How Climate Models Work
News headlines about climate change research often are based on climate models. But to most people, how these models function is mysterious.
Researchers are continually improving the ability of climate models to simulate realistic climate. The models are imperfect descriptions of the Earth’s climate system, but their progress has advanced the understanding of future climate change and expanded the use of models in climate change planning. The most sophisticated climate models, called Global Climate Models (GCMs), often are used to generate data that are incorporated in to regional studies that assess, for example, future streamflow or vegetation change. They are also used to help people understand the relation between greenhouse gas (GHG) emissions and future climate, providing valuable information for policy makers who may set GHG reduction targets.
Despite widespread references made to climate models, there is a great deal of confusion about them. The following topics provide a window into the anatomy of climate models and how they work:
- The building blocks of global climate models
- Time and space in models
- Driving the global climate model with greenhouse gas emission scenarios
- Evaluating models
In a virtual system that evolves similarly to the real world, climate models attempt to integrate all of the known factors that influence climate, from the transfer of atmospheric heat into the oceans to the reflection of solar rays by polar and mountain ice. From the climate modeler’s standpoint, the processes that control the climate can be expressed by mathematical equations derived from scientific laws, empirical data, and observations. These equations are converted into computer language and, along with information about the Earth’s geography, such as topography and vegetation, form the basis of a climate model.
Figure 1. Schematic view of the components, processes, and interactions of the climate system.
| Enlarge This Figure |
Credit: Intergovernmental Panel on Climate Change, 2007
To understand how a climate model is constructed, it helps to think of the Earth’s climate as a complex system of many interacting parts that include the atmosphere, oceans, land surface, and sea and land ice (Figure 1). “Component models” for each of these parts have been developed and are continually refined at more than a dozen scientific centers worldwide. Atmosphere models are the oldest and evolved during the 1960s. They have at their core the equations for fluid motion, which describe air movement, and the first law of thermodynamics, which relates to the conservation of energy.
Ocean component models followed atmospheric models and were built to simulate ocean currents, salinities, and temperatures. By 1970, the first model integrated the atmosphere and ocean components into what is commonly referred to as atmosphere-ocean general circulation models (AOGCMs).
Some AOGCMS also include land surface components. Surface hydrologic processes such as evaporation, changes in snowpack, and infiltration of water into soil are typically found in these models.
Figure 2. The evolution of computer models has increased the number of vertical ocean and atmospheric layers, as well and reduced the size of each horizontal gridbox. Current models now include up to 50 vertical layers, an increase in resolution from the time this image was generated.
| Enlarge This Figure |
Credit: Adapted from Hadley Centre graphic.
Climate is a global phenomenon, and processes that occur far off can impact local conditions. As a result, climate models must encompass the entire Earth. They also must be able to simulate tens to hundreds of years of time in order to properly include the effects of heat transfer to the oceans. Time evolves in the models at discrete intervals called “timesteps,” which can range from a few minutes to an hour, depending on the spatial resolution of the model—the finer the spatial resolution, the shorter the timestep. Each model simulation generates enormous amounts of data output that can easily amount to hundreds of terabytes—the equivalent of the storage capacity of roughly a thousand typical desktop computers.
Because of the complexity of the mathematical equations in climate models, these equations can only be solved approximately. To determine the most precise result within this limitation, climate models typically divide the globe into a grid, creating “gridboxes” for the oceans, land, and atmosphere. The finer the grid, the higher the spatial resolution and the more computer power required to run the simulations. The finest resolution AOGCMs slice the Earth into horizontal gridboxes that are roughly 50 miles on a side (Figure 2). For both the oceans and the atmosphere, these gridboxes are stacked vertically, with as many as 50 or so boxes stacked on each other.
However, many climate processes take place at spatial scales that are smaller than a model gridbox, such as evaporation from the ocean, or the development of thunderstorms. As a result, these processes are “parameterized,” which means the average effect of the physics of the process and its sensitivity to change are captured but without going into the small scale details.
The scenarios of future greenhouse gas emissions drive the current generation of climate models. Greenhouse gas emissions, along with other natural factors, such as incoming solar radiation and volcanic aerosols, alter the amount of energy within the climate system and cause changes to the climate.
Emissions scenarios are estimates of how greenhouse gas emissions, such as carbon dioxide, methane, nitrous oxide, and their accumulation in the atmosphere might unfold over the next century. The International Panel on Climate Change (IPCC) has developed a suite of emissions scenarios that are widely used to generate climate projections from GCMs. The first scenarios were published in 1992, and a revised version was published in 2000 in the IPCC Special Report on Emissions Scenarios (SRES). The SRES scenarios are based in part on assumptions about demographic development, socio-economic development, and technological change. Researchers are again working on refining the emissions scenarios to reflect the most up-to-date information. This new set of scenarios will drive the next generation of climate model projections and likely will be included in the next IPCC assessment report.
The accuracy of global climate model simulations is evaluated by how well they reproduce climate statistics as opposed to individual events. This arises because model projections are not periodically reset to observed conditions but rather run freely through time. Consequently, simulations cannot reproduce the weather on any specific day, but they should reproduce averages and other weather statistics that reflect real conditions. For example, the climate models cannot reproduce a specific event such as the 1997–98 El Niño, but they are designed to create virtual El Niño and La Niña events that have similar magnitudes, durations, and recurrences as the real ones.
The coarse spatial resolution of models makes evaluating them difficult, particularly in mountainous regions like the western U.S. The large grid sizes (several thousand square miles) force mountains to be smoothed; each gridbox has an average elevation of all the topography in that grid. As a result, snowpack is poorly represented because the smoothed topography reduces the elevation of mountain peaks. However, the climate models do simulate the large-scale climate trends affecting mountainous regions. Current climate models produce a winter storm track and a summertime zone of high pressure that impact the Southwest, and they broadly show the differences in annual observed precipitation across the Great Plains, Rockies, and Intermountain West. Models, however, do not simulate monsoon precipitation well.
The approximately two dozen different climate models developed at the world’s climate modeling centers produce somewhat different climate states and projections of the future climate. For example, both the CM2.1 climate model developed at NOAA’s Geophysical Fluid Dynamics Laboratory and the CCSM3 model developed at the National Center for Atmospheric Research project a drier climate in the Southwest by the year 2100, but the CM2.1 model projects a greater degree of drying. Ultimately, the main reason for the differences among climate model results is an incomplete scientific understanding of many climate-related processes, particularly at smaller spatial scales. Even for processes that are comparatively well understood, such as the absorption and reflection of sunlight at the Earth’s surface, there can be legitimate scientific differences about the best way to represent these processes in the models through parameterization. This situation has led many researchers to analyze multi-model ensembles, which are collections of simulations from many different climate models, to better encompass the range of plausible future climates.
Produced by the Western Water Assessment for the Colorado Water Conservation Board. The full report is available online at:
| http://cwcb.state.co.us/Home/ClimateChange/ClimateChangeInColoradoReport |
This content was adapted from:
Ray, A.J., et al. 2008. Climate Change in Colorado: A Synthesis to Support Water Resources Management and Adaptation, 52 pgs.