The world’s largest climate modeling initiative is quietly ramping up its next project, but U.S. participation is a wild card.
The Trump administration’s dramatic funding reductions in climate science are throwing American involvement into question, after U.S. modeling groups led the international collaboration for years. Some agencies plan to participate, but face new hurdles from staff cuts, financial uncertainties and reorganizations. Others could withdraw entirely.
Known as the Coupled Model Intercomparison Project, or CMIP, the program involves contributions from dozens of research groups around the world whose models form the backbone of climate research. Studies based on these simulations help scientists understand how greenhouse gas emissions are affecting the planet’s oceans, clouds, ice sheets and more. And they help policymakers prepare for the consequences.
Diminished U.S. involvement could undermine these studies by shrinking the pool of public modeling data and reducing confidence in the simulations, according to scientists.
But the U.S. itself might be the biggest loser, they say. Withdrawing from the project, and others, could cost the country its influence over large international collaborations. It can also rob the government of free feedback on its own models and easy access to giant pools of climate data, which are useful for everything from national security to extreme weather forecasting.
Historically, the federal government has enjoyed many times its return on investment in the project — American modeling groups contributed about 1 petabyte of data to CMIP’s last major initiative and gained access to about 15 petabytes in return. That’s a staggering amount of data — 1 petabyte is the equivalent of about 200 million good-quality photo files.
This data helps fuel scientific research that’s used to improve weather models, prepare for disasters and help the military incorporate climate impacts into national security plans.
“The U.S. has expended a lot of resources in this space for decades and decades,” said Paul Durack, a climate scientist with Western Australia’s Department of Water and Environmental Regulation and a member of the panel that helps oversee and coordinate CMIP’s activities.
Durack previously worked for the Department of Energy’s Lawrence Livermore National Laboratory, which houses one of the most prominent U.S. modeling groups historically involved in CMIP.
“This is definitely proud science history [for] America,” Durack said. “Through that leadership, there’s been a huge amount of payback.”
Links to IPCC
CMIP will carry on with or without the United States. It’s coordinated by the U.N. World Climate Research Programme and involves scientists from around the globe.
The program has existed since the 1990s, helping researchers hone computer models, coordinate experiments and share results with international colleagues. It starts a new project approximately every seven years — organizers have been planning the current cycle for several years, and participants are getting ready to start running their models.
The program isn’t well known outside scientific circles, but its outputs have underpinned global climate science for decades, including landmark reports by the U.N. Intergovernmental Panel on Climate Change.
Research groups have spent decades building and refining their computer models, designed to capture the Earth’s environmental processes and simulate the effects of greenhouse gases on the planet. Different models have different strengths and weaknesses when it comes to simulating aspects of the climate system, such as clouds, precipitation and ocean currents.
CMIP helps research groups run coordinated simulations with standardized settings and goals. Participants provide feedback on each other’s models and share results from their own simulations. This huge pool of data gives researchers greater confidence in their conclusions about climate change — one reason studies based on CMIP data are so heavily cited in IPCC reports.
The project operates on a cyclical basis, timed to coincide with IPCC assessment reports, which are released approximately every five to 10 years.
These efforts have helped models grow more advanced over the years. Each cycle identifies different priorities for modelers to work on. The current initiative, CMIP7, is focused on issues such as improving the representation of sea surface temperatures and climate extremes — like unprecedented heat or rainfall — in the simulations.
“Over each phase, the project has gotten more and more international and comprehensive and the user community has grown,” said John Dunne, co-chair of the CMIP panel that oversees the project and a researcher at GFDL, one of the U.S. modeling groups. “It’s way more effort than anyone could do in their spare time.”
‘We plan to continue’
But the U.S. role in CMIP7 is unclear.
Four American modeling groups typically participate, including scientists at NOAA’s Geophysical Fluid Dynamics Laboratory, NASA’s Goddard Institute for Space Studies, the National Science Foundation’s National Center for Atmospheric Research and the Department of Energy.
Each one has its own models and research priorities. NOAA, for instance, focuses largely on weather and climate predictions while much of NASA’s work aims to analyze and interpret satellite observations of the planet.
Some of the modeling groups are planning to carry on as usual, to the best of their abilities. But staff reductions, reorganizations and funding uncertainties have raised unexpected challenges over the past year.
“Until we are told otherwise, we plan to continue based on what we’ve been funded for by Congress,” Dunne said. A CMIP workshop will take place later this month in Kyoto, Japan, he said, and he’s received approval from the Trump administration to attend.
NOAA’s modeling work has proven useful to the federal government and the private sector, Dunne added, noting that the laboratory receives occasional visits from members of the military, the Army Corps of Engineers, insurance experts and other decision-makers.
“We get lots of internal U.S. interest in the CMIP data for a whole variety of applications,” he said. “And so the case is pretty strong that this information is considered very useful for planning.”
But Dunne added that staff reductions over the past year have taken their toll. More than 2,000 NOAA employees left the agency in 2025 through layoffs, early retirements and other reductions in force. That includes some modeling experts who would have helped with the CMIP simulations.
“We’ve lost a lot of expertise, so that’s gonna be a scramble to try and make sure that we’re not missing competence,” Dunne said. “There’s a huge amount of memory, the institutional memory, that we rely on.”
The other three modeling groups face additional uncertainties.
‘Waiting on … guidance’
A NASA communications officer confirmed that scientists from the agency will be represented in the CMIP7 cycle. But the White House last month withheld funding for NASA science missions, flouting a federal spending bill passed by Congress in January. It’s unclear how long the pause will last or how it might impact the agency’s participation in CMIP.
The National Center for Atmospheric Research, meanwhile, is facing existential threats. The White House announced in December its intent to dismantle the institution, which is funded by the National Science Foundation and famed for its weather and climate research.
That hasn’t happened yet. But NSF announced last month that it would transfer NCAR’s Wyoming-based supercomputing facility to a different manager. The new operator has not yet been publicly named. And it’s still unclear how the move will affect NCAR’s ability to run its usual CMIP simulations.
David Hosansky, a media relations manager for NCAR, said in an email that the institution is “waiting on NSF guidance and in the meantime operating as normal.”
The Department of Energy historically contributes fewer simulations to CMIP than the other modeling centers. But it’s unclear whether the agency will participate in CMIP7. A spokesperson for Lawrence Livermore National Laboratory, which houses the modeling group that typically contributes, said several staff members were out of the office and he was unable to answer questions about CMIP before publication.
Potential reductions in U.S. involvement could ripple across the research world, Dunne said. In a worst-case scenario, the current cycle, CMIP7, could fail to improve upon the model outputs produced in CMIP6. That would mean “there’s no progress,” Dunne said.
U.S. modeling groups who reduce or curtail their participation would also miss out on the benefits of free feedback from experts, according to another federal scientist familiar with the CMIP process.
“That model would suffer from not being exposed to scrutiny by researchers all over,” said the scientist, who was granted anonymity for fear of facing retaliation by the Trump administration.
‘Diminished’
DOE historically has played another major role in the CMIP cycles. It helps set standards for the quality of the project’s data, and it stores that data on its own servers. It’s one of just a few hubs worldwide that serves as a global CMIP data repository.
That’s an important service to the international community. Scientists who want to use CMIP data in their research would otherwise have to request it from individual participants. Accessing everything from a single node is faster, reduces the risk of disruptions and ensures all the data is formatted in a standardized way.
The U.S. benefits, too. It gives the country global influence when it comes to designing standards for data quality and formatting. And quick access to a giant pool of global simulations makes it easier to work on big-data projects, like AI models.
It’s unclear if DOE will provide these services for CMIP7.
“There’s no other country on this planet that has the same kind of computer firepower the U.S. has,” said Durack, who formerly worked for DOE’s Lawrence Livermore National Laboratory. “Not turning on those resources means that data is never gonna reside in the U.S., which really hurts your intention to have a big data pool for your AI tools.”
CMIP7 likely won’t conclude for a few more years. So there’s still time for U.S. agencies to make decisions about the extent of their contributions.
Still, Durack noted, U.S. leadership in CMIP has already been “quite diminished over the last 12 months at least.”
