PRESS OFFICE: (202) 628-3622

—LOBAL 0 4

BACKGROUNDER

1331 PENNSYLVANIA AVE. NW • SUITE 1500 - NORTH TOWER • WASHINGTON, DC 20004-1703

Climate Models: Shortcomings and Limitations
General Circulation Models
General circulation models (GCMs) are computer representations of global climate. They are
based on mathematical equations derived from our knowledge of the physics that govern the
Earth-atmosphere system. By definition, "climate" encompasses a vast number of factors
(cloud cover, air and ocean temperatures, rain and snowfall, air and ocean currents,
barometric pressure, atmospheric composition, etc.). In GCMs, climate scientists typically
divide the Earth's atmosphere into a three-dimensional grid, consisting of thousands of small
boxes. Within each box, they attempt to represent mathematically the interactions among
these and other factors (including geography and topography) that determine climate. The
result is essentially a computer program that can create a picture of the climate based on
variables entered by the modelers.
The picture is far from perfect, however. When testing the models, scientists attempt to
reproduce existing climate conditions by feeding the computer climate data they know to be
accurate. In an attempt to make the computer programs mirror current climate, scientists are
forced to calibrate their models (i.e., adjust their equations) to better reflect actual climate.
Calibrating the models does not mean eliminating their imperfections, but rather
compensating for them.
Even after the models are calibrated, however, only the most general predictions (e.g.,
equator is warm, poles are cool, seasons change) are reasonably accurate. On a regional level
(continents and subcontinents), the models cannot accurately predict climate. These
limitations must be acknowledged when interpreting the climate simulations.
Greenhouse Scenario: Doubling Atmospheric Carbon Dioxide
The starting point for all predictions of catastrophic global warming is an assumption that
levels of carbon dioxide (CO2) in the atmosphere will double in the next century as a result
of continuing CO2 emissions from human activity (e.g., burning fossil fuel). While some
scientists question the validity of this assumption, it is the basis for the predictions of
increasing average global temperatures made by advanced GCMs. Even so, the various
GCMs disagree significantly over the magnitude, ranging from 1.5 to 4.5 degrees Celsius
(1992 IPCC report).

 Limitations in the GCMs
GCMs are remarkably complex, often the products of teams of scientists and requiring the
most powerful computers to operate. But, nevertheless, the models are not nearly as complex
as the physics that determine climate, much of which is still not understood by climate
scientists. Ultimately, GCMs are only crude representations of the real world climate system.
In particular, the models do not adequately describe several key factors that will influence the
climate's response to an assumed doubling of CO2:
Role of the Oceans
First, the role of the oceans is not fully understood by scientists, and therefore cannot be
reliably accounted for in models. Oceans cover about 70 percent of Earth's surface and
currently hold about 300 times more CO2 than the atmosphere. In addition, the ocean is
responsible for transporting heat, which will greatly affect the magnitude and regional
distribution of the response to a doubling of CO2. Yet, modeling the dynamics of ocean and
atmospheric interaction is more difficult and even less accurate than modeling either the
ocean or the atmosphere. Improving confidence in the models will demand that ocean
models be better developed and more fully integrated with atmospheric models.
Cloud Changes
Another major factor not well accounted for in GCMs is the response associated with clouds.
As CO2 doubles, changes are likely in cloud types, cloud coverage, and water content and
reflectivity of clouds. These changes will certainly complicate the climate's response to a
doubling of CO2; increased cloud cover during the day has a significant cooling effect. Yet,
the physics is not understood well enough for models to incorporate the effects of clouds with
significant confidence. This is particularly important since natural water vapor is by far the
most influential greenhouse gas, accounting for 70-90 percent of the heat absorbing capacity
of the atmosphere.
Small-Scale Atmospheric Processes
Another source of inaccuracies in the models is the use of large blocks of land and
atmosphere (as defined by the grid into which the models divide the atmosphere) whose
characteristics are represented by only one set of data that does not accurately account for
variations within that block. The size of the base measurement typically is about 90,000
square miles (about the size of Colorado), which precludes many small-scale, but important,
climate related processes from being included in the models' calculations. This treatment
contributes to the models' inability to predict climate in specific continental or subcontinental regions.

 Improving the Models
General circulation models must be improved substantially before more confidence can be placed
in the scenarios they generate. How much improvement is necessary? During the past 15 years,
according to climate models, greenhouse gas emissions should have led to an increase in global
temperatures of 0.3 to 0.5 degrees Celsius and an increase in U.S. temperatures of 0.5 to 1 degree
Celsius. But NASA-analyzed satellite temperature readings for the same period show a slight
cooling trend of a few hundredths of a degree Celsius.
Dr. Stephen Schneider, a climate modeler writing in Global Warming. The Greenpeace
&pan (1990), indicates that much improvement is still necessary. Schneider emphasizes the
importance of developing and running models that accurately "couple," or integrate, the
atmospheric and oceanic models and that can accurately represent regional climate changes:
"Our findings suggest that it is imperative to run high-resolution, coupled atmospheric,
oceanic, cryospheric [ice cover], and land-surface sub-model models if regional, timeevolving scenarios of climatic changes are to have any hope of credibility. On the other
hand, the more complex, three-dimensional models are not yet at an adequate phase of
development to have trustworthy coupled atmosphere/ocean models that are both well
verified and economical enough to be run over the hundred years needed for
greenhouse-gas-transient simulations.... Since the agricultural and other environmental
impacts of increasing greenhouse gases depend on the specific regional and seasonal
distribution of climatic change, resolution of the transient-climate-response debate,
among others, is critical for climatic-impact assessment and ultimate adaptive responses
to the advent or prospect of increasing greenhouse gases."
Schneider and other scientists suggest that reliable regional climate projections are at least 10
years in the future, and probably more. They agree that an expanded research effort is necessary
to get the required computer power and to develop and routinely run the coupled climate models
over the needed 50- to 100-year simulated time frames.
This research effort is already under way. The National Oceanic and Atmospheric
Administration recently hosted a conference attended by climate modelers from around the world
where they discussed the failure of current model predictions to match observed climate changes,
as well as ways to improve the models.
Implications for Policymakers
How should policymakers treat the scenarios generated by climate models when developing a
policy response? Policymakers must be aware of the limitations of computer models and their
assumptions, recognizing that while they may provide scientists with some useful tools to
explore the dynamics of future global climate, the scenarios cannot be considered an accurate
picture of the future. The models are certainly not yet reliable enough to serve as a basis for
billion-dollar public policy decisions that would affect energy prices, consumer costs and U.S.
competitiveness in world markets.
[7182]