Positioning rod floats for river discharge measurements based on video images taken from different viewing angles

The float method is a standard protocol for estimating the flow rate of an open channel. Floating rods are released in water flow, and the time duration to pass through two cross-sections is counted to estimate mean streamwise velocity. For flooding steep-sloped rivers, water depth becomes high, the flow is highly turbulent, and a significant volume of debris and gravel is transported, making flow rate measurements using sophisticated instruments challenging. The float method relies on very simple instruments and is useful for such challenging conditions. However, the method is labour-intensive and at least four people are required for field measurements. Non-contact methods, including image analysis methods and radar velocimeters, both measure water surface flow velocity and serve as alternatives to traditional float methods for measuring flood flow rates. The application of non-contact methods in regular hydrometry is on the rise. While investigating replacing measurement methods, flow rates obtained using the float method and alternative methods were compared, and some bias was confirmed. To explore the reason for bias, here, we focus on rod float trajectory and propose a method for easily and accurately estimating the location of floats. The location of a float was calculated using a video capturing flowing floats obtained from a riverbank and bridge. By accounting for the viewing angle of the video shoot and location uncertainty, which depends on the viewing angle, an algorithm for estimating float location based on the maximum likelihood method is proposed. The developed method was used to analyse the trajectories of floats dropped from different locations on a bridge. Convergence and divergence of float trajectories were confirmed, and flowing floats with similar pathways displayed similar local and mean velocities. The obtained accurate float trajectories will help improve the accuracy of flow rate estimation.