Geomorphology 59 (2004) 149 164

www.elsevier.com/locate/geomorph

Numerical modelling of flow structures over idealized transverse

aeolian dunes of varying geometry
Daniel R. Parsons a,*, Ian J. Walker b, Giles F.S. Wiggs c
b

a
School of Earth Sciences, University of Leeds, Woodhouse Lane, Leeds LS2 9JJT, UK
Department of Geography, University of Victoria, P.O. Box 3050, Station CSC, Victoria, British Columbia, Canada V8W3P5
c
Department of Geography, University of Sheffield, Western Bank, Sheffield S10 2TN, UK

Accepted 16 July 2003

Abstract
A Computational Fluid Dynamics (CFD) model (PHOENICSk 3.5) previously validated for wind tunnel measurements is
used to simulate the streamwise and vertical velocity flow fields over idealized transverse dunes of varying height (h) and stoss
slope basal length (L). The model accurately reproduced patterns of: flow deceleration at the dune toe; stoss flow acceleration;
vertical lift in the crest region; lee-side flow separation, re-attachment and reversal; and flow recovery distance. Results indicate
that the flow field over transverse dunes is particularly sensitive to changes in dune height, with an increase in height resulting
in flow deceleration at the toe, streamwise acceleration and vertical lift at the crest, and an increase in the extent of, and strength
of reversed flows within, the lee-side separation cell. In general, the length of the separation zone varied from 3 to 15 h from the
crest and increased over taller, steeper dunes. Similarly, the flow recovery distance ranged from 45 to >75 h and was more
sensitive to changes in dune height. For the range of dune shapes investigated in this study, the differing effects of height and
stoss slope length raise questions regarding the applicability of dune aspect ratio as a parameter for explaining airflow over
transverse dunes. Evidence is also provided to support existing research on: streamline curvature and the maintenance of sand
transport in the toe region; vertical lift in the crest region and its effect on grainfall delivery; relations between the turbulent
shear layer and downward forcing of flow re-attachment; and extended flow recovery distances beyond the separation cell. Field
validation is required to test these findings in natural settings. Future applications of the model will characterize turbulence and
shear stress fields, examine the effects of more complex isolated dune forms and investigate flow over multiple dunes.
D 2003 Elsevier B.V. All rights reserved.
Keywords: Aeolian; Dunes; Computational Fluid Dynamics (CFD); Flow acceleration; Flow separation; Flow reversal; Flow recovery; Aspect
ratio

1. Introduction
The role of secondary flow structures in the morphology, dynamics and spacing of desert sand dunes
* Corresponding author. Tel.: +44-113-343-6624; fax: +44-113343-5259.
E-mail address: parsons@earth.leeds.ac.uk (D.R. Parsons).
0169-555X/$ - see front matter D 2003 Elsevier B.V. All rights reserved.
doi:10.1016/j.geomorph.2003.09.012

has been the focus of much recent research (McKenna

Neuman et al., 1997, 2000; Wiggs, 2001; Walker and
Nickling, 2002) and has been complimentary to similar investigations of bedforms in fluvial environments
(e.g., Nelson et al., 1993; Bennett and Best, 1995).
Building upon earlier work over low-angled hills
(e.g., Jackson and Hunt, 1975; Bowen and Lindley,
1977; Bradley, 1980; Britter et al., 1981; Zeman and

150

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

Jensen, 1987; Raithby et al., 1987), this recent research has greatly improved our understanding of
dune form flow interactions (stoss flow acceleration,
crestal separation, lee re-circulation, re-attachment,
etc.). This progress has been achieved with field
studies of windward flow dynamics (Lancaster et al.,
1996; Frank and Kocurek, 1996a; Wiggs et al., 1996;
McKenna Neuman et al., 2000) and lee-side flow
separation and recovery (Frank and Kocurek, 1996b;
Walker and Nickling, 2002, in press (a,b)). However,
while field and laboratory studies have succeeded in
providing some imprecise relationships between dune
aspect ratio and flow acceleration (e.g., Lancaster,
1994) and have provided detailed empirical relationships characterizing the flow field over transverse
dunes and related these to dune height and lee-side
flow re-attachment and recovery (e.g., Frank and
Kocurek, 1996b; Walker and Nickling, in press (a)),
questions remain as to the presence and the sensitivity
of these secondary flow structures to changes in dune
geometry.
Progress in this regard is hampered by paucity of
additional field and laboratory studies to validate such
relationships and by the relatively small number of
dune geometries investigated. In particular, the complex turbulent structure in the lee side of dunes has
generally precluded the measurement of flow structure
in this region, largely because of limitations in instrumentation (Nickling and McKenna Neuman, 1999;
McKenna Neuman, 2002). Mathematical modelling of
airflow over dunes has provided additional data for
the investigation of secondary flow regimes, but
studies to date have also suffered from an inability
to simulate the highly turbulent flow in the lee of
dunes (e.g., Walmsley et al., 1982; Raithby et al.,
1987; Stam, 1997). For example, Stam (1997) applied
an analytical flow model based on Jackson and Hunts
(1975) boundary layer model, which is unable to
solve the reverse flow lee-side eddy. This limits the
calculation of flow structures to low angle dunes
where lee-side eddies are not present. Stam (1997)
notes that numerical techniques are required successfully to simulate flows over a greater range of dune
forms.
Numerical flow models have been widely applied
in engineering disciplines for many years. In the last
few years, there has been a proliferation of the use of
Computational Fluid Dynamics (CFD) in the fields of

geomorphology and hydrology (see Bates and Lane,

1998). These models provide spatially rich data on
flow field properties that facilitate considerable insight and understanding of the distribution of complex
flow processes. Indeed, these models can provide
details of the flow field that are often difficult to
measure and offer controlled conditions in which
certain aspects of the experimental set up can be
varied rapidly. This paper applies a CFD model to
flow over idealized transverse aeolian dunes and
describes the sensitivity of different elements of the
flow field to variations in geomorphic parameters. The
model used is capable of simulating the highly turbulent reverse flow vortex in the lee of the dune and so is
able to provide an acceptable solution of the downwind distance to flow re-attachment given variations
in dune height, windward slope length and, thus,
aspect ratio.

2. Methods
2.1. Numerical model
This paper employs a numerical model based upon
the PHOENICSk 3.5 code, which is one of several
commercially available CFD programs. The model
solves the elliptic form of the Reynolds-averaged
Navier Stokes equations in two dimensions with a
finite volume method: a cuboidal grid in a Cartesian
frame. The form of the dune was represented within
the model using a relatively new cut-cell porosity
treatment, where the intersections of the inserted
geometry with the grid lines are determined and the
areas and volumes of partially blocked cells are
calculated to a high degree of accuracy (Spalding
and Zhang, 1996; Yang et al., 1997a,b). The equation
formulation is modified to account for the local nonorthogonal intersection of the dune with the grid cells,
resulting in significantly enhanced predictions of nearsurface flow dynamics.
The hybrid-upwind interpolation scheme (Peclet
number = 2) applied in the model is only first order
accurate and can suffer from numerical diffusion
when flow is highly skewed relative to the grid.
Nevertheless, it is more stable than higher-order
schemes, and investigations analogous to this present
one have indicated that errors due to the interpolation

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

151

scheme are not likely to be significant (e.g., Waterson,

1994). The pressure and momentum equations were
coupled through the SIMPLEST algorithm (a variation of SIMPLE; Pantaankar and Spalding, 1972)
where pressure and velocity fields were iteratively
calculated until continuity errors in mass and momentum were adequately small (residuals were < 0.01%
of inlet flux). Turbulence closure was achieved
through application of a two-equation k e model,
modified by renormalisation group theory (Yakhot et
al., 1992). This turbulence model is recommended for
simulating flows with significant mean strain and
shear. For example, it has been shown to perform
better in the prediction of sheared and re-circulating
flows over backward facing steps (e.g., Bradbrook et
al., 1998).
2.2. Model application and assessment
Fig. 1. Simulated incoming plane bed velocity profile.

The model was initially applied to the experimental

set-up of Walker and Nickling (in press (a)) (see
Parsons et al., in press, for full details). Mass flux
values were specified for each grid cell in the upstream inflow, providing an incoming velocity profile
for the model. In order to simulate upwind effects of
the dune (e.g., pressure stagnation and flow deceleration), this profile had to be specified far enough
upstream of the dune. A modelled inflow profile was
specified that implicitly produced the measured plane
bed boundary layer (with a free stream velocity of 13
m s 1) at the point of dune intersection (see Walker
and Nickling, in press (a)). This inflow profile was
used in all the experiments in this paper (Fig. 1). At
the outlet profile, a zero pressure boundary condition
was applied, and, thus, calculated pressure values for
all cells in the domain were defined relative to this.
The length of the simulation domain was 960 cm and
flow depth was 76 cm, with the dune toe positioned at
500 cm into the domain, matching the conditions and
dimensions of the wind tunnel simulation of Walker
and Nickling (in press (a)). In the cell at the fluid
solid interface, it was necessary to prescribe conditions for the velocity and turbulence parameters. For
this purpose, the universal Law of the Wall was
applied in the interface cells. In this experiment,
smooth wall conditions were applied, matching the
roughness experimental set-up of Walker and Nickling (in press (a)).

Full verification and validation of the numerical

model to these experimental conditions and the flow
measurements obtained was demonstrated and discussed by Parsons et al. (in press). Validation was
based on 415 predicted points within the model
domain, which coincided with the locations of measurements taken in the wind tunnel experiment. Although excellent agreement between the measured
and predicted velocities was established, with correlation coefficients for streamwise and vertical velocity
components of 0.97 and 0.83, respectively (Table 1;
Fig. 2), there are significant zones of disagreement,
particularly for the lower velocities (Fig. 2). Parsons et
al. (in press) identified that the majority of these
points are from the lee separation zone where, due
to design limitations, the measuring probe (a TSIR
IFA 300 constant temperature hot-film anemometer)
applied in the wind tunnel experiment (Walker and
Nickling, in press (a)) was unable to resolve the
highly turbulent and negative velocities within this
region. Thus, although the validation process identified some notable differences between the measured
and modelled results, they are primarily due to the
limitations of the measuring instrument rather than
that of the numerical model. In regions where the
instrument is known to perform well, the match is
very good. Indeed, the removal of validation points in
the lee re-circulation zone improves the relationships

152

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

Table 1
Linear regression results between predicted and measured variables
for all validation points (n = 415) and for all points excluding those
within the dune lee separation zone (212)
Variable

b Coefficient

Correlation
coefficient

Streamwise velocity
(all points)
Vertical velocity
(all points)
Streamwise velocity
(excluding separation zone)
Vertical velocity
(excluding separation zone)

1.41

0.98

1.03

0.83

1.29

0.95

1.03

0.88

(Table 1), with a considerable increase in the vertical

velocity correlation, and although there is slight
decline in the streamwise velocity correlation, there
is a significant movement of the regression line
towards that of equality. Furthermore, qualitative
assessment of the predicted flow patterns and the
indications given by flow streamers (see Walker and
Nickling, in press (a)) confirm the presence and the
extent of the separation zone, which is successfully
simulated by the numerical model.
The model is able to simulate areas of flow
stagnation at the toe, acceleration up the stoss slope

Table 2
Geometric properties of Experiments 1 9
Experiment
number

Dune
height
(h)

Stoss
base
length
(L)

Stoss
angle

Aspect
ratio
(h/L)

Lee
base
length

Lee
slope
angle

1
2
3
4
5
6
7
8
9

8.00
8.00
8.00
8.00
8.00
4.00
16.00
6.00
12.00

56.00
112.00
28.00
84.00
42.00
56.00
56.00
56.00
56.00

8.13
4.09
15.95
5.44
10.78
4.09
15.95
6.12
12.10

0.143
0.071
0.286
0.095
0.190
0.071
0.286
0.107
0.214

12.80
12.80
12.80
12.80
12.80
6.40
25.61
9.60
19.20

32.0
32.0
32.0
32.0
32.0
32.0
32.0
32.0
32.0

and flow reversal in the lee, which closely match

measurements obtained in the wind tunnel. The model
is therefore deemed able to provide a realistic and
complete 2D picture of the flow structure over idealized dune forms, providing prediction fields that are
spatially much richer than results produced by current
wind tunnel experiments and field studies. The use of
numerical modelling allows rapid alteration the geometry of the dune under controlled conditions, permitting analysis of the interactive effect of dune form
on the flow field.

Fig. 2. Comparison of modelled to measured velocities (a) streamwise (U) and (b) vertical (V) for the wind tunnel data of Walker and Nickling
(2002) (Experiment 1 in Table 2).

D.R. Parsons et al. / Geomorphology 59 (2004) 149164



2.3. Experiments

Based on the successful verification and validation

of the model (Parsons et al., in press), it was deemed
appropriate to use the model to test the effect of
simple dune geometry variations on streamwise and
vertical velocity flow fields. In particular, certain
elements of the flow field were investigated including:


velocity profiles at the dune toe, dune crest and at

three dune heights downstream of the crest;
 streamwise and vertical velocity profiles at 1 cm
above the dune toe, crest and three dune heights
downstream of the crest;

Fig. 3. Isovel contour plots of streamwise velocity (U, m s

in Table 2).

153

lee-side separation zone length; and

lee-side distance to flow recovery.

Simulated velocity profiles and the streamwise

velocities near the bed are of interest for predicting
the effects of dune aspect ratio on stoss flow acceleration and the strength of flow within the separation
cell, particularly for examining the implications for
sediment transport. Vertical velocities provide insight
on the presence of streamwise curvature effects (i.e.,
flow stabilization) as well as the occurrence and
magnitude of vertical lift or downdrafts over different
dune forms. The last two parameters were of particular interest due to the ability of the model to predict

) calculated for five different dune geometry scenarios (Experiments 1, 2, 5, 6 and 7

154

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

separated and reversed flow in the highly turbulent

lee-side eddy. The lee-side separation zone length was
determined from the location where the simulated
near-surface streamwise velocity at 1 cm changed
from negative (upstream) to positive (downstream)
in orientation. Distance to flow recovery in the lee
was determined as the point at which the simulated
near-surface streamwise velocity at 1 cm above the
surface was within 99% of its unperturbed upwind
value. This point was not necessarily the point where
the full boundary layer had recovered; nevertheless, it
does provide an indication of flow recovery. A fixed
1-cm distance from the surface was applied in each
experiment as the uniform grid resolution used in the
modelling precluded the variable setting this nearsurface distance.
Details of the differing dune geometries that were
used in the model runs are shown in Table 2. All units
are in centimetres and degrees, allowing testing and
comparisons with simulated experiment of the wind
tunnel data of Walker and Nickling (in press (a)).

3. Results
The model output for streamwise velocity (U, m
s 1) is shown as isovel contour plots in Fig. 3. The
results shown here correspond to Experiments 1, 2, 5,
6 and 7 in Table 2 and cover a range of dune height,
stoss length and aspect ratios simulated in this investigation. Flow separation length and distance to flow
recovery results are summarized in Table 3.

Table 3
Distances from dune crest to flow re-attachment and flow recovery
over each of the experimental dune geometries
Experiment
number

Length to flow
re-attachment, cm
(x/h) from crest

Length to flow
recovery, cm
(x/h) from crest

1
2
3
4
5
6
7
8
9

73
59
96
65
82
13
234
34
148

558
526
594
544
568
306
724
424
624

(9.13)
(7.34)
(12.00)
(8.13)
(10.25)
(3.25)
(14.63)
(5.67)
(12.33)

(69.75)
(65.75)
(74.25)
(68.00)
(71.00)
(76.50)
(45.25)
(70.67)
(52.00)

In each experiment, it is clear that the intrusion of

the dune into the simulated boundary layer has a large
effect on flow structure. In each case, the model
predicts flow deceleration immediately upwind of the
dune followed by windward slope acceleration to a
maximum velocity at the crest. These results correspond to findings in previous investigations (e.g.,
Jackson and Hunt, 1975; Bowen and Lindley, 1977;
Lancaster et al., 1996; McKenna Neuman et al., 1997;
Walker and Nickling, 2002; Wiggs, 1993; Wiggs et al.,
1996). In the lee of the dunes, the simulations shown in
Fig. 3 show large disturbances in streamwise velocity
with flow re-attachment occurring within approximately 3 15 dune heights and flow recovery occurring several tens of dune heights downwind (Table 3).
The dune in Experiment 7 was steep-sided, and, here,
near-bed flow recovery occurs just within the boundaries of the simulation. Another interesting observation
is the convergence in the upper (faster) isovels of the
flow field, which corresponds to a zone or jet of
accelerated, overshot flow extending from the crest
above the flow separation cell (Walker and Nickling,
2002). This effect is observable for Experiments 1, 2
and 5 where dune height (h) has been maintained and
is less apparent for Experiments 6 and 7.
Fig. 4 shows the vertical velocity field (V, m s 1)
for the same group of experiments. For each experiment, a zone of positive V exists on the upper stoss
increasing toward a maximum at the crest. This relates
to the topographic (upward) forcing of the dune on
near-surface streamlines. Small pockets of positive V
are also evident on the mid-lee slope indicating
vertical lift in this region.
A zone of strong downward flow delineated by the
0.8 m s 1 isovel exists in the lee extending above
the flow separation region from the base of the lee
slope to beyond the flow re-attachment point. This
zone of downward flow can be seen to shift further
downstream with increases in dune height (Fig. 4) and
appears to vary with the size of the lee-side separation
cell. Indeed, this zone of downdraft aligns closely
above the point of flow re-attachment as was measured in Walker and Nicklings (in press (a)) study.
Interestingly, the height of this zone extends to approximately 5 h for all runs (with the exception of the
steep dune in Experiment 7). These observations
confirm that dune height (and not necessarily aspect
ratio) plays an important role in perturbing the pres-

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

Fig. 4. Isovel contour plots of vertical velocity (V, m s

Table 2).

155

) calculated for five different dune geometry scenarios (Experiments 1, 2, 5, 6 and 7 in

sure filed over the dune and influencing the vertical

velocity distributions and flow re-attachment (Walker
and Nickling, in press (a)).
The sensitivity of flow patterns to changing dune
geometry are highlighted in more detail in Figs. 5 and
6 and in Figs. 7 12. Full velocity profiles at the dune
toe, crest and lee with changing dune stoss slope
length and changing dune height are provided in Figs.
5 and 6, respectively. Figs. 7 10 identify changes in
the near-bed velocity components with changes in the
dune stoss length and the dune height. The variability
of separation re-attachment length and distance to
flow recovery with changing dune geometry are
detailed in Figs. 11 and 12.

Figs. 6 and 7 indicate that both deceleration at the

toe and acceleration at the crest are sensitive to
changes in dune height while maintaining dune stoss
slope length (i.e., as dune stoss angle increases). Both
figures indicate that increasing dune height appears to
have a greater impact on acceleration at the crest than
on deceleration at the toe. Furthermore, near-bed flow
velocities at the toe decelerate almost linearly with
increasing dune height as near-bed acceleration at the
crest follows a power function (Fig. 7). Streamwise
velocity in the lee of the dune decreases rapidly with
increasing dune height, before becoming negative as
the separated lee-side eddy reverses flow at the foot
of the lee slope (Figs. 6 and 7). The magnitude of the

156

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

Fig. 5. Velocity profiles at the dune toe, crest and lee for changing dune stoss slope lengths (Experiments 1, 6, 7, 8 and 9 in Table 2).

lee-side sheltering effect near the bed increases dramatically with dune height for the range of dune sizes
simulated here (Fig. 7). When the flow is reversed,
increasing dune height has the effect of increasing the
velocity of near-surface reversed flow at 3 h downstream of the crest in the separation cell, although this
increase appears rather minor.
The effect of changing stoss slope basal length (L)
while maintaining dune height is shown in Figs. 5 and
8. Increasing stoss slope length appears to have a
negligible impact upon streamwise velocities at the
crest, although minor effects are clear for velocities at
the toe and in the lee (i.e., near-bed lee-side velocities
become less negative). Such results are expected given
that stoss slope angle is less sensitive to a change in
stoss slope length than a change in dune. A steepening
of this windward angle leads to both an increase in
flow acceleration at the dune crest and flow deceleration in the upwind toe region due to increased

streamline compression and flow stagnation effects

respectively (Wiggs et al., 1996).
Figs. 9 and 10 are similar to Figs. 7 and 8, except
that they focus on vertical velocity at the crest, toe and
lee side of each of the experimental dune geometries.
Dune height is shown to have a significant impact on
vertical velocity in the crestal regions of the dunes due
to topographic forcing and a small, but significant,
effect on vertical velocities at the toe (Fig. 9). This
increase in V in the toe region with increasing dune
height provides some support for the streamline curvature model of Wiggs et al. (1996). Sediment transport is maintained through this toe region, despite a
reduction in time-averaged streamwise flow velocity,
as a consequence of concave streamline curvature
resulting in increased turbulence intensity and Reynolds stresses (Wiggs et al., 1996). The small increase
in vertical velocity with increasing dune height shown
in these experiments (Fig. 9) corresponds to an

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

Fig. 6. Velocity profiles at the dune toe, crest and lee for changing dune heights (Experiments 1, 6, 7, 8 and 9 in Table 2).

Fig. 7. Streamwise velocity as a function of dune height.

157

158

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

Fig. 8. Streamwise velocity as a function of dune stoss slope length.

increase in streamline angle from 2.5j to 8.0j in the

toe region.
As expected, dune height appears to have only a
small effect on vertical velocities close to the surface
in the lee (at 3 h downstream of the crest) (Fig. 9). At
low dune heights (4 and 6 cm), the negative vertical
velocities indicate that downward flow occurs closer
to the form as flow reattaches (Fig. 4). At dune heights
above 8 cm, there is no evidence of a vertical velocity
component in the flow structure, with the zone of

negative vertical velocity shifting downstream. This is

attributed to re-attachment lengths being much greater
at higher dune heights (Fig. 11) than the measurement
point 3 h downstream of the crest, and, hence, nearsurface flow at 3 h is dominated by the reversed
streamwise component in larger dunes.
Fig. 10 shows that vertical velocities decline rapidly with increasing stoss length. It appears that
changes in stoss slope length have a greater impact
on vertical velocities at the crest than on streamwise

Fig. 9. Vertical velocity as a function of dune height.

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

159

Fig. 10. Vertical velocity as a function of dune stoss slope length.

velocities (Fig. 8). This is expected given that changes

in stoss slope length have an immediate impact on
streamline angles at the dune crest. In addition, shorter
stoss slope lengths result in steeper windward slopes,
and, hence, higher vertical velocities at the crest (Fig.
10). This appears to have a lesser effect at the dune toe
where V decreases only slightly with increasing stoss
slope length. Similar to Fig. 9, the results in Fig. 10
suggest that vertical velocity at the bed in the separation cell is independent of dune geometry.

The length of the flow separation zone with

changing dune aspect ratio is shown in Fig. 11.
In general, flow re-attaches within 3 15 dune
heights downwind (Table 3), which fits within
previously documented estimates of 4 10 h (Frank
and Kocurek, 1996b; Walker and Nickling, 2002).
These data confirm Walker and Nicklings (2002)
suggestion that an increase in dune height (i.e., an
increase in aspect ratio) causes a corresponding
increase in the extent of the lee-side separation cell

Fig. 11. Separation zone length (cm) as a function of dune aspect ratio.

160

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

(i.e., flow re-attachment occurs further downwind

from the dune crest). The wide range in values
from 3 < x/h < 15 (i.e., 3 15 dune heights downwind from the crest) indicates the sensitivity of the
structure of the lee-side eddy to this geometrical
parameter. The data also show that a similar, though
less steep, relation occurs if dune height is maintained at a constant but stoss slope length is
decreased (i.e., also an increase in aspect ratio).
The different gradients of the relations evident in
Fig. 11 for changing dune height and changing
stoss length indicate that the length of the separation zone is more sensitive to the former.
Similarly, Fig. 12 shows the effect of dune aspect
ratio on the downwind distance to flow recovery in
the lee side of the dune. In general, streamwise
velocities at the surface do not recover to 99% of
upwind values until approximately 52 77 dune
heights downwind. However, even at these distances,
only the near-surface velocity values have recovered,
with the full boundary layer profile often still recovering. Although actual recovery distances increase
with aspect ratio, interestingly, the shortest height
normalised recovery distances occur over the tallest
dunes with the steepest stoss slope angles (Experiments 7 and 9, Table 3). These flow recovery lengths
exceed Lancasters (1988) estimate of 10 15 h and
Walker and Nicklings (in press (b)) estimate of 25
30 h, and they are closer to distances of 30 50 h for

flow over a backward-facing step and sub-aqueous

dunes (Bradshaw and Wong, 1972; McLean and
Smith, 1986, respectively). The relationship shown
(Fig. 12) for increasing dune height in the aspect ratio
demonstrates a power function increase in recovery
distance. Similarly, a decrease in stoss slope length
(resulting in an increase in the aspect ratio) increases
recovery distance in a linear relation that is less steep
than that for changing dune height (Fig. 12).
The results in Figs. 11 and 12 are interesting in that
while variations in dune height and stoss slope length
both influence the aspect ratio of the dune, they have
differing impacts on the downwind distance to flow
recovery and the separation zone length. In both
cases, the airflow structure is more sensitive to
changes in dune height than changes in stoss slope
length.

4. Discussion
To date, logistical and instrumentation limitations
have prevented effective characterization of flow field
response over transverse dunes of varying geometry.
Only recently have CFD models become available
and validated for use in simulating the flow field over
isolated transverse dunes (see Parsons et al., in press).
It is clear from this simulation and other previous
wind tunnel and field studies that variations in stream-

Fig. 12. Distance to flow recovery (cm) as a function of dune aspect ratio.

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

wise and vertical velocities over transverse dunes are

the outcome of the intrusion of the dune into the
atmospheric boundary layer. The result is a perturbation in the near-surface pressure field resulting in
fluid momentum changes that, in turn, cause secondary flow effects such as: flow stagnation and deceleration at the upwind toe; streamline compression,
flow acceleration and vertical lift up the windward
slope; and, in the lee, flow separation, streamline
expansion, flow re-attachment and reversal, a zone
of downward vertical flow and a lengthy flow recovery distance. These secondary flow effects are shown
in this study to vary significantly with dune geometrynamely dune height, stoss basal length and, thus,
aspect ratio.
4.1. Streamwise velocity distribution
The CFD model used in this study reliably predicts
streamwise flow deceleration immediately upwind of
the dune followed by windward slope acceleration to
a maximum velocity at the crest. At the dune toe,
velocities increase only slightly with stoss slope
length (i.e., for less steep dunes) and decrease only
slightly with increases in dune height for the range of
forms investigated. This is likely the result of a
reduced stagnation effect imposed on the flow by
dune with lesser aspect ratios (i.e., less steep dunes)
(Wiggs et al., 1996). Toward the crest, flow acceleration is more sensitive to increases in dune height than
to stoss slope length. This is primarily due to stoss
slope angle being more sensitive to changes in dune
height than basal length and thus dune height exerting
a greater perturbation on the windward flow field,
which results in enhanced streamline convergence
and, hence, flow acceleration over taller dunes.
The model also characterizes lee-side streamwise
velocity variations very effectively (Parsons et al., in
press). Reversed near-surface velocities inside the
separation zone increase slightly (i.e., become less
negative) with increasing stoss slope length (Figs. 5
and 8) but are highly sensitive to changes in dune
height (Figs. 6 and 7). This difference is highlighted
through comparison of the lee profiles in Figs. 5 and
6. Taller dunes having a larger lee-side separation
cell and stronger reversed flow near the surface can
be explained for two main reasons. First, stoss flow
is accelerated more toward, and overshot faster from,

161

the crest of taller, steeper dunes. Second, the lee-side

velocity gradient in the separation zone is steeper,
and as a result, momentum exchange and resultant
recycling of fluid mass back toward the dune is
greater (Walker and Nickling, in press (a)). Therefore, for the range of dune forms investigated in
this study, taller dunes with greater aspect ratios (i.e.,
h/L>0.14) have larger lee-side flow separation
regions (Fig. 11) and stronger near-surface reversed
flows (Figs. 5 and 6).
Dune height also has a greater effect than stoss
slope length on streamwise flow velocity recovery
distance (Fig. 12). Interestingly, the shortest normalised recovery distances occur over the tallest dunes
with the steepest stoss slope angles (Experiments 7
and 9, Table 3) while longer distances (up to 76.5 h)
are required for shorter dunes. This is mainly due to
the effects normalising for height (Table 3) and Fig.
12 demonstrates the effect of h and L and, thus, aspect
ratio on the actual recovery distances. The power law
function with increases in dune height shows that
recovery distance increases begin to diminish with
larger heights (Experiments 1 and 9). This may relate
to enhanced turbulent momentum exchange and dissipation in the larger flow separation region in the lee
of steeper obstacles, thereby reducing the distance for
boundary layer recovery. It may also be due to
increases in negative vertical velocity magnitudes in
the lee forcing higher streamwise velocity towards the
surface and producing recovery sooner. This will be
explored further in a related paper.
4.2. Vertical velocity distribution
This study shows that vertical velocity variations
on the windward face are most affected by dune
height and stoss slope length (i.e., increasing stoss
slope angle), especially at the crest. This confirms
the widely held view that dunes with steeper windward slopes experience greater topographic forcing,
streamline convergence and flow acceleration toward
the crest (Lancaster, 1985; Tsoar, 1985; Wiggs,
1993; Lancaster et al., 1996). It is shown here that
as dune height increases, flow is overshot with
greater vertical velocity (upward lift) from the crest
(Figs. 3 and 9).
In terms of sand transport at the dune toe, the
decrease in streamwise velocity (and hence sand

162

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

transport) with increasing dune height (Fig. 7) may be

offset by an increase in streamline angle (controlled
by stoss slope length) and a rise in vertical velocity
(Figs. 9 and 10). This, in turn, may result in higher
turbulence intensities, Reynolds stresses and vertical
lift in the toe region that might be sufficient to
maintain sand transport in this region, which supports
recent wind tunnel and field studies by McKenna
Neuman et al. (2000) and Walker and Nickling (in
press (a,b)) and the streamline curvature model of
Wiggs et al. (1996).
In the lee, near-surface vertical velocities are only
slightly sensitive to changes in dune height for the
range of forms simulated. The effect is greatest for
shorter dunes (i.e., h < 8 cm) where the 1-cm measurement height 3 h downstream of the crest at in the
dune lee shows greater downward velocity as it is
closer to the upper boundary of the smaller separation
cell (Figs. 4 and 10). This may also reflect a small
pocket of vertical lift (i.e., positive V) evident over
the mid lee slope for most runs (Fig. 4). This occurs
due to a steep favourable (negative) pressure gradient
that causes slower lee flow to rise toward the faster
flow in the shear layer bounding the separation cell
(Walker and Nickling, in press (a)). This phenomenon
is found in flow over roofs (Ginger and Letchford,
1993) and over low-angle fluvial dunes (Best and
Kostaschuk, 2002). Walker and Nickling (in press (a))
suggest that this effect (combined with slope effects
and impact from fallout grains, though slight) is able
to reduce transport thresholds in the lee slope region.
They also indicated that vertical updrafts in the upper
lee enhance modified suspension of grains into the lee
to distances well beyond typical saltation trajectories
(cf. Nickling et al., 2002). This reinforces the thought
that saltation is not likely the dominant mechanism
for sediment delivery into the lee and that secondary
lee-side airflow patterns have a significant effect on
dune sedimentary dynamics (Walker and Nickling, in
press (a)).
The flow field simulation (Fig. 4) shows that
beyond the lift region immediately leeward of the
crest, flow becomes downward (i.e., vertical velocities
become negative) beyond the lee slope. The progressive shift from stoss-upward to lee-downward motion
reflects a wave-like, dune-generated perturbation in
the flow field and confirms measured patterns over
sub-aqueous dunes (Best and Kostaschuk, 2002) and

over an idealized aeolian form (Walker and Nickling,

in press (a)). This wave-like influence of dune form
on vertical velocity extends to a height of approximately 3 5 h and the shift to downward flow translates further downwind over the lee-slope as dune
height increases (Fig. 4). In general, the extent of the
downward flow zone appears to be relatively independent of stoss slope length and, hence, aspect ratio.
It extends from the crest (for shorter dunes) to tens of
dune heights downwind (e.g., >14 h for the taller dune
in Experiment 7) and the flow re-attachment point is
found near the downwind edge of the faster core of is
zone. Though this paper does not characterize turbulence, this finding corroborates Walker and Nicklings
(in press (a)) claim that downward flow from a
turbulent shear zone (their zone G) overlying the
separation region drives flow re-attachment at the
surface. This study also shows that, like the separation
cell, the extent of this zone is more dependent on dune
height and not necessarily explained the aspect ratio
alone. Therefore, the influence of dune form on
vertical velocity plays an important role in sediment
delivery into the lee via grainfall in separated airflows
(Nickling et al., 2002) as well as in determining the
point of flow re-attachment, boundary layer recovery
and subsequent saltation development distance (Walker and Nickling, in press (a,b)).

5. Conclusion
Analysis of CFD-derived flow structures over
idealized transverse dunes has shown the potential
to quickly and reliably test relations between dune
geometry and wind flow structure. Data confirm that
the flow field over transverse dunes is particularly
sensitive to changes in dune height, with an increase
in height resulting in flow deceleration at the toe,
acceleration at the crest and an increase in the size of
the lee-side separation zone. Evidence is provided to
support the streamline curvature model of Wiggs et al.
(1996) that explains the maintenance of sand transport
in the toe region of the dune despite declining streamwise velocities. This study also confirms patterns of
vertical lift in the crest region and downward flow
beyond the lee slope documented by Walker and
Nickling (in press (a)), which respectively have influence on lee-side sediment delivery via grainfall (Nick-

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

ling et al., 2002) and flow re-attachment, boundary

layer recovery and the re-development of saltation at
distances tens of dune heights downwind of the dune
(Walker and Nickling, in press (b)). The structure of
lee-side airflow has been shown to be very sensitive to
changes in dune height (h) and stoss slope basal
length (L). The length of the lee-side separation zone
varied from approximately 3 15 h downwind, increasing with height or shorter stoss slope lengths.
Similarly, the downwind distance to flow recovery
ranged from 45 to >75 dune heights downwind with
changes in dune height.
The differing effects of height and stoss slope
length in these experiments raises questions regarding
the applicability of dune aspect ratio as a parameter
for explaining airflow over transverse dunes. Though
changes in either L or h can produce the same aspect
ratio (h/L), this study shows that dune height has a
greater effect on streamwise flow perturbations (e.g.,
flow deceleration at the dune toe, flow acceleration at
the dune crest, reversed lee-side surface flow) and on
lee-side flow structure (e.g., separation cell length,
flow recovery distance). In general, steeper (and not
longer) dunes of the same aspect ratio have a greater
effect on the flow field, particularly in the lee, for the
range of dune shapes investigated in this study. This is
because the important factor is the size of the dune in
relation to the boundary layer rather than the actual
shape of the dune, which is described by the aspect
ratio.
Further experimentation using CFD for calculating
airflow over transverse dunes is planned and extended
analyses will include shear stress development and
turbulent momentum exchange on simple triangular
dune profiles before extending experimentation to
more complex and realistic dune geometries consisting of concave convex-shaped windward slopes and
multi-dune geometries.

Acknowledgements
This work was undertaken while Daniel Parsons
was in receipt of a NERC studentship GR16/99/FS/2
with additional financial support from the British
Geomorphological Research Group to attend the
International Conference on Aeolian Research 5 at
Lubbock, TX.

163

References
Bates, P.D., Lane, S.N. (Eds.), 1998. High Resolution Flow Modelling. Hydrological Processes 12, 1129 1396 (Special issue).
Bennett, S.J., Best, J.L., 1995. Mean flow and turbulence structure
over fixed, two-dimensional dunes: implications for sediment
transport and bedform stability. Sedimentology 42, 491 513.
Best, J.L., Kostaschuk, R.A., 2002. An experimental study of
turbulent flow over a low-angle dune. Journal of Geophysical
Research 107, 3135 3154.
Bowen, A.J., Lindley, D., 1977. A wind tunnel investigation of the
wind speed and turbulence characteristics close to the ground
over various escarpment shapes. Boundary - Layer Meteorology
12, 259 271.
Bradbrook, K.F., Biron, P., Lane, S.N., Richards, K.S., Roy, A.G.,
1998. Investigation of controls on secondary circulation and
mixing processes in a simple confluence geometry using a
three-dimensional numerical model. Hydrological Processes
12, 1371 1396.
Bradley, F., 1980. An experimental study of the profiles of wind
speed, shearing stress and turbulence at the crest of a large
hill. Quarterly Journal of the Royal Meteorological Society
106, 101.
Bradshaw, P., Wong, F.Y.F., 1972. The reattachment and relaxation of
a turbulent shear layer. Journal of Fluid Mechanics 52, 113 135.
Britter, R.E., Hunt, J.C.R., Richards, K.J., 1981. Analysis and wind
tunnel studies of speed up, roughness effects and turbulence
over a two-dimensional hill. Quarterly Journal of the Royal
Meteorological Society 107, 91 110.
Frank, A., Kocurek, G., 1996a. Airflow up the stoss slope of sand
dunes: limitations of current understanding. Geomorphology 17,
47 54.
Frank, A., Kocurek, G., 1996b. Toward a model of airflow on the
lee side of aeolian dunes. Sedimentology 43, 451 458.
Ginger, J.D., Letchford, C.W., 1993. Characteristics of large pressures in regions of flow separation. Journal of Wind Engineering
and Industrial Aerodynamics 49, 301 310.
Jackson, P.S., Hunt, J.C.R., 1975. Turbulent wind flow over a low
hill. Quarterly Journal of the Royal Meteorological Society 101,
929 955.
Lancaster, N., 1985. Variations in wind velocity and sand transport
rates on the windward flanks of desert sand dunes. Sedimentology 32, 581 593.
Lancaster, N., 1988. Controls of aeolian dune size and spacing.
Geology 16, 972 975.
Lancaster, N., 1994. Dune morphology and dynamics. In: Abrahams, A.D., Parsons, A.J. (Eds.), Geomorphology of Desert
Environments. Chapman & Hall, London, pp. 474 505.
Lancaster, N., Nickling, W.G., McKenna Neuman, C., Wyatt, V.E.,
1996. Sediment flux and airflow on the stoss slope of a barchan
dune. Geomorphology 17, 55 62.
McKenna Neuman, C., 2002. The role of instrumentation in aeolian
research: recent advances and future challenges. Joint Meeting
of the International Conference on Aeolian Research (ICAR5)
and The Global Change and Terrestrial Ecosystem Soil Erosion
Network (GCTE-SEN), Texas Tech University, Lubbock, Texas,
22 25 July 2002.

164

D.R. Parsons et al. / Geomorphology 59 (2004) 149164

McKenna Neuman, C., Lancaster, N., Nickling, W.G., 1997. Relations between dune morphology, air flow, and sediment flux on
reversing dunes, Silver Peak, Nevada. Sedimentology 44,
1103 1113.
McKenna Neuman, C., Lancaster, N., Nickling, W.G., 2000. The
effect of unsteady winds on sediment transport on the stoss
slope of a transverse dune, Silver Peak, NV, USA. Sedimentology 47, 211 226.
McLean, S.R., Smith, J.D., 1986. A model for flow over twodimensional bed forms. Journal of Hydraulic Engineering 112,
300 317.
Nelson, J.M., McLean, S.R., Wolfe, S.R., 1993. Mean flow and
turbulence fields over two-dimensional bedforms. Water Resources Research 29, 3935 3953.
Nickling, W.G., McKenna Neuman, C., 1999. Recent investigations
of airflow and sediment transport over desert dunes. In: Goudie,
A.S., Livingstone, I., Stokes, S. (Eds.), Aeolian Environments,
Sediments and Landforms. Wiley, Chichester, pp. 15 47.
Nickling, W.G., McKenna Neuman, C., Lancaster, N., 2002. Grainfall processes in the lee of transverse dunes, Silver Peak, Nevada. Sedimentology 49, 191 209.
Pantaankar, S.V., Spalding, D.B., 1972. A calculation procedure for
heat, mass and momentum transport in three-dimensional parabolic flows. International Journal of Heat and Mass Transfer 15,
1782 1799.
Parsons, D.R., Wiggs, G.F.S., Walker, I.J., Garvey, B.G., in press.
Time-averaged numerical modelling of airflow over an idealised
transverse dune.
Raithby, G.D., Stubley, G.D., Taylor, P.A., 1987. The Askervein
hill Project: a finite control volume prediction of three dimensional flow over the hill. Boundary - Layer Meteorology 39,
247 267.
Spalding, D.B., Zhang, Q., 1996. ASAPsimulating flows with a
Cartesian mesh. Proceedings of the PHOENICS User Conference, Tokyo, 1996.
Stam, J.M.T., 1997. On the modelling of two-dimensional aeolian
dunes. Sedimentology 44, 127 141.
Tsoar, H., 1985. Profile analysis of sand dunes and their steady state
significance. Geografiska Annaler 67A, 47 59.

Walker, I.J., Nickling, W.G., 2002. Dynamics of secondary airflow

and sediment transport over and in the lee of transverse dunes.
Progress in Physical Geography 26, 47 75.
Walker, I.J., Nickling, W.G., in press (a). Mean flow and turbulence
characteristics of simulated airflow over isolated and closely
spaced transverse dunes. Sedimentology.
Walker, I.J., Nickling, W.G., in press (b). Simulation and measurement of surface shear stress over isolated and closely spaced
transverse dunes. Earth Surface Processes and Landforms.
Walmsley, J.L., Salmon, J.R., Taylor, P.A., 1982. On the application
of a model of boundary-layer flow over low hills to real terrain.
Boundary - Layer Meteorology 23, 17 46.
Waterson, N.P., 1994. Validation of convection discretisation
schemes. VKI Dip Rep. 1994-33. Von Karmen Institute for
Fluid Dynamics, Rhode-Saint-Genese, Belgium, pp. 1 177.
Wiggs, G.F.S., 1993. Desert dune dynamics and evaluation of
wind velocity: an integrated approach. In: Pye, K. (Ed.), The
Dynmics and Environmental Context of Aeolian Sedimentary
Systems. Geological Society Special Publication, vol. 72,
pp. 37 46.
Wiggs, G.F.S., 2001. Desert dune processes and dynamics. Progress
in Physical Geography 25, 53 79.
Wiggs, G.F.S., Livingstone, I., Warren, A., 1996. The role of
streamline curvature in sand dune dynamics: evidence from field
and wind tunnel measurements. Geomorphology 17, 29 46.
Yakhot, V., Orszag, S.A., Thangam, S., Gatshi, T.B., Speziale, C.G.,
1992. Development of a turbulence model for shear flow by a
double expansion technique. Physics of Fluids. A, Fluid Dynamics 4, 1510 1520.
Yang, G., Causon, D.M., Ingram, D.M., 1997a. Calculation of 3-D
compressible flows around moving bodies. 21st International
Symposium on Shock Waves, Australia, July 20 25.
Yang, G., Causon, D.M., Ingram, D.M., Saunders, R., Batten, P.,
1997b. A Cartesian cut-cell method for compressible problems:
Part A. Static-body problems; Part B. Moving-body problems.
Aero Journal of the Royal Aeronautical Society, 47 65.
Zeman, O., Jensen, N.O., 1987. Modification of turbulence characteristics in flow over hills. Quarterly Journal of the Royal Meteorological Society 113, 55 80.