Multi-objective optimization of sensor placement in urban drainage systems using flood observations derived from social media

Urban flooding poses increasing risks to cities under climate change and rapid urbanization, making effective monitoring of Urban Drainage Systems (UDS) essential for flood early warning. Existing sensor placement methods largely depend on simulated hydraulic data and often neglect the trade-off between detection reliability and information redundancy. To address these limitations, this study constructs an empirical dataset by integrating social media texts, historical waterlogging records, and municipal drainage network data. A multi-objective optimization framework is then proposed, jointly considering sensing cost, coverage, detection reliability, and redundancy. The case study in Wuhan demonstrates that, under the same sensor number and coverage range, the proposed framework achieves nearly six times higher detection reliability than the re-clustering algorithm, providing a practical solution for UDS monitoring.