AI Resilience Report forAtmospheric and Space Scientists

Today’s weather and climate scientists are using more AI tools to help with data and forecasts. For example, NASA and IBM built a new AI “foundation model” trained on 40 years of weather data to improve storm predictions and climate analysis ^[1]. Private companies also use AI: one startup (Atmo) trains AI on 60 years of climate measurements and real-time satellite data, producing forecasts in seconds that are much faster and even more accurate than old methods ^[2].

NOAA experts report that early AI models can better predict extreme events like hurricanes or heat waves ^[3]. On the data-gathering side, satellites and automated instruments already collect weather measurements around the globe. New tools add to this: for example, a Swiss company uses drones to automatically measure weather conditions and feed data into models ^[4].

Researchers are also blending computer graphics and physics: Google’s new “NeuralGCM” model mixes physics equations and AI to improve short-range forecasts ^[4]. In short, many core tasks – interpreting data and running forecasts – are now helped by AI-driven models. Other tasks like writing formal reports or talking with officials are still mostly done by people.

Humans still make the final judgments and explain the weather for the public.

AI is being adopted in weather science because there is lots of data and a big payoff for faster forecasts. Governments and businesses are investing in the technology. For example, NOAA is spending \$100 million on new supercomputers to run AI and machine learning models ^[4].

Workshops with NOAA and White House experts noted that AI tools can improve forecasts and urged NOAA to responsibly integrate these new methods ^{[3] [3]}. At the same time, experts say these AI tools will supplement – not replace – traditional methods for now ^[4]. Weather forecasting must be very reliable (it protects lives), so scientists and the public want to make sure AI is tested carefully ^{[4] [3]}.

In practice, this means adoption is happening but with caution. In summary, new AI tools are growing fast in meteorology, speeding up data analysis and forecasts. The good news for students is that human expertise remains vital – scientists are still needed to check the models, understand results, and communicate about the weather and climate.