NCAR’s new climate model is running hot

Study examines possible reasons

Aug 5, 2019 - by Laura Snider

A new study takes a detailed look at possible reasons why the latest version of the flagship climate model based at the National Center for Atmospheric Research (NCAR) appears to be “running hot,” projecting an even more pronounced warming response to increased carbon dioxide in the atmosphere than previous versions.

For decades, climate models – including those managed by NCAR – have shown that a doubling in atmospheric carbon dioxide from preindustrial levels would result in somewhere between 1.5 and 4.5 degrees Celsius (2.7 – 8.1 degrees Fahrenheit) of warming.

NCAR’s newest iteration of the Community Earth System Model (CESM2), which was released to the research community last year, is now projecting 5.3°C (10.1°F) of warming if carbon dioxide is doubled. Similar increases in climate sensitivity are also being reported by research teams operating other leading climate models, which are running experiments in preparation for the next report of the Intergovernmental Panel on Climate Change (IPCC).

“Multiple independent models are coming up with answers that are outside the canonical range from the last 30 years,” said NCAR scientist Andrew Gettelman, who led the new study. “We are concerned there could be something wrong in the model, but we’re even more worried that the model might be right. This magnitude of increase in climate sensitivity increases the risk for extreme climate change impacts.”

The study, published in Geophysical Research Letters, is a methodical analysis of the new model aimed at better understanding why the climate sensitivity jumped out of the historical bounds.

“In our paper we picked our model apart,” Gettelman said. “We went under the hood, and isolated different processes so we could zero in on what changed.”

The scientists were ultimately able to show that the biggest difference comes from the way clouds, as well as cloud interactions with the tiny atmospheric particles called aerosols, are represented in the new model. While it’s possible that these representations are inaccurate and are contributing too much to surface warming, the scientists can’t yet determine that with certainty. In fact, the clouds simulated in CESM2 look more realistic than those in its predecessor when compared to observations.

To get a fuller picture of whether the model’s new, higher climate sensitivity is an actual possibility or the result of an error, CESM2 scientists will likely require the help of the broader scientific community. They will also need more detailed observations of clouds, which could improve our understanding of how they may change along with the climate.

“We’ve put the data out there from this study, and it’s all publicly available,” Gettelman said. “We hope other people take a look and try to figure out if anything is broken.”

Sorting through change drivers and feedback loops

How much the climate warms as human emissions of greenhouse gases continue to increase depends not only on the heat-trapping quality of the gases themselves, but also on other factors that force changes in the climate system – including emissions of sunlight-reflecting aerosols from both volcanic eruptions and industrial activity. Aerosols also affect the climate system by interacting with clouds. Scientist think these interactions cause clouds to be more reflective, which works to cool the planet, but the size of that impact is still uncertain.

A number of feedback loops in the climate system also play a role in determining how warm the climate will become in the future. Feedbacks are responses to changes in temperature that can amplify or reduce the warming caused by the greenhouse gases emitted by burning fossil fuels. For example, a warmer climate can reduce snow cover, which in turn reduces the reflectivity of Earth’s surface and causes even more surface warming. A warmer climate also allows the atmosphere to hold more water vapor, which acts as a greenhouse gas and causes more warming.

Clouds are also an important source of feedback in the climate system, but their impacts are complicated. An increase in some kinds of clouds could reflect more incoming sunlight back into space, causing a cooling effect, while an increase in other kinds could trap more heat at the surface. Where clouds form is also important – clouds in the tropics can reflect more direct sunlight due to the relative angle of the Sun than clouds at higher latitudes, for example.

Complicating this picture, scientists are still working to figure out precisely how clouds and aerosols interact. In addition, cloud processes unfold on tiny scales, which are hard to capture in climate models that simulate the entire globe. Scientists are also uncertain how those intricate processes will change in response to global warming.

Despite these challenges, researchers at NCAR and their collaborators at universities and federal agencies worked to improve the way clouds and their interactions with aerosols were simulated in CESM2 compared to its predecessor, CESM. Gettelman said they anticipated that some of these upgrades, based on new observations, would increase climate sensitivity. For example, a recent field campaign aimed at observing clouds over the Southern Ocean, known as SOCRATES, revealed that the clouds there contained more supercooled water and less ice than previously thought – a difference that could be expected to increase climate sensitivity somewhat.

But these kinds of expected outcomes do not explain the entire magnitude of CESM2’s increase in climate sensitivity, which is 1.3°C higher than CESM.

A new emissions dataset

The issues with CESM2 and climate sensitivity became apparent during testing with a new international dataset of emissions. The dataset includes updated estimates of aerosol emissions released by burning fossil fuels. These types of industrial aerosols, which interact with clouds to block more incoming sunlight, are thought to be a significant contributor to the cooling that occurred in the mid-20th century and to have damped overall warming throughout the 20th century and up through today.

The scientists who developed CESM2 tested the model repeatedly during model development to make sure that it matched the observed climate of the last century and a half. However, when they switched from an older emissions dataset to the new one, the resulting simulations did not warm the climate in the 20th century as much as observed.

The scientists determined that the main problem stemmed from the more detailed way CESM2 handled interactions between aerosols and clouds. They set about investigating approaches to adjust the model’s response to aerosol emissions so that model simulations would again accurately simulate 20th century climate. Some of the adjustments necessarily also affected clouds in the model, though the impacts of that are not fully understood, Gettelman said.

The scientists successfully fixed CESM2’s simulations of past climate before its official release, but the climate sensitivity also rose to 5.3 degrees C. Because other climate models that began using the emissions dataset are also seeing similarly high climate sensitivities, scientists will conduct a more thorough investigation of the dataset to see if there's a problem with it. The dataset will be used for an international climate research project known as the sixth Coupled Model Intercomparison Project (CMIP6), which will inform the next IPCC assessment.

Gettelman said other modeling groups are now in the process of writing their own diagnostic papers on climate sensitivity. Having more detailed information across models may help the larger modeling community understand whether the increased sensitivity is a real increase in the threat of climate change or a problem in the models. It will also be helpful to learn how the models that don’t include advanced aerosol interactions are reacting to the dataset, he said.

“Many of the more advanced models have added aerosol cloud interactions in their most recent versions,” Gettelman said. “They all deal with the processes in different ways, but the one thing they have in common is the CMIP6 dataset.”

See all News