Your current browser version is out of date, in order not to affect your use, it is recommended that you use the latest Google Browser, Firefox browser, 360 browser, after changing the browser use more smoothly! (note! Dual-core browser please switch to speed mode)
12 July 2023
Climate scientists push for access to world's biggest supercomputers to build better Earth models

How quickly will Earth warm, and what will that mean for the planet? To answer those questions, researchers have spent decades building increasingly sophisticated global-climate models — but those models are straining the limits of available computing power.

Now a cadre of scientists is pushing for an ambitious solution: a network of next-generation modelling centres, dubbed Earth Visualization Engines (EVE), where hundreds of staff scientists would use the latest supercomputers to run ultra-high-resolution climate models at full scale. Such models are starting to come online, and many researchers think that they could markedly improve climate projections.

“Maybe I’m a romantic, but climate change is a global problem in dire need of internationalism,” says Bjorn Stevens, director of the Max Planck Institute for Meterology in Hamburg, Germany, and a leading proponent of EVE. “What’s lacking is a coordinated strategy to link the biggest supercomputing machines to one of the biggest human problems.”

Supporters formally launched the proposal at a climate summit in Berlin last week, and hope to introduce it to government leaders at the United Nations climate change conference in Dubai later this year. But the idea has yet to win over everybody. Even in the climate-science community, some researchers question the proposal’s cost, practicality and scientific pay-off.

“There are a lot of good ideas in there, and it can be galvanizing to talk about these things,” says Gavin Schmidt, who leads NASA’s climate-modelling team at the Goddard Institute for Space Studies in New York. “But ultimately, I don’t think it’s the way forward.”

Simulating Earth

Climate simulations model the conditions inside the ‘cells’ of a 3D grid that wraps around a virtual Earth, often coupled with separate models representing components such as oceans and ice sheets. Many models run on a grid with cells that average 50–100 kilometres per side, but numerous teams around the world are developing higher-resolution models with grid cells of one kilometre or even less.

Studies have already shown that higher-resolution models can produce more realistic simulations of clouds, ocean eddies and more. Many researchers hope that these models can help resolve fundamental questions about the magnitude of global warming as well as details about the role of clouds, dust and air pollution.

Some climate scientists now have access to supercomputers that can perform trillions or even quadrillions of operations per second. But running high-resolution models on current machines remains both costly and time-consuming.

Exascale power

EVE’s backers seek to change that equation by creating between three and five modelling centres, each equipped with the latest generation of the world’s fastest ‘exascale’ supercomputers, which can perform quintillions (one million trillions, or 1018) of operations per second. These centres would be staffed by 200–300 scientists. With an estimated price tag of €300 million (US$330 million) annually for each centre, the total cost would be between €10 billion and €15 billion over 10 years.

These centres would produce freely accessible data that would be easy for other researchers to access, visualize and use. At the same time, supporters say, EVE centres would help to democratize climate science by involving researchers from the global south. Climate modelling doesn’t have to be a field that is dominated by “boys with toys” in wealthy countries, Stevens says. “We can do this together.”

The first large-scale test of these ideas is already under way in Europe, where the European Commission has invested an initial €150 million in an initiative called Destination Earth. Project scientists are developing high-resolution models and tools that will allow scientists to easily use and visualize the data.

Strife about EVE

Some question whether such efforts will bear fruit, arguing that increasing the resolution of climate models is no panacea for climate science. Scientists have learnt that the differences between models can reveal insights about how the real world functions, says Schmidt. He and others worry that EVEs would ultimately centralize modelling — and, at least in the early days, pull expertise from existing modelling centres.

Climate scientists are pursuing a range of ideas to improve models, says Ruby Leung, a climate scientist at the Pacific Northwest National Laboratory in Richland, Washington. “Zooming in on a single approach and a small number of centres at this point in time could limit other efforts and their potentials,” she says.

Such concerns are overblown, says Peter Bauer, who headed Destination Earth at the European Centre for Medium-Range Weather Forecasts in Bonn, Germany, before retiring this month. EVE is not intended to centralize existing science, he says, but rather to provide a platform for new science. Nor are its supporters worried about the price tag. They point to the roughly 1.4-billion-Swiss-franc ($1.5 billion) annual budget of CERN, Europe’s particle-physics laboratory near Geneva, and note that the cost of catastrophic global warming will be much larger.

“It’s really not a lot to ask when you think about the climate problem more broadly,” Stevens says.