This ongoing research direction treats flood sensing as an identifiability and decision-support problem, not just a monitoring problem. The goal is to infer hidden hydraulic states and fields under sparse sensing while also designing sensor networks that reduce uncertainty where it matters most for operational decisions.

Key directions

• Developing physics-consistent Bayesian inversion methods that produce posterior ensembles rather than fragile point estimates
• Designing value-of-information objectives for sensor placement under budget limits, sensor outages, and heterogeneous observation modalities
• Coupling inference and sensing in a closed-loop framework so the observing system is optimized for decision-relevant uncertainty reduction

Why it matters

Flood emergencies are often managed with sparse and brittle sensing networks. This work aims to make flood inference more trustworthy and flood monitoring more useful by aligning both with the actual decisions operators need to make in the field.