A wind-tunnel boundary-layer study of the effects of a surface roughness change: Rough to smooth

The changes imposed on mean velocities and turbulence statistics in the lower atmosphere by an abrupt change in surface roughness, from very rough to smooth, were modelled in a wind tunnel. The influence of a change in the effective surface level, which often accompanies such a variation in surface roughness, was also studied. A deep, turbulent flow was generated upstream of the change, which had a logarithmic mean velocity profile and constant shear-stress for approximately 200 mm above the floor, except for a region near the surface which was influenced by the three-dimensional nature of the random rough surface.When the surface roughness change coincided with a change in surface level, the downstream flow close to the surface was in the wake of the upstream roughness elements, and measured Reynolds shear-stress values were lower than those obtained when the downstream surface was raised. Otherwise, the influence of a change in surface level was small.In all cases, Reynolds shear-stress varied approximately linearly with height in the lower two-thirds of the internal layer and no constant stress region was apparent near the surface, even 2 m downstream of the roughness change. When the roughness change was not accompanied by a change in level, Reynolds shear-stress values extrapolated to the surface agreed well with surface shear-stress inferred from the law of the wall.Changes in mean squares of vertical and lateral velocity fluctuations and in integral time scales, as the flow passed downstream of the roughness change, were surprisingly small.     

A Wind-Tunnel Investigation of Wind-Turbine Wakes: Boundary-Layer Turbulence Effects

Wind-tunnel experiments were performed to study turbulence in the wake of a model wind turbine placed in a boundary layer developed over rough and smooth surfaces. Hot-wire anemometry was used to characterize the cross-sectional distribution of mean velocity, turbulence intensity and kinematic shear stress at different locations downwind of the turbine for both surface roughness cases. Special emphasis was placed on the spatial distribution of the velocity deficit and the turbulence intensity, which are important factors affecting turbine power generation and fatigue loads in wind energy parks. Non-axisymmetric behaviour of the wake is observed over both roughness types in response to the non-uniform incoming boundary-layer flow and the effect of the surface. Nonetheless, the velocity deficit with respect to the incoming velocity profile is nearly axisymmetric, except near the ground in the far wake where the wake interacts with the surface. It is found that the wind turbine induces a large enhancement of turbulence levels (positive added turbulence intensity) in the upper part of the wake. This is due to the effect of relatively large velocity fluctuations associated with helicoidal tip vortices near the wake edge, where the mean shear is strong. In the lower part of the wake, the mean shear and turbulence intensity are reduced with respect to the incoming flow. The non-axisymmetry of the turbulence intensity distribution of the wake is found to be stronger over the rough surface, where the incoming flow is less uniform at the turbine level. In the far wake the added turbulent intensity, its positive and negative contributions and its local maximum decay as a power law of downwind distance (with an exponent ranging from −0.3 to −0.5 for the rough surface, and with a wider variation for the smooth surface). Nevertheless, the effect of the turbine on the velocity defect and added turbulence intensity is not negligible even in the very far wake, at a distance of fifteen times the rotor diameter.     

Near Wall Flow over Urban-like Roughness

In this study, comprehensive measurements over a number of urban-type surfaces with the same area density of 25% have been performed in a wind tunnel. The experiments were conducted at a free stream velocity of 10 m s^-1 and the main instrumentation was 120 ° x-wire anemometry, but measurement accuracy was checked using laser Doppler anemometry.The results haveconfirmed the strong three-dimensionalityof the turbulent flow inthe roughness sublayer and the depths of the inertial sublayer (log-law region) and roughness sublayer for each surface have been determined. Spatial averaging has been used to remove the variability of the flow in the roughness sublayer due to individual obstacles and it is shown that the spatially averaged mean velocity in the inertial sublayer and roughness sublayer can,together, be described by a single log-law with a mean zero-plane displacement and roughness length for the surface, provided that the proper surface stress is known. The spatially averaged shear stresses in the inertial sublayer and roughness sublayer are compared with the surface stress deduced from form drag measurements on the roughness elements themselves.The dispersive stress arising from the spatial inhomogeneity in the mean flow profiles was deduced from the data and is shown to be negligible compared with the usual Reynolds stresses in the roughness sublayer. Comparisons have been made between a homogeneous (regular element array) surface and one consisting of random height elements of the same total volume. Although the upper limits of the inertial sublayer for both surfaces were almost identical at equivalent fetch, the roughness sublayer was much thicker for the random surface than for the uniform surface, the friction velocity and the roughness length were significantly larger and the `roughness efficiency' was greater. It is argued that the inertial sublayer may not exist at all in some of the more extreme rough urban areas. These results will provide fundamental information for modelling urban air quality and forecasting urban wind climates.     

The flow in a turbulent boundary layer upstream of a change in surface roughness

Modification of a turbulent flow upstream of a change in surface roughness has been studied by means of a stream function-vorticity model.A flow reduction is found upstream of a step change in surface roughness when a fluid flows from a smooth onto a rough surface. Above that layer and above the region of flow reduction downstream of a smooth-rough transition, a flow acceleration is observed. Similar flow modification can be seen at a rough-smooth transition with the exception that flow reduction and flow acceleration are reversed. Within a fetch of –500 < x/z ₀< + 500 (z ₀ is the maximum roughness length, the roughness transition is located at x/z ₀ = 0), flow reduction (flow acceleration) upstream of a roughness transition is one order of magnitude smaller than the flow reduction (flow acceleration) downstream of a smooth-rough (rough-smooth) transition. The flow acceleration (flow reduction) above that layer is two orders of magnitude.The internal boundary layer (IBL) for horizontal mean velocity extends to roughly 300z ₀ upstream of a roughness transition, whereas the IBL for turbulent shear stress as well as the distortion of flow equilibrium extend almost twice as far. For the friction velocity, an undershooting (overshooting) with respect to upstream equilibrium is predicted which precedes overshooting (undershooting) over new equilibrium just behind a roughness transition.The flow modification over a finite fetch of modified roughness is weaker than over a corresponding fetch downstream of a single step change in roughness and the flow stays closer to upstream equilibrium. Even in front of the first roughness change of a finite fetch of modified roughness, a distortion of flow equilibrium due to the second, downwind roughness change can be observed.     

Mean Flow and Turbulence Characteristics in an Urban Roughness Sublayer

In this study, a detailed model of an urban landscape has been re-constructed inthe wind tunnel and the flow structure inside and above the urban canopy has beeninvestigated. Vertical profiles of all three velocity components have been measuredwith a Laser-Doppler velocimeter, and an extensive analysis of the measured meanflow and turbulence profiles carried out. With respect to the flow structure inside thecanopy, two types of velocity profiles can be distinguished. Within street canyons,the mean wind velocities are almost zero or negative below roof level, while closeto intersections or open squares, significantly higher mean velocities are observed.In the latter case, the turbulent velocities inside the canopy also tend to be higherthan at street-canyon locations. For both types, turbulence kinetic energy and shearstress profiles show pronounced maxima in the flow region immediately above rooflevel.Based on the experimental data, a shear-stress parameterization is proposed, inwhich the velocity scale, u_s, and length scale, z_s, are based on the level and magnitude of the shear stress peak value. In order to account for a flow region inside the canopy with negligible momentum transport, a shear stress displacement height, d_s, is introduced. The proposed scaling and parameterization perform well for the measured profiles and shear-stress data published in the literature.The length scales derived from the shear-stress parameterization also allowdetermination of appropriate scales for the mean wind profile. The roughnesslength, z₀, and displacement height, d₀, can both be described as fractions of the distance, z_s - d_s, between the level of the shear-stress peak and the shear-stress displacement height. This result can be interpreted in such a way that the flow only feels the zone of depth z_s - d_s as the roughness layer. With respect to the lower part of the canopy (z < d_s) the flow behaves as a skimming flow. Correlations between the length scales z_s and d_s and morphometric parameters are discussed.The mean wind profiles above the urban structure follow a logarithmic windlaw. A combination of morphometric estimation methods for d₀ and z₀ with wind velocity measurements at a reference height, which allow calculation of the shear-stress velocity, u_*, appears to be the most reliable and easiest procedure to determine mean wind profile parameters. Inside the roughnesssublayer, a local scaling approach results in good agreement between measuredand predicted mean wind profiles.     

A wind-tunnel study of turbulent flow close to regularly arrayed rough surfaces

Recent observations of flux-gradient anomalies in atmospheric flow close to forests, and similar rough surfaces, prompted a wind-tunnel investigation in which cross-wire anemometry was used to study the vertical development and horizontal variability of adiabatic flow over five regularly arrayed rough surfaces, encompassing a 32-fold range of roughness concentration . The roughness elements were cylinders, 6 mm in both height and diameter.Below a layer in which the velocity profile is semi-logarithmic, two surface influences upon the mean velocity field can be distinguished: wake diffusion and horizontal inhomogeneity. The wake diffusion effect causes non-dimensional vertical velocity gradients to be smaller than in the semi-logarithmic region; at least for elements with aspect ratios l/h 1, it is governed by the transverse dimension l of the roughness elements, and is observed when z > h + 1.5l (where z is height above the underlying surface, and h is the height of the roughness elements). A simple diffusivity model successfully describes the horizontally averaged velocity profiles in the region of wake influence, despite conceptual disadvantages. The horizontal inhomogeneity of the flow is negligible when z > h + D (D being the inter-element spacing), and does not entirely mask the wake diffusion effect except over very sparsely roughened surfaces ( 0.02). A criterion for negligibility of both effects, and hence for applicability of conventional turbulent diffusivity theory for momentum, is z > h + 1.5D. These results are compared with atmospheric data, and indicate that wake diffusion may well cause some underestimation of the zero-plane displacement d over typical vegetated surfaces.     

Drag due to regular arrays of roughness elements of varying geometry

Comparisons are made of experimental studies on the drag, at high Reynolds number, due to regular arrays of roughness elements of various shapes immersed in a turbulent boundary layer. Using a variant of Millikan's dimensional analysis, the form of the velocity profile is deduced in terms of the dimensions and concentration of the roughness elements. A drag formula results which is shown to be in good agreement with data. Available measurements of the partition of drag between the elements and the intervening surface indicates that equipartition occurs at quite low concentrations. The interaction between elements is then small, so that the drag coefficient of a typical roughness element is nearly constant. A re-examination of some of O'Loughlin's velocity-profile data, obtained below the tops of the roughness elements, suggests the existence of a nearly constant-stress layer scaled to the shear stress of the intervening surface. Above the roughness elements, the mean-velocity profile undergoes a transition to the form appropriate to the total shear stress exerted by the roughened surface. A formula is given which describes the one-dimensional velocity profile over the entire range, excluding the viscous sublayer on the intervening surface. The viscous sublayer appears to correspond quite closely to that on a smooth plate. This work was initiated while the authors were with the Division of Plant Industry, CSIRO. At present on leave at the Natural Resource Ecology Laboratory, Colorado State University, Fort Collins, Colorado, U.S.A.     

A Flume Experiment on the Adjustment of the Mean and Turbulent Statistics to a Transition from Short to Tall Sparse Canopies

Water-flume experiments are conducted to study the structure of turbulent flow within and above a sparse model canopy consisting of two rigid canopies of different heights. This difference in height specifies a two-dimensional step change from a rough to a rougher surface, as opposed to a smooth-to-rough transition. Despite the fact that the flow is in transition from a rough to a rougher surface, the thickness of the internal boundary layer scales as x ^4/5, consistent with smooth-to-rough boundary layer adjustment studies, where x is the downstream distance from the step change. However, the analogy with smooth-to-rough transitions no longer holds when the flow inside the canopy and near the canopy top is considered. Results show that the step change in surface roughness significantly increases turbulence intensities and shear stress. In particular, there is an adjustment of the mean horizontal velocity and shear stress as the flow passes over the rougher canopy, so that their vertical profiles adjust to give maximum values at the top of this canopy. We also observe that the magnitude and shape of the inflection in the mean horizontal velocity profile is significantly affected by the transition. The horizontal and vertical turbulence spectra compare well with Kolmogorov’s theory, although a small deviation at high frequencies is observed in the horizontal spectrum within the canopy. Here, for relatively low leaf area index, shear is found to be a more effective mechanism for momentum transfer through the canopy structure than vortex shedding.     

Aerodynamic Properties of Rough Surfaces with High Aspect-Ratio Roughness Elements: Effect of Aspect Ratio and Arrangements

We examine the effect of varying roughness-element aspect ratio on the mean velocity distributions of turbulent flow over arrays of rectangular-prism-shaped elements. Large-eddy simulations (LES) in conjunction with a sharp-interface immersed boundary method are used to simulate spatially-growing turbulent boundary layers over these rough surfaces. Arrays of aligned and staggered rectangular roughness elements with aspect ratio >1 are considered. First the temporally- and spatially-averaged velocity profiles are used to illustrate the aspect-ratio effects. For aligned prisms, the roughness length (\(z_\mathrm{o}\)) and the friction velocity (\(u_*\)) increase initially with an increase in the roughness-element aspect ratio, until the values reach a plateau at a particular aspect ratio. The exact value of this aspect ratio depends on the coverage density. Further increase in the aspect ratio changes neither \(z_\mathrm{o}\), \(u_*\) nor the bulk flow above the roughness elements. For the staggered cases, \(z_\mathrm{o}\) and \(u_*\) continue to increase for the surface coverage density and the aspect ratios investigated. To model the flow response to variations in roughness aspect ratio, we turn to a previously developed phenomenological volumetric sheltering model (Yang et al., in J Fluid Mech 789:127–165, 2016), which was intended for low to moderate aspect-ratio roughness elements. Here, we extend this model to account for high aspect-ratio roughness elements. We find that for aligned cases, the model predicts strong mutual sheltering among the roughness elements, while the effect is much weaker for staggered cases. The model-predicted \(z_\mathrm{o}\) and \(u_*\) agree well with the LES results. Results show that the model, which takes explicit account of the mutual sheltering effects, provides a rapid and reliable prediction method of roughness effects in turbulent boundary-layer flows over arrays of rectangular-prism roughness elements.     

Observations of the barotropic Ekman layer over an urban terrain

Data from low-level soundings over Cambridge, U.S.A. were selected on the basis of an Ekman-like variation of the wind vector with altitude combined with evidence of a barotropic atmosphere. The method of geostrophic departure was used to determine the shear-stress distribution. The analysis yields the dimensionless properties of the barotropic Ekman layer under neutral and stable stratification. Some important results include: the geostrophic drag coefficient displays no dependence on the degree of static stability; the dimensionless height of the boundary layer decreases with increasing stability in agreement with the prediction of Zilitinkevich; the properties of the urban surface layer, where the roughness elements are multistory buildings, show no dependence on atmospheric stability under the moderate wind conditions which display the Ekman-like wind profile; and the directions of the horizontal shear stress and the vertical derivative of the velocity vector usually tend to be parallel only near the surface layer. Values of the two constants of the Rossby number similarity theory are found for the neutral barotropic Ekman layer at a surface Rossby number equal to 2 × 10⁵. The implications of the work with respect to wind-tunnel simulation of the flow over models of urban areas are discussed.     

An extended Miles' theory for wave generation by wind

Miles' inviscid theory of surface wave generation by wind is (a) modified by replacing the logarithmic shear velocity profile with one which applies right down to the wave surface and which exhibits an explicit dependence on the roughness of the surface, and (b) extended to include the effects of the interaction of wave with air flow turbulence by considering the wave-modified mean flow as the mean of the actual turbulent air flow over water waves and using this in a mixing-length model.The surface pressure is shown to depend significantly on the flow conditions being aerodynamically smooth or rough. Its component in phase with the surface elevation is practically unaffected by the wave-turbulence interaction. However, such interaction tends to increase the rate of energy input ß from wind to waves travelling in the same direction, e.g., the increase is 2gk ² for aerodynamically rough flow, where gk is the Von Karman constant. It also provides damping of waves in an adverse wind which can be about 10% of the growth rate in a favourable wind.     

Flow over high roughness elements

The mean velocity and longitudinal turbulence-intensity distributions inside the zone of and above high roughness elements were investigated experimentally. This was accomplished by using a model forest canopy. The results indicate that the flow may be divided into transition and fully-developed flow regions, followed by a short adjustment distance near the downstream terminus of the rough boundary. The transition region has a strong effect on the flow characteristics within and above the layer of roughness elements. Generally, a similar qualitative variation for both velocity and turbulence was found inside and above the roughness zone, whose influence extends to more than three times the roughness height.Investigation of the modified universal logarithmic law for describing the velocity variation above the roughness zone revealed that both of the so-called similarity parameters, i.e., friction velocity and roughness length, are not local constants. On the contrary, for a given flow and local conditions they vary drastically with height. It is suspected that this is due to the fact that the classical assumption of constant shear stress throughout the boundary layer or significant portions of it is not satisfied in the case of roughness elements many times greater in height than the thickness of the viscous wall zone.     

Mean Flow and Turbulence Statistics Over Groups of Urban-like Cubical Obstacles

Direct numerical simulations of turbulent flow over regular arrays of urban-like, cubical obstacles are reported. Results are analysed in terms of a formal spatial averaging procedure to enable interpretation of the flow within the arrays as a canopy flow, and of the flow above as a rough wall boundary layer. Spatial averages of the mean velocity, turbulent stresses and pressure drag are computed. The statistics compare very well with data from wind-tunnel experiments. Within the arrays the time-averaged flow structure gives rise to significant ‘dispersive stress’ whereas above the Reynolds stress dominates. The mean flow structure and turbulence statistics depend significantly on the layout of the cubes. Unsteady effects are important, especially in the lower canopy layer where turbulent fluctuations dominate over the mean flow.     

Large-Eddy Simulation Of Neutral Turbulent Flow Over Rough Sinusoidal Ridges

The requirements for a credible large-eddy simulation of neutral, turbulent flow over hills with an aerodynamically rough surface are discussed, in order to select a suitable case for simulation. As well as providing adequate resolution within the dynamically important inner region, obtaining a realistic upstream or undisturbed mean velocity profile is also of critical importance. A distributed drag canopy formulation has been introduced to the model to allow it to obtain such a profile while resolving very close to the rough surface. Simulations have then been performed of flow over ridges of varying heights. The results from the steepest case, which is just on the verge of separation, are compared with wind-tunnel observations. It is shown that the large-eddy simulation results are in much better agreement with the experimental data than are the results from a simple first-order mixing-length closure model. An encouraging lack of sensitivity of the simulation results to numerical resolution is also demonstrated.     

Experimental Study on Effects of Ground Roughness on Flow Characteristics of Tornado-Like Vortices

The three-dimensional wind velocity and dynamic pressure for stationary tornado-like vortices that developed over ground of different roughness categories were investigated to clarify the effects of ground roughness. Measurements were performed for various roughness categories and two swirl ratios. Variations of the vertical and horizontal distributions of velocity and pressure with roughness are presented, with the results showing that the tangential, radial, and axial velocity components increase inside the vortex core near the ground under rough surface conditions. Meanwhile, clearly decreased tangential components are found outside the core radius at low elevations. The high axial velocity inside the vortex core over rough ground surface indicates that roughness produces an effect similar to a reduced swirl ratio. In addition, the pressure drop accompanying a tornado is more significant at elevations closer to the ground under rough compared with smooth surface conditions. We show that the variations of the flow characteristics with roughness are dependent on the vortex-generating mechanism, indicating the need for appropriate modelling of tornado-like vortices.     

Large-eddy Simulation of Turbulent Flow Across a Forest Edge. Part I: Flow Statistics

The statistics of turbulent flow across a forest edge have been examined using large-eddy simulation, and results compared with field and wind-tunnel observations. The moorland-to-forest transition is characterized by flow deceleration in the streamwise direction, upward distortion of the mean flow, formation of a high pressure zone immediately in front of the edge, suppression of the standard deviations and covariance of velocity components, and enhancement of velocity skewnesses. For the selected forest density, it is observed that the maximum distortion angle is about 8 degrees from the horizontal. Instead of approaching a downwind equilibrium state in a monotonic manner, turbulence (standard deviations and covariances of velocity components) and mean streamwise velocity undershoot in the transition zone behind the edge. Evolution of flow statistics clearly reveals the growth of an internal boundary layer, and the establishment of an equilibrium layer downwind of the edge. It is evident that lower-order moments generally adjust more quickly over the new rough surface than do higher-order moments. We also show that the streamwise velocity standard deviation at canopy height starts its recovery over the rough surface sooner than does the vertical velocity standard deviation, but completes full adjustment later than the latter. Despite the limited domain size upstream of the edge, large-eddy simulation has successfully reproduced turbulent statistics in good agreement with field and wind-tunnel measurements.     

Measurements of turbulence structure in the boundary layer on a rough surface

Mean products of velocity fluctuations up to fourth order have been measured in a wind tunnel at the trailing edge of a flat plate, one side of which was covered with floor-sanding paper to produce a fully rough surface. This set-up permits easy comparison of structural parameters in smooth-wall and rough-wall boundary layers. The Reynolds-stress profiles and second-order parameters are closely the same on the rough and smooth surfaces; in particular the decrease in Reynolds shear stress near the rough surface, encountered by several other laboratory workers, was not found in the present results. The triple products are spectacularly altered for a distance of up to 10 roughness heights from the rough surface, and imply a large net rate of transport of turbulent energy and shear stress towards the surface. Comparison with other published data shows that the behaviour of this modified region depends on roughness geometry as well as on the roughness height itself; for example, the mean cube of the normal-component fluctuation remains positive (energy transport away from the surface) over sand or gravel roughness but goes negative, like the other energy-transport terms, over crop canopies.     

Scaling of the Vertical Spreading of a Plume of a Passive Tracer in a Rough-Wall Neutral Boundary Layer

The influence of surface roughness on the dispersion of a passive scalar in a rough wall turbulent boundary layer has been studied using wind-tunnel experiments. The surface roughness was varied using different sizes of roughness elements, and different spacings between the elements. Vertical profiles of average concentration were measured at different distances downwind of the source, and the vertical spread of the plume was computed by fitting a double Gaussian profile to the data. An estimate of the integral length scale is derived from the turbulence characteristics of the boundary layer and is then used to scale the measured values of plume spread. This scaling reduces the variability in the data, confirming the validity of the model for the Lagrangian integral time scale, but does not remove it entirely. The scaled plume spreading shows significant differences from predictions of theoretical models both in the near and in the far field. In the region immediately downwind of the source this is due to the influence of the wake of the injector for which we have developed a simple model. In the far field we explain that the differences are mainly due to the absence of large-scale motions. Finally, further downwind of the source the scaled values of plume spread fall into two distinct groups. It is suggested that the difference between the two groups may be related to the lack of dynamical similarity between the boundary-layer flows for varying surface roughness or to biased estimates of the plume spread.     

Neutral surface-layer flow over isolated roughness strips

Numerical simulations are used to study neutral surface-layer flow which passes over isolated or widely separated strips of modified roughness embedded in an otherwise homogeneous surface.Simple power laws are given for the maximum height and horizontal extent of turbulent momentum and horizontal mean velocity wakes with an assessment of the range of validity of this proposal. Furthermore, it is indicated, how vertical profiles of energy budget and horizontal mean velocity are distorted by a roughness strip.