Making Sense of Uncertainty: What do Climate Models Tell Us?

“Models are always wrong, but they can be useful,” said Claudia Tebaldi, a statistician at the National Center for Atmospheric Research, during her presentation on the uncertainty of climate models at the 2015 Metcalf Institute’s Public Lecture Series.


An animation of CM2 Global Coupled Climate Models (CM2.X) average decadal July maximum daily surface air temperatures for the 21st century. Source NOAA

The physics of heat and temperature ultimately determine how our environment changes over time, but that doesn’t stop climate scientists from assessing the future through prediction, said Tebaldi, explaining that by combining quantitative mathematical equations with the parameters of Earth’s physical systems, climate models aim to project the future path of our global climate. Using temperature, precipitation, wind, and vegetation-cover information, a model simulates a climate scenario that can help produce predictions spanning as far as an entire century, and even further into the future.

Tebaldi illuminated some of the finer points on the role of models in climate prediction, including what uncertainties they often encounter. Ultimately the goal of all climate models is to understand the implications of climate mechanics both now and in the future, she said. Essentially, they become virtual laboratories where settings can be shifted across space and time in order to look at specific research questions.

As computing abilities have continued to improve, scientists can simulate an increasing number of scenarios. This has also led to increasing changeability in climate predictions, something that Tebaldi noted could cause difficulty amongst researchers.

“Models are always approximating,” she said. “They are encouraged to provide robust results based on different assumptions of future conditions.”

A great deal of debate exists surrounding which types of models are the most accurate. Evaluating the performance of a model based on its predictions, however, leaves room for misinterpretation.

Tebaldi’s message stressed that reliance on model predictions is not the answer. Rather, models should be used as a reminder and a tool to be used when asking, ‘What changes do we see in our world today?


Evan Ridley | Sea Grant Science Communications Intern and Marine Affairs Graduate student at the University of Rhode Island