The model scientist who fixed the greenhouse effect

Syukuro (“Suki”) Manabe in the 1960s at Princeton University, New Jersey, where he taught from 1968-1997. He was working on weather prediction in Tokyo during the difficult postwar years when he was invited to come to the US Weather Bureau’s unit working on the general circulation of the atmosphere. He was assigned programmers to write computer code so he could concentrate on the physical concepts and mathematics. Image copyright: AIP Emilio Segrè Visual Archives, used with permission.

In 1963, using one of the world’s first transistor-based supercomputers, Syukuro Manabe was supposed to be simulating how Earth’s atmosphere behaves in more detail than ever before. Instead, the young US Weather Bureau scientist felt the frustration, far more common today, of a crashed system. But resolving that problem would lead ‘Suki’ Manabe to produce the first computerised greenhouse effect simulations, and lay the foundations for some of today’s most widely used climate models.

After growing up during the Second World War, studying in bomb shelters, Suki entered the University of Tokyo in 1949 to become a doctor like his father and grandfather. The same year Japanese physicist Hideki Yukawa won a Nobel Prize, and helped drive many students into his subject, including Suki. “I gradually realized, ‘Oh my God, I despise biology,’” he told interviewer Paul Edwards in 1998. But to start with, he wasn’t very successful in his new subject. “At the beginning my physics grade was miserable – straight C,” he recalled.

Those grades came about because Suki’s main interest was in the mathematical parts of the subjects, but he hadn’t been thinking about what the maths really meant. When he realised this he concentrated on the physics he found most interesting, in subjects related to the atmosphere and oceans, and his grades started to improve. “By the time I graduated from geophysics and went on to a Master’s course at the University of Tokyo, I was getting a pretty solid way of thinking about the issues,” he said.

Suki went on to get a PhD, but when he finished the kinds of jobs in meteorology he was qualified for were hard to find in Japan. But he had applied his interests to rainfall, in an approach known as numerical weather prediction pioneered by scientists like John von Neumann, Carl-Gustaf Rossby and Bert Bolin. Another leader in the field, Joe Smagorinsky, was looking at rainfall in a similar way, and had read Suki’s research. Joe was setting up a numerical weather prediction team at the US Weather Bureau in Washington, DC, and in 1958 invited Suki to join him.

Their early models split the world into grids reaching into the air and across its surface, calculating what happens within and between each cube as today’s versions still do. But Joe wanted Suki to go further in preparation for the arrival of a transistorised IBM ‘Stretch’ computer in 1963. Joe wanted to develop complex system models that included the role of water movements, the structure of the atmosphere, and heat from the Sun. In particular Joe wanted to push from simulating two layers in the atmosphere to nine.

Breakdowns averted

Climate models are systems of mathematical formulae known as differential equations based on the basic laws of physics, fluid motion, and chemistry. To “run” a model, scientists divide the planet into a 3-dimensional grid, apply the basic equations, and evaluate the results. Atmospheric models calculate winds, heat transfer, radiation, relative humidity, and surface hydrology within each grid and evaluate interactions with neighboring points. Image credit: NOAA

So as well as adapting to the US, and fixing the programming mistakes that arose, Suki had to develop new ways to simulate climate physics on their existing computers. For example, he had to create ‘parameterisations’ from direct measurements to stand in for processes smaller than the cubes in his grid, making them as simple as possible. Under these combined stresses, Suki came close to a nervous breakdown, avoided only narrowly thanks to the arrival of less error-prone programmers.

When the Stretch machine arrived in what had become the Geophysical Fluid Dynamics Laboratory (GFDL) of the National Oceanic and Atmospheric Administration (NOAA) its response to the calcuations was similar to Suki’s. It became unstable and crashed. And Joe was suddenly away a lot, organising the Global Atmospheric Research Programme alongside Bert Bolin and others, which would collect data needed to build better models. That left Suki in charge of trouble-shooting, feeling bad that the expensive new computer was sat around idle. So he started simplifying the calculations enough to get it working. “I just single-handedly tossed off some complexity every time he was travelling,” Suki remembered.

One element of the physics their general circulation model (GCM) needed was the greenhouse effect, not to study global warming, but to simulate how heat spreads through the atmosphere. “Greenhouse gases are the second most important factor for climate, after the Sun,” Suki explained. “They warm Earth’s surface by as much as 30°C.” But modelling the greenhouse effect using approaches concentrating on heat radiation from some of the scientists most important in highlighting it, like Guy Callendar, caused big errors. Including the magnifying effect of water vapour into the radiative heat budget of the Earth’s surface, sometimes doubling CO2 would cause 10°C cooling, other times it would cause 15°C warming.

The problem, Suki realised, was that the models needed to look at more than just heat energy in the form of radiation. Scientists had recently realised that processes related to clouds, called cumulus convection, moved heat upwards from the Earth’s surface into the low atmosphere above, the troposphere. When water vapour in clouds turns into rain it releases latent heat, which can cause a temperature rise known as sensible heat, in the upper troposphere. Suki therefore included cumulus convection and large-scale atmospheric circulation, which also helps heat flow upwards in similar ways, in his models.

500K – what does that get you today?

A UNIVAC 1108 computer, circa 1970. In the foreground is the operator’s console. In the rear are its tape drives, and on the right is the CPU with its maintenance panel. GFDL’s version had 500K of memory. Credit: Unisys

Doing this was so complicated at the time he and teammate Richard Weatherald used a single column of air to model the whole planet’s atmosphere. But the approach, published in 1967, did provide a stable early answer to the question of what effect doubling the level of CO2 in the air would have. They found the amount of warming at Earth’s surface this caused, known as climate sensitivity, would be around 2°C. That was just a third of the first estimates made by Svante Arrhenius. The approach was basic, and without more data it couldn’t be confirmed, but it was a start.

Up to that point, Suki’s atmosphere models had been kept separate from ocean models built elsewhere in GFDL by Kirk Bryan. But because oceans can store a lot of heat energy global warming is slower than it would be without them. This fact made Kirk and Suki realise they had to come together. They set up a calculation that would first work out the condition of the atmosphere, then use the results of that stage to model ocean processes. Then, in a final stage it allowed interaction between the atmosphere and ocean. It needed to be simple, as the Univac 1108 computer they were now using had just half a megabyte of memory – smaller than most modern digital photographs. As a result the calculation had to be run for 50 days continuously. That marathon effort would be selected as one of the top 10 achievements in NOAA’s 200 year history in 2007.

Running the model for weeks on end took both optimism and vigilance. The scientists monitored it closely to ensure factors like heat and water vapour levels stayed in balance worldwide, Suki told me. “Although we did our best to construct a bug-free program, it is practically impossible to prove that it is absolutely bug-free,” he stressed. Their success marked a turning point in his career – afterwards he would focus more on using ocean-coupled models to study global warming than building them.

And though what can be done with climate models has progressed greatly since then, other elements remain remarkably similar. “Thanks to the remarkable advancement of computer technology, the computational resolution of models has improved greatly,” Suki said. “Meanwhile, parameterisation of sub-grid scale processes, like cloud-microphysics, land surface processes and sea ice dynamics, has become much more detailed. However, the basic model structure remains unchanged.”

• This is the first part of a two part profile. Now read part two.

Further reading:

Apart from the quotes in the last two paragraphs, which are from questions I asked Suki directly, all the quotes in this blog entry are from an interview of Suki Manabe by Paul Edwards on March 14-15 1998, published by the Archives for the History of Quantum Physics Collection, Niels Bohr Library & Archives, American Institute of Physics, College Park, MD USA. They are copyrighted by the AIP, and used here with its permission.

This year I’ve already written about the following pivotal climate scientists who came before Suki Manabe, or were around at the same time: Svante Arrhenius, Milutin Milanković, Guy Callendar part I, Guy Callendar part II, Hans Suess, Willi Dansgaard, Dave Keeling part I, Dave Keeling part II, Wally Broecker part I, Wally Broecker part II, Bert Bolin part I, Bert Bolin Part II

Syukuro Manabe and Richard T Wetherald (1967). Thermal equilibrium of the atmosphere with a given distribution of relative humidity Journal of the Atmospheric Sciences DOI: 10.1175/1520-0469(1967)024
Syukuro Manabe and Kirk Bryan (1969). Climate calculations with a combined ocean-atmosphere model Journal of the Atmospheric Sciences DOI: 10.1175/1520-0469(1969)026

Spencer Weart’s book, ‘The Discovery of Global Warming’ has been the starting point for this series of blog posts on scientists who played leading roles in climate science.