Our Science Using AI

NSF NCAR is reimagining how Earth system science is conducted through the transformative use of AI. We integrate AI and ML with observations, data assimilation, and physics-based models to uncover hidden patterns, accelerate discovery, and expand the limits of predictability. From hybrid AI–physics modeling and AI-driven parameterizations to emulators, explainable AI frameworks, and advanced uncertainty quantification, our work advances fundamental understanding and enables more skillful, actionable predictions across the Earth-Sun system.

The projects highlighted below demonstrate how AI is reshaping research across NSF NCAR—driven by internal innovation and strengthened through collaboration with external partners and the broader research community.

Atmospheric Chemistry Research to Understand Wildfires and Earth System Impacts:

AI improves the representation of atmospheric chemistry and aerosols, enhancing understanding of wildfire impacts and coupled Earth system feedbacks.

Atmospheric Chemistry and Aerosol Emulation for Earth System Models

Revealing the Global Expansion of the Wildland–Urban Interface (WUI) Fires with Machine Learning

Data Assimilation and Observational Integration

AI enhances data assimilation by improving how models and observations are combined. AI-augmented workflows reduce computational costs, accelerate the use of new satellite datasets, and support higher-resolution reanalyses.

AI-Enabled Data Assimilation for Next-Generation Weather Prediction

Hydro-DART for AI-Ready Flood Prediction

Weather-to-Subseasonal Predictability Research to Improve High-Impact Event Prediction

AI enhances high-impact hazard forecasts and advances research to extend predictability from weather to seasonal timescales.

Real-time AI-Driven Medium-Range Convective Hazard Forecasts

ML for Subseasonal-to-Seasonal (S2S) Prediction

Geospace and Solar Research to Advance Space Weather Predictability

AI enhances modeling of solar magnetic processes and Sun–Earth coupling, strengthening prediction of space weather and its relevance to critical infrastructure.

PINNBARDS: Physics-Informed AI for Predicting Solar Storms

Digital Twin of the Sun

Urban-Scale Earth System Research to Inform Urban Risk and Resilience

AI enables high-resolution modeling of urban processes and interactions between infrastructure, improving understanding of risk and system dynamics in complex built environments.

Investigating Air, Heat, and Turbulence in City Corridors

Understanding Processes and Predictability

AI helps researchers uncover sources of predictability, identify model biases, and reveal new process-level insights. Explainable AI and causal discovery methods provide tools for understanding, not just forecasting, the Earth system.

Explainable and Interpretable AI (XAI/IAI) Research Examples

Calibrating Terrestrial Models with AI

AI-Driven Field Campaigns and Data Collection

AI and machine learning create new possibilities for designing, conducting, and analyzing field campaigns. NSF NCAR is actively exploring how AI can strengthen observational strategies, improve data usability, and accelerate scientific insight. Emerging directions include:

Applying AI to help design targeted field campaigns that address critical data and knowledge gaps relevant to modeling and data assimilation
Automating quality assurance and quality control of raw sensor data in real time or near-real time
Developing natural language–based tools to expand discovery and use of observational datasets across scientific domains
Enabling AI-ready observational data to support advanced analytics and model development
Advancing data merging and gap-filling approaches

Project Spotlights:

For more information or opportunities to get involved, please contact: ncarai@ucar.edu