Present climate is determined from analysis of meteorological
observations. Climates of the past may be deduced by studying proxy
data such as ice-cores, tree-rings, pollen, etc. Future climate changes
may be estimated by means of computer simulations using programs known
as General Circulation Models or Climate Models.
People have built physical models having some properties in common with the atmosphere. These 'rotating dish-pans' do contain wave motions and instabilities similar in some respects to those observed. But no-one has constructed a mechanical model which behaves in detail like the atmosphere. The difficulties with such a task are obvious: for example, we cannot replicate the radial gravitational force of the spherical earth in the laboratory. So, we design instead a mathematical model in which the geometry of the atmosphere can be precisely represented and a variety of detailed physical processes incorporated. The computer program to solve or integrate this system of mathematical equations is our climate model.
The essence of modelling is to use the known laws of physics to design a computer program capable of simulating the atmospheric flow. With realistic initial observational data the program can produce a short-range weather forecast, typically for up to a week ahead. If the solution is carried forward for an extended period - months or years of simulated flow - the forecast will diverge rapidly from the true evolution of the atmosphere. However, its statistical characteristics - mean values, frequency of extremes, etc. - may resemble those of the real climate provided the model includes a representation of all the processes which determine climate.
The model has as its basis the fundamental principles of physics - conservation of mass and energy and Newton's laws of motion. These determine the overall behaviour of the atmosphere. Many physical processes must also be allowed for: the phase changes of water, incoming solar radiation, frictional drag at the earth's surface, sub-grid-scale turbulence and so on. The details of some of these micro-physical processes are poorly understood with the consequence that there are inherent inaccuracies and uncertainties in all climate models.
In studying climate change, we must also include chemistry in the model. Small changes in the concentration of some species may dramatically alter the radiation balance and profoundly influence the climate. Details of the interaction of solar and terrestrial radiation with compounds such as methane and ozone must be taken into account. A vast complex of chemical reactions occurs in the atmosphere. Current models have relatively simple treatment of chemical processes. Future models may need to consider the concentrations of hundreds of compounds, their reaction rates, and so on.
The most advanced climate models are capable of simulating the observed climate with a good level of accuracy. The zonal mean winds and temperatures in each season are generally well simulated, and the statistics of extremes are realistic. Thus, a reasonable picture of the climate of a given geographical location emerges from the models. However, it must be acknowledged that there are still serious shortcomings. In a recent comparative study of 14 climate models, it was found that the temperature was on average too cold, particularly in the polar upper troposphere and tropical lower troposphere. It was conjectured that all the models are misrepresenting or even omitting some mechanism, resulting in this deficiency.
So, we can simulate the current climate fairly faithfully. But why bother? Why spend large resources finding out what we already know? Well, our fond hope is that a climate model can do more than simulate the status quo. Like a laboratory model, the climate model can be used in experiments to study how the climate may change in the future. For example, we may study the sensitivity of model climate to a doubling of carbon dioxide simply by changing a single number (representing CO2 concentration) and re-running the model. This may give us an indication of what is likely to happen in the real world under such circumstances. Indeed, just such experiments are the basis of current predictions of global warming resulting from the burning of fossil fuels. However, model results have been found to vary widely when details of the physical parameterisations are adjusted: in recent experiments with the Hadley Centre model, the predicted increase in global mean surface temperature ranged from 1.9C to 5.2C using different cloud schemes. The following quotation is from a report on these experiments:
The guidance provided by model sensitivity studies is about the best we have for now. But there is some danger in placing too much reliance on it. The equations governing the atmosphere are non-linear, their solutions are hyper-sensitive to small perturbations and minor changes can have major implications. Thus, seemingly negligible deficiencies in a climate model may render its output useless or misleading. Moreover, there are certainly physical and chemical processes which are insignificant in present conditions and are therefore ignored in current models, but which may be of critical importance in altered circumstances. As an example, the detailed micro-physics of ice clouds is involved in the depletion of Antarctic ozone. This was not foreseen by the modellers, with the result that we were all caught on the hop. I fear there may be more unpleasant surprises to come.
In conclusion, the overall success of climate models in simulating the present climate of the atmosphere is impressive. Although there are shortcomings in all models, they give a generally accurate picture of reality. They provide a valuable means for estimating the likely climatic consequences of changes induced by mankind's activities. At present, we must interpret their guidance with some caution; in particular, the details of geographical variations in climate impact may be unreliable. We can be confident that the dependability of model guidance will grow with their increasing sophistication, so that detailed regional climate impact predictions may be reliable in the future, perhaps shortly after the turn of the millennium. But in an uncertain world we can never completely rule out those 'nasty surprises'.
(1) Basic Text: An Introduction to Three-dimensional Climate Modelling. Warren M Washington and Claire L Parkinson, University Science Books, 1986.
(2) Recent CO2-Sensitivity Study: The Hadley Centre Transient Climate Change Experiment. Hadley Ceentre Publication, August, 1992.
(3) Recent Model Intercomparison Study: An Intercomparison of Climates Simulated by 14 Atmospheric General Circulation Models. G J Boer et al., July, 1991, CAS/JSC WGNE, WMO T.D.-No.425.