Vous êtes sur la page 1sur 21

Aquacultural Engineering 6 (1987) 237-257

Modeling Temperature Distribution in Freshwater Ponds*


T h o m a s P. Cathcart and Frederick W. W h e a t o n
Department of Agricultural Engineering, University of Maryland, College Park, MD 20742, USA

ABSTRACT A model for predicting the vertical temperature profile of turbid freshwater ponds is described. High resolution in space and time and a reduction in the number of meteorological inputs usually required by such models is achieved by using pond surface temperature as a model input. A method to estimate hourly surface temperature is proposed and demonstrated.

NOMENCLATURE A B
el

E F(t) H
m l?

Q|OSS O,

O m
r

S(z)

Surface area (m 2) Fraction of the solar energy spectrum having wavelengths greater than 700 nm. Slope of regression line Specific heat of water (kJ (kgC)- 1) Molecular diffusivity of water (m ~-h- ~) Surface temperature interpolation function Heat content of pond layer (J) Number of model elements Extinction coefficient (m- ~) Surface heat loss (J) Total heat flux across the boundaries of element 1 (J) Heat flux across the lower boundary of element 1 (J) Fraction of the incident solar radiation reflected by the water surface Insolation at depth z below the surface (J m-2)

*Scientific Article Number A-4460 Contribution Number 7452 of the Maryland Agricultural Experiment Station (Department of Agricultural Engineering). 237 Aquacultural Engineering 0144-8609/87/S03.50-- Elsevier Applied Science Publishers Ltd, England, 1987. Printed in Great Britain

238

T. P. Cathcart, F. W. Wheaton

So
T

tL tH

V(z) x;
Z

zl /9

Insolation at the water surface (J m-2) Temperature (C) Daily low temperature (C) Daily high temperature (C) Time (h) Time of day of TL (h) Time of day of TH(h) Volume of pond elements (m3) Heat added to element i due to wind mixing (J) Depth, increasing downward (m) Depth of the bottom of the surface element (m) Water density (kg m-3) INTRODUCTION

There is currently a large and growing body of literature associated with attempts to predict temperature distribution in aquatic systems. In general, these efforts have been focused on salt-stratified (non-convecting) solar ponds, cooling ponds used to dissipate waste heat, and large water bodies, such as lakes, reservoirs, and locally defined portions of oceans (Jirka et al., 1978; Delnore, 1980; Harleman, 1982; Meyer, 1983; Sturm et al., 1983; Patterson et al., 1984). The prediction of water column temperature is a matter of some importance to aquaculturists. For cold blooded organisms, temperature directly affects growth rate, the onset of reproduction, and the occurrence of thermal stress. It also may influence resistance to disease and behavior. Temperature gradients influence pond mixing processes through the concomitant creation of a vertical density gradient. The density gradient tends to inhibit vertical mixing and so may be an important determinant of dissolved oxygen and waste metabolite distribution in the pond environment. Because most aquacultural ponds are highly turbid, temperature gradients (as great as 4C m-1) frequently occur during the growing season. A reliable temperature distribution model would enhance the capacity of aquaculturists to monitor pond conditions, predict periods of sub-optimal growth, and make intelligent management decisions. There is, at present, no temperature distribution model specifically designed for the types of ponds frequently encountered in aquaculture. The applicability of existing models to the aquacultural problem is somewhat limited. Salt-stratified solar ponds are characterized by very low turbidity and an unusual (i.e. 'reversed') temperature gradient. Models of

Modeling temperature in fresh water ponds

2 39

large water bodies are frequently limited by low resolutions in space and time (on the order of meters and days). For aquaculturists, resolutions in hours and fractions of meters are desirable. One possible strategy is to adapt an appropriate model to the special conditions of aquacultural ponds, but here too there are problems. Most temperature distribution models require a number of meteorological inputs to estimate the heat flux across the air-water interface and the shear stress placed on the water surface by the wind (Jirka et al., 1978; Sturm et al., 1983). Some of the inputs, such as incident solar radiation, tend to remain relatively constant over substantial geographic areas, and so may be derived from nearby weather stations (Jirka et al., 1978). Others, however, such as wind velocity and direction, are subject to considerable local variation. If use of empirical constants is to be avoided, measurements must be taken very near to the site of interest (Bloss and Harleman, 1979; Sturm et al., 1983). Acquisition of weather monitoring equipment (and the requirements of maintenance) is probably not a practical option for most aquacultural operations due to cost. For a model to be useful in the private sector, some (perhaps substantial) simplification is required over most conventional temperature distribution models; and this simplification should be accomplished without a corresponding decrease in accuracy! The strategy adopted in this paper is relatively simple: pond surface temperature is assumed known and is used as a model input. This approach obviates the need to estimate surface heat flux and provides an alternate method for computing the effect of wind on temperature distribution. Clearly, this has changed the nature of the model. The problem now is to estimate not the temperature gradient of the entire water column, but rather the gradient beneath a surface element of some prescribed depth. The rationale of this approach is that, unlike many other types of water bodies, aquacultural ponds receive frequent attention during the growing season. Use of this model requires that some method be utilized to provide hourly estimates of surface water temperature. Following a description of the model, a method to provide reasonable hourly estimates is proposed and demonstrated. MODEL DEVELOPMENT The temperature distribution model presented here consists of three parts. The first part accounts for molecular diffusion and internal absorption of solar radiation. The second portion identifies and removes buoyant instability from the water column and so approximates thermal

240

T. P. Cathcart, F. W. Wheaton

convection. The third part estimates the quantity and distribution of heat mixed down into the water column by wind action. The first two parts correspond rather closely to other models and are summarized only briefly. The approach used to estimate the quantity and distribution of wind mixed heat is original to this model and is treated with more depth.

Molecular diffusion and solar absorption


Following Jirka eta/. (1978), the one-dimensional heat equation (eqn ( 1 )) with a heat source is used to estimate the molecular diffusion of heat and the variable absorption of solar radiation with depth.

a T= E V(z---)) a(v(z) a_s) A Cp 1 p ffz (S (z )A ) c3t Oz

(1)

The function S(z) corresponds to Beer's Law for the internal absorption of solar radiation: S(z) = (1 - r)(1 -

B)Soe-nzl

(2)

The term 1 - B represents the portion of the solar energy spectrum between 300 and 700 ran. The value of 1 - B appears to vary between 0.4 and 0.45. Jirka et aL (1978) assume a value for 1 - B of 0.45 for their model. The same assumption is made in the present model. The use of element volumes (V(z)) is required to conserve energy in the heat equation (1). Element volumes are recalculated for every change in pond elevation. Use of the model presupposes that an accurate bottom contour map of the pond is available. An explicit finite difference numerical method is used to solve eqn (1) for each time step. The locations of element nodes are redetermined for each change in pond elevation. The surface boundary condition is the known temperature of the surface element. It is assumed that there is no heat flux through the bottom or sides of the pond and that the pond temperature structure is horizontally homogeneous.

Thermal convection
When the surface heat loss exceeds the amount of radiation entering the pond from above, the water column becomes buoyantly unstable (i.e. cooler water overlays warmer). Adjacent elements that are buoyantly unstable are located and mixed in the present model. As in the case of the heat equation, knowledge of element volumes is required for energy to be conserved.

Modeling temperature in freshwater ponds

241

Wind-mixed heat

In order to estimate the amount of energy mixed down into the water column by the wind, the model first calculates an upper limit, or reservoir, of energy which cannot otherwise be accounted for in the pond heat balance (eqn (3)):
O= SoA( 1 - r ) - AH~ - A H 2 - Q~oss

(3)

where A H~ refers to the heat change in the surface element during a given time step, A H2 is the heat change in the remainder of the water column, and Q~os,is the heat flux across the air-water interface. The heat change in the surface element may be written:
AHI=SoA(1-r)(1-B)(a-e-"~)+SoA(1-r)B-Q]

(4)

to account for the wavelength-dependent nature of energy absorption in the pond. The first term on the right hand side (RHS) of eqn (4) represents solar energy absorption of wavelengths in the range 300-700 nm (Rabl and Nielson, 1975). Absorption in this region may be approximated using a Beer's Law function. The second term on the RHS of eqn (4) represents absorption at wavelengths greater than 700 nm, which occurs within a few cm of the surface in turbid waters. The third term, Q,, represents transport out of the surface layer, either across the air-water interface or down into the water column. The heat change in the remainder of the water column (AH2) may be written:
A H 2 = SoA( 1 - r)(1 - B)e'"ZJ + Qm

(5)

The first term on the RHS represents absorption of solar radiation in the range of 300-700 mn. The second term (Qm) accounts for additional transport of heat from the surface layer. Rewriting eqn (3) with eqns (4) and (5) and clearing terms gives: Q] = Qm + Qio~s Heat transport out of element 1 (Q~) may be divided into two parts:
AH] = ( S o A ( 1 - r)(a - B ) ( 1 - e - " ~ ) - Q11)+(SoA(1- r)B-Q12)

(6)

(7)

where Q~--QI~ + Q12. We then assume that all long-wave (> 700 nm) solar radiation, having been absorbed close to the surface, is almost immediately returned to the atmosphere through surface losses:
QI2= S o A ( 1 - r)B

(8)

242

T. P. Cathcart, F. W. Wheaton

so that A H 1= SoA(1 - r)(1 - B)(1 -e-nZl) - Qll (9)

Approximating observed heat changes in the surface element with a linear function of SoA yields:
Qn=SoA(1-r)(1-B)(1-e-nZl)-CISoA(1-r)

(I0)

where C1 is the slope of the regression line. The quantity of heat energy, which is equal to
Qll = Qm +

{~oss-Q12

(11)

provides the upper limit on the amount of energy which may be mixed down into the water column. A number of separate processes, caused ultimately by the effects of the wind on the water column, may result in the downward transport of heat. These include the creation of currents which lead to turbulence, turbulence caused by the breaking or billowing of internal waves, seiching, and boundary turbulence, to name a few (Fischer et aL, 1979). Clearly, the topic is complex. Many questions remain about the mechanisms of mixing, and their relative importance, in aquatic bodies. It is not within the scope of the present model to predict the mechanisms which may be active in an aquacultural pond. Rather, the attempt has been made to use the single piece of information available to us, the density gradient, to estimate how the reservoir of 'wind-mixed heat' may be distributed in the water column. The approach used to distribute the wind-mixed heat is summarized by the set of m - 1 simultaneous equations: For i--- 1 to m - 1,

Ex, Ev,
J=i+ l J~i+ l m

e -p/z

(12)

E
k=i

E
k=i

where m is the number of model elements and X is the amount of heat added to each element. Using this approach, the absence of a density gradient (Op/Oz = 0) results in uniform distribution of wind-mixed heat on a per unit volume basis. Use of an exponential distribution of windmixed heat in the presence of a density gradient reflects the cost in kinetic energy of transporting heat along a density gradient. The specific use of an exponential distribution is purely a mathematical convenience

Modeling temperature in freshwaterponds

24 3

and is a first approximation without implying any special insight into the physical problem. Equation (12) results in m - 1 simultaneous equations. Since there are m unknowns (X~,..., Xm), an mth equation is required. This is provided by the summation: ZX~= Qm+ Q~oss- Q12 i=~ The equations are then solved using Gaussian elimination. (13)

Estimation of hourly surface temperature


The model, as here presented, requires that hourly surface water temperatures be available as model inputs. It is probably not reasonable to assume that single channel temperature recorders will be used to provide actual measurements. It is necessary, then, to substitute a method for estimating hourly temperatures on the basis of some conveniently measured physical parameter. A possible solution to this problem may be the use of daily maximum-minimum thermometers. Such a thermometer, if held slightly below the surface (on the order of 10-15 cm) and read on a daily basis, will provide values for daily high and low surface water temperatures. If these values could be reliably associated with hours of the day (i.e. average hours of low and high temperature) then a basis would exist for interpolating surface temperature for the remaining 22 hours. The present model estimates hourly temperatures between daily high and low water temperatures using a function F(t) having the derivative: dF dt
K(t--tL)(t--tH)

(14)

where K is an equation constant, t represents the time of day (h), and tL and tH are the times of day when the temperature is at its daily minimum or maximum value, respectively. Then, using the conditions:

F(/L)-- rL (15) F(t.) = r .


Eqn (14) can be integrated and the values of K and the integration constant ascertained. This yields a cubic function having first derivatives of

244
TH

T. P. Cathcart, F. W. Wheaton

Temperature

TL

. . . .

',

tm

tH

Fig. 1.

Demonstration interpolation between hypothetical daily high and low temperatures.

zero at t L and tH, a minimum of TL at tL, and a maximum of THat tH. The results of a hypothetical interpolation are illustrated in Fig. 1. MODEL VALIDATION Initial validation of the pond temperature model was performed using temperature data from a pond located on the University of Maryland's Wye Research and Education Center. Temperatures were recorded continuously from thermocouples on five stakes in the pond. Pond contours and the locations of the stakes are illustrated in Fig. 2. The pond had a surface area of 0.17 ha and a maximum depth of 2.2 m. The central stake and the two stakes located along the short axis of the pond (Fig. 2) had thermocouples mounted vertically at 30-cm intervals. On the two remaining stakes, the thermocouples were separated by 60 cm. On all stakes, the top thermocouple was 15 cm below the surface at the commencement of measurements. The temperature recording system was in operation from 15 August to 31 October. Interruptions to the record, as a result of power failures and mechanical malfunctions, resulted in the loss of approximately 20% of the data and necessitated a repetition of the procedure. This repetition is currently underway. For present purposes, gaps in the record of surface water temperature are filled using cubic spline interpolation or the maximum-minimum procedure outlined above.

Modeling temperature in freshwater ponds

245

(a)

(b) Fig. 2. Schematicrepresentation of the pond (L-- 78 m, W-- 22 m) used to validate the pond temperature model. (a): Contour map of pond bottom. The outer line is at depth 0, the second line is at depth 0.15 m. Remaining lines correspond to depth increments of 0.3 m. (b): Crosses mark the approximate locations of stakes used to support thermocouples. The penetration of light into the water column was measured using a Lycor LI-188B light meter with a submersible sensor. Measurements were made approximately twice a week and were used to calculate extinction coefficients. Obse.ryations of pond elevation were made, approximately twice a week, using a depth marker mounted near the edge of the pond. Extinction coefficients and pond elevations were estimated by linear interpolation on days when measurements were not taken. Hourly values of incident solar radiation were taken at a site approximately 65 km from the Wye Institute.

One dimensionality
Use of eqn ( 1 ) in the present model requires that the temperature distribution in the pond be horizontally homogeneous. There is some agreement among modelers (e.g. Jirka et al., 1978; Patterson et al., 1984) that the volume flow rate into and out of the pond may be used to assess the

246

T. P. Cathcart, F. IV. Wheaton

one-dimensional assumption. A high flow rate, relative to pond volume, may result in mixing and/or the creation of currents which could lead to horizontal temperature variations. Except during periods of rain, this was probably not a factor in the pond used for model validation, Inflow to the pond was, in general, less than the evaporation rate as evidenced by the slow but steady decrease in pond elevation between periods of precipitation. During and immediately after storms, flows into and out of the pond were observed to dramatically increase. There was, however, no way to quantitatively measure these flows and the degree and duration of perturbation could not be calculated. A qualitative examination of the pond temperature record suggests that the one-dimensional assumption was valid most of the time. During dry periods, horizontal variations appear to be less than 0.5C. During storm activity, variations are, for the most part, less than 0.5C although occasional variations of IC and more were observed. Gaps in the temperature record, which frequently occurred during periods of stormy weather, limit the completeness of such observations. A statistical comparison of horizontal temperatures has not, as yet, been undertaken.

Surface losses
The pond temperature model assumes that all of the absorbed solar radiation having a wavelength greater than 700 nm is almost immediately lost to the atmosphere through the air-water interface. Figure 3 represents daily absorption of solar radiation in the pond as a fraction of the total incident radiation on the pond surface. The quantity of radiation absorbed is based on the observed temperature changes in the pond during periods of insolation. The mean fraction of solar radiation that was absorbed was 0.29 with most of the values less than 0.40. The fraction of the solar energy spectrum having wavelengths greater than 700 nm is, in general, 0.55-0.60, suggesting that surface losses during insolation are indeed greater in magnitude than the energy contained in the long-wave portion of the solar spectrum.

Temperature prediction
During the period when temperatures were recorded (15 August to 31 October) the water column experienced a large seasonal decrease in temperature. Figure 4 represents observed and predicted temperatures plotted against time for this period. For illustrative purposes, the plot is limited to temperatures at 4 p.m. from the fourth element (approximately 1 m in depth). A rigorous evaluation of the model requires that the seasonal effect be limited in so far as it is possible to do so. This was

Modefing temperature in freshwater ponds


~.5

247

~.0

OH /

SR

0.5
-

. . . . . oo . . . . . -. _- . . . .

.e .

_. e.
a

0.40 O. 29

ol~l~

%oo o+
oo

oooo ~
c~
I

0.0

220

240

260
Time (Julian Days)

280

300

320

Fig. 3.

Change of pond heat energy during insolation as a fraction of incident solar energy.
30

25

Temperature
(C)

20

t5

10

'

'

'

5001

lO001

15001

20100

Time

(hours)

Fig. 4.

Observed and predicted temperatures showing seasonal cooling. Temperatures are for 4 p.m. at a depth of approximately 1 m.

accomplished by fitting a linear function to the observed temperatures from each of the 6 subsurface elements. The regression was then subtracted from both observed and predicted temperatures. Figure 5 illustrates the data of Fig. 4 after seasonal adjustment. Because of the large

248
6

T. P. Cathcart, F. W. Wheaton

LEGEND : OBSERVED - x PREDICTED

T (C)

xx l

-2

-4

i ~000[

~5[00

500

' 2000
Time (Mours)

Fig. 5.

Observed and predicted seasonally adjusted temperatures (T*). The data in this graph corresponds to that of Fig. 4.

eee

~..

"

PreOLcteO Tw (C] 0
"

0 - "

"

LEGEND

. ~

- -

REG~EBBION

LINE I

,IT LINE ]
@ -~
-8
= I l I i I , I 2 T'(C} , I i I

-4

O
ODserved

Fig. 6.

Predicted vs observed seasonally adjusted temperatures (T*) for element 2 (depth approximately 0.45 m) at 4 p.m. (r= 0.97).

number of data points, it is not possible to represent predicted versus observed temperatures for all hours and elements. Figures 6-11 offer a sample of typical curves. Correlation coefficients shown on the figures are those for all of the data points for that element.

Modeling temperature in fresh water ponds

249

L E G E N D

o,77-REGRESSIONLINE J IDEAL FIT LINE J

~
0

//~$
"
. "

Predicted

.~0

TK (C)
-2

-4

E -6

i -

14 -

12

l
0 Observed

I
2

t
4

I 6

T~ (C)

Fig. 7.

Predicted vs observed seasonally adjusted temperatures (T*) for element 3 (depth approximately 0.75 m) at 4 p.m. (r= 0-97).
6

Pre0lcted

24

0@

" @""0 @e @~. .@@

[C)

o ~.2=

LEGE.O

I
-dJ -4
Fig. 8.

.~

J
, I 0

I-: I- i 2

REG.EGGION L:NEI

I -2

1 6

Predicted vs observed seasonally adjusted temperatures (T*) for element 4 (depth approximately 1.05 m) at 4 p.m. (r-- 0-96).

The model exhibited a systematic tendency to underpredict pond temperatures in most elements. This pattern is especially prominent in Figs 7 - 9 and is formalized in the comparison of the mean observed and predicted temperatures for each element. Variances for each element were

250

T. P. Cathcart, F. W. Wheaton

LEGEND e - DATA REGRESSIONLINE| IDEAL FIT LINE [

/-

/ ~cj .
~0" (~ O 00 ~ - O ~ O" ; 0 O

//

/
0

Predicted
T" 0
Q

(c}

-4 -4

, m

12

~ 0

Observed T"

(el

Fig. 9.

Predicted vs observed seasonally adjusted temperatures (T*) for element 5 (depth approximately 1.35 m) at 4 p.m. (r-- 0-95).

LEGEND
/

/
/

//
.

DATA
- REGRESSION LINE ~

~"

IDEAL FIT LINE

Predicted T~

y:o-

~/~

."

(el

-2

- 4 4 i J * , _l3 . . . .

_12 . . . .

i ....
-i

i .... 0 Observed T

i .... ~ (C)

i .... 2

i ....
3

i 4

Fig. 10.

Predicted vs observed seasonally adjusted temperatures (T*) for element 6 (depth approximately 1.65 m) at 4 p.m. (r= 0"93).

not significantly different ( P > 0.05). Predicted and observed means of elements 3, 4, 5, and 7 were significantly different from one another

(P < 0-05).
M e a n temperatures for each element are compared in Fig. 12. Despite

Modeling temperature in freshwater ponds


41
LEGEND

2 51

]
e

@
2 - --

DATA
REGRESSIONLINE] IDEAL FIT LINE[_j ~ "

Predicted

T* (C)

(~ 0@@ O f " ~ @O . " @ . "@ @ @ @ -"


@

-2

.e"

-4 -3

. . . .

I -2

. . . .

. . . .

. . . .

. . . .

I 2

. . . .

I 3

-~

0
Observed T *

i (C)

Fig. 11.

Predicted vs observed seasonally adjusted temperatures (T*) for element 7 (depth approximately 2.00 m) at 4 p.m. (r-- 0"90).
0.20 F

0.

~.0~

I:"
-

LEGEND
o sE.vEO

PREDICTED

0.00!~,
Mean T

-c.loh

4D

r r -o.2o I
-0,30!
@

-0.40

I 2

Element

Fig. 12.

A comparison of mean and predicted observed seasonally adjusted temperatures for the 6 subsurface elements of the pond.

the systematic underprediction, the magnitude of the difference is 0.5C or less. Figure 13 represents a more qualitative view of the model. Using 1 C as a benchmark, Fig. 13 shows the percentage of predictions which fall

252

T. P. Cathcart, F. W. Wheaton

I
Percent Within I C

I
40 20

0
Hour: I 2 24 1 3 24 ~ 4 Element 24 I 5 24 1 6 24 ! 7 24

Fig. 13. Percentage of predictions which deviate from observed temperatures by IC or less.
within 'reasonable' limits of the observed temperature for each element at each hour of the day. The model appears to perform worst for elements 2 and 4, with agreement falling below 80% for certain hours of the day. For other hours and elements, predictions are within IC 80-100% of the time.

Wind mixing
The importance of the wind mixing portion of the model is highlighted by Figs 14 and 15. Figure 14 compares the mean predicted temperatures for each element, with and without wind mixing, to the observed mean temperatures. The tendency to underpredict pond temperature is much exacerbated in the absence of the wind mixing procedure. Figure 15 shows that the improved agreement of the predicted temperatures with wind mixing is relatively uniform for all hours of the day.

Estimating surface temperatures


The usefulness of the pond temperature model is, to a degree, predicated on the availability of hourly surface water temperatures. The maximum-minimum procedure outlined in the preceding section was used to estimate hourly temperatures. It was assumed that daily minimum and maximum water temperatures occurred at 8 a.m. and 4 p.m. respectively

Modeling temperature in fresh water ponds


0.5

253

C.O Mean T [C)

W-----

LEGEND

\
-0.5

"\
\
\, "x.

D OBSERVED - PREDICTED - WIND NIXING PREDICTED NO WIND NIXING

Ix

- 1 . 0

4 Element

Fig. 14.

A comparison of mean predicted (seasonally adjusted) temperatures with and without wind mixing.

Percent
E
Withln
! C

Hour:

~ 5

24

% B

24

1 7

24

Element

Fig. 15.

Percentage agreement with observed temperatures: solid b a r s = n o wind mixing; open bars = wind mixed.

(daylight savings time). The estimated temperatures were then adjusted to eliminate the seasonal bias and compared to observed surface temperatures. The means and variances of the two data sets were not significantly different ( P > 0 ' 0 5 ) . The estimated mean temperature was

254

T. P. Cathcart, F. W. Wheaton

approximately 0.17C greater than the observed mean. The correlation coefficient was 0.99. Predictions based on estimated surface temperatures are compared to observed temperatures in Figs 16 and 17. Figure 16 illustrates the rela-

Mean T CC)

i1
-0. ~ \

X---(:ffiI(RVEDI:~EO iCTrn C-ENO____ I.E- ACTU~. NTERPOLA TEO S.T. - -X~' ~ CT rn S.T.
m $
P

'

'\ \ '\
\
"e

i"

)<

-0,4

-0.6

'

"~t~'

4
Element

Fig. 16.

A comparison of mean predicted temperatures (seasonally adjusted) using actual and interpolated surface temperatures.
100

80

Percent

Wlt~Ln toc

60

4O

20

0
Hour:

t 3

24
Element

I 8

24

Fig. 17.

Percentage agreement with observed temperatures: solid bars = interpolated surface temperatures; open bars = actual surface temperatures.

Modeling temperature in freshwaterponds

255

tively small differences observed in the predicted means using observed and estimated surface temperatures. Figure 17 shows the percentage of predictions which agreed with observed temperatures to within IC. Using estimated surface temperatures, the greatest decline in accuracy appears to be within elements 4-6. Overall, agreement to within IC declined by about 7% using estimated surface temperatures.

DISCUSSION Error and sensitivity analyses have not yet been performed. There are, however, a few points that should be made about the pond temperature model. One assumption that is made in this and in most other temperature models is that of a very small (i.e. instantaneous) time scale for the removal of buoyant instability from the water column. It was a matter of some interest, then, when the temperature record revealed that cool surface water frequently overlays warmer water. The difference was small, never more than about 0.5C, but it occurred in approximately 19% of the observations and persisted for up to 11 hours at a time. The fact that the magnitude of the temperature difference was on the order of the resolution of the recorder raises the possibility that the instability was not real. The systematic nature of the observations (instabilities occurred only at surface elements, only during periods of cooling, and were observed at more than one location in the pond), however, suggests that the temperature differences were present. The depths of the unstable layer varied between a minimum of 15 cm to a maximum of perhaps 70 cm. The presence and persistence of a buoyantly unstable surface layer appeared to contribute significantly to the error in the model. When run with a surface temperature record that had been altered such that subsurface temperatures never exceeded that at the top of the water column, almost 30% of the error was removed. Figure 18 illustrates the type of improvement that was encountered. The observation of persistent buoyant instability is probably due to the relatively fine resolution of the model. There does not appear to be a convenient way to remove the error at the present time. As the difficulty occurs most frequently during periods of rapid cooling, the error may be relatively uncommon during summer months. One of the objectives of the work which led to the pond temperature model was to find a way to simplify the environmental variables required to make accurate predictioris about temperature distribution. A method has been suggested here by which it may be possible to avoid the neces-

256
I00

T. P. Cathcart, F. W. Wheaton

BO-

i
Percent B0-

Within ioC
40-

20-

0 12 IB 24

Hour of Oay

Fig. lB. Percentage agreement with observed temperatures (element 4): hatched bars--actual surface temperatures; open bars = surface temperatures altered to eliminate thermal instability.

sity for on-site measurement of wind velocity and direction. Another issue that needs to be addressed, however, is the penetration of shortwave radiation into the water column. Although hourly, or even daily, measurements are not required, weekly or biweekly observations are. Currently, accurate measurements are only possible using a submersible light meter. Secchi disks offer an inexpensive alternative, but have an uncertain accuracy and insert an additional level of inconvenience to implementing the model (Bloss and Harleman, 1979). Research designed to investigate this question is currently at the planning stage at the University of Maryland. As mentioned, data collection is presently underway to provide additional validation of the pond temperature model. A fully operational weather station at the V~ye Institute will offer, not only on-site solar radiation measurements, but the opportunity to compare assumptions about wind mixing and surface heat loss to actual meteorological measurements.

ACKNOWI =EDGEMENTS We gratefully acknowledge Dr Russell Brinsfield and Mr James Wood of the Wye Research and Education Center for financial support, use of

Modeling temperature in fresh waterponds

257

equipment, and much help; Dr Larry Harding of the Chesapeake Bay Institute for the generous use of equipment; Professor Norman Martin, Mr Bob Dixon, and Mr Gary Seibel of the Department of Agricultural Engineering for hours of valuable assistance; and the University of Maryland Department of Computer Science for computer support.

REFERENCES Bloss, S. & Harleman, D. R. E (1979). Effect of wind-mixing on the thermocline formation in lakes and reservoirs. Report No. 249, R. M. Parsons Laboratory, MIT, Cambridge, Mass. Delnore, V. E. (1980). Numerical simulation of thermohaline convection in the upper ocean. J. Fluid Mech., 96 (4), 803-26. Fischer, H. B., List, E. J., Koh, R. C., Imberger, J. & Brooks, N. H. (1979). Mixing in Inland and Coastal Waters, Academic Press, New York, 483 pp. Harleman, D. R. F. (1982). Hydrothermal analysis of lakes and reservoirs. J. Hydraulics Division, Proc. ASCE, 108,302-25. Jirka, G. H., Watanabe, M., Octavio, K. H., Cerco, C. E & Harleman, D. R. E (1978). Mathematical predictive models for cooling ponds and lakes, Part A: Model development and design considerations. Report No. 238, R. M. Parsons Laboratory, MIT, Cambridge Mass. Meyer, K. A. (1983). A numerical model to describe the layer behavior of saltgradient solar ponds. J. Solar Energy Engng, 105,341-7. Patterson, J. C., Hamblin, P. E & Imberger, J. (1984). Classification and dynamic simulation of the vertical density structure of lakes. Limnol. Oceanogr., 29 (4), 845-61. Rabl, A. & Nielsen, C. E. (1975). Solar ponds for space heating. Solar Energy, 17, 1-12. Sturm, T. W., Fulford, J. M. & Fay, K. J. (1983). Lake temperature dynamics in hybrid cooling systems. J. Environ. Engng, 109 (3), 668-84.

Vous aimerez peut-être aussi