Physical Limnology
Physical Limnology
Physical Limnology
Physical Limnology
JoRG IMBERGER
and
JOHN C. PATTERSON
Centre for Water Research
The University of Western Australia
Nedlands, Western Australia
1. Introduction
(1987), and Muller and Garwood (1988), which discuss the motion of a
stratified fluid and mixing due to turbulence. There also have been two
recent major conferences dealing with the oceanic internal gravity waves
(Muller and Pujalet, 1984) and the dynamics of the oceanic surface mixed
layer (Muller and Henderson, 1987). Much of this material is relevant to
an understanding of the hydrodynamics of reservoirs and lakes. To avoid
duplication, this review concentrates on smaller scale motions and the
resulting mixing. A description of the action of rotation has therefore
been almost completely omitted since it concerns the dynamics of internal
waves, and it is adequately described in Hutter (1984,1986) and Stocker
and Hutter (1986). Circulation and mixing in ice-covered waters is an
extremely important subject for a large number of lakes, but this has also
been omitted because of the existence of the review by Carmack (1986).
Because of the rapid development of the subject, much of the
description in this review is speculative and concentrates on findings at
the Center for Water Research. Data from the Wellington, Canning, and
Harding Reservoirs in Western Australia are used as illustrations of
processes. These data sets have been used to set the stage for the review
and literature on the dynamics displayed by these data has been
referenced extensively.
The review begins with a discussion of seasonal behavior and the role
of the geographic location of a lake. Through a new, dimensionless
parameter the dynamics of a lake in the tropics at the time of peak
stratification can be compared with the dynamics of a lake in the
temperate zones at the beginning and end of the stratification cycle. This
section is followed by four sections on the dynamics of the surface layer
(formerly called the mixed layer) beginning with a discussion of the
surface fluxes, followed by the vertical dynamics of the surface layer.
Upwelling, differential deepening, and differential heating and cooling
are presented in separate sections as deviations from the vertical picture.
Our knowledge on outflow has increased considerably since the last
review by Imberger (1980) and this material is updated here. There are
also a number of significant developments in our understanding of the
inflow dynamics, particularly where the inflow separates from the river
channel.
Mixing below the surface layer is also discussed. Because of recent
reviews by Gregg (1987), Hopfinger (1987), and Thorpe (1987), in-
dividual mechanisms are not discussed, and the discussion concentrates
on the underlying dynamics contributing towards mixing in the hypolim-
306 Jorg Zmberger and John C. Patterson
200
-
N
2
X
i *
+
z
w
m
YI
-200
-400 I I I 1 I I I I I I I I I I
J J A S O N D J F M A M J J A S
MONTH
1986 1987
FIG. 1. Net heat flux at the surface of the Canning Reservoir. The data were collected
with a meteorological station mounted on a specially designed buoy kept in position with a
taut mooring. The sensors were 2.0 m above the water level at all times. The smooth line
represents the averaged flux, and the tick marks represent the start of the month.
50.0
I
J
I
J A
I
S
I
O
I I
N
I
D
I
J
1
F
I
M
1
A
I
M
I
J
I
J A
I
MONM
1986 1987
FIG.2. The thermal characteristicsof the water in the Canning Reservoir corresponding
to the net heat flux shown in Figure 1. Drops were collected at fortnightly intervals
throughout the 15 months.
end of December, the temperature at the surface had reached 24°C over
a depth of nearly 10 m. The stratification during this period appeared to
be unable to suppress the turbulence in the deep part of the lake between
20 and 30m depth, where the temperature rose from a low just above
11°C in September to 12°C in December. After December, the warmer
surface temperatures increased the back radiation, the evaporation, and
the sensible heat loss, and the surface temperatures rose only slightly,
reaching a maximum of 25°C in late February.
After this time, penetrative convection, introduced by a net surface
cooling, led to a characteristic decrease of the surface water temperature
of the top 15m, which continued from late February to June. Also
noticeable was the gradual deepening of the isotherms at the base of the
surface layer, induced by the erosion at the base by both natural
convection and wind-induced turbulence. Surface cooling has two effects.
First, the kinetic energy of the penetrative convection thermals erodes
the stable structure underlying the surface layer. Second, the cooling
decreases the stability (the density difference between the surface layer
and the underlying water), thus making the base more susceptible to
erosion by turbulence from within the surface layer. By June, the lake
again had become isothermal.
In limnology textbooks (Hutchinson, 1957; Wetzel, 1975; Henderson-
Sellers, 1984b), the upper warm layer is called the epilimnion, the strong
temperature gradient region below is called the metalimnion, and the
colder water in the lowest part of the lake the hypolimnion. These terms
are useful for description, but as lmberger (1985) pointed out, the
understanding of the properties of water behavior in these regions has
recently been revised.
As the following sections will show, the surface layer is not as well
mixed as is generally believed, nor is it in a state of uniform and constant
turbulence. The surface layer, or epilimnion, responds to diurnal changes
in surface fluxes. The metalimnion contains the major temperature
changes between the surface and bottom water. It is made up of many
temperature steps, and it is unrealistic to define the thermocline as that
depth at which the temperature gradient is maximum, as has been
suggested by previous authors (see Wetzel (1975) and Straskraba (1980)).
Typical profiles from the data set of Figure 2 are shown in Figure 3. The
temperature gradient is a strong function of the resolution of the
instrument. For the smoothed profile with a resolution roughly matching
the vertical extent of the interface (Figure 3b), the gradient reaches only
w
b C
5;
L I I I I I I I I 1140
11 18 25 -15 15 -15 0 15 -15 0 15
TEMPERATURE StK dED FINESCALE MICROSTRUCTURE
(.C) dTMz ('C/rn) dT/dz ('C/rn) dT/& ( Q n )
FIG.3. Typical profile data obtained in the Canning Reservoir on 26 January 1987 at a central location. (a) Temperature obtained
with a standard conductivity, temperature, depth profiling instrument. (b) Smoothed temperature gradient. (c) Enhanced temperature
gradient with an effeective resolution of 0.02 m. (d) Temperature gradient obtained with a microstructure profiler which had a resolution
of 1 mm. This profile was taken a little earlier than that shown in (a), (b), and (c) and only extended to a depth of 20m.
Physical Limnology 311
TEMPERATURE (“C)
RG. 4. Evolution of a diurnal surface layer. Data taken in the Wellington Reservoir on
25 February 1985. Solid line: data collected at 0824 hours during calm conditions. Dashed
line: data collected at 1501 hours at time of maximum surface temperature gradient. Dotted
line: data collected at 1701 hours; cooling had commenced.
[ 60
y = 240
for fp > 0 (northern hemisphere),
for fp < 0 (southern hemisphere),
and t is the Julian day.
Physical Limnology 313
LAKE
ARGYLE >
J F M A M J J A S O N D
MOMH
FIG.5. Mean surface water temperature for the Wellington Reservoir and Lake Argyle.
Solid line: data. Dashed line: empirical relationship (2.1) suggested by Straskraba (1980).
314 Jorg Zmberger and John C . Patterson
40.001
where z is the vertical coordinate from the bottom of the lake, A(z) is the
area of the lake at height z , p(z) is the water density at a height z , z, is
the water depth, and zg is the height to the center of volume of the lake,
defined by
Straskraba (1980) suggests that the depth of the center of the meta-
limnion and the rate of turbulent mixing depend on the magnitude of S,.
He also observed that for a particular lake basin undergoing temperature
stratification given by (2.1) and (2.2), S, should be small near the
equator, rise to a maximum at middle latitudes, and again reach zero at
latitudes of approximately 70"; the actual stability being dependent on the
density of the water which is a highly nonlinear function of the
temperature (UNESCO, 1981). Idso (1973) reintroduced this concept of
stability, but multiplied S, by gravity g and divided (2.3) by A(z,),
removing the absolute size of the lake from the definition of the stability.
However, even with these modifications, S, is still dimensional. Further,
the concept of stability expressed by (2.3) is deficient since it only
accounts for the stabilizing influence of the stratification and not the
destabilizing influence of the disturbances of wind, inflows, and outflows.
The problem can be addressed by generalizing the surface layer
stability criterion first introduced by Sverdrup (1949, developed further
by Hurley Octavio et al. (1977), Spigel and Imberger (1980), Thompson
316 Jorg Imberger and John C. Patterson
W = g’h2
- (2.5)
u2,L ’
where g’ is the modified acceleration due to gravity across the base of the
surface layer, h is the thickness of the surface layer, u , is the water
friction velocity due to wind stress, and L is the fetch length. Here
g’ = g Aplp,, where Ap is the density jump across the base of the
surface layer and po is the hypolimnion density. The friction velocity u, is
defined by (3.1) below. This number is dimensionless and represents, as
explained in Imberger (1985) and Imberger (1987), the ratio of the
baroclinic pressure force g’h2 at the point of upwelling and the surface
force u:L imposed by the wind stress; both forces being calculated per
unit width of the lake. This number will be elaborated upon in Section
IV, but a simple generalization is suggested to describe the behavior of
the lake as a whole.
Consider a general lake with an arbitrary stratification p ( z ) being acted
upon by a general wind field with a surface friction velocity u , ( x , y ) ,
where x and y are the horizontal coordinates embedded in the lake’s
surface. As this wind stress is imposed on the surface layer, there will be
a net force acting to overturn the density structure of the water column.
Taking moments about the center of volume located at zg, we obtain for
equilibrium
where zo is the center of gravity of the actual water mass with a density
stratification p ( z ) , M is the total mass of water, and /3 is the angle
subtended to the vertical by the center of mass at the center of volume.
This leads to the ratio we shall call the Lake number
(2.7)
Following the work of Spigel and Imberger (1980), the angle /3 may be
fixed so that LN is calculated at the point where upwelling commences,
that is when the metalimnion intersects the surface, so that
(2.10)
where zT is the height to the center of the metalimnion, and the fetch is
scaled with A'/*(z,,,), yielding the relationship
(2.11)
Assuming that the wind stress is constant over the surface, then LN
reduces to
(2.12)
where A. is the surface area of the lake A&). For large Lake numbers,
the stratification will be severe and dominate the forces introduced by the
surface wind stress. Under these circumstances, stratification is expected
to be horizontal, with little or no seiching and little turbulent mixing in
the metalimnion or the hypolimnion. Horizontality of the isotherms at
large LN is illustrated by the data set taken at Canning Reservoir on the
18 January 1987, which is depicted in Figure 7. A weak, warm lense of
water resident in the sheltered upstream part of the basin and a very
weak slope of "C m-l towards the dam wall, attributable to a first
mode seiche, are the only departures from this horizontality. As
discussed in Shay and Imberger (1988), the turbulence levels in the
epilimnion and the hypolimnion were extremely low. At the time, the
wind mean shear velocity (averaged over 24 hours) was 0.0033 m s-',
leading to a value of L , = 86. The full implication of LN with respect to
deep mixing will be discussed in Section X.
The interplay between the temperature distribution variation with
latitude and the changing water density due to the elevated temperatures
318 Jorg Zmberger and John C . Patterson
TEMPERATURE ("C)
0.0
- 26.0'
10.0
--E
f 20.0
x
30.0
40.0 470 CA65 CA60 CA50 CA45 CAM CA35 CA30 CM5 CA20 CAiC
A j A a A t u
I I
1000.0 2000.0 3000.0 4000.0 5000.0 6000.0
CUMULATIVE DISTANCE
9000
8000
7000
POISON GULLY
6000
- 5000
w
VI
+
a
z0
4
K
400C
8
J
a
s
0
300C
2000
CANNING RESERVOIR
1000
0000
FIG. 7. (conld.)
320 Jorg Zmberger and John C. Patterson
FIG.8. An example of the variation of the Lake number as a function of season and
latitude. The surface was constructed by computing the Lake number from data gathered in
the Canning Reservoir, but which was transposed to different latitudes using the empirical
relationship (2.1) suggested by Straskraba (1980). A constant wind speed of 8.0ms-' was
applied throughout.
Steinhorn and Gat, 1983); to small shallow lakes (MacIntyre and Melack,
1982; Bunn and Edward, 1984). The Wellington Reservoir is an example
(see Imberger, 1987) where a strong variability of the salinity of the
inflow led in some years to meromictic behavior.
In general, the dissolved salts influence the dynamics of the lake only
through the equation of state, and provided the Lake number and other
parameters allow for salinity variation, their dynamics are similar to
single species stratification. In the section on deep mixing, we discuss an
exception to this statement, and under certain circumstances, such as in
Lake Kivu (Newman, 1976), double diffusive instabilities may arise in
salt-stratified lakes due to different rates of molecular diffusion between
temperature and salt (see Turner (1985) for a complete review on
multi-components).
Many attempts exist to empirically correlate the dynamics of a lake to
variables such as the surface area (Ward, 1977) and geographical trends
such as altitude (Hutchinson, 1957; Wright, 1961; Loffler, 1968; Khom-
skis, 1969; Lerman and Stiller, 1969; Bella, 1970; Lerman, 1971;
Darbyshire and Colclough, 1972; Khomskis and Filatova, 1972; Blanton,
1973; Idso and Cole, 1973; Lewis, 1973; Sundaram and Rehm, 1973;
Tzur, 1973) and length of the lake (Yoshimura, 1936; Patalas, 1960 and
1961; Arai, 1964; Ventz, 1973). These correlations are useful only in that
they are special cases of the lake number LN formulation.
The interaction between the atmosphere and a lake occurs at the lake’s
surface. A great deal of work has been done to determine the fluxes of
momentum, energy and mass across the air-water interface and many
numerical formulae have been developed. These methods are reviewed
by Tennessee Valley Authority (1972), Straskraba (1980), and
Henderson-Sellers (1986).
Short wave radiation (300 nm to 1000 nm) is usually measured directly.
Long wave radiation (greater than 1000nm) emitted from clouds and
atmospheric water vapor can be measured directly or calculated from
cloud cover, air temperature, and humidity (Tennessee Valley Authority,
1972). The reflection coefficient, or albedo, of the short wave and long
wave radiation varies from lake to lake and depends on the angle of the
sun, the color of the water, and the surface wave state. Back radiation
322 Jorg Zmberger and John C. Patterson
from the warm water surface may be calculated from the black body
radiation law, it also may be measured by pointing a radiation instrument
towards the water surface.
It was shown by Strub and Powell (1987), Marti and Imboden (1986),
Keijman (1974), Sadhuram et al., (1988), and many other authors that
simple aerodynamic bulk .,formulae, with constant transfer coefficients,
can be used to calculate the momentum, the sensible heat, and the latent
heat fluxes on a seasonal basis:
E -
-= q ' w ) = -u*q* = -C,U,(q, - q.), (3.3)
PLV
where z is the surface stress, p is the density of the air, u' and w' are the
horizontal and vertical fluctuation of velocity, the overbar is a time
average long enough to average out the short term energy but short
enough to allow for synoptic variability (Busch, 1977; Smith, 1980;
Geernaert et al., 1987), u , is the shear velocity in the air, CD is the
momentum or drag coefficient, U, is the air velocity at a certain height z
above the water surface, H is the sensible heat transfer, Cp is the specific
heat of water, 6" is the temperature fluctuation, 8, is the temperature
scale, CHis the heat transfer coefficient, often called the Stanton number
(Geernaert et al., 1987), 8, is' the water surface temperature, 8, is the
temperature of the air at the height z above the water, E is the latent heat
flux, Lv is the latent heat of vaporization, q' is the specific humidity, q* is
the specific humidity scale, Cw is the latent heat transfer coefficient or
Dalton number (Geernaert et al., 1987), q, is the specific humidity at a
height of z above the water surface, and q. is the specific humidity at
saturation pressure at the water surface temperature.
Equations (3.1), (3.2), and (3.3) are only valid for a stationary surface.
Since the water moves, U, should be replaced by U, - Us, where Us is the
water surface mean velocity (see Businger (1973) for a full discussion),
although for most wind speeds, this is a small correction.
The value of the transfer coefficients CH and Cw for such bulk
modeling has received a great deal of attention and, in the case of lakes,
may be estimated by carrying out a heat and water budget over seasonal
Physical Limnology 323
time scales (for example, Strub and Powell (1987) for Castle Lake, Marti
and Imboden (1986) for Lake Sempach, Myrup el al. (1979) for Lake
Tahoe, and Taylor and Aquise (1984) for Lake Titicaca). These
references suggest a value of CH= Cw = 1.9 X (referenced to 10 m)
yields an adequate description of the total stored energy over a seasonal
time scale. This may be compared with a lower value of 1.45 x
suggested by Hicks (1972). Constant values are sufficient for seasonal
lake modeling since the thermal budget is self-regulating; an underestim-
ate of the heat loss will cause the water surface to heat, thus increasing
heat loss, while an overestimate of heat loss will cause the water surface
to cool and, by (3.2) and (3.3), this will lead to the heat loss being
decreased.
At smaller time scales (from hours to days), this negative feedback
mechanism no longer suffices because the thermal inertia of the water in
the surface layer is too great. The influence of the air column stability
and the water surface roughness must therefore be accounted for, and
these may introduce considerable variability to all three transfer
coefficients. Excellent reviews exist on this topic: Businger (1973), Dyer
(1974), Stewart (1974), Garratt (1977), Wu (1980), Donelan (1982), and
Blanc (1985). There are three main measurement techniques to deter-
mine the instantaneous surface fluxes at the water surface: eddy
correlation, profile, and dissipation. These have been used to calibrate
(3.1) to (3.3) thus giving values for CD, CH, and Cw. A description of
each technique is given by Dobson et al. (1980) and Blanc (1983). The
eddy correlation method is direct; the profile method uses mean velocity,
temperature, and humidity profile data; and the dissipation method uses
the spectral characteristics of the high wave number turbulence. Results
from the eddy correlation technique usually agree with each other within
10% (Blanc 1987); profile techniques usually agree with eddy correlation
methods to within 25% (Miyake et al., 1970; Wucknitz, 1976). Large and
Pond (1981) and Smith and Anderson (1984), in comparisons between
the dissipation method and the eddy correlation technique, found that
the two agreed to within 40%. Within similar schemes, there can be even
greater variability (see Lo and McBean (1978) for interprofile
comparisons).
The reliability of the bulk aerodynamic equations (3.1) to (3.3)
depends on the sensors being located in the internal boundary layer of
the lake. Air coming from land will meet a different roughness, humidity
and temperature over the lake and the internal boundary layer will be
324 Jorg Imberger and John C. Patterson
kz d e
(3.7)
e, dz
kz dq
Dyer (1974) summarized the various proposed forms for the functions
@ M , @ H , and GW. For a convecting boundary layer ( z / L< 0),
(3.10)
Hicks (1976) reanalyzed the Wangara data (Clarke et al., 1971) and
found (3.11) to hold for only weakly stable (0 < z / L < 0.5) boundary
326 Jorg Imberger and John C . Patterson
@M [
= 8 - 4.25 ):( I'-+):( . (3.13)
Paulson (1970) substituted (3.9) to (3.13) into (3.6) to (3.8) and carried
out the integration to obtain
(3.14)
(3.15)
(3.16)
= [ - 16(;)] 114
, (3.17)
for the unstable case,
l+x
+ In( y - 2 tan-'(x)
VM= 2 In( 7)
1+x2
) + -,n2 (3.18)
-5(3
z
0 < - < 0.5,
L
(3.20)
0.5(i)-2-4.25(;)-1-71n(i)
These relationships appear to have been tested over the range from
-15 < z / L < 15, but for the values larger than 15, the similarity functions
must be extended as suggested by Sorbjan (1986).
Equations (3.18) to (3.22) can now be used to relate the neutral to the
actual transfer coefficients, leaving only the surface roughness to be
estimated. The roughness lengths are related directly to the drag
coefficients, as can be seen by substituting (3.14) to (3.16) into (3.1) to
(3.4). Under neutral conditions,
(3.23)
(3.24)
(3.25)
(3.27)
(3.28)
(3.29)
(3.30)
328 Jorg Zmberger and John C. Patterson
(3.31)
(3.33)
(3.34)
(3.35)
C W - e-24 Rie
7 (3.36)
CWN
(3.37)
(3.38)
Physical Limnology 329
a = 0.83C;:62, (3.39)
b = 0.25C,$'. (3.40)
32
B 31
30
_-11
70
15 19 23 27 31
TEMPERATURE ib)
-Q
2.L
fw/m2)
-
110-31S.W. RADIATION (10-3) NET RADIATION LATENT HEAT FLUX SENSIBLE HEAT FLUX
(W/m2)
(10-3) TOTAL HEAT
FLUX (W/m2)
31 -
30 -
between 0.0123 and 0.351. Garratt (1977) added the data from Kitaigor-
odski et al. (1973), Wieringa (1974), Kondo (1975), and Smith and Banke
(1975) to obtain a value of (Y equal to 0.0144; a value that also seemed to
fit inferred drag coefficients applicable to wind speeds up to 50ms-'.
More recently, Wu (1980) arrived at a value of 0.0185. Large and Pond
(1981) derived a value of CDN= 0.012 for wind speeds between 4 and
11m s-l, but increasing with wind speed beyond 11m s-'. Equating the
two formulations at a wind speed of 25 m s-' by using (3.27) leads to a
value of (Y = 0.0097, considerably smaller than suggested by most other
investigators. At low wind speeds (less than 5ms-'), the situation is
unclear, but in general, the results from the above investigators suggest a
constant value of 1 X for C,, (see Hicks (1972)). In general, the
inter-technique variability and data scatter (Blanc, 1985, 1987) outweigh
the above differences, but the most recent values of a; estimated by Wu
(1980), (0.0185); Geernaert et al. (1986), (0.0178); and Geernaert et al.
(1987), (0.0165) appear to suggest a mean value of (Y equal to 0.0175 for
a fully developed wave field.
The large variability in the surface roughness has recently received
considerable attention. Kitaigorodskii (1968) and Kitaigorodskii and
Zaslavskii (1974) already indicated that Charnock's (1955) relationship
(3.29) should be modified to include the wave age Co/u,:
(3.41)
FIG. 9. Example of typical heat flux variations computed from data shown in Figure 1.
Day 29 is 29 January 1987. (a) Surface water temperature. (b) Air temperature. (c)
Short-wave incoming radiation. (d) Net total radiation. (e) Latent heat flux. (f) Sensible
heat flux. (g) Total heat flux at the water surface. (h) Drag coefficient C,. (i) Heat and
mass transfer coefficients CH= C, . (j) The bulk Richardson number.
332 Jorg Imberger and John C . Patterson
duration winds) generally require higher values for the neutral drag
coefficient.
Hsu (1974, 1986) and Geernaert et al. (1987) derived a relationship
between the wave age and wave steepness:
(3.42)
where is defined above and Lw is the wave length at the peak of the
energy spectrum. By using both statistics from the JONSWAP formula-
tion and from Chapter 3, Section 4, of the U.S. Army Corps of Engineers
(1977), the wave steepness for equilibrium seas becomes 0.023, yielding a
value of a slightly higher than that proposed by Wu (1980) (ais equal to
0.0185). Graf et al. (1984) investigated drag coefficients for shoaling
waves and large wave steepness and showed that water roughness using
(3.42) yielded good results. Geernaert et al. (1987) carried out a direct
correlation between the neutral drag coefficient C D N and the wave age:
indicating again an increased coefficient for small values of the wave age
(a developing wave surface). The relationships (3.41), (3.42), and (3.43)
also explain the reduction of drag coefficient in shallow water where
Co/u, tends to be larger (Hicks et al., 1974; Emmanuel, 1975; Geernaert
et al., 1987).
Donelan (1982) partitioned the water surface roughness into two parts:
2, = p[ E ( w )dw]lD, (3.44)
and
(3.45)
where wp is the frequency at the peak of the spectrum and /Iis a constant
equal to 0.0125. He then defined a neutral drag coefficient
Physical Limnology 333
that the wind field over a lake can be extremely variable: this has
enormous consequences for the behavior of the surface layer, as will be
seen in Section VI.
FREQUENCY (cyclesh)
Ro. 10. Power spectral density of the wind speed during the period shown in Figure 9.
Seventeen days of data were used to compute the spectrum.
Physical Limnology 335
3
dT/dz (“C/m)
FIG. 11. (a) Examples of microstructure profiles collected at the entrance of Salmon
Brook, Wellington Reservoir on 26 February 1985. See Figure 26a for station locations.
Temperature:
I. 0928 SB20
11. 0939 SB20
111. 1002 c10
IV. 1021 c10
V. 1057 C10
VI. 1726 SBlO
(b) Microstructure temperature gradient signals corresponding to data in Figure 1la.
Physical Limnology 337
turbulence was established at small scales and was not the result of a
large overturn as has been normally associated with turbulence in a
stratified fluid (Gregg, 1987). Imberger (1988) attributed this rapid
spread of the energy to the leakage by internal wave action from the base
of the small surface mixed layer. Twenty minutes later, and at a
somewhat more exposed site further away from the shore, the surface
layer had deepened to about 0.3 m and the temperature distribution
within the lake was more uniform (Figure lla-111). There is clear
evidence of active turbulence extending below the base of the surface
layer to almost 2 m (Figure llb-111). This deepening process continued
and the profile taken another twenty minutes later reveals a “uniform”
surface layer extending to nearly 0.7m. However, as described by de
Szoeke and Rhines (1976), Imberger (1985), and Spigel ef al. (1986), the
interface at the base of the surface layer, initially sharp and sustained by
the surface-introduced turbulence, is now smeared by billow activity
energized by the momentum of the surface layer, which is clearly
illustrated in Figures lla-IV and llb-IV. The leakage through the base of
the mixed layer remained and the turbulent activity extended to nearly
2 m.
During the measuring period, the lake was exposed to a very strong net
heating flux of around 750 W m-*. At approximately 1030 hours, the
wind intensified, reaching a peak of nearly 6 m s-l at about 1100 hours. A
well-defined diurnal thermocline formed at 1.2 m, reflected both in the
temperature (Figure 1la-V) and the temperature gradient signal (Figure
llb-V), but two features are noticeable. First, the heat flux introduced
sufficient buoyancy at the surface to prevent a completely mixed surface
layer from forming, but the temperature gradient signal shows strong
turbulence in this surface layer. Second, the temperature gradient signal
continued to show turbulence below the base of the surface layer,
extending to about 2.0 m.
The wind strength remained at about 4-6 m s-l for the remainder of
the day and only began to decrease at 1700 hours. At the time the last
profile was collected, the wind had decreased to about 3ms-’. As seen
from Figure lla-VI, the surface layer was not well mixed, even though
the turbulence obviously extended down to 2.3m (see Figure llb-VI).
The gradient signal shows an intermittent, patchy turbulence within the
surface layer, although there was a well-defined wind-stirred near-surface
region extending down to 0.06m. The water column below 2.2m
appeared to have become quiescent.
338 Jorg Imberger and John C. Patterson
a b
I I I 1 I
I I I 1 I I
L2:.5 ’ 22.0 22.5 23.0 -50 0 50
FIG. 12. Temperature microstructure data from the Wellington Reservoir (13 March
1982) at the central basin after a very strong sea breeze had deepened and tilted the surface
layer. (a) Temperature. (b) Temperature gradient. (After Imberger (1985).)
Now consider an example (Figures 12a and 12b) from the diurnal
sequence documented in Imberger (1985). The profile was taken at a
time when there was a net heat loss of about 250Wm-’, a strong
wind-induced surface layer velocity of about 0.12 m s-’, and an active
wind stress. As seen in Figure 12, the surface layer was extremely
well-defined (variation less than O.O3”C),the base of the mixed layer was
diffuse due to billow activity (Imberger, 1985), and the water below the
base of the surface layer was relatively quiescent.
The third example, shown in Figure 13, also comes from the diurnal
data set presented in Imberger (1985) but was collected when natural
convection completely dominated the surface layer dynamics. The profile
is characterized by a well-mixed surface layer with obvious thermals
Physical Limnology 339
a b
falling through the full extent of the layer and impinging on a sharp
interface.
In summary, these examples show that it is difficult to define the
surface layer in terms of the temperature profile. It is better to define the
layer as that depth of water directly energized by the surface fluxes and
the mean shear of the surface layer. The data also show that the
temperature, and thus the density, is rarely absolutely uniform in the
surface layer, especially when there is strong surface heating or when the
layer is retreating (wind stress reducing with time). Lastly, the surface
fluxes not only introduce turbulence into the surface layer but also, via
leakage through the base of the surface layer, into the deeper waters
making the above definition of the surface layer somewhat ambiguous.
Some of this surface-introduced energy energizes the water column as a
340 Jorg Irnberger and John C . Patterson
whole, but the exact fraction has not been documented in field
measurements.
The overall dynamics of the surface layer is determined by the
magnitude of the surface Wedderburn number (Imberger and Hamblin,
1982) given by (2.5). For W>>1, tilting of the isotherms due to the
applied wind stress will be small and horizontal variations are negligible.
This corresponds to strong stratification, light winds, and slow deepening
of the mixed layer dominated by surface-introduced turbulence reaching
the base of the mixed layer and eroding the interface. For W ((1,
deepening is dominated by internal shear production and occurs on a
time scale much shorter than horizontal convection in the surface layer.
This leads to a sharp interface downwind and a broad upwelling at the
upwind end of the lake (Monismith, 1986). In the initial classification (see
Spigel et al. (1986)), this regime was misinterpreted as having a
horizontal interface due to the envisaged rapid deepening. This question
is addressed in the next section. For intermediate values of the
-
Wedderburn number, W 1, upwelling and horizontal mixing become
important (Spigel and Imberger, 1980). This was confirmed by the model
of Imberger and Monismith (1986), the laboratory finding of Monismith
(1986) and by the field data of Imberger (1985) and Strub and Powell
(1986).
In the case where W > > l , the processes are essentially one-
dimensional, and as shown by Spigel (1980), the layer deepening can be
uncoupled from any internal seiching that may exist. There is a long
history of interest in such a one-dimensional surface layer. Munk and
Anderson (1948) discussed earlier observations and presented a theory
based on an eddy diffusion coefficient, dependent on stability, which
enabled the prediction of the surface layer behavior. The central task in
understanding the surface mixed layer is to quantify the rate at which the
kinetic energy of the turbulent velocity fluctuations is converted to
potential energy within and at the base of the surface layer as denser
water from beneath the layer is entrained and then mixed into the surface
layer. The turbulence available for this mixing may be generated at the
surface through pressure work, shear production, and wave breaking
(Kraus and Turner, 1967; Kraus, 1977; Cavaleri and Zecchetto, 1987;
Muller and Garwood, 1988) or in the surface layer and the base by shear
production (Pollard et al., 1973; Imberger, 1985). The transport and
redistribution of the turbulent kinetic energy within the surface layer is
dominated by large scale or secondary motions if present. The most
Physical Limnology 341
0.10
0.20
n
a
5
0.30
BURST 19
0.40
0.10
0.20
0.30
0.40
0.50
me-’
FIG.14. Two examples of dissipation as a function of depth taken in the ocean and
compared with the similarity scaling for turbulence produced by both wind stress and by
convection. (From Muller and Garwood (1987); data originally from Gregg, unpublished.)
the water column; i.e., both processes cause mixing. Billowing utilizes
mean flow kinetic energy that would otherwise be available to produce
mixed-layer turbulence for entrainment, but at the same time weakens
the density gradient at the base of the mixed layer. Billowing reduces the
energy required for further mixed-layer deepening, since the energy
required to entrain heavier fluid is roughly proportional to the strength of
the density gradient across the base of the mixed layer. There thus arises
a rather complex and unsteady interaction whereby mixed-layer deepen-
ing sharpens a gradient, making it unstable to shear so that billowing
occurs. Billowing weakens the gradient, which is then more easily eroded
by further deepening, which leads to further billowing.
One of the principal goals of the model by Rayner (1981), and later
Spigel et al. (1986), was to demonstrate that the interaction between
billowing and deepening can be successfully modeled by accounting
simultaneously for the energetics of surface stirring, shear production,
billowing, and variability in the store of turbulent kinetic energy.
In these simple one-dimensional integral models, it is assumed that the
surface layer may be approximated in the integral sense by three distinct
zones (Niiler and Kraus, 1977; Spigel et al., 1986): a comparatively thin
constant stress surface layer of thickness y (see Figure 15) where
turbulent kinetic energy is produced and then exported to the fluid
below; a uniform central layer in which part of the surface energy is used
to mix the fluid and is lost to dissipation; and a thin front of thickness
ATMOSPHERK: TRANSFERS
p s r/”L p
3
/fWlHL
?
V
Z-H
-
F
rJ SURFACE
DRIFT
ZONE
WELL MMED LAYER
S, (FULLY TURBULENT)
HYPOLIMNION
Definition sketch of the surface layer model. (After Spigel et al. (1986))
346 Jorg Imberger and John C. Patterson
6 at the base of the surface layer. This transition marks the pronounced
density jump described above as the diurnal thermocline. Here, the
remainder of the turbulent kinetic energy generated at the surface plus
that generated internally by shear and less that which is locally dissipated
or radiated downwards by internal waves (Linden, 1975; E and Hopfin-
ger, 1986; Imberger, 1988) is used to entrain underlying fluid into the
central layer above. Using the symbols defined in Figure 15, the
governing differential equations (Niiler and Kraus, 1977; Spigel et al.,
1986) may be written as
dS
-=
a-
- -(s‘w’), (4.3)
at az
ap - - - ( ap ’ w-
_ ’)+--, a aQ
at dz cpaz (4.4)
au
-=
a-
- -(u’w’),
at 3.2 (4.5)
The leakage or transfer of turbulent kinetic energy through the base may
be written
(4.12)
As seen from Figures 11, 12, and 13, this is usually nonzero. Imberger
(1988) postulated that the mechanism for turbulent energy transport was
by internal wave propagation, away from the base of the surface layer
along rays. Linden (1975) estimated from laboratory experiments that up
to 50% of the turbulent kinetic energy budget arriving at the base of the
mixed layer seeps through the base. By constrast, E and Hopfinger
(1986) suggested that while internal waves were generated, energy
radiation generally did not affect the entrainment rate. Scaling suggests
that the leakage A L be parameterized by
AL = CLA3N3 (4.13)
where CL is a coefficient, A is the amplitude of the internal waves
generated, N is the buoyancy frequency below the base, and h is the
depth of the mixed layer. At present, there are no conclusive data that
allows the determination of the correctness of (4.13) or the magnitude of
CL,and for this reason, Spigel et al. (1986) assumed CL= 0. It is worth
remarking that since their model gave very good comparisons with
observed field data, the leakage term must be proportional to the flux of
turbulent kinetic energy arriving at the interface, which is seen by noting
-
that A E:n/N.
Spigel et al. (1986) followed Niiler and Kraus (1977) and integrated
(4.2) to (4.6) from the bottom to the surface of the lake using the
boundary conditions given by (4.7) to (4.12). Further, closure was
achieved by introducing a set of efficiencies for each of the source and
sink terms. Shear production in the base layer minus the dissipation was
348 Jorg Zmberger and John C . Patterson
approximated by
Uzdh 1 d U2dh 1 d
--+--(U?h)- cbh = Cs 2-+--(Uzh)], (4.14)
2 dt 12dt 2 dt 12dt
where Eb is the average dissipation in the base layer and Cs is the
efficiency of production of turbulent kinetic energy. Experimental evi-
dence that allows the evaluation of Cs is sparse. Sherman et al. (1978)
reanalyzed the experiments of Kato and Phillips (1969) and Kantha et al.
(1977) to arrive at a value of between 0.2 and 0.5. Imberger (1985)
presented field data giving a value of 0.24 and Spigel et al. (1986)
obtained best simulation results with a value of 0.2.
The turbulence introduced at the surface constant stress layer may be
parameterized (Kraus and Turner, 1967; Niiler and Kraus, 1977) as
(4.15)
where c is the velocity scale in the surface drift layer (see Figure 15), CN
is the efficiency of energy production at the water surface, and u* is the
water shear velocity.
The source of buoyancy flux due to surface heating may be written
(Deardorff, 1970; Zeman and Tennekes, 1977; Rayner, 1981) in the
form
w: = g ( h + !){
2
[
Ly Q ( H ) - Q ( 5 -
POCP
): ] + ( y e " ( H ) }
+ pwss- -
2*g
POCP
1"
5-m
Qdz. (4.16)
The dissipation in the mixed layer may be written (Spigel et al., 1986)
where E, is the average dissipation in the mixed layer. This form was
introduced by Mahrt and Lenschow (1976); similar forms were suggested
by Garwood (1977) and Zeman and Tennekes (1977), as well as by the
results of Willis and Deardorff (1974). The alternative parameterization
used by Niiler and Kraus (1977), Sherman et al. (1978), and others
consists of introducing a set of efficiencies to reduce each energy source
by a certain fraction; an approach that does not allow for the storage of
Physical Limnology 349
Given these assumptions and carrying out the integration yields (Spigel et
al., 1986) the necessary set of equations for the unknowns U s , h, 0, , S, ,
6, and E,:
dEs
h-=-(C,+
dt
CE)E,'"+ W: + C&:, (4.19)
(4.21)
(4.22)
(4.24)
Comparison with (4.1) indicates that the depth of the mixed layer scales
the same as 6, which means that for a particular Us the depth is fixed and
cannot continue to deepen, contrary to (4.23). This was the basis of the
original Pollard et al. (1973) model, and may be explained by noting that
once a particular surface layer has a certain shear across its base, both h
and 6 will adjust until (4.1) and (4.24) are satisfied, after which time
there will be no further adjustment (Sherman et al., 1978; Lawrence et
al., 1987; Thorpe, 1987). The experiments of Narimousa et al. (1986) and
Narimousa and Fernando (1987) rely on a surface energy source u', that
will continually sharpen the interface and reduce 6, and the entrainment
law (4.23) therefore depends on the type of experiment and must be used
with caution.
This is best seen by substituting (4.22) into (4.24) and noting that g'h is
constant for no net heat flux, which then yields
1 dh Cy"
where Ri = g'hlu:. (4.25a)
u, dt -Ri'"'
the spin-up term in (4.25) and defining ii as the mean (over the depth of
the mixed layer) root mean square velocity allows (4.25) to be rewritten
ldh
--== CF
(4.26)
ii dt Ri’
where
(4.27)
Nokes (1988) has done a thorough review of data from all available
grid experiments and concluded that the rate of non-dimensional
deepening ranges in proportionality from Ri-’ to Ri-’.75. In this context,
the results from Kranenburg (1984) and Murota and Michioku (1986b)
give convincing evidence that the non-dimensional deepening is propor-
tional to Ri-l. In all recent work, investigators have defined the
Richardson number Ri=g’l/u*, where 1 is the integral scale of the
turbulence at the position of the interface (but in the absence of the
interface) and u is the root mean square velocity of the turbulence
arriving at the position of the interface. Unfortunately, there are no data
available on the relationship between the depth of the mixed layer and
the integral scale for surface layers so we cannot take this comparison any
further. It would be constructive to carry out an integral analysis of the
grid configuration to determine the relationship between ii and u, and to
establish if (4.26) would lead to the relationship suggested by Nokes
(1988):
l d h - 0.15
u dt Ri-’.’ * (4.28)
This is now possible as all the necessary spatial variations can be derived
from Hopfinger and Toly (1976), Hannoun and List (1988), and Hannoun
et al. (1988). Last, it is important to reconcile, as discussed by Nokes
(1988), the difference in the measurement of the entrainment rate dhldt:
indirectly from a measure of the change of buoyancy in the surface layer
and directly from the propagation rate of the interface.
Murota and Michioku (1986b) also carried out an experiment where
they used grid stirring together with a convective energy source and
successfully compared their results with the model put forward by
Sherman et al. (1978) and embodied in (4.19), although their value for
C , was 2.9, not 1.33.
b
0.0 w
VI
h,
1.0 -
2.0 -
7.0 1 ,/ /
21.50 21.75 22.00 22.25 22.50 22.75 23.00 23.25 23.50 23.75 24.00 21.50 21.75 22.00 22.25 22.50 22.75 23.00 23.25 23.50 23.75 24.00
TEMPERATURE ("C) TEMPERATURE ('C)
C d
0.0 0.0
1.o 1.o
2.0 2.0
x
E
3.0
-E 3.0
4.0 4.0
+ +
w
0 5.0 x 5.0
6.0 6.0
7.0 7.0
8.0
21.50 21.75 22.00 22.25 22.50 22.75 23.00 23.25 23.50 23.75
-
24.00
830
..
21.50 21.75 22.00 22.25 22.50 22.75 23.00 23.25 23.50 23.75 24.00
TEMPERATURE ('C) TEMPEAATURE (%)
FIG.16. Comparison of measured (solid line) and predicted (dashed line) temperature profiles. The profiles b, c, and
d have been corrected for effects of advection. (After Spigel el ul. (1986).)
Physical Limnology 353
V. Upwelling
When the surface wind stress increases, the isopycnals surface at the
upwind end and deepen at the downwind end (Wedderburn, 1912;
Keulegan and Brame, 1960; Blanton, 1973; Stefan and Ford, 1975; Spigel
and Imberger, 1980; Monismith, 1986; and many others). This phenome-
non is called upwelling.
A demonstration of upwelling is given by the longitudinal transects
taken in the Wellington Reservoir (Figure 17a) in August 1988. The
354 Jorg Imberger and John C . Patterson
isopycnals prior to the start of a wind event are shown in Figure 17d; the
wind event started at about 1000hours and lasted for about 6hours
(Figure 17b), with a direction of 320" (Figure 17c), the direction of the
transect. The isopycnals after 4 hours are shown in Figure 17e, indicating
a high degree of disturbance. The hypolimnion was also disturbed but
retained a degree of horizontality. The upwind surface layer showed
strong upwelling with a well-defined horizontal gradient supported by
cold water brought up from the metalimnion.
The idealized density profile assumed by Heaps and Ramsbottom
(1966) and later used by Spigel and Imberger (1980) in their lake
classification scheme consisted of only two layers. This two-layer ap-
proximation was used by Thompson and Imberger (1980) and Imberger
and Hamblin (1982) to formulate the Wedderburn number (2.5) as the
nondimensional parameter determining the response of the surface layer
to an impulsive wind stress.
The initial value problem for the two-layer situation was solved by
Spigel(l980). For the case with rotation, the reader is referred to Csanady
(1982), Heaps (1984), Kielman and Simons (1984), and Horn et al.
(1986); the influence of rotation will not be discussed here. From Spigel
(1980), in the first stages of development of the surface layer following
the startup of the wind stress, the surface layer moves according to
(4.22). This surface layer velocity was shown by Imberger (1985) to be
applicable until the interface tilt reaches the center of the lake and
induces a baroclinic pressure gradient, which retards the motion. The
period of the surface layer seiche was given by ( h , < H)
T,= (5.1)
where L is the length of the lake, Ap is the density difference across the
FIG.17. Upwelling as recorded along the central channel of the Wellington Reservoir in
response to a strong northwesterly wind. Data collected on 24 August 1988. (a) Location
map showing plane of projection and measuring stations. (b) Wind speed as recorded in the
central valley of the Wellington at 1.4m above the water surface. (c) Wind direction
recorded as above. (d) Isopycnals at 0930, before wind had commenced. (e) Isopycnals at
1330 extending from the Gervase River to the main basin. (f) Example of a temperature
microstructure record, collected during the strong upwelling event shown in Figure 17e. (g)
Corresponding temperature gradient. (h) Corresponding dissipation as computed from the
Batchelor spectra fitting technique (Bar graph) and the Wigner-Ville technique (sticks).
23 000
22 000
21 000
20 000
-p 19000
GI
c
Y
LINE OF PROJECTDN
z l8OW AND WIND DIRECTION
:
8 17000
WELLINGTON RESERVOIR
2
s 16 000
15000
14 000
13 000
12 000
14 I 15000 16000 17000 18000 19000 20000 21000 22000 23000
LOCAL CO-ORDINATES Iml
-ur
-
E
0
6
L
fn
4
0
f 2
a
360
0
E
'f 270
:a:
k
0
g 180
5
90 u . .
I . ,. . . . .
2 '.O 237.5 238.0 238.5 239.0
DAY
356 Jorg Imberger and John C. Patterson
0.0
d DENSllY (kg/m3
2.0 --
4.0 --
6.0 --
E
E 8.0 --
12.0
’O’O 1
3000.0 4000.0 5000.0
PROJECTED DISTANCE (m)
0.0
2.0
4.0
--E
6.0
Eul 8.0
0
10.0
\
12.0 /*,*
‘+AO
o \ ,~
14.0
205
A
C60
A
C50
L
a2 /
A
’& A
2000 0 3000 0 4000 0 5000 0
PROJECTED DISTANCE (m)
A-
b
tilting of the base of the surface layer becomes stronger with decreasing
W (Monismith, 1986). The three-layer model has been used successfully
to explain why upwelling is induced for moderately small Wedderburn
numbers. With increasing wind stress relative to the buoyancy influence,
W decreases and the amplitude of the mode 1 response increases, causing
the metalimnion (layer 2) to tilt, which induces upwelling not only from
the metalimnion but also from the hypolimnion. However, it is not the
Wedderburn number that determines the mode 1 response, but the Lake
number LN (2.12); LN is the mode 1 counterpart to W in mode 2. This
may be illustrated by examining in more detail the experiments of
Monismith (1986). The stratification used in these experiments is well
approximated by a three-layer stratification with a constant density in
layers 1 and 3 and a linear transition between the two. For a rectangular
basin of length L, width B and total density difference between the top
and bottom layers of Ap, the Lake number takes the form
where
I
0.0-
O ' O P
5.0 -
-
-b
I
I 10.0-
n.
I-
w
n
15.0-
I , I I I ~o,ol-~---ll,oo~
20.0- I I I I I
50 100 150 200 250 300 50 100 150 200 250 300
DISTANCE (cm) DISTANCE (cm)
0.0 0.0
5.0- 5.0-
-E
x
--5
F
a
w
0
i
w
O
15.0- 15.0-
20.0 I I I I I T 20.0 I I I I I
FIG. 19. Contours of density field measured by Monismith (1986) in a laboratory tank.
The stress was introduced at the bottom with a belt moving from left to right.
(a) Experiment 8: W = 3.9, L , = 6.4, t = 0.32 T, .
(b) -
. , Experiment 8: W = 3.9, L , = 6.4, I = 0.64 T, .
(c) Experiment 8: W = 3.9, L , = 6.4, = 0.96T,.
(d) Experiment 8: W = 3.9, L , = 6.4, = 8.0 T , .
(e) Experiment 22: W = 0.4, L , = 0.15, = 0.35 T, .
( f ) Experiment 22: W = 0.4, L , = 0.15, = 0.53 T, .
(9) Experiment 22: W = 0.4, L , = 0.15, = 0.70 TI.
(h) Experiment 22: W = 0.4, L , = 0.15, = 1.05 TI.
362 Jorg Imberger and John C. Patterson
6 f
0.0
1
FIG.19. (conid.)
surface layer. By contrast, the Lake number for the data shown in Figure
17 was considerably smaller (0.24), explaining the greater mixing
observed at depth (see Section X ) .
W and LN are separate indicators, and for the case where W is small
but LN is large, only the surface layers respond to wind stress. Where W
and LN are small, the lake as a whole responds and we may expect that
vertical mixing will greatly increase throughout the lake.
The response in the case of continuous stratification (Monismith, 1986)
can now be interpreted in terms of these two nondimensional numbers.
For a linear stratification that reaches the surface, the mixed layer depth
is zero, as is the density jump across the base of the surface layer
(continuous stratification may be viewed as the limit where both
approach zero). This would mean that W is zero but LN is finite. The
expected response is a strong tilt of the surface isopycnals with active
upwelling, but at depth, the isopycnals remain horizontal. This case was
recently solved by Monismith (1987), using the same decomposition
technique mentioned above and assuming that the horizontal velocity
profile u(x, z, t) and the force field f ( x , z , t) could be decomposed as
follows:
n=l
where &(z) are solutions of the long wave form of the internal
eigenvalue problem (Gill, 1966) such that
and u,, fn, and 5, are the amplitude functions. Comparison with
experimental results from Monismith (1986) indicate that the fundamen-
tal mode was predicted well, but higher orders, which depend strongly on
the exact nature of the assumed stress distribution within the water
column were not as well predicted, although qualitatively, the agreement
was good.
So far the discussion has concentrated completely on the response of
times up to the quarter internal wave period for either mode 1 or 2.
Beyond this time, the internal displacement degenerates into seiches ( W
large, LN large) or damped motions (W and LN small). The damping of
the oscillations is caused by internal and boundary turbulence and we
364 Jorg Imberger and John C . Patterson
(a) The interface tilts and opens at the upwind end over a period of
time approximately equal to the internal seiche period, (given
by (5.1)). The data suggest that the isopycnals are almost
horizontal at the bottom of the interface but slope considerably
upwards at the top of the interface, establishing a weak horizontal
density gradient in the mixed layer by a combination of upwelling
and horizontal variation in turbulent entrainment. The internal
circulation within the surface layer causes a general upwelling and
divergence of the flow at the upwind end, leading to a diffuse
interface. At the downwind end, there is downwelling, causing a
convergence at the interface, leading to an extremely sharp,
well-defined entraining interface.
(b) The velocity profile in the surface layer is such that the velocity at
the stress surface is approximately 20u, downwind (Kranenburg,
1985; Monismith, 1986) and about 224, downwind at the interface.
The flow at the interface was observed by Monismith (1986) to be
jet-like, concentrated a very short distance above the interface and
contained by a weak density gradient there. This differs from the
assumption made by Spigel and Imberger (1980) that the recircula-
tion flow in the mixed layer could be neglected for W > 1.
(c) The upwelling region is confined to the upwind end of the basin.
The density in the surface layer increases with time; for T > T,,
where T, is the characteristic longitudinal mixing time, the density
profile in the epilimnion varies linearly with distance along the
length of the surface layer.
(d) The net entrainment rule which best fits the experimental data
from Keulegan and Brame (1960), Kranenburg (1985), and Moni-
Physical Limnology 365
dh 1
-= C l u e Ri-', (5.7)
dt
where C1 is a constant between 0.07 (Kranenburg, 1985; Moni-
smith, 1986) and 0.23 (Wu, 1973).
These conclusions permit the construction of a simple model of
upwelling and mixed layer deepening for reservoirs where W 1 and -
L N>> 1 (Imberger and Monismith, 1986, Figure 20).
For this case, as the stress is applied to the free surface, the interface
tilt is set up by the mode 2 response so that the upper isopycnals take on
an angle given by the dynamic balance developed by Wu (1973), and the
boundary layer thickness is proportional to the displacement of the upper
isopycnals
I
STREAMLINESOF UPWELLINGFLOW
u UPWELLING REGION -
SHEARSTRESS
I
I\ \ \...
.....I.\ .................. .).
............................
......\I ....
.............
.I... ..-. ............
A,... ........
4 4
I ISOPYCNALS
I PZ
I
I I
x-0 XIL
FIG.20. Definition sketch for model of mixed-layer deepening due to upwelling. (After
Imberger and Monismith (1986).)
366 Jorg Imberger and John C. Patterson
The model assumes that the return flow of 2u, drags interfacial fluid with
it from the downwind end to the upwelling region. It is assumed that this
fluid, which can be visualized as being planed off the interface, enters the
upwelling region and then is distributed longitudinally by shear flow
dispersion. A simple mass balance would indicate that
(5.10)
ag:,
- B
(0, t ) = - H ( t ) , (5.13)
ax EX
ag:,
- (L, t) = 0, (5.14)
ax
g:,(x, 0) = 0. (5.15)
The required solution can be found in Carslaw and Jaeger (1978) as
2 [ierfc{(2n~+ L - ~ ) ( 2 ( ~ ~ t ) l ~ ) - ’ }
m
gk(x, r ) = 2/3(t/&,)ln
n=O
+ ierfc{(2nl+ x ) ( 2 ( ~ ~ t ) ’ ” ) - ’ } ] , (5.16)
where ierfc(x) = {exp(-x’)/dn} - x erfc(x) and erfc(x) is the com-
plementary error function. Equation (5.16) is plotted in Figure 21 in
terms of the quantity r = &(x, t ) ~ ~ ’ * / 3 - ’ t - ~for
’ ’ different values of
0
- ,
‘1L.10
L\\
0.2 a4 0.6 0.8 1
- 4Id
t = Tf - (“,W.($)
c: . (5.20)
If t > 0.44 TO, the factor appearing in (5.20) will not be n/4.
Figure 22 shows the data from three experiments reported in Keulegan
and Brame (1960), from two experiments reported in Kranenburg (1985),
and from Experiment 8 of Monismith (1986). In reducing these data,
Imberger and Monismith (1986) carried out the following operations:
(a) C,was first chosen to be 0.07 and, for all except Experiment 8, C2
was chosen to be 10.
(b) The resulting “raw” curves were plotted.
(c) So as to best fit the data to the theoretical curves, the dimension-
less time was offset by an amount A x and C1 was altered.
Physical Limnology 369
r
-
'1L
10-
10:
102 10-1 1 I
4%)
FIG.22. Plot of values of (I'/qL)calculated from densities measured with correction for
the initial conditions and adjusted for a best-fit of data at two values of 5. The solid lines
are the theoretical solutions while the symbols represent:
Monismith (1986), W= 3. 9, At'=0.060, C,=O.O7, C,=6;
A Keulegan and Brame (1960), W = 3.0, AT' = 0.015, C , -0.11, C , = 10;
0 Keulegan and Brame (1960), W = 2.4, AT'= 0.025, C , = 0.11, C , = 10;
0 Keulegan and Brame (1960), W = 1.5, AT' = 0.030, C, = 0.05, C , = 10;
+ Kranenburg (1985), W = 3.7, AT' = -0.015, C , = 0.11, C , = 10;
x Kranenburg (1985), W = 2.3, AT'= +0.010, C , = 0.11, C , = 10.
(After Imberger and Monismith (1986).)
The last step was interactive and performed first for the data from the c
position, say el, c
closest to = 0. Once a satisfactory fit was obtained for
el, the data at a second position, c2
> 0.5, was reduced using the final
values of C1 and AT. From trial and error, it became apparent that for
each experiment analyzed, a best set of values of AT, C1, and C2 existed
that appeared to minimize the differences between theory and measure-
370 Jorg Imberger and John C. Patterson
ments for both values of t;. In all cases, reasonable values of all three
parameters were obtained: the time offsets were always less than 0.04 To,
and the entrainment coefficient C, was between 0.05 and 0.133, well
within the range reported in the literature and cited above. The
difference between values of C2 chosen for the belt-driven flow and the
wind-driven flow may reflect differences in the structure of those flows,
especially in the way the shear stress and surface velocity vary with x .
Recently Imberger and Spigel (1987) have carried out measurements in
Lake Rotognaio in New Zealand during conditions of strong heating,
weak winds, and very turbid water (active algal bloom reduced the light
extinction depth to 0.5 m). The general upwelling circulation remained,
but a strong stratification persisted within the whole surface layer,
inhibiting turbulence and eliminating the entrainment surface. The
surface buoyancy flux outpaced the mixing caused by wind-induced
circulation. This type of upwelling is characterized by W being very small
but L N very large, and the circulation appeared as that described by
Monismith (1986) for the continuously stratified case. The major
difference observed was that with the buoyancy flux present a steady state
circulation appeared to form quite rapidly.
The other important variation involves the case of a variable wind
field. If the wind changes direction (Dickey and Simpson, 1983) or speed,
gravitational adjustments lead to large intrusions that may degenerate
into solitons (Kao et al., 1978; Maxworthy, 1980) or undular bores
(Thorpe, 1974; Farmer, 1978; Smyth and Holloway, 1988).
In summary, the upwelling response of the epilimnion (diurnal
thermocline) is determined by the magnitude of W, and that of the
metalimnion by the magnitude of L N . Where W is small, but LN large, it
is possible to model the entrainment process at steady state as simple
recirculation, where the entrained fluid is swept along the interface into
the upwind upwelling region, then is mixed in the surface layer downwind
by shear dispersion. If the stratification is complicated and admits
numerous eigenfunctions of equal importance, then an equivalent num-
ber of nondimensional numbers become important.
The wind field over a lake surface is rarely uniform. Mesoscale wind
variability (McBean and Paterson, 1975; Bean et al., 1975) causes
Physical Limnology 371
variation over large lakes, and wind sheltering by the surrounding terrain
is the major factor affecting smaller lakes (Parker and Imberger, 1986).
Wind variability has two major effects. First, in areas with relatively high
wind speed, evaporation, and thus latent heat transfer, will be greater,
introducing a horizontal gradient in the surface heat flux. This phenome-
non, called differential cooling (Imberger, 1982), is discussed in Section
VII. Second, higher wind speeds on the exposed parts of the lake cause
both a more rapid local deepening of the surface layer and a greater
introduction of momentum. Both must ultimately be averaged across the
lake surface by gravitational adjustments of the isopycnals. The variabi-
lity in the surface layer deepening is called differential deepening
(Imberger and Parker, 1985).
Data from a field investigation near the headland of Salmon Brook in
the Wellington Reservoir illustrate these three phenomena extremely
well (Parker and Imberger, 1986). The data were collected along a
transect extending from a station in a sheltered embayment, station W4,
to the center of the basin exposed to the wind, Station C10 (Figure 23a).
The morning was warm and there was no wind. Shortly thereafter, a
southerly wind began to blow (wind at C10 was 2 m s-' and at W4 was
1m s-I); the surface layer was warm in the sheltered part of the transect
with some noticeable cooling out towards C10 (Figure 23b). This
difference was due to greater evaporative cooling in the open parts of the
lake. The wind intensified sharply at around 1540hours (rose to 4ms-'
at C10, but remained less than 0.6ms-' at W4) leading to noticeable
deepening at C10 (Figure 23c). The wind continued strongly until about
2000hours, when it decreased markedly to less than l m s - ' . The
horizontal density gradient associated with the isotherms is shown in
Figure 23d, implying a further deepening than that at 1805 hours. After
this time, the isotherms progressively relaxed (Figure 23e) until at
2300 hours (Figure 23f), the isotherms were once again virtually
horizontal.
Parker and Imberger (1986) investigated the evolution of the water
profiles by using the surface layer models described in the previous
section. They showed that the local deepening laws could be used to
describe the observations if local wind conditions were used and the
effects of horizontal advection were accounted for, although this ap-
peared to be a relatively minor correction. The results from Parker and
Imberger (1986) suggest that to obtain an average deepening law, careful
averaging must be done over the surface of the lake.
372 Jorg Imberger and John C. Patterson
15000
MAIN BASIN OF
1420
N M K C10
1.0
1.0-
-,..- *-.- .*.-
--E
2.0-
h
X I
N M WClO K C10
0.0
PROJECTED DISTANCE (rn)
4.0
100.0 200.0 300.0 400.0 500.0
PROJECTED DISTANCE
FIG.23. (conrd.)
374 Jorg Imberger and John C. Patterson
4.0
100 0 200.0 300.0 4000 500.0
PROJECTEDDISTANCE (rn)
FIG.23. (conrd.)
The case just discussed was for conditions of moderate W and large
LN. In Figure 24a, we show a much larger transect extending along the
length of the Wellington Reservoir approximately six hours after a
westerly wind had begun. The speed of the wind was 2ms-' at
0600 hours, then rose steadily and peaked at 6 m s-l at 1430hours. The
direction remained quite steady at due west (see Figures 17b and 17c).
The dramatic two-dimensionality of the structure is observed in the
isopycnals in Figure 24a with severe deepening between Stations C10 and
C45 and again between Stations C80 and C90; both sections were more
exposed to the strong westerly. The deepening of the isotherms between
C10 and C45 was due to the setdown associated with upwelling at C10,
whereas the slope of the isopycnals between C40 and C60 was due to the
more sheltered nature of this part of the valley. Confirmation of the
general tilt west-east due to upwelling is given in Figure 24b, which
shows data from a west-east transect at Station C40. The water in both
exposed areas was warmer than its surroundings, which is distinct from
what is observed in Figures 23b to 23f. This indicates that advection was
dominating the deepening process (Imberger, 1985). Further, the average
Wedderburn number for the central basin surface flow at 0927 hours was
0.25, and LN was approximately 1, implying a much less stable lake as a
whole, with mixing expected throughout; the shear associated with
upwelling (4.20) was sufficient even at Station C40, close to the end of
a DENSWY (kg / m3) b DENSITY (kg I m3)
0.0. 0.0.
5.0 5.0.
10.0- 10.0-
-
E E
--
15.0. 15.0.
E
I
I-
n
y1
n l !k
20.0- 20.0-
25.0- 25.0-
30.0.
:loo,
2000.0
CAW Iy
4000.0
CB", "I" I
6000.0 8000.0
Ac;\y"ycL
10000.0 12000.0
30.0-
0.0
A
200.0
A A
300.0
*
4dO.O
.I
500.0
A
6bO.O
A
7[
EAST WE
CUMULATIVE DISTANCE (m)
FIG.24. (a) Transect along the Wellington Valley (see Figure 17a). The wind was coming from the west (as shown in Figures 17b and
17c) and blowing across the lake between the 2000 and 5OOO m mark as well as between the 12000-16000m mark. The isopycnals show
extreme differential deepening, the wind strength at the time was 6ms-'. (b) Cross section at Station C40, from east to west, the
differential deepening, shown in Figure 24a was accompanied by strong upwelling. (c) Temperature. (d) Gradient temperature. (e)
Dissipation as recorded with microstructure instrument at Station C40 at 0947 hours. Dissipation as measured by both the Wigner-Vie
technique (sticks) and the Batchelor spectrum fitting technique (bar chart).
376 Jorg Imberger and John C. Patterson
F
0 1
OL
1 I I I I
0 (D N
l- z P
N
(u)Hld30
Physical Limnology 377
the basin, to induce very active mixing throughout the water column (see
Figures 24c, 24d, and 24e).
De Szoeke (1980) addressed the problem of a variable wind stress over
the ocean. However, he aimed to determine the variability induced by
the Ekman pumping driven by the curl of the surface wind stress and not
by the differential rates of vertical entrainment. Maxworthy and Moni-
smith (1988) were motivated, in part, by the observations in the
Wellington (Parker and Imberger, 1986) and conducted a grid stirring
experiment where the grid extended over only part of a long, stratified
tank. Their results illustrated the sequence of events immediately after
the commencement of the grid oscillations. First, the turbulence intro-
duced by the grid mixed the underlying fluid and set up a downward
propagating entraining front immediately below the grid. Second, this
front continued to propagate downwards until either the turbulence
became weaker due to the greater distance from the grid (Hopfinger and
Toly, 1976), or the mixed fluid collapsed and intruded horizontally into
the neighboring quiescent ambient fluid. Third, as the intrusion propag-
ated out into the main tank, it reached a stage where the end of the tank
began to slow its progress, once again influencing the rate of descent of
the entrainment front immediately below the grid.
Using the various relationships for intrusion speed and entrainment
rates, Maxworthy and Monismith (1988) derived a relationship for the
speed of propagation of the entrainment front under the grid and
satisfactorily verified these relationships with the experiments. The grid
mechanism, while convenient, is a poor analogue for wind-induced
turbulence. Thus, even though the flow scenario will be the same, it is
more useful, as Maxworthy and Monismith (1988) did, to write down the
relationships for the case where the mixing is carried out by a wind stress.
Consider a strongly stratified surface layer over part of which a strong
wind stress is suddenly initiated with a shear velocity of u , . Under the
wind stress, the initial deepening will be given by (4.25b)
1 dh - C,
u , dt Ri* ’
which is the relationship relevant for a linearly stratified fluid and where
C , = 0.1 (Kranenburg, 1985). As shown by de Szoeke and Rhines (1976),
surface-induced turbulence and deepening dominates initially, until the
shear at the base of the mixed layer (5.2) builds up. For longer times,
378 Jorg Zmberger and John C. Patterson
EXPOSED
- SHELTERED
~~
ENTRAINMENT
FRONT
STRATIFIED 4-
where g’ is the reduced gravity across the turbulent front and Lf is the
length of the density variation at the outflow end; this length increases
with time as the intrusion increases to bring outer fluid into the turbulent
zone. This balance may be expected to be valid provided Lf > h2ule,
(Fischer et al., 1979).
The effective longitudinal dispersion coefficient is given by Fischer et
al. (1979) for such a velocity profile:
h2u2
K, = 5.3 x
€2
L‘ = (Kxt)1’2 (6.8)
for times which are short enough so that L > Lf.Combining (6.5) and
(6.8) yields the relationships for both the dispersion coefficient and the
380 Jorg Zmberger and John C . Patterson
peak velocity:
(6.9)
12 3 114
u=1.2(g@)
u*t , (6.10)
where it was assumed that the depth h adjusts slowly compared to the
time it takes for the flow to adjust to the changing Lf.
Unfortunately, the data presented in Linden and Simpson (1986) are
not sufficiently detailed to allow comparison. Maxworthy and Monismith
(1988) balanced the deepening given by (4.23) with the vertical advection
induced by the outflowing intrusional such that
(6.11)
1moo-
--
E
5
I -
14600-
> SALMON
14200-
N
14b00 15dOO 1 seoo 16hOO
X COORDINATE (m)
?‘
2.0-
,
4.0-
-E-E 6.0-
0
8.0-
10.0- 10.0-
12.0- - 0 . 21
I 200.0 4000 600.0 800.0 1000.0 12 0 0.0 100.0 200.0 300.0 400.0 51 0
FIG.26. An example of differential heating and cooling, illustrated with data collected
in Salmon Brook on 25 February 1985. (a) Bottom bathymetry of the site showing site and
transect positions. (b) Longitudinal transect taken early morning, starting at 0849 hours. (c)
Transect across Salmon Brook at Station SB20, early morning, commencing at 0909 hours.
(d) Longitudinal transect illustrating midday heating, taken at 1333hours. (e) Cross-
sectional transect at Station SB20, showing thermal heating near boundaries, data taken at
1340 hours. (f) Transect taken at SB25 at 1639 hours, illustrating the influence of a very
weak westerly wind. (g) Transect taken at SB20 at 1536 hours, the wind at this station was
somewhat stronger. (h) Transect taken at SBOO, commencing at 1639 hours. This section
was almost completely exposed to the wind, which now had reached a value of 4 m s-’. (i)
The net heat flux variation over the five days of the experiment as computed by the rate of
change of heat of the total water column at Station SB30.
Physical Limnology 383
d TEMPERATURECC)
“-I 10.0
12.0
5830
2dO.O
--
4ob.O Sdo.0
A
----..._
- .-
8dO.O
L
4.01 I 2.01
-..._ ,.-*.----.
10.0
FIG.26. (cod.)
from the sheltered SB25 station to the more exposed SBOO (Figure 26h)
station, and to compare this with the discussion in Section VI. At SB25,
the Wedderburn number was around 5-10, as indicated by the very weak
surface upwelling and the almost horizontal isotherms immediately
below. At SB20, the whole diurnal thermocline was tilted, surfacing at
the west end of the side arm, indicating that the Wedderburn number was
around 1. Lastly, at SBOO, the Wedderburn number was calculated as 0.2
and the isotherms indicate a very similar pattern to that shown in Figure
19h. where the Wedderburn number was 0.4.
384 Jorg Imberger and John C. Patterson
TEMPERATURE (‘C)
g
SB2OW SBBOE
6.0
200.0 3W.O 400.0 500.0 600.0
PROJECTED DISTANCE (m)
TEMPERATURE(‘C)
h TEMPERATURE (‘C) i
0 0-
I
I
c
I
I
-
TOTALNETHEATFLUXAT
iI
10-
2 0-
-1000 MUOBSERVED
HEAT INPUT AT WSi
-
, .
m DAY 65
I
,
7
DAY 68
DAY 67
DAVU
O A V ~
-E v-
cc
’=5.
z30-
a -
- - - - - M- - -
U OQSERMD
Lu * 8URFACECOOLING
4 0-
+I00
5 0- 0 0600 1200 I800 2400 0600
SBlOE
A
l.,/= 0- *O-
./--_*..-,..
,,
.-.o-
A
SBlOW
A
TIME (H)
60
I I I
FIG.26. (conrd.)
The consequences for the heat budget of this cyclic flow pattern are
illustrated in Figure 26i. The data show that the water column at Station
SB30 (the head of the side arm) responded as if twice the surface heat
load had been applied in both the heating and cooling phases. This is
further strong evidence that the inertia of the water leads to a large phase
lag between the thermal forcing and the flow response.
Flow patterns associated with the cooling phase are driven by the
greater temperature drop around the boundary; it is yet to be established
whether a variable heat flux also contributed to the excess temperature.
Physical Limnology 385
On the heating phase, both the reduced depth and the reduced latent
heat loss (lower wind speed in the sheltered areas) around the perimeter
lead to the greater temperatures around the boundaries. Steady convec-
tion in a long, shallow cavity, an approximation of the side arm, has
received a great deal of attention for two-dimensional configurations
(Hart, 1972; Cormack et al., 1974; Imberger, 1974; Bejan and Tien,
1978; Bejan and Rossie, 1981; Bejan et al., 1981; Inaba et al., 1981;
Simpkins and Dudderar, 1981; Hart, 1983a, 1983b; Drummond and
Korpela, 1987) and in three-dimensions by Imberger (1976) and Scott
and Imberger (1988). These solutions rely on a viscous buoyancy balance
at first order and all have stable stratification that changes from one with
vertical isotherms at low Rayleigh numbers to one with horizontal
isotherms and boundary jets along the top and bottom boundaries at very
large Rayleigh numbers, such as those that operate in a side arm of a
reservoir. These solutions are of limited value, however, as they require
an imposed steady horizontal density gradient, with zero heat flux at the
upper and lower boundaries. Surface heat fluxes can be incorporated into
the solution technique (Cormack et al., 1975; Scott and Imberger, 1988)
but unless vertical velocities are allowed at first order, these heat fluxes
can only be incorporated at second order, making the solution techniques
less applicable.
Most importantly, in these diurnally cycled flows the inertia of the
water leads to a velocity field almost 90" out of phase with the thermal
forcing. In the data shown in Figure 26, the velocities did not reverse
until about 1500 hours, seven hours after the net heat flux had changed
from a loss to a gain (Monismith and Imberger, 1988). The explanation
of this observation lies in the following scaling arguments. A thermal
gradient is instantaneously imposed on a long cavity of fluid. This
gradient would induce a flow which would be governed by the balance
_
dP -
dz
- -gP,
T=-U (7.3)
Kgh'
386 Jorg Imberger and John C. Patterson
and h is the depth of the flow. Inserting typical values derived from
Figure 26d leads to a time scale of two hours, which is reasonable given
the gradual build-up of the gradient.
The unsteady cavity problem has also been investigated by Patterson
and Imberger (1980) and Patterson (1984). The latter paper used internal
radiative heating to drive the flow rather than heating and cooling at the
vertical end boundaries. The critical parameter to emerge from the long
box problem is Gr A4, where
gcu ABh3
Gr = (7.4)
Y2 ’
A = -h (7.5)
L’
b SALINITY (psfi)
DISTANCE (m)
FIG.27. (a) Isotherms (“C) along a transect from Station CA45 (distance = 4377 m) in
the main Canning River tributary to Station CA24D (distance=9147m). Data was
collected between 1704 hours and 1813 hours, 15 December 1983. For station locations, see
Figure 7b. (From Imberger (1985).) (b) Isohalines (pss) for the same data set as in (a).
(From Imberger (1989.) (c) Map of Canning Reservoir showing measurement stations.
Physical Limnology 389
POISON GULLY
CANNING RESERVOIR
CA180
CA230
equation became
ae de ae
-+ u -+ W - = K V20 + Q(x, Z, t ) . (7.7)
at ax az
A scaling analysis revealed that the flow development, for the initial
value problem, depended critically on the magnitude of the parameter
P = Gr A4, (7.8)
390 Jorg Zmberger and John C . Patterson
where
gaQoh4L3
Gr = (7.9)
Y3
is the Grashof number for the problem and A is defined by (7.5). Eight
regimes of flow were defined. There are two extremes: P > Pr4 (where
Pr = Y / K , the Prandtl number) is the regime where inertia and convec-
tion dominate the final stages of flow. Second, P<Pr-'A4 gives a
thermal field described by a simple balance between conduction and
radiative heating.
Introducing typical values for the problem illustrated in Figures 27a
-
and 27b leads to the first regime (assuming Pr 1) as P is much greater
than one. For this regime Patterson (1984) observed a single gyre
centered at the transition between heating and cooling. The eddy was flat
and supported horizontal isotherms.
Trevisan and Bejan (1986) modeled the turbidity cloud as the perfect
reflector. This is the extreme in differential absorption and, once again, a
simple gyre was generated at the boundary between the absorbing and
reflecting areas although the gyre computed by these authors was not
dissimilar to that found by Patterson (1984). Neither model describes the
real situation; the absorption is differential in horizontal and vertical
directions and the patches are usually three-dimensional.
The original problem that motivated research into convective motions
was in cold lakes and concentrated on the thermal bar problems
(Rodgers, 1971; Spain et al., 1976; Mortimer, 1987). This phenomenon
occurs when a lake, initially at a temperature less than 4"C, begins to
warm. Similar to the observations described above, this warming occurs
more rapidly around the perimeter due to the shallow depth and the
warmer river inflows. In this sense, it is identical to what is believed to be
a good model for the data shown in Figures 26d and 26e. The extra
complication is that the equation of state is nonlinear at these low
temperatures, with a maximum density at 4°C.
Once the boundary fluid has a temperature greater than 4"C, it is
lighter and propagates along the surface. As it meets the colder (<4"C)
water offshore, mixing produces some 4°C water that sinks (Marmoush et
al. , 1984) and produces a complicated frontal mixing 'pattern, distinct
from the normal buoyant surface overflow which has a well-defined
frontal roller (Luketina and Imberger, 1987). Elliott and Elliott (1970)
and Ivey and Hamblin (1988) have carried out laboratory experiments in
Physical Limnology 391
long cavities and Ivey and Hamblin (1988) in particular present graphic
data for the formation of a strong region of convergence, where the 4°C
water plunged in a plumelike sheet to the bottom of the tank. By analogy
with the long box problem, Ivey and Hamblin (1988) suggest the thermal
bar (the region of strong convergence) has a width of O(h Ra-"4), where
gtu( A q 2 h 3
Ra = 9 (7.10)
YK
and
p = po[ 1 - a(8 - 80)21 (7.11)
describes the equation of state in the range 0°C I8 5 8"C, Bo = 4°C.
Bennett (1971) included rotation to show that a long shore shear was
generated by the thermal bar density gradients. Further, a transverse
gyre motion supporting the convergence reinforced the thermal bar. This
analysis showed that the time scales for the establishment and propaga-
tion into the center of the lake is of the order of months, distinct from the
behavior in temperate lakes where, as shown above, it is a matter of
hours.
In summary, we have illustrated a new class of convective flows driven
by what has been called differential heating and cooling. The main
features of this flow are that it is three-dimensional and unsteady with the
inertia of the water introducing a large phase lag between the thermal
forcing and the resulting motions. The impact of these motions should
not be underestimated. In the smaller lakes discussed above, these
convective transports will communicate boundary water to the center of
the lake within hours and in large lakes, where rotation is important,
Ookubo et al. (1984) have shown that the boundary thermal inputs lead
to basin scale gyres, which again ensure a greatly enhanced horizontal
transport.
vm. outflow
The outflow from a stratified fluid is influenced by the buoyancy force,
which inhibits differential vertical motion, causing tilting of the isopyc-
nals. When an outlet is suddenly opened, the pressure wave instan-
taneously sets up a radial flow pattern towards the outlet. However, the
radial convergence near the sink quickly distorts the isopycnals, leading
392 Jorg Zmberger and John C. Patterson
such that
F 2 = 8 n 2 ~ 2 (-
1 E),
where E is the ratio h l H . He postulated that the critical Froude number
(the Froude number at which the upper surface is drawn down into the
cusp) was given by
dF2
-- - 0.
dE
This leads to the solution E = $ and
l$ = 3.42. (8.5)
394 Jorg Imberger and John C . Patterson
4
\
CUSPED SOLUTIONS
\.
1.
1 .
\.
A.
LL"2
CUSPED SOLUTIONS
WITH WAVES -
UNIQUE WAVELESS SPECULATIONONLY
SOLUTIONS
#tEXPERIMENT
HOCKING
(1 988)
0
-l:o -0.5 0.0 0.5 1 .o
(H - h)l H
FIG. 28. Solid line: theoretical results from Vanden Broeck and Keller (1987).
Dot-dash line: theoretical results with a sink (Hocking, 1988). The Froude number F, in
this diagram is based on the depth of water and not the sink depth.
1
H = h + - - Q2 (8.9)
(4nh2)' 2g'
F, = 5.09, (8.10)
396 Jorg Imberger and John C . Patterson
where
(8.11)
This result was first derived by Craya (1949). More recently Wood (1978)
generalized (8.10) to flows in a sector of a sphere and Bryant and Wood
(1976) treated the flow of multiple layers in a sector. Forbes and Hocking
(1988) derived mathematical solutions for the stagnation flow solutions
with no cusps up to a Froude number of 6.4. Experimental verification
was carried out by Lawrence and Imberger (1979) who found a value of
F, = 4.9, which was confirmed by Wood (1978). Harleman et al. (1959)
had found a value of 3.4, whereas the experiments of Jirka and Katavola
(1979) yielded a value of 3.15.
Blake and Ivey (1986a,b) also looked at the case where viscosity is the
controlling force and not inertia to derive the result
h c = 2 . 1 (VgQ
l ) 'I4
9
(8.12)
In both the line and point sinks a diffuse interface will lead to lower
critical Froude numbers (Jirka and Katavola, 1979). This may be
explained by comparing the flow with the upwelling flow in a high
Wedderburn number regime; the three-layer solution of Monismith
(1986) showed a distinct divergence of the interface at the upwind end.
Such a mode 2 response is identical to the interface spreading near the
sink.
For discharges greater than the critical value given by (8.10) or (8.12),
both layers flow toward the sink, such that the interface makes a
well-defined angle at the sink and has a critical point at which the flow in
both layers is controlled. Wood (1978) derived an expression A = Q E / Q H
as a function of the Froude number
where QE is the flow from the epilimnion and QH is the flow from the
Physical Limnology 397
\
\
““I
fly
50
40 -
KINEMATIC APPROACH
KINEMATIC APPROACH
I
30 -
20 -
0
1 01 0 2 °
05 1 2L
10 20 50 100 200 500 1000 2000 5000 10000
F3
FIG.29. The ratio 1 of epilimnion to hypolimnion discharge as a function of the Froude
number for flow towards a point sink. (After Lawrence and Imberger (1979).)
(8.14)
@=
1
2>
2 “ 2
(4- + n = l -
nn
sin 2nnz H(nnx + t). (8.18)
-6 --_
An-
- O(Gr;1'6), (8.19)
x 2 x
where Gr, = N2x4/v2. Thus, a viscosity dominated layer has an x'13
upstream growth. This layer was first derived theoretically and verified
experimentally by Koh (1966a). The velocity profile is bell shaped with a
weak return flow immediately above and below the main layer.
Second, as a shear adds to its predecessor, the central velocity growth
is of order 2nxlNt. Advection will dominate once the induced velocity
O(v/An) is equal to the nth wave celerity O(AnN/2n). Successive waves
of order greater than n will no longer be able to propagate upstream.
This condition leads to a layer thickness
6
- = O(F;I2), (8.20)
X
where F, = q / 2 N x 2 and q is the total flow from both sides. Hence, in the
inertial case, the layer is of constant thickness converging only in the
potential flow region near the sink.
Combining (8.19) and (8.20) shows that the withdrawal layer changes
from one dominated by viscous forces to an inertially dominated layer at
a distance x , given by:
(8.21)
VqJ' = gz
- 7 +7 ,
d p dH
(8.22)
PodW d v
where q ~ =
' (p/po)1'2qJand A is the total upstream head. The first term
Physical Limnology 40 1
lated, since the layer thickness was independent of the discharge, that the
thickness should be the same as in the two-dimensional case. This is
borne out by the data from Koh (1966a), who found
6
- = 5.8 Gr-116 pr-1/6 (8.25)
L
for the case where inertia was negligible. Spigel and Farrant (1984)
investigated the inertia-buoyancy limit and verified that the layer
thickness was given by the scale presented by Imberger (1980):
113
6=C@ 9 (8.26)
where C1= 1.58 (Bohan and Grace, 1973), 1.32 (Lawrence and Imber-
ger, 1979), 1.6 (Spigel and Farrant, 1984), and 1.32 (Ivey and Blake,
1985). Ivey and Blake (1985) also showed there was a layer thickness
6 = 4 . 2 (vQ
3)
(8.27)
that described selective withdrawal when inertia was small but species
convection was important.
The transition between these various layers is more complicated than
in the two-dimensional case. Ivey and Blake (1985) found a transition
parameter
(8.28)
but this parameter is now independent of distance from the sink so that
the inertia-free convective regime thickness (8.27) is also no longer
dependent on distance from the sink. A double criterion thus arises
depending on whether the convective layer can exist at all. These are
A: S > 3, S > (Gr Pr-2)1130 Inertial layer (8.26);
B: S > 3, S < (Gr P Y -~ )” ~ ’ Diffusive layer (8.25);
C: S < 3, S > (Gr Pr-2)1118 Convective layer (8.27);
D: S C 3, S < (Gr Pr-2)1118 Diffusive layer (8.25);
The difference here is that in Cases C and D, the inertial layer is thinner
than the convective layer and thus cannot exist. The first criterion on S is
a measure of the relative size of the inertial and convective layers, while
the second criterion on S differentiates the inertia and diffusive layers in
Cases A and B and the convective and diffusive layers in Cases C and D.
404 Jorg Imberger and John C . Patterson
(8.29)
(where Q is the discharge, b the radius of the cylinder, and f the Coriolis
parameter) approached 0.261. Shih and Pao (1971) carried out experi-
ments and found that the withdrawal layer formed with a radius
(8.30)
R --.Q (8.32)
-YL
For Rf > 1, (8.30) applies; for Rf < 1, (8.31) is relevant.
The above formulae predict a vertical withdrawal layer whereas in the
stratified case, the withdrawal layer is horizontal. Investigations are
underway (McDonald, personal communication) in an effort to reveal the
transition of the flow as the parameter f /N is varied from 0 to infinity.
Physical Limnology 405
Ix. Inflow
Reservoirs and lakes are supplied from rivers; as the river water meets
the relatively stagnant reservoir water, it usually encounters water of a
slightly different temperature, salinity or turbidity. These variables lead
to an inflow that is either lighter or heavier than the surface water of the
lake. When the inflowing water is very much heavier (Hebbert et al.,
1979; Ford et al., 1980) the river water enters the reservoir, and plunges
to the bottom of the lake. Alternatively, examples of where the inflow
water was not sufficiently heavy to take it all the way to the bottom are
given by Howard (1953), Wunderlich and Elder (1969), Ford (1978),
Ford and Johnson (1981), Fischer and Smith (1983), and Imberger
(1985). In all these cases, the inflowing water plunged beneath the lake
water at the entrance then flowed down the drowned river valley,
entraining lake water as it moved downstream. At some depth, the total
entrainment made the underflow neutrally buoyant relative to the
immediately adjacent water, and the underflow became an intrusion
406 Jorg Zmberger and John C . Patterson
(Figure 30). Last, there is the case where the inflowing water is lighter
than the resident surface water and the river overflows the lake as a
buoyant surface flow (Figure 30). The dynamics of such an overflow are
similar to the situation where a source of fresh water enters an estuary or
coastal environment (Chen, 1980; Ookubo and Muramota, 1981; Chao,
1988; Luketina and Imberger, 1987, 1988).
The surface buoyant jet occurs only rarely in a lake in tropical and
temperate regions but is normal in high latitude cold lakes, where such
inflows induce thermal bars (see Section VII). The dynamics of these
flows is a subject in itself and has been extensively treated by Chen
(1980), Sargent and Jirka (1982), Fischer and Smith (1983), O’Donnell
and Garvine (1983), Chu and Jirka (1985), Arita et al. (1986), and
Luketina and Imberger (1987, 1988), with an excellent review of the
frontal region being given by Simpson (1982). The reader is referred to
these articles for further information on the overflow case.
Consider the case where the density of the inflowing water is somewhat
heavier than the resident surface water in the lake and the river water
will plunge down the river bed until it reaches neutral depth. The
Physical Limnology 407
propagation down the inclining plane of the leading head of the gravity
current has been investigated by Stacey and Bowen (1988) and Wilkinson
and Wood (1972). However, in most cases, the inflowing discharge is
sufficiently steady to allow the underflowing river to be analyzed as a
steady or quasi-steady gradually varying flow.
In this case, a balance is struck in the underflow such that bottom
friction, interfacial entrainment and inertia retard the flow, and the
downslope component of gravity accelerates it. This momentum balance
is influenced by the slope and flatness of the river bed and the rotation of
the earth. A discussion of the latter influence is omitted here.
As mentioned above, most authors have assumed the underflow
discharge to be steady, and if it is further assumed that the underflow is
confined laterally by the drowned river valley, a gradually varying flow
analysis yields satisfactory results (Ellison and Turner, 1959; Savage and
Brimberg, 1975; Hebbert et al., 1979; Akiyama and Stefan, 1984). The
main conclusion reached in these studies (see Fischer et al. (1979)) is that
the flow quickly adjusts, so that as it propagates downstream along a
particular bed slope, the Froude number of the underflow becomes a
constant equal to the normal flow Froude number F,, where the Froude
number is given by
and Edinger 1984). However, these all use closure schemes so that the
estimates of entrainment would be overestimated (Franke et al., 1987).
For a divergent river valley with a relatively flat bottom, the flow may
separate from the side walls. Johnson et al. (1987a,b) observed six
different flow regimes depending on the entrance aspect ratio, the Froude
number, and the angle of divergence. These are schematically re-
produced in Figure 31. In general, the higher the Froude number, the
closer the flow is to a simple, entering jet. Once separated from the side
wall, the inflow behaves as a free jet and Johnson ef al. (1988) found that
the entrainment was much higher for such separated flows and could be
estimated from simple jet analysis (Fischer et al., 1979; Rodi, 1982).
The case of inflow into a lake with a flat bed morphology is similar to
the large divergence case and is characterized by a central momentum jet
followed by a radially spreading buoyant plume, originating from a
virtual origin. The virtual origin occurs at a distance from the inflow of
order M3/4B-112(Fischer et al., 1979; Luketina and Imberger, 1987),
where M is the entrance momentum flux and B is the corresponding
buoyancy flux. Luketina and Imberger (1987, 1988) presented an analysis
and a comparison with field data of a buoyant overflow, whereas
Hauenstein and Dracos (1984) presented a similar analysis and a
comparison with laboratory data for the case when the inflow is denser
III-IMMhA
SMALL DIFFUSER ANGLE LARGE
S T R O N P BUOYANCY WEAK
SURFACE
PLUNGELINE (REGION)
A1 A2 A3 B C E
STREAMLINES
FIG.31. Principal regimes of negatively buoyant flow into horizontal diverging channel.
(After Johnson er al. (1987a).)
410 Jorg Zmberger and John C. Patterson
than the lake water. Recently, Tsihrintzis and Alavian (1987) developed
a further integral model similar to that of Hauenstein and Dracos (1984)
for three-dimensional flow down an inclined plane.
Once the inflow reaches a depth of neutral buoyancy, the inflowing
water leaves the river valley and intrudes horizontally into a stratified
lake. Imberger (1985) presents field data from the Wellington Reservoir
that illustrate this lift-off process. The dynamics of the intrusion appear to
have only been thoroughly investigated for two-dimensional flow. In this
case, an intrusion entering a linearly stratified ambient water body will
have a length and width as summarized by Imberger et al. (1976) with the
propagation length being given by
i
0.44 LRli2t’, t’ 5 R, (9.4)
I = 0.57 LR2/3t’5i6, R -=t’ <PI?, (9.5)
CLR 3/4f’3i4, P? < t‘ < R-’, (9.6)
where the coefficients in (9.4) to (9.6) were determined from experimen-
tal data by Manins (1976) and Maxworthy (1980), but where the constant
C in (9.6) is still unknown. In these formulations, the parameters are
defined as
R = F Gr’I3, (9.7)
N2L4
Gr=- (9.9)
Y2 ’
(9.10)
and I is the length of the intrusion at a time t after initiation of the flow.
As already discussed for intrusions formed by differential deepening,
(9.4) to (9.6) are not valid when the intrusion approaches the rear wall
(Fischer et al., 1979).
Darden et al. (1975) showed that the accompanying velocity structure
due to such an intrusion consists of multiple reversing jets above and
below the entering intrusion, and Maxworthy (1980) demonstrated that
simple intrusions can degenerate into solitons if the volume flux to the
intrusion is terminated.
The second main case in a lake is where the downflow encounters a
Physical Limnology 411
where d , is the depth of the upper layer, d2 is the depth of the lower
layer, p1 is the density of the upper layer, p2 is the density of the lower
layer, hl is the depth of the upper fluid above the intrusion, h2 is the
depth of the lower layer below the intrusion, and Ap = p2 - p,. Faust
(1981) generalized the same analysis and showed that a similar expression
could be derived for an intrusion into a three-layer fluid. However, in this
case, he found that internal waves generated at the interfaces could lead
to a slow-down of the intrusion.
The above intrusion dynamics have only considered laminar flow.
However, in lakes, the downflow is strongly turbulent (Imberger, 1985)
and it is important to assess the collapse of this turbulence as the
intrusion enters the lake. Lin and Pao (1979) and later Hopfinger (1987)
have reviewed this subject and concluded that the collapse takes place in
times of O(N-') and that the vertical motions decay most quickly,
leaving, in the absence of additional shear, two-dimensional turbulence
with a vorticity vector orientated predominantly vertically. The boundary
mixing experiments of Browand ef al. (1987) suggest that the decay is
extremely rapid and that we would not expect to find turbulence within
the intrusion at a horizontal distance from the lift-off point much greater
than the thickness of the intrusion. This has been confirmed by Roberts
and Matthews (1987) who injected a horizontally neutral buoyant
momentum source from a round nozzle into a linearly stratified fluid and
found that the mean shape of the jets spread horizontally at a distance of
M"4N-1",where M is the momentum flux of the jet and N is the
buoyancy frequency of the ambient water. Although these authors did
not measure the turbulent flux directly, they inferred that the largest
turbulent eddies became anisotropic at the same distance.
Now in the case of an underflow, the Froude number based on the
density difference between the underflow and the neighboring ambient
412 Jorg Imberger and John C . Patterson
for the two-dimensional flow case. Assuming that the collapse starts at a
distance order (MN-2)”3, which is the equivalent result for the two-
dimensional flow, the turbulence would commence the decay at a
distance O(QN-l”) or one intrusional layer thickness into the lake.
Given these two pieces of evidence, one would expect not to find
turbulence in an intrusion. However, recently Imberger (1987) has
shown, by carrying out temperature microstructure measurements, that
the shear from the intrusions together with background shear from
internal wave activity can sustain a low level of turbulence. In a further
analysis of the same data, Imberger (1988) has shown that the turbulence
was at the point of decay, with
Mixing below the surface layer is generally patchy and sporadic, with
individual events occupying no more than tens of cubic meters and
remaining at full intensity for no more than a few minutes (Imberger,
1985). The two central questions remain: What is the efficiency of energy
conversion at turbulent sites and what is the patch statistical distribution
(Imberger and Hamblin, 1982)?
No direct measurements of the local value of the vertical diffusion
coefficient K , exist, but many investigators have taken advantage of the
closed nature of the lake basin and obtained estimates of the vertical
diffusion coefficient averaged over the whole basin (Kullenburg et al.,
1974; Jassby and Powell, 1975; Imboden and Emerson, 1977; Imboden et
al., 1977; Robarts and Ward, 1978; Weiss et al., 1979; Lewis and Perkin,
1982; Imboden et al., 1983; Imboden and Joller, 1984; Sanderson et al.,
1986; Colman and Armstrong, 1987; Lake Biwa Research Institute, 1987;
Wuest, 1987; Wuest et a/., 1988). The data confirmed (Imberger and
Hamblin, 1982) that the lake-wide averaged vertical exchange coefficient
fluctuates between 2 x m2 s-l (near molecular) during periods of
very strong stratification and weak winds (Wuest, 1987) to values as high
as m2s-l during periods of very weak stratification in tropical lakes
(Lewis and Perkin, 1982). It has also been shown (Ward, 1977;
Straskraba, 1980; Wuest, 1987) that the vertical diffusivity depends on
the buoyancy frequency, the area of the lake and the depth of the lake.
Wuest (1987) attempted a correlation of the rate of mixing with the
Wedderburn number (2.5) and the buoyancy frequency, writing the
vertical diffusion coefficient as
(10.1)
K, (rn21s)
his profile shapes, this would have been close to the depth of the seasonal
thermocline, which would make W close to LN. His W is really an
indicator of the stability of the seasonal thermocline and not the surface
layer.
We saw in Section I1 that in the majority of cases a lake is adequately
approximated by a three-layer stratification, so there are two numbers, W
and LN , which determine its response to a wind force. Small W and large
LN lead to a strong surface layer response, and small LN causes the main
thermocline to tilt with the associated stronger mixing in the hypolim-
nion. Three examples provide an illustration. In Figure 33, data from the
Canning Reservoir show a case where W =0.1 and LN=288. The
dissipation levels are close to zero everywhere, except in the top two
meters, where a very weakly stratified surface layer resided. During the
time of this profile, nearly 200 microstructure casts were taken over a
period of 10 days and only one profile showed the slightest sign of
turbulence in the hypolimnion. By contrast, Figure 17f depicts a
microstructure profile at W = 2.1 and LN = 0.24, whereas in Figure 24c
the conditions were such that W = 4.5 and LN = 0.44; in both cases the
whole profile had become turbulent with dissipation values in the
hypolimnion reaching m2s - ~ . These data are presently being
analyzed more fully, but this preliminary analysis, together with the
evidence from Wuest (1987), strongly suggests that the response of the
Physical Limnology 415
FIG.33. Microstructure taken in the Canning Reservoir during a period of large Lake
number L , = 288. (a) Temperature. (b) Temperature gradient. (c) Dissipation as calculated
from the Wigner-Ville technique (sticks) and the Batchelor spectra fitting technique (bar
chart). Note that there is no turbulence below the top 2 m.
(10.3)
where Qo is the air flow rate of the aerator, Qi is the river inflow rate at
the point where the intrusion forms, U is the inflow velocity of the
intrusion, and zi is the height of the intruding intrusion. All other terms
are defined in Sections I1 and XII. The additional terms are the
additional disturbing moments offered by the bubble flow pog2nQ$'3~g3A0
(Section XII) and the river momentum flux pOUQiacting at a moment of
(zg- zi) (Section XI). Whenever LN is small, the seasonal thermocline
may be expected to respond more actively, thus leading to higher
transport rates.
The origins of the mechanisms leading to internal mixing are numer-
ous, consisting of a range of instabilities, wave-wave, wave-shear,
Kelvin-Helmholtz billows, Holmboe instabilities, critical layer absorp-
tion, double diffusion, bores, and hydraulic jumps. These mechanisms
have recently been thoroughly reviewed and the reader is referred to
Hopfinger (1987) and Thorpe (1987). Mixing in the metalimnion is also
reviewed by Gregg (1987). For a review of double diffusion, the reader is
referred to Turner (1985).
The nature of the turbulence in a lake can be determined from the
turbulent Froude versus Reynolds number diagrams (Gibson, 1980, 1982,
1986, 1987a,b, 1988; Imberger and Boashash, 1986; Imberger, 1988). The
Froude number is defined
(10.4)
Physical Limnology 417
(10.5)
(10.6)
FIG.34. The Froude versus Reynolds number plot containing the data from microstruc-
ture segments with well-defined spectral components protruding above a step gradient form.
All dissipations used in this plot were estimated by fitting a Batchelor spectrum to the
spectra computed for the whole segment.
0 Bubble plume thermal PT
0 Thermal falling in a weak ambient stratification
0 Thermal impinging on a sharp interface
A Penetrative convection
V Surface buoyant plume roller region
V Surface buoyant plume wake
x Well mixed surface region
+ Strong stratification, strong shear
Strong stratification, weak unknown shear
+ Strong stratification, no known shear
* Gravitational underflow
A Penetrative convection thermals impinging on an interface
0 Intrusion flows.
0-
3-
-
r 6-
t
0 .
9- -19
26 26-30
TEMPERATURE (“C)
0
n/dz (“Ch)
3010”0
- 10’
& (m2/s3)
10.’ lo4
112
TEMPERATURE (“C)
0
dT/dz (“Urn)
--
3010’0
hw
lo4
Ws3)
10‘ loJ
’I2
h i (1406)
7r
I
I,
-3
w
D
. -
-
E
I
6k
0
- 9
I
’ZLi4 ‘
-
I
- l L
26 28 -30 0 3d10’0’”’;06 ‘;‘Iz
TEMPERATURE (“C) dTidz ( T i m ) EuIx ( m Z / s 3 )
3x m2s-l for the interior of the ocean. This result was derived from
the formula
Kl
K , =- (10.7)
BL ’
where K is the diffusion coefficient in the benthic boundary layer, 1 is the
turbulent benthic boundary layer thickness, 8 is the bottom slope, and L
is the basin scale.
The other mechanism is breaking internal waves (Cacchione and
Wunsch, 1974; Murota and Hirata, 1979; Murota et al., 1980; Thorpe,
1987; Garrett and Gilbert, 1988; Ivey and Nokes, 1989; Wallace and
Wilkinson, 1988). Internal waves shoal and break as they propagate over
a sloping bottom, especially if the bottom slope is at the same angle as
the rays of the internal wave field (a critical slope). Ivey and Nokes
(1989) found the turbulence in the benthic boundary layer resulting from
the breaking waves to be sustained by shear leading to Richardson
numbers as low as 0.01. For a critical slope, the steady state boundary
layer thickness resulting from the wave breaking was given by
1 = 5I;, (10.8)
where I; is the internal wave amplitude.
The mixing efficiency within this active boundary layer measured as the
flux-Richardson number (buoyancy flux divided by production) reached a
maximum of 20%. The diffusion coefficient K , can now be estimated
from (10.7) and the estimate of the diffusion coefficient in the benthic
boundary layer (Ivey and Nokes, 1989):
K = 0.09 cot2, ( 10.9)
where I; is the wave amplitude and w the wave frequency. Thus, given
(10.7)-(10.9), a recipe for estimations of the vertical diffusion coefficient
due to boundary mixing is now available. However, application still
requires a knowledge of the internal wave spectrum in a lake as a
function of L N, in order to determine what percentage of the bottom is at
critical slope. This is an urgent priority. Further, given the importance of
the first mode seiche, it is not clear that boundary mixing plays the same
important role in lakes as it appears to do in the ocean.
In summary, a great deal of progress has been made in the area of
hypolimnetic mixing. Recent measurements have conclusively shown the
turbulence level in the interior of the lake increased as LN decreases and
422 Jorg Zmberger and John C . Patterson
also that the effective basin average diffusion coefficient correlates to N- "
where a is approximately 1. The role of the benthic boundary layer and
the turbulence due to shear and internal wave breaking has been
quantified in laboratory experiments and these results are now ready to
be tested in field experiments with the data set from Harding Reservoir.
This should determine the functional relationship between the proportion
of transport in the interior compared with that in the benthic boundary
layer as a function of the Lake number LN .
XI. Modeling
(1970), Huber et al. (1972), Sundaram and Rehm (1973), and Markofsky
and Harleman (1973) all developed models that simulated the vertical
transport of heat in the epilimnion by means of a diffusion-like process,
characterized by an eddy difhsivity E,. The definition of E, was the
major difference in these models. For example, Huber et al. (1972)
assumed a constant value of E, , calibrated from a data sequence. Orlob
and Selna (1970) used a constant value of E, in the epilimnion, with a
value exponentially decaying with depth in the hypolimnion, also
calibrated against data. These models were poor representatives of the
actual processes. Sundaram and Rehm (1973) incorporated some of the
physics into the representation by defining an E, that depended on the
stratification and the wind, through an overall Richardson number.
Similar formulations for E, have been suggested by, for example,
Thomas (1975) and Witten and Thomas (1976). Henderson-Sellers (1976)
has further developed these ideas for deriving E, for thermocline models
based on boundary layer theory and a mixing length turbulence closure
scheme.
Although the early models based on diffusivities were moderately
successful and, being relatively insensitive to the value of E,, were
simple to calibrate, Harleman (1982) showed that this was the result of
the dominance of other processes for the vertical transport of heat, and
the insensitivity may not occur in all cases. Further, in all cases,
calibration was required and transfer of the model from one lake or
reservoir to another required recalibration. Even within one lake,
conditions outside the range of calibration could not be modeled with
confidence. More importantly, the description of the turbulent transport
processes by a single bulk eddy transport coefficient meant that the
effects of the individual processes, such as turbulence generated by the
wind, convective mixing, internal waves, boundary mixing, and basin
scale seiching for example, were integrated into a single parameteriza-
tion. The coefficient was particular to both the lake and the combination
of processes acting at the time, and could have no real physical
interpretation. Because of its integrated, non-physical nature, there was
no independent means of verifying the parameter value or dependence.
Finally, the application of these diffusivities to other lake properties, for
example salinity or dissolved oxygen, was not necessarily correct. Fischer
et al. (1979) pointed out that the distribution of different properties may
arise from different combinations of physical processes, and the same
diffusivity may therefore not be appropriate.
424 Jorg Imberger and John C. Patterson
U
Fr =- (11.1)
(g’H)”’ ’
where g’ was based on the density difference between inflow and surface
water, and H was the total depth. For Fr < l/n, they concluded that the
reservoir was sufficiently stratified for the one-dimensional assumption to
hold. Bloss and Harleman (1979) pointed out that the surface fluxes,
particularly the effects of the surface wind, were not represented in this
criterion and suggested that a limitation on wind speed U, be used in
addition. This was
(11.2)
426 Jorg Imberger and John C. Patterson
where d was the mixed layer depth, L the length of the lake, H the total
depth, and Ap and p o the density jump at the thermocline and the mean
hypolimnion density respectively. If this inequality was not satisfied, it
was assumed that the deviation of the thermocline from the horizontal
was sufficient to prevent the one-dimensional assumption from holding.
The numerical parameter was based on earlier empirical data relating
surface set up to wind speed.
This criterion was similar in character to that established by the
analysis of Spigel and Imberger (1980), based on rigorous dynamical
grounds. Here, as discussed in detail in Section IV, the stratification and
wind forcing were related through the Wedderburn number W
(Thompson and Imberger, 1980) in a series of criteria which categorized
the response of the lake to external wind forcing. These criteria, together
with modified inflow and outflow criteria and a parameter that charac-
terized the importance of the earth’s rotation, were discussed in
Patterson et al. (1984). These may be summarized as follows.
The deviation of the mixed layer was characterized by the value of W
(defined in Section IV). For W > 10, the deviation was negligible, and
the mixed layer deepening was characterized by the one-dimensional
effects of surface stirring. For 3 < W < 10, increasing evidence of shear
production was likely, however the one-dimensional model, with a shear
production algorithm was still valid. For W <3, the interface was
sufficiently close to the surface at the upwind end to be affected by
surface stirring, and the lake became mixed through an upwelling front
moving downwind. These criteria were based on the first mode wave
behavior in a rectangular basin containing a fluid with two densities.
Even with this simplification, the criteria proved a useful classification of
lakes (Patterson et al., 1984).
The inflow criterion was effectively the original Water Resources
Engineers form above (11.1). The approximate nature of the calculation
of Fr meant that there was little significance in the factor l/n, and Fr < 1
was taken to indicate one-dimensional behavior. Similarly, an outflow
Froude number was formed as
F, = Q (11.3)
H2(g’H)”2 ’
where Q was the discharge, which again reflected the relative importance
of inertial and gravitational adjustments. For F, < 1, the one-dimensional
structure was not disturbed. A criterion based on comparing the Rossby
Physical Limnology 427
radius of deformation with the lake scale was used to characterize the
effects of the earth’s rotation.
The difficulty with the Wedderburn number and Froude number
calculations is that they are based on simple geometries and simple
two-layer density structures. To calculate the value of W appropriate to a
real lake involved solving the eigenvalue problem for the internal wave
field in the actual stratification and utilizing this to form an equivalent
two layer structure; depending on the stratification ambiguity could arise.
Likewise, the Froude number criteria were based on very simple
geometries, and took no account of, for example, the ratio of inflow to
storage.
The Lake number LN introduced above minimizes these difficulties of
application. The form discussed in Section I1 has been shown to be
equivalent, in the two layer rectangular case, to W, and its interpretation
is clear. For LN>> 1, the stratification is sufficiently strong to minimize
any disturbance of the metalimnion from the surface wind, and the para-
meterizations of Section IV are appropriate. Thus in an equivalent way to
W, the value of LN characterizes the assumption of the overall one-
dimensionality with respect to surface wind rather than just the surface
layer. Further, by redefining the maximum angle of deviation, the higher
mode responses may also be characterized.
The values of LN calculated for the Canning Reservoir, based on actual
values of u, (rather than the fixed value used in Section 11) and the field
profiles of temperature and salinity, are shown in Figure 36a. This shows
the same trend as Figure 8: LN values of order 50 for the summer period
when the stratification is strong and LN values of order 0-2 in winter
when the stratification is weak and the winds high. Clearly the one-
dimensional assumption is appropriate in summer, but there are periods
during winter when LN is close to 0, and the assumption may be invalid.
Examination of those times, however, reveals that the stratification is
extremely weak, and the error made by a vertical mixing model
compared to a model based on upwelling is small; both produce a
homogeneous reservoir over short time scales. On the other hand, if LN
were small in times of moderate to strong stratification, or W is small but
LN is large, upwelling will occur at the upwind end from the hypolimnion
in the former case and the surface layer in the latter. In both cases, the
one-dimensional assumption does not hold. In these cases, a two- or
three-dimensional circulation model may be more appropriate.
The form of LN for river inflow may also be easily determined.
428 Jorg Imberger and John C. Patterson
+
3w +
1W- ++
+ +
+
+ + t +
50- t
+ + +
+ +++ + + +
+ + + +
3 1.50-
L
9
3 0.00-
ol
0.30
0.00 I I-.
~ *++
I
+
+
+
t
I
+
+
+
t +
I
; k S OI N D J I F M I A I M J I J
MONTH
FIG.36. (a) The Lake number LNfor Canning Reservoir for those days on which field
data are available, based on the actual values of wind speed for the period 11 June 1986 to 7
September 1987. (b) The Lake number LN,[for Canning Reservoir, based on the inflows for
the same period as Figure 36a.
Physical Limnology 429
(11.4)
where the length scale for the reservoir is taken as Ah”. Thus, large
values of LN,I correspond to small interface deviations.
The values of LN,, for the Canning Reservoir on those days on which
profile data are available are shown in Figure 36b. These are based on
the conservative assumption that zi = 0, that is, that the river underflows.
Clearly, LN.1 increases to relatively large values in the summer when the
stratification, and therefore, S, are large, and the inflows are small. LN,l
becomes small in periods of very weak stratification, and as in the
previous case, the weakness of the one-dimensional assumption causes
little damage.
Using LN and LN,, then gives an indication of the appropriateness of
the one-dimensional assumption for surface and inflow driven deforma-
tions of the density structure. The extension of LN to outflows and
deformations resulting from the earth’s rotations are not obvious and
have not been pursued here.
Given that the indicators for a lake or reservoir suggest that a
one-dimensional parameterization is appropriate and that a process-based
mixing model is preferred to a diffusivity model, the remaining question
is which of the available models should be chosen. As indicated above,
most of the development has been in the context of the MIT and
DYRESM models. In addition, the U.S. Army Corps of Engineers
model CE-QUAL-R1 is available. The temperature prediction com-
ponent of this model is, in its present form (U.S. Army Corps of
Engineers, 1982), very similar to DYRESM. In the following, the
structure and performance of DYRESM only will be discussed in detail.
As noted above, DYRESM is based on a Lagrangian representation of
the lake, with each horizontal layer of uniform property but variable
thickness and location. Thus, each layer expands or contracts as inflow
and outflow affect its volume, and those above move up or down. In this
way all vertical advection of mass is accounted for by layer movement,
and problems of numerical diffusion are not present. Further, conserva-
tion of mass is simpler to achieve without the necessity of computing
vertical velocities.
The model contains five basic process descriptions: surface fluxes of
heat, mass, and momentum; mixed layer dynamics; mixing below the
430 Jorg Imberger and John C . Patterson
The fluxes of heat, mass, and momentum at the surface are described
in detail in Section 111. Following the comments in that section with
respect to the self-regulation of the thermal budget, the bulk formulae
equations (3.1)-(3.3) are appropriate for modeling that reports on a daily
time scale, and frequently utilizes meteorological data from a single site,
averaged over a full day. The evaporative heat flux is calculated from
(3.3), using a bulk value of Cw of 2.6 X loF3, within the range of values
suggested. This corresponds to the formulation arising from the Lake
Hefner data (Tennessee Valley Authority, 1972) and is similar to the
form used by Orlob and Selna (1970). Sensible heat transfer (3.2) uses
CH= Cw and is related by the Bowen ratio (Henderson-Sellers, 1986).
The value of CD for momentum transfer (3.1) is wind speed dependent,
in the form (Donelan, 1982; Ivey and Patterson, 1984)
1.124 X lop3, u < 4 m s-', (11.5)
c,=( (0.96 + 0.041 U)x U >4 m s-'.
The remaining heat transfers are treated in the usual way. Incoming
short wave radiation is absorbed by the water column, after reflection at
the surface. The reflection is characterized by the albedo, which is
measured for the particular site. The albedo A, is a function of sun angle,
surface roughness, water color, etc., but is usually taken, for averaged
data, as a constant. Typically, A, is of order 0.04, corresponding to 4%
reflection. The absorption of short wave radiation is characterized by
Beer's Law
(11.6)
Physical Limnology 43 1
cPi==, (11.7)
i
and the qi are the attenuation coefficients for each band. Although two
and higher band relationships have been used, for example for the
absorption of short wave radiation in ice (Patterson and Hamblin, 1988),
in general a single band formulation is sufficient for use with daily data.
Incoming long wave radiation is either measured or predicted from
formulae such as the Swinbank equation (Swinbank, 1963), modified to
include the effect of cloud (Henderson-Sellers, 1986). Likewise, the long
wave emission from the surface may be measured or calculated from the
Stefan-Boltzmann Law
QL= m T 4 , (11.8)
where E is the surface emissivity, CT the Stefan-Boltzmann constant, and
T the surface temperature in "IS.The emissivity is fixed at 0.975.
These heat transfers are implemented on the layer structure of the
model, with all fluxes except short wave affecting the surface layer only.
The time step of the heating and surface mixing calculations is also set
here; the time is limited to that in which the surface layer temperature
changes by 3"C, up to a maximum of 12 hours. An additional limitation is
placed on the time step in terms of momentum transfer; the change in the
mean mixed layer velocity from the previous value is also limited.
+--
Apgh
2p0 24p0 dh
+--I
gb2 d ( A p ) g A p b d b dh
12p0 dh dt
CK
=-(w:
2
+ q3u:)+- u:+--+--
"'2[
U f d d U,dd(U,)
6 dh 3 dh 1;'
dh
(11.9)
432 Jorg Imberger and John C. Patterson
where
q: = w: + q3u:, (11.10)
and the following equivalences apply:
77 = CN, (11.11)
(11.12)
(11.13)
The parameter a in (10.2) is given the value 0.5 (Imberger and Hamblin,
1982).
4. Inflow
The inflow of the rivers into the reservoir is modeled in three separate
processes, as described in Section IX: determination of the plunge point,
underflow, and intrusion. Following Section IX, the plunge point
determination is parameterized by (9.1) and (9.3), the underflow by
(9.2), and the intrusions by (9.4), (9.9, and (9.7)-(9.11). The implemen-
tation of (9.4) and (9.5) is described in Imberger and Patterson (1981);
-
briefly the balance is changed from inertia-buoyancy at R 1 (R defined
by (9.7)). The cases of steep slopes, unconfined, or broad river valleys
are not modeled by DYRESM. The impact of a downflow on a sudden
density change is treated by the method of calculation of N2, as for the
withdrawal case below.
The downflow is modeled by tracking the day’s inflow down the slope
and locating its position and flowing depth. Thus, the underfiow is made
up of a number of parcels of inflows placed along the slope, moving down
on each day until their level of neutral buoyancy is reached. Insertion
following (9.4) or (9.5) generates the intrusion thickness from conserva-
tion of volume. The insertion lengths calculated from (9.4) and (9.5)
together with the downflow characteristics are essentially two-
dimensional parameterizations of the inflow; these are later utilized in a
quasi two-dimensional version (see below).
5. outfiow
which is a cosine profile about the level of the sink with a linear decay
away from the sink to the far end of the lake, where L is the lake length,
dIn the layer half thickness, uo the centerline velocity, and z, the height
of the sink.
Invoking conservation of volume enables computation of the vertical
velocities, and the streamlines were shown by Hocking et al. (1988) to
have the form
Physical Limnology 435
50.0
40.0
- 30.0
N
20.0
10.0
I I I I I I I J I I I I I I
J A S O N D J F M A M J J A
DAY
'7
0
1974 1975 1978 wn 1978 1979 1980 1981 1982
T I M PERKID
FIG. 38. The offtake salinities at Wellington Reservoir, measured (dashed line) and
predicted by DYRESM (dotted line), over the extended period 1974 to 1982: (a) mid-level
offtake and (b) scour offtake.
--
E
N
30-
20-
1C-
b SALINITY (ppt)
FIG. 39. (a) The isohalines resulting from salinity measurements at a sequence of
stations in Canning Reservoir on 27 October 1986. (b) The isohalines drawn from the
simulation of Canning Reservoir by two-dimensional DYRESM, after a short (30-day)
simulation.
440 Jorg tmberger and John C . Patterson
(12.1)
(12.2)
(12.3)
For PA < 30, the plume broke through the interface and mixed with the
upper layer. For PA > 30, the stratification was strong enough to prevent
the entrained fluid from penetrating the interface, and detrainment
occurred at or below the interface. In the first case, mixing occurred from
above, with a deepening of the interface; in the second, mixing was
weak, and characterized by a thickening of the interface.
These results were consistent with others; the McDougall (1978)
experimental results were of moderate PN, with steady subsurface
intrusions. The two layer results were similar to those obtained by
Graham (1980). While insufficient details are given by Graham to
calculate P A , it is clear that in the early part of the experiments, the
“diffusive” profiles are characteristic of PA > 30; the density jump at the
interface is sufficiently strong to force detrainment below the interface.
At later times, the stratification has been reduced to allow the effective
PA<30, and the plume penetrates fully, resulting in a sharper interface
as mixing from above deepens the upper layer. The experiments of
Kranenburg (1979) may also be put in the context of Pa. The series of
profiles shown in Kranenburg indicate, after the initial stages, a rapidly
deepening interface, consistent with a calculated value of PA of 0.1, much
less than the suggested value of 30 for penetration of the interface. This is
consistent with the qualitative description given by Kranenburg of an
intrusion plunging from the surface to the interface. The experiments of
Dortch and Holland (1980) were directed at comparisons with hydraulic
methods and insufficient details were given for a comparison to be made.
These conclusions have been incorporated into the reservoir model
described above. Although integral models of the plume itself exist, none
Physical Limnology 445
have been coupled to a far field model, other than that of Kranenburg
(1979). Kranenburg developed a three component model; the plume, the
intrusion, and the far field two-layer stratification, in a cylindrical
geometry. Good comparisons with the experimental results were ob-
tained, but the model was not intended for reservoir applications.
The development of the coupled bubble plume and reservoir model is
fully described in Patterson and Imberger (1989); briefly, the single
plume model of McDougall (1978) was discretized using the layer
structure determined by the reservoir model, with the plume entraining
from each layer as it passes. As the plume rises, the effective buoyancy
anomaly decreases as entrainment lowers the plume density and the
ambient density decreases. At some level, the vertical velocity of the
entrained fluid becomes zero, the fluid is ejected and mixes with the
underlying reservoir water. It is assumed that the resulting horizontal
gravitational adjustment is sufficiently rapid for the one-dimensional
adjustment to remain in force. The bubble plume continues, starting a
new entrainment cycle.
The coupled model was verified against a one-year data series from
Myponga Reservoir in South Australia in which an aerator had been
operated over the summer at a depth of 14m below the surface
(Patterson and Imberger, 1989). The model in general reproduced the
resulting thermal structure, and it was concluded that it adequately
represented the coupling of the bubble plume dynamics with the
processes acting in the reservoir.
This enabled the use of the coupled model to determine various aspects
of plume behavior. Of particular interest was the maximum efficiency
airflow rate found in the Asaeda and Imberger (1988) experiments for a
simple geometry and simple stratifications. The question of the ap-
plicability of this to real geometries and stratifications is fundamental to
the optimum design of destratification systems.
One measure of performance of a destratification system is the
efficiency q, defined above. Asaeda and Imberger (1988) showed that the
efficiency depended on the parameters PN and Pa for the two specific
stratifications described. For the more general case the stability of the
system is characterized by S, and the airflow rate by Q,.Similar to the
cases for wind stirring and inflow, it is possible to form a version of the
Lake number appropriate to the bubble plume. Again, following Section
I1 and taking moments about the center of volume yields Stg/3, where /3 is
the angle of deviation of an isopycnal. If the disturbing force is taken as
446 Jorg Imberger and John C. Patterson
(12.6)
where Bmax is the maximum deviation. For the two-layer case Bmax-
(2, - zT)/L;for the linearly stratified case & , - H / L . In general, if LN,B
is large, the deviation is small compared with Bmax and the density
structure will not be disturbed by the bubble plume; if LN,B is small, the
deviation is large. In the first case, the airflow rate is insufficient to
penetrate the stratification and cause mixing. The second case cor-
responds to an airflow rate well in excess of that required to penetrate the
stratification. If LN,B is large the efficiency will clearly be small since very
little mixing in a relatively strongly stratified environment will occur; if
LN,B is small, although mixing is occurring, the disturbance is more than
is required, and the efficiency is again low. The manifestation of the small
LN,B case is a plume detraining at the surface, which would also have
detrained at the surface with a much lower airflow rate. This qualitative
argument indicates that a maximum value of efficiency should occur at
intermediate values of LN,B and that, following the experiments of
Asaeda and Imberger (1988), the achievable efficiency should be of the
order of 0.12.
The parameters derived by Asaeda and Imberger (1988) are directly
related to L N , B . It is straightforward to show, using the definition of Bmax
above, that, for the linear stratification,
(12.7)
Physical Limnology 447
where A. is the surface area of the experimental tank, and for the two
layer case,
(12.8)
-
For the linear stratification experiments, the value of PN lo3 gives
LN,B -8. For the two-layer experiments that yielded a transition of
character at PA-30, this puts transition at LN,B-5. The definition of
LN,B then unifies the two parameters defined by the experiments of
Asaeda and Imberger, and suggests that for other stratifications, LN,B -
5-8 will maximize efficiency of the bubble plume-induced mixing.
The numerical experiments with reservoir data by Patterson and
Imberger (1989) described above yielded single plume efficiency curves
which indicated a maximum efficiency of order 0.15, consistent with the
Asaeda and Imberger experiments. Patterson and Imberger introduced
the parameter P, where
(12.9)
and M is the total mass of the reservoir, derived from the ratio of energy
stored in the stratification to the energy input by the plume.
The simulations in Patterson and Imberger (1989) showed that a
maximum efficiency occurred at P - 1, as shown in Figure 41a for
Myponga Reservoir. Over the simulation period, a wide range of
stratification is present, and by varying Qo , a wide range of P values may
be obtained. Figure 41 is constructed from the results of several
simulations, with the efficiency plotted as a function of the P calculated
for a particular stratification and airflow without regard to the temporal
order of occurrence.
Myponga Reservoir is a small reservoir of storage 26.8 x lo6m3 and
surface area 2.8 x 106m2.The same calculation has been carried out for
the Harding Reservoir in the northwest of Western Australia. The
Harding Reservoir is substantially larger (storage 63.8 x lo6 m3 and
surface area 14.1 x lo6m2) and is exposed to quite different meteorologi-
cal forces, being located at latitude 20"s (compared with Myponga
Reservoir at 35"s). The calculation of P, however, yields a maximum
-
efficiency again at P 1, with a peak efficiency of 0.15 (Figure 41b). This
parameter therefore unifies the simulation data.
448 Jorg Imberger and John C . Patterson
I TEMPERATURE ("C)
0.0
2.0
--
E
4.0
E
0
6.0
8.0
10.0 I I I
21 I 40.0 60.0 80.0 100.0 120.0
FIG.40. Transverse transect across the center of a 400 m long bubble diffuser (1 mm
diameter holes spaced at 4 m) located at a depth of 9 m. Data was obtained by yo-yoing a
temperature probe through the water column at intervals of approximately 8 meters from
the surface to within 1 m of the bottom. Gas flow rate was 40 I s-'. (a) Morning with weak
stratification (April 1989) and (b) Afternoon with strong stratification (April 1989).
For the reservoir, the appropriate Pmax is (H - zT)/L, and the relationship
between L N , B and P is not direct, as both M and zT may change daily.
The calculation of L N , B would require new simulations; these have not
been performed.
The performance of a single plume is characterized by the value of P
and therefore indirectly by ; for optimum performance, a single
-
plume should be operated such that P 1. As the stratification changes,
the airflow rate should change. Further, the degree of destratification
obtained by a plume is given by A(gS,), which will generally be much less
Physical Limnology 449
b TEMPERATURE (OC)
--
E
E
W
n
I
I
28.1
28 1
-280.
./
.BUBBLER
, 280
219-x
FIG.40. (conrd.)
than gS,. Either more plumes are required or the time over which the
destratification occurs must be extended. An estimate for the number of
-
plumes required is given by n gS,/A(gS,). Patterson and Imberger
(1989), however, show that the efficiency of multiple plumes is higher
than a single plume carrying the same total airflow. The optimum
operational strategy then is to operate as many separate plumes as
-
possible, each with P 1. This of couse involves altering the airflow rate
at perhaps daily intervals, depending on the stratification, a level of
control which may not be possible. In reality, both the number of plumes
and the airflow rate are usually fixed, or variable only in the simplest
sense, such as stopping and starting the aerator.
The effect of this nonvariable operational strategy was demonstrated
by Patterson and Imberger (1989); for Myponga Reservoir, the overall
efficiency achieved was only 0.025 for a 501s-' airflow, the result of
many times when the stratification was weak and the P value extremely
low. This result was strongly dependent on the number of plumes and
450 Jorg Imberger and John C . Patterson
a
20
- 1
I o9 1b-2 1b-1 102
FIG.41. (a) The efficiency of a single bubble plume as a function of the parameter P for
-
Myponga Reservoir, showing a clear maximum at P 1. (b) The efficiency for a single
bubble plume as a function of P for Harding Reservoir. The maximum is again at P 1.-
Physical Limnology 45 1
XnI. Summary
Advances in physical limnology in the last 10 years, both in the more
traditional area of internal wave seiching and in the newer area of
small-scale motions and mixing, has created a very large discipline. We
have highlighted this progress in this review, and now wish to draw
together those areas which we feel require more work.
(a) Seasonal behavior. Lakes in the tropics may be compared with
those in temperate regions if the stability and forcing are matched. This
may be done by evaluating two nondimensional numbers: the Wedder-
burn number W ,which describes the behavior of the surface layer, and
the Lake number L N , which describes the response of the entire lake.
This hypothesis was formed from a review of earlier work on the
intercomparison of lakes; a detailed intercomparison study of a large
range of lakes in different locations should now be conducted. On the
basis of this it may be found that further nondimensional numbers are
required to specify higher mode responses.
(b) Surface fluxes. A great deal of progress has been made, mainly by
oceanographers, in determining the exchange of momentum, heat, and
water vapor at the air-water interface. Provided the atmospheric internal
boundary layer is well established and the air is either not too stable or
unstable, there are adequate procedures for calculating surface fluxes.
However, few lakes studies have data of sufficient quality to enable the
use of the stability correction. This, together with the well-documented
fact that surface roughness has a strong influence, requires that meteoro-
logical stations be located in the internal boundary layer and that both
stability and roughness be incorporated through a model for the fetch
duration and radiation of waves. Last, the wind is spatially highly
variable over the surface of the lake. This leads to energy input at a
basin-scale length that causes deep hypolimnetic mixing and contributes
452 Jorg Imberger and John C . Patterson
Acknowledgements
Sections I to X were written by Jorg Imberger and Sections XI and XI1 were written by
John Patterson. We would like to thank Greg Ivey and Graeme Hocking for their
comments on the manuscript. Carolyn Oldham prepared the figures and performed the
computations, and Colleen Henry-Hall edited the manuscript and prepared it for
publication. This assistance is gratefully acknowledged. This work was supported by the
Centre for Environmental Fluid Dynamics, the Australian Research Council, the Australian
Water Research Advisory Council, and the West Australian Water Authority.
References
Akiyama, J., and Stefan, H. G. (1984). Plunging flow into a reservoir: Theory, J . Hydr.
Eng. 110,484-499.
Akiyama, J., and Stefan, H. G. (1985). Turbidity current with erosion and deposition. J .
Hydr. Eng. 111, 1473-1496.
Akiyama, J., and Stefan, H. G. (1987). Gravity currents in lakes, reservoirs and coastal
regions: Two-layer stratified flow analysis. Proj. Rept. No. 253, University of
Minnesota.
Anderson, R. J., and Smith, S. D. (1981). Evaporation coefficient for the sea surface from
eddy flux measurement. J . Geophys. Res. 86, 449-456.
Andre, J. E., and Lecarrere, P. (1985). Mean and turbulent structures of the oceanic
surface layer as determined from one-dimensional third-order simulations. J . Phys.
Oceanogr. 15, 121-132.
Arai, J. (1964). Some relations between the thermal property of lake and its fetch size.
Geogr. Rev. of Japan 37, 131-137.
Arita, M., Jirka, G. H., and Tamai, N. (1986). Classification and mixing of two-
dimensional buoyant surface discharges. 1. Hydr. Res. 24, 333-345.
Asaeda, T., and Imberger, J. (1988). Structures of bubble plumes in stratified environ-
ments. Environmental Dynamics Rept. ED-88-250, Centre for Water Research,
University of Western Australia.
Atkinson, J. F. (1988). Interfacial mixing in stratified flows. J. Hydr. Res. 26, 27-31.
Atkinson, J. F.,and Harleman, D. R. F. (1983). A wind-mixer layer model for solar ponds.
Solar Energy 31,243-259.
Bean, B. R., Emmanuel, C. B., Gilmer, R. O., and McGavin, R. E. (1975). The spatial
and temporal variations of the turbulent fluxes of heat, momentum and water vapor
over Lake Ontario. J. Phys. Oceanogr. 5 , 532-540.
Bejan, A., Al-Homoud, A. A., and Imberger, J. (1981). Experimental study of
high-Rayleigh number convection in a horizontal cavity with different end tempera-
tures. J. Fluid Mech. 109, 283-299.
Bejan, A., and Rossie, A. N. (1981). Natural convection in a horizontal duct connecting
two fluid reservoirs. J . Heat Transfer 103, 108-113.
Bejan, A., and Tien, C. L. (1978). Laminar natural convection heat transfer in a horizontal
cavity with different end temperatures. J . Heat Transfer 100,641-647.
Bella, D. A. (1970). Simulating the effect of sinking and vertical mixing on algal population
dynamics. J. Water Pollution Control Fed. 42(5), 140-152.
Bennett, J. R. (1971). Thermally driven lake currents during the spring and fall transition
periods. In “Proc. 14th C o d . Great Lakes Res.,” Ann Arbor p. 535. Intl. Assoc. for
Great Lakes Res.
456 Jorg Zmberger and John C . Patterson
Benney, D. J. (1966). Long nonlinear waves in fluid flows. J. Math. Phys. Sci. 45, 52-63.
Blake, S., and hey, G. N. (1986a). Density and viscosity gradients in zoned magma
chambers, and their influence on withdrawal dynamics. J . Volcanology and Geothermal
Res. 30,201-230.
Blake, S . , and hey, G. N. (1986). Magma-mixing and the dynamics of withdrawal from
stratified reservoirs. J . Volcanology and Geothermal Res. 27, 153-178.
Blanc, T. V. (1983). A practical approach to flux measurements of long duration in the
marine atmospheric surface layer. J. Clim. Appl. Met. 22, 1093-1110.
Blanc, T. V. (1985). Variation of bulk-derived surface flux, stability, and roughness results
due to the use of different transfer coefficient schemes. J . Phys. Oceanogr. 15,
650-669.
Blanc, T. V. (1987). Accuracy of bulk-method-determined flux, stability and sea surface
roughness. J . Geophys. Res. 92,3867-3876.
Blanton, J. 0.(1973). Vertical entrainment into the epilimnia of stratified lakes. Limnol.
Oceanogr. 18,697-704.
Bloss, S., and Harleman, D. R. F. (1979). Effect of wind-induced mixing on the seasonal
thermocline in lakes and reservoirs. Rept. No. 249, Massachusetts Institute of
Technology.
Bohan, J. P., and Grace, J. L. (1973). Selective withdrawal from man-made lakes. Tech.
Rept. No. H-73-4,Waterways Experiment Station, Vicksburg, Mississippi.
Boyce, F. M., Robertson, D. G., and hey, G. N. (1983). Summer thermal structure of
Lake Ontario off Toronto: Cooling the big city. Atmos. Ocean. 21,397-417.
Brady, J. A., Davis, J. M., and Douglas, E. W.(1983). First filling of Kielder Reservoir
and future operational considerations. J. Inst. Water Eng. Sci. 37, 295-312.
Brim, W. D., and Beard, J. D. (1980). Reservoir management by pneumatic induced
circulation. Paper No. 5-16.In “Proc. Symp. on Surface Water Impoundments,” (H.
G. Stefan, ed.), ASCE, pp. 877-885. Minneapolis, Minnesota.
Britter, R. E., and Simpson, J. E. (1978). Experiments on the dynamics of a gravity current
head. J. Fluid Mech. 88, 223-240.
Britter, R. E., and Simpson, J. E. (1981). A note on the structure of the head of an
intrusive gravity current. J . Fluid Mech. 112,459-466.
Brocard, D. N., and Harleman, D. R. F. (1980). Two-layer model for shallow horizontal
convective circulation. I. Fluid Mech. 100, 129-146.
Browand, F. K.,Guyomar, D., and Yoon, S. C.(1987). The behavior of a turbulent front
in a stratified fluid: Experiments with an oscillating grid. J . Geophys. Res. 92,
5329-5341.
Brown, I. K., Wooley, D. A., and Jory, A. G. (1982). Artificial destratification of Lake
Moms to improve water quality. In “Proc. Hydr. and Water Resources Symp.,”
Institution of Engineers, Australia, Melbourne, Victoria.
Brubaker, J. M. (1987). Similarity structure in the convective boundary layer of a lake.
Nature 330,742-745.
Bryant, P. J., and Wood, I. R. (1976). Selective withdrawal from a layered fluid. J. Fluid
Mech. 77,581-591.
Buchak, E. M., and Edinger, J. E. (1984). Generalized, longitudinal-vertical hydrodynam-
ics and transport: Development, programming and applications. Doc. No. 84-WR,
Waterways Experiment Station, Vicksburg, Mississippi.
Bulson, P. S. (1967).Theory and design of bubble breakwaters. Dock and Harbour Aurh.
48(560), 47-54.
Bunn, S . E., and Edward, D. H. D. (1984). Seasonal meromixis in three hypersaline lakes
on Rottnest Island, Western Australia. Aust. J . Mar. Freshw. Res. 35,261-265.
Physical Limnology 457
Bums, F. L. (1977). Localized destratification of large reservoirs to control discharge
temperatures. Prog. Waf. Tech. 9, 53-63.
Busch, N. E. (1977). Fluxes in the surface boundary layer over the sea. In “Modeling and
Prediction of the Upper Layers of the Ocean” (E. B. Kraus ed.), pp. 72-91, Pergamon
Press, New York.
Businger, J. A. (1955). On the structure of the atmospheric surface layer. J . Meteor. U ,
553-561.
Businger, J. A. (1973). Turbulent transfer in the atmospheric surface layer. In “Workshop
on Micrometeorology” (D. A. Hauger ed.), pp. 67-100. Science Press, Ephrata,
Pennsylvania.
Cacchione, D., and Wunsch, C. (1974). Experimental study of internal waves over a slope.
J, Fluid Mech. 66, 223-239.
Caldwell, D. R. (1983). Oceanic turbulence: Big bangs or continuous creation? 1. Geophys.
Res. 88,7543-7550.
Carmack, E. C. (1986). Circulation and mixing in ice-covered waters. In “The Geophysics
of Sea Ice” (N. Untersteiner ed.), pp. 641-712. Plenum Press, New York.
Carslaw, H. S., and Jaeger, J. C. (1978). Conduction of Heat in Solids, Third Edition.
Oxford University Press.
Carson, D. J., and Richards, P. J. R. (1978). Modeling surface turbulent fluxes in stable
conditions. Boundary Layer Met. 14, 67-81.
Cavaleri, L., and Zecchetto, S. (1987). Reynolds stresses under wind waves. J. Geophys.
Res. 92,3894-3904.
Chao, S . Y. (1988). River-forced estuarine plumes. J. Phys. Oceanogr. 18, 72-88.
Charnock, H. (1955). Wind stress on a water surface. Quart. J . Roy. Meteor. SOC. 82,
639-640.
Chen, J. C. (1980). Studies on gravitational spreading currents, Rept. No. KH-R-40,
California Institute of Technology.
Cheung, F. B., and Epstein, M. (1987). Two-phase gas bubble-liquid boundary layer flow
along vertical and inclined surfaces. Nuclear Eng. and Design. 99, 93-100.
Christodoulou, G. G. (1986). Interfacial mixing in stratified flows. J . Hydr. Res. 24, 77-92.
Chu, V. H., and Baddour, R. E. (1984). Turbulent gravity-stratified shear flows. 1. Fluid
Mech. Us,353-378.
Chu, V. H., and Jirka, G. H. (1985). Surface buoyant jets and plumes. In “Encyclopedia of
Fluid Mechanics” (N. Cheremisinoff ed.), pp. 1053-1084. Gulf Publishing, Houston,
Texas.
Clarke, R. H., Dyer, A. J., Brook, R. R., Reid, D. G., and Troup, A. J. (1971). The
Wangara experiment: Boundary layer data. Tech. Paper No. 19, CSIRO Div. Met.
Phys., Melbourne, Victoria.
Claussen, M. (1987). The flow in a turbulent boundary layer upstream of a change in
surface roughness. Boundary Layer Met. 40,31-86.
Cohen, A. (1979). Existence and regularity for solutions of the Korteweg-de Vries
equation. Arch. Rat. Mech. Anal. 71, 143-175.
Callings, I. L. (1985). Two large Froude numbers cusped free-surface flows due to a
submerged line source or sink. Manuscript. Deakin University, Victoria.
Colman, J. A,, and Armstrong, D. E. (1987). Radon-222 determined vertical eddy
diffusivity in the benthic boundary layer of ice-covered lakes. Limnol. Oceanogr. 32,
577-591.
Cooley, P., and Harris, S. L. (1954). The prevention of stratification in reservoirs. J. Inst.
Water Eng. 8, 517-537.
Cormack, D. E., Leal, L. G., and Imberger, J. (1974). Natural convection in a shallow
458 Jorg Imberger and John C. Patterson
cavity with differentially heated end walls. Part 1: Asymptotic theory. J. Fluid Mech.
65,209-230.
Cormack, D. E., Stone, G. P., and Leal, L. G. (1975). The effect of upper surface
conditions on convection in a shallow cavity with differentially heated end walls. Zntl. J .
Heat Mass Transfer 18,635-648.
Craya, A. (1949). Recherches theoriques sur I’ecoulement de couches superposees de
fluides de densites differentes. La Houille Blanche Jan.-Feb., 44-55.
Csanady, G. T. (1972). Response of large stratified lakes to wind. J . Phys. Oceanogr. 2,
3-13.
Csanady, G . T. (1982). “Circulation in the Coastal Ocean.” D. Reidel Publishing, Boston.
Dake, J. M. K., and Harleman, D. R. F. (1969). Thermal stratification in lakes: Analytical
and laboratory studies. J. Water Resources Res. 5 , 484-495.
Darbyshire, J., and Colclough, M. (1972). Measurement of thermal conductivities in a
freshwater lake. Geofiica Pura i Applicata 93,151-158.
Darden, R. B., Imberger, J., and Fischer, H. (1975). Jet discharge into a stratified
reservoir. J. Hydr. Diu. ASCE 101, 1211-1220.
Davis, J. M. (1980). Destratification of reservoirs-A design approach for perforated-pipe
compressed-air systems. Water Services 84, 497-504.
de Szoeke, R. A. (1980). On the effects of horizontal variability of wind stress on the
dynamics of the ocean mixed layer. J . Phys. Oceanogr. 10,1439-1454.
de Szoeke, R. A., and Rhines, P. B. (1976). Asymptotic regimes in mixed-layer deepening.
J . Marine Res. 34, 111-116.
Deardorff, J. W. (1968). Dependence of air-sea transfer coefficients on bulk stability. J .
Geophys. Res. 73,2549-2557.
Deardorff, J. W. (1970). Convective velocity and temperature scales for the unstable
planetary boundary layer. J. Atinos. Sci. 27, 1211-1213.
Deardorff, J. W. (1974). Three-dimensional numerical study of the height and mean
structure of a heated planetary boundary layer. Boundary Layer Met. 7,81-106.
Debler, W. R. (1959). Stratified flow into a line sink. J . Eng. Mech. Diu. 85, 51-65.
Denton, R. A,, and Wood, I. R. (1981). Penetrative convection at low Peclet number. J .
Fluid Mech. 1l3, 1-21.
Dickey, T. D., and Simpson, J. J. (1983). The influence of optical water type on the diurnal
response of the upper ocean. Tellus 35B, 142-154.
Dillon, T. M., Richman, J. G., Hansen, C. G., and Pearson, M. D. (1981). Near surface
turbulence measurements in a lake. Nature 290, 390-392.
Ditmars, J. D., and Cedenvall, K. (1974). Analysis of air-bubble plumes. In “Proc. Coastal
Eng. Conf.,” Copenhagen. Ch. 128, pp. 2209-2226.
Djordjevic, V. D., and Redekopp, L. G. (1978). The fission and disintegration
- of internal
solitary waves moving over two-dimensional topography. J. Phys. Oceanogr. 8 ,
1016-1024.
Dobson, F. W., Hasse, L., and Davis, R. (1980). “Air-Sea Interaction Instruments and
Methods.” Plenum Press, New York.
Donelan, M. A. (1980). Similarity theory applied to the forecasting of wave heights,
periods and directions. In “Proc. Canadian Coastal Conf. ,” pp. 47-61. Burlington,
Ontario.
Donelan, M. A. (1982). The dependence of the aerodynamic drag coefficient on wave
parameters. In “First Intl. Conf. on Meteorology and Air/Sea Interaction of the
Coastal Zone” (preprint volume), pp. 381-387. The Hague, American Met. SOC.,
Boston.
Donelan, M. A., and Peirson, W. J. (1987). Radar scattering and equilibrium ranges in
Physical Limnology 459
(1985). Three dimensional flow in the upper ocean. Science 227, 1552-1556.
Wetzel, R. G. (1975). “Limnology.” W. B. Saunders, Philadelphia, Pennsylvania.
Wieringa, J. (1974). Comparison of three methods for determining strong wind stress over
Lake Flevo. Boundary Layer Met. 7,3-19.
Wilkinson, D. L. (1972). Dynamics of contained oil slicks. J. Hydr. Diu. ASCE 98,
1013-1030.
Wilkinson, D. L. (1979). Two-dimensional bubble plumes. J. Hydr. Diu. ASCE 105,
139-154.
Wilkinson, D. L., and Wood, I. R. (1972). Some observations on the motion of the head of
a density current. J. Hydr. Res. 10, 305-324.
Willis, G. E., and Deardorf€, J. W. (1974). A laboratory model of the unstable planetary
boundary layer. J. Atmos. Sci. 31, 1297-1307.
Witten, A. J., and Thomas, J. H. (1976). Steady-wind driven currents in a large lake with
depth-dependent eddy viscosity. J. Phys. Oceanogr. 6, 85-92.
Wood, I. R. (1978). Selective withdrawal from two-layer fluid. 1. Hydr. Diu. ASCE 104,
1647- 1659.
Wright, J. C. (1961). The limnology of Canyon Ferry Reservoir. IV. The estimation of
primary production from physical limnological data. Limnol. Oceanogr. 6 , 330-337.
Wu, J. (1973). Wind-induced turbulent entrainment across a stable density interface. J.
Fluid Mech. 61, 275-287.
Wu, J. (1980). Wind-stress coefficientsover sea surface near neutral conditions-A revisit. J.
Phys. Oceanogr. 10,727-740.
Wu,J. (1985). Parameterization of wind-stress coefficientsover water surfaces. J. Geophys.
Res. 90,9069-9072.
Wucknitz, J. (1976). Determination of turbulent fluxes of momentum and sensible heat
from fluctuation measurements and the structure of wind field over waves above the
tropical Atlantic during ATEX. “Meteor” Forschungsergab 11, 25-50.
Wuest, A. (1987). Ursprung und grosse von mischungsprozessen im hypolimnion natur-
licher seen. Rept. No. 8350, Eidgenossischen Technischen Hoschschule, Zurich.
Wuest, A., Imboden, D. M., and Schurter, M. (1988). Origin and size of hypolimnic mixing
in Urnersee, the southern basin of Vienvaldstattersee (Lake Lucerne). Schweiz. Z .
Hydrol. 50.
Wunderlich, W. D., and Elder, R. A. (1969). Selective withdrawal from density stratified
reservoirs. Rept. No. 21, Water Resources Res. Advance.
Yih, C. S. (1965). “Dynamics of non-homogeneous fluids.” Macmillan, New York.
Yih, C. S. (1980). “Stratified Flows.’’ Academic Press, New York.
Yoshimura, S. (1936). A contribution to the knowledge of deep water temperature of
Japanese lakes. Japanese J. Astro. Geophys. 13, 61-120.
Zeman, O., and Tennekes, H. (1977). Parameterization of the turbulent energy budget at
the top of the daytime atmospheric boundary layer. J. Amos. Sci. 34, 111-123.