When Hurricane Sandy pushed 14 feet of storm surge into Lower Manhattan in 2012, the water didn’t just flood subway tunnels—it exposed the fragility of one of the world’s most iconic coastal cities. Now, a new AI model developed by researchers including Matteo Lo Schiavo and Francesco Paparella at the CMCC Foundation in Italy is offering New York and other vulnerable coastal hubs a faster, more precise way to anticipate the next big surge—before it hits. Their AI emulator can predict extreme storm surges with remarkable accuracy, even under future climate conditions, and it does so in minutes instead of the weeks traditional models require. For cities where every inch of sea level rise multiplies risk, this speed isn’t just convenient—it’s transformative.
More than 780 million people live in low-lying coastal zones, where rising seas and intensifying storms are turning rare disasters into recurring threats. Traditional physics-based models have long been the gold standard for simulating storm surges, but their computational demands limit how many scenarios planners can explore. Running just one simulation can take days, making it nearly impossible to map the full range of possible futures or quantify the uncertainty behind worst-case outcomes. That’s where AI steps in. By training machine learning models to emulate the outputs of physics-based systems, researchers have created a hybrid tool that captures complex ocean dynamics—tides, wind, pressure, coastal topography—without the processing lag.
The team’s model was tested on projections for The Battery in New York City, a critical monitoring site at the southern tip of Manhattan. Under future climate scenarios, the AI predicted storm surge levels that closely matched those of high-fidelity physics models, including for extreme events exceeding historical records. Most strikingly, the AI achieved this with a 99% correlation to the physics-based results while reducing computation time from weeks to minutes. This leap in efficiency allows planners to run thousands of simulations, testing how different emissions pathways, sea level rise trajectories, or infrastructure changes might alter flood risk. For example, city engineers could rapidly assess whether a proposed seawall would hold under a 2°C warmer world or how storm surge might shift if wetlands are restored along Jamaica Bay.
The implications extend far beyond New York. Coastal cities from Mumbai to Miami, Lagos to Shanghai, face similar challenges in preparing for uncertain extremes. By making high-resolution risk assessment scalable, this AI tool helps democratize access to forward-looking flood planning. It doesn’t replace physics models—it amplifies them, turning rare projections into routine analyses. As sea levels continue to rise, the ability to explore ‘what if’ scenarios at speed could mean the difference between resilience and ruin.
This isn’t about predicting the future with certainty, but about mapping its probabilities with confidence. And with each simulation, the model sharpens our vision of what lies ahead.