AI predicts Florida canal floods in seconds, giving water managers real-time options

AI flood model gives Florida water managers up-to-the-minute decisions

Florida's 2,175-mile canal network is the state's safety valve when hurricanes close in. Water managers drop canal levels ahead of landfall to absorb rain and surge. But storms shift fast, and conventional tools can miss turning points. That raises exposure to both false alarms and surprise flooding.

Why traditional models fall short in a storm

Physics-based models mirror gates, flows and operations with high precision, but they take close to an hour to run. In a fast-moving event, that delay can cost options. Managers need results in seconds, not after the window to act has closed.

A faster path: FIU's AI model for real-time flood decisions

A collaborative team at Florida International University built an AI model that simulates flood scenarios almost instantly and suggests actionable strategies. It was trained on nearly a decade of data from the South Florida Water Management District, learning how rainfall, tides, groundwater and storm surge interact across the region. Historic events such as Hurricane Irma (2017), Hurricane Sandy (2012) and Tropical Storm Isaias (2020) helped fine-tune reliability. Researchers tested it on the Miami River, which drains into Biscayne Bay.

"Accuracy is obviously very important to us, because overestimation of water stages can cause false alarms and panic while underestimation can result in unexpected and dangerous flooding events," said Giri Narasimhan, an FIU Knight Foundation School of Computing and Information Sciences professor. "We were able to create a tool that provides water managers with the information to either eliminate a flood event or drastically reduce it."

Recent FIU graduate Jimeng Shi led the research as a Ph.D. student in Narasimhan's group, building a system that runs complex or worst-case scenarios in seconds. The work reflects a broader push to make AI trustworthy in high-stakes, real-world operations.

What managers can do with it

• Cut decision latency from ~1 hour to seconds during storm operations.
• Run "what-if" gate and pump strategies before committing field crews.
• Balance risk: avoid overreacting (costly false alarms) and underreacting (dangerous flooding).
• Create a shared operating picture between control rooms and incident command.
• Prioritize resources to the most sensitive reaches along the canal network.

Strategic value beyond a single storm

Study co-author Jayantha Obeysekera, director of FIU's Sea Level Solutions Center and former chief modeler at the South Florida Water Management District, sees longer-range value. "The model also holds a lot of potential as a tool to help agencies make longer-term decisions," he said. "It could guide 20- or 30-year infrastructure investments, such as whether to build new pumps, reservoirs or levees by screening potential solutions efficiently."

For executives, that means faster pre-screening of capital plans, clearer ROI scenarios and tighter alignment between operations and capital planning. The result: fewer expensive missteps and a stronger case for funding the right projects.

Implementation checklist for leadership

• Integrate data pipelines: rainfall, tide gauges, groundwater, surge forecasts, gate/pump telemetry.
• Establish model governance: version control, audit trails, and sign-off criteria before deployment.
• Run side-by-side drills: compare AI outputs with physics-based runs and field observations.
• Keep a human-in-the-loop: define thresholds where operators must review and approve actions.
• Track KPIs: forecast error, decision lead time, false-alarm rate, avoided damage estimates.
• Plan cyber resilience: backups, access controls, and failover to conventional models.
• Train teams: briefings for ops, emergency management and leadership on interpreting outputs.

The study is detailed in the Journal of Water Resources Planning and Management. For organizations building AI capability across leadership and operations, consider practical upskilling paths such as AI courses by job function.

Real-time situational awareness is becoming a baseline requirement for storm response. This model shows how AI can compress time-to-decision and improve outcomes across Florida's canal system-without sacrificing accuracy.