Rice AI Tool Simulates 10,000 Synthetic Cyclones to Improve Hurricane Risk Modeling

Researchers at Rice University are developing an open-source AI tool designed to enhance hurricane risk modeling by simulating thousands of potential tropical cyclones. The project, titled “Leveraging Large Earth Foundation Models for Probabilistic Hurricane Hazard Assessment,” leverages NASA’s Prithvi-WxC weather and climate foundation model, created in partnership with IBM. The tool aims to overcome a major hurdle in hurricane risk assessment: the limited historical data available to predict the frequency and severity of future storms.

The Rice team, led by Avantika Gori and Guha Balakrishnan, plans to generate a dataset of approximately 10,000 synthetic cyclones. These simulations will represent a broader range of possible storm tracks, wind conditions, and rainfall patterns than historical records alone can provide. The researchers will test the tool’s efficiency by simulating storms from their formation to dissipation, accounting for surrounding weather conditions, and refining the data with physics-informed models to capture complex hurricane characteristics like spiral rainbands.

The project could offer valuable insights for researchers and insurers by estimating the probability of hurricane landfalls and the severity of wind and rainfall impacts on specific locations. It may also support portfolio-level risk assessments, helping insurers evaluate combined financial exposures across multiple properties. The team collaborates with CERCat, an industry consortium involving Rice and Lehigh universities, and plans to make the tool open-source for broader use.

Gori, an assistant professor of civil and environmental engineering, emphasized the potential of AI to explore realistic storms beyond historical records. Balakrishnan, an assistant professor of electrical and computer engineering, highlighted the integration of computer vision techniques with physical storm behavior to build accurate models. The researchers hope their work will reduce uncertainty in natural hazard risk estimates, leading to more transparent pricing and greater willingness to insure complex risks.