The random sweeping decorrelation hypothesis was analysed theoretically and experimentally in terms of the higher-order velocity structure functions $D_{u_i}^{(m)}(r) = \left< [u_i^m(x + r) - u_i^m(x)]^2\right>$. Measurements in two high Reynolds number laboratory shear flows were used: in the return channel (Rλ ≈ 3.2 × 103) and in the mixing layer (Rλ ≈ 2.0 × 103) of a large wind tunnel. Two velocity components (in the direction of the mean flow, u1, and in the direction of the mean shear, u2) were processed for m = 1−4. The effect of using Taylor's hypothesis was estimated by a specially developed method, and found to be insignificant. It was found that all the higher-order structure functions scale, in the inertial subrange, as r2/3. Such a scaling has been argued as supporting evidence for the sweeping hypothesis. However, our experiments also established a strong correlation between energy- and inertial-range excitation. This finding leads to the conclusion that the sweeping decorrelation hypothesis cannot be exactly valid.The hypothesis of statistical independence of large- and small-scale excitation was directly checked with conditionally averaged moments of the velocity difference $\left< [u_i(x + r) - u_i(x)]^l\right>_{u_i^*}, l = 2-4$, at a fixed value of the large-scale parameter u*i. Clear dependence of the conditionally averaged moments on the level of averaging was found. In spite of a strong correlation between the energy-containing and the inertial-scale excitation, universality of the intrinsic structure of the inertial subrange was shown.

Characteristics of turbulence in the inertial range are experimentally studied in the atmospheric surface layer over the range of the Taylor microscale based Reynolds number Rλ≊(2.8–12.7)×103 and in a large wind tunnel (in a mixing layer at Rλ≊2.0×103 and a return channel at Rλ≊3.2×103). The intermittency exponent μ, estimated from the correlation function of energy dissipation Rεε(r)=〈ε(x)ε(x+r)〉∝r−μ, is found to be independent of Reynolds number and approximately equal to 0.20. No ‘‘measurable’’ deviation from the −5/3 exponent in the ‘‘five-thirds’’ law is found. On the other hand, the Kolmogorov constant C in this law is found to be weakly dependent on Rλ. This dependence is well described by the power law C∝R−μ/2λ≊R−0.10λ at μ≊0.20.

Temperature time traces are obtained in turbulent thermal convection at high Rayleigh numbers. Measurements are made in the midplane of the apparatus, near the sidewall but outside the boundary layer. A telegraph approximation for temperature traces is generated by setting the fluctuation amplitude to 1 or 0 depending on whether or not it exceeds the mean value. Unlike the standard diagnostics of intermittency, the telegraph approximation allows one to distinguish the tendency of events to cluster (clusterization) from their large-scale variability in amplitude. A qualitative conclusion is that amplitude intermittency might mitigate clusterization effects.

An experimental investigation of the fine-scale structure of turbulence was carried out. Five different shear flows were studied: three in a wind tunnel with an open working section and an elliptical nozzle and two in a wind tunnel of closed working section and square cross-section. The experiments tested two approaches to the theory of fine-scale turbulence structure: one based on the Navier-Stokes equations and the other on some similarity hypotheses. The variability of all fine-scale constants (including exponents in inertial-subrange power laws and the Kolmogorov constant) is revealed. A correlation between all fine-scale constants and the external intermittency coefficient is established.

Kolmogorov’s refined similarity hypothesis (RSH) was challenged by Hosokawa and Yamamoto [Phys. Fluids A 4, 457 (1992)] on the basis of a direct numerical simulation data set at Rλ&lt;100. The authors considered that their results were ‘‘a clear case of violation of the RSH.’’ However, the hypothesis was proposed for very high Reynolds number flows and is not expected to apply at the low Reynolds numbers typical of current simulations. The Hosokawa and Yamamoto tests have been repeated with a high Reynolds number (Rλ≊3.2×103) experimental data set. The results are in good agreement with the Kolmogorov RSH.