Lentz JFM2002
Lentz JFM2002
Lentz JFM2002
1. Introduction
The dynamics of gravity currents in a rotating reference frame is a classical problem
that is relevant to a variety of geophysical applications (Simpson 1997). For example,
river water discharging into saltier, and hence denser, ocean water turns cyclonically
and forms a narrow buoyant gravity current that can flow hundreds of kilometres
along the coast before dispersing (e.g. Munchow & Garvine 1993; Hickey et al.
1998; Rennie, Largier & Lentz 1999). The characteristics and dynamics of buoyant
gravity currents along a vertical wall are relatively well understood from laboratory,
theoretical and numerical model studies (Stern, Whitehead & Hu 1982; Griffiths
& Hopfinger 1983; Kubokawa & Hanawa 1984a, b; Griffiths 1986; Helfrich, Kuo
& Pratt 1999). However, the relevance of these studies to the typical geophysical
applications where the buoyant gravity current propagates along a sloping bottom is
unclear. Previous laboratory (Whitehead & Chapman 1986) and numerical modelling
studies (Chao 1988; Chapman & Lentz 1994; Kourafalou et al. 1996; Yankovsky &
Chapman 1997; Xing & Davies 1999; Garvine 1999) suggest a sloping bottom has a
profound influence on the characteristics and dynamics of buoyant gravity currents
(hereafter referred to as simply gravity currents).
Studies have shown that gravity currents against a vertical wall in a rotating fluid
are characterised by a blunt, turbulent nose region, followed by a region of geostrophic
252 S. J. Lentz and K. R. Helfrich
Ww
Nose region
hp
Ww
cw
(a) (b)
Figure 1. Schematic of a buoyant gravity current along a wall: (a) cross-section and (b) plan
view. The circulation pattern in (b) is in a coordinate frame moving with the nose and is based on
Griffiths & Hopfinger (1983).
Wα
vE
α hp
flow (figure 1) (Stern et al. 1982; Griffiths & Hopfinger 1983; Kubokawa & Hanawa
1984a, b; Griffiths 1986). In the geostrophicpregion behind the nose, the thickness
of the buoyant fluid along the wallp is hp = 2Qf/g 0 . The width Ww = b1 LR scales
with the deformation radius LR = g 0 hp /f and the propagation speed of the nose
p
cp = b2 cw scales with the internal wave speed cw = g 0 hp . Here the volume transport
of the gravity current is Q, f is the Coriolis frequency, g 0 = g∆ρ/ρo is reduced gravity,
g is the gravitational acceleration, ∆ρ = (ρs − ρo ) is the density difference between
the plume (ρs ) and ambient fluids (ρo ). The scaling coefficients b1 and b2 are found
to be O(1) from laboratory experiments (e.g. b1 = 0.42 and b2 = 1.1 from Stern et
al. (1982); b1 = 0.6 and b2 = 1.3 from Griffiths & Hopfinger (1983); b1 = 0.8 and
b2 = 1.0 from Kubokawa & Hanawa (1984b)). In a reference frame moving with the
nose, the flow component parallel to the wall is toward the nose near the wall (except
for a narrow frictional boundary layer at the wall) and away from nose over the
outer portion of the gravity current (figure 1b). In the laboratory studies, these flows
rapidly develop instabilities which lead to lateral spreading of the current behind the
nose.
A dynamical framework, similar to that for gravity currents along a wall, does not
exist for the case of a sloping bottom. Whitehead & Chapman (1986) found in a
laboratory study that a gravity current on a sloping bottom was more stable, wider,
and the nose propagated more slowly than a gravity current along a wall, but they did
not offer a dynamical explanation for their results. Garvine (1999) found empirically
from numerical model results that gravity current widths over a slope did not scale
simply with the deformation radius based on the source conditions, but instead also
depended on the bottom slope. A recent laboratory study by Avicola & Huq (2002)
also demonstrates the importance of bottom slope to the buoyant gravity current
response.
Chapman & Lentz (1994) and Yankovsky & Chapman (1997) used numerical
models to study a ‘slope-controlled’ (or ‘bottom-trapped’) gravity current, where most
of the gravity current is in contact with a sloping bottom (figure 2). Slope-controlled
Buoyant gravity currents along a sloping bottom 253
currents have a different dynamical character than ‘surface-trapped’ gravity currents
along a wall because bottom friction is an important component of the dynamics
(Chapman & Lentz 1994). The along-slope flow onshore of the foot of the front drives
an offshore transport in the bottom Ekman layer that widens the gravity current until
the vertically sheared, geostrophic flow at the density front separating the buoyant
and ambient fluid is zero at the bottom. At this point, the offshore Ekman transport
is zero because there is no longer a bottom stress. The gravity current reaches a
geostrophic equilibrium and stops spreading offshore. Behind the nose region, the
equilibrium gravity current transport is concentrated at the front and the flow is
weak onshore of the front where the water is homogeneous. Yankovsky & Chapman
(1997)pshow that in equilibrium the depth where the front intersects the bottom is
hp = 2Qf/g 0 . This is the same as the gravity current depth along a vertical wall
because in both cases the basic assumptions are that the flow is geostrophic and the
flow is zero where the front intersects the bottom or the wall. For a constant bottom
slope α, the offshore distance to where the front intersects the bottom is Wα = hp /α.
The resulting steady gravity current can be significantly wider than the baroclinic
deformation radius typical of a surface-trapped gravity current along a wall.
There is no criterion for evaluating whether the structure of a gravity current
over an arbitrary slope will more closely resemble the surface-trapped (figure 1a) or
slope-controlled (figure 2) case. Yankovsky & Chapman (1997) examine this issue
for the region near the source, but not for the coastal gravity current away from
the source. Neither is there a theory for the nose propagation speed in the presence
of a finite bottom slope. To address these problems, a scaling theory for gravity
currents over a sloping bottom is proposed in § 2. The theory builds on the previous
results from studies of surface-trapped (along a wall) and slope-controlled gravity
currents described above and provides estimates of the geometry (thickness and
width), propagation speed and flow field characteristics of gravity currents along
a slope as functions of Q, g 0 , f and α. The theory recovers the earlier results
for rotating gravity currents against a vertical wall in the limit of infinite bottom
slope and the earlier results for slope-controlled gravity currents in the limit of a
small bottom slope. The transition between these two limits is characterized by a
single non-dimensional parameter. In § 3 the proposed scaling is evaluated with a set
of new laboratory experiments, and the results of previous laboratory experiments
(Whitehead & Chapman 1986) and numerical modelling studies (Chapman & Lentz
1994; Yankovsky & Chapman 1997; Garvine 1999). The results are discussed in § 4
and summarized in § 5.
(b)
Wp
Ww
Wα Ww Q
cp
hp
α Wα u$0
Figure 3. Schematic cross-sections and plan views of the buoyant gravity current geometry and
circulation for (a) surface-trapped cw /cα 1 and (b) slope-controlled cw /cα 1 limits. The
circulation patterns are in a stationary reference frame, in contrast to figure 1.
As noted above, (1) is equally applicable to the limit of a vertical wall. The offshore
distance to where the front intersects the bottom is
hp cw cw
Wα ∼ = , (2)
α f cα
where
p
cw = g 0 hp = (2Qg 0 f)1/4 (3)
is, to within an O(1) constant, the propagation speed of a gravity current nose along
a vertical wall. The parameter
αg 0
cα = (4)
f
has units of velocity and is shown below to be the nose propagation speed scale in
the limit of small bottom slope. Since the along-slope flow is geostrophic, assume
that there is a geostrophic adjustment offshore of the foot of the front so that the
distance from the foot of the front to the offshore edge of the gravity current scales
with the deformation radius based on hp , that is
p
g 0 hp cw
Ww ∼ = . (5)
f f
The total width of the gravity current at the surface is, from (2) and (5),
cw cw
Wp = Ww + Wα ∼ 1+ . (6)
f cα
There are two implied coefficients on the right-hand-side of (6). However, for simplicity
and clarity, the coefficient that multiplies cw /cα is assumed to be 1. The evaluation of
the scaling in § 3 suggests this is a reasonable choice.
The propagation speed of the nose of the buoyancy current cp may be estimated
by assuming that the ageostrophic nose region has a finite extent, the nose shape is
steady as it propagates along the coast, and that there is minimal mixing of the gravity
current fluid with the ambient fluid. Mixing and detrainment at the nose of gravity
Buoyant gravity currents along a sloping bottom 255
currents along a vertical wall is observed in laboratory experiments (Stern et al. 1982;
Griffiths & Hopfinger 1983; Kubokawa & Hanawa 1984b). However, Whitehead &
Chapman (1986) found that mixing at the nose was nearly eliminated over a slope.
With these assumptions, volume conservation at the nose of the buoyancy current
implies Q∆t = Acp ∆t, where A ∼ 12 hp Wp is the cross-sectional area of the plume
(figure 3) and ∆t is a small time increment. Using (1) and (6),
Q cw
cp = ∼ . (7)
A 1 + cw /cα
Thus, the propagation speed of the gravity current nose is always less than cw and
tends to scale with the slower of the two propagation speeds cw or cα . Equations (1),
(2), (6) and (7) give scales for the depth hp , offshore distance to the foot of the front
Wα , width Wp , and nose propagation speed cp of a gravity current given Q, g 0 , α
and f.
The key non-dimensional parameter is the ratio
1/4
cw 2Qf 5 Wα
= ∼ . (8)
cα αg
4 03 Ww
Note that cw /cα is also the scale of the ratio of the isopycnal slope ∼ hp /Ww to the
bottom slope α. The ratio cw /cα may be p thought of as the inverse of a slope Burger
number of the form αN/f, with N = g 0 /hp . If cw /cα 1, then cp ∼ cw , Wp ∼ Ww ,
and the buoyancy current is surface-trapped (figure 3a). In this limit the gravity
current is independent of α and the scaling theory recovers the previous results for
gravity currents along a wall. If cw /cα 1, then cp ∼ cα , Wp ∼ Wα , and the buoyancy
current is slope-controlled (figure 3b). In this limit cp ∼ αg 0 /f is independent of
Q. The slope-controlled limit will occur for buoyancy currents with large transports
(though the 1/4-power dependence is weak), at faster rotation rates (higher latitudes),
smaller density anomalies, and smaller bottom slopes. The two limits for the nose
speed, cw and cα , are analogous to the linear wave speeds of Kelvin waves and
high-wavenumber, coastal-trapped topographic waves, respectively (e.g. LeBlond &
Mysak 1978). However, the dynamical connection is subtle since these gravity current
flows are fundamentally nonlinear.
The character of the associated flow field varies with cw /cα . Onshore of the foot of
the front, the flow should be weak (Chapman & Lentz 1994). This follows from the
assumption that the along-slope flow is geostrophic, that in equilibrium the bottom
stress is zero, and that cross-slope density gradients are small onshore of the front,
so that there cannot be a substantial geostrophic vertical shear in this region. If the
transport Q is confined to the frontal region, then taking the area of the frontal region
to be approximately Ww hp /2 and using (5) and (1), conservation of volume transport
implies that the average velocity, up , is
2Q
up ∼ ∼ cw . (9)
Ww hp
This provides a scale for the average velocity, but does not provide any insight into
the structure of the flow within the frontal region (as seen for example in figure 1),
which presumably depends on the shape of the interface, instabilities and mixing
(Stern et al. 1982; Griffiths & Hopfinger 1983). Thus, behind the gravity current nose
the flow is divided into a region near the front moving at an average velocity cw and
an onshore quiescent region (figure 3). Since from (7) and (9) up /cp ∼ (1 + cw /cα ) > 1,
256 S. J. Lentz and K. R. Helfrich
the approaching flow must turn shoreward within the ageostrophic nose region and
come to rest to fill the quiescent onshore region with buoyant water. In a frame
propagating with the nose, the flow approaches the nose in the offshore region and
flows back upstream near the coast. The flow structure provides a simple kinematic
reason for the reduction in the nose propagation speed cp over a sloping bottom. The
flow up is confined near the front by geostrophy and is independent of the bottom
slope. In the wall limit (figure 1), the entire gravity current is moving at an average
flow speed of up . Since there is no quiescent onshore region, the nose propagates at
the average flow speed, i.e. cp = up . Over a sloping bottom, cp must be less than up
because as the flow approaches the nose, it takes time to fill in the quiescent onshore
region. Both hp and Q are independent of α, so as the bottom slope decreases, the
cross-sectional area of this quiescent onshore region increases, more time is required
to fill in this region, and cp must decrease.
The scaling results are independent of the fluid viscosity or bottom stress, but the
adjustment to an equilibrium state over a slope depends on the frictional, bottom
Ekman layer. The time it takes for the gravity current to reach equilibrium is
presumably related to two time scales: the time scale f −1 for geostrophic adjustment
of the front and the time it takes the foot of the front to move offshore a distance
Wα . The latter is approximately Wα divided by the average velocity in the bottom
Ekman layer vE . However, vE decreases as the front moves offshore because the
near-bottom velocity and the bottom stress decrease as the front approaches its
equilibrium position. Hence the rate at which the front moves offshore decreases with
time. Assuming a constant kinematic viscosity, the offshore velocity in the bottom
Ekman layer scales as vE ∼ ub /2, where ub is the near-bottom geostrophic velocity.
For a narrow front with the along-slope velocity in thermal wind balance the velocity
difference across the front is
g 0 h(t)
up − ub (t) ∼ , (10)
W (t)f
√
where h(t) is the frontal depth which is increasing with time, W (t) ∼ g 0 h(t)/f, h/W
is the slope of the interface, and we assume the velocity scale in the frontal region up
does not change with time. Since up ∼ cw , ub → 0 as h(t) → hp . The rate at which the
foot of the front moves offshore is assumed to scale as the offshore Ekman velocity
1 dh(t) ub
∼ vE ∼ . (11)
α dt 2
Using (10) for ub in (11), using (9), (5) and (3), and integrating yields
p p
t̂ ∼ −2 ĥ − ln(1 + ĥ − 2 ĥ) (12)
where ĥ = h(t)/hp , hp is as defined in (1), t̂ = t/tadj and
2Wα 2cw
tadj = = (13)
cw fcα
is the adjustment time scale for the foot of the front to reach its equilibrium position.
The equilibrium scaling is only valid for times longer than both tadj and f −1 after
the initiation of the flow. Another natural time scale of the proposed scaling that will
be used in the subsequent analysis is tp = Wp /cp = (1 + cw /cα )2 /f. The dynamical
relevance of this time scale to interpretation of the laboratory experiments appears to
be related to interfacial drag between the plume and ambient fluids and is discussed
in § 4.
Buoyant gravity currents along a sloping bottom 257
p
hp = p2Qf/g 0 Maximum thickness of gravity current
cw = g 0 hp Propagation speed in limit of steep bottom slope
cα = αg 0 /f Propagation speed in limit of small bottom slope
cp ∼ cw (1 + cw /cα )−1 Propagation speed
Ww ∼ cw /f Width offshore of the foot of the front (deformation radius)
Wα ∼ hp /α = (cw /f)(cw /cα ) Offshore distance to the foot of the front
Wp ∼ Ww + Wα = (cw /f)(1 + cw /cα ) Total width
up ∼ cw Average velocity in frontal region
tadj = 2cw /fcα Adjustment time scale for foot of front to reach equilibrium
tp = Wp /cp ∼ (1 + cw /cα )2 /f Adjustment time scale associated with interfacial drag
Table 1. Glossary of scales. Input parameters are the gravity current transport Q,
reduced gravity g 0 , Coriolis parameter f, and bottom slope α.
x
Top
view
Sloping
bottom
215 cm
Side
view α
f /2
buoyant spheres (diameter 4 mm) were placed in the gravity current to determine the
characteristics of the near-surface velocity field.
The non-dimensional parameter cw /cα and scales for the widths Wp and Wα , the
propagation speed cp , and the average flow speed up were estimated using (8), (6), (2),
(7), (9) and the source values of Q and g 0 , α and f (table 2). The observed propagation
speed of the nose cobsp as a function of time was determined by differencing the
along-slope positions of the leading edge of the nose in an image sequence and
dividing by the time between images. The widths Wpobs and Wαobs as functions of time
were measured at a fixed location about halfway (≈ 90 cm from the source) along
the slope. For LHa, Wαobs was estimated as the cross-slope location of the darkest
portion (8-bit grey scale) of the gravity current in individual images assuming that
the darkest value corresponds to the thickest point of the dyed gravity current. These
estimates from LHa agree well with the more direct estimates of Wαobs from LHb
(see figure 8). Buoyant sphere displacements were tracked from image to image to
determine the surface water velocity within the gravity current. Average cross-slope
profiles of along-slope velocity were determined within the region between 40 and
160 cm downstream of the source. This region was chosen to avoid complex flows
near the source (discussed below) and at the end of the slope where the gravity
current encounters the wall. The maximum of the average along-slope flow umax
260 S. J. Lentz and K. R. Helfrich
and its cross-slope position ymax were determined and compared to up and Wα ,
respectively.
Whitehead and Chapman (1986) experiments
Observations from two sets of laboratory experiments by Whitehead & Chapman
(1986) are included in the evaluation of the scaling. The first set of experiments
(designated WCa) were similar to ours. Buoyant fluid was pumped into an 89.8 cm
diameter rotating tank at a constant rate (Q = 38 cm3 s−1 ) along a vertical wall.
The resulting gravity current flowed along the wall, encountered a sloping bottom
(α = 0.2), and then flowed along the slope. The values of f and ∆ρ were varied in 12
different runs and the arrival times of the gravity current nose at 10 cm increments
along the slope over a distance of 50 cm were recorded providing estimates of the
propagation speed cobs
p . The widths Wp and Wα were not recorded for these runs.
The second set of 19 runs (designated WCb), conducted in a 2 m diameter tank,
were similar to the first set except that the flow was generated by the release of a
reservoir of buoyant fluid, i.e. a dam break. Again, the resulting gravity current first
propagated along the vertical wall, encountered the slope, and then propagated along
the slope. For this second set f, ∆ρ and the reservoir depth were varied. Because
of the dam break initial condition the gravity current transport Q was unknown.
Therefore, the observed propagation speed along the wall cobs w is used to estimate
√
cw /cα , Wp and cp by replacing cw with cobs
w in (8), (6) and (7). (The use of cw = g 0 H,
where H is the reservoir depth, yields similar results.) The width of the gravity current
was measured by Whitehead & Chapman ‘soon’ after the current reached the slope.
Photographs suggest the gravity current width over the slope continued to increase
with time. There were 17 runs in which both Wpobs and Wp could be estimated. (There
were no reported measurements of cobs w in run 13 and Wp
obs
in run 19 (see table 2
in Whitehead & Chapman 1986).) Whitehead & Chapman (1986) documented the
position of the nose along the wall and slope as a function of time for 10 of the
19 runs (their figure 7). These data were used to estimate cobs p as a function of time
after the gravity current began propagating along the slope. Thus, these experiments
provide 17 estimates of Wpobs (from WCb) and 22 estimates of cobs p (12 from WCa and
10 from WCb). The parameter cw /cα ranged from 1.27 to 10.97, with one run along
a vertical wall.
3.2. Evaluation of the scaling using laboratory results
In each of our laboratory runs (LHa and LHb) a gravity current propagates along
the slope. Examples of two runs with small and large values of cw /cα are shown
in figure 5. Over the sloping bottom there was no obvious evidence of entrainment,
except near the source, and little evidence for the development of frontal instabilities,
as previously found by Whitehead & Chapman (1986). This is in sharp contrast to the
studies of gravity currents along a wall in which turbulent mixing and development
of instabilities are common features (Stern et al. 1982; Griffiths & Hopfinger 1983).
The run in figure 5(a) has cw /cα = 0.17, well into the wall regime, yet shows no
evidence of instability along the front. However, instabilities were observed once the
gravity current reached the end of the slope region and began flowing along the
vertical wall of the tank (not shown). Comparisons of the buoyant plume shape at
different times (figure 6) show that the nose region propagates with essentially steady
shape as assumed in the scaling development. For both extremes of cw /cα , and for
all the intermediate runs, the width and along-slope structure roughly scale as Wp .
The gravity current width is approximately Wp a distance of 1–2Wp behind the nose.
Buoyant gravity currents along a sloping bottom 261
(a) (b)
t = 20 s t = 40 s
t = 60 s t = 120 s
t = 150 s t = 320 s
Figure 5. Examples of two laboratory runs, (a) run 16 cw /cα = 0.17 and (b) run 4 cw /cα = 13, at
three times after the source was turned on. The lower panels include some of the surface particle
tracks. The background grid has 10 cm spacing.
While it seems reasonable that the alongshore structure scales with Wp , this is not an
explicit assumption of the scaling development.
Characteristics near the source varied. For cw /cα > 1 a recirculating bulge often
developed near the source (run 4, figure 5b). The presence and structure of the
recirculating bulge depends in part on the location of the source. Runs 2 and 3 were
nearly identical (table 2) except that the source was near the water’s edge in run 2
and about 15 cm offshore in run 3. A bulge developed near the source in run 3 but
not in run 2. However, downstream of the source differences in the gravity current
images and characteristics for the two runs were small. This suggests that the bulge
caused little alteration of the gravity current transport or density. Water samples
from the gravity current were taken during most runs to determine the density of
the plume (ρp ). Densities in the gravity current were similar to the source values,
(ρp − ρs )/(ρo − ρp ) < 0.4 and typically < 0.1, indicating there was little mixing with the
ambient water. The higher values of (ρp − ρs )/(ρo − ρp ) occurred for smaller values of
cw /cα (i.e. surface-trapped gravity currents). Whether this has to do with conditions
at the source, the proximity of the density samples to the foot of the front, or implies
surface-trapped gravity currents entrain more than slope-controlled gravity currents
is unclear. Thus, while the near-field region of the source shows interesting variations,
it does not appear to have a substantial influence on the downstream characteristics
262 S. J. Lentz and K. R. Helfrich
2.0
(a)
1.5
y 1.0
Wp
0.5
y 1.0
Wp
0.5
Figure 6. Outlines of plumes relative to the nose at several times during the two runs shown in
figure 5, (a) cw /cα = 0.17 at t = 38–138 s at intervals of 20 s and (b) cw /cα = 13.1 at t = 56–336 s
at intervals of 40 s. The x (along-slope) and y (cross-slope) dimensions are scaled by Wp . In (b) the
plume shapes have been rotated so that the x-axis is parallel to the shore near the nose to reduce
effects of the shoreline curvature at this high rotation rate.
of the gravity current. In the analysis below we focus on the region downstream of
the source.
Bottom width
Since hp and Wα are simply related by (2) given α, this comparison focuses on
measurements of Wα . Values of Wαobs as functions of time t since the nose reached
the measurement position increase rapidly initially, then increase more gradually,
approaching a constant value ranging from 2 to 25 cm depending on the particular
run (figure 7a). (Runs LHb are shown in figure 7 because they provide clearer
estimates of Wαobs than runs LHa, which show the same time dependence but are
noisier.) Normalizing time by tadj , the time scale for the foot of the front to move
offshore to its equilibrium position, and Wαobs by Wα tends to collapse the observations
(figure 7b). The normalized widths approach roughly constant values between 0.7 and
0.9 for t > (10 − 30)tadj . Runs with cw /cα > 7 were too short (t/tadj < 30) to have
reached steady state. For these runs tadj is relatively long (15–25 s) because these
gravity currents are wide at the bottom and consequently it takes longer for the
foot of the front to move offshore to the equilibrium position. The observed time
dependence of the normalized widths is similar to the theoretical time dependence
(dashed line figure 7) given by (12), though the normalized widths approach a value
that is less than 1.
Buoyant gravity currents along a sloping bottom 263
30
25
20
Wαobs (cm)
15
10
5
(a)
1.0
0.8
Wαobs/Wα
0.6
0.4
0.2
(b)
0 50 100 150
t/tadj
Figure 7. (a) The offshore distance to where the front intersects the bottom Wαobs as a function of
time for runs LHb. (b) Wαobs normalized by Wα as a function of time normalized by tadj . The dashed
line in (b) is the theoretical time dependence from (12). The normalized observed widths reach a
roughly constant value between 0.7Wα and 0.9Wα .
To quantify the relationship between Wαobs and Wα , averages of Wαobs over the last
five observations of each run were computed for LHa and LHb (figure 8). Including
only runs with cw /cα < 7 that appear to have reached a steady state (these are the
runs with Wα < 25 in figure 8), regression analysis yields a slope of 0.67 ± 0.09 and
a correlation between Wαobs and Wα of 0.96. Including all 28 runs yields a regression
slope of 0.54 ± 0.07 and a correlation of 0.95. (Table 3 contains the correlations and
regression slopes for this and all subsequent comparisons.) Note that the results are
similar for runs LHa (buoyant water dyed) and LHb (ambient water dyed).
Surface width
In contrast to Wαobs , the total gravity current widths Wpobs continue to increase and
do not approach a steady width over the duration of each run (figure 9a). Both Wpobs
and the rate of increase vary substantially between runs. For example, 100 s after the
264 S. J. Lentz and K. R. Helfrich
30
20
13
Wαobs (cm)
10
1
0.11
0 10 20 30 40
Wα (cm)
Figure 8. Comparison of Wαobs (averaged over the last five samples of each run) and Wα for runs
LHa (open circles) and LHb (solid circles). For runs that are near equilibrium (cw /cα < 7) the
correlation is 0.96 and the regression slope (dashed line) is 0.68. In this and all subsequent plots
the error bars are standard error of the mean (standard deviation divided by the square root of
number of samples). The value of cw /cα is noted next to selected points.
40
30
Wpobs (cm)
20
10
3.0
(b)
2.5
3.1
c =1
2.0 c w/ α
Wpobs/Wp
1.5 .11
c =0
c w/ α
1.0
0.5
0 5 10 15 20 25
t/tp
Figure 9. (a) The total width Wpobs of the gravity current as a function of time for runs LHa.
(b) Wpobs normalized by Wp as a function of time normalized by tp . The dashed lines in (b) are the
estimated time dependence due to interfacial drag from (17) discussed in § 4 for the two experimental
limits of cw /cα . The time dependence scales as t1/2 .
dependence of Wpobs /Wp may be due to interfacial drag between the buoyant plume
and the ambient fluid causing the continual increase in gravity current width. The
dashed lines are theoretical curves for the time dependence due to the combination
of interfacial drag and offshore movement of the foot of the front derived in § 4.
To evaluate the width scaling, Wpobs was averaged over the time period 1 < t/tp < 3,
and compared to Wp . The time interval for the averaging was chosen to include all the
runs. The shortest run in terms of t/tp was run 4 (cw /cα = 13.1), which had a duration
of 1.4tp . The correlation between the average Wpobs over this time interval and Wp is
0.97 and the regression slope is 1.01 ± 0.14 (figure 10). It is evident from figure 9(b)
that the correlation is not sensitive to the averaging interval, but the regression slope
is sensitive to the choice of averaging interval. The scaling also collapses the gravity
current widths reported by Whitehead & Chapman (1986) from WCb (figure 10 and
table 3). However, for WCb the regression coefficient is 0.35, possibly because the
266 S. J. Lentz and K. R. Helfrich
13
40
30
Wpobs (cm)
20
0.11
10
0 10 20 30 40
Wp (cm)
Figure 10. Comparison of Wpobs (averaged over the time interval 1 < t/tp < 3) and Wp for runs
LHa (solid circles). The correlation is 0.97 and the regression slope (dashed line) is 1.01 for LHa.
Also shown are the results from WCb (open circles) which have a correlation of 0.92 and regression
slope (dashed line) of 0.35. The value of cw /cα is noted next to selected points.
widths were measured shortly after the gravity current encountered the slope, i.e. at
an earlier normalized time than in the present laboratory study. As noted previously,
their photographs indicate that the gravity current widths continued to increase after
the time the width was defined, as also observed in our laboratory runs (figure 9).
The buoyancy anomaly (g 0 ) may also be reduced relative to its source value in their
runs due to entrainment during the period the gravity current flowed along the wall.
Recall that the scaling for Wp (6) has two implied coefficients. The comparisons in
figure 10 support the choice of 1 for the coefficient multiplying cw /cα .
Propagation speed
Estimates of cobs
p in runs LHa and LHb decrease with time from initiation of
the source inflow (or equivalently distance from the source) for each run, with the
largest variations at the start of the run (figure 11a). Scaling time by tp and cobs p
by cp , the large initial variability between runs is primarily confined to t/tp < 2
(figure 11b). (In figure 11b the noisy time series of cobs
p were filtered (five-point low-
pass filter) to clarify the trends.) At longer times, normalized propagation speeds
fall on approximately the same curve, with cobs p /cp decreasing roughly as t
−1/2
. The
same time dependence is observed in WCa and WCb (not shown, but see figure 7
in Whitehead & Chapman 1986). The dashed lines again indicate the expected time
dependence due to the combination of interfacial drag and offshore movement of the
foot of the front (§ 4).
To quantify the relationship between cobs obs
p and cp , cp (unfiltered data) were averaged
over the time period 2 < t/tp < 4 (figure 12). (For WCa and WCb the time origin was
taken to be when the gravity current began flowing along the slope.) The averaging
period was a compromise between obtaining enough of the noisy cobs p estimates to
Buoyant gravity currents along a sloping bottom 267
4
(a)
3
cpobs (cm s–1)
3
cpobs/cp
0 5 10 15 20 25 30
t/tp
Figure 11. (a) The nose propagation speed cobs p of the gravity current as a function of time for runs
LHa. (b) Filtered cobs
p normalized by c p as a function of time normalized by tp . The dashed lines in
(b) are the estimated time dependence due to interfacial drag from (18) discussed in § 4 for the two
experimental limits of cw /cα . The time dependence scales as t−1/2 .
form an accurate average and not including too long a period over which cobs p is
decreasing. Two runs from LHa and LHb and four runs from WCb had durations
less than 2tp ; for these runs the averaging interval was 1 < t/tp < 2 (open symbols
figure 12). Three of the runs have large values of cobs
p relative to cp , consistent with
the runs not having reached equilibrium. For the WCa runs the last data point is
chosen because the duration was less than 2tp for all but one of these runs. There
is close agreement between cp and the average cobs p for both the present laboratory
experiment and the Whitehead & Chapman study (figure 12, table 3). The one outlier
from WCa (cp = 4.5 cm s−1 ) corresponds to a run along a vertical wall (which was
erroneously identified as over a slope in table 1 in Whitehead & Chapman (1986) (J.
Whitehead, personal communication)) and is not included in the regression analysis
(table 3).
268 S. J. Lentz and K. R. Helfrich
4
cpobs (cm s–1)
0.11
1
13
0 1 2 3 4 5
cp (cm s–1)
50
(b)
40
30
20
10
0
40 80 120 160 200
Along-slope distance rel. to source (cm)
Cross-slope distance (cm)
50
(c)
40
30
20
10
0
–140 –120 –100 –80 –60 –40 –20 0
Along-slope distance rel. to nose (cm)
Figure 13. The flow field from run 16, cw /cα = 0.17, (a) individual particle tracks, (b) velocity
vectors in the stationary reference frame, and (c) velocity vectors in a reference frame moving with
the nose of the gravity current. The outline of the plume near the end of the run is also shown for
reference in each panel.
50
(b)
40
30
20
10
0
20 40 60 80 100 120 140 160 180
Along-slope distance rel. to source (cm)
Cross-slope distance (cm)
50
40 (c)
30
20
10
0
–120 –100 –80 –60 –40 –20 0
Along-slope distance rel. to nose (cm)
Figure 14. The flow field from run 7, cw /cα = 3.8, (a) individual particle tracks, (b) velocity vectors
in the stationary reference frame, and (c) velocity vectors in a reference frame moving with the nose
of the gravity current. The outline of the plume near the end of the run is also shown for reference
in each panel.
30 30 30
y (cm)
20 20 20
Wαobs
ymax
10 10 10
Wαobs
Wαobs
umax
0 1 2 0 1 2 3 0 0.5 1.0 1.5
u (cm s–1) u (cm s–1) u (cm s–1)
Figure 15. Cross-slope profiles of along-slope velocity u for (a) cw /cα = 0.17 (run 16),
(b) cw /cα = 1.01 (run 6), and (c) cw /cα = 5.91 (run 10). Small dots represent individual velocity
estimates from the difference in particle positions between frames. Large circles represent aver-
age velocity in 2 cm cross-slope bins. The maximum along-slope velocity umax and the cross-slope
location ymax for the bin-averaged profiles are determined for each run. The locations of Wαobs are
indicated by the triangles.
than 1. The numerical model runs by Chapman & Lentz (1994) (12 model runs) and
Yankovsky & Chapman (1997) (nine model runs) focused on slope-controlled gravity
currents for which cw /cα ranged from 1 to 13.
Garvine (1999) varied Q, ∆ρ, α, source widths and latitude (f). The gravity currents
only penetrate a finite distance along the coast because of mixing. Garvine estimated
the width of the gravity current at a position that was about halfway between the
source and the total along-coast penetration distance of the gravity current. Garvine
then determined empirically the dependence of gravity current width on various
input parameters for those runs with a constant bottom slope of 0.002 (summarized
in his figure 13). To determine whether Wp collapses Garvine’s numerical model
results, estimates of Wp from (6) are compared to the model buoyancy current widths
Wpmodel (figure 17). The model widths are proportional to Wp for the model runs in
which Q, ∆ρ, α and inflow width were varied, but not for the model runs at small
f (latitudes less than 25◦ ). Excluding the low latitude runs, the correlation is 0.98
(table 3). Moreover, (6) accounts for the dependence on bottom slope that was not
included in Garvine’s empirical relationship. The regression coefficient of about 4
between Wpmodel and Wp presumably accounts for the relationship between the source
and local gravity current values of Q and g 0 . There is a substantial reduction in
g 0 between the source and where the width was measured due to mixing with the
ambient fluid. The failure of the scaling at low latitudes is probably because the
relationship between the inlet transport and the coastal gravity current transport is
272 S. J. Lentz and K. R. Helfrich
2.5
(a)
2.0 1
umax (cm s–1)
0.11
1.5
1.0
0.5 13
0 1 2 3 4 5 6
up = cw (cm s–1)
(b) 13
20
15
ymax (cm)
10
0.11
0 10 20 30 40
Wα or Wαobs (cm)
Figure 16. (a) Comparison of maximum along-slope flow from the mean cross-slope profile umax
and up for runs LHa. The correlation is 0.74 and the regression slope (dashed line) is 0.38.
(b) Comparisons of cross-slope position of current maximum ymax with Wα (solid circles) and with
Wαobs (open circles) for runs LHa. The correlation between ymax and Wαobs is 0.94 and the regression
slope (dashed line) is 1.04. The value of cw /cα is noted next to selected points.
50
Wpmodel (km)
40
30
20
10
0 5 10 15 20 25
Wp (km)
Figure 17. Comparison of gravity current width Wpmodel with Wp from Garvine’s (1999) numerical
model runs. The various run types are indicated in the inset. Excluding runs with latitudes (f) less
than 25◦ , the correlation is 0.98 and the regression slope (dashed line) is 3.97.
in the buoyant gravity current equalled the source transport. The resulting gravity
current widths from their model studies, Wpmodel , are proportional to Wp (figure 18)
with regression slopes of 0.55 and 0.92 (table 3). The smaller regression slope for the
Chapman & Lentz (1994) results may be related to the upstream spreading of some of
the buoyant water from the source. Both studies also reported the offshore distances
to the foot of the front Wαmodel , which agree well with Wα (table 3), as previously
reported by Yankovsky & Chapman (1997). There is relatively little change in the
density of the buoyant gravity current in the numerical model runs of both Chapman
& Lentz (1994) and Yankovsky & Chapman (1997) compared to the model runs of
Garvine (1999), indicating much less mixing with the ambient shelf water. Therefore,
the local Q and g 0 of the buoyant current are closer to the inflow (source) values.
4. Discussion
The laboratory experiments and the reanalysed numerical model results support
the proposed scaling in almost every measure over two orders of magnitude in
the parameter cw /cα . The demonstration that the front adjusts to a steady offshore
position that scales as Wα (figures 7 and 8), is the first laboratory confirmation of the
slope-controlled gravity current equilibrium (Chapman & Lentz 1994; Yankovsky &
Chapman 1997).
However, one aspect of the laboratory observations that is not consistent with the
scaling theory is the temporal increase in Wpobs and corresponding decrease in cobs p .
In the absence of entrainment, the assumption was made that the gravity current
would reach an equilibrium and then the geometry and propagation speed would not
change. The fact that Wpobs increases as t1/2 (figure 9b) and cobs p decreases as t−1/2
(figure 11b) is consistent with conserving volume transport and hp , or equivalently
Wαobs , reaching a steady value (figure 7). Stern et al. (1982) and Griffiths & Hopfinger
274 S. J. Lentz and K. R. Helfrich
60
8.5
50
Wpmodel (km)
40
6.3
30 4.7
20
10 1.1
0 10 20 30 40 50 60
Wp (km)
Figure 18. Comparison of Wpmodel with Wp from Chapman & Lentz’s (1994) and Yankovsky &
Chapman’s (1997) numerical model runs. Values of cw /cα are noted for selected numerical model
runs.
(1983) observed a weak exponential decay of the nose speed of gravity currents
along a wall that Griffiths & Hopfinger (1983) attributed to radiation of inertial
waves. Griffiths (1986) suggests that continual detrainment of buoyant plume fluid
from the nose region also may lead to a slow decay of speed. The finite size of the
upstream reservoirs also may lead to an eventual reduction in nose speed as the
initially upstream-propagating Kelvin wave generated by the dam break is reflected
downstream and eventually catches up with the nose (cf. Rottman & Simpson 1983).
Kubokawa’s (1984b) experiments along a wall showed little evidence of nose speed
decay, perhaps because of the relatively large upstream basin.
The nose propagation speed along the slope in Whitehead & Chapman’s exper-
iments also exhibited a t−1/2 dependence that they attributed to downstream radiation
of topographic shelf waves. Interestingly, the increasing width is not evident in
numerical model runs (Chapman & Lentz 1994; Yankovsky & Chapman 1997)
and the decreasing propagation speed with time is not evident in limited oceanic
observations (Rennie et al. 1997). This suggests that the temporal evolution of cp
and Wp may be associated with a process specific to the laboratory. One possible
explanation for the t−1/2 dependence of cp is continuous entrainment of ambient fluid
into the buoyant current that reduces g 0 (Rennie et al. 1997). However, there was
little evidence of significant entrainment in the new laboratory runs from either the
video images or the measurements of downstream plume density. Furthermore, the
agreement between the time dependence of Wαobs and the theoretical time dependence
(2) that assumes no change in g 0 (figure 7) suggests entrainment is not the cause of
the time dependence.
An alternative explanation is that viscous, interfacial drag between the buoyant
Buoyant gravity currents along a sloping bottom 275
current and ambient fluid at rest is widening and slowing the gravity current. The
resulting stress-driven Ekman flow would move the front offshore near the surface
(increase Wp , but not affect Wα ) and retard the along-slope flow. From Ekman layer
dynamics, if the geostrophic flow in the gravity current is up , the stress-driven com-
ponent of flow (uE , vE ) ∼ (−up /4, up /4), oriented at 45◦ to the along-slope direction,
and it occurs in an Ekman layer of thickness (2ν/f)1/2 = 1.5–3 mm, where ν is the
kinematic viscosity of water. This is qualitatively consistent with the observed flow
pattern in figure 13. To test this hypothesis, consider first the influence of this stress-
driven flow on the width of the gravity current offshore of the foot of the front Ws (t)
(to distinguish this time-dependent width from the width scale Ww ). Assume a fixed
transport Q confined offshore of the foot of the front, as before, so
W s hp
Q∼u , (14)
2
where u is the along-slope geostrophic velocity in the frontal region. Assume hp is
fixed, but that the width Ws increases at a rate proportional to the offshore Ekman
drift
∂Ws u
∼ vE = . (15)
∂t 4
Using (15) to substitute for u in (14), noting that Q/hp ∼ fWw2 /2 (see (9)) and
integrating in time yields
1/2
tf
W s ∼ Ww , (16)
2
where we assume that Ws = 0 at t = 0. The total width W (t) ∼ Ws (t)+Wα normalized
by Wp ∼ Ww (1 + cw /cα ) is
1/2
W t cw /cα
∼ + , (17)
Wp 2tp 1 + cw /cα
where tp = Wp /cp = (1 + cw /cα )2 /f. For simplicity, we focus on times long compared
to the rapid adjustment of the foot of the front to its offshore equilibrium position.
Therefore we neglect the time dependence of h given by (12) and assume the depth
of the foot of the front is constant and equal to hp . The time dependence associated
with the adjustment of the foot of the front can be incorporated into this analysis
by integrating numerically, but the changes are small and do not alter the basic t1/2
behaviour. The time dependence of the nose propagation speed c(t) can be estimated
by assuming c(t) ∼ Q/A(t) as in (7), where, in this case, A(t) ∼ W (t)hp /2. Using (17)
and (12) yields
" #−1
1/2
c t cw /cα
∼ + . (18)
cp 2tp 1 + cw /cα
The estimated time dependence due to interfacial drag, W (t)/Wp from (17) and
c(t)/cp from (18) are shown in figures 9 and 11 (dashed lines) for the minimum
and maximum values of cw /cα . A proportionality constant of 0.75 for W , and the
inverse for c have been used to roughly fit the data. In both cases the essential
feature is the t1/2 behaviour of Ws in (16) and t−1/2 dependence of c in (18). The
agreement supports the hypothesis that the continual increase in the gravity current
widths observed in both our experiments and those of Whitehead & Chapman (1986)
276 S. J. Lentz and K. R. Helfrich
are due to interfacial viscous drag between the buoyant plume and ambient fluids.
While this process is important in the laboratory studies, it may not be important in
the ocean (or numerical models with stratification-dependent mixing schemes) where
viscous drag is small and turbulent stresses at the interface may be inhibited by the
strong stratification separating the gravity current and the ambient fluid.
For our laboratory experiments, the constants associated with the width Wp and
nose propagation speed cp are approximately 1 based on either the regression analyses
(table 3) or the coefficients used for the derived time dependence (figures 9 and 11).
This is consistent with the corresponding constants found in the previous laboratory
studies of gravity currents along a wall (Stern et al. 1982; Griffiths & Hopfinger 1983;
Kubokawa & Hanawa 1984b). One of the striking contrasts to previous studies of
gravity currents along a wall is the stability of the flow in the presence of a slope
(figure 5). Over a sloping bottom there was little evidence of mixing in the nose
region. Griffiths & Hopfinger (1983) concluded that mixing near the nose was due
to Kelvin–Helmholtz instability for gravity currents against a vertical wall. Neither
was there indication of baroclinic instability in the trailing geostrophic current over
a sloping bottom, in contrast to the vertical wall studies (Griffiths & Linden 1981,
1982; Griffiths & Hopfinger 1983). The observed stability is consistent with linear
stability analysis of coastal upwelling fronts over a sloping bottom (Barth 1989a, b;
Reszka & Swaters 1999). The elimination of the Kelvin–Helmholtz billows is not due
solely to Reynolds number effects. In our experiments an estimate of the Reynolds
number in the nose region is Re = (up hp )/ν ≈ 800–2000, comparable to those of Stern
et al. (1982) and Griffiths & Hopfinger (1883). One possible explanation is that the
shallow layer under the nose region prevents billows from growing to large amplitude.
Another possibility is that the vertical shear is reduced by the presence of the slope.
Recall that some experiments were conducted for cw /cα < 0.2, well into the wall
regime, and that the nose speeds and current widths approached the earlier results
for gravity currents along a vertical wall. However, even for these low values of cw /cα
the slope was at most α = 0.29. It remains to be seen if nose region turbulence and
baroclinic instabilities occur for larger, but finite, values of α, or whether the vertical
wall is a special case.
It is premature to claim that the scaling coefficients that have been determined
from the experiments are valid at oceanographic scales. However, it is useful to ask
where parameters typical of the coastal ocean lie in the scaling. Continental shelves
generally have bottom slopes in the range α = 10−4 to 10−2 . For a typical mid-latitude
estuary with a moderate outflow Q = 500 m3 s−1 , f = 10−4 s−1 , and g 0 = 0.02 m s−2
(representative of Delaware Bay; Garvine 1999). For these values, with α = 10−4 to
10−2 , (8) gives 10 > cw /cα > 0.1. The weak Q1/4 dependence of cw /cα makes this
estimate relatively insensitive to the source transport, i.e. increasing the transport by
a factor of 100 (representative of a river such as the Mississippi) only increases cw /cα
by a factor of 3. Thus, oceanic buoyant coastal currents are often in the intermediate
range between the surface-trapped and slope-controlled limits. The vertical wall limit,
while providing insight, is probably not directly applicable to most oceanographic
situations. One consequence is that in oceanographic settings, the flow field may
be quite different from that observed in the classical laboratory studies of gravity
currents along a vertical wall, with the along-slope flow concentrated offshore near
the front and onshore flow near the nose (figure 14). This flow pattern and the rate at
which fluid moves along the shelf may have implications for the coastal ocean since
buoyant coastal currents transport constituents, such as sediment, marine organisms,
nutrients and chemical pollutants.
Buoyant gravity currents along a sloping bottom 277
5. Summary
A simple scaling theory is proposed for buoyant gravity currents over a sloping
bottom that extends and builds on previous studies of gravity currents along a
wall (Stern et al. 1982; Griffiths & Hopfinger 1983; Kubokawa & Hanawa, 1984b)
and gravity currents in contact with a sloping bottom (Chapman & Lentz 1994;
Yankovsky & Chapman 1997). The proposed theory provides scales for the buoyant
current thickness hp (1), the width Wp (6), nose propagation speed cp (7), and the
flow field characteristics ((9), and figure 3) given the gravity current transport Q, the
density anomaly g 0 (= g∆ρ/ρ◦ ), the Coriolis frequency f and the bottom slope α. Key
assumptions are: the cross-slope momentum balance is geostrophic; the foot of the
front moves offshore until the near-bottom flow, and hence the bottom stress, are zero;
and the gravity current nose shape remains roughly steady as it propagates along
the coast. The key non-dimensional parameter characterizing the buoyant gravity
current response is cw /cα , where cw = (2Qg 0 f)1/4 is the nose speed limit over a steep
bottom slope (vertical wall) and cα = αg 0 /f is the nose speed limit over a gently
sloping bottom. If cw /cα 1, then the gravity current is surface-trapped, independent
of the bottom slope and resembles a gravity current along a vertical wall. In this
case the gravity current width is approximately the deformation radius based on the
gravity current thickness hp and both the nose speed and the average flow speed
are approximately cw . If cw /cα 1, then the gravity current is slope-controlled, i.e.
most of the buoyant water is in contact with the bottom and bottom friction is
important in the establishment of the current. In this case the gravity current width is
1 + cw /cα times the deformation radius based on hp and the nose propagation speed
is approximately cα . Behind the nose, the along-slope flow is concentrated at the
front separating the buoyant and ambient fluids and the flow is weak in the region
onshore of the front. Near the nose, the flow turns onshore to fill in this quiescent
nearshore region. A scaling for the temporal adjustment to the equilibrated front
is proposed. Results from new and previous laboratory experiments spanning two
orders of magnitude in cw /cα , and previous numerical modelling results all strongly
support the scaling theory. Experimentally observed time-dependence of the gravity
current width and nose speed is shown to be consistent with viscous interfacial Ekman
processes.
We thank Rich Garvine for graciously providing his numerical model results and
helpful discussions. We also thank Dave Chapman and Jack Whitehead for helpful
discussions. This research was supported by NSF grant OCE-0095059. This is Woods
Hole Oceanographic Institution Contribution Number 10575.
REFERENCES
Avicola, G. & Huq, P. 2002 Scaling analysis for the interaction between a buoyant coastal current
and the continental shelf: experiments and observations. J. Phys. Oceanogr. (to appear).
Barth, J. 1989a Stability of a coastal upwelling front, 1, model development and a stability theorem.
J. Phys. Oceanogr. 94(C8), 10844–10856.
Barth, J. 1989b Stability of a coastal upwelling front, 2, model results and comparison with
observations. J. Phys. Oceanogr. 94(C8), 10857–10883.
Chao, S.-Y. 1988 River-forced estuarine plumes. J. Phys. Oceanogr. 18, 72–88.
Chapman, D. C. & Lentz, S. J. 1994 Trapping of a coastal density front by the bottom boundary
layer. J. Phys. Oceanogr. 24, 1464–1479.
Fong, D. 1998 Dynamics of freshwater plumes: observations and numerical modeling of the wind-
278 S. J. Lentz and K. R. Helfrich
forced response and alongshore freshwater transport. PhD Thesis, MIT/WHOI Joint Program
in Oceanography.
Garvine, R. W. 1999 Penetration of buoyant coastal discharge onto the continental shelf: a numerical
model experiment. J. Phys. Oceanogr. 29, 1892–1909.
Garvine, R. W. 2001 Artifacts in buoyant coastal discharge models: an observational and model
study. J. Mar. Res. 59, 193–225.
Griffiths, R. W. 1986 Gravity currents in rotating systems. Annu. Rev. Fluid Mech. 18, 59–89.
Griffiths, R. W. & Linden, P. F. 1981 The stability of buoyancy driven coastal currents. Dyn.
Atmos. Oceans 5, 281–306.
Griffiths, R. W. & Linden, P. F. 1982 Density-driven boundary currents. Geophys. Astrophys. Fluid
Dyn. 19, 159–187.
Griffiths, R. W. & Hopfinger, E. J. 1983 Gravity currents moving along a lateral boundary in a
rotating frame. J. Fluid Mech. 134, 357–399.
Helfrich, K. R., Kuo, A. C. & Pratt, L. J. 1999 Nonlinear Rossby adjustment in a channel.
J. Fluid Mech. 390, 187–222.
Hickey, B. M., Pietrafesa, L. J., Jay, D. A. & W. C. Boicourt, W. C. 1998 The Columbia River
Plume Study: subtidal variability in the velocity and salinity fields. J. Geophys. Res. 103(C5),
10339–10368.
Kourafalou, V. H., Oey, L.-Y., Wang, J. D. & Lee, T. N. 1996 The fate of river discharge on the
continental shelf, 1, Modeling the river plume and the inner shelf coastal current. J. Geophys.
Res. 101(C2), 3415–3434.
Kubokawa, A. & K. Hanawa, K. 1984a A theory of semigeostrophic gravity waves and its
application to the intrusion of a density current along a coast. Part 1. Semigeostrophic gravity
waves. J. Oceanogr. Soc. Japan 40, 247–259.
Kubokawa, A. & K. Hanawa, K. 1984b A theory of semigeostrophic gravity waves and its
application to the intrusion of a density current along a coast. Part 2. Intrusion of a density
current along a coast in a rotating fluid. J. Oceanogr. Soc. Japan 40, 260–270.
LeBlond, P. H. & Mysak, L. A. 1978 Waves in the Ocean. Elsevier.
Munchow, A. & Garvine, R. W. 1993 Dynamical properties of a buoyancy-driven coastal current.
J. Geophys. Res. 98(C11), 20063–20077.
Pichevin, T. & Nof, D. 1997 The momentum imbalance paradox. Tellus 49A, 298–319.
Rennie, S., Largier, J. L. & Lentz, S. J. 1999 Observations of low-salinity coastal current pulses
downstream of Chesapeake Bay. J. Geophys. Res. 104(C8), 18227–18240.
Reszka, M. K. & Swaters, G. E. 1999 Eddy formation and interaction in a baroclinic frontal
geostrophic model. J. Phys. Oceanogr. 29, 3025–3042.
Rottman, J. W. & Simpson, J. E. 1983 Gravity currents produced by instantaneous releases of heavy
fluid in a rectangular channel. J. Fluid Mech. 135, 95–110.
Stern, M. E., Whitehead, J. A. & Hu, B. L. 1982 The intrusion of the head of a gravity current
along the coast of a rotating fluid. J. Fluid Mech. 123, 237–266.
Simpson, J. H. 1997 Gravity Currents: in the Environment and the Laboratory, 2nd Edn. Cambridge
University Press.
Whitehead, J. A. & Chapman, D. C. 1986 Laboratory observations of a gravity current on a
sloping bottom: the generation of shelf waves. J. Fluid Mech. 172, 373–399.
Xing, J. & Davies, A. M. 1999 The effect of wind direction and mixing upon the spreading of a
buoyant plume in a non-tidal regime. Cont. Shelf Res. 19, 1437–1483.
Yankovsky, A. E. & Chapman, D. C. 1997 A simple theory for the fate of buoyant coastal
discharges. J. Phys. Oceanogr. 27, 1386–1401.