Enhancing inundation monitoring of road networks using crowdsourced flood reports

Abstract As the risk of flooding continues to impose on transportation infrastructure systems, enhancing the ability to monitor flood inundations in road networks becomes more significant. This research explores the potential crowdsourced data, specifically 3-1-1 reports, has to complement information from physical flood sensors. Three flooding events in two watersheds in Harris County, Texas were studied. To assess 3-1-1 report's capabilities, a time series analysis between reports and water elevation of physical sensor data was performed to capture channel overflow. Next, a graph-based observability analysis identified multiple combinations of minimum additional sensor locations needed for complete network monitoring in the study area. To find the optimal combination, a principal component analysis assigns a criticality score based on exposure, road importance, neighborhood vulnerability, and frequency of 3-1-1 reports near a node. The results indicate that 3-1-1 reports effectively improve flood monitoring by reducing the need for physical sensors by 32% in areas that lack flood sensors. This approach can help city managers improve flood monitoring by leveraging socially sensed data to supplement physical sensors, especially in blind spots where no flood gauge exists.