Closing the Loop: AI-Enabled Forecasting, Modeling, and Operations for Urban Flood Resilience

Utilities and municipalities are under increasing pressure to make faster, better-informed decisions before, during, and after storm events. Yet in many organizations, rainfall forecasting, system operations, model analysis, and resilience planning still occur in separate workflows. This presentation describes a more connected approach: an AI-enabled operational framework that brings these functions together to support both immediate flood response and longer-term stormwater system improvement. The work centers on urban flooding and sewer system applications, where short lead times, rapidly changing rainfall patterns, and network complexity demand timely and reliable decision support. Digital ecosystems leveraging AI layers that integrate radar-based rainfall ingestion, short-horizon nowcasting, field observations, and physics-based hydrologic and hydraulic models can provide a continuously updated picture of expected system conditions. These tools can be used to translate gridded rainfall into model-ready inputs, accelerate scenario evaluation, and support operational decisions related to conveyance, storage, real-time control, and flood management. A key message of the presentation is that AI delivers the most value when paired with established engineering models and domain expertise. In this setting, machine learning is used to improve rainfall processing, forecast skill, uncertainty characterization, and computational efficiency, while hydraulic models preserve physical realism and operational credibility. The result is a practical workflow that helps teams move from raw data to actionable guidance on the timescales required for urban flood management. The same workflow also supports post-event learning. By comparing forecasts, observations, and modeled system performance, practitioners can identify where predictions were accurate, where system behavior differed from expectations, and how model parameters, operating strategies, or resilience priorities should be updated. This creates a repeatable feedback loop in which each storm improves the next forecast, the next operational response, and the next planning decision.