Trump Wants Weather Without Climate. The Code Says No.

Key Takeaways

• On June 1, 2026, a federal judge blocked the hand-off of the supercomputing center behind America’s hurricane, wildfire, and water models, and found its operator likely to win the case.
• The lab at stake costs roughly 55 cents per American per year. Its flagship forecasting model had logged over 36,000 registered users across 162 countries as of the last published census.
• An internal OMB memo, a NOAA termination letter, and NSF’s own budget request each explain the breakup differently. One of them says the quiet part in writing.
• The plan promises to keep weather research and delete climate research. Whether the code can actually be split that way is the question this article answers.
• Washington has priced a public earth-data system for the market once before. A scene of satellite data went from $650 to $4,400, and Congress ended up reversing it by statute.

Two Government Documents That Cannot Both Be True

Start with a document the government publishes every year. The National Science Foundation’s (NSF) budget request to Congress describes the National Center for Atmospheric Research (NCAR) in Boulder, Colorado as “highly integrated programs organized around three overlapping primary areas of activity”: observational facilities, “community weather and climate models with many thousands of users,” and petascale supercomputing.

Now the newer document. In a draft memo dated November 19, 2025, staff at the White House Office of Management and Budget (OMB) directed NSF to “rescope the research and modeling of NCAR to focus on weather and not on climate modeling,” according to internal documents obtained and published by the Union of Concerned Scientists, an advocacy group that opposes the plan. Follow-up OMB communications flagged NCAR’s work on “climate variability, long-term fossil fuel-caused climate change, and atmospheric chemistry” for action because that research “informs regulations on emissions that the Administration does not support.”

One document says the weather work and the climate work are a single integrated system. The other orders a scalpel drawn between them. Both came from the same government. This article is about what happens when the second document wins.

Twenty-Four Hours in December

The public version arrived with none of the memo’s careful wording. On December 15, 2025, President Trump attacked Colorado Governor Jared Polis as a “weak and pathetic man” for refusing to grant clemency to Tina Peters, the former Mesa County clerk serving a nine-year sentence for her role in a scheme to defeat election security protocols after 2020.

What followed, according to the federal lawsuit later filed by NCAR’s managing nonprofit, was a cascade. On December 16, the Department of Transportation terminated $109 million in transportation grants for Colorado, and OMB Director Russell Vought posted on X that NSF “will be breaking up the National Center for Atmospheric Research in Boulder, Colorado,” calling it “one of the largest sources of climate alarmism in the country” and promising that “any vital activities such as weather research will be moved to another entity or location.” A White House spokesperson tied the actions to the governor directly: “Maybe if Colorado had a governor who actually wanted to work with President Trump, his constituents would be better served.”

On December 17, NSF sent formal notice that it would strip the center’s operator of its supercomputing facility. On December 19, the National Oceanic and Atmospheric Administration (NOAA) terminated a climate adaptation research agreement with the same organization, effective immediately.

That termination letter is worth pausing on, because it is the rare document that states a philosophy plainly. NOAA alleged no misconduct and no performance failure. It wrote that the program was “no longer aligned with effectuating current programmatic goals and agency priorities” because it relied on theoretical research and long-term academic partnership rather than “applied, commercializable outcomes.” The research’s offense was not being wrong. It was not being for sale.

What 55 Cents a Year Actually Buys

NCAR is not a household name, but its output is in your pocket. The center employs over 800 scientists, engineers, and support staff, and its work underpins hurricane forecasting, wildfire modeling, the National Water Model, and systems relied on by the Department of Defense, the Federal Aviation Administration, and NASA.

Its best-known product is the Weather Research and Forecasting model, or WRF: a physics engine that simulates the atmosphere at scales from a single thunderstorm to a hurricane. WRF is distributed “without cost, copyright encumbrance, or restrictions on modification,” and by 2017, the last comprehensive census published in the peer-reviewed literature, it had logged more than 36,000 registered users across 162 countries, making it “arguably the world’s most-used atmospheric model.” Wind farm operators run WRF variants to predict tomorrow’s output. Air-quality agencies run it for smoke. National weather services around the world run it operationally. NSF’s own budget documents list “wildfire management, road and aviation safety, public health, and renewable power generation” among the applications built on NCAR’s research.

The price tag for all of this is small enough to be a rounding error in federal terms. The current five-year cooperative agreement, signed in September 2023, carries a budget ceiling of roughly $938 million including interagency transfers. Against a U.S. population of about 340 million:

$\frac{\$938\text{M} \div 5\ \text{years}}{340\text{M people}} \approx \$0.55\ \text{per American per year}$

Fifty-five cents a year, at the ceiling, for the national center. NSF’s budget line for NCAR itself was $116.2 million in the FY2023 request, flat in the agency’s own estimates through FY2028.

The Scalpel Problem

Here is the technical claim at the heart of the breakup, and why working scientists treat it as incoherent: there is no “weather code” and “climate code” to sort into separate boxes.

A weather model and a climate model are the same category of machine. Both divide the atmosphere into a three-dimensional grid and march the equations of fluid motion and heat forward in time. The differences are configuration, not kind. A forecast run is initialized from the latest observations and stops at day ten, while a climate run couples the same style of atmospheric core to ocean and ice models and runs for a century. NCAR’s model families make the overlap explicit: WRF handles storm-scale weather, while the Community Earth System Model (CESM), with about 6,000 registered users of its own as of the same 2017 census, handles the century scale. They share physics research, verification datasets, staff, and the supercomputing complex in Cheyenne, Wyoming, where the Derecho supercomputer, the Casper analysis cluster, and their data systems form what court filings describe as a “highly integrated, high-performance supercomputing ecosystem.” That facility serves approximately 1,500 researchers from more than 500 universities.

Gavin Schmidt, director of NASA’s Goddard Institute for Space Studies, put the integration bluntly: “NCAR is a unique & valuable asset—far more than a climate model, or observations, or technology, or training ground, or gathering space. It covers weather, space weather, data, climate, paleo-climate, and everything in-between.”

The artificial intelligence (AI) boom does not dissolve this dependency; it deepens it. Google DeepMind’s GraphCast, the flagship of the new generation of AI forecasters, was “trained on four decades of weather reanalysis data, from the ECMWF’s ERA5 dataset,” and that dataset is itself built by running a traditional physics model to fill the gaps between historical observations. Strip out the physics-based modeling enterprise and you have not replaced it with AI. You have unplugged the machine that generates AI’s training data and its accuracy benchmark.

So when the plan promises that “vital activities such as weather research will be moved,” ask the engineering question: moved as what? The bidders want pieces. NSF’s January 2026 solicitation invited proposals for NCAR’s components and even invited interest in its Mesa Laboratory headquarters “for either private or public use.” The government simultaneously moved to hand the supercomputing center to the University of Wyoming, which backed the transfer in court. Each piece is plausible on its own. The integrated system that makes the pieces valuable is exactly what a piecewise auction destroys.

A Judge Reads the Timeline

The organization that runs NCAR, the University Corporation for Atmospheric Research (UCAR), sued NSF, NOAA, OMB, and the Commerce Department on March 16, 2026, calling the campaign unconstitutional retaliation against Colorado.

On June 1, Senior U.S. District Judge R. Brooke Jackson granted a preliminary injunction. The order is narrow in scope but not in language. It bars the government from “divesting UCAR or NCAR of any rights, resources, or responsibilities related to” the Wyoming supercomputing center while the case proceeds, and the court found UCAR “likely to succeed” on its claim that the transfer decision was “arbitrary, capricious, an abuse of discretion, or otherwise not in accordance with law.” On the motive question, the judge wrote that “the inference that retaliation played at least some role in the transfer decision is considerably strengthened by the fact that the federal government simultaneously undertook several other actions adverse to Colorado.”

The injunction protects one facility. The process for the rest of the center rolled on, and according to the complaint, an NSF program director told a workshop of more than 183 attendees in March that the majority of public input received on the restructuring “would not be considered meaningful to OMB.”

The Landsat Rerun

If you want to know how privatizing a public earth-science asset ends, you do not need a model. You need the 1980s.

In 1984, Congress decided the Landsat earth-observation satellites could be privatized, and NOAA selected a commercial vendor for the data: a venture called EOSAT. What followed, in NASA’s own retelling: “NOAA and then EOSAT restricted distribution of Landsat images and raised prices from $650 to a high of $4400 per scene,” a jump NASA characterizes as a 600% price increase that “priced out many data users.” Sales of the data “never reached predicted levels.”

Congress performed the autopsy in statute. The Land Remote Sensing Policy Act of 1992 found that “the cost of Landsat data has impeded the use of such data for scientific purposes” and that “full commercialization of the Landsat program cannot be achieved within the foreseeable future, and thus should not serve as the near-term goal of national policy on land remote sensing.” Washington ordered the next satellite built government-owned.

The structural rhyme with 2026 is uncomfortable. Then as now, an integrated public data system was pushed toward commercial operators on the theory that the market would fund what taxpayers had been funding, and the parts of the mission with no commercial buyer, which is to say the science, were priced out first. NOAA’s termination letter to UCAR, with its demand for “applied, commercializable outcomes,” reads like the EOSAT theory typed up thirty years later.

The Honest Case for Blowing Things Up

There is a serious argument buried under this wrecking job, and pretending otherwise would be its own dishonesty.

U.S. operational weather prediction genuinely lags. Cliff Mass, the University of Washington atmospheric scientist who has spent a decade documenting the problem, puts NOAA’s global model “in third or fourth place behind the European Center, the UK Meteorology Office, and often the Canadians.” He blames fragmentation: American weather modeling is scattered across five federal agencies, and NOAA declined years ago to adopt NCAR’s more modern MPAS model, a choice Mass calls “the loss of nearly a decade of effort and tens of millions of dollars.” Even UCAR’s own complaint concedes the gap, arguing NCAR’s research is needed to help American weather prediction models “catch up to superior European models.”

So the status quo deserved disruption. If the administration had consolidated the fragmented modeling effort into one independent national center, drawing on NCAR and NOAA together with one accountable leader, serious scientists would have applauded, because that is roughly what reformers like Mass have proposed for years.

That is what makes the actual plan legible. A reform consolidates. This plan fragments: pieces solicited from separate bidders, the integrated supercomputing ecosystem pried loose, the climate half deleted for openly stated regulatory reasons, all announced on social media one day after the president’s feud with the state’s governor boiled over. The reform case is the camouflage. The timeline, the memo, and the termination letter are the act.

And here the steelman leaves a mark that should worry reformers most of all. If the courts fully unwind this campaign, the verdict will read as a defense of the status quo, and the real fragmentation Mass has spent a decade documenting survives untouched. Either outcome, wreckage or reversal, leaves honest consolidation politically radioactive. The people who most needed American forecasting to get better, from wind developers already fighting federal radar-interference claims to the grid planners betting on energy infrastructure the White House treats as leverage, lose in both branches.

The Sentence Already on the Books

Litigation will grind on. UCAR still seeks broader relief, NSF is still weighing what to do with the pieces, and the Atlantic hurricane season is already underway while the people who maintain the models update their resumes.

But the ending of this experiment may already be written, because Congress wrote it last time. The finding is still sitting in the U.S. Code, in the notes to 51 U.S.C. § 60101, where lawmakers recorded what happened when America priced its own earth data like a product: the cost “impeded the use of such data for scientific purposes.” It took a failed commercial satellite, a gutted user base, and seven years of drift before that sentence became law. The models now on the auction table are better than Landsat ever was at seeing what is coming. The government dismantling them has chosen not to look.