Climate change modelling

General Circulation Models (GCMs) are the basic tool used for modelling climate change. There are 30 modelling groups around the world using a variety of GCMs to investigate the potential impact of human activity on the climate. These GCMs model the atmospheric and oceanic interactions by solving the numerical equations that govern the physics of these interactions. The key equations are those relating to the conservation of mass, momentum and energy in the atmosphere and ocean, and are solved at a large number of points on a three-dimensional grid covering the whole world.

The equations used for the GCMs are essentially the same as those solved in the numerical weather models used by the Bureau of Meteorology to forecast the weather over a 4–10 day period; however, the GCMs use a much coarser grid, and the models are run for a much longer time period. Many important atmospheric phenomena (e.g. individual cumulus clouds) influence the way the large-scale flow will develop. These phenomena are often too small to be resolved by the computational grid over the time period that the models are run. To account for these phenomena, the models ‘parameterise’ their effect on the characteristics of the large-scale flow (BoM 2003).

Incoming solar radiation is absorbed by the Earth, and this energy is redistributed around the globe by atmospheric and oceanic circulations. The absorbed energy is radiated back to space at longer wavelengths (infrared band), and an energy balance between the incoming and outgoing radiation is established. Any perturbation to the net radiative energy available to the global Earth-atmosphere is termed a ‘radiative forcing’ by the Intergovernmental Panel on Climate Change (IPCC 2001); positive forcings warm the Earth’s surface and lower atmosphere and negative forcings have a cooling effect.
These forcings are incorporated into the GCMs by changing the concentrations of greenhouse gases and aerosols throughout the model simulation runs. The level of understanding of the impact of radiative forcing agents varies greatly, as does the uncertainty around the effect of each; for example, there is a very low level of scientific understanding of the indirect effect of aerosols in the troposphere, compared with a high level of understanding of the direct greenhouse effect of carbon dioxide, methane, water vapour and halocarbons.

A schematic representation of the horizontal and vertical grid structure for a general circulation model

A schematic representation of the horizontal and vertical grid structure for a relatively coarse-resolution general circulation model. The east-west cross-section in the right panel corresponds to the boxed area of the grid in the left panel, and indicates the terrain-following grid on which the numerical calculations are carried out (BoM 2003).

There are several sequential steps involved in developing and running a GCM. First, the validity of the climate model is established by comparing a simulation of the current climate (based on the present-day greenhouse levels) with the observed climate. The greenhouse gas levels are doubled for the next run, and the change in global mean temperature is used as a measure of the sensitivity of the model. A transient experiment is then conducted, in which the greenhouse gas levels are increased gradually in accordance with the gas emission scenario that is adopted (equivalent to a particular forcing).
Knowledge of how the atmosphere responds to various radiative forcing agents informs the development of the range of emission scenarios used in experimental runs of the GCMs. Most models, however, cannot be run for the full range of possible scenarios , owing to both the complexity of computation and the consequent processing time required to run transient coupled GCMs. To overcome this difficulty, many GCMs are run with a 1% per year compound increase in carbon dioxide, which is close to the current growth rate of equivalent carbon dioxide in the atmosphere (BoM 2003). Simpler climate models are then used to investigate the range of scenarios, using the results from this ‘transient climate response’ model.

One of the areas of greatest uncertainty associated with the current generation of GCMs is their treatment of feedback mechanisms. Feedback processes can act to amplify the response of the climate to the radiative forcing (positive feedback), or they can counteract it (negative feedback). In the BoM’s GCM, the dominant positive feedback is due to water vapour (BoM 2003). Other major feedbacks relate to the changing albedo resulting from changing ice, snow cover and vegetative cover.
In many parts of Australia, the El Niño-Southern Oscillation phenomenon has a dominant influence on many industries, particularly agriculture and fisheries. It is therefore important to understand how global warming will affect this oscillation. Many GCMs simulate a faster rate of warming in the central and eastern tropical pacific compared with the western tropical pacific – influences which favour the forming of El Niño conditions. However, there is still a considerable amount of research required to fully understand all the interactions before the projections of frequency, amplitude and pattern of El Niño events, as a result of climate change, can be used with confidence.

Regional scale climate change modelling

The contribution of various agents to global, annual-mean radiative forcing (Wm-2) since the mid-1700s. The vertical lines about the bars indicate the range of uncertainty and the words across the bottom axis indicate the level of scientific understanding supporting each of the estimates (BoM 2003)

More investment has been made, both nationally and internationally, in modelling global climate change scenarios than in understanding the impacts of these scenarios at the regional level. This relatively small investment has now resulted in a scientific community with limited capacity to meet the demand for information and understanding of the potential impacts on a regional scale.
The global-scale assessments of climate variables that are simulated by the GCMs are generally not appropriate for assessing the impact of climate change at the regional and local levels. The interactions of the climate with local topographic features, such as coastlines and mountains, result in large-scale changes in atmospheric circulations, and significant variation in the impact on local and regional scales (BoM 2003).

The techniques that have been developed to derive regional-scale climate projections range from applying the GCMs at a finer horizontal resolution (which is very computationally intensive) through to statistical ‘downscaling’. To date, the large differences between the various models in regional-model climate projections suggest a low level of confidence in their reliability for producing realistic climate projections (BoM 2003). The results of these models, however, are useful for undertaking an analysis of the sensitivity of a particular region to climate change.

Statistical downscaling techniques use the coarse grid of the GCM output to establish statistical relationships between local climate variables and global-scale atmospheric variables. Variables from the GCM output, such as mean sea-level pressure, are considered reliable. The established statistical relationships are used to infer local variables from the reliable GCM output variables, at a high temporal resolution (e.g. daily).

The statistical downscaling techniques require a long time-series of surface climate observations over a relatively dense network. The meteorological networks that meet these requirements in Australia are generally restricted to the south-east and south-west. The most common variables used in these types of studies are daily extremes of rainfall and temperature. Other derived variables that are particularly important to agriculture include growing degree day (cumulative amount of time in a season that the temperature is between particular thresholds important for plant growth) and stream flow.