๐ฌ๐งThe Guardian TechnologyโขStalecollected in 30m
Trump budget cuts threaten NOAA's AI weather forecasting models
๐กUnderstand how federal budget shifts impact the availability of critical training data for large-scale AI models.
โก 30-Second TL;DR
What Changed
NOAA recently deployed AI-powered global weather models to improve forecasting speed and efficiency.
Why It Matters
The degradation of high-quality training data could lead to 'model drift' or reduced accuracy in critical climate forecasting, impacting public safety and disaster preparedness.
What To Do Next
If you are building models on public datasets, implement robust data versioning and caching to mitigate risks from potential upstream data source outages.
Who should care:Researchers & Academics
Key Points
- โขNOAA recently deployed AI-powered global weather models to improve forecasting speed and efficiency.
- โขModels are trained on centuries of historical weather data to ensure predictive accuracy.
- โขProposed budget cuts to climate data programs threaten the data pipeline required for model training.
- โขReliability of hurricane and extreme heat forecasts is at risk due to potential infrastructure degradation.
๐ง Deep Insight
Web-grounded analysis with 21 cited sources.
๐ Enhanced Key Takeaways
- โขNOAA's recently operationalized AI-driven global weather models, AIGFS, AIGEFS, and HGEFS, significantly reduce computational resource usage (AIGFS uses 0.3% of traditional GFS resources) while enhancing forecast speed and accuracy, particularly for tropical cyclone tracks.
- โขThese advanced AI models were developed under Project EAGLE, a collaborative initiative involving NOAA Research, the Earth Prediction Innovation Center (EPIC), academia, and private industry, notably leveraging Google DeepMind's GraphCast model as a foundational architecture.
- โขThe 'Trump budget cuts' refer to actions taken during a hypothetical second Trump administration (starting January 2025), which have resulted in hundreds of NOAA employee layoffs, cessation of weather balloon launches in some regions, and the discontinuation of updates for critical climate data products like the 'Billion Dollar Weather and Climate Disasters' database.
- โขProposed budget cuts for 2026 under this administration target national laboratories and observatories, including the Mauna Loa Observatory, which has maintained the longest continuous record of atmospheric CO2 levels globally since 1958.
- โขBeyond operational cuts, the administration has also terminated grants and contracts vital for climate research, such as funding for the CarbonTracker Program at Harvard University and the Cooperative Institute for Modeling the Earth System (CIMES) at Princeton University.
๐ Competitor Analysisโธ Show
| Competitor/Entity | Key Features/Models | Benchmarks/Performance | Notes |
|---|---|---|---|
| NOAA (U.S.) | AIGFS, AIGEFS, HGEFS (Hybrid-GEFS), HRRR-Cast | AIGFS uses 0.3% compute of GFS for 16-day forecast; HGEFS extends forecast skill by 18-24 hours over traditional GEFS; HRRR-Cast 100-1000x more efficient than operational HRRR. | Hybrid AI-physics approach; leverages Google DeepMind's GraphCast. |
| European Centre for Medium-Range Weather Forecasts (ECMWF) | AIFS (AI Forecast System) | Launched AIFS in Feb 2025; considered a world leader in atmospheric prediction. | NOAA's hybrid ensemble approach is seen as more robust for extreme outliers. |
| Google DeepMind | GraphCast, GenCast | GraphCast used as foundation for NOAA's models; GenCast extends reliable forecasting from 10 to 15 days, outperforming ECMWF forecasts 97.2% of the time. | AI-first approach, highly efficient, requires less computational power. |
| Microsoft | Aurora | More accurate than ECMWF 92% of the time in 10-day forecasts during 2023. | Actively developing global AI models. |
| IBM Weather Company | Advanced AI, big data analytics, cloud computing | Offers extremely accurate, live weather information globally. | Caters to agriculture, transportation, insurance, retail with hyperlocal forecasts. |
| AccuWeather | AI, data-driven analysis | Focus on accurate local forecasts and severe weather warnings. | Informs business and personal decision-making. |
| DTN | Sophisticated modeling, satellite imagery, AI-based analytics | Provides accurate, industry-specific weather data and analytics. | Serves agriculture, energy, transportation, and aviation. |
๐ ๏ธ Technical Deep Dive
- NOAA's new operational suite includes three distinct AI-driven global weather prediction models: AIGFS (Artificial Intelligence Global Forecast System), AIGEFS (Artificial Intelligence Global Ensemble Forecast System), and HGEFS (Hybrid-GEFS).
- AIGFS is an AI-based system designed for improved speed and efficiency, capable of generating a 16-day forecast using only 0.3% of the computing resources of the traditional GFS and completing it in approximately 40 minutes.
- AIGEFS is an AI-based 31-member ensemble system that provides a range of probable forecast outcomes, showing improved performance over the traditional GEFS by extending forecast skill by 18 to 24 hours.
- HGEFS is a pioneering 62-member 'grand ensemble' that combines the 31 AI-based members of AIGEFS with the 31 physics-based members of NOAA's flagship Global Ensemble Forecast System (GEFS), consistently outperforming both AI-only and physics-only ensemble systems.
- The core AI models, particularly those under Project EAGLE, are based on Google DeepMind's GraphCast architecture, which NOAA's Environmental Modeling Center (EMC) fine-tuned using NOAA's own Global Data Assimilation System (GDAS) analyses.
- HRRR-Cast, NOAA's first regional experimental AI forecast system, utilizes a ResNet architecture (specifically ResHRRR) with convolutional neural networks enhanced by squeeze-and-excitation blocks and Feature-wise Linear Modulation. It supports probabilistic forecasting via the Denoising Diffusion Implicit Model (DDIM).
- HRRR-Cast is significantly more computationally efficient, being 100 to 1000 times faster than the operational HRRR, and was trained on three years of HRRR analysis data (2021-2024).
- The global AI models operate on a 0.25-degree latitude-longitude grid (approximately 28 km) and 13 pressure levels, producing 16-day forecasts twice daily.
๐ฎ Future ImplicationsAI analysis grounded in cited sources
The accuracy and reliability of future weather and climate forecasts will decline.
AI models, despite their efficiency, are heavily dependent on continuous streams of high-quality observational and historical data for training and validation, which is directly threatened by cuts to data collection programs and observatories.
The United States risks losing its leadership in global weather prediction and climate science.
While NOAA has achieved significant advancements with its new AI models, sustained cuts to research funding and critical data infrastructure could impede further innovation, allowing other global meteorological agencies and private sector entities to surpass U.S. capabilities.
Public safety and economic resilience against extreme weather events will be compromised.
Less accurate and timely forecasts, coupled with reduced data on the frequency and intensity of extreme weather events, will hinder effective disaster preparedness, emergency response, and economic planning for industries reliant on precise weather information.
โณ Timeline
2022-10
NOAA discusses integrating AI/ML into forecast systems
2024-12
Google DeepMind's GenCast demonstrates superior AI weather forecasting capabilities
2025-01
Donald Trump's second administration begins, establishing the Department of Government Efficiency (DOGE)
2025-02
Approximately 1,300 NOAA employees are laid off as part of federal workforce reductions
2025-05
Proposed NOAA budget cuts are leaked, targeting climate functions and ending the 'Billion Dollar Weather and Climate Disasters' product
2025-08
NOAA Research develops HRRR-Cast, its first regional experimental AI forecast system under Project EAGLE
2025-12
NOAA officially operationalizes its new suite of AI-driven global weather models (AIGFS, AIGEFS, HGEFS) under Project EAGLE
2026-03
Further budget cuts lead to reduced weather balloon launches and threats to other data sources
๐ Sources (21)
Factual claims are grounded in the sources below. Forward-looking analysis is AI-generated interpretation.
๐ฐ
Weekly AI Recap
Read this week's curated digest of top AI events โ
๐Related Updates
AI-curated news aggregator. All content rights belong to original publishers.
Original source: The Guardian Technology โ