The 2,175-mile system of interconnected artificial canals play a particularly important role when hurricanes are pinwheeled towards the peninsula to cross Florida from Orlando to keys.
Before the storm sets ashore, water managers strategically lower the canal levels to absorb incoming rainfall and storm surges. But, as the tragedy of Texas and Spain shows, floods are as unpredictable as they are dangerous. And they're getting the handles difficult. Weather conditions change in one dime, and sudden shifts can overwhelm traditional flood models.
Traditional physics-based models replicate canal systems in fine detail, taking into account factors from water flow to gate operation. Although accurate, these simulations require enormous computing power and can take nearly an hour to complete. In rapidly moving storm conditions, delays can hinder timely decision-making.
A real-time system is required. Therefore, a joint FIU team has developed an advanced AI model that can change the way floods are predicted and responded to Florida's vast canal systems. The breakthrough detailed in Journal of Water Resources Planning and Managementprovides near-simulations for flood scenarios. Then we go further and propose practical strategies.
“Accuracies are clearly very important as overestimating water stages can cause false alarms and panic, which can lead to unexpected and dangerous flood events.
“We were able to create tools that would inform water managers that would eliminate or dramatically reduce flood events.”
Students from the research group of recent FIU alumni Jimeng Shi Narasimhan, who led the research as a doctoral degree, developed a model that would run complex or worst-case scenarios in seconds.
Using nearly a decade of historic environmental and weather data collected by South Florida's Water Management District, AI systems were trained to recognize how rainfall, tides, groundwater and storm surges interact across the region. Historic storms, including Hurricane Irma (2017), Hurricane Sandy (2012), and Tropical Storms (2020), were used to fine-tune the reliability of the model. Researchers tested it on the Miami River, which runs through downtown Miami and flows into Biscayne Bay.
Researchers say this is all part of their broad efforts to make AI a reliable tool in real-world high-stakes scenarios.
Co-author and director of Jayantha Obeysekera, FIU's Sea Level Solutions Center, who previously served as the chief modeler in the South Florida Water Management District, said the technology could have surpassed immediate flood control.
“This model also has many potential as a tool to help agencies make long-term decisions,” he said. “Efficient screening of potential solutions can guide 20 or 30 years of infrastructure investment, including whether to build new pumps, reservoirs, or levees.”
