You are on page 1of 12

Majid Jandaghi Alaee Greg Ivey

Charitha Pattiaratchi
Secondary circulation induced by ow curvature and Coriolis effects
around headlands and islands
Received: 2 February 2003 / Accepted: 9 July 2003
Springer-Verlag 2004
Abstract Previous studies have shown that ow curva-
ture in river bends generates a secondary circulation in
the plane normal to the mean ow direction. A similar
circulation pattern is shown to exist in oceanic situations
when ows are subject to curvature, mainly due to
interaction with topographic features. However, it is
shown that, due to dierences between oceanic condi-
tions and river bends, theory and prediction methods
based on the assumptions for river bends are invalid for
oceanic ows. Via scaling arguments based on the
equations of motion, that include both the eects of ow
curvature and the Coriolis force, parameters that govern
the dierent ow regimes are identied. The maximum
strength of the secondary ow is derived for each ow
regime and is veried using a three-dimensional (3-D)
numerical model applied to an idealized island. It is also
shown that upwelling, due to the generation of second-
ary ow, occurs o the tips of the headland or island,
and its inuence can extend far downstream.
Keywords Secondary circulation Flow curvature
Coriolis eects Headlands Islands
1 Introduction
Studies on the ow structure in river bends have con-
rmed that vertically sheared, curved ows produce
secondary circulation in the plane normal to the mean
ow direction (e.g. Rosovski 1957; Thorne and Hey
1979; de Vriend 1981). The curvature-induced secondary
circulation is the result of a local imbalance between the
vertically varying centrifugal force and the cross-stream
pressure gradient (e.g. Kalkwijk and Booij 1986,
henceforth KB). This force imbalance always results in
ow towards the inside of the bend in the lower part of
the water column and an outward ow near the surface.
In large-scale ows where rotation is important, any
imbalance between the cross-stream pressure gradient
and the Coriolis acceleration can also result in a sec-
ondary ow. The orientation of this type of secondary
circulation associated with the Coriolis acceleration
depends on the direction of the ow.
KB proposed an analytical model to investigate the
secondary ow for the case of steady, constant curvature
ows. Dealing with the transverse momentum equation
in the curvilinear system, KB suggested a simple balance
between transverse friction and deviation of the driving
force (curvature and/or Coriolis) from its mean. Taking
a logarithmic prole for the ow velocity over the depth,
model predictions of transverse velocity showed good
agreement with experimental data from curved umes
with shallow rectangular cross-sections.
Garrett and Loucks (1976) suggested that oceanic
curving ows around headlands may also result in the
generation of secondary circulation. Tidal current data
in the vicinity of Gay Head, USA, reported by Geyer
(1993), clearly showed the presence of the secondary
circulation. The observed transverse secondary ow of
0.13 m s
1
(corresponding to 1520% of the streamwise
ow) was much stronger than the 0.03 m s
1
, computed
by the model of KB. This result revealed what could be
a signicant inuence of the secondary ow on the ow
dynamics around headlands. Geyer (1993) suggested
that two factors were most likely responsible for the
discrepancy: (1) enhanced shear due to the dierence
between the actual velocity prole and the logarithmic
prole in the streamwise ow; and, (2) reduced vertical
mixing caused by stratication. However, for oscilla-
tory oceanic curving ows such as tidal ows around
headlands, the time-dependent term and, more impor-
tantly, the varying characteristics of the ow in the
streamwise direction, neglected in KB, may have con-
tributed to the discrepancy. The observations by Geyer
(1993) indicate that the model by KB may not be
Ocean Dynamics (2004) 54: 2738
DOI 10.1007/s10236-003-0058-3
Responsible Editor: Richard Signell
M.J. Alaee G. Ivey C. Pattiaratchi (&)
Centre for Water Research,
The University of Western Australia,
35, Stirling Highway, Crawley 6009, Australia
e-mail: pattiara@cwr.uwa.edu.au
applicable to the oceanic conditions such as Gay Head
and similar sites.
Garrett and Loucks (1976) suggested that secondary
ow at headlands may be suciently strong to induce
upwelling, the oshore owin the upper part of the water
column being replaced locally by onshore ow in the
bottom boundary layer. So far, only one mechanism has
been proposed for upwelling in island or headland wakes:
Ekman pumping (Greenspan 1968: Pedlosky 1979),
causing upwelling in the centre of island wake eddies
(Wolanski et al. 1984, 1994, 1996; Deleersnijder et al.
1993). However, there have been observations of
upwelling o headlands and islands in the absence of an
eddy (e.g. St John and Pond 1992). Alaee et al. (1998)
reported upwelling in the wake of Rottnest Island, where
an eddy was unlikely to be generated in the wake region.
Thus, upwelling as a result of the secondary circulation
may be responsible for the observed vertical water mo-
tion in wakes in the absence of an eddy.
This paper focuses on the generation of secondary
circulation due to ow curvature with the aim of
developing a prediction method to estimate the strength
of the secondary circulation. Dynamics of secondary
ows and the models utilized in the previous studies are
reviewed in Section 2. In this section, two non-dimen-
sional numbers are identied and a classication scheme
in the form of a ow regime diagram is suggested. The
transverse velocities are predicted for each of the ow
regimes using simple scaling arguments. The 3-D
numerical model (HAMSOM) is described in Section 3
and the remainder of the paper presents the results of
numerical simulations over a wide parameter range in
order to compare the results with prediction methods.
Finally, the applicability of the models and the occur-
rence of upwelling are discussed in Section 5.
2 The dynamics of secondary circulation
A curvilinear coordinate system s, n and z is considered
(Fig. 1) with the streamwise coordinate s oriented in the
direction of vertically averaged ow and the cross-
stream coordinate n oriented normal to the ow. Fol-
lowing KB, the momentum equation in the n direction
can be written as
@u
n
@t

@u
s
u
n
@s

@u
2
n
@n

@u
z
u
n
@z
2
u
s
u
n
R
n

u
2
n
u
2
s
R
s
fu
s

1
q
@p
@n
friction terms = 0 ; (1)
where R
s
and R
n
are the radii of curvature of the
streamlines of the depth-averaged ow in the streamwise
and normal directions, respectively, p is the pressure, q
the mass density and f the Coriolis parameter. Assuming
u
n
u
s
, uniform density, neglecting friction in vertical
planes and vertical advection Eq. (1) can be simplied,
using the hydrostatic pressure assumption and friction
terms parameterized in terms of an eddy viscosity:
@u
n
@t
u
s
@u
n
@s

u
2
s
R
s
g
@g
@n
fu
s

@
@z
K
z
@u
n
@z

= 0 : (2)
Here, g is the water level, g is the acceleration due to
gravity and K
z
the vertical eddy viscosity. By denition,
the depth-averaged transverse velocity, u
n
, is zero. KB
averaged Eq. (2) over depth and obtained:
u
s
@u
n
@s

u
2
s
R
s
g
@g
@n
f u
s

s
n
qh
= 0 ; (3)
where s
n
is the bottom friction in the n direction and h is
the water depth. Eliminating the free surface slope term
by combining Eqs. (2) and (3), KB found an expression
for the secondary ow as:
@u
n
@t
u
s
@u
n
@s
u
s
@u
n
@s

@
@z
K
z
@u
n
@z

s
n
qh
=
u
2
s
u
2
s
R
s
f (u
s
u
s
) : (4)
The right-hand side of Eq. (4) identies two distinct
mechanisms leading to transverse motion: ow curva-
ture and the Coriolis eects due to the deviation of u
s
from the vertical mean, in both cases.
To obtain an analytical solution for steady ows in
curved open channels, KB ignored all streamwise vari-
ations and assumed a logarithmic prole for u
s
as well as
a parabolic form for the vertical eddy viscosity. Under
these assumptions Eq. (4) simplies to:

@
@z
K
z
@u
n
@z

s
n
qh
= driving force ; (5)
which was solved analytically by KB. The results were in
good agreement with the laboratory measurements of
constant curvature channels with uniform rectangular
cross-sections. It was proposed that beyond an initial
relaxation length in a circular channel bend, the sec-
ondary circulation reaches its fully developed state cor-
responding to the maximum transverse velocity, u
n;max
.
From the KB model, under the assumption of logarith-
mic proles in the vertical, the transverse velocity in-
duced by either ow curvature (u
nb;max
) or the Coriolis Fig. 1 The curvilinear coordinate system
28
acceleration (u
nc;max
) in the surface layer can be estimated
(for a given drag coecient C
D
= 3 10
3
) as follows:
u
nb;max
~ k
b
(s)
6[u
s
[h
R
s
(6a)
u
nc;max
~ k
c
(s)3fh ; (6b)
where k
b
(s) and k
c
(s) are functions of the streamwise
coordinate and vary between 0 and 1. In fact, they are
the ratio of local u
n
to the fully developed u
n
. Based on
these assumptions, Geyer (1993) estimated the maxi-
mum strength of u
n
for tidal ow at Gay Head and
found that the observed secondary ow was consistently
up to four times stronger than the predicted values from
Eq. (6). This indicates that the omission or retention of
some of the terms on the LHS of Eq. (4) may not be
justied in many eld situations. Comparison between
the terms on the LHS of Eq. (4) using data from Gay
Head indicates that the non-linear terms play a key role
while viscous forces are negligible compared to bottom
friction. Neglecting the viscous term and the vertically
averaged non-linear term in comparison to other terms,
Eq. (4) can be rewritten in the steady state as:
u
s
@u
n
@s

s
n
qh
=
u
2
s
u
2
s
R
s
f (u
s
u
s
) : (7)
The relative importance of advection to friction in the
LHS of Eq. (7) is given by:
u
s
@u
n
@s
=
s
n
qh
~
h
LC
D
= R
ef
; (8)
where L is the streamwise length scale and R
ef
has been
dened as the equivalent Reynolds number (e.g. Pingree
and Maddock 1980; Pattiaratchi et al. 1987; Signell and
Geyer 1991). In a similar manner, the relative importance
of the twopotential drivingforces onthe RHSof Eq. (7) is:
u
2
s
u
2
s
R
s
=f (u
s
u
s
) ~
a
b

u
s
fR
s
~ R
om
: (9)
Here, we have introduced a and b as characteristic
parameters of the streamwise velocity dened by:
a = 1
u
2
s
u
2
s
(10)
b = 1
u
s
u
s
(11)
and R
om
as the modied Rossby number. Although it is
explained later that a and b may be approximated by
two constants so that R
om
is equivalent to the Rossby
number (R
o
= 2u
s
=fR
s
), we keep R
om
in order to compare
the numerical results with the predictions. It means that
for the comparison, we consider the eects of the local
prole of the streamwise velocity in the prediction of u
n
,
while in practice these eects will be included in a con-
stant coecient (see below).
R
om
and R
ef
represent the relative magnitudes of the
generation and dissipation of the momentum in the
secondary ow. On this basis, ow regimes in relation to
secondary ow may be categorized by considering values
of R
om
and R
ef
. For either of these non-dimensional
numbers there would be a value indicating an approxi-
mate balance between the forcing mechanisms. There-
fore, four ow regimes may be dened as shown in
Table 1 (see also Fig. 2). Now, for each of the ow
regimes the strength of the secondary circulation, u
n
, can
be predicted considering a balance between the remain-
ing forces in Eq. (7). The strength of the secondary ow
(u
n
) associated with each of the ow regimes A, B, C and
D (Fig. 2) can be predicted from the primary force
balance as follows:
Regime A: u
n
~
fbh
C
D
(13)
Regime B: u
n
~ fbL (14)
Regime C: u
n
~
ahU
s
C
D
R
s
(15)
Regime D: u
n
~
abU
s
R
s
: (16)
We now examine these predictions of u
n
(Table 1) using
a three-dimensional numerical model to simulate the
conditions for the dierent ow regimes. The corre-
sponding results, comparison and discussion are pre-
sented in the subsequent sections.
Fig. 2 The ow regime diagram based upon the two non-dimensional
numbers, R
om
= (a=b)(bU=Rs) and R
ef
= h=bC
D
Table 1 Flow regimes and the
corresponding dominant
force balances
Regime A Balance between bottom friction and Coriolis forces R
om
< 1 and R
ef
< 1
Regime B Balance between inertia and Coriolis forces R
om
< 1 and R
ef
> 1
Regime C Balance between bottom friction and centrifugal forces R
om
> 1 and R
ef
< 1
Regime D Balance between inertia and centrifugal forces R
om
> 1 and R
ef
> 1
29
3 The numerical model
The three-dimensional primitive equation model,
HAMSOM (HAMburg Shelf Ocean Model), was
developed by Backhaus (1985) for the North Sea
(Backhaus and Hainbucher 1986). The model is based
upon a semi-implicit numerical scheme described by
Stronach et al. (1993). However, for the barotropic
mode, the model is implicit so that relatively large
computational time steps can be employed. The HAM-
SOM model has been widely adapted to the German
Bight; Juan de Fuca system (Backhaus et al. 1987;
Stronach et al. 1993); the South China Sea (Pohlmann
1986), Atlantic coast o Spain and Baja California; and
Southwestern Australia (Pattiaratchi et al. 1996).
The model uses xed permeable horizontal interfaces
between layers and the equations of continuity and
momentum are vertically integrated over each model
layer. The distribution of pressure is assumed to be
hydrostatic and the Boussinesq approximation is in-
voked. The model uses a nite dierence scheme with
the variables distributed over the grid according to the
Arakawa C grid (Arakawa and Lamb 1977). The solu-
tion technique is described in Backhaus (1985) and
Stronach et al. (1993).
HAMSOM includes the choice of either one of three
independent turbulent closure schemes for the computa-
tion of the vertical eddy viscosity (K
z
). These three
schemes are based on the Richardson principle (Stronach
et al. 1993), the Kochergin closure scheme (Kochergin
1987) and constant K
z
in the water column. The hori-
zontal eddy viscosity (K
x
) is considered constant every-
where. At the surface and bottom, boundary conditions
are applied which include surface wind stress and bottom
shear stress. The former is specied using a drag rela-
tionship dependent on the wind speed and direction. The
bottom boundary condition is of the form of the qua-
dratic bottomstress. At lateral solid boundaries, the usual
no-ux condition is used while the roughness of a solid
boundary can be considered by choosing no-slip, half-slip
or full-slip boundary conditions. The sea-surface eleva-
tion or the ow must be prescribed at open boundaries.
4 The idealized domain
An idealized domain was used to isolate the inuence of
the important terms in Eq. (7) such as friction, centrif-
ugal force and the Coriolis force on the development of
the secondary ow. The model domain (Fig. 3a) con-
sisted of a constant depth (h) channel with a 180 160
grid (283 185 km
2
). An elliptic island with sloping
sides was specied at the channel centre (Fig. 3b). The
island was dened by:
y = b 2
x x
0
a

2
1
2
y
0
; (17)
where a and b denote the major and minor axes of the
ellipse at the surface and x
0
, y
0
are the coordinates at the
centre of the channel.
The ability of the HAMSOM model to handle a
variable grid was used to rene resolution around the
island, where the secondary ow is likely to be strong.
Hence, utilizing four dierent grid sizes (250, 1000,
5000 and 15 000 m) in the domain ensured that the
boundary eects did not inuence the area of interest
(Fig. 3a). It was also possible to choose vertical layers
with dierent thickness. Thus, in all experiments ve
layers (5, 10, 10, 10 and 5 m thick, from the surface to
the bottom, respectively) were considered. These runs
were conducted in barotropic mode (see Stronach et al.
1993).
The model included two open boundary conditions,
the northern and southern boundaries as seen in
Fig. 3a. In the upstream open boundary an inow,
uniform along both the width and depth of the
channel, was introduced while the model was allowed
to compute the outow characteristics using the Or-
lanski radiation condition (Orlanski 1976). A no-slip
boundary condition was used for all the lateral
solid boundaries. Each experimental run was started
from rest and continued until the ow was in a steady
state.
5 The numerical experiments
For each of the ow regimes dened in Section 2, several
model runs were conducted. In all the runs, the idealized
geometry was used with h = 16 or 40 m and using
a = 20 km, b = 10 km, S = 1 km (Fig. 3b); The forcing
was a uniform inow of either 3.2 or 8 m
2
s
1
(m
1
width) equivalent to a uniform ow velocity of
0.20 m s
1
which resulted in current speeds o tips of the
idealized island similar to the typical current speeds
around Rottnest Island (see Alaee et al. 1998) which
provided the motivation for the present work. To avoid
instabilities due to surface waves, the inow was
increased gradually from zero to its nal value and
remained constant afterwards. In all of the model runs,
the results indicated an attached ow condition, where
the ow separation either does not occur or does not
result in eddy generation in the wake region (see Figs. 4a
and 5a).
As explained below, the interaction between the
eects of Coriolis and ow curvature leads to two dif-
ferent ow structures around the tips of the island. In
order to assess the prediction method for the strength of
the secondary ow, 38 sampling locations were selected
around each tip of the island, where the presence of
strong secondary circulation was expected (Fig. 3b). At
these locations the predicted transverse velocity
u
n(predicted)
or u
np
; was calculated based upon the scaling
estimates given in Eqs. (13) to (16), and compared to
u
n(model)
, or u
nm
; obtained from the numerical modelling
results, for each of the dierent regimes.
30
5.1 Regime A (R
om
< 1 and R
ef
< 1)
Five numerical experiments were conducted corre-
sponding to regime A, as described in Table 2. To satisfy
the conditions, the vertical eddy viscosity and the drag
coecient for bottom friction were set to 0.005 m
2
s
1
and 0.01, respectively. Various Coriolis parameters, f
(for the Southern Hemisphere), were chosen for these
experiments to produce R
om
of the order of 0.1. Since
R
om
is dependent on the local velocity prole and local
radius of ow curvature, we computed R
om
based on
Eq. (9) for all the sampling locations. It was found that
R
om
was O(0.1) for locations away from the island.
For each numerical run, the distribution of the
transverse velocity in the surface layer was obtained. For
example, Fig. 4a and b show the velocity eld and the
distribution of the u
nm
in the surface layer, for the run
using f = 9:6 10
4
s
1
(run Infcorbf26). To predict the
transverse velocity, u
np
, from Eq. (13), b was computed
from the local vertical prole of the streamwise velocity,
obtained from the numerical results, at each of the
sampling locations. The results indicated that b varied
over a range of 0.15 to 0.2, for various locations and in
between dierent runs. The predicted transverse velocity
u
np
was compared to u
nm
; measured directly from the
numerical results. This comparison implied that for all
the locations around the island, the ow curvature still
inuences the ensuing secondary circulation. With dis-
tance away from the island, the streamlines tend to be
straight, R
s
(similar to cases with b ) and thus
R
ef
0 and R
om
0. The comparison between the
predicted and the measured transverse velocity for these
far-oshore locations showed good agreement between
u
np
and u
nm
(Fig. 6).
5.2 Regime B (R
om
< 1 and R
ef
> 1)
R
ef
> 1 indicates that the non-linear term is the domi-
nant term (Table 1) in the LHS of Eq. (7), implying
Fig. 3a Plan view of the
idealized domain indicating the
dierent grid sizes used in the
numerical simulations. A
uniform depth of h was adopted
for the whole domain. b The area
with the nest grid size (250 m)
and the elliptical island with the
greater and less diameter of a
and b. The sloping side of the
island reaches the depth of H
(dashed line) in a distance of
S. The sampling locations are
shown in black and white dots
31
@u
n
=@s ,= 0. Since f is constant over the domain, the
variation of u
n
in the streamwise direction results from
variations in ow-curvature eects. Thus, the sampled
locations must be in the region of the ow curvature, but
from Section 1, the locations may not be very close to
the island either, because R
om
increases as R
s
decreases in
the immediate vicinity of the island boundaries.
Four numerical runs were conducted with parameters
similar to those used in the regime A except that the
bottom drag coecient was set to 0.001 to ensure
R
ef
> 1. To estimate the transverse velocity from
Eq. (14), b was computed for all the selected locations.
Figure 7 shows the comparison between u
nm
and u
np
for
locations corresponding to the four numerical runs. This
comparison indicates that the predicted transverse
velocities are in good agreement with that from the
numerical results.
5.3 Regime C (R
om
> 1 and R
ef
< 1)
Two numerical experiments, with diering water
depths and horizontal eddy viscosities, were conducted
corresponding to the ow regime C (Table 1). In all the
runs, the Coriolis parameter was set to zero to ensure
that the centrifugal force was the only driving force in
the RHS of Eq. (7). To predict the corresponding u
np
from Eq. (15), a was computed at the sampling loca-
tions from the local prole of u
s
using the numerical
results. Since u
np
for the regimes C and D is inversely
proportional to the radius of ow curvature, R
s
was
calculated based on the orientation of velocity vectors
in two successive numerical grids, at each sampling
location (see Kalkwijk and Booij 1986; Chant and
Wilson 1997) and the results show that the predicted
transverse velocity u
np
is in good agreement with u
nm
(Fig. 8). However, for a few locations adjacent to the
island or far oshore, the predicted transverse velocities
do not correspond with the numerical results. This
discrepancy is likely to be due to inaccuracies in the
determination of R
s
.
5.4 Regime D (R
om
> 1 and R
ef
> 1)
Several numerical runs with various parameter values
were conducted (Table 2). Similar to the preceding sec-
tion, the Coriolis term was neglected (f = 0) and
therefore the ow curvature term was the only driving
force in Eq. (7). In order to compare the predicted
transverse velocity from Eq. (16) to the numerical
results, a and R
s
were computed for each sampling
location. Figure 9 shows that the predicted transverse
velocity from Eq. (16) has reasonable agreement with
u
nm
obtained from the numerical results, despite inac-
curacy in the estimation of R
s
for the neighbouring
locations to the island or far oshore sampling points.
6 Discussion
The ow classication described in Table 1 can be
shown as a ow regime diagram (Fig. 1). The vertical
diusion coecient inuences the transverse velocity
through the streamwise velocity proles and hence the
coecients a and b.
For the four regimes in the ow-regime diagram
(Fig. 1), the maximum transverse velocity can be
estimated, using Eqs. (13) to (16), as:
Fig. 4a The velocity eld in the surface layer from the numerical
modelling results of the case corresponding to the ow regime A with
h = 16 m, K
z
= 0:005 m
2
s
1
, K
x
= 1 m
2
s
1
, C
D
= 0:01 and
f = 9:6 10
5
s
1
(run Infcorbf26). b Distribution of the transverse
velocity in the surface layer for run Infcorbf26. Positive and negative
velocities indicate the secondary circulation owing in the positive and
negative n direction, respectively
32
U
n
= K
A
fh
C
D
(for R
om
< 1 and R
ef
< 1) (17)
U
n
= K
B
fb (for R
om
< 1 and R
ef
> 1) (18)
U
n
= K
C
hU
s
C
D
R
s
(for R
om
> 1 and R
ef
< 1) (19)
U
n
= K
D
bU
s
R
s
(for R
om
> 1 and R
ef
> 1) ; (20)
where U
n
represents the maximum transverse velocity
near the surface, the streamwise length scale L is taken as
the island ellipse scale b, and the coecients K
A
, K
B
, K
C
and K
D
the prole factors a and b. Flow with a specic
vertical prole and a given bottom drag coecient
results in constant values for the prole factors. For
instance, assuming a logarithmic prole (see Kalkwijk
and Booij 1986), C
D
= 3 10
3
yields a = 0:22 and
b = 0:12 and hence a=b = 1:8. Note that the approach
here has been to avoid the a priori assumption that all
streamwise velocities are logarithmic over depth and to
use the simple but plausible values of bottom drag and
vertical viscosity in our model runs with water depths of
16 and 40 m. In this respect, the numerical experiments
indicated as a and b do not vary widely and the ratio a=b
may be approximated by 2. Therefore the modied
Rossby number, dened in Eq. (9), is simplied to
R
om
= 2U
s
=fR
s
.
6.1 Flow regimes A and B
In the ow regimes A and B, the inuence of ow cur-
vature in the generation of the secondary circulation is
negligible compared to Coriolis eects. Figure 10a
demonstrates the comparison between u
np
and u
nm
,
corresponding to the ow regime A. As explained in
Section 4.1, the locations matching the regime A are
located far oshore from the island to be free from the
eects of the ow curvature. Since the vertical structure
of the ow is uniform in the far-oshore region, all the
locations possess identical streamwise ow structure as
well as transverse circulation. On this basis each data
point in Fig. 10a represents the whole oshore region
from the results of one numerical experiment. The best
t through the data indicatesK
A
= 0:026 for in Eq. (17)
and taking C
D
= 0:003, similar to the case of Gay Head
(Geyer 1993), Eq. (17) is simplied to:
U
n
~ 9fh : (21)
Equation (21) is similar to Eq. (6b) from KB for the fully
developed secondary circulation except that the constant
Fig. 5a Velocity vectors in the surface layer from the numerical
results of the run InfB2 with h = 40 m, K
z
= 0:005 m
2
s
1
,
K
x
= 1 m
2
s
1
, C
D
= 0:001. b The transverse velocities in the surface
layer corresponding to ow structure in InfB2. Positive and negative
velocities indicate the secondary circulation owing in the positive and
negative n direction, respectively. c The transverse velocities in the
bottom layer corresponding to the run InfB2
b
33
value inEq. (21) is three times larger thanthat in Eq. (6b).
For the case of the observations around Gay Head (Geyer
1993), Eq. (6b) yields a value of 0.006 and Eq. (21) a value
of 0.018 m s
1
, with the latter prediction being much
closer to the observed U
n
of about 0.13 m s
1
.
Figure 10b illustrates the comparison between u
np
from Eq. (18) and u
nm
measured from the numerical
results at the locations with the regime-B conditions and
from the least-squares line for the data we obtain a value
of K
B
= 0:02.
In the runs associated with regime B, variation of the
transverse velocity u
nm
near the island, at which
the curvature eects become important, is interesting.
The transition between curvature- and Coriolis-induced
secondary circulation can be seen in Fig. 4b: close to the
left tip of the island the transverse circulation induced by
ow curvature is stronger than that induced by the
Coriolis eects due to the small radius of the curvature
near the tip and high local ow speed. However, away
from the island tip the curvature inuences weaken
considerably due to the very large R
s
and thus the
Coriolis-induced secondary circulation dominates. The
transition between these two driving forces is also clear
Table 2 Description of exam-
ples of the numerical experi-
ments that were conducted in
order to examine the prediction
method for the strength of the
secondary circulation
Regime Run h (m) C
D
k
x (
m
2
s
)1
) Inow
(
m
2
s
)1
) f 10
5
(s
)1
)
A Infcorbf22 16 10
)2
1 3.2 8
Infcorbf23 16 10
)2
1 3.2 11.5
Infcorbf25 16 10
)2
1 3.2 14
Infcorbf26 16 10
)2
1 3.2 9.6
Infcorbf2 40 10
)2
1 8 8
B Infcorb22 16 10
)3
1 3.2 8
Infcorb23 16 10
)3
1 3.2 11.5
Infcorb25 16 10
)3
1 3.2 14
Infcorb26 16 10
)3
1 3.2 9.6
C Infbf22 16 10
)3
1 3.2
Infbf2)2 40 10
)2
1 8
D InfB2 40 10
)3
10 8
Infbf1 40 0.00 1 8
Infbf2 40 0.003 1 8
Infbfkx1 40 0.00 1 8
Infkx1 40 10
)3
10 8
Infkx2 40 10
)3
10
)1
8
Fig. 6 Comparison between the transverse velocity from the predic-
tion method and numerical results (for the ve runs described in the
text), corresponding to the regime A
Fig. 7 As Fig. 6 for the regime B
Fig. 8 As Fig. 6 for the regime C
34
in the vicinity of the right tip of the island, where the
Coriolis- and curvature-induced secondary circulation
are of opposite sign. It can be seen that close to the right
tip, the curvature eects overcomes the Coriolis eects
causing an oshore u
nm
of 0.01 cm s
1
, while at a dis-
tance of about 2 km from the tip, an onshore u
nm
of 0.02
cm s
1
is generated. This demonstrates that an asym-
metry can occur in the distribution of transverse velocity
around an island due to the relative magnitudes of the
Coriolis and ow curvature eects.
6.2 Flow regimes C and D
In ow regimes C and D, the centrifugal force due to
ow curvature is the dominant forcing in the RHS of
Eq. (7). This forcing is relatively strong o the tip of the
island due to the occurrence of the high ow speed and
small radius of curvature o the tip. This can be clearly
seen in Fig. 5b, which shows the distribution of the
curvature-induced u
nm
in the surface layer: the strongest
transverse velocities are generated o the tip of the
island, which weaken oshore and downstream.
Figure 11 show the results of the comparison of u
np
from Eqs. (19) and (20) for the ow regimes C and D,
respectively, with the corresponding numerical results of
u
nm
. In both cases, spurious data most likely due to the
unrealistic computed radius of curvature were not con-
sidered. From Fig. 11, two values of 0.11 and 0.27 are
obtained for K
C
and K
D
, respectively.
Using, C
D
= 0:003, Eq. (19) can be simplied to:
U
n
= 35
hU
s
R
s
: (22)
Equation (22) can be compared to Eq. (6a) which was
suggested for the estimate of the maximum transverse
velocity induced by ow curvature (Geyer 1993). Equa-
tion (6a) predicts U
n
= 0:028 m s
1
, whilst Eq. (22)
yields U
n
= 0:166, which compares well with the observed
U
n
~ 0:13 m s
1
around Gay Head. According to the
owregime diagram, however, we suggest that the case of
Gay Head coincides with regime D as discussed below.
6.3 Application
To use the regime diagram to estimate the strength of
the secondary circulation in a eld situation, it is
essential to determine R
ef
= h=C
D
b and R
om
= 2U
s
=fR
s
for the ow. Then, the maximum transverse velocity in
the surface layer can be estimated using the corre-
sponding equation selected from Eqs. (17) to (20).
Consider the case of secondary ow around Gay
Head, one where the eld data is available (Geyer 1993)
and where the observed secondary circulation was
Fig. 9 As Fig. 6 for the regime D
Fig. 10a, b The comparison between the transverse velocities resulted
from the numerical simulations and the predictions for the ow
regimes A (a) and B (b) in order to determine the corresponding
constant factors
Fig. 11a, b As Fig. 10 but for the ow regimes C and D
35
forced by tidal ow around a headland. Geyer (1993)
used the results of the KB analytical model and showed
that the streamwise ow could be approximated to
quasi-steady state. For values of U
s
= 0:7 m s
1
,
h = 20 m, b = 2000 m, R
s
= 3000 m, f = 1 10
4
and
C
D
= 0:003, we obtain R
ef
= 3 and R
om
= 5, which
suggest ow regime D. Using Eq. (20) with K
D
= 0:27,
we obtain U
n
= 0:13 m s
1
, essentially the same as the
observations (~ 0:13 m s
1
).
As another example, consider the strength of the
secondary circulation around Rottnest Island o the
West Australian coast. During the summer wake of
Rottnest Island, a northerly quasi-steady wind-driven
current prevails in the Rottnest Island region
(Pattiaratchi et al. 1996; Alaee et al. 1998). Field data
(Alaee et al. 1998) indicate that wind stress in the
transverse direction is negligible. For values of
U
s
= 0:35 m s
1
, h = 50 m, b = 4500 m, R
s
= 3000 m,
f = 7:7 10
5
(Alaee et al. 1998) and C
D
= 0:003, we
obtain R
ef
= 3:7 and R
om
= 3, which coincides with the
ow regime D. Using Eq. (20), a maximum transverse
velocity of U
n
= 0:14 m s
1
is predicted for the summer
in the vicinity of Rottnest Island.
6.4 Upwelling and downwelling
Garrett and Loucks (1976) suggested that the secondary
ow around headlands may result in localized topo-
graphic-induced upwelling. In the plane of the secondary
circulation normal to the main ow, the onshore ow
near the bottom is drawn upwards and replaces the o-
shore transverse ow near the surface (Fig. 12a). Con-
servation of mass equation may be written as:
@u
s
@s

@u
n
@n

W
h
= 0 ; (23)
where W is the vertical velocity and h is the total depth.
Numerical results presented earlier show that the
streamwise velocity changes slightly over short distances
in the s direction and may neglected in the balance in
Eq. (23) and the two-dimensional continuity equation is
obtained in the plane of the secondary ow.
@u
n
@n
=
W
h
: (24)
Figure 13 shows the distribution of the vertical
velocity in layer 4 (30 m below the surface) corre-
sponding to the numerical results of run InfB2 shown in
Fig. 5 and reveals that upwelling driven by the oshore-
directed transverse velocity is strongest o the tip of the
island and extends a considerable distance downstream.
Using this numerical simulation, and extracting data
near the island tip, we obtain DU
n
= 0:05 m s
1
,
Dn = 750 m (three numerical grids) and with h = 50 m,
we obtain W ~ 0:003 m s
1
which compares favourably
with the range values in Fig. 13.
Topographic-induced downwelling, in the upstream
side of islands, has been reported by several researchers
(e.g. Doty and Oguri 1956; Hamner and Haury 1981).
Fig. 12a, b Schematic diagram indicating the interaction between the
secondary circulation and lateral solid boundary causing upwelling (a)
and downwelling (b)
Fig. 13 Distribution of vertical velocity (in m s
1
) in the layer 4
obtained from the numerical results of InfB2. For grid resolution see
text
36
Figure 13 clearly shows a region of downwelling in the
upstream side of the idealized island. Related to this
downwelling, Fig. 5b and c reveals that in the upstream
of the island the transverse velocity is directed onshore
near the surface and oshore near the bottom and the
downwelling is a result of downslope motion on the side
slope of the island, as shown in Fig. 11b. The numerical
results indicate that the downwelling is about 1 order of
magnitude weaker than upwelling o the tip of the island.
7 Summary and conclusions
In this study, using the numerical simulation of ow
structure around an idealized island or headland, we
investigated the development of the secondary circula-
tion induced by both the ow curvature and the Coriolis
eect, for quasi-steady oceanic ows. Simple scaling of
the transverse momentum equation, in the curvilinear
coordinate system, led us to a ow regime diagram using
two non-dimensional numbers, a Reynolds number
R
ef
= h=C
D
b and a Rossby number R
om
= 2U
s
=fR
s
. A
predictive method was proposed for the strength of the
secondary ow U
n
for each ow regime and it was shown
that the predictions were in good agreement with the
numerical results. Nevertheless, a broader range of
simulations will result in more accurate constant factors
used in the predictive formula.
From the numerical results, the three-dimensional
structure of the ow implied that the interaction between
the secondary circulation and lateral solid boundaries
results in either upwelling or downwelling. It was con-
cluded that the strong secondary ow due to ow cur-
vature around the island (or headland) tips causes strong
upwelling in this vicinity (up to 250 m day
1
from our
results), which can extend far downstream. Downwel-
ling, based on the same mechanism but with reverse
curvature, was found along the upstream side of the
island and is highly localized. The results show that
downwelling is much weaker than upwelling near the tip
of the island (or headland). The occurrence of vertical
water motion in the simulations, with no eddy in the
wake of the island or headland, implies that an island or
headland may induce upwelling or downwelling even
under the attached ow condition.
Acknowledgements The rst author wishes to thank the Ministry of
Culture and Higher Education and Amir Kabir University of
Technology for providing a scholarship. The work was undertaken
whilst the rst author was a student at the Centre for Water
Research, The University of Western Australia. This is Centre for
Water Research reference ED1269MA.
Notation
a; b semimajor and semiminor axes idealized elliptical
island
f Coriolis parameter
g acceleration due to gravity
h total water depth
k
b
; k
c
function of s
k
x
horizontal eddy viscosity coecient
k
z
vertical eddy viscosity coecient
n coordinate perpendicular to streamline
p pressure
s streamwise coordinate
t time
u
n
velocity in n direction (transverse velocity)
u
n;b
transverse velocity due to curvature eect
u
n;c
transverse velocity due to Coriolis eect
u
nm
transverse velocity from the numerical results
u
np
transverse velocity from the prediction method
x; y rectangular orthogonal coordinates
x
0
; y
0
coordinates of centre of elliptical island
z vertical coordinate
C
D
bottom drag coecient
L streamwise length scale
R
n
radius of curvature in n direction
R
s
radius of curvature in s direction
R
ef
equivalent Reynolds number (= h=C
D
L)
R
om
modied Rossby number (= a=b)(U
s
=fR
s
)
U
n
transverse velocity near the surface
U
s
streamwise velocity near the surface
W vertical velocity
a shape factor of streamwise velocity prole
(= 1 u
2
s
=u
2
s
)
b shape factor of streamwise velocity prole
(= 1 u
s
=u
s
)
g water level
s
n
bottom shear stress in n direction
References
Alaee MJ, Pattiaratchi C, Ivey G (1998) A eld study of the three-
dimensional structure of the Rottnest Island wake. In: Dron-
kers J, Scheers MBAM (eds) Physics of estuaries and coastal
seas. Balkema, Rotterdam, 239245
Arakawa A, Lamb V (1977) Computational design of the basic
dynamical processes of the UCLA general circulation model.
Method Comput Phys 16: 173263
Backhaus J (1985) A three-dimensional model for the simulation of
shelf sea dynamics. Dt Hydrogr Z 38: 165187
Backhaus J, Hainbucher D (1986) A nite dierence general cir-
culation model for shelf seas and its application to low fre-
quency variability on the North European Shelf. In: Nihoul
JCJ, Jamart BM (eds) Three-dimensional models of marine and
estuarine dynamics. Elsevier Oceanography Series, vol 45,
Elsevier Amterdam, pp. 221244
Backhaus J, Crean PB, Lee DK (1987) On the application of a
three-dimensional numerical model to the waters between
Vancouver Island and the main land coast of British Columbia
and Washington State. In: Heaps NM (ed) Coastal and estua-
rine sciences, vol 4, American Geophysical Union, Washington,
DC, pp. 149176
Chant RJ, Wilson RE (1997) Secondary circulation in a highly
stratied estuary. J Geophys Res 102: 2320723215
Deleersnijder E, Norrow A, Wolanski E (1992) A three-dimen-
sional model of the water circulation around and island in
shallow water. Continental Shelf Res 12: 891806
Garrett CJR, Loucks RH (1976) Upwelling along the Yarmouth
Shore of Nova Scotia. J Fish Res Board Can 33: 116117
37
Geyer WR (1993) Three-dimensional tidal ow around headlands.
J Geophys Res 98: 955966
Greenspan HP (1968) The theory of a rotating uid. Cambridge
University Press, New York
Hamner WM, Hauri IR (1981) Eects of island mass: water ow
and plankton pattern around reef in the Great Barrier Reef
lagoon, Australia. Limnol Oceanogr 26: 10841102
Kalkwijk JPT, Booij R (1986) Adaptation of secondary ow in
nearly-horizontal ow. J Hydraul Res 24: 1937
Kochergin VP (1987) Three-dimensional prognostic models. In:
Heaps NM (ed) Coastal and estuarine sciences, vol. 4. Ameri-
can Geophysical Union, Washington DC
Orlanski I (1976) A simple boundary condition for unbounded
hyperbolic ows. J Comput Phys 21: 251269
Pattiaratchi C, James A, Collins M (1987) Island wakes and
headland eddies: a comparison between remotely sensed data
and laboratory experiments. J Geophys Res 92: 783794
Pattiaratchi C, Backhaus J, Abu Shamle B, Alaee MJ, Burling M,
Gersbach G, Pang D, Ranasinghe R (1996) Application of the
three-dimensional numerical model for the study of coastal
phenomena in Southwestern Australia. Proceedings of Ocean-
Atmosphere-Pacic Conference, 1996
Pedlosky J (1979) Geophysical uid dynamics. Springer, Berlin
Heidelberg, New York
Pingree RD, Maddock L (1979) Tidal ow around an island with a
regularly sloping bottomtopography. J Mar Bio Ass 59: 699710
Pohlmann T (1986) A three-dimensional circulation model of the
south China Sea. In: Nihoul JCJ, Jamart BM (eds) Three-
dimensional models of marine and estuarine dynamics. Elsevier
Oceanography Series vol. 45, Elsevier Amsterdam. pp. 245268
Rosovski IL (1957) Flow of water in bends of open channels. Israel
program for scientic translation, Ghods
Signell RP, Geyer WR (1991) Transient eddy formation around
headlands. J Geophys Res 96: 25612575
St John MA, Pond S (1992) Tidal plume generation around a
promontory: eects on nutrient concentrations and primary
productivity. Continental Shelf Res 12: 339354
Stronach JA, Backhaus JO, Murty TS (1993) An update on the
numerical simulation of oceanographic processes in the waters
between Vancouver Island and the Mainland. Oceanogr Mar
Biol Annu Rev 31: 186
Thorne CR, Hey RD (1979) Direct measurements of secondary
currents at river inection point. Nature 280: 226228
Vriend HJ, de (1981) Steady ow in shallow channel bends, PhD
Thesis, Delft University of Technology
Wolanski E, Imberger J, Heron ML (1984) Island wakes in shallow
coastal waters. J Geophys Res 89: 1055310569
Wolanski E (1994) Physical oceanographic processes of the Great
Barrier Reef. CRC Marine Science Series, Tokyo, 194 pp
Wolanski E, Asaeda T, Tanaka A, Deleersnijder E (1996) Three-
dimensional island wakes in the eld, laboratory and numerical
models. Continental Shelf Res 16, 11: 14371452
38

You might also like