Though the 2025 hurricane season was relatively quiet for the United States, researchers are combining massive amounts of observational data with wildly complex computer models to predict the impact of future storms on coastal communities.
The University of Florida’s Engineering School of Sustainable Infrastructure & Environment, or ESSIE, is part of a project that recently received the Excellence in Partnering Award by the National Oceanographic Partnership Program for its collaborative, multi-institutional effort aimed at improving prediction of coastal storm impacts. ESSIE’s team focused on creating models to predict such complex storm effects as breaching, erosion, water levels and property damage.
Researchers hope the project will improve the accuracy of the models and enable residents and emergency managers to target specific risks and make storm preparations more effective and timelier.
A key element of the project was testing models for accuracy, comparing predicted outcomes created with their models with actual outcomes. They used pre- and post-storm data from six hurricanes to test their models and predictions.
Consisting of nine teams, each with a lead principal investigator, researchers on the project represented 13 unique academic institutions, 13 government agencies or divisions, and 11 industry performers.
The Excellence in Partnering Award is given annually to a National Oceanographic Partnership Program project that best exemplifies the program’s objective of developing a successful network of partnerships to advance ocean sciences.
Maitane Olabarrieta, Ph.D., professor in the UF Department of Civil & Coastal Engineering, was one of nine lead principal investigators on the project. She led the Waves, Sediment, Surge and Structure Response Forecasting System team. She is working alongside Arthriya Subgranon, Ph.D., a former assistant professor at the Herbert Wertheim College of Engineering.
The massive project spanned four years and wrapped up in April. While Floridians are accustomed to predictions of a storm’s path and associated surge, researchers wanted to take hurricane prediction to the next level, providing reliable and useful predictions of storm effects that really matter — infrastructure and housing damage, erosion and accretion (the accumulation of sediment or matter). Olabarrieta said these real-world effects are notoriously hard to predict.
So, how did the teams expand the scale of their predictions?
Data.
So much data.
As part of the research, nine teams would spring into action each time a storm threatened to make landfall in Florida. The teams worked across a diverse set of modalities: atmospheric forecasting, high-resolution digital elevation mapping, remote sensing using synthetic aperture radar, drifting wave buoy deployment, nearshore sensor deployment and oceanic forecasting.
As part of the oceanic forecasting effort, Olabarrieta’s team created models designed to predict coastal erosion and damage to infrastructure, combining data collected across all the teams and applying machine learning models — trained on masses of historical hurricane data — to ultimately create the prediction.
One innovation was the inclusion of data gleaned from public sources.
Subgranon and her doctoral student, Steven Klepac, mined data from local building permit offices for information about documented damage from previous storms. Building permits (applied for in rebuilding efforts) typically include information such as first-floor elevation, roof and wall materials and even neighborhood density. Their machine learning model, dubbed BuildForce, was trained on this historical data and ultimately fed information into the forecasting model.
The combination of such varied and granular data with powerful prediction models, all running on UF’s HiPerGator supercomputer, enabled researchers for the first time to begin to predict specific storm effects such as water levels, building damage and coastal morphology changes.
Leveraging the reams of data researchers had already collected, the pre- and post-storm comparisons revealed three of their models performed strongly in predicting storm waves, water levels, breaching, erosion and damage during recent hurricanes.
As the goal is collaboration and improving predictions, all project data are publicly available on DesignSafe, a data repository funded by the National Science Foundation. The hope is that releasing the data to researchers worldwide will invite further collaboration and continued model improvements.
To date, the effort has produced more than 10 scientific papers and over 50 conference/seminar presentations, underscoring its visibility and leadership in coastal resilience research.
Researchers also collaborated with government agencies to create an engaging map to visualize the data and methods for the project. View it here.
Olabarrieta sees this work as nothing short of transformative.
“In an era of accelerating climate change, rising sea levels and increasing storm intensity, improving coastal impact prediction is an urgent societal need,” she said. “This project highlights the power of large-scale, collaborative science to meet that challenge. We are deeply thankful to our sponsors and collaborators for their exceptional partnership and for making this transformative project possible.”