Hurricanes remain the most destructive natural hazards in the United States, resulting in over $1.5 trillion in damage and more than 7,000 deaths since 1980. The main cause of these losses is storm surge, which occurs when ocean water levels rise due to strong winds and lower air pressure within hurricanes. Additional factors such as waves breaking near the coast, known as wave setup, also contribute to increased sea levels during storms.
Accurate storm surge forecasts are essential for giving coastal residents time to evacuate and allowing emergency responders to prepare. However, producing high-resolution forecasts can be slow, potentially limiting the time available for action.
Navid Tahvildari, Associate Professor of Civil and Environmental Engineering at Florida International University, stated: “As a coastal engineer, I study how storm surge and waves interact with natural and human-made features on the ocean floor and coast and ways to mitigate their impact. I have used physics-based models for coastal flooding and have recently been exploring ways that artificial intelligence can improve the speed of storm surge forecasting.”
Currently, operational storm surge forecasts use hydrodynamic models based on water flow physics. These models factor in environmental conditions such as wind speed and direction, tide timing, seafloor shape, and landscape features to estimate projected surge heights. While advances in computing power have improved these models’ speed at low resolutions across large areas, running high-resolution simulations that provide neighborhood-level detail still takes several hours.
“Those hours can be critical for communities at risk to evacuate safely and for emergency responders to prepare adequately,” Tahvildari said.
Modelers create a computational grid by dividing target areas into small cells; higher resolution means smaller cells but requires more computing power and time. Using lower resolution speeds up forecasting but decreases accuracy.
Artificial intelligence (AI) has shown promise in addressing these challenges by producing detailed predictions faster. AI models based on deep neural networks can quickly predict water levels along coastlines using wind field data. In some cases, these AI-driven predictions have been more accurate than traditional methods.
“There are two main sources of uncertainty in storm surge predictions,” Tahvildari explained. “One involves the data fed into the computer model… The second uncertainty involves the resolution of the computational grid…”
AI can also help generate forecasts for areas lacking historical data or analyze extreme conditions not previously recorded. By training on both historic and synthetic data generated from physics-based models, AI systems can quickly produce surge forecasts tailored to current wind and atmospheric conditions.
Additionally, training AI with hydrodynamic model data improves its ability to generate inundation risk maps that identify streets or houses likely to flood during extreme events without historical precedent.
Tahvildari noted: “AI is already being used in operational storm surge forecasts in a limited way, mainly to augment the commonly used physics-based models.”
He added that ongoing research efforts include using AI for storm surge prediction with observed data, assessing hurricane damage after storms, and processing images to determine flood intensity—helping validate models at detailed scales.
“As artificial intelligence models rapidly spread through every aspect of our lives and more data becomes available for training them, the technology offers potential to improve hurricane and storm surge forecasting in the future, giving coastal communities faster and more detailed warnings about the risks on the way,” Tahvildari said.
###
