Real-Time Culvert Blockage Detection Using a Low-Cost IoT System: Field Validation and Decision Framework

Abstract

Increasingly frequent extreme rainfall events are stressing roadway drainage networks worldwide, as culvert blockages caused by debris, sediment, and vegetation lead to flooding and infrastructure damage. Despite their importance, most culverts are still inspected manually at periodic intervals, without the capability for real-time blockage detection. This limitation reduces the ability of transport agencies to identify rapidly developing blockage-related hydraulic risks and plan proactive maintenance. This study presents and field-validates a low-cost Internet of things (IoT) system for continuous and real-time culvert blockage detection using ultrasonic water-level monitoring. The system was field-tested at an operational culvert in Canterbury, New Zealand, under four discharges ( 0.10 − 0.25    m 3   s − 1 ) and five blockage ratios (0%–100%), extending previously established laboratory-derived normalized discharge-stage relationships. The laboratory-derived five-level blockage classifier achieved 65% overall accuracy across the field dataset, while the calibrated three-tier framework categorizing blockage levels as low, medium, and high achieved 82% accuracy on the withheld validation subset. These findings demonstrate the feasibility of real-time blockage-state detection using a low-cost ultrasonic sensing system under field conditions. The proposed three-tier decision framework translates normalized inlet water-level and discharge information into actionable maintenance priorities: monitor (low), inspect soon (medium), and respond (high). The field-tested IoT-based framework provides a low-cost basis for integrating real-time blockage monitoring into digital asset-management workflows and demonstrates potential for broader deployment following further multisite and long-term validation.