Climate Science · Ocean Modeling

Faster Ocean Models and a Major Carbon Correction Emerge From Five Years of Climate Research

Nine research teams backed by Schmidt Sciences have sped up climate simulations, sharpened ocean models, and revised the math on how much carbon fire returns to the atmosphere.

By SEVENSEAS Media  |  Climate & Ocean Science

Climate models are the quiet machinery behind almost every serious decision about our warming world. They tell coastal planners where the water will rise, tell disaster officials when to prepare, and tell policymakers what a given emissions choice will actually cost. Yet for decades those models have carried the same stubborn limitations. They demand so much computing power that only a handful of institutions can run them, they blur or miss the small-scale processes where clouds form and carbon burns, and the science they produce rarely reaches the people who need it most.

A five-year research effort backed by Schmidt Sciences set out to close those gaps, and its results are now reshaping how climate systems are modeled and who gets to use them. The Virtual Earth Systems Research Institute, or VESRI, supported nine research teams working on some of the hardest problems in climate modeling. Simulations that once needed thousands of computing processors now run on just a handful of GPUs in roughly half the time, and one ocean emulator that captures crucial ocean dynamics now runs dramatically faster than the model it replaced.

A Major Correction on Fire and Carbon

One of the most consequential findings came from CALIPSO, a founding VESRI project led by Philippe Ciais, which set out to model what he calls the most unifying and important blind spot in climate science, namely how carbon is lost from plants, soils, and oceans. By reconstructing a global record of burned area stretching back more than a century, the team found that fire has been contributing roughly 60 percent more carbon to the atmosphere than previous estimates reflected.

That correction is not a minor footnote. It fed into a new version of the Global Fire Emissions Database, known as GFED5, and it ripples through the carbon budgets and emissions pathways that governments and planners rely on. A separate CALIPSO investigation in northern Siberia found that roughly a third of the fires there burn peatlands, which have dried out in hotter summers and in some cases smolder through the winter. That is a carbon source most models simply do not account for.

Faster and More Accurate Ocean Models

For an ocean audience, the modeling advances are just as striking. Current climate models divide the Earth into grid boxes typically about 25 kilometers across, which means anything smaller, including cloud formation, turbulence, radiation, and ocean mixing, goes unresolved. The M²LInES project, led by climate physicist Laure Zanna, used machine learning to capture those missing processes and embed them inside existing models. The result was a 30 to 50 percent reduction in North Atlantic sea-surface-temperature bias, a meaningful gain in a region that drives weather across much of the Northern Hemisphere.

The team also built Samudra, an ocean emulator that runs roughly 400 times faster than the model it was trained on. Faster emulators matter because they let researchers explore many more scenarios in far less time, which is exactly what is needed when the questions involve heat, currents, and a changing ocean.

Other founding VESRI projects incorporated brittle sea-ice physics into climate models, showed that stratospheric gravity waves, the ripples that move through the atmosphere, have a larger effect on wind and weather than previously believed, and improved the scientific understanding of climate extremes.

Putting Climate Science in Decision-Makers’ Hands

Better models only help if people can use them. The MIT-led BC3 project integrated fast climate emulators into En-ROADS, a climate solutions simulator that now reaches more than 305,000 users across 169 countries. Those users can explore the regional consequences of different emissions choices, from heat and flooding to air quality and storm risk, in the places they actually live and plan for.

A New Round of the Challenge

Building on these results, Schmidt Sciences is opening expressions of interest for a new round of the VESRI Climate Modeling Challenge, and researchers can submit applications through September 11. The program launched in 2020 with a simple but ambitious goal, which was to bring the right researchers together, break down the silos that slow science down, and tackle climate modeling’s hardest problems. It now spans nine research projects across 17 countries at 65 institutions.

VESRI was founded on the belief that coordinated, team-based science could move faster and further than fragmented individual efforts. More than five years in, we’ve seen that belief prove true, and the new VESRI Climate Modeling Challenge reflects our commitment to pushing the frontier further. V. Balaji, Climate Center Lead, Schmidt Sciences

For a field that has long been constrained by computing power, hidden blind spots, and a gap between the science and the people who use it, five years of coordinated work has moved the frontier on all three at once.

Learn more about VESRI’s nine initial projects at schmidtsciences.org/vesri. This article is based on a press release from Schmidt Sciences.

SEVENSEAS Media · Conservation Journalism for the Ocean