Eur. Phys. J.

E (2018) 41: 53
DOI 10.1140/epje/i2018-11660-5
THE EUROPEAN
PHYSICAL JOURNAL E
Regular Article

Numerical simulation of wind-sand movement in the reversed

flow region of a sand dune with a bridge built downstream
Wei He, Ning Huanga , Bin Xu, and Wenbo Wang
Key Laboratory of Ministry for Education on Western Disaster and Environment, Lanzhou University, Lanzhou 730000, China

Received 28 December 2017 and Received in final form 31 March 2018

Published online: 23 April 2018 – 
c EDP Sciences / Società Italiana di Fisica / Springer-Verlag 2018

Abstract. A bridge built inside the reversed flow region of a sand dune will change the characteristics of
wind-sand movement in this region. The Reynolds-averaged Navier-Stokes simulation and discrete particle
tracing are used to simulate the wind-sand movement around a sand dune with a bridge built inside the
reversed region. Three cases with different bridge positions are studied. The results show that 1) compared
with the isolated dune case, a tall bridge built at the leeward toe leads to an increase in the deposition
rate on the leeward slope and a longer reversed flow region downstream of the sand dune; meanwhile, the
high speed of crosswind on the bridge indicates that some measures should be taken to protect trains from
strong crosswind; 2) a low bridge at the leeward toe has little effect on the sand deposition and reversed
flow region of the dune; however, low sand transport rate and crosswind speed on the bridge show that
anti-crosswind/sand measures should be taken according to the actual situation and 3) a low bridge on
the leeward slope has little effect on the length of reversed flow region, however, high crosswind speed
and sand flux on the bridge reveal the need of anti-crosswind/sand measures on the bridge. Moreover, the
bridges in the reversed flow region increase the sand flux near the leeward crest; as a result, the moving
patterns of the sand dune are changed.

1 Introduction
Railways suffer from strong crosswind and sand hazards
when going through regions of Gobi and desert. The cur-
rent length of sand hazard affected railway sections is more
than 1000 km in China, and the second Lan-Xin line [1],
Qinghai-Tibet line [2] and Baotou-Lanzhou line [3] are the
most representative. Wind-sand disasters endanger the
running trains, increase the maintenance costs and cause
huge financial loss. To reduce the damage by wind-sand
disaster, simulation studies are required before a railway
is constructed, and the results may provide good guide-
line before the planning of a railway line and help to avoid
extreme sand accumulation over the railbed. The rapid de-
velopment of numerical simulations has actually prepared
convenient tools for early evaluation.
To connect the Dunhuang city with the Qinghai-Tibet
railway, a new 520-km-long line is under planning. In the
south of the Dunhuang city, the railway will go through
a 10.7-km-long valley with moving sand dunes on both
sides (fig. 1(a)). Long-term observation data of wind di-
rection in the region is still lacking. To obtain the wind
rose, we used anemometers to measure wind speed at 2
meters above the ground during May 19th to June 11st,
2015. It is shown in fig. 1(b) that the main wind directions
a
e-mail: huangn@lzu.edu.cn (corresponding author)
are west and northwest. The dispersion in the sand val-
ley ranges from 60 m to 100 m, and to cross this complex
landform region, designers plan to build the railway on a
bridge. Because the radius of turning circle and track ir-
regularity are strictly specified, the options for the design
of the railway line are limited, so the bridge will inevitably
get close to the sand dunes; in some extreme cases the
bridge will pass by the sand dune inside the reversed flow
region. Some field observation results showed serious sand
accumulation on the bridge if it is too close to the leeward
slope of a sand dune. Since sand particles on the railway
will endanger the safety of the operation of trains and re-
duce the service life of the tracks, some problems should
be studied regarding wind-sand prevention: 1) the inter-
action between the sand dune and the part of the bridge
inside the reversed flow region and 2) the mechanism of
sand accumulation on the bridge.
Recently many studies have been conducted on the
wind flow over dunes, turbulence structure and sand move-
ment. It is known that the reversed flow region plays a sig-
nificant role in maintaining the shapes of sand dunes [4].
A mixing layer and highly turbulent shear zone were also
identified which affected the flow reattachment and sub-
sequent sediment transport in the leeward region of the
dune [5,6]. Because of the complex flow structure and the
existence of reversed sand transport, the leeward region
was generally neglected in the previous experiments [7].
Page 2 of 9 Eur. Phys. J. E (2018) 41: 53

Fig. 1. Landform of the sand valley for the railway to go through (a) and the wind rose (b).

Fig. 2. Four cases of different bridge-dune positions.

Instead, a simulation is used to better understand the (RANS) method is used to calculate the turbulent flow
wind-sand flow motion in the leeward region. Recently, over the slope and bridge. Particle movement is calculated
many studies on the wind structure over dunes have been by the discrete particle tracing and the grain-bed collision
conducted using the computational fluid dynamic (CFD) is simulated with the splash function. Three cases with
methods [8–16]. However, sand particles were not consid- different bridge-dune positions are simulated based on the
ered in these models. In the wind-sand flow, sand parti- actual conditions (fig. 2), including case 1, a high bridge
cles react to the air flow, so the wind field can be very built at the leeward toe of the dune; case 2, a low bridge
different when coupled with sand particles; thus the sand at the leeward toe of the dune and case 3, a low bridge
particles should not be ignored. Some experimental re- on the leeward slope. Case 4 is an isolated sand dune act-
sults showed the relationship between flow separation and ing as a control group. The wind flow, sand transport and
sand transport over dunes [17,18]. The effects of turbu- sand accumulation over the leeward slope and bridge are
lence on sediment transport and the development of an discussed in detail. At last, some advices on wind-sand
aeolian bed form were also investigated [4,19,20] and the prevention are proposed to protect the railway from sand
instantaneous turbulent velocity was found to have a large hazards and harmful crosswind.
influence on sand transport [21].
The results of the aforementioned studies show that
the flow structure in the leeward region is complex, and 2 Method
the reversed sand transport and reattachment effect are
important to maintain the shape of the sand dune. As a 2.1 Equations of wind flow field
result, a bridge inside the reversed flow region will further
complicate the wind-sand field: the stable shape of the The governing equations of incompressible turbulent flow
sand dune will be disturbed; moreover, sand hazards and with constant viscosity under the influence of moving sand
strong crosswind will occur on the bridge. particles are [22]:
This paper aims to investigate the flow structure, sand ∂ui
transport rate and sand deposition over the leeward slope = 0, (1)
∂xi
of the dune when a bridge is built inside the reversed
flow region. The sand flux and crosswind on the bridge ∂ui ∂uj 1 ∂p ∂ 2 ui ∂ui uj
floor is also studied. The Reynolds-averaged Navier-Stokes + uj = fi − +ν − , (2)
∂t ∂xj ρ ∂xj ∂xj ∂xj ∂xj
Eur. Phys. J. E (2018) 41: 53 Page 3 of 9

where t stands for the time; xi (x1 , x2 , x3 ) stands for the 1

Fg = π(ρp − ρ)D3 g, (12)
directions (the stream wise, span wise and vertical); ui 6
stands for the transient speed in the i-th direction; p, ρ and 24 6
ν are the fluid pressure, density and molecular viscosity CD = + √ + 0.4, Rep ≤ 1000,
Rep 1 + Rep
coefficient, respectively; fi is the volume force which is the
CD = 0.424, Rep > 1000, (13)
reaction force of sand particles to the wind:
|Vr |D
Q  1 Rep = (14)
npcell
fi = Cdp Ap |Vr |Vr , (3) ν
Vcell n=1 2
where D stands for the average diameter of particles, up
where Vcell stands for the volume of grids, Q is a constant is the speed of sand particles in wind-sand flow, ρp =
defined in Dupont et al. [22], npcell stands for the parti- 2650 kg/m3 is the sand density, ρ is the fluid density, g
cle number in a cell, Cdp is the drag coefficient of sand is the gravity acceleration, CD is the drag coefficient for
particles, Ap is the projected area of grains and Vr is the sphere, and Rep is the Reynolds number of sand move-
relative velocity. ment.
According to the basic idea of Reynolds-averaged
Navier-Stokes (RANS) equations, the standard k−ε model
is a model based on model transport equations for the tur- 2.3 Sand bed splash process
bulence kinetic energy (k) and its dissipation rate (ε). The
k and ε are obtained from the following transport equa- In the wind-sand flow, sand particles take off and enter the
tions [23]: saltation layer before eventually falling back to the surface
   due to the gravity and impacting the bed. Sand particles
∂ ∂ ∂ μt ∂k
(ρk) + (ρkui ) = μ+ + Gk − ρε, may rebound and eject other particles on the bed, and
∂t ∂xi ∂xj σk ∂xj this process is called the sand-bed collision. The splash
(4) equations are defined as follows.
  
∂ ∂ ∂ μt ∂ε
(ρε) + (ρεui ) = μ+
∂t ∂xi ∂xj σε ∂xj
2.3.1 Movement of impacting particles
ε ε2
+C1ε Gk − C2ε ρ , (5)
k k The probability for a sand particle to rebound after im-
where σk and σε are the Prandtl number of the turbulence pacting the bed is [24]:
kinetic energy and its dissipation rate, μ is the viscosity
of the fluid. The remaining variables in eqs. (4) and (5) Preb = 0.95[1 − exp(−γreb vimp )], (15)
are defined as follows:
where vimp is the impact velocity, and γreb = 2 sm−1 is an
Gk = μt S 2 , (6) empirical parameter.

S = 2Sij Sij , (7) The rebound velocity and angle of impact particles
k2 follow a normal distribution [25]:
μt = ρCμ , (8)  
ε
1 (vreb − vreb )2
where the Sij is the shearing rate tensor. The model con- prob(vreb ) = √ exp − 2 , (16)
2πσreb 2σreb
stants C1ε , C2 , Cμ , σk and σε have the following default
values: where vreb  = 0.6vimp is the averaged rebound velocity,
C1ε = 1.44, and σreb = 0.25vimp is the standard deviation. The re-
bound angle follows the same distribution, αvreb  = 30◦
C2ε = 1.92,
and σαvreb = 15◦ . Variation in the impact angle has little
Cμ = 0.09, influence on the splash process.
σk = 1.0,
σε = 1.3. (9)
2.3.2 Movement of splashing particles

2.2 Equations of saltation movement of sand particles The number of splashing particles is related to velocity of
the impacting particle [26–28]:
The movement equations of saltating particles subject to
the gravity Fg and aerodynamic drag FD are: a
Nej =  vimp , (17)
dup  gdp 
mp = FD + Fg , (10)
dt
1 where the constant a ranges from 0.01 to 0.05 and dp  is
FD = CD πD2 ρ|Vr |Vr , (11) the averaged sand diameter on the ground.
8
Page 4 of 9
Fig. 3. Schematic diagram of the computational domain (a) and grid system used in the simulations (b).
The velocity of splashing particles follows an exponen-
tial distribution [25]:
 
1 vej
prob(vej ) = exp − , (18)
vej  vej 
where vej  = 0.08vimp [29]. The splashing angle follows a
normal distribution, αvej  = 30◦ , αhej  = 30◦ , σαvej =
15◦ , σhej = 15◦ .
2.4 Computational domain and boundary conditions
Figure 3 shows the parameters in the model. The com-
putational domain is a 300 m × 32.6 m × 100 m cube. The
sand dune is a triangle sand bed whose size is 10 m in
height and 60 m in length. To obtain a fully developed
wind-sand flow, the inlet boundary is set 10 m upstream
of the windward toe. The coordinate system is specified in
fig. 3(a). The boundary conditions of the wind field are: 1)
on the ground, sand dune and bridge surface, the no-slip
boundary condition is used; 2) at the top and sides of the
domain, the symmetric boundary condition is used, and
the derivatives of variables u, v, w and p with respect to
the vertical ordinate z are set to be zero; 3) at the inlet
boundary, the logarithmic profile u = u∗ /κ ln(z/z0 ) is set
for the initial condition and 4) at the outlet boundary, zero
diffusion is assumed for all flow variables. The diameters
of sand particles in nature obey the logarithmic normal
distribution [30]:
 
1 (ln d − μ)2
p(d) = √ exp − (19)
dσ 2π 2σ 2
where the average diameter μ = 2.5e−4 m and σ = 7e−5 m.
To ensure the accuracy and iteration speed of the
simulation, structured mesh is used. The sizes of the
grids attaching to the ground and sand dune are 0.5 m ×
0.5 m × 0.005 m. The grids attaching to the bridge are
0.5 m × 0.5 m × 0.125 m. The biggest grids in the domain
are 2 m × 2 m × 2 m and the grid number is 1.87 million
(fig. 3(b)). The total calculation time is 270 s, the parti-
cle time step is Δtp = 0.001 s and the fluid time step is
Δt = 0.01 s.
2.5 Simulation process
According to Shao 1993 [31] and Huang 2012 [30], aerody-
namic entrainment is usually considered to be ignorable
Eur. Phys. J. E (2018) 41: 53
Fig. 4. Comparison between the calculated results and wind
tunnel experimental results of the variations of the sand trans-
port rate with the height.
and the particles entrained directly by air are treated as
an incentive of wind-blown sand movement. In our work,
we introduced a few induced particles with initial verti-
cal speed of 0.1 m/s to trigger the sand flow. Because the
concentration for solid particles in the air is really small,
particle-particle collision in the air was neglected [32]. The
whole computation scheme can be described as follows:
1) Set the logarithmic wind profile as the initial con-
dition for air variables and set the initial position and
velocity of the induced particles.
2) The equations (eqs. (1) to (9)) of gas phase were
calculated to convergence.
3) Inject the induced particles. The wind velocity ob-
tained in step 2 was used to calculate the particle motion
(eqs. (10) to (14)) between air and sand particles.
4) Identify the particles that reach the sand bed and
apply the splash functions (eqs. (15) to (18)). Random
sequences are obtained from these functions and assigned
to vreb and vej . In this way, the rebound and splashed
velocity of particles are introduced.
5) Go to step 2 and calculate the equations of air with
particle-fluid interaction force for the next time steps.
6) End the calculation when the stop time is reached.
Previous works indicate that a limit exists on the
amount of particle that can be splashed and that can fly
as a consequence of the fluid flow transport [33,34]. When
Eur. Phys. J. E (2018) 41: 53 Page 5 of 9

Fig. 5. Simulated wind field at the bridge pier plane (x-z plane of y = 16.3 m).

we simulate air flow with sand particles, a two-way cou-

pling is used for wind-sand interaction, which means the
particle is affected by fluid, and fluid is also affected by
particles [7,30]. As shown in eq. (11), the relative speed
between particles and locate fluid is related to the drag
force that the fluid acts on sand particles. As shown in
eqs. (2), (3), when considering the particles reactive force,
a body force term is added in the momentum equation. In
other words, as the amount of sand particles in the airflow
increases, the reaction force of sand particles to fluid in-
creases, leading to decreased wind speed. Decreased wind
speed leads to a decreased relative speed between the par-
ticles and the local fluid, when they impact the soil, the
speed of impacting particles decreases. As the speed of im-
pacting particles decreases, the amount of splashed parti- Fig. 6. Horizontal velocity profiles over the slope at the x-z
cles and the splashed speed decrease. Finally, the number plane of y = 16.3 m. U is the horizontal velocity, U e is the ex-
of sand particles in the airflow decreases. In this way, the ternal velocity and H is the slope terrain height. The positions
amount of sand particles in the wind-sand flow becomes x = 10 m, 50 m, 70 m are the windward toe, the windward crest
stable. and the leeward toe, respectively.

2.6 Experimental validation of the model 3 Calculation results and discussion

To validate the model for simulation, the calculated sand
flux of steady saltation in the wind-sand flow is compared
with the experimental result with the wind-sand flow at
the same wind speed on a smooth sand bed surface in a
multifunctional environment wind tunnel of Lanzhou Uni-
versity, Gansu Province, China. The incoming wind field
is logarithmic with friction velocity of 0.5 m/s, and the
sand is collected with a sand trap. Several types of sand
traps are widely used in wind tunnel experiments though
sand traps may change the flow field and affect the flux
measurement [35]. The sand trap used in this paper was a
vertical multi-channel sand trap, which consists of a pro-
tection cover, the wedge shape inlet, bracket and collection
box. The average sampling efficiency is 0.91 [36]. Figure 4
shows the simulated sand transport rate versus the height
at 10 m downstream of the inlet, together with the exper-
imental results obtained in the wind tunnel. The good
agreement of both results suggests that the calculation is
accurate.
3.1 Changes in the flow velocities
Figure 5 shows the simulated contour maps of the trans-
verse velocity of the coupled wind field and the vector
diagram in the pier plane (x-z plane of y = 16.3 m). It
clearly shows that the bridge pier separates the reversed
flow region into two parts: a smaller eddy above the lee-
ward slope and a larger eddy downstream of the bridge.
It also shows that the length of the reversed flow region
varies in different cases: case 1 has a 70-m-long reversed
flow region, which is longer than that in case 2 (55 m),
case 3 (50 m) and case 4 (50 m), indicating that the tall
bridge has a greater influence on the reversed flow region.
Because the bridge is low in cases 2 and 3, the length of
the reversed flow region changes little.
Figure 6 shows the horizontal velocity distribution of
the flow field at the x-z plane of y = 16.3 m. The hori-
zontal velocities have obvious differences below 2 H and
Page 6 of 9
Fig. 7. Simulated wind field at the bridge beam plane (x-z plane of y = 0 m).
Fig. 8. Horizontal velocity profiles over the slope terrain at
the x-z plane of y = 0 m. U is the horizontal velocity, U e is the
external velocity and H is the slope terrain height. The posi-
tions x = 10 m, 50 m, 70 m are the windward toe, the windward
crest and the leeward toe, respectively.
approach to the same above 3 H. In case 1, the horizon-
tal velocity near the surface is increased by 25% at the
crest. Above the leeward slope, the reversed flow is 10%
higher than that in case 4. Differently from cases 2, 3 and
4, the horizontal velocity is positive in case 1 below 0.6
H at 80 m. Compared with the other cases, the horizontal
velocity near the surface is reduced by the bridge in case
1 between 90 m and 150 m. At 200 m, the influence of the
bridge on horizontal velocities almost disappears.
Figure 7 shows the simulated contour maps of the
transverse velocities of the coupled wind field and the vec-
tor diagram at the beam plane (x-z plane at y = 0 m). It
shows that the bridge in case 1 is tall enough to enter and
separate the region with accelerated wind speed. The air
flow with high speed is forced downward by the beam and
an eddy is formed between the bridge and the sand dune.
Cases 2 and 3 show little change in the wind field because
the bridge is low and the cross section area of the beam is
small.
Eur. Phys. J. E (2018) 41: 53
Fig. 9. Vertical distribution of horizontal wind velocity in the
x-direction on the bridge.
Figure 8 shows the horizontal velocity distribution of
the flow field at the x-z plane of y = 0 m. In case 1, the
horizontal velocity at the slope crest changes little. Above
the leeward slope, the velocity of the reversed flow near the
dune surface is 52% higher than in case 4. Differently from
cases 2, 3 and 4, the horizontal velocity is positive in case 1
below 0.44 H at 80 m. Downstream of the bridge, the hor-
izontal velocity decreases near the bed between 90 m and
150 m and the influence of the bridge on horizontal veloc-
ities almost disappears at 200 m.
The strong crosswind endangers the running trains,
and fig. 9 shows the relationship between the crosswind
speed and the height relative to the bridge floor. In case
1, the crosswind is larger than in cases 2 and 3 because
the bridge floor is in the accelerated region and the wind
velocity at 2 m above the bridge can reach 20 m/s. The
wind velocity at 2 m above the bridge is 18 m/s in case 3,
hence it is also a strong crosswind. In case 2, because the
bridge is low and located near the core of the reversed flow
region, the wind speed at 2 m above the bridge is 2 m/s,
indicating a weak crosswind over the bridge floor.
Eur. Phys. J. E (2018) 41: 53
Fig. 10. Spatial distribution of sand particles at the bridge pier plane (y = 16.3 m).
Fig. 11. Vertical distribution of sand concentration and horizontal velocity on the leeward slope at x = 60 m.
3.2 Changes in sand transport
In the calculation process, the two-way coupling effect is
considered. Figure 10 shows a snapshot about the distri-
bution of particles at t = 270 s. Affected by the bridge, the
moving pattern of sand particles shows some differences.
In case 1, most sand particles will be gathered and stay
between the leeward slope and the bridge because of the
existence of an eddy in fig. 5; however, few sand parti-
cles can be found on the bridge floor. Since the bridge in
case 2 and case 3 is lower, the sand movement within the
reversed flow region of the dune does not change signifi-
cantly compared with case 4, which has an isolated sand
dune. However, sand deposition is serious on the bridge
floor in cases 2 and 3, especially in case 2, because par-
ticles with lower moving speed suggest that the sand will
deposit on the bridge floor easily.
Figure 11 shows the variations of sand concentration
and horizontal velocity with the height at the middle of
the leeward slope (x = 60 m). Different from the isolated
sand dune in case 4, the concentration and horizontal ve-
locity of sand in case 1 are changed significantly by the
bridge. In case 1, the sand concentration close to the sand
dune surface is 0.02 kg/m3 and approaches 0.04 kg/m3 at
3.5 m, twice as large as in case 4. The sand horizontal
velocity in case 1 is closer to zero, indicating that sand
particles with small kinetic energy are gathered above the
leeward slope. The particles with higher concentration and
Page 7 of 9
lower moving speeds in case 1 form favorable conditions
for sand depositing on the leeward slope. In case 2, the
sand concentration near the surface is 0.003 kg/m3 and
above 3.5 m, the concentration varies little compared with
that in case 4. Moreover, the horizontal velocities of parti-
cles in case 2 change little from case 4; as a result, the sand
depositing condition in case 2 is not changed remarkably
by the bridge.
Dune motion consists of saltating grains climbing up
the windward side, accumulating on the crest and there-
after rolling down the slip face after being deposited at
the bottom on the lee side of the dune [37]. Thus the rate
of sand transport at the crest is an important parameter
of dune motion. Because the gap between the bridge and
the dune is narrow, the sand transport rate and wind field
near the crest are disturbed by the bridge. Figure 12(a)
shows the relationship between the rate of sand transport
and the height at the crest, where solid lines stand for the
windward side and dashed lines stand for the leeward side.
Compared with the isolated dune in case 4, the bridge in-
side the reversed flow region will not change the sand flux
in the windward slope; however, within 0 m to 0.5 m above
the dune surface, the sand flux at the leeward side of the
crest is increased significantly. According to Lima et al.
2017 [38], the threshold friction velocity u∗t is 0.25 m/s
(consists in medium sand). The friction velocity of parti-
cles on the slope needs to be corrected and in this paper,
the corrected threshold shear velocity on the leeward slope
Page 8 of 9
Fig. 12. Sand flux at the crest of sand dune (a) and wall shear stress distribution on the leeward slope (b).
Fig. 13. Sand flux above the bridge.
is 0.32 m/s [39], corresponding to a threshold shear stress
of 0.125 Pa. Figure 12(b) shows the distribution of the wall
shear stress on the leeward slope. The wall shear stress is
lower than the threshold shear stress in cases 2 to 4 and
in case 1 at y = 16.3 m; this indicates that sand move-
ment on the leeward slope surface is impossible in these
cases. However, in case 1 at y = 0 m, the wall shear stress
is higher than the threshold shear stress on the leeward
slope near the toe (x = 63 m to 69 m); this indicates that
affected by the bridge beam, sand particles in this region
will be drifted by the reversed air flow.
Figure 13 shows the relationship between the rate of
sand transport and the height over the bridge in 3 cases.
Case 1 shows a zero rate of sand transport because the
bridge is higher than the sand transport layer. Case 2
shows a low rate of sand transport and the peak value
of 0.005 kg/(m2 s) occurs at 2.5 m, so the sand hazard
is slight. Case 3 shows a high rate of sand transport,
0.03 kg/(m2 s) at 0.3 m, so the sand hazard is serious com-
pared with those in cases 1 and 2. As a result, to avoid
uncontrollable sand hazards, the railway bridge should not
be built on the leeward slope of a dune.
Eur. Phys. J. E (2018) 41: 53
4 Conclusion
In this paper, the Reynolds-averaged Navier-Stokes
(RANS) equations and discrete particle tracing are used
to simulate the wind-sand movement on the leeward slope
with a bridge inside the reversed flow region. The con-
tour map of the transverse velocities in the coupled wind
field and the crosswind on the bridge are obtained, and
the characteristics of sand particle movement are stud-
ied, including the rate of sand transport, mass concentra-
tion and moving speed on the leeward slope and bridge
floor. The result shows that: 1) compared with the iso-
lated dune case, the bridge separates the reversed flow re-
gion downwind the dune into 2 parts; in case 1 the length
of the reversed flow region is increased by the bridge and a
new eddy is formed above the leeward slope; however, the
lengths of the reversed flow region in cases 2 and 3 change
little; 2) because the bridge floor is inside the accelerated
region above the dune crest, the crosswind on the bridges
in cases 1 and 3 is strong; on the contrary, the crosswind
in case 2 is weak because the bridge is inside the reversed
flow region; 3) in case 1, affected by the eddy above the
leeward slope, sand particles with higher concentration
and lower moving speeds of floating are detected and 4)
the rate of sand transport on the bridge floor in case 3 is
higher than those in the other cases.
Engineers concern about sand accumulation and
strong crosswind on the bridge floor. Based on the nu-
merical results in this paper, we put forward some ad-
vices: 1) when a tall bridge is built near the leeward toe
(case 1), anti-crosswind measures should be taken on the
bridge floor; however, because the bridge floor is beyond
the sand transport layer, sand prevention measures are not
required; 2) when a low bridge is built near the leeward
toe (case 2), both the rate of sand transport and sand
velocity are low on the bridge floor, and sand prevention
measures are recommended according to the actual situ-
ation, but anti-crosswind measures are not required and
3) when a bridge is built on the leeward slope (case 3),
large rate of sand transport and strong crosswind occur on
the bridge floor, and measures should be taken to prevent
Eur. Phys. J. E (2018) 41: 53
them both. In engineering applications, sand fence and
windbreak wall are good ways to reduce crosswind and
sand transport. Many works of CFD simulations about
sand fence and windbreak wall have been proposed [38,
40,41]. These works provides good examples and can be
introduced into wind-sand prevention on railways in the
future.
This work is supported by the Scientific Research Special of
Forestry Public Welfare Industry of China (052400001), the
National Natural Science Foundation of China (11772143), and
the Innovative Research Group of the National Natural Science
Foundation of China (11421062).
Author contribution statement
Wei He and Bin Xu designed research and performed the
experiment. Wenbo Wang provided data of terrain and
bridge size. All the authors contributed to the writing of
the manuscript.
References
1. J.P. Zhang, Y.S. Wang, F.Q. Jiang, China Railw. Sci. 32,
14 (2011).
2. K.C. Zhang, J.J. Qu, Q.J. Han, Z.S. An, Sediment. Geol.
273, 91 (2012).
3. N. Zhou, Sci. Geogr. Sin. 32, 1391 (2014).
4. I.J. Walker, Earth Surf. Process. Landf. 24, 437 (2015).
5. A. Frank, K. Gary, Sedimentology 43, 451 (2010).
6. I.J. Walker, W.G. Nickling, Prog. Phys. Geogr. 26, 47
(2002).
7. H. Jiang et al., Geomorphology 283, 41 (2017).
8. D.R. Parsons, G.F.S. Wiggs, I.J. Walker, R.I. Ferguson,
B.G. Garvey, Environ. Model. Softw. 19, 153 (2004).
9. D.R. Parsons, I.J. Walker, G.F.S. Wiggs, Geomorphology
59, 149 (2004).
10. H.J. Herrmann, J.S. Andrade jr., V. Schatz, G. Sauer-
mann, E.J.R. Parteli, Physica A 357, 44 (2005).
11. S. Feng, N. Huang, Environ. Model. Softw. 25, 362 (2010).
12. S.J. Wakes, T. Maegli, K.J. Dickinson, M.J. Hilton, Envi-
ron. Model. Softw. 25, 337 (2010).
Page 9 of 9
13. M. Porcar, P. Caballero, J. Wind Eng. Ind. Aerodyn. 99,
879 (2011).
14. A.D. Araujo, J.S. Andrade Jr., L.P. Maia, H.J. Herrmann,
Sci. Rep. 3, 2858 (2013).
15. M.C. Baddock, G.F.S. Wiggs, Earth Surf. Process. Landf.
36, 1435 (2011).
16. J.A. Palmer, R. Mejia-Alvarez, J.L. Best, K.T. Chris-
tensen, Exp. Fluids 52, 809 (2012).
17. K. Lynch, D.W.T. Jackson, A. Cooper, Geomorphology
105, 139 (2009).
18. K. Lynch, D.W.T. Jackson, A. Cooper, Earth Surf. Pro-
cess. Landf. 35, 344 (2010).
19. G. Sterk, A.F.G. Jacobs, J.H. Van Boxel, Earth Surf. Pro-
cess. Landf. 23, 877 (2015).
20. H. Schonfeldt, S. Von Lowis, Meteorol. Z. 12, 257 (2003).
21. G. Wiggs, C. Weaver, Meteorol. Z. 39, 1 (2012).
22. S. Dupont et al., J. Geophys. Res.: Atmospheres 118, 7109
(2013).
23. B. Launder, D. Spalding, Lectures in Mathematical Models
of Turbulence (Academic Press, 1972).
24. R. Anderson, P. Haff, J. Chem. Inf. Model. 53, 21 (1991).
25. R. Anderson, Proc. R. Soc. Edinburgh 96B, 149 (1989).
26. B. Andreotti, J. Fluid Mech. 510, 47 (2005).
27. D. Beladjine, M. Ammi, L. Oger, A. Valance, Phys. Rev.
E 75, 061305 (2007).
28. J.F. Kok, N.O. Renno, J. Geophys. Res.: Atmospheres
114, D17204 (2009).
29. M.A. Rice et al., Sedimentology. 42, 695 (1995).
30. D. Tong, N. Huang, J. Geophys. Res.: Atmospheres 117,
D16205 (2012).
31. Y. Shao, M.R. Raupach, P.A. Findlater, J. Geophys. Res.
98, 12719 (1993).
32. G.J. Brown, Appl. Math. Model. 26, 155 (2002).
33. K. Lettau, H. Lettau, Exploring the Worlds Driest Climate
(Center for Climatic Research, University of Wisconsin,
Madison, 1978).
34. M. Almeida et al., Phys. Rev. Lett. 96, 018001 (2006).
35. A.D. Araujo et al., Granular Matter 11, 193 (2009).
36. A. Zhao et al., J. Desert Res. 59, 119 (2004).
37. J. Kok et al., Rep. Prog. Phys. 75, 106901 (2012).
38. I. Lima et al., Sci. Rep. 7, 45148 (2017).
39. Y. Shao, Physics and Modelling of Wind Erosion
(Springer, 2000) p. 380.
40. B. Li, D. Sherman, Aeolian Res. 17, 33 (2015).
41. J. Cheng, G. Xin, L. Zhi, F. Jiang, Sci. Rep. 7, 41462
(2017).