ABSTRACT
Frequent passes of high-speed boats in navigable rivers can have significant contribution to bank erosion. The velocities associated with the boat-generated waves can be much larger then mean flow velocities. Wave impact along the riverbanks can quickly initiate the erosion and retreat process by undercutting and subsequent failures. In this study, the boat-generated waves and boat traffic were measured at two reaches along the Connecticut River and Tennessee River. Water surface displacements were measured by using self-logging wave staffs and boat passes were monitored by time-lapse cameras. Wind speed and direction were also measured for some cases. The measured wave time series for each logger was converted to the frequency domain, and the chirp-like transient wave patterns left by the boat passes on the spectrogram were identified using sound filtering tools, and object detection methods with deep convolutional neural networks. In this paper, the wave data analysis procedures, and methods for identification of boat passes are presented. The results show that, high-speed boat traffic can generate waves that are much larger than the wind generated waves. The methods described here offer a fast and economical method with minimum intrusion for boat monitoring in navigable rivers.
