views updated



A simulation is a collection of numbers and equations that are represented in a computer to mimic the behavior of a physical system. Simulations are based on the laws of physics that are relevant to the physical objects being considered, whether natural or artificial. For example, a simulation of the solar system might consist of a collection of numbers recording the masses and states of motion of the sun, planets, moons, and perhaps a number of comets, asteroids, and dwarf planets. These numbers would be changed by the computer according to Newton's laws of motion. If a simulation is accurate, it allows the user to predict the future behavior of the physical system being simulated—in this case, the orbits of the planets and other objects.

In climate studies, simulations are written to describe the behavior of the oceans, atmosphere, and other aspects of the climate system. Some simulations describe only one aspect or fraction of the climate system, while others seek to model the entire system at once.

Since Earth's climate system is complex, climate simulations—or, as they are usually called by scientists, climate models—must also be complex. Today's atmosphere-ocean general circulation models (AOGCMs) push supercomputer capacities to the limit, and will continue to do so as climate modelers attempt to simulate Earth's climate system with ever greater accuracy. Predictions from AOGCMs, despite their limitations and uncertainties, are our only source of knowledge about future climate.

Historical Background and Scientific Foundations

Simulation depends on computers. Powerful digital computers were first built during World War II (1939– 1945) for military purposes such as code-breaking and logarithm tables. Computers were sold commercially for the first time in 1950s and soon became indispensable tools in business and science. Computer power has increased exponentially ever since.

In a typical computer simulation or model, Earth, or some part of it, is represented by a grid of elements somewhat like the squares on a chessboard. Each element has its own budget of solar energy, water, vegetation, air, and other quantities, represented by a collection of variables (numbers that may change). The behavior of these variables is described by a set of equations that delineate their physics, that is, the behavior of their real-world counterparts. For example, equations might describe the rate at which heat is exchanged between the air and the sea in a particular element, or the rate at which growing plants absorb carbon dioxide from the atmosphere. Other equations describe the relationships between adjacent squares of the model.

At any given moment, the state of the model—the picture of the world it gives—is stored as a collection of numbers in computer memory. To give the model a time dimension, models discretize time, that is, break it up into steps. These time-steps are similar to the still frames that are shown one after the other to create video images. If the time-steps are close enough together, the flow of time can be represented with acceptable accuracy.

To give a simplified example, a time-step might take the model forward one day in time: the amount of heat energy transferred between the air to the ocean in an element would be specified by an equation that says how fast energy flows between the two. The numbers recording the heat energy in the air and water would change, one increasing and the other increasing (unless there happened to be zero total heat exchange that day). All the other variables associated with that element of the model, and with all its other millions of elements, would be updated in the same way, so that their new values would give a picture of the climate one day later.

The growth of computer power has been reflected in the field of climate simulation. In the mid 1970s, early climate models took into account only carbon dioxide emissions, a simple model of the atmosphere, sun, and rain. By a decade later, in the mid 1980s, cloud cover, fixed ice, and land surfaces had been added. By the time of the Intergovernmental Panel on Climate Change's (IPCC's) First Assessment Report (1990), simple oceans without circulation were simulated by models typically chopping up the globe's surface into about 2,000 squares; by the Second Assessment Report (1995), volcanic activities, sulfate emissions, and oceans with depth had been added, and over 8,000 squares were being simulated. By the Third Assessment Report (2001), aerosols, rivers, the carbon cycle, and the overturning thermohaline circulation of the oceans were being simulated over about 16,000 squares; and by the Fourth Assessment Report (2007), atmospheric chemistry and interactive vegetation had been added to simulation grids of over 42,000 squares.

This is a simplified account. In practice, today's AOGCMs cover the globe with grids of stacked or layered cubes, rather than with a quilt of flat squares, so the numbers given are much smaller than the total number of model elements. Typically, one grid of cubes models the ocean and another models the atmosphere, while two-dimensional (flat) grids model sea-ice and land-surface processes. Interconnected, these various grids model the interactions of surface processes with atmospheric and oceanic flows, clouds, and other factors to simulate the real world as closely as possible. Similar techniques are used in aircraft and rocket design, weather forecasting, solar physics, astronomy, nuclear weapons design, and many other fields.

With each increase in resolution and complexity, models have become better capable of predicting climate change and of hindcasting past climate change—that is, describing past changes in climate without the benefit of consulting actual climate records. These predictions are compared to what actually happened to see how well the model would have foretold the future. In this way, models can be tested against measurements, even though simulations of the future must wait to be tested against reality.

Although more complex models are a fundamental advance, their growth has not led to the disappearance of simpler models using fewer equations, fewer dimensions, or coarser grids. Simpler models, precisely because they are simpler, can sometimes provide insights that almost-realistic models cannot. Climate scientists work with a hierarchy of models: AOGCMs are used in conjunction with more-detailed models of regional climate and with process-oriented models that simulate cloud formation, smaller oceanic eddies, or particular weather events such as hurricanes. There is no one best climate model.

Impacts and Issues

Climate simulations are an indispensable tool of climate science and of decision-making by citizens and government officials. In its 2007 Assessment Report on climate change, the IPCC consulted at least 14 separate AOGCMs, often averaging their predictions of future climate or simulations of paleoclimate to produce estimates with less uncertainty.


DISCRETIZE: To represent a phenonemon that occurs smoothly over some range, such as wind speed or temperature, as a set of numbers separated by fixed gaps. “Discrete” means separate, not connected. For instance, if temperature is measured only to the closest tenth of a degree, then temperature is discretized to jumps of 0.1 degree. All numerical representations of continuous natural phenomenon discretize them to some extent. In climate modeling, treating Earth's environment as a mass of blocks or squares, each of which experiences uniform conditions, is a necesssary form of discretization. The finer the discretization (i.e., the smaller the jumps or units), the more closely the real, continuous world is represented.

INTERGOVERNMENTAL PANEL ON CLIMATE CHANGE: Panel of scientists established by the World Meteorological Organization (WMO) and the United Nations Environment Programme (UNEP) in 1988 to assess the science, technology, and socioeconomic information needed to understand the risk of human-induced climate change.

PALEOCLIMATE: The climate of a given period of time in the geologic past.

THERMOHALINE CIRCULATION: Large-scale circulation of the world ocean that exchanges warm, low-density surface waters with cooler, higher-density deep waters. Driven by differences in temperature and saltiness (halinity) as well as, to a lesser degree, winds and tides. Also termed meridional overturning circulation.

A common objection to climate modeling is that computer models often fail. Such critics often mention the fallibility of weather prediction, which also depends on computer simulations: “They can't even say what the weather will be next week,” the argument goes, “so how can they say what it will be like in 50 years?” This challenge depends on a confusion of weather and climate. Weather is the behavior of the atmosphere on short time scales like hours or months; climate is the behavior of the atmosphere on the scale of years or even decades. The details of particular weather events are much more chaotic and difficult to predict than climate. Predictions of global warming or other climate changes are a fundamentally different kind of prediction than weather forecasting.

Also, climate models are constantly tested for reliability by checking their power to predict past climate. In 2007, the IPCC used 14 models to simulate the climate history of the twentieth century. Researchers drove the models with natural and human inputs such as volcanic eruptions and human releases of CO2 but did not constrain them to match actual climate records: the goal was to see whether the simulations would describe climate changes that matched the actual records. As a group, the simulations did a remarkably good job of modeling global climate, including sharp dips in global temperature after major volcanic eruptions such as Agung (1963) and Pinatubo (1991) and the overall rapid rise of global temperature in the late twentieth century and beyond. According to the IPCC, “There is considerable confidence that climate models provide credible quantitative estimates of future climate change.”

See Also Chaos Theory and Meteorological Predictions; General Circulation Model (GCM).



McGuffie, Kendall, and Ann Henderson-Sellers. A Global Warming Primer. New York: Wiley, 2005.

Solomon, S., et al, eds. Climate Change 2007: The Physical Science Basis: Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change. New York: Cambridge University Press, 2007.


Delworth, Thomas, et al. “Simulation of Early 20th Century Global Warming.” Science 287 (2000): 2246–2250.

Gillett, Nathan P. “Simulation of Recent Southern Hemisphere Climate Change.” Science 302 (2003): 273–275.

Hansen, James E., et al. “Global Climate Data and Models: A Reconciliation.” Science 281 (1997): 930–932.

Stocker, Thomas F. “Models Change Their Tune.” Nature 430 (2004): 737–738.

Trenberth, Kevin E. “The Use and Abuse of Climate Models.” Nature 386 (1997): 131–133.

Web Sites

“Modeling Climate.” National Oceanic and Atmospheric Administration (NOAA), June 12, 2007. <> (accessed November 18, 2007).

Larry Gilman