This paper presents a laboratory study of the dynamic properties of air flow over small wind-generated water waves. On the basis of measurements of mean velocity profiles, turbulent intensity profiles and energy spectra, the detailed structure of turbulent wind immediately above and between the crests of progressive water waves has been examined.
The velocity sensor (a hot-wire anemometer) was fastened to a self-adjusting positioner to measure instantaneous air velocities at a fixed distance from a moving water surface. The waves had a dominant frequency, 2·4 Hz, and a ratio of wave celerity and air friction velocity close to one. With the aid of a digital computer, the desired parameters of air flow were obtained by a statistical technique which was developed to sample and average simultaneous recordings of water surface displacements and instantaneous air velocities.
The results of the wind field measurements over representative waves indicate that, on the average, the air flow separates from the wavy water surface just behind crests, and reattaches somewhere on the windward face of the next wave. The measured turbulent quantities consistently show the characteristics of separated air flow. The separation phenomenon suggests that, without some modification, the Benjamin-Miles shearing flow mechanism is inapplicable to the growth of fully-developed small water waves, at least when the ratio of the phase speed to air friction velocity is of order unity. The observed flow configuration tends to support the separation mechanism of energy transfer originally outlined by Jeffreys and later explored further by Stewart and Deardorff.
Mean properties of the turbulent air flow referred to the mean water level were obtained by continuous sampling of the air flow over many waves with a sensing probe, either at a fixed distance from the mean water level (fixed probe measurement) or at a constant distance from the moving water surface (moving probe measurement). It was found that for continuously averaged measurements, the fixed probe yielded results which deviate less from the local mean than the moving probe results. This holds for the mean velocity distributions and especially for the turbulent quantities.