We detail three techniques which can be used to separate wave-associated and turbulent motions in two-dimensional, time-unresolved velocity fields measured beneath moderately-broadbanded wind-generated waves. Two methods to extract wave-associated quantities, employing a conditional average of measured data contingent on the amplitude and phase of the surface elevation or the spectral coherence between the measured data and the surface elevation, are based solely on the obtained data; a third method employs linear wave theory to find the potential flow which conforms to the measured surface elevation. We first compare the results of the two data-driven decomposition techniques to a ground truth imposed in synthetic data with realistic surface elevation profiles, finding slightly better performance with the spectral coherence method given its ability to infer the role of a range of wavenumbers. Next, we show that the two methods perform similarly when constructing the orbital velocities beneath wind-driven waves measured experimentally with particle image velocimetry in a laboratory. We assess the error which is induced in turbulence statistics by the tendency of the decomposition methods to misclassify orbital motions as turbulence and discuss the applicability of each to common laboratory settings. • Flow under wind-driven waves is decomposed into turbulent and orbital parts. • Phase-locked averaging and spectral coherence perform similarly well. • Employing linear wave theory introduces artifacts into statistics.
Ruth et al. (Wed,) studied this question.