Author(s): Brian Crookston; Caitlin Arnold

Linked Author(s): Brian Crookston

Keywords: Hydraulic structures; Reverse roller; Deep learning; Aerial imagery; Low-head dams inventory

Abstract: Currently in the USA most low head dams are not inventoried, unregulated and without safety signage. Some of these structures present a public safety hazard, with an average of 5-6 drownings per month occurring from recreationalists being entrapped in the reverse roller that can form at the toe of the structure. In an effort to improve public safety, a national joint task force was recently formed between the American Society of Civil Engineers, The United States Society on Dams and Levees, and the Association of State Dam Safety Officials. Though crowdsourcing and partnerships with private, civic, university, and professional organizations a volunteer effort is underway to inventory all low head dams throughout the entire USA. Herein is presented an overview of the danger, this task committee’s efforts, and an identification technique using aerial imagery, deep learning, and ArcGIS Pro so that computers can scan river corridors and identify these drowning machines, thus populating the national inventory. Results of this study show that leveraging artificial intelligence technology is a viable and accurate approach to low head dam identification, but algorithm training is dependent upon image resolution, vegetation, and topography.

DOI: https://doi.org/10.3850/IAHR-39WC252171192022957

Year: 2022