You are on page 1of 15

Aquat. Sci.

71 (2009) 65 – 79
1015-1621/09/010065-15 Aquatic Sciences
DOI 10.1007/s00027-009-8063-3
 Birkhuser Verlag, Basel, 2009

Research Article

Predicting the onset of thermal stratification in shallow inland


waterbodies
Brett F. Branco* and Thomas Torgersen

University of Connecticut, Department of Marine Sciences, 1080 Shennecossett Road, Groton, CT 06340, USA

Received: 11 October 2007; revised manuscript accepted: 20 October 2008

Abstract. Diurnal thermal stratification in shallow profiles, standard weather data and measured light
lakes and ponds is well documented and affects attenuation coefficients. Though uncertainties in B
biogeochemical gradients. The onset of thermal and u* can produce errors in IS on the order of 40 –
stratification should occur when the stratification 60 %, the parameter is robust for predicting the onset
parameter IS = -HB/u* rises above a critical value, of stratification. Accuracy can be improved by site-
Icrit. Previous applications of the stratification param- specific determinations of wind sheltering. The critical
eter lack a common definition for the onset stratifi- value for the onset of stratification was found to be
cation and include assumptions inappropriate for approximately 1. The stratification parameter can be
shallow inland waterbodies. Here, the onset of strat- an effective screening tool for 1) predicting the
ification is defined relative to a practical threshold presence of significant chemical gradients, and 2)
given by the Rohr-Stillinger criterion, N2 = e/(20n). determining the minimum wind speed required to
Using this definition, the utility of IS for predicting prevent diurnal thermal stratification from develop-
stratification in two shallow ponds was evaluated ing.
using select periods from over 600 days of temperature

Keywords. Shallow lakes; buoyancy flux; diurnal mixed layer; chemical stratification.

Introduction coupling between the diurnal physical dynamics and


the dynamics of phytoplankton(Ganf, 1974), dissolved
Shallow lakes and ponds can stratify and destratify oxygen (Gu et al., 1996; Ford et al., 2002; Branco,
(with regard to both density and chemistry) on a daily 2007), methane (Ford et al. 2002), and other biogeo-
basis, and may even remain persistently stratified over chemical processes has been demonstrated, yet it is
multiple days (MacIntyre, 1993; Gu et al., 1996; still common in the literature to sample shallow lakes
Condie and Webster, 2001; Ford et al., 2002). Thus, and ponds as if they were well-mixed with little or no
the entire water column of a shallow lake or pond may justification. The well-mixed assumption is likely to be
behave as a diurnal mixed layer (diurnal stratification/ false in many instances and can lead to inappropriate
destratification cycle), the physical dynamics of which sampling strategies or misleading interpretation of
are described in detail by Imberger (1985). Strong results.
A simple parameter for predicting the onset of
thermal stratification in wind-mixed waterbodies can
* Corresponding author phone: +1 718 951-5000 x6441;
fax: +1 718 951-4753; e-mail: bbranco@brooklyn.cuny.edu be used to assess the physical dynamics in a small
Published Online First: February 12, 2009 shallow lake or pond prior to sampling or to assess how
66 B. F. Branco and T. Torgersen The onset of stratification

stratification might have impacted the sample taken. Table 1. Definitions and units of variables used.
Such a parameter has been defined previously by Variable Definition Units
Holloway (1980) for estuaries and by Imberger (2001)
B buoyancy flux m2 s-3
for lakes. Hereafter, the term stratification will be CD drag coefficient -
used to indicate density stratification caused by cp specific heat capacity of water J kg-1 C-1
thermal gradients unless indicated otherwise. In this g gravitational constant m s-2
H maximum water depth m
paper, we (i) evaluate the utility of this parameter for Icrit critical value of the stratification parameter -
predicting the onset of stratification in two shallow IS stratification parameter -
ponds, (ii) determine the critical value of the strat- LM Monin-Obukhov length scale m
m’ net wind power transmission coefficient -
ification parameter that defines the practical onset of Q heat flux per unit area W m-2
stratification, (iii) discuss the limitations and uncer- u* water-side friction velocity m s-1
tainties associated with applying the stratification z depth from the air-water interface (positive m
down)
parameter to shallow lakes and ponds, and (iv)
N2 buoyancy frequency squared s-2
provide a method for calculating the minimum wind NRS2 stratification threshold from the Rohr- s-2
speed necessary to prevent stratification. The link Stillinger criterion
between the onset of stratification and the onset of Y solar penetration function -
a thermal expansion coefficient C-1
dissolved oxygen gradients in the study ponds is used e rate of turbulent kinetic energy dissipaton m2 s-3
as an example of the biogeochemical importance of h attenuation coefficient m-1
the stratification parameter. hVIS attenuation coefficient for visible light m-1
k von Karmans constant -
n viscosity m2 s-1
The stratification parameter 1 density kg m-3
Assuming water temperature is above the temperature
of maximum density, a net heat flux into the water
column across the air-water interface creates a negative Though the value of Holloway (1980) was for a
buoyancy flux (B) that tends to stratify the water shallow estuary, it was based on an empirical deter-
column and lower its potential energy (Kraus and mination of m and lacked a definition of stratification
Turner, 1967; Simpson and Hunter, 1974; Kim, 1976). against which to assess the data.
Note that the z-axis is defined as positive downwards. In The stratification parameter (IS) is similar to the
order to maintain a well-mixed water column, the rate well-known Monin-Obukhov length scale (LM) as
of mixing energy input must offset the stratifying applied to the surface layer of lakes (Imberger, 1985):
influence of the buoyancy flux (Simpson and Hunter,
1974; Simpson et al., 1978; Simpson and Bowers, 1981). u3
LM ¼  (3)
The balance between the stratifying and mixing energy B
inputs can be expressed as a dimensionless stratifica-
tion parameter (Holloway, 1980; Imberger, 2001): Thus, the relationship between IS and LM is:

HB H
IS ¼  (1) ¼ IS (4)
u3 LM

where H is the water column depth and u* is the The onset of stratification is expected to occur when
friction velocity (see Table 1 for units). The onset of the ratio of H to LM is order 1 because LM indicates the
stratification begins when IS rises above a critical value depth scale over which the shear production and the
given by: buoyant destruction of turbulent kinetic energy are
roughly equal (Imberger, 1985). The difficulty in
Icrit = 2 m (2) applying eq. 1 and eq. 2 as a precise criterion for the
onset of stratification in shallow lakes and ponds is
where m’is the wind power transmission coefficient, determining the values of m’ and B. In particular,
discussed later. Imberger (2001) recommends a value many of the assumptions and simplifications used in
of Icrit = 0.2 for lakes and Holloway (1980) determined deeper systems do not apply to shallow systems, as
a value of Icrit = 3.6 for a shallow wind mixed estuary. discussed in the following sections.
The value of Imberger (2001) was based on surface
layers of lakes with seasonal thermoclines, an arbi- Wind power transmission coefficient
trary definition of stratification corresponding with a Some of the wind energy entering a waterbody is
0.2 8C temperature differential and an analysis that transferred to seiches, surface waves or internal waves
included data from late afternoon destratification. and is subsequently lost to bottom friction. More
Aquat. Sci. Vol. 71, 2009 Research Article 67

Table 2. Shortwave radiation bandwidth fractions of the total energy (f) and composite attenuation coefficients (h)
wavelength range (nm) f h (m-1) 95 % absorption depth (m)
< 400 (UV) 0.046 assume same as VIS variable
400 – 700 (VIS) 0.430 measured variable
700 – 910 0.214 2.92 1.0
910 – 950 0.020 20.4 0.15
950 – 1090 0.089 29.5 0.10
1090 – 1350 0.092 98.4 0.03
> 1350 0.109 2880 < 110-4

importantly, a considerable fraction of the wind surface heat flux at the air-water interface (Q0net )
energy is lost to viscous dissipation that can account includes the net longwave radiation emission, the
for 80 – 90 % of the surface energy fluxes into a surface evaporative heat flux, and the sensible heat flux
mixed layer at equilibrium (Imberger, 1985). The wind (Henderson-Sellers, 1984). The penetrating solar
power transmission coefficient (m’) can be defined as shortwave radiation (SWR) flux (Qswr) is treated
“the ratio between the rate of increase of potential separately from the surface heat fluxes because it
energy in the water column and the wind energy” causes internal heating of the water column. The SWR
input to the system (Holloway, 1980). The value of m’ flux can be approximated for heat budget analyses and
has been determined empirically, often in relatively buoyancy flux calculations by parsing the full spec-
deep lakes or ocean waters that are stratified. Some of trum into discrete bandwidths with composite attenu-
the wind energy lost from the wind-mixed surface ation coefficients (Jassby and Powell, 1975; Jellison
layer in deep stratified systems leaks through the and Melack, 1993; Bean, 2005). Each bandwidth is
thermocline. In shallow, initially unstratified lakes, the attenuated as a function of depth according to Beers
surface-mixed layer terminates at the sediment-water Law (Mobley, 1994):
interface. Thus, it is unclear if previous empirical
determinations of m’ in deeper water columns are X
7
applicable to the onset of stratification in shallow Qzswr ¼ Qoswr fi expðhi zÞ (6)
lakes and ponds. i¼1

One of the few attempts to quantify m’ in an


initially unstratified shallow wind mixed system was where fi is the fraction of energy contained in the ith
made by Holloway (1980). He estimated a mean value bandwidth, and hi is the composite attenuation
of m’ = 1.8 based on four independent estimates coefficient of the ith bandwidth.
ranging from 0.73 to 4.5 in a shallow, wind-mixed When eq. 6 is substituted into eq. 5, the expression
estuary during periods of weak and ephemeral strat- for the buoyancy flux becomes:
ification. The large uncertainty in previous estimates
of m’ and the questionable generality of its value over ga  o 
B¼ Qnet þ ð1 þ Y ÞQoswr (7)
different depth scales mean that Icrit can not be 1cp
calculated precisely from eq. 2. Rather, the value for
Icrit needs to be determined directly through empirical where Y is a solar penetration function (Kim 1976):
methods, as shown in this work. However, from
Holloway (1980), m’ and thus Icrit, are expected to be X7  
1  expðhi H Þ expðhi H Þ
order 1. Y ¼2 fi  (8)
i¼1
hi H 2
Buoyancy flux in sallow water
The buoyancy flux (B) is given by the expression: Applying the seven bandwidth approximation of Bean
(2005; reproduced here in Table 2), the terms in the
0 1 summation of eq. 8 associated with the four band-
Q0net þ Q0swr  QHswr
ga B Z H C widths > 910 nm are negligible because they attenuate
B¼ @ dQzswr A (5)
1cp þ 2 ðH  zÞdz rapidly within the top 10 – 15 cm of the water column
H 0 dz (Table 2). At a water depth of 1 m and a light
attenuation coefficient of 2 m-1, neglecting these
where 1 is the density of water, g is the gravitational terms introduces an error of 2 % in Y.
constant, a is the thermal expansion coefficient, cp is The attenuation of the ultraviolet (UV: < 400 nm)
the specific heat capacity of water, and z is depth into and visible (VIS: 400 – 700 nm) fractions of the total
the water column (positive downwards). The net SWR spectrum is strongly influenced by the concen-
68 B. F. Branco and T. Torgersen The onset of stratification

tration of colored dissolved organic matter, chloro- Methods


phyll and suspended particulate matter (Mobley,
1994; Morris et al., 1995). The composite attenuation Study sites
coefficient of visible light (hVIS) was routinely meas- The stratification parameter was applied to Mirror
ured during our work, but the attenuation coefficient Lake (N 41.8078, W 72.2478) and Swan Lake (N
of UV was not. The UV and VIS bandwidths are 41.8118, W 72.2528), two impoundments on the main
treated as a single fraction with the measured compo- campus of the University of Connecticut (Storrs, CT,
site attenuation coefficient of visible light. The error in USA; Fig. 1 and Table 3), using data from 2002, 2003
this simplification is variable depending on the and 2004. Both of these small, shallow ponds are
characteristics of the particular waterbody in ques- mesotrophic to eutrophic. Mirror Lake is a phyto-
tion, but is limited by the small percentage of the total plankton-dominated pond with no submerged macro-
energy contained in the UV fraction (< 5 %). The phytes observed. Chlorophyll a concentrations in
attenuation of the 700 – 910 nm fraction (21.4 %) is not Mirror Lake typically range from 1 – 100 mg L-1. The
significantly affected by the constituents in the water attenuation coefficient for visible light typically rang-
column (Ohlmann et al., 2000) and should be treated es from 2 to 5 m-1. Chlorophyll a values up to 400 mg L-1
separately. After accounting for the assumptions and and attenuation coefficients up to 10 m-1 have been
simplifications discussed above, eq. 8 becomes: observed during intense cyanobacteria blooms. The
bottom of Swan Lake was blanketed by the macro-
  phytes Elodea until 2003 when herbicides were
1  ehVIS H ehVIS H
Y ¼ 0:952  applied. During 2004, no macrophytes were observed
hVIS H 2
  (9) in Swan Lake and chlorophyll a concentrations ranged
1  e292H e2:92H from < 1 to 20 mg L-1. The attenuation coefficient
þ 0:428 
2:92H 2 ranged from 1 to 2.5 m-1 during this period. The
sediments of Mirror Lake are loosely consolidated
Including a separate term for the 700 – 910 nm and fine and can be described as a black ooze. The
bandwidth improves the accuracy of Y by almost bottom of Swan Lake is generally sandier with some
10 % over an expression that includes this bandwidth gravel regions.
with the UV and visible portion of the spectrum (for
H = 1 m and hVIS = 2 m-1). The improvement in accu- Table 3. Physical characteristics of Mirror Lake and Swan Lake on
racy increases with increasing depth and decreasing the campus of the University of Connecticut (Storrs, CT, USA).
light attenuation. Parameter Mirror Lake Swan Lake
Note that the buoyancy flux expression presented
here is a simplification that does not consider the area (ha) 2.05 0.99
volume (m3) 14643 9710
effects of rain, resuspended sediments and dissolved mean depth (m) 0.72 0.98
constituents. At our sites, the water column was always maximum depth (m) 1.5 1.8
unstratified during rain events when cloudy skies length (m) 229 178
width (m) 120 78
reduce the incident SWR fluxes and higher winds and maximum fetch (m) 200 177
inflows increase the energy available to maintain well-
mixed conditions. Thus, the stratification parameter
has no utility on rainy days at our sites. The simple case
of thermal density stratification onset is considered Over 600 days of high temporal and spatial
here and the contributions of sediment and chemical resolution temperature and oxygen data were col-
contributions to the onset of stratification are neglect- lected during the spring, summer and fall periods of
ed. This approach is justified for this work because 2002 – 2004 in the locations indicated by the white
examination of over four years of data from the study circles in figure 1. Temperature was measured with a
sites indicates that the appearance of gradients in chain of thermistors spaced 10 to 20 cm apart
dissolved constituents or turbidity (as a proxy for (Branco et al., 2005). Each thermistor has a precision
suspended sediments) always lags the onset of thermal of  0.01 8C, but the variability amongst thermistors
gradients. While suspended sediments and gradients was  0.02 8C (Bean, 2005). During the spring and
in dissolved constituents may reinforce the stratifica- fall, it is typical to observe temperature stratification
tion once it begins or contribute to local stratification on some days and unstratified conditions on others
within the sediment-water boundary layer, they are (Fig. 2). Oxygen data were collected using the pump-
not major factors in driving the onset of stratification profiler system described in Branco et al. (2005).
at the scale of the entire water column. Light attenuation coefficients were determined from
Aquat. Sci. Vol. 71, 2009 Research Article 69

Figure 2. Thermistor data in Mirror Lake for 23 – 30 March 2003.


Only the surface (115 cm above bottom) and bottom thermistors
(10 cm above bottom) are shown for clarity. The total water depth
at this location in the pond was 120 cm.

intervals (approximately one-hour intervals) were


obtained from the Windham Airport station (N
41.7428, W72.1848), which is approximately 9 km
southeast of Mirror Lake and 9.7 km southeast of
Swan Lake. The relative humidity data were interpo-
lated onto a 30 min grid to match the frequency of the
farm data. Surface heat fluxes were calculated from
bulk formulae (Appendix I) using weather data from
the farm and airport and surface water temperature
from the highest thermistor in the chain.

Practical threshold of stratification


Weak and ephemeral stratification may be detectable,
yet have negligible impact on the biogeochemistry on
hourly timescales. We define stratification in terms of
a practical threshold above which the time scale of
mixing is reduced relative to the time scale of diurnal
biogeochemical dynamics. Temperature gradients are
Figure 1. Mirror Lake (ML) and Swan Lake (SL) on the campus of used often as a measure of stratification (Jellison and
the University of Connecticut (Storrs, CT, USA). White circles Melack, 1993; Fee et al., 1996) because they are easy
indicate the location where temperature and oxygen profiles were to measure and the precision of typical thermistors
acquired.
used in field studies allows detection of temperature
differences < 0.02 8C. However, since density is a non-
light profiles taken with a Li-CORTM spherical linear function of temperature, identical thermal
Quantum PAR Sensor. gradients may correspond to different density gradi-
Air temperature, wind speed at 2 m, wind direc- ents for different mean water column temperatures.
tion, SWR (wavelengths 280 to 2800 nm) and rainfall The square of the buoyancy frequency (N2 in s-2) is a
data were obtained from the University of Connect- preferred measure of stratification since it is a direct
icuts Plant Science Farm (N 41.7958, W 72.2288), expression of the density gradient:
located approximately 2 km southeast of Mirror Lake
and 2.7 km southeast of Swan Lake. The data are g d1
N2 ¼ (10)
30 min averages except for the rainfall data, which are 1o dz
30 min totals. Relative humidity data at irregular time
70 B. F. Branco and T. Torgersen The onset of stratification

Table 4. Mean temperature (Tmean), temperature range (Trange), and light attenuation coefficient (hVIS) used in the calculation of Y (eq. 17).
Period Pond Tmean1 Trange2 hVIS (m-1)
5 – 10 October 2003 Mirror 17.5 14.7 – 22.1 2.23
23 – 29 March 2003 Mirror 12.1 9.3 – 15.8 2.04
9 – 11 May 2003 Mirror 17.8 15.1 – 21.9 2.25
24 – 25 April 2004 Swan 14.8 13.6 – 17.0 1.66
1
simple average of all thermistor temperatures over all depths during the specified period
2
the minimum and the maximum temperature recorded by any thermistor at any depth during the specified period
3
measured on 3 October
4
typical early spring value
5
measured on 13 May
6
average of all values from 2004

Here, N2 is calculated assuming a linear density isothermal water column subject only to penetrating
gradient between adjacent thermistors where density SWR. For a light attenuation coefficient of 2 m-1, it
is calculated as a function of temperature. Typical would take almost an hour to create a temperature
spacings between thermistors ranged from 10 to differential of 0.5 8C over the top 1 m with SWR = 400
20 cm. The Rohr-Stillinger active turbulence criterion W m-2, a typical midmorning value for Mirror Lake.
as described in West and Lorke (2003) can be used to The periods selected for analysis from over 600 days of
specify a stratification threshold based on N2. The data demonstrated sensitivity to changing meteoro-
criterion states that mixing is suppressed by stratifi- logical conditions and contained both thermally
cation when the rate of turbulent kinetic energy stratified and unstratified days. The ranges of wind
dissipation is below a transitional value (etr) found speed and incident SWR (Fig. 3), as well as mean
from the expression: water column temperature and light attenuation
(Table 4) are representative of the two ponds during
etr ¼ 20nN 2 (11) spring and fall.

where n is the viscosity. Assuming that the rate of


dissipation follows the law of the wall (West and
Lorke, 2003):

e ¼ u3 ðkzÞ1 (12)

where k is von Karmans constant, the threshold for N2


above which active turbulence is suppressed is found
by substituting eq. 12 into eq. 11:

2
NRS ¼ u3 ð20nkzÞ1 (13)

Practical stratification occurs when N2 is greater than


NRS2 at z = H/2. Values of n = 1.210-6 m2 s-1 and k = 0.4
were used.
Figure 3. Wind speed (dark line) and shortwave radiation (SWR;
hatched areas) for the periods of (a) 5 – 10 October 2002, (b) 23 – 29
Calculating stratification parameter (IS) from data March 2003, (c) 9 – 11 May 2003 and (d) 24 – 25 April 2004.
The values of the stratification parameter (IS) for
select periods in Mirror Lake and Swan Lake were
calculated for half-hour intervals to correspond with The friction velocity (u*) was calculated from the
the temporal resolution of the weather data. This is wind speed at 10 m:
consistent with our application of the practical strat-  
ification threshold, which is never reached instanta- 1 0:5
u ¼ CD a U10 (14)
neously. Rather, stratification will build over time to 1w
the practical threshold when surface warming be-
comes more important than wind mixing. The time- The drag coefficient was assumed to be 1.210-3 for
scale over which this occurs depends on the relative winds > 3 m s-1 and was calculated using the expres-
strength of u* and B. This is evident by considering an sion of West and Lorke (2003) for winds < 3 m s-1:
Aquat. Sci. Vol. 71, 2009 Research Article 71

1:15
CD ¼ 0:0044U10 (15) March and 26 March. On these two days, IS was
frequently > 1. On 28 March, the stratification rose
The measured wind speed height at the plant farm above the practical threshold in the morning when IS
(2 m) was converted to a height of 10 m assuming a was > 1, but quickly fell below the threshold by noon
logarithmic velocity profile (Panofsky and Dutton, when IS decreased below 1. In contrast, the stratifica-
1984): tion was below the threshold on the morning of 27
  March when IS was < 1, but was greater than the
U z threshold in the afternoon. In this case, IS increased
Uz ¼ ln (16)
k z0 above 1 only briefly, though the timing of this event
coincides with stratification rising above the thresh-
where z0 is the surface roughness, taken to be 0.01 m, a old. The results for this one-week period strongly
typical value for a grassy field over which the wind suggest that the critical value of IS is approximately 1
anemometer was located. (Icrit  1) for Mirror Lake. The results from the period
The ponds are likely to experience wind sheltering of 5 – 10 October 2002 (Fig. 5) and 9 – 11 May 2003
effects due to their small size and the proximity of (Fig. 6) support this finding. For example, on 7, 8 and
trees and buildings. A reduction in u* was estimated 10 October, stratification does not rise above the
from Rubbert and Kçngeter (2005). For Mirror Lake, threshold until the afternoon when the values of IS
u* was multiplied by a factor of 0.77 for northerly and increase above 1.
northeasterly winds due to the close proximity of The decrease of IS below 1 does not necessarily
buildings and trees to the shore. For all other wind result in immediate destratification because it takes
directions, it was assumed that sheltering was insig- time for mixing energy to degrade the thermocline.
nificant. For Swan Lake, u* was multiplied by a factor This is evident on 27 March 2003 in Mirror Lake
of 0.45 for all wind directions since the pond is more (Fig. 4) when stratification remained above the
sheltered than Mirror Lake on all sides. The assump- threshold for many hours after IS fell below 1. In this
tion for Mirror Lake is supported by the agreement case, the winds became calm around sunset leaving
between temperature data and modeling results convective overturn during surface cooling as the
(Branco, 2007), but the Swan Lake assumption primary driver for destratification. Thus, IS < Icrit is a
remains untested. The buoyancy flux was determined prerequisite for destratification, though it is not a
using eq. 7 and eq. 9. robust predictor of destratification.
When wind speed is zero, the stratification param- Application of the stratification parameter to
eter becomes undefined (eq. 1). However, the mini- Swan Lake is more difficult since the wind sheltering
mum wind speed in the farm data corresponds with the coefficient was never directly measured or modeled.
anemometers mechanical threshold of approximately Using a sheltering factor of 0.45 on 24 April 2004,
0.2 m s-1, and these values were used as recorded. The stratification remains below the threshold through-
time period of 23 – 29 March 2003 in Mirror Lake is out the day (Fig. 7). The value of IS is between 1 and 3
used to determine the critical value of IS for the onset from early morning until noon before decreasing
of practical stratification. below 1 in the late afternoon. If the wind sheltering
coefficient is accurate, this result suggests that Icrit
may be closer to 3 in Swan Lake, though still order 1
Results and consistent with the value of 3.6 specified by
Holloway (1980). However, the sheltering effect may
The values of the stratification parameter (IS) can be be overestimated. If we apply the same wind
positive or negative depending on the sign of the sheltering factor from Mirror Lake to Swan Lake,
buoyancy flux. The stratification parameter is neg- the result for 24 April 2004 is consistent with Icrit  1.
ative at night and under cloudy daytime conditions Note that changes in the wind sheltering factor
when surface cooling is greater than solar heating. For change both the stratification parameter and the
display purposes, the values of the stratification stratification threshold as they are both dependent
parameter are plotted using a log scale in figures 4 – on u*. Regardless of the wind sheltering coefficient
7 and thus negative values are not shown. used, the data from 25 April 2004 in Swan Lake are
Stratification remained below the practical thresh- consistent with Icrit  1 (Fig. 7).
old defined earlier for the entirety of the daylight
period on 23 March, 25 March and 29 March (Fig. 4).
On these three days, IS never rose above 1. In contrast,
stratification was strong and clearly exceeded the
practical threshold for most of the daylight hours on 24
72 B. F. Branco and T. Torgersen The onset of stratification

Figure 5. Same as figure 4 for the period 5 – 10 October 2002 in


Mirror Lake.

Figure 4. (a) The value of the stratification parameter (IS); and (b)
the bulk buoyancy frequency squared (N2 : black line) and the waterbody or the diurnal mixed layer (Imberger,
stratification threshold from eq. 13 (NRS2 : gray line) for the period 1985) bounded by a metalimnion or seasonal ther-
23 – 29 March 2003 in Mirror Lake. The solid line in (a) indicates
Icrit = 1.0. The vertical dashed lines indicate the onset of stratified
mocline (Imberger, 2001; Kraus and Turner 1967;
daytime periods where N2 exceeds NRS2. Daytime (black) and Kim 1976). Unattenuated SWR can be significant in
nighttime (white) periods are indicated by the bar at the top of a shallow lake or pond. When hH < 2 (for example, a
panel (b).
1 m deep water column with hVIS = 2 m-1), more than
10 % of the associated SWR reaches the bottom of
Discussion the water column. Some of the radiation will be
reflected and scattered and some will be absorbed by
Sources of uncertainty the sediments. The relative importance of each
Major uncertainties in the calculation of IS can be process is determined by the physical, biological
attributed to 1) uncertainty in the wind sheltering and chemical characteristics of the interface (Mar-
effect when using remote winds, 2) uncertainty in itorena et al., 1994).
sediment heat fluxes, 3) uncertainty in uncalibrated On hourly timescales, the exchange of sensible
heat flux formulae, and 4) uncertainty in the light heat between the sediments and water column is
attenuation coefficient. The uncertainty associated driven by diurnal changes in the water temperature
with the wind sheltering effect was discussed above. (Smith, 2002). The diurnal sensible flux is typically
The buoyancy flux calculation (eq. 8) does not into the sediments during water column heating and
consider sediment-water heat exchange and assumes out of the sediments during water column cooling,
the sediment-water interface is an open boundary for with peak fluxes into and out of the sediments
unattenuated SWR. These omissions/simplifications occurring at dusk and dawn, respectively. Thus, the
are insignificant when calculating B for a deep sediment heat flux tends to destabilize the water
Aquat. Sci. Vol. 71, 2009 Research Article 73

Figure 6. Same as figure 4 for the period 9 – 11 May 2003 in Mirror


Lake.
Figure 7. Same as figure 4 for 24 and 25 April 2004 in Swan Lake.
The line with filled squares in (a) and (b) are for a wind sheltering
coefficient of 0.45. The line with open squares in (a) and (b) are for
column in the morning and stabilize the water column a wind sheltering coefficient of 0.77. See text for explanation. The
in the afternoon. vertical dashed lines indicate the onset of stratified daytime periods
where N2 exceeds NRS2 on 25 April.
The errors associated with our treatment of
unattenuated radiation and sediment heat fluxes
depend on many factors such as the depth of the The uncertainties associated with heat flux
water column, the attenuation coefficient, the albedo equations are low based on the modeling results of
of the sediment surface, the thermal conductivity of Branco (2007). The model was able to predict the
the sediments, and the water temperature near the vertical temperature distribution and the integrated
bottom. However, some estimate of the error associ- heat budget of Mirror Lake with no systematic bias
ated with neglecting sediment heat fluxes can be on the weekly timescale and small errors on the
made. For example, the maximum sediment sensible hourly timescale using the heat flux equations in
heat flux one might expect in a 0.5-m deep water Appendix I as forcing functions. The maximum
column is approximately 10 W m-2 at the dawn and instantaneous error associated with any of the heat
dusk peaks (Smith, 2002). The SWR flux typically flux terms is estimated to be order 10 W m -2 and
ranges from 100 to 600 W m-2 during the day at usually much less. Since the daily mean surface heat
temperate latitudes, with values < 100 W m-2 occurring fluxes to Mirror Lake over the period 23 – 30 March
near dawn and dusk. Thus, the mean daily sediment 2003 were -150 W m -2, it can further be estimated
heat flux values will be orders of magnitude lower than that the maximum error in B due to heat flux
the surface heat fluxes and SWR. The uncertainty uncertainties (including sediments) is < 10 % for
associated with excluding the sediment heat flux is this period. The uncertainty or natural variability in
expected to be insignificant for most of the day, though the light attenuation coefficient (hVIS) can also add
it may be significant for the first hour after sunrise uncertainty to the value of B. For a depth of 1.2 m
when the SWR flux is low. and a light attenuation coefficient of 2 m -1, an
74 B. F. Branco and T. Torgersen The onset of stratification

uncertainty of  0.5 m -1 will result in an error on the oxygen stratification in the water column that is
order of 10 %. Simple checks on the sensitivity of B controlled by density stratification.
to combined uncertainties in the light attenuation In the small, shallow ponds studied here, a
coefficient and the heat fluxes indicate that errors stratification threshold based on the Rohr-Stillinger
on the order of 40 – 60 % can be realized. However, active turbulence criterion (eq. 13) provides a clear
since the parameter IS is changing over orders of and practical distinction between weak thermal strat-
magnitude, the uncertainties will have little effect ification and strong thermal stratification. Further-
on the value of IS relative to Icrit unless they are of the more, the practical onset of stratification has been
same order for extended periods. demonstrated to be associated with values of the
stratification parameter rising above a critical value of
Downslope convection Icrit  1 for Mirror Lake and that Icrit is order 1 for
Down-slope convection is not included in the calcu- Swan Lake. A value of Icrit = 1 corresponds with
lation of IS, but may cause stratification to increase LM = H. This is a satisfying result since stratification
early in the morning before SWR becomes significant would be expected to begin when the Monin-Obu-
and IS rises above Icrit. Downslope convection is driven khov lengthscale becomes less than the total water
by downwelling near the pond edges where water depth, and this occurs when IS increases above 1
cools and increases its density at a faster rate than the (MacIntyre et al., 2002). Thus, the critical conditions
water over deeper areas in the interior (Sturman et al., necessary for thermal stratification on the diurnal
1999). Evidence of downslope convection is found timescale are quantified in the stratification parame-
when bottom temperature continues to decrease even ter. This can now be used to identify periods when the
after surface temperature begins to increase after water column might exhibit significant chemical
sunrise. This phenomenon is evident in the ponds on stratification as well.
the mornings of 9 and 10 May 2003 in Mirror Lake and The top to bottom difference in oxygen concen-
24 and 25 April 2004 in Swan Lake. After the tration was calculated for a series of water column
downslope convection ceases early in the morning, profiles obtained with an automated pump-profiling
stratification should only continue to increase above system (Branco et al, 2005) during 9 May 2003 in
the threshold if IS rises above Icrit. Otherwise, the weak Mirror Lake and 25 April 2004 in Swan Lake (Fig. 8).
stratification induced by downslope convection will be The oxygen difference can be compared with the
short-lived. excess stratification found by subtracting the prac-
tical threshold from N2. Negative values of excess
Biogeochemical implications of stratification onset stratification indicate that stratification is below the
Smith et al. (2002) estimate that there are 2.6 million threshold, and negative oxygen differences indicate
small (< 1 ha) artificial impoundments in the con- that the surface oxygen concentration is lower than
terminous United States, with the number increasing that near the bottom. The latter condition is fre-
at an annual rate of 1 – 3% y-1. Downing et al. (2006) quently found in our ponds in the spring whenever
estimate the global abundance of natural waterbod- they become supersaturated with oxygen. Because
ies < 10 ha in size at over 300 million. Thus, small the ponds are shallow, there is adequate sunlight for
waterbodies represent over 99 % of the total number photosynthesis to occur throughout the water col-
of natural inland waterbodies and contain 31 % of the umn. The negative oxygen differentials in figure 8
total surface area (Downing et al., 2006). Many of may be due to higher rates of oxygen loss across the
these small waterbodies are likely to be very shallow air-water interface than across the sediment water
with mean depths of order 1 m or less (Wetzel, 2001; interface. The gradient may also be reinforced by the
Fig. 3 – 2). The cumulative biogeochemical impact of photosynthesis of benthic algae. Note that the oxy-
these small, shallow waterbodies is poorly defined, gen difference is generally near zero when the excess
but likely significant at the local, regional and global stratification is zero or negative, but increases in
scales. For example, small waterbodies are likely an magnitude with the onset of positive excess stratifi-
important component of the global carbon budget cation.
(Cole et al., 2007; Torgersen and Branco, 2008). The Of course, the magnitude of the oxygen difference
means and mechanisms by which shallow water- at any given time is a function of both stratification
bodies process their autochthonous and allochtho- and the biotic rates of oxygen production (photosyn-
nous carbon loading is governed in part by microbial thesis) and consumption (respiration and chemical
activity in the sediments (e.g., Torgersen and Branco, oxygen demand). However, the point to note is that
2008). Thus, the pathways and rates of the carbon large oxygen concentration gradients can develop in
processing are altered by redox and pH conditions of very shallow waters in the presence of diurnal
the sediment, which in turn are directly impacted by stratification. If using oxygen as the basis for calcu-
Aquat. Sci. Vol. 71, 2009 Research Article 75

no matter how small in magnitude or duration) were


used as criteria for IS, a lower value of Icrit would likely
be determined, and thus a lower value of m’.

The minimum wind speed to prevent stratification


In general, when IS < Icrit stratification can not rise
above the practical threshold given by the Rohr-
Stillinger criterion. Thus, the minimum wind speed
required to maintain an unstratified water column can
be predicted for a given water depth and light
attenuation coefficient by substituting eq. 14 into
eq. 1 and solving for U10 :

 1=2
min 1a CD 1
U10 ¼ ðHBIcrit Þ =3 (17)
1ref

Figure 8. Stratification above the practical threshold (N2 – NRS2 : Applying constant densities for water (1000 kg m-3)
solid line) and surface to bottom difference in dissolved oxygen
(DO) at half-hour intervals (open squares) for 25 April 2004 in and air (1.2 kg m-3) and a constant drag coefficient for
Swan Lake (top panel) and 9 May 2003 in Mirror Lake (bottom winds at 10 m (1.2x10-3), the expression for the critical
panel). Negative DO differences indicate lower oxygen concen- wind speed at 10 m (Ucrit) is:
tration near the surface driven by deep water supersaturation and
surface gas exchange.
1
Ucrit ¼ 833ðHBÞ =3 (18)
lation of ecosystem metabolism or as an indicator of
redox condition (for example), the sampling strategy The critical wind speed for a waterbody (Fig. 9) is a
must account for the stratification. A series of single function of the water depth and the buoyancy flux,
water samples taken at mid-depth would not provide which is primarily controlled by the SWR flux during
an adequate assessment of the chemical and ecolog- the day. The range of B shown in figure 9 is typical for
ical conditions in the waterbody. the spring, summer and fall at our study sites. We
emphasize that depth alone does not determine the
The upper limit on the value of m propensity of a system to stratify or remain unstrati-
The critical value for the stratification parameter fied. The ambient meteorological conditions that
(Icrit  1) suggests that an upper limit of 0.5 can be determine wind speed and the buoyancy fluxes are
assigned to the value of m’. This value lies between 0.1 equally important. As more sunlight energy is avail-
and 0.73, the implicit value of Imberger (2001) able (increasing B) it takes more wind energy to keep
(applying eq. 2 and Icrit = 0.2) and the low end of the the system unstratified for any given depth. This
Holloway (1980) estimates cited earlier. Holloway information is critical to environmental engineers,
(1980) computed m’ by observing changes in the water managers and researchers when designing artificial
column potential energy and constructing an energy ponds for storm water retention/detention, aquacul-
balance including the buoyancy flux B. However, his ture, recreation or aesthetics. For example, prevention
simplifications when formulating B would tend to of biogeochemical gradients may be an important
produce overestimates of m’ As pointed out earlier, design criterion, and we have shown that oxygen
Imberger (2001) used data associated with diurnal distribution in the water column is sensitive to the
surface layers overlying a seasonal thermocline to practical stratification threshold defined here. Com-
determine the critical value for the stratification bining estimates of light attenuation with historical
parameter. Since some of the surface energy can meteorological data to estimate U10 and B can allow
leak through the bottom of the diurnal surface layer one to determine the likely percentage of days that a
into the seasonal thermocline (Imberger, 1985), the waterbody of a specified depth will stratify, and thus
value of m’ may be lower in deep lakes than in the establish one of the conditions necessary for biogeo-
shallow waterbodies considered in this work where chemical gradients to develop.
the diurnal surface layer terminates at the sediment- The timescale of stratification onset requires that
water interface. At the same time, our estimate of weather data be available at the resolution of one hour
m’= 0.5 should be strictly considered an upper limit. If or better in order to apply the stratification parameter
measurable stratification (i.e. any gradient in density effectively. At our sites, wind speed often follows a
76 B. F. Branco and T. Torgersen The onset of stratification

parameterize the effects on wind stress and buoyancy


fluxes so that the stratification parameter can be
extended to macrophyte-dominated waters and to the
role of physics in controlling resilient trophic states
(Scheffer et al., 1993; Scheffer et al., 2001).

Persistent stratification and destratification


Finally, the stratification parameter is not applicable
in the form presented here on days where stratifica-
tion persists above the practical threshold from the
previous day. For example, the destratification over
the night of 10 – 11 May 2003 in Mirror Lake (Fig. 6)
was incomplete and N2 was still above the practical
stratification threshold by sunrise on the 11th. These
periods of persistent stratification (incomplete night-
time overturn) can lead to rapid depletions of the
oxygen at the sediment-water interface and subse-
quent increases of reduced compounds into the water
column from the sediments (Branco, 2007). However,
Figure 9. The critical wind speed to prevent stratification as a the stratification parameter can only indicate if the
function of buoyancy flux (B) for water depths of 1 m, 2 m and 3 m.
The critical wind speed was calculated from eq. 18. If wind speed is
waterbody would likely have stratified from a well-
below the critical range in an unstratified waterbody, the water- mixed condition, and can not predict persistent
body will tend to stratify. If wind speed is above the critical range, stratification. Destratification is a separate problem
the waterbody is likely to remain unstratified. and the daily sequencing of stratification/destratifica-
tion for shallow waterbodies needs to be evaluated
with daily information.
diurnal pattern (low during the night, higher during
the day) and generally ranges from 0 to 5 m s-1 (Fig. 3). Conclusions
Since the mean depths of Mirror Lake and Swan Lake The stratification parameter, IS, has utility for predict-
are both order of 1 m, the critical wind speed falls ing the onset of thermal stratification in the two
within this typical range throughout the day, and the shallow ponds of this work, with a critical value of
timing of the wind speed increases relative to the Icrit  1. The critical value for the onset of stratification
increase in SWR (and thus B) is critical in determining is likely to be order 1 over a range of sizes and shallow
the onset of stratification. Mean daily wind speed data depths (< 3 m) because IS is based on a general energy
is inadequate for characterizing the propensity of a balance over the full depth of the water column.
shallow waterbody to stratify (Condie and Webster, Recognition of the controls exerted by physical
2001). dynamics in shallow lakes and ponds is critical when
evaluating their biogeochemical dynamics. Because of
Macrophytes and algae mats the excessively large numbers of shallow lakes and
The buoyancy flux expression presented in this work ponds, individual sampling to quantify cumulative
here does not account for the presence of macrophytes biogeochemical function in the landscape is not
or macroalgae mats that dissipate some of the mixing possible. Tools such as the stratification parameter
energy supplied by the wind (Herb and Stefan, 2004; then become necessary for identifying likely modes of
Herb and Stefan, 2005) and also shade the water system dynamics and provide a means to parameterize
column from SWR (Westlake, 1964). Both the energy these shallow systems in, e.g., regional and watershed-
dissipation and the buoyancy flux associated with the scale models. This requires more work and modeling
shading can be parameterized provided the plant type, to link physical processes to biogeochemical process-
density, and canopy height is known (Herb and Stefan, es.
2005). Floating macrophytes or macroalgae will also
impact the buoyancy through modification of the heat
fluxes and wind stress at the air-water interface (Dale
and Gillepsie, 1976; Chimney et al., 2006). Though the
impacts of macrophytes on diurnal stratification have
been recognized for at least 30 years (e.g. Dale and
Gillepsie, 1976), more work is needed to quantify and
Aquat. Sci. Vol. 71, 2009 Research Article 77

Appendix I: Heat flux formulations for the removal of latent heat by the evaporation of
water is:
Long-wave radiation
The long-wave radiation from the atmosphere to the Qevap ¼ 1w Lv E (I-7)
water surface is a function of the air temperature
(Henderson-Sellers, 1986): where Lv is the latent heat of vaporization for water [J
kg-1] and E is the evaporation rate [m s-1]. The
Qal ¼ ð1  AL Þea s TaK
4
(I-1) evaporation rate is commonly expressed as a function
of the forced convection provided by wind. However
where AL is the reflectivity (0.03), ea is the atmospheric for sites like Mirror Lake, evaporation can still be
emissivity, s is the Stefan-Boltzmann constant: significant under light or even negligible winds when
(5.669710-8 W m-2 K-4) and TaK is the air temperature the boundary layer near the air-water interface is
[K]. unstable and free convection is present (Ryan et al.,
Atmospheric emissivity can be expressed as a 1974; Sill, 1983). The latent heat flux formula devel-
function of water vapor pressure in air and the fraction oped by Ryan et al. (1974) and improved by Adams
of cloud cover. Henderson-Sellers (1986) recom- et al. (1990) includes both wind-forced convection and
mends the equations presented by Raphael (1962). free convection under an unstable atmosphere. The
For a cloud cover fraction (C) between 0.0 and 0.6 the expression is (Adams et al., 1990):
equation is:
h 1
2  2 i1=2
  Qe ¼ lðTwv  Tav Þ =3 þ 5:1  102 A0:05 U2
ea ¼ 0:84  0:1  9:973  106 ea ð1  CÞ
(I-2)  ðesw  ea Þ (I-8)
þ 3:491  105 ea
where l is a constant equal to 2.710- W m-2 Pa-1
and for C > 0.6: (8C)-1/3, Twv and Tav are the virtual temperature of the
surface water and air, respectively, at 2 m, A is the
  surface area of the waterbody [ha], U2 is the wind
ea ¼ 0:87  0:175  29:92  106 ea ð1  CÞ
(I-3) speed at 2 m height, and esw is the saturated vapor
þ 2:693  105 ea pressure at the air water interface [Pa]. The virtual
water or air temperature (Tv) is given by:
where ea is the vapor pressure at air temperature [Pa].
Since C is a difficult parameter to measure and is not TK
Tv ¼  (I-9)
routinely available, C = 0.5 is assumed to avoid 1
0:378e
P
systematic bias and apply eq. I-2:

ea ¼ 0:79 þ 3:990  105 ea (I-4) where P is the atmospheric pressure [Pa] and TK is
either water temperature or air temperature [K]. For
The water surface is treated as a black body emitter: Tav, e is equal to ea and for Tsv, e is equal to esw. The
constant l contains the kinematic mass transfer
Qwl ¼ ew s TwK
4
(I-5) coefficient obtained through scaling arguments, latent
heat of vaporization, and the conversion from vapor
where TwK is the surface water temperature [K] and ew density to vapor pressure using the ideal gas law. See
is the water emissivity ( 0.972) (Henderson-Sellers, Ryan et al. (1974) for details.
1986). Water temperature is computed by the model. For the saturated vapor pressure, Henderson-
The net long wave radiation flux across the air-water Sellers (1986) recommends:
interface is:

a w es ðTÞ ¼ 2:718  1010  exp 4157=ðTK  33:91Þ (I-10)
Ql ¼ Q þ Ql
l (I-6)

Evaporative (latent) heat flux The vapor pressure in air can be calculated by
The evaporative heat flux includes both the latent heat multiplying the saturated vapor pressure at air tem-
loss from the water and the heat carried away by water perature by the relative humidity:
vapor itself, though the latter can be neglected
(Henderson-Sellers, 1986). A common expression ea ¼ RH  esa (I-11)
78 B. F. Branco and T. Torgersen The onset of stratification

Sensible heat flux in shallow aquatic systems: observations, theory and models.
The University of Connecticut, 223 pp.
The sensible heat flux is driven by temperature Branco, B., T. Torgersen, J. R. Bean, G. Grenier and D. Arbige,
difference between the air and water. Empirical 2005. A new water column profiler for shallow aquatic systems.
studies have shown that the sensible heat flux is a Limnol. Oceanogr. Meth. 3: 190 – 202.
Chimney, M. J., L. Wenkert and K. C. Pietro, 2006. Patterns of
function of wind speed. The expression cited in vertical stratification in a subtropical constructed wetland in
Culberson and Piedrahita (1996) was used: south Florida (USA). Ecological Engineering 27: 322 – 330.
Cole, J. J., Y. T. Prairie, W. H. Caraco, W. H. McDowell, L. J.
Tranvik, R. G. Striegl, C. M. Duarte, P. Kortelainen, J. A.
Qs ¼ 1:5701U2 ðTw  Ta Þ (I-12) Dowining, J. J. Middleburg and J. Melack, 2007. Plumbing the
global carbon cycle: integrating inland waters into the terres-
Shortwave radiation trial carbon budget. Ecosystems 10: 172 – 185.
Condie, S. A. and I. T. Webster, 2001. Estimating stratification in
Some of the solar short-wave radiation incident at the
shallow water bodies from mean meteorological conditions. J.
air-water interface is reflected back to the atmos- Hydraul. Eng. 127: 286 – 292.
phere. Thus the solar short-wave radiation (SWR) that Culberson, S. D. and R. H. Piedrahita, 1996. Aquaculture pond
penetrates the air-water interface is: ecosystem model: Temperature and dissolved oxygen predic-
tion – mechanism and application. Ecological Modelling 89:
231 – 258.
Qoswr ¼ QIswr ð1  AS Þ (I-13) Dale, H. M. and T. Gillepsie, 1976. The influence of floating
vascular plants on the diurnal fluctuations of temperature near
the water surface in early spring. Hydrobiologia 49: 245 – 256.
where QIswr is the incident shortwave radiation AS is Downing, J. A., Y. T. Prairie, J. J. Cole, C. M. Duarte, L. J. Tranvik,
the surface albedo. The surface albedo varies widely R. G. Striegl, W. H. McDowell, P. Kortelainen, N. F. Caraco, J.
with water conditions, the zenith angle of the sun and M. Melack and J. J. Middelburg, 2006. The global abundance
and size distribution of lakes, ponds, and impoundments.
cloud cover. A simple empirical relationship from Limnol. Oceanogr. 51: 2388 – 2397.
Henderson-Sellers (1986) was chosen based on the Fee, E. J., R. E. Hecky, S. E. M. Kasian and D. R. Cruikshank, 1996.
assumption that cloud cover fraction was in the range Effects of lake size, water clarity, and climatic variability on
mixing depths in Canadian Shield lakes. Limnol. Oceanogr. 41:
of 0.05 to 0.5:
912 – 920.
Ford, P. W., P. I. Boon and K. Lee, 2002. Methane and oxygen
AS ¼ 0:0434q0:95 (I-14) dynamics in a shallow floodplain lake: the significance of
periodic stratification. Hydrobiologia 485: 97 – 110.
Ganf, G. G., 1974. Diurnal mixing and the vertical distribution of
where q is the sun zenith angle [radians]. phytoplankton in a shallow equatorial lake (Lake George,
Uganda). The Journal of Ecology 62: 611 – 629.
Gu, R., F. N. Luck and H. G. Stefan, 1996. Water quality
stratification in shallow wastewater stabilization ponds.
Acknowledgments Water Res. Bull. 32: 831 – 844.
Henderson-Sellers, B., 1984. Engineering Limnology, Pitman
This material is based upon work supported by the Advanced Publishing Program, Boston, 356 pp.
Henderson-Sellers, B., 1986. Calculating the surface energy
National Science Foundation under Grant No. balance for lake and reservoir modeling: A review. Rev.
EAR0002993 and Grant No. EAR0003408. Support Geophys. 24: 625 – 649.
was also provided by the Connecticut Institute of Herb, W. R. and H. G. Stefan, 2004. Temperature stratification and
mixing dynamics in a shallow lake with submersed macro-
Water Resources and the University of Connecticut.
phytes. Lake and Reservoir Management 20: 296 – 308.
John Bean, Peter Rich, Jeffrey Saunders, Suzanne Herb, W. R. and H. G. Stefan, 2005. Dynamics of vertical mixing in
Auer, Heather Pugh, and David Banker contributed a shallow lake with submersed macrophytes. Water Resour.
to the field work. We thank Steve Olsen for providing Res. 41: W02023.
Holloway, P. E., 1980. A criterion for thermal stratification in a
access to the weather data from the University of wind-mixed system. J. Phys. Oceanogr. 10: 861 – 869.
Connecticuts Plant Science Farm. We also thank Imberger, J., 1985. The diurnal mixed layer. Limnol. Oceanogr. 30:
Alfred West and two anonymous reviewers for their 737 – 770.
Imberger, J., 2001. Lakes. In: A. H. E.-S. a. W. W. Piegorsch, The
thoughtful comments, critiques and suggestions, Encyclopedia on Environmetrics, John Wiley & Sons Ltd,
which have improved the manuscript significantly. pp. 1121 – 1130.
Jassby, A. and T. Powell, 1975. Vertical patterns of eddy diffusion
during stratification in Castle Lake, California. Limnol.
Oceanogr. 20: 530 – 543.
References Jellison, R. and J. M. Melack, 1993. Meromixis in hypersaline Mono
Lake, California .1. Stratification and vertical mixing during
Adams, E. E., D. J. Cosler and K. R. Helfrich, 1990. Evaporation the onset, persistence, and breakdown of meromixis. Limnol.
from heated water bodies: predicting combined forced plus Oceanogr. 38: 1008 – 1019.
free convection. Water Resour. Res. 26: 425 – 435. Kim, J.-W., 1976. A generalized bulk model of the oceanic mixed
Bean, J. R., 2005. Effective vertical diffusivity in shallow lakes by layer. J. Phys. Oceanogr. 6: 686 – 695.
an enhanced flux-gradient method: the dependence of diffu- Kraus, E. B. and J. S. Turner, 1967. A one-dimensional model of the
sivity on density gradient and an example application to seasonal thermocline. II. The general theory and its conse-
ecosystem stability. University of Connecticut, 143 pp. quences. Tellus 19: 98 – 105.
Branco, B., 2007. Coupled physical and biogeochemical dynamics MacIntyre, S., 1993. Vertical mixing in a shallow, eutrophic lake:
Aquat. Sci. Vol. 71, 2009 Research Article 79

possible consequences for the light climate of phytoplankton. Scheffer, M., S. R. Carpenter, J. A. Foley, C. Folke and B. Walker,
Limnol. Oceanogr. 38: 798 – 817. 2001. Catastrophic shifts in ecosystems. Nature 413: 591 – 596.
MacIntyre, S., J. R. Romero and G. W. Kling, 2002. Spatial- Sill, B. L., 1983. Free and forced convection effects on evaporation.
temporal variability in surface layer deepening and lateral J. Hydraul. Eng. 109: 1216 – 1231.
advection in an embayment of Lake Victoria, East Africa. Simpson, J. H. and J. R. Hunter, 1974. Fronts in the Irish Sea.
Limnol. Oceanogr. 47: 656 – 671. Nature 250: 404 – 406.
Maritorena, S., A. Morel and B. Gentili, 1994. Diffuse reflectance Simpson, J. H. and D. Bowers, 1981. Models of stratification and
of oceanic shallow waters: influence of water depth and bottom frontal movement in the shelf seas. Deep-Sea Research 28:
albedo. Limnol. Oceanogr. 39: 1698 – 1703. 727 – 738.
Mobley, C. D., 1994. Light and Water : Radiative Transfer in Simpson, J. H., C. M. Allen and N. C. G. Morris, 1978. Fronts on the
Natural Waters Academic Press, San Diego, 592 pp. continental shelf. J. of Geophys. Res. 83: 4607 – 4614.
Morris, D. P., H. Zagarese, C. E. Williamson, E. G. Balseiro, B. R. Smith, N. P., 2002. Observations and simulations of water-sediment
Hargreaves, B. Modenutti, R. Moeller and C. Queimalinos, heat exchange in a shallow coastal lagoon. Estuaries 25: 483 –
1995. The attentuation of solar UV radiation in lakes and the 487.
role of dissolved organic carbon. Limnol. Oceanogr. 40: 1381 – Smith, S. V., W. H. Renwick, J. D. Bartley and R. W. Buddemeier,
1391. 2002. Distribution and significance of small, artificial water
Ohlmann, J. C., D. A. Siegel and C. D. Mobley, 2000. Ocean radiant bodies across the United States landscape. Sci. Total Envi-
heating. Part I: optical influences. J. Phys. Oceanogr. 30: 1833 – ron. 299: 21 – 36.
1848. Sturman, J. J., C. E. Oldham and G. N. Ivey, 1999. Steady convective
Panofsky, H. A. and J. A. Dutton, 1984. Atmospheric Turbulence : exchange flows down slopes. Aquat. Sci. 61: 260 – 278.
Models and Methods for Engineering Applications, Wiley, Torgersen, T. and B. F. Branco, 2008. Carbon and oxygen fluxes
New York, 397 pp. from a small pond: temporal variability and the CO2/O2
Raphael, J. M., 1962. Prediction of Temperature in Rivers and imbalance. Water Resour. Res., 44, W02417, doi:10.1029/
Reservoirs. J. Power Div. Am. Soc. Civ. Eng. 88: 157 – 181. 2006WR005634.
Rubbert, S. and J. Kçngeter, 2005. Measurement and three- Westlake, D. F., 1964. Light extinction, standing crop and photo-
dimensional simulation of flow in a shallow reservoir subject synthesis within weed beds. 15: 415 – 425.
to small-scale wind field inhomogeneities induced by shelter- Wetzel, R. G., 2001. Limnology: Lake and River Ecosystems,
ing. Aquat. Sci. 67: 104 – 121. Academic Press, Boston, 1006 pp.
Ryan, P. J., D. R. F. Harleman and K. D. Stolzenbach, 1974. Surface West, A. and A. Lorke, 2003. Small-scale hydrodynamics in lakes.
heat loss from cooling ponds. Water Resour. Res. 10: 930 – 938. Annual Review of Fluid Mechanics 35: 373 – 412.
Scheffer, M., S. H. Hosper, M.-L. Meijer, B. Moss and E. Jeppesen,
1993. Alternative equilibria in shallow lakes. Trends Ecol.
Evol. 8: 275 – 279.

To access this journal online:


http://www.birkhauser.ch/AS

You might also like