A new method for detecting spikes in acoustic Doppler velocimeter data sequences is suggested. The method combines three concepts: ͑1͒ that differentiation enhances the high frequency portion of a signal, ͑2͒ that the expected maximum of a random series is given by the Universal threshold, and ͑3͒ that good data cluster in a dense cloud in phase space or Poincaré maps. These concepts are used to construct an ellipsoid in three-dimensional phase space, then points lying outside the ellipsoid are designated as spikes. The new method is shown to have superior performance to various other methods and it has the added advantage that it requires no parameters. Several methods for replacing sequences of spurious data are presented. A polynomial fitted to good data on either side of the spike event, then interpolated across the event, is preferred by the authors.

In this paper it is suggested that the double-averaged (in temporal and in spatial domains) momentum equations should be used as a natural basis for the hydraulics of rough-bed open-channel flows, especially with small relative submergence. The relationships for the vertical distribution of the total stress for the simplest case of 2D, steady, uniform, spatially averaged flow over a rough bed with flat free surface are derived. These relationships explicitly include the form-induced stresses and form drag as components of the total stress. Using this approach, we define three types of rough-bed flows: (1) Flow with high relative submergence; (2) flow with small relative submergence; and (3) flow over a partially inundated rough bed. The relationships for the double-averaged velocity distribution and hydraulic resistance for all three flow types are derived and compared with measurements where possible. The double-averaged turbulent and form-induced intensities and stresses for the case of regular spherical-segment-type roughness show the dominant role of the double-averaged turbulence stresses and form drag in momentum transfer in the near-bed region.

Abstract. The random field approach for gravel-bed roughness characterization, which is based on the presentation of bed elevations as a three-dimensional random field, is justified as an alternative to the characteristic particle size approach. We first show that the bed elevation distribution is close to Gaussian and then investigate gravel-bed roughness using the second-order structure function. The latter reveals two distinct regions: a scaling region at small spatial lags and a saturation region at large scales. The scaling exponent H (a form of Hurst exponent) appears to be isotropic and universal for both manually created "unworked" gravel beds (H = 0.5) and natural water-worked gravel beds (H -0.79). However, the gravel-bed roughness, in general, is not isotropic and should be characterized by three independent characteristic scales. A simple model of gravel-bed roughness based on the structure function parameterization is developed and compared with the characteristic particle size approach.