simulate the prevailing Earth surface conditions as closely as possible for selected time periods. The early CLIMAP and COHMAP models were of this type (CLIMAP Project Members, 1981; COHMAP Members, 1988). Recent models have become more sophisticated, allowing for annual or seasonal variations of input data. Important linkages in the climate system have been demonstrated by the analogue modelling experiments, like the relationship between enhanced summer insolation and monsoon strength in the northern hemisphere, and the deflection of the jet stream by the build-up of the Laurentide ice sheet.
The purpose of the sensitivity experiments, however, is to test the relative importance of the different model components. Input variables are changed individually to elaborate the most responsive factors of the model. By comparing the modelling results with palaeoclimatic reconstructions, it may be possible to discover the most important forcing factor(s). However, some of the modelling outputs have clearly been wrong. Hansen et ah (1984) found, for example, that land ice was the most important factor in global cooling during the last glacial maximum, while Broccoli and Manabe (1987) in their experiments found that reduced C02 content was of greatest importance.
The most severe limitations with GCMs are the scale problems, such as the topographic generalizations to conform with the large grid cells. Another problem is the varying quality of the input data derived from proxy records. In addition, since GCMs are constructed for fixed time periods, they are considered to be static. Simulation and modelling of highly complex and rapid global and regional climatic changes lies beyond the capacity of existing GCMs.
Figure 7.3 Boundary conditions for the COHMAP simulation for the last 18,000 years. External forcing is shown for northern hemisphere solar radiation in June-August (SjjA) and December-February (SDjF) as the percentage difference from present-day radiation receipts. Internal boundary conditions include land ice, global mean sea-surface temperatures (SST) expressed as difference from the present, aerosols, and concentrations of CO2 in ppmv. (Modified from Kutzbach and Webb, 1993)
Was this article helpful?