AI: The Future of Floodplain and Flood Risk Management

As intense storms and flooding become more common throughout the United States, communities need better ways to address existing flooding challenges while mitigating their risk against future severe weather events. With communities demanding immediate attention and action, floodplain and flood risk management professionals face increased pressure to efficiently deliver innovative solutions. Artificial Intelligence (AI) and Machine Learning (ML) can help.

By integrating AI into existing workflows, we can streamline processes to improve floodplain management. For example, new patterns analyzing economic impacts of mitigation investments identified by ML are creating novel ways of evaluating and measuring how various mitigation options affect return on investment. This technology is also enabling highly-specialized disciplines like hydrology and hydraulics to produce better results faster. In the past, these disciplines were overly complex or expensive for most communities.

Additionally, AI-driven solutions are being used to solve complex asset management, operational decision making and damage assessment challenges. In response to Hurricanes Irma and Maria in Puerto Rico and the U.S. Virgin Islands, we worked with the Federal Emergency Management Agency to remotely assess damage using ML. Doing so reduced on-the-ground field inspections by 80%. Likewise, our team relied on satellite imagery to determine roof types, superstructures and other building infrastructure throughout Puerto Rico. This enabled us to refine the island’s digital building inventory, including specific structural characteristics, to improve future hazard modeling.

Furthermore, flood modeling tools incorporating these advanced technologies such as digital twins allow communities to see the potential impact of future severe weather events and prepare for them. These tools use actual community characteristics as a baseline and mimic the patterns of more intensive calculations. By including additional scenarios and climate or development conditions, their models can provide better real-time forecasts. In 2013, a severe storm flooded Colorado’s Boulder County, dumping a year’s worth of rain in just a few hours. This prompted county officials to launch a resiliency study to enhance floodplain management and protect critical infrastructure. As part of the study, we ran a simulation using digital-twin technology demonstrating the impact future rainfall projections would have on the county through 2050. Our models showed more severe floods were possible in the coming decades and indicated infrastructure originally built to withstand once-per-century weather events may need upgrades. Today, we are partnering with government agencies to further accelerate flood modeling nationwide through the adoption of emerging ML technologies.

As the use of AI and ML technologies continues to grow, we must quickly identify how to align their capabilities with the needs of engineers, community planners and floodplain managers. Technical products, such as flood maps or consequence projections, rely on careful coordination and validation of underlying sector-specific data and tools. However, without proper validation, AI-based flood models can provide realistic-looking results that are misleading at best and inaccurate at worst. This is particularly true in edge cases, like flooding near bridges, culverts or highly-urbanized areas. If underlying sector-specific data and tools are not validated, these models may not meet required regulations, generate a wide range of possible outcomes, or produce false results. In addition, sometimes novel frameworks must be developed so known limitations, such as geographic conditions or difficult hydraulic situations, will be considered. Experienced flood risk management practitioners can create these frameworks and directly apply regulatory or physical constraints within them to ensure accurate results.

To effectively align these advanced technologies with day-to-day floodplain and flood risk management tasks, getting feedback from industry professionals is vital. Our team participates in events such as the Association of State Floodplain Managers’ annual conference and the Gilbert F. White National Flood Policy Forum to engage in productive conversations around the use of AI in floodplain and flood risk management. Moving forward, the insights gained from these intensive, detailed conversations will help foster the responsible deployment of AI technologies in hazard identification, response, mitigation and communication activities.

By implementing AI in floodplain management workflows, we can help communities better respond to current severe weather events and mitigate their risk against them in the future. This will enhance flood safety and resilience throughout the United States and worldwide for years to come.