Vous êtes sur la page 1sur 13

Geomorphology 139–140 (2012) 258–270

Contents lists available at SciVerse ScienceDirect

Geomorphology
journal homepage: www.elsevier.com/locate/geomorph

Wind tunnel simulation of the three-dimensional airflow patterns behind cuboid


obstacles at different angles of wind incidence, and their significance for the
formation of sand shadows
Wanyin Luo ⁎, Zhibao Dong, Guangqiang Qian, Junfeng Lu
Key Laboratory of Desert and Desertification, Cold and Arid Regions Environmental and Engineering Research Institute, Chinese Academy of Sciences, China

a r t i c l e i n f o a b s t r a c t

Article history: Sand shadows commonly form downwind of obstacles in arid and semi-arid regions. Understanding their dy-
Received 22 March 2011 namics requires insights into the airflow patterns behind the obstacles under different wind regimes. Here,
Received in revised form 24 October 2011 we studied models of cuboid obstacles to characterize the three-dimensional responses of airflow behind ob-
Accepted 27 October 2011
stacles with different shape ratios to variations in the incident flow in a wind-tunnel simulation. Wind veloc-
Available online 6 November 2011
ity was measured using particle image velocimetry (PIV). The flow patterns behind cuboid obstacles were
Keywords:
complicated by changes in the incidence angle of the approaching flow and in the obstacle's shape ratio.
Dune dynamics The flows separated both horizontally and vertically, creating reverse-flow cells downwind of the obstacles.
Obstacles to wind flow The horizontal cells were characterized by two asymmetrical, opposing reverse vortices at flow incidence an-
Secondary airflow patterns gles greater than 60 to 65°, whereas the vertical cells formed a single vortex at all angles. The cells shifted po-
Shadow dunes sition in response to changes in the incidence angle and shape ratio, reflecting the interaction between outer
Wind tunnel simulation and reverse flows. The parameters of the horizontal and vertical reverse cells depended on the shape ratios
and incidence angles but not on the wind velocity. The sizes of the horizontal and vertical reverse cells
were correlated, reflecting their interactions. The low-velocity “shadow” behind the obstacle caused sand de-
position and determined the sand shadow's evolution. Field experiments will be required to relate these air-
flow patterns to sand shadow development.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction al., 1993; Livingstone and Warren, 1996). Understanding the dynamics
of these dunes thus requires insights into the airflow fields behind
Sand shadows (also called shadow dunes) form in the lee of obstacles, obstacles.
and are widely distributed in arid and semi-arid regions and in some As the name implies, sand shadows represent sediment accumula-
coastal regions around the world. The most frequently reported obsta- tions that form in the shelter (wind shadow) of, and immediately be-
cles that create such sediment deposits are topographic obstacles such hind, an obstacle. Bagnold (1941) first used this name to describe the
as hills and rocks, human structures such as cairns or buildings, and sed- accumulation of eolian sand in the lee of an obstacle that reduces the
iments trapped by vegetation (e.g., Karcz, 1968; Carter, 1978; Hesp, local wind velocity. Similar structures, associated with both wind and
1981; Clemmensen, 1986; Wilson, 1988; Gunatilaka and Mwango, water, have subsequently been recognized in a variety of depositional
1989). Consequently, sand shadows have received much attention environments (e.g., Wilson, 1988). Allen (1982) suggested the less
from eolian geomorphologists (Bagnold, 1941; Clemmensen, 1986; specific term current shadow for such features in both eolian and
Hesp and McLachlan, 2000; Hesp et al., 2005; Dong et al., 2008). During aqueous environments. However, related terms such as lee dune,
the formation and subsequent evolution of sand shadows, there is an it- shadow dune and coppice dune are more widely used in the eolian lit-
erative interaction between the obstacle's geometry, the characteristics erature to describe such deposits (Hesp, 1981; Pye and Tsoar, 1990;
of the secondary airflow fields created by that geometry combined with Cooke et al., 1993).
the direction of the incident wind, the morphology of the sand shadow, Although sand shadows have attracted research attention because
and the interaction between neighboring dunes or obstacles (Cooke et of their occurrence outside Martian craters, they are the least well-
understood of anchored dunes (Greeley and Iversen, 1985). The litera-
ture suggests that these deposits cannot survive large changes in wind
direction, such as the ones that occur where the wind regime is highly
⁎ Corresponding author at: Cold and Arid Regions Environmental and Engineering
Research Institute, Chinese Academy of Sciences, No. 320, West Donggang Road, Lanzhou,
variable in direction, and can only grow to meso- or mega-size under
Gansu Province 730000, China. Tel.: +86 931 4967498; fax: +86 931 8277169. constant wind regimes (Cooke et al., 1993; Livingstone and Warren,
E-mail address: wyluo@lzb.ac.cn (W. Luo). 1996). Obstacles generate different sand shadows depending on the

0169-555X/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.geomorph.2011.10.027
W. Luo et al. / Geomorphology 139–140 (2012) 258–270 259

Fig. 3. Illustration of the obstacle model and coordinate system used in the wind tunnel
tests. Model parameters: H = height, W = width, B = thickness, U = free-stream wind
velocity, z = height above the bed, and x = distance downwind of the obstacle model.

1990; Cooke et al., 1993). Sand shadows can form after a storm or
can be created by a weaker wind that exceeds the threshold frictional
velocity required to entrain sand particles (Bagnold, 1941;
Clemmensen, 1986; Pye and Tsoar, 1990; Cooke et al., 1993;
Livingstone and Warren, 1996).
Fig. 1. Successive stages in the growth of sand shadows, based on the illustrations of The morphology and evolution of a sand shadow are sensitive to
Bagnold (1941). Airflow around the obstacle creates two wings (A), which gradually the angle of the incident wind and the shape of the obstacle, which
coalesce (B) as the secondary air flow and reverse flows within the wind shadow strongly affects the secondary flow patterns in its lee (Bagnold,
carry the grains towards the middle of the obstacle, producing an early sand shadow
1941; Gunatilaka and Mwango, 1989; Dong et al., 2008). Wind tunnel
(C) under an ideal wind which is relatively constant in direction. The sand shadow is
modified to something that resembles the structure shown in (D) when the wind is
studies have shown that the shear stress exerted on the sand bed in
supposed to swing between two limiting directions indicated by the two arrows. the wake of an obstacle is a function of the obstacle's dimensions
and the friction wind velocity (Greeley and Iversen, 1985). Iversen
et al. (1990) found that the intensity of turbulence and erosion was
sand supply, wind regime, and the geometry of the obstacle, and can related to the obstacle's aspect ratio (the ratio of its height to its
generate double or single sand shadows. The eolian literature on sand width perpendicular to the flow). Dong et al. (2008) wind tunnel sim-
shadows mainly concerns their internal sedimentary structures and ulation revealed that the flow patterns around shrubs were compli-
morphological descriptions (Clemmensen, 1986; Gunatilaka and cated by the presence of bleed flows (which pass through the
Mwango, 1989), and little attention has been paid to the aerodynamic obstacle) and displaced flows (which flow around the obstacle)
processes that govern their development. Although there is an obvious when the shrub density equals or exceeds a critical density (which
association with turbulence patterns in the lee of an obstacle, there is ranged between 0.05 and 0.08). Thus, sand shadows develop behind
still much to learn about sand shadows. dense shrubs when the dominant reversed flow leads to sediment ac-
The wind pattern around the solid cuboid obstacle has been stud- cumulation on the lee side. If the obstacle's shape varies in three di-
ied intensively in the field of civil engineering through experimental mensions, some of its facets will be oblique to the flow direction.
and numerical simulation (Castro, 1979; Hunt et al., 1979; Hunt, Bagnold (1941) illustrated the ideal stages in the evolution of sand
1982; Castro and Dianat, 1984; Martinuzzi and Tropea, 1993; Lamb, shadows (Fig. 1), and noted that the size of the sand shadow is very
1994; Chou and Chao, 2000). There exists a considerable amount of sensitive to the shape of the bottom of the obstacle. However, the
published data for flows around obstacles. However, most literature characteristics of the wakes that develop behind angular bodies, and
is concerned with the flow structure upstream of surface-mounted especially the secondary airflow patterns behind these bodies, and
rectangular cylinders in channel flow, there are few studies concern- their relationship with their shape and changes in the angle of inci-
ing the three-dimensional phenomena in the lee of an obstacle with dence of the wind are not yet fully understood. To investigate the
different shapes and, in particular, the sedimentological process in- mechanisms responsible for the formation and evolution of sand
volved are still unclear. When a sufficiently strong wind flows around shadows under complicated secondary airflows, we performed a
a rectilinear obstacle, flow separation occurs in the lee of the obstacle. scaled wind-tunnel simulation. Our goal was to improve understand-
The nature and significance of this and related secondary flows are ing of the three-dimensional airflow patterns behind cuboid obstacles
controlled by the interaction of at least two factors: the obstacle's using models with different shapes and under different angles of inci-
shape and the angle of incidence of the flow (Sweet and Kocurek, dence of the wind. We characterized the airflow fields using particle

Fig. 2. Experimental set-up used to measure (A) the vertical flow patterns and (B) the horizontal flow patterns behind an obstacle.
260 W. Luo et al. / Geomorphology 139–140 (2012) 258–270

image velocimetry (PIV), and used the results to define the horizontal flow patterns in our tests. Based on the dimensions of the above-
and vertical variations in airflow patterns as a function of the wind in- mentioned real-world obstacles, the geometric scaling of the obstacle
cidence angle and the obstacle's shape ratio, taking advantage of the models was therefore between 1:40 and 1:120. The ratio of model
ability of PIV to provide non-intrusive velocity measurements within height to the boundary layer height, H/δ, was 0.21 (i.e., 25:120).
the whole field of the target area. Based on our results, we discuss the Thus, the tunnel's flow depth posed little restriction on boundary
geomorphological significance of the lee vortex and wake. layer development over the obstacle models (White, 1996). The
ratio of the model's width to the width of the working section of
2. Scaled wind-tunnel simulation the wind tunnel was 0.125 (i.e., 50:400). This ensured that wind tun-
nel constraints (“wall effects”) on the flow patterns around the
2.1. Experimental set-up models would be insignificant. In the tests, we positioned the models
in the middle of the working section of the wind tunnel.
The scaled simulation experiments were carried out in a wind tunnel For small-scale experimental models to adequately simulate field
at the Key Laboratory of Desert and Desertification of the Chinese phenomena, a number of dimensionless variables should not differ
Academy of Sciences. The blow-type, non-circulating wind tunnel significantly between the model and the corresponding full-scale
(Fig. 2) has a total length of 10.5 m, with a 4 m-long test section. The values (Kind, 1976; Iversen, 1981). The constraints imposed by the
cross-sectional area of the test section is 0.4 m tall by 0.4 m wide. The conditions in wind-tunnel studies rarely allow all of these similarity
free-stream wind velocity in the wind tunnel can be varied from 1 to requirements to be met simultaneously, and some compromises
35 m s− 1. To ensure that the test airflow is aerodynamically rough, a must thus be adopted (Musick et al., 1996). The geometric scale of
roughness array at the front of the working section enhances boundary the test models (between 1:40 to 1:120) in the present study rough-
layer development in the working section, and the boundary layer in ly met the strict Reynolds number requirements for dynamic simi-
the test section is generally more than 120 mm. larity. Nonetheless, we tried to generally meet the requirements of
The sizes and shapes of sand shadows in arid and semi-arid re- “Reynolds number independence” recommended by White (1996)
gions are highly variable, ranging from meters to kilometers in length, for solid models. We investigated air flow at three free-stream
depending on the obstacles that create them, and both parameters wind velocities (U = 6, 8 and 10 m s − 1) so that the flow Reynolds
are controlled primarily by the frontal area (height × width) of the ob- numbers (Re = Uδ/ν) would be sufficiently high, ranging from
stacle (Cooke et al., 1993). There is considerable variation in the 5.14 × 10 4 to 8.57 × 10 4, where δ is the thickness of the boundary
shapes of obstacles in the real world, so it is impossible to create a sin- layer and v is the kinematic viscosity of the air. In addition, the Reyn-
gle representative model that would be geometrically similar to all olds number calculated using the width of the obstacle model
the obstacles found in the field. We instead chose a scale model com- (50 mm) ranged from 2.14 × 10 4 to 3.57 × 10 4. During the testing,
parable to the kinds of anthropogenic obstacles that are widely dis- the models were positioned 3.0 m downwind from the leading
tributed at the edges of deserts in northern China and that are edge of the test section of the wind tunnel, at a distance equal to
threatened by sand deposition. These obstacles (cairns and buildings) 25 times the boundary layer thickness, which meets the general
generally have a height of 1.0 to 3.0 m and a frontal area of approxi- rule for matching mean velocity profiles (White, 1996). Under
mately 1 to 20 m 2. these conditions, Reynolds number independence was attained for
Fig. 3 shows a schematic diagram of the wind obstacle model and the solid models used in our study (Townsend, 1956; Snyder,
the coordinate system used in this study. The height (H) and width 1972; White, 1996).
(W) of the test model were 25 and 50 mm, respectively, resulting in Seginer's (1975) experiment showed that the size of the lee vortex
a frontal area of 1250 mm 2. The thickness (B) was set to 5, 10, and decreased greatly as the wind flow became increasingly parallel to a
20 mm to generate different shape ratios (λ = 0.2, 0.4 and 0.8), barrier, and at angles of incidence between 30° and 50°, velocities in
which is defined as the ratio of the top (or basal) area of the test this vortex flow can exceed velocities in the unaffected upwind
models to its side area perpendicular to the wind. Thus, we examined flow. To examine the effect of obstacle orientation (here, represented
the effects of three cuboid models with different shape ratios on wind by the wind incidence angle), we chose 10 incidence angles (α = 45°,

Fig. 4. (A) Horizontal velocity and (B) turbulence intensity profiles for the approaching flow as a function of height in the absence of the model obstacles. u(z) = wind velocity at
height z above the surface; U = free-stream wind velocity. Heights represent multiples of H, the height of the obstacle.
W. Luo et al. / Geomorphology 139–140 (2012) 258–270 261

50°, 55°, 60°, 65°, 70°, 75°, 80°, 85°, and 90°), which we achieved by exposed by the laser pulses, a synchronizer (a timing controller that
rotating the model in the x–y plane on the wind tunnel floor. used highly accurate electronics to synchronize the laser with the
We measured the wind field around the obstacle models using a camera), and software and a data-acquisition card that captured,
PIV device provided by the Beijing Cubic World Science & Technology stored, processed and displayed the data on a computer. This mea-
Development Co., Ltd. It has five key components (Fig. 2): a double- surement technique is based on a combination of laser-imaging tech-
pulsed laser (two laser pulses that illuminated the seeding material nology with digital image processing. For more details concerning the
in the target area within a short time interval), light-sheet optics (op- measurement principle, see Stanislas et al. (2000) and Dong et al.
tics that transformed the laser light into a thin light plane that could (2007a). We used a mist of olive oil droplets (with a mean particle di-
be guided into the target area), a CCD (charge-coupled device) cam- ameter less than 1 μm) diffused by a thermal aerosol generator at the
era with a fast frame-transfer CCD that captured the two frames entrance of the drive section of the wind tunnel as the seeding

Fig. 5. Illustrations of the horizontal streamline patterns around the model obstacles (λ = 0.8, U = 8 m s− 1) as a function of the incidence angle (α). Units on the two axes of the
graphs represent multiples of W, the width of the obstacle.
262 W. Luo et al. / Geomorphology 139–140 (2012) 258–270

material. The wind velocities measured by the PIV equipment were


calibrated using a static Pitot tube before the test, and the difference
between the two measurements was less than 0.5%.
Fig. 2 shows the experimental set-up and the layout of the instru-
mentation. We selected a vertical plane parallel to the wind direction
and that passed through the centerline of the models, and a horizontal
plane that was parallel to and 5 mm above the floor of the wind tunnel,
the lowest level that our PIV system could resolve, and investigated the
flow patterns in these two planes as functions of the model's shape ratio
and the wind velocity. To measure airflow fields in the vertical plane,
we projected a light sheet from the ceiling of the wind tunnel
(Fig. 2A), and positioned the CCD camera of the PIV system 0.25 m
from the light sheet, resulting in a target measurement area 104.4 mm
long by 104.4 mm tall (producing a 2048 × 2048 pixel image with a
magnification coefficient of 0.051). To measure airflow fields in the hor-
izontal plane, we projected the light sheet from the side of the wind
tunnel (Fig. 2B), and positioned the CCD camera with a wide-angle
lens 0.50 m above the light sheet, resulting in a target measurement Fig. 6. Variation of the deflection angle for the horizontal reverse cells (θ, the angle be-
area 161.8 mm long by 161.8 mm wide (producing a 2048 × 2048 tween the x-axis and a line connecting the flow separation and reconnection points) as
pixel image with a magnification coefficient of 0.079). The image acqui- a function of the wind incidence angle (α) and wind velocity.
sition rate was set at 10 frames per second. The estimated wind velocity
represented the average of the values recorded over a period of 40 s
(i.e., 400 frames). Each pair of two frames (separated by an interval of horizontal and vertical planes. The displaced flows in the two planes in-
80 to 120 μs) yields a wind-velocity dataset. The final measurement re- evitably interfere with each other, forming layers with different flow
sults thus represent the average of 200 records, ensuring its statistical patterns (Hesp, 1981; Dong et al., 2008). Since the flow patterns were
significance. similar at different shape ratios and wind velocities (independent of
the Reynolds number), we have focused in this paper on the flow pat-
2.2. Approaching flow terns behind a typical obstacle (with λ = 0.8) as an example of the
flow separation and reattachment that occur in both horizontal and ver-
The secondary flows that develop in the presence of obstacles are tical planes in the remainder of our discussion.
formed by modification of the approaching flow. We measured the ve-
locity profiles of the approaching wind above the bare tunnel floor be- 3.1.1. Horizontal flow patterns
fore testing of the model obstacles, and calculated the turbulence To avoid the complexity created by changes in the obstacle geometry
intensity as a function of height (Fig. 4). The vertical and lateral velocity and orientation, many previous studies of the horizontal flow patterns
components of the approaching wind were negligible compared with around an obstacle have focused on the flow dynamics around typical
the longitudinal (horizontal) velocity component. The velocity profiles simple, symmetrical solid obstacles such as cylinders and spheres. Most
in the turbulent boundary layer of the approaching flow were success- of the classical work was thus concerned with idealized two-
fully fitted using a logarithmic expression (r2 > 0.99, P b 0.05): dimensional horizontal flows, and the researchers calculated wake
widths, points of flow separation, and velocity profiles for these obstacles
uðzÞ u z (e.g., Hesp, 1981; White, 2004). This research found that the horizontal
¼ ln
U k z0 flow patterns around cylinders and spheres were functions of the obsta-
cle's Reynolds number (White, 2004). However, for a Reynolds number
where u(z) is the mean horizontal wind velocity at height z, U is the greater than 6.96 ×104, as in the present study, the horizontal flow pat-
free-stream wind velocity, z0 is the aerodynamic roughness length, u* tern should be independent of the Reynolds number, and horizontal flow
is the friction velocity, and k is the von Karman constant. The friction ve- separation results in the formation of two opposing vortices in the lee of
locities corresponding to the three free-stream wind velocities (6, 8 and the obstacle (White, 2004). For three-dimensional solid obstacles, the
10 m s− 1) were 0.26, 0.33 and 0.37 m s − 1, respectively. The aerody-
namic roughness lengths corresponding to these velocities were
1.2 × 10− 5, 0.8 × 10− 5 and 0.4 × 10 − 5 m, and decreased with increasing
qffiffiffiffiffiffiffi
free-stream wind velocity. The turbulence intensity (defined as u′ 2 =U,
u′ is the fluctuation velocity of the streamwise velocity components)
generally decreased with increasing height, ranging from 0.08 to 0.14
close to the ground surface.

3. Results and discussion

3.1. Flow separation and reattachment

Unlike the airflow produced behind porous obstacles (e.g., fences


and shrubs), the airflow fields behind solid obstacles such as those in
the present study are simplified by the absence of a bleed flow that
passes through gaps in a porous obstacle. As is the case for two- Fig. 7. Illustration of the parameters used to characterize the horizontal reverse cells
that formed behind an obstacle (λ = 0.8, α = 90°, U = 8 m s− 1). S and C are the flow di-
dimensional solid obstacles (i.e., obstacles that are thin perpendicular vergence and convergence points, respectively; lc and w are the flow convergence po-
to the wind direction) no bleed flow exists, so flow patterns behind sition and width of the horizontal reverse cell, respectively. All values are expressed as
the obstacles were dominated by simultaneous displaced flows in the multiples of the width (W) of the obstacle.
W. Luo et al. / Geomorphology 139–140 (2012) 258–270 263

flow pattern depends on both the width of the obstacle and its orienta- The results of our tests indicated that although the dimensions of
tion with respect to the wind (Howard, 1985; Cooke et al., 1993). the flow patterns were affected by the free-stream wind velocity, the
Fig. 5 shows the typical horizontal streamline patterns around the pattern itself remained similar for the range of wind velocities (6 to
obstacle models (λ = 0.8) for the different flow incidence angles at a 10 m s − 1) and Reynolds numbers (5.14 × 10 4 to 8.57 × 10 4) that we
free-stream wind velocity of 8 m s − 1. The flow patterns behind the ob- investigated. The flow structures illustrated in Fig. 5 thus appear to
stacles can be divided into two parts based on a somewhat ill-defined be generally representative of the range of free-stream wind veloci-
discontinuity surface according to Bagnold's (1941) method; this sur- ties and modeled obstacles that we examined. The shape of the re-
face represents the boundary between the external (outer) flow region verse cells was roughly elliptical, and the flow began to recover
that mostly passes by the obstacle without significant modification, and some distance downwind from the separation cell.
the internal flow region behind the obstacle in which a reverse flow de- We chose four parameters to characterize the separation cell: the
velops. For all incidence angles, the lee airflow separated just behind the flow convergence position (lc), width (w), area (sH), and aspect ratio
downwind flank of the obstacle and recovered at some distance (typi- (w/lc) of the horizontal reverse cell (Fig. 7). The flow convergence posi-
cally around 2W) downwind of the obstacle. Immediately downwind tions are expressed as the distance of the convergence point from the
of the point of flow separation, a reverse cell forms in the wind shadow center of the downwind edge of the obstacle, and could be identified ac-
of the obstacle and fills this space with vortices of reversed airflow. The curately from the streamlines. This parameter characterizes the long
airflow is characterized by two opposing vortices when the incidence axis of the reverse cell. The width of the reverse cell can be roughly de-
angle is 60° or greater. In contrast with permeable obstacles, the two fined by the position of the outermost boundaries of the two areas with
vortices were asymmetrical at all incidence angles except 90° (perpen- closed streamlines. This width thus characterizes the short axis of the
dicular to the wind), although the values at the angles of 80° and 85° reverse cells. The area of the reverse cells is defined as the area enclosed
were nearly symmetrical. The pattern of flow separation and reversal by the outermost boundaries of the flow lines that define the reverse
behind porous obstacles is different because of bleed flows that push cells, and represents the magnitude of the development of a horizontal
the reverse cells so that they form farther behind the obstacle. The de- flow reversal. The aspect ratio is defined as the ratio of the width to the
flection angle θ (the angle between the x-axis and the line that connects length of the reverse cells, and thus characterizes the degree of horizon-
the flow divergence point S with the flow convergence point C) for the tal flatness (i.e., along the x-axis) of the reverse cell. The flow conver-
reverse cells behind the obstacles decreased with decreasing incidence gence position and the width of the horizontal reverse cell are
angle until one of the reverse vortices disappears, at an incidence angle expressed as multiples of the obstacle's width (W), whereas the area
less than 65° (Fig. 6). As the incidence angle decreases, crosswinds of the horizontal reverse cell is converted into a dimensionless relative
move the reverse cell laterally and downwind (Fig. 5). This suggests area expressed as sH/Ab (where Ab = W × B = 50 mm× 20 mm, repre-
that the incidence angle does not determine the existence of a flow sep- sents the horizontal basal area of the obstacle model).
aration and a flow reversal behind obstacles with a constant shape ratio, Fig. 8 shows the variation in the four parameters as a function of the
although it may influence the characteristics of the separation cell. The incidence angle and the free-stream wind velocity behind obstacles
positions of the centers of the two opposing vortices are affected by the with shape ratio of 0.8. Because the flow convergence point tended to
incidence angle; the distance between the two points increases to a move farther downwind as the incidence angle decreased, the width
value greater than the width of the obstacle, ranging from 1.05W to of the reverse cell generally decreased with increasing incidence
1.23W, as the incidence angle decreases, but does not appear to differ angle, indicating that the reverse cell develops more strongly and be-
between wind velocities. comes more symmetrical as the wind becomes more perpendicular to

Fig. 8. Variations in the parameters of the horizontal reverse cells as a function of the incidence angle and wind velocity (λ = 0.8).
264 W. Luo et al. / Geomorphology 139–140 (2012) 258–270

the obstacle, resulting in both narrowing and shortening of the reverse Fig. 9 shows the horizontal isovelocity contours behind obstacles
cell. The area of the reverse cell increased as the incidence angle in- with a shape ratio of 0.8 under a free-stream wind velocity of
creased, whereas the aspect ratio generally decreased, indicating that 8 m s− 1. For all incidence angles, a negative or zero wind velocity oc-
the increasing separation of the flow at larger incidence angles flattens curs immediately behind the obstacle. As the incidence angle increases,
the reverse cells and increases the sheltered area in the lee of the obsta- the isovelocity lines become denser, indicating a greater velocity reduc-
cle, where sand can accumulate. The response of the four parameters of tion. A negative-velocity bubble (i.e., a vortex of reversed flow) caused
the reverse cell to wind velocity does not show a clear trend. by flow separation formed at a distance starting from 0.5W to 1.5W

Fig. 9. Illustration of typical horizontal isovelocity contours behind model obstacles with different incidence angles (λ = 0.8, U = 8 m s− 1). All distance values are expressed as mul-
tiples of the width (W) of the obstacle.
W. Luo et al. / Geomorphology 139–140 (2012) 258–270 265

downwind in the lee of the obstacles. The velocity in the bubble became move farther from the x-axis as the incidence angle decreased, which
increasingly negative and the bubble moved closer to the obstacle as the may lead to different forms of sediment deposition behind the obstacle.
incidence angle increased. The area of the negative-velocity bubble de- As was the case for the shift in the position of the reverse cells, the shift
creased as the incidence angle decreased. Using the −0.5 m s− 1 isove- in position of the negative-velocity bubbles and the magnitude of the
locity line as an example, the center of the bubble enclosed by this line is velocity reduction reveal the effects of an oblique wind flow.
located at 1.2W to 1.3W downwind of the obstacle for an incidence
angle less than 60°, but is located at 0.6W to 0.7W downwind for an in- 3.1.2. Vertical flow patterns
cidence angle greater than or equal to 80°. The shape of the negative- Previous research on two-dimensional vertical flow patterns, and es-
velocity bubble changed in complex ways, and the bubble seemed to pecially on the vertical flow separation over transverse dunes and porous

Fig. 10. Illustration of the vertical streamline patterns around the model obstacles (λ = 0.8, U = 8 m s− 1). Units on the two axes of the graphs represent multiples of H, the height of
the obstacle. The dashed line indicates the change in position of the center of the vertical reverse cell.
266 W. Luo et al. / Geomorphology 139–140 (2012) 258–270

center of the vertical reverse cells shifts downwind, away from the ob-
stacle, with increasing incidence angle. We chose four parameters to
characterize the vertical reverse cell (Fig. 11): the reattachment dis-
tance (lR), height of the reverse cell (h), area of the cell (Sv), and the as-
pect ratio (h/lR). These parameters are defined as follows.
The reattachment distance is the horizontal distance between the
downwind edge of the obstacle and the reattachment point (point R
in Fig. 11), where the separated flow begins to reattach and move for-
ward along the surface. This parameter characterizes the long (hori-
zontal) axis of the separation cell. Theoretically, the downwind apex
Fig. 11. Illustration of the parameters used to characterize the vertical reverse cell
(λ = 0.8, α = 90°, U = 8 m s− 1). R is the reattachment point; lR is the reattachment dis-
of the reverse cell and the reattachment point should be located at
tance; h is the height of the vertical reverse cell. Units on the two axes of the graphs different positions because the former can occur above the surface
represent multiples of H, the height of the obstacle. whereas the latter must be at the surface. However, the difference
in their positions appears to be sufficiently small that it can be
neglected, and in practice, both the downwind apex of the reverse
fences, mainly focused on the dynamical effect of the stoss (windward) cell and the reattachment point can be characterized by the reattach-
slope on the airflow patterns in the lee of the obstacle and on the effect ment distance (Walker, 1999; Dong et al., 2007a). We determined the
of fence porosity on the shelter distance downwind of the barrier so as reattachment point by examining the horizontal velocity component
to determine the optimal design of sand-protection fences (e.g., Walker of the flow along the ground in the lee of the obstacle. This point
and Nickling, 2002; Dong et al., 2007a,b). Flows that separated above was located where the horizontal velocity changes sign, moving
dune ridges, cliff bodies, and porous fences generally reattached at some from upwind to downwind. The height of the reverse cell in Fig. 11
distance in the obstacle's lee, which would lead to the deposition of wind- can be roughly defined by the maximum height of the upper bound-
blown sand upwind of the reattachment point. If the stoss slope of the ob- ary of the closed streamlines, which characterizes the short (vertical)
stacle is greater than 50°, a vortex forms immediately upwind of the axis of the separation cell. The area of the reverse cell is defined as the
obstacle, producing circulation patterns that create an “echo dune” area enclosed by the outer boundary of the reversed-flow stream-
(Tsoar, 1983). The flow pattern in a vertical plane (here, the x–z plane) lines, and represents the extent of the development of the vertical
therefore has strong significance for the formation of sand shadows. flow reversal. The aspect ratio is defined as the ratio of the height of
Fig. 10 shows the typical vertical flow patterns behind obstacles the reverse cell to the reattachment distance. This parameter charac-
with a shape ratio of 0.8 under a free stream wind velocity of 8 m s− 1 terizes the vertical flatness of the reverse cell. The reattachment dis-
at different incidence angles. For all incidence angles, flow separation tance and height of the vertical reverse cell are expressed as
occurred at the downwind edge of the obstacle, leading to the develop- multiples of the obstacle height (H), and the area of the vertical cell
ment of a reverse flow in the lee of the obstacle, immediately down- is converted to a dimensionless relative area expressed as sV/Ab
wind of the separation point. The shape of the reverse cell was (where Ab equals the base area of the obstacle model).
roughly elliptical, and the flow began to recover at some distance Fig. 12 shows the variations in these four parameters as a function of
(2.5H to 3.0H) downwind from the separation cell. The position of the the incidence angle and wind velocity for obstacles with a shape ratio of

Fig. 12. Variations in the parameters of the vertical reverse cells as a function of the incidence angle and wind velocity (λ = 0.8).
W. Luo et al. / Geomorphology 139–140 (2012) 258–270 267

0.8. All four parameters varied more with respect to the incidence angle formation of a negative-velocity bubble some distance (between
than with respect to the wind velocity. The reattachment distance in- 0 and 3.0H) downwind of the obstacle, which would lead to the deposi-
creased linearly from 2.5H to 3.5H with increasing incidence angle. tion of eolian sediment. The bubble of negative velocities moved down-
These reattachment distances are much shorter than those reported for wind and decreased in width as the incidence angle decreased and
two-dimensional transverse dunes (Dong et al., 2007a) and porous expanded farther downwind as the incidence angle increased, similar
fences (Dong et al., 2007b) and slightly shorter than those reported for to the pattern for the vertical reverse cells shown in Fig. 10. The area
three-dimensional model shrubs (Dong et al., 2008), but longer than of the negative velocity bubble increased with increasing incidence
those reported for three-dimensional artificially-flexible plant clumps angle, reaching its maximum size at an incidence angle of 90°, when it
(Hesp, 1981). Dong et al.'s (2007a, 2007b, 2008) wind-tunnel simulation covered a distance of about 3.0H along the surface. Both the form and
showed that the reattachment distance was a function of the stoss slope the evolution of the vertical and horizontal isovelocity patterns with re-
for transverse dunes and a function of the porosity (density) of porous spect to the incidence angle were comparable, which also suggests that
fences and shrubs. It generally ranged from 4.8H to 10.8H for transverse the development of flow velocity patterns in three dimensions are
dunes as the stoss angle increased, and ranged from 8.4H to 16.0H for po- related.
rous fences as porosity decreased (for porosity less than 0.2), but varied
within a narrow range (from 3.7H to 4.4H) when the shrub density was 3.2. Effect of the shape ratio
greater than 0.8. Hesp's (1981) wind-tunnel test showed a shorter reat-
tachment distance (approximately between 1.5H to 2.1H, estimated Even though the flow patterns behind the obstacle were most
from the three-dimensional flow pattern diagram (Fig. 2in his paper)) strongly affected by the flow incidence angle, the shape ratios also
than our result, mainly because of the large size and the flexibility of influenced the flow separation. The aspect ratio of obstacles has
plant clumps that he used in the simulation caused the flow to be reat- been shown to influence the intensity of erosion (Iversen et al.,
tached near to the plant. Furthermore, the flow visualizing techniques 1990). For example, wakes in the lee of sharp, continuous scarps are
in Hesp's research were not as accurately as the PIV system in the present much longer than those in the lee of smooth, isolated obstacles
work, which also caused the flow pattern from those earlier methods to (Engel, 1981; Cooke et al., 1993).
be roughly estimated. This difference suggests both that the flow pat- Fig. 15 shows the variations in the mean parameters of the hori-
tern's complexity increases around porous obstacles as a result of bleed zontal reverse cells as a function of incidence angles for obstacles
flows and that the three-dimensionality of vegetated or solid obstacles with different shape ratios. All parameters changed with respect to
creates interference between the horizontal and vertical flow patterns the incidence angle for all three shape ratios, confirming that the
that shortens the reattachment distance. We did not find a significant flow patterns are similar for the three ratios. Except for the aspect
correlation between the reattachment distance and the position of the ratio, the values of the other three parameters increased as the
horizontal flow convergence in our data (Fig. 13A), possibly because shape ratio decreased; that is, the thinner (streamwise direction)
these values were measured at different heights above the surface. the obstacle, the greater the acceleration of the flow, the stronger
The height and area of the vertical reverse cell both increased as the the separation effect, and the greater the convergence distance down-
incidence angle increased (Fig. 12B, C), but the height increased more wind from the obstacle. Thus, the area with a horizontal reverse flow
slowly than the area. The aspect ratio decreased with increasing inci- is wider behind a thin obstacle than behind a thick obstacle, and the
dence angle, but fell within a narrow range (from 0.5H to 0.68H) for flatness of the cell increases behind the thicker obstacle.
all incidence angles, suggesting that the incidence angle affects both Fig. 16 shows the variations in the mean parameters of the vertical
the height and the area of the reverse cell. The area of the horizontal re- reverse cells as a function of the incidence angle for obstacles with
verse cell was roughly related to the area of the vertical reverse cell different shape ratios. The four parameters of the vertical reverse
(Fig. 13B), which means that the three-dimensional reverse cell behind cells showed more complex patterns than with the horizontal cells,
the obstacles expanded both in lateral and vertical directions as the in- but the overall trends with respect to the incidence angle were simi-
cidence angle increased. Our results also show that the shape ratio of lar to those of the horizontal reverse cells. This again suggests that the
the solid obstacles influenced the parameters of both the horizontal development of horizontal and vertical velocity patterns was related.
and the vertical reverse cells. For example, the reattachment distance was largest for a shape ratio
Fig. 14 shows the vertical isovelocity contours behind obstacles with of 0.8 at incidence angles less than 60°, but lowest for this shape
a shape ratio of 0.8 under a free-stream wind velocity of 8 m s − 1. The ratio at incidence angles greater than 65°. This may be because the
streamlines became denser as the flow incidence angle increased. As shape ratio has less effect than the oblique flows (around the obsta-
the airflow passed over the angular bodies, the compressed streamlines cle's flanks) on the flow separation when the incidence angle is less
spread out, leading to a reduction in the mean flow velocity and the than 65°, whereas the shape ratio has a greater effect on the flow

Fig. 13. Relationships between (A) the horizontal reattachment distance and the position of the horizontal flow convergence and (B) between the areas of the horizontal and ver-
tical reverse cells (λ = 0.8, U = 8 m s− 1).
268 W. Luo et al. / Geomorphology 139–140 (2012) 258–270

Fig. 14. Illustration of typical vertical isovelocity contours behind the model obstacles with different incidence angles (λ = 0.8, U = 8 m s− 1). Units on the two axes of the graphs
represent multiples of H, the height of the obstacle.

separation when the airflow is more perpendicular to the obstacle, 4. Conclusions


especially for obstacles with a large shape ratio, for which the wide
top would interfere with flow separation and lead to flow reattach- We simulated the three-dimensional flow velocity field behind cu-
ment near the surface (White, 2004). boid obstacles with different shape ratios as a function of the flow
W. Luo et al. / Geomorphology 139–140 (2012) 258–270 269

Fig. 15. Variations in the mean parameters of the horizontal reverse cells as a function of incidence angle and shape ratio.

incidence angle using scale models in a wind tunnel to provide a deeper with different shape ratios, the incidence angle had a stronger influence
understanding of the aerodynamic factors that control the formation of on the flow patterns that developed around the obstacles. The wind ve-
sand shadows behind the obstacles. Using PIV, we were able to obtain locity also had a relatively small effect on the flow patterns. The airflow
detailed measurements of the airflow patterns behind the model obsta- separated both above and beside the obstacles, forming a negative-
cles. This non-intrusive measurement technique provided rich data on velocity bubble (reversed flow) in the “wind shadow”. This bubble
the flow patterns in both the horizontal and the vertical planes. Our would lead to the deposition of eolian sediments in the shadow zone.
simulation tests appear to be approximately independent of the Reyn- The shadow disappeared at some distance downwind from the obstacle
olds number. Because the flow patterns were similar for obstacles that depended on the incidence angle, and the flow once again merged

Fig. 16. Variations in the mean parameters of the vertical reverse cells as a function of incidence angle and shape ratio.
270 W. Luo et al. / Geomorphology 139–140 (2012) 258–270

with the general flow of the wind, which reflects the overall evolution Chou, J.H., Chao, S.Y., 2000. Branching of a horseshoe vortex around surface-mounted
rectangular cylinders. Experiments in Fluids 28, 394–402.
of sand shadows. Clemmensen, L.B., 1986. Storm-generated eolian sand shadows and their sedimentary
An incidence angle of 60° or 65° appears to be the critical angle structures. Vejers Strand, Denmark. Journal of Sedimentary Petrology 56, 520–527.
below which horizontal reverse cells seem to fade away, and instead Cooke, R., Warren, A., Goudie, A., 1993. Desert Geomorphology. UCL Press, London.
(526 pp.).
of paired reverse cells, only a single small vortex develops. We chose Dong, Z., Qian, G., Luo, W., Wang, H., 2007a. A wind tunnel simulation of the effects of
four parameters to characterize both the horizontal and vertical re- stoss slope on the lee airflow pattern over a two-dimensional transverse dune.
verse cells, and found that all four affected the characteristics of Journal of Geophysical Research 112, F03019. doi:10.1029/2006JF000686.
Dong, Z., Luo, W., Qian, G., Wang, H., 2007b. A wind tunnel simulation of the mean
these cells, depending mainly on the incidence angle. The obstacle's velocity fields behind upright porous fences. Agricultural and Forest Meteorology
shape ratio also affected the characteristics of the reverse cell, espe- 146, 82–83.
cially the flow convergence and reattachment positions and the as- Dong, Z., Luo, W., Qian, G., Lu, P., 2008. Wind tunnel simulation of the three-
dimensional airflow patterns around shrubs. Journal of Geophysical Research
pect ratios in both the horizontal and the vertical planes. The
113, F02016. doi:10.1029/2007JF000880.
reverse cells developed in both planes, and the areas of the two Engel, P., 1981. Length of flow separation over dunes. Journal of the Hydraulics Division
cells were correlated, but not strongly. 107, 1133–1143.
As this was an exploratory study, we analyzed the airflow patterns Greeley, R., Iversen, J.D., 1985. Wind as a Geological Process on Earth, Mars, Venus and
Titan. Cambridge University Press, UK. (322 pp.).
behind the model obstacles with an emphasis on their relationship to Gunatilaka, A., Mwango, S.B., 1989. Flow separation and the internal structure of
the incidence angle and the shape ratio of the obstacle. We found that shadow dunes. Sedimentary Geology 61, 125–134.
the incidence angle was the most important factor that affected the Hesp, P.A., 1981. The formation of shadow dunes. Sedimentary Petrology 51, 101–112.
Hesp, P.A., McLachlan, A., 2000. Morphology, dynamics, ecology and fauna of Arctotheca
flow patterns behind the obstacles, but that the shape ratio was also populifolia and Gazania rigens nabkha dunes. Journal of Arid Environments 44,
important. On this basis, the size and shape of the sand shadows 155–172.
that evolve behind such obstacles should be related to the obstacle's Hesp, P.A., Davidson-Amott, R., Walker, I.J., Ollerhead, J., 2005. Flow dynamics over a
foredune at Prince Edward Island, Canada. Geomorphology 65, 71–84.
shape ratio and the wind's incidence angle. This hypothesis should Howard, A.D., 1985. Interaction of sand transport with topography and local winds in
be confirmed by future tests using eolian sediments to determine the northern Peruvian coastal desert. In: Barndorff-Nielson, O.E., et al. (Ed.),
the shape and size of the sand shadows that form behind the obsta- Proceedings of the International Workshop on the Physics of Blown Sand.
University of Aarhus Press, Aarhus, pp. 511–543.
cles. This will provide stronger evidence of the geomorphologic sig- Hunt, A., 1982. Wind tunnel measurements of surface pressure on cube buiding models
nificance of our results for the formation and development of sand at several scales. Journal of Wind Engineering and Industrial Aerodynamics 10,
shadows. However, the limited space available in wind tunnels, in- 137–163.
Hunt, J.C.R., Abell, C.J., Peterka, J.A., Woo, H., 1979. Kinematical studies of the flows
cluding the one we used for the present study, usually restricts our
around free or surface-mounted obstacles, applying topology to flow visualization.
ability to simulate dune morphology. Additionally, flow patterns be- Journal of Fluid Mechanics 86, 179–200.
hind an obstacle may be changed when a sand mound has formed Iversen, J.D., 1981. Comparison of wind tunnel modeling and full-scale snow fence
in its lee (as indicated in Fig. 1). The present study applies only to drifts. Journal of Wind Engineering and Industrial Aerodynamics 8, 231–249.
Iversen, J.D., Wang, W.P., Rasmussen, K.R., Mikkelsen, H.E., Hasiuk, J.F., Leach, R.N.,
the initial stage in that there is no sand mound in the lee of the obsta- 1990. The effect of a roughness element on local saltation transport. Journal of
cle. Thus, field experiments to study the development of sand Wind Engineering and Aerodynamics 36, 845–854.
shadows will be required to relate the airflow patterns observed in Karcz, I., 1968. Fluviatile obstacle marks from the wadis of Negev (southern Israel).
Journal of Sedimentary Petrology 38, 1000–1012.
the present study to dune morphology and evolution in the field. Fur- Kind, R.J., 1976. A critical examination of the requirements for model simulation of
thermore, this paper only discussed the responses of mean velocity wind-induced erosion/deposition phenomena such as snow drifting. Atmospheric
field behind obstacles to variations in the incident flow and its signif- Environment 10, 219–227.
Lamb, K.G., 1994. Numerical simulations of stratified inviscid flow over a smooth obstacle.
icance on the formation of sand shadows. There should be a causal re- Journal of Fluid Mechanics 260, 1–22.
lationship between the sand shadow formation and other parameters Livingstone, I., Warren, A., 1996. Aeolian Geomorphology: an Introduction. Longman
such as the turbulence structures, shear stress and pressure. Analysis Singapore Publishers (Pte) Ltd., Singapore. (211 pp.).
Martinuzzi, R., Tropea, C., 1993. The flow around surface-mounted, prismatic obstacles
of different aspects of all these parameters in future research will pro- placed in a fully developed channel flow. Journal of Fluids Engineering 115, 85–92.
vide a deeper understanding of the aerodynamics of sand shadow for- Musick, H.B., Trujillo, S.M., Truman, C.R., 1996. Wind tunnel modeling of the influence
mation and its evolvement process. of vegetation structure on saltation threshold. Earth Surface Processes and Landforms
21, 589–605.
Pye, K., Tsoar, H., 1990. Aeolian Sand and Sand Dunes. Unwin Hyman Ltd., London.
Acknowledgments (396 pp.).
Seginer, I., 1975. Flow around a windbreak in oblique wind. Boundary-Layer Meteorology
9, 133–141.
This study was supported by the National Natural Science Foundation Snyder, W.H., 1972. Similarity criteria for the application of fluid models to the study of
of China (40901003) and the West Light Foundation for Doctoral air pollution meteorology. Boundary-Layer Meteorology 3, 113–134.
Stanislas, M., Kompenhas, J., Westerweel, J., 2000. Particle Image Velocimetry: Progress
Program of The Chinese Academy of Sciences (Y028701001). We
Towards Industrial Application. Kluwer Academic Publishers, Dordrecht.
would like to acknowledge the two anonymous referees and the editor Sweet, M.L., Kocurek, G., 1990. An empirical model of aeolian dune lee-face airflow.
Andrew Plater for their constructive comments and suggestions that Sedimentology 37, 1023–1038.
Townsend, A.A., 1956. The Structure of Turbulent Shear Flow. Cambridge University
helped to improve the manuscript.
Press, Cambridge. (315 pp.).
Tsoar, H., 1983. Wind tunnel modeling of echo and climbing dunes. Sedimentology 38,
247–259.
References Walker, I.J., 1999. Secondary airflow and sediment transport in the lee of a reversing
dune. Earth Surface Processes and Landforms 24, 437–448.
Allen, J.R.L., 1982. Sedimentary structures: their character and physical basis.
Walker, I.J., Nickling, W.G., 2002. Dynamics of secondary airflow and sediment
Developments in Sedimentology, volume II. Elsevier, Amsterdam, p. 30B.
transport over and in the lee of transverse dunes. Progress in Physical Geography
Bagnold, R.A., 1941. The Physics of Blown Sand and Desert Dunes. Methuen and Co.
26, 47–75.
Ltd., London. (265 pp.).
White, B.R., 1996. Laboratory simulation of aeolian sand transport and physical
Carter, R.W.G., 1978. Ephemeral sedimentary structures formed during Aeolian
modeling of flow around dunes. Annals of Arid Zone 35, 187–213.
deflation of beaches. Geological Magazine 115, 379–382.
White, F.M., 2004. Fluid Mechanics (Fifth Edition. McGraw-Hill Education (Asia) Co.
Castro, I.P., 1979. Relaxing wakes behind surface-mounted obstacles in rough wall
and Tsinghua University Press, Beijing. 866 pp.
boundary layers. Journal of Fluid Mechanics 93, 631–659.
Wilson, P., 1988. Recent sand shadow development on Muckish Mountain, Co Donegal.
Castro, I.P., Dianat, M., 1984. Surface flow patterns on rectangular bodies in thick
The Irish Naturalists' Journal 22, 529–531.
boundary layers. Journal of Wind Engineering and Industrial Aerodynamics 11,
107–119.

Vous aimerez peut-être aussi