November 1966: Syukuro Manabe makes the first modern climate model

In 1946, a year after the creation of the first programmable modern computer, mathematician John von Neumann had an idea: scientists should use the machine to predict the weather.

Neumann formed a meteorology group at Princeton University and got to work. In 1950, after more than 24 hours of runtime, the ENIAC computer generated a 24-hour forecast based on a simplified model of the atmosphere.

The forecast arrived late and wasn’t entirely accurate, but the government took notice of this promising work. In 1955, the U.S. Weather Bureau created the General Circulation Research Section, headed by Joseph Smagorinsky, with the goal of modeling the atmosphere and, hopefully, better predicting the weather. (The section was later renamed the Geophysical Fluid Dynamics Laboratory, or GFDL.)

Smagorsky recruited Syukuro “Suki” Manabe, a promising meteorology Ph.D. grad from Tokyo University, to work on the project.

Smagorinsky “wanted to build a global numerical model,” Manabe recalled in a 2005 interview — an ambitious vision. But the climate is remarkably complex, and representing its multiple interconnected physical processes with the team’s “feeble” computer was a challenge. “If we put everything in the model at once, the computer couldn’t handle it,” Manabe said. “I was watching the model blow up all the time.”

For example, to describe how gases and heat move, the team needed to determine whether they could couple the equations of motion with equations describing energy transfer from radiation. They’d then need to run the model and generate a realistic description of how temperatures are distributed in the atmosphere, across its layers and around the globe.

This kind of complex prediction was a lot to ask of early computers, especially in the early stages, when the team just needed to see whether their basic approach was working.

Manabe decided to simplify things. He and his colleague Robert Strickler started by modeling how heat transfers and gases move through a vertical column of the atmosphere. Because this model looks only at motion along one axis, up and down, it’s described as one-dimensional.

The model worked: “I was pleased to get a nice stratosphere and near-perfect surface temperature,” Manabe said. The two published a paper on this work in 1964.

To test the one-dimensional model, Manabe and his colleague Richard Wetherald decided to look at the effects of different concentrations of carbon dioxide and other greenhouse gases. “Before I coupled this simple model with the three-dimensional one, I wanted to see how sensitive the model is to cloudiness, water vapor, ozone, and to CO₂,” Manabe later recalled.

“I realized that CO₂ is important,” he said. “As it turned out, I changed the right variable and hit the jackpot.” When the duo doubled CO₂ in their model, global average temperature increased by about 2.4 °C.

Manabe and Wetherald submitted their findings to the Journal of the Atmospheric Sciences on Nov. 2, 1966, and the study was published the following year. It’s widely regarded as the most influential climate change paper ever published, and laid the foundation for all modern climate models.

“I had no idea how important the idea of greenhouse gases would become,” Manabe said in a 2015 interview. “I had no idea it would have such a great impact on society.”

Global warming wasn’t a completely new idea: In the mid-1800s, scientists like John Tyndall and Svante Arrhenius realized that burning fossil fuels could cause global warming. But Manabe’s contemporaries disagreed about the magnitude of CO₂’s warming power. “At the time, no one cared about global warming,” he said. “Only gradually people began to realize that global warming is really an issue.”

Despite the simplifications in their model, Manabe and Wetherald’s estimation of the climate’s sensitivity to carbon dioxide concentrations still stands today.

Manabe’s model included two key factors that others had overlooked when estimating the greenhouse effect, says atmospheric physicist Joanna Haigh, an emeritus professor at Imperial College London. One factor was the so-called “convective adjustment,” which accounts for how gases in the atmosphere move in response to incoming solar radiation and the reflection of heat from the ground.

The other factor key to Manabe’s model is water vapor. When the atmosphere warms, it can hold more water vapor, which is a greenhouse gas. This causes a positive feedback loop: The warmer the atmosphere gets, the more water vapor it can hold, and the warmer it gets. As a result, scientists who made “dry calculations” underestimated the warming effect of CO₂, says Haigh.

Tapio Schneider, a climate modeler at Caltech, says this kind of work requires good intuition. After years of experience working on climate models, scientists build up knowledge that helps them make hypotheses and decide what to leave out and keep in. But when Manabe did his work, he didn’t have that kind of experience — climate models didn’t exist.

“Why was Einstein sure the speed of light was constant? With Suki it was similar,” Schneider says. The model was accurate in part because Manabe assumed that relative humidity would stay constant. “How did he know what to simplify? He just guessed, and he got all these key things right.”

Schneider says the Manabe paper is so important, and so simple, that he has his students recreate the model. Today’s computers are at least 1012 more powerful than the one Manabe used in 1966, but his model still works, and serves as a powerful teaching tool.

Manabe soon expanded his model to cover the globe, and he led GFDL’s efforts to couple their atmospheric model with ocean models. GFDL moved to Princeton University in 1968 and became part of the National Oceanic and Atmospheric Administration (NOAA) two years later.

Now a professor emeritus at Princeton, Manabe shared the 2021 Nobel Prize in Physics for his work on climate modeling.

According to the 2024 United Nations Environment Program Emissions Gap Report, the planet is on course for an average temperature increase of 2.6 to 3.1°C in this century. To limit that increase to the 1.5°C pledged in the Paris Agreement, the world needs to cut emissions by 42% by 2030 and 57% by 2035.

Now, says Schneider, climate modelers can help people understand how these changes will play out locally. How likely are extreme weather events? How can localities plan for this and protect themselves? In the 1960s, atmospheric modeling was about studying an interesting physical process. Today, says Haigh, “it’s almost existential.”

Modelers have access to a tremendous amount of data, powerful computers, and new computing tools, like machine learning. “It’s on us to rethink how to use the resources we have,” he says. “Suki made good assumptions. Now we have data to test all of these things.”