Role

First Author

Team

Yeji Kim, Jeil Oh, Matthew Bartos

Platforms

Python, PipeDream, Extended Kalman Filter

Date

Jun 2022 – Dec 2023

Published: Kim, Y., Oh, J., & Bartos, M. (2025). Sustainable Cities and Society, 105982. DOI →

Live Dashboard: Waller Creek Stormwater Digital Twin →

Overview

Urban drainage systems face growing flood hazards from climate change, while existing monitoring is complicated by unreliable sensor data and imperfect hydrologic models. This work introduces a stormwater digital twin that fuses real-time sensor data with a hydraulic-hydrologic model to estimate water depths and discharges under sensor and model uncertainty.

Approach

• Developed a novel Extended Kalman Filter (EKF) state estimation scheme that simultaneously assimilates sensor data and detects faulty measurements
• Long-term real-world deployment in Austin’s Waller Creek watershed
• Open-source Python software implementation enabling real-time monitoring and active control

Key Results

• ROC AUC > 0.99 for sensor fault detection — substantially reducing false flood alarms
• Improved water-depth estimation at ungauged locations
• More accurate near-term flood forecasts compared to a base hydraulic model
• Provides a complete framework for rapid flood response, predictive maintenance, and active control of sewer systems

Reference

Kim, Y., Oh, J., & Bartos, M. (2025). Stormwater digital twin with online quality control detects urban flood hazards under uncertainty. Sustainable Cities and Society, 105982. https://doi.org/10.1016/j.scs.2024.105982