Hurricane Forecasting and the FY-26 NOAA Budget Proposal

If NOAA were really interested in providing the American public with “gold-standard, cutting-edge data, models, and forecasts”, they wouldn’t be trying to get rid of the people who make it happen.

NOAA’s FY26 budget proposal shuts down the vast majority of NOAA’s research side (Oceanic and Atmospheric Research, OAR, or “NOAA Research”), with wide-ranging implications. Substack posts by former colleagues Michael Lowry and Alan Gerard nicely cover the background and broader implications of the proposal to shutter the NOAA Labs and Cooperative Institutes, with more depth than I want to get into here and I recommend these posts highly. In addition, this USA Today piece provides some nice background. My objective here is to amplify on a Bluesky post of mine that listed some of the major contributions of NOAA Research to operations at the National Hurricane Center (NHC), and share a few thoughts on how the budget proposal would affect future forecasts.

The NOAA Labs and Cooperative Institutes (CIs) have been in place for decades and have been responsible for some of the most important advances in hurricane forecasting and products at the NHC. The organizations most relevant to hurricane operations today would be the Atlantic Oceanographic and Meteorological Laboratory (AOML), the Cooperative Institute for Research in the Atmosphere (CIRA), the Cooperative Institute for Meteorological Satellite Studies (CIMSS), and the Cooperative Institute for Marine and Atmospheric Studies (CIMAS). During my time at OAR I contributed to a couple of those advances, and while at NHC I was the beneficiary of many more. Without the Labs and CIs, NHC would not have seen the following major advances and innovations:

• The GFDL model, which was NOAA's premier regional tropical cyclone model for over a decade.

• The Hurricane Analysis and Forecasting System (HAFS), NOAA’s state-of-the-art, high-resolution, coupled modeling system for tropical cyclones.

• SHIPS and LGEM, the leading statistical tropical cyclone intensity forecast models.

• SHIPS-RI, the first successful tool for forecasting rapid intensification.

• Atmospheric motion vectors for assimilation into global forecast models.

• Satellite-based shear analyses, and analyses of the Saharan Air Layer (SAL).

• The Advanced Dvorak Technique (ADT), for automated estimates of tropical cyclone intensity based on satellite imagery.

• Use of dropwindsondes in the hurricane environment for improving track forecasts. It was initial research from OAR that led to NOAA’s acquisition of the Gulfstream-IV jet.

• Use of dropwindsondes to document the vertical structure of the hurricane eyewall, and the development of the "90% rule" to estimate TC intensity from reconnaissance flight-level observations.

• Doppler radar processing and analysis techniques that improve real-time model track and intensity forecasts by up to 30%.

• The Stepped-Frequency Microwave Radiometer (SFMR) to measure surface winds in TCs, developed on the NOAA-P3s but now installed on all Hurricane Hunter aircraft.

• The Wind Speed Probability and Time of Arrival maps, the foundation of NHC's move from deterministic to probabilistic decision-support tools.

This is by no means a comprehensive list of OAR’s contributions, but just the first things that came to mind when thinking back about the day-to-day life of an NHC hurricane forecaster. This list touches on nearly all aspects of the hurricane forecast process, from data gathering to analysis to modeling to forecasting to products and communications: When forecasters sit down to prepare the Tropical Weather Outlook, they’ll likely consult SAL and shear analyses from CIMSS. When they estimate a storm’s strength and size, they’ll consult the CIMSS ADT, or adjust flight-level winds to the surface with the 90% rule, or use SFMR data developed and calibrated by AOML. No NHC Hurricane Specialist prepares their TC forecast without consulting one or more of the models developed at least in part at a NOAA Lab or CI. When storms threaten landfall, their suite of models will have been enhanced through the assimilation of dropwindsonde and AOML-processed Doppler radar data. And when the forecast is done and it’s time to communicate risks to the media, emergency managers, and the public, they’re now able to do so with probabilistic products developed by CIRA scientists.

I don’t really want to get deep into the financial merits of the budget proposal because that’s not my area of expertise. But a couple things are worth noting. Over the past 25 years, NHC’s track errors have been cut in half and their intensity errors by about a third. Not all of this improvement is attributable to NOAA Research, of course, and some of it comes from advances completely outside of the United States (e.g., the ECMWF model). But a healthy chunk of the improvement is, and I’d even wager that a majority of the intensity progress flowed from NOAA Research funding. A 2025 study estimated that the intensity forecast improvements that occurred from 2007-2022 produces savings of $2B per U.S.-landfalling hurricane.* In contrast, the proposed annual budget savings from shuttering all of the NOAA Labs and Cooperative Institutes is $90M.

So what happens to forecasters if the NOAA budget proposal becomes reality? Here are a few of my short-term expectations and longer-term concerns:

• Immediately the flow of Doppler radar data from NOAA aircraft into numerical models would stop, because OAR-funded staff are needed to process the data onboard the aircraft. These “tail-Doppler Radar” (TDR) missions are tasked by the National Weather Service when tropical cyclones threaten landfall. Regional model forecasts (e.g., from HAFS) would be degraded without the TDR data.

• Existing operational models that depend on NOAA Research support (e.g., SHIPS, LGEM, SHIPS-RI) could fall out of service as things break or as configuration changes are required by operations.

• Hurricane forecasters would immediately lose access to an array of tools and analyses currently accessed through the websites of the NOAA CIs (e.g., satellite analyses from CIMSS). Situational awareness and initial analyses of storms out of reconnaissance range would be degraded.

• Longer term, opportunities to develop and test new aircraft observing systems would be eliminated or diminished. Who is going to develop the next generation of observing tools (like the small drones currently being tested at AOML), and on what platform? In addition to cutting OAR, the NOAA budget also eliminates efforts to replace the aging P-3’s, which have been an incredibly productive test bed for in-situ instrumentation over the past half-century, and a source of ground truth validation for satellite-based instruments.

• In a chronically short-staffed National Weather Service, who is going to be left to collaborate with forecasters to develop operationally useful tools? It’s hard to express just how much interaction occurs between product developers at the Labs and the CIs with their forecaster and technical counterparts at NHC. A community of scientists now exists across the country who have come to understand operational needs and how to meet them. That community is about to be scattered to the winds. Once lost, it would take years or decades to rebuild.

Nearly every season NHC touts new accuracy records, and the forecast cone keeps shrinking. In meteorology, we sometimes talk about the “limits of predictability” - that theoretical future time when forecasts become as good as they can get, and we would no longer be able to set new records. Since my career began over 40 years ago, I’ve seen hurricane forecast accuracy bust through estimate after estimate of this limit. And in just the last decade, we finally started making progress on one of the hardest hurricane problems of all - rapid intensification. So I expect records could continue to fall for years to come.

Knowing that these improvements save billions of dollars with each landfalling hurricane, shouldn’t we want to keep trying to make forecasts better? Twenty-five years from now, will we want our children to say that we reached the limit of predictability, not because nature stopped us, but because we decided to stop trying?

* Note - The original version of this post misquoted a savings of $5B per landfalling hurricane; that value actually only applied to major hurricanes but I missed that detail. The latest version of the study, which extended the analysis to all hurricanes and added an additional two years of data, estimates the savings per hurricane landfall to be $2B.