Key points are not available for this paper at this time.
ABSTRACT Flow velocity measurement is fundamental to hydrological and hydraulic studies, providing essential data for streamflow estimation and river dynamics analysis. Traditional in situ methods like propeller gauges and acoustic Doppler current profilers are accurate but intrusive and labour‐intensive, while non‐intrusive image processing methods like large‐scale particle image velocimetry (LSPIV) offer a safer, more efficient alternative for measuring flow in challenging environments. Through a combination of controlled flume experiments and field deployments, the method's accuracy, sensitivity, and operational limitations were assessed in this study. Field experiments demonstrated the influence of environmental factors such as wind, reflections, and tracer distribution on velocity estimation. Laboratory tests provided controlled conditions to evaluate software performance across different flow regimes. Across both controlled flume experiments and multiple field deployments, LSPIV performance was found to be most consistent under medium flow conditions within the tested range. By evaluating identical workflows across scales and environments, the analysis demonstrates how tracer type and distribution, key software parameters (particularly interrogation and search area settings), and post‐processing and filtering strategies jointly influence velocity estimation accuracy. Within this multi‐scale framework, wood‐chip tracers provided more robust surface coverage and visibility than biodegradable cornstarch under both laboratory and field conditions. While the method performed well in both laboratory and natural settings, further research is needed to refine image processing under challenging field conditions and to develop capabilities for fully tracer‐free velocity estimation. Overall, the LSPIV method represents a robust and adaptable solution for non‐contact flow monitoring, with strong potential for broader application in hydrological research and river management.
Matin et al. (Fri,) studied this question.