Bridging Flood Sensing and Traffic State Estimation for Roadside Flood Sensor Placement: A Bayesian Network-based Approach

Abstract

Urban flooding's impact on road networks and traffic is becoming a more serious problem in the context of climate change. Estimating traffic states under flood impacts requires information about road inundation. Roadside flood sensors are becoming an important source of this information. This study proposes a Bayesian Network-based approach to strategically place roadside flood sensors to enhance the information input for traffic state estimation. By integrating historical traffic speed data with flood-induced road closures, the method quantifies the uncertainty reduction in traffic estimation through entropy analysis and applies a greedy algorithm to optimize sensor placement. In the results, flood information from optimally placed sensors reduced the uncertainty of traffic states in the Bayesian Network, thereby helping traffic state estimation and prediction achieve more certain estimation results. Beyond proposing a sensor placement method, this study also enhances the information value of flood sensing and is expected to promote larger-scale application of roadside flood sensors.