Thermal Mediation in a Natural Littoral Wetland: Measurements and Modeling Please share

advertisement
Thermal Mediation in a Natural Littoral Wetland:
Measurements and Modeling
The MIT Faculty has made this article openly available. Please share
how this access benefits you. Your story matters.
Citation
Andradóttir, Hrund Ó., and Heidi M. Nepf. “Thermal Mediation in
a Natural Littoral Wetland: Measurements and Modeling.” Water
Resources Research 36.10 (2000): 2937–2946. ©2000 American
Geophysical Union.
As Published
http://dx.doi.org/10.1029/2000WR900201
Publisher
American Geophysical Union
Version
Final published version
Accessed
Wed May 25 17:59:30 EDT 2016
Citable Link
http://hdl.handle.net/1721.1/69012
Terms of Use
Article is made available in accordance with the publisher's policy
and may be subject to US copyright law. Please refer to the
publisher's site for terms of use.
Detailed Terms
WATER RESOURCES
Thermal
mediation
RESEARCH,
in a natural
VOL. 36, NO. 10, PAGES 2937-2946, OCTOBER 2000
littoral
wetland:
Measurements and modeling
Hrund0. Andrad6ttir
andHeidiM. Nepf
Ralph M. ParsonsLaboratory,Departmentof Civil and EnvironmentalEngineering
Massachusetts
Institute of Technology,Cambridge
Abstract. As a river flowsthroughshallowlittoral regionssuchas wetlands,forebays,and
side arms,the temperatureof the water is modifiedthroughatmosphericheat exchange.
This process,whichwe call thermal mediation,can controlthe initial fate of river-borne
nutrient and contaminantfluxeswithin a lake or reservoir.This paper presents
temperatureobservationsthat demonstratethe occurrenceof thermal mediationand
directlysupportthe theoreticalresultsderivedby AndradOttirand Nepf [2000]. The
measurements
showthat the wetlandwarmsthe river inflowby approximately1-3øC
during summerand fall nonstormconditions.Lessthermal mediation occursduring
storms,both becausethe residencetime is significantlyreducedand becausethe wetland
circulationshiftsfrom laterallywell mixed(low flows)to short-circuiting
(storms).The
dead-zonemodel can simulateboth theseregimesand the transitionbetweenthe regimes
and is therefore a good choicefor wetland modeling.
1.
This paper presentsdetailed observationsof thermal mediation in a natural wetland that receivesunregulatedriver inflow. The major contributionsof the paper are to (1) demonstrate through field observation that wetland thermal
mediationoccursin a smallwatershedand howit is affectedby
Introduction
Wetlands can play an important role in improving downstream water quality. Numerousmassbalance studieshave
shownthat suspendedsediments,nutrients,metals, and anthropogenicchemicalsare efficientlyremovedin natural and flow conditions,(2) comparewetland thermal structureand
constructedwetlands through a variety of sink mechanisms, circulationduringlow flowsand storms,(3) validatethe useof
suchas bacterial conversion,sorption,sedimentation,natural the dead-zonemodel in wetlands,both by confirmingthe andecay,volatilization,and chemicalreactions[Tchobanoglous, alytical dead-zone model results derived by Andrad6ttir and
1993].Yet other studieshaveshownthat wetlandsmay alsoact Nepf [2000] and by showingthat the model simulateswell
asa temporalsourceof nutrientsand pollutantsastheyrelease wetland thermal behaviorduringvariable flow conditions,and
storedmaterials[Mitschand Gosselink,1993,p. 157].To date, (4) demonstratethe effect of wetlandthermal mediationon
extensive research has been conducted to understand
the
lake intrusiondynamics.This paper is the observationalcounchemical,biological,and physicalprocessesunderlyingthe terpart to the theorypresentedbyAndrad6ttirandNepf [2000].
sink/sourcepotential of these complexecosystems.
However,
one processthat so far has receivedlittle attentionis thermal
mediation,i.e., the temperaturemodificationof the water flow- 2. Theoretical Background
Thermal mediation is a well-known processin cooling
ing through the wetland. For littoral wetlandsthe outflow
temperature determinesthe lake intrusion depth which, in ponds, in which atmosphericheat exchangeis exploited to
turn, affects the initial fate of nutrients/contaminants in the attenuatethe wasteheat from power plants.Andrad6ttirand
lake, aswell as the residencetime and mixingdynamicswithin Nepf [2000] showedthat thermal mediationis also an importhe lake. Using a dead-zonemodel, Andrad6ttir and Nepf tant processin littoral wetlands,when the river inflow follows
[2000]showedthat wetlandthermalmediationcanprofoundly a different seasonaltemperaturecyclethan the wetlandwater.
affect intrusiondepth.For example,this processcan shift the
timescaleof lake intrusion depth variability from predominantly seasonalto synopticand diurnal. Moreover, wetlands
can prolongsurfaceintrusionsduringsummer,which can lead
to increasedhuman exposureto river-borne contaminants.
Similarly,the increasednutrientsupplyto the epilimnionduring the growing seasoncan acceleratelake eutrophication
[Carmacket al., 1986;Metropolitan
Council,1997].On the other
hand,more surfaceintrusionslead to quickerflushing,potentially reducingthe long-termdepositsof nutrientsand contam-
This condition is met in small or forested watersheds, where
Copyright2000 by the American GeophysicalUnion.
H, and the surface heat transfer coefficientK. The ratio of the
Paper number2000WR900201.
nominalresidencetime J to the thermalinertia, alsocalled
0043-1397/00/2000WR900201 $09.00
thermal capacity,
the river followsa dampedseasonalcyclebecauseof groundwater recharge[Gu e! al., 1996] and/or sun shading[Sinokro!
and Stefan, 1993]. Thermal mediation is less significantin
largerwatersheds,in whichthe river hasenoughtime to ½quilibrat½with the atmosphereand in which sunshadingand wind
shelteringare lessprominent.
While the inflow conditionsetsthe potential for wetland
thermal mediation,the degree of thermal mediation actually
occurring is governed by the residence time distribution
(RTD) withinthe wetlandandthe thermalinertiaof the water
inants in the lake. Wetland thermal mediation can therefore
/heat'The thermal inertia representsthe heating timescaleof
havecomplexshort-andlong-termeffectson lakewaterquality.
the water columnand is definedasthe ratio of the water depth
2937
2938
ANDRAD(STTIR AND NEPF: THERMAL MEDIATION IN NATURAL LITTORAL WETLAND
Table 1. VegetationDrag Estimatesin the Upper Forebaya
Vegetation Flow Re = UH/v
Yes
Yes
No
low
high
high
O (103)
O (105)
O (105)
Cfb
0.2-2(1, 2)
0.008-0.02
(2,3)
0.004(4)
A
mixed reactorwhere the RTD has a large variancearoundthe
meannominalresidence
timeJ (Figurela) [Kadlec,1994].
During highflows,however,the circulationbecomesinertia- or
(A, •, andFi -2 < 1), andsubstantial
short
28-280 jet-dominated
1.1-2.8 circuitingoccurs;that is, a large portion of the river flow exits
0.55
the wetland in much less time than the nominal residence time
abeddragdecreases
significantly
at highReynoldsnumbers,Re, due
to pronationof vegetationstem.
J (Figurelb). Boththeseregimes,
andespecially
the short-
circuitingregime, produce less thermal mediation than the
bEstimates
are adaptedfrom(1) KadlecandKnight[1996,p. 201], ideal plug flow [Andrad6ttirand Nepf, 2000].
(2) Chen[1976],(3) Dunnetal. [1996,p. 54],and(4)Andrad6ttir[1997,
pp. 69, 74].
3.
Methods
3.1.
F = J/theat
,
(1)
Site Description
The Aberjonawatershedis a medium-sized
watershed(65
km2 surface
area)located
in suburban
Boston,
Massachusetts.
is a major factor controlling how much thermal mediation
occurs.For "ideal" plug flow, r = 0 producesno thermal
mediation,whereasr > 3 producesover90% of the maximum
thermalmediationpossiblein the system[e.g.,Jirkaand Watanabe,1980].Wetland flow regimes,however,are rarely ideal
[Kadlec,1994],and water circulation,whichgovernsboth the
skewnessand varianceof the RTD, will generallymodify how
efficientlythe wetland mediatesthe water temperatures.
Wetland water circulationdependson three meteorological
Becauseof long term industrial activitiesthe watershedis
heavily contaminatedwith heavy metals such as arsenicand
lead whichare routinelytransporteddownstreamby the Aberjona River until they are finally depositedin the Upper Mystic
Lake [Solo-Gabriele,
1995;•4urilioet al., 1994].Beforeentering
the lake, the river flowsthroughtwo littoral wetlands,first the
largerupperforebayand then the smallerlowerforebay(Figure 2a). Both wetlandsare vegetatedwith water lilies and
submergedcoontail,and their mean water depth rangesbeforces: wind, which scaleson the mean wind shear stressand tween 1.3 and 2.0 m dependingupon season.In comparison,
wetland length as •wL; river buoyancy,which scaleson the the lake epilimnionis approximately5 m deep, and the therdensityanomalybetweenthe inflow and wetland water and moclineis 5 m wide duringthe summer.
waterdepthasAp#H2;andrivermomentum
perunitwidth,
In this paper, we focuspredominantlyon the thermal mepUSH, whereUo is the inflowvelocityand p is the water diation occurringin the upper forebay shownon Figure 2b.
density.In addition,the water circulationis influencedby veg- This wetland providesmost of the thermal mediationin this
etationdrag,whichscales
onthefrictionfactoras• p
. system,first, becausethe river dischargesinto it, second,beThe relative importanceof vegetationdrag, wind, and water causeits largersize(thuslongerresidencetime) allowsmore
buoyancyto the riverinertiacanbe summarized
by the ratio of thermal alterations to occur, and, third, because it receives
scales:
limited return flow from the lake. In contrast, the lower fore-
Vegetationdragparameter[DePaoli,1999]
bayexchanges
significantly
with the lake throughthe 60 m wide
channelconnectingthis wetland to the lake as illustratedon
Figure 2a. The lower forebaycan thus be consideredto be an
C•L
A = 2 •q'
(2)
Wind parameter
n =
•w L
(3)
(InternalFroudenumber)
-2
Fi-2= IAP/P
#H
V• '
(4)
Noticethatsinceboth• andFi-2 crU•-2, thentheinfluence
of wind and buoyancyon water circulationchangeswith flow
conditions.Similarly,A is indirectlyrelatedto Uo throughthe
I
I
RTD(O
RTD(O
frictionfactorCf, whichdepends
strongly
onwaterdepthand
velocity, as well as vegetationtype and density[Petrykand
BosmajianIII, 1975; Chen, 1976; Shih and Rahi, 1982]. In
particular,A candecreasedrasticallyduringstorms,asa result
of increasedwater depthsand stempronation.This is shownby
_
)
t
r
:
t
t
a)
b)
)
t
the characteristic
valuesof Cf givenin Table1. Watercircu- Figure 1. Bimodal circulationin free water surfacewetlands.
lation is thereforeexpectedto changebetweenlow flowsand
storms[DePaoli, 1999;•4ndrad6ttir,1997]: During low flows,
vegetationdrag,wind, and buoyancydominatethe water cir-
(a) Underlowflows,drag,wind,and/orbuoyancy
dominatethe
circulation
(A, •, andFi-2 >> 1); thewetlandislaterally
well
mixedproducinga relativelysymmetricresidencetime distri-
bution(RTD) aroundthenominalresidence
timeJ. (b) Durculation(A, •, andFi-2 >> 1). Uniformvegetation
dragwill ing high flowsthe circulationis river-dominated(A, •, and
help distributethe riverflowoverthewidthof thewetland[Wu Fi -2 < 1); shortcircuiting
occurs
producing
a skewed
RTD
and Tsanis,1994], and wind will enhancelateral and vertical with much of the flow exitingthe wetland in muchshortertime
mixing.Consequently,the wetlandbehavesas a partiallywell than t.
ANDRADOTTIR
AND NEPF: THERMAL MEDIATION
extensionof the lake surfacelayer, and the outflowtemperature of the upper forebaycanbe taken asrepresentativeof the
temperatureof the lake inflow.
3.2.
Field
a)
Water temperatureswere monitoredevery5 min at different
locationsin the Upper Mystic Lake systemfrom July to November in 1997 and 1998. Temperature measurementswere
cm below the surface and 10-20
cm above the bed
2939
Upper
Førebayj•J•
....._2•
Aberjona
(•
Observations
made 10-20
IN NATURAL LITTORAL WETLAND
,•
River
Lower
i•0.053
•\
Upper
Mysti•
•
at each site in the upper forebayshownon Figure 2b using
Onset temperatureloggerswith 0.2øCresolution.In 1997 the
temperatureloggersat the outlet of the wetlandwere repeatedly stolen, so the following year, the measurementswere
'97,'98
repeatedwith additionalloggersdeployedat the outletandin
the shallowvegetatedzone. Lake water temperatureswere
measuredin 1997 at 1-1.5 m depthintervalswithin the surface
mixedlayer and thermocline(1.8-10.9 m depth) at the locab)
tions shownon Figure 2a usingthermistorchainswith 0.1øC
resolution[Frickerand Nepf, 2000]. Lake intrusiondepthwith
andwithout the wetlandswas estimatedby matchingthe averageof the near-surfaceand near-bedwetlandchanneltemperature and the river bed temperature,respectively,to the lake
thermistordata. Surfaceintrusionswithin the top 2 m in the
J•Coontail
lake
were
not
resolved
because
no thermistor
.8m
was located
there. Changesin suspendedsedimentlevelsdid not significantlyaffect the water densityand were neglectedin the calculations.
Wind speed and direction were measured 10 m above
Zone
2 ¾•__••_-•'•
groundat the southernend of the lake at 10 min intervalsin
1997 and 1998 (Figure 2a). Hourly Aberjona River flow was
0.8
metres
measuredby the U.S. GeologicalSurveyapproximately800 m
upstreamof the inlet to the upper forebay.In addition, air
temperature,relativehumidity,and solarradiationwere monitored every5 min during 1997 (Figure 2a). Cloud coverwas Figure 2. Monitoring program in the Upper Mystic Lake
(a) Site overviewand surmeasuredat 3 hour intervalsat the Boston Logan Airport system,Winchester,Massachusetts.
face
areas.
(b)
Detailed
map
of
the
upperforebaywith mean
located 15 km east of the studysite.
water depthsduring the 1998 monitoringperiod. Thermistor
locationsare denotedby T, the anemometeris denotedby W,
3.3. Dead-Zone Model Application
and the weatherstationis denotedby S.
3.3.1.
Model formulation.
The dead-zone model was de-
Dead
Zone
Qr
0
scribedin detail byAndrad6ttirand Nepf [2000].In short,the
wetlandareais dividedintotwozones,a flowzone(or channel)
with cross-sectional
areaA c and a stationarydead zone with
cross-sectional
areaAa. As shownon Figure2b,while the river
traverses
the wetlandchannel(shadedarea),it exchanges
continuouslywith two stationarydeadzoneson either sideof the
channel. Note that the southern dead zone is smaller, shal-
lower, and more vegetatedthan the northern dead zone. The
circulationregime in the wetland is describedby the areal
ratio, q = Ac/(A c + Ad) , and the nondimensional
lateral
exchangecoefficient a* - AQ/Q r, where AQ is the total
lateralexchange
rate betweenthe channelanddeadzonesand
Qr is the inflow flow rate. Neglectinglongitudinaldispersion,
the simplifiedgoverningequationsfor the depth-averaged
water temperaturein the channel,Tc(X, t), anddeadzone,Td(X,
t), are
a7'c
a7'c a*u
--+ u --=at
O•'d
Ox
L
a*U q
rc) +-
rk
pCpHc'
(5)
q•
t) = r0(t),
O•c
Ox
are
=0.
(7)
(8)
x=L
Here u = Q•(t)/Ac is the averagespeedin the channel,L is
the length of the wetland, T O is the wetland inflow tempera-
ture,rkisthenetsurface
heatfluxpersurface
area,andCp is
the specificheat of water.
The parametersq and a* are site-dependentand are typically evaluatedby conductingdye experiments[e.g.,Bencala
and Walters,1983]. For preliminary designand assessment,
however, it is useful to derive expressionsbased upon the
inflow and site geometry alone. First, consider the riverdominatedcirculationregime,whenthe riverflowsthroughthe
wetlandlike a jet (Figure lb). The laboratoryexperimentsof
Chu and Baines[1989] suggestthat the jet spreadslinearly
from its initial width Wo, with spreadingangle/• • 0.1, such
that the channelwidth is W• = Wo + 2/•x. This yields the
at= L 1- q(•d--•c)+pCpHd.(6) mean
The boundaryconditions
at the inlet (x - 0) andoutlet(x - L)
100 200 300
wetland
areal ratio of
q=
W
• •/• L
W'
(Wo+I•L)
He
(9)
2938
ANDRADOTTIR
AND NEPF: THERMAL MEDIATION
Table 1. Vegetation Drag Estimatesin the Upper Forebaya
Vegetation Flow Re = UH/v
Yes
Yes
No
low
high
high
O (103)
O (10s)
O (10s)
Cf•
0.2-2(1, 2)
0.008-0.02
(2, 3)
0.004(4)
A
IN NATURAL LITTORAL WETLAND
mixedreactorwhere the RTD has a large variancearound the
meannominalresidence
timeJ (Figurela) [Kadlec,1994].
During highflows,however,the circulationbecomesinertia- or
(A, •, andFi -2 < 1), andsubstantial
short
28-280 jet-dominated
1.1-2.8 circuitingoccurs;that is, a large portion of the river flow exits
0.55
the wetland in much less time than the nominal residence time
abeddragdecreasessignificantly
at high Reynoldsnumbers,Re, due
to pronationof vegetationstem.
J (Figurelb). Boththeseregimes,
andespecially
the short-
circuitingregime, produce less thermal mediation than the
bEstimates
are adaptedfrom(1) KadlecandKnight[1996,p. 201], ideal plug flow [Andrad6ttirand Nepf, 2000].
(2) Chen[1976],(3) Dunn et al. [1996,p. 54], and (4)Andrad6ttir[1997,
pp. 69, 74].
3.
Methods
3.1.
r -' J/theat
,
(1)
Site Description
The Aberjonawatershedis a medium-sizedwatershed(65
km2 surface
area)locatedin suburban
Boston,
Massachusetts.
is a major factor controlling how much thermal mediation
occurs.For "ideal" plug flow, r = 0 producesno thermal
mediation,whereasr > 3 producesover 90% of the maximum
thermal mediationpossiblein the system[e.g.,Jirka and Watanabe,1980].Wetland flow regimes,however,are rarely ideal
[Kadlec,1994], and water circulation,which governsboth the
skewness
and varianceof the RTD, will generallymodify how
efficientlythe wetland mediatesthe water temperatures.
Wetland water circulationdependson three meteorological
forces: wind, which scaleson the mean wind shear stressand
wetland length as •wL; river buoyancy,which scaleson the
densityanomaly between the inflow and wetland water and
Because of long term industrial activities the watershed is
heavily contaminatedwith heavy metals such as arsenic and
leadwhich are routinelytransporteddownstreamby the Aberjona River until they are finally depositedin the Upper Mystic
Lake [Solo-Gabriele,
1995;/1urilioet al., 1994].Beforeentering
the lake, the river flowsthroughtwo littoral wetlands,first the
largerupperforebayand then the smallerlowerforebay(Figure 2a). Both wetlandsare vegetatedwith water lilies and
submergedcoontail,and their mean water depth rangesbetween 1.3 and 2.0 m dependingupon season.In comparison,
the lake epilimnion is approximately5 m deep, and the thermoclineis 5 m wide during the summer.
waterdepthasAp#H2;andrivermomentum
perunitwidth,
In this paper, we focuspredominantlyon the thermal mepUSH, whereUo is the inflowvelocityand p is the water diation occurringin the upper forebay shownon Figure 2b.
density.In addition,the water circulationis influencedby veg- This wetland providesmost of the thermal mediation in this
etationdrag,whichscales
onthefrictionfactoras•l pCœU•L
. system,first, becausethe river dischargesinto it, second,beThe relative importanceof vegetationdrag, wind, and water causeits larger size (thus longerresidencetime) allowsmore
buoyancyto the river inertia canbe summarizedby the ratio of thermal alterations to occur, and, third, because it receives
scales:
limited return flow from the lake. In contrast, the lower fore-
Vegetationdrag parameter[DeP•oli, 1999]
bayexchanges
significantly
with the lake throughthe 60 m wide
channelconnectingthis wetland to the lake as illustrated on
Figure 2a. The lower forebaycan thusbe consideredto be an
C•L
A = 2 H'
(2)
Wind parameter
•w L
1•= pU•H'
(3)
(InternalFroudenumber)
-2
Fi-2=IAP/PI#
H
U• '
(4)
Noticethatsinceboth• andFi-2 crU• 2, thentheinfluence
of wind and buoyancyon water circulationchangeswith flow
conditions.Similarly,A is indirectlyrelated to Uo throughthe
I
I
RTD(O
RTD(t)
frictionfactorCf, whichdepends
strongly
onwaterdepthand
velocity,as well as vegetationtype and density [Petlykand
BosmajianIII, 1975; Chen, 1976; Shih and Rahi, 1982]. In
particular,A can decreasedrasticallyduringstorms,as a result
of increasedwater depthsand stempronation.This is shownby
_
> t
r
,_
t
t
a)
b)
>
t
the characteristic
valuesof Cf givenin Table1. Watercircu- Figure 1. Bimodal circulationin free water surfacewetlands.
lation is therefore expectedto changebetweenlow flows and
storms[DePaoli, 1999;Andrad6ttir,1997]: During low flows,
vegetationdrag, wind, and buoyancydominatethe water cir-
(a) Underlowflows,drag,wind,and/orbuoyancy
dominatethe
circulation
(A, •, andFi-2 >> 1); thewetlandislaterally
well
mixed producinga relatively symmetricresidencetime distri-
bution(RTD) aroundthe nominalresidencetime t. (b) Dur-
culation
(A, •, andFi-2 >> 1). Uniformvegetation
dragwill ing high flows the circulationis river-dominated(A, •, and
help distributethe river flow overthe width of the wetland[Wu Fi -2 < 1); shortcircuiting
occurs
producing
a skewed
RTD
and Tsanis,1994], and wind will enhancelateral and vertical with muchof the flow exitingthe wetland in much shortertime
mixing. Consequently,the wetland behavesas a partially well than t.
ANDRADOTTIR
AND NEPF: THERMAL MEDIATION
extensionof the lake surfacelayer, and the outflow temperature of the upperforebaycanbe taken asrepresentativeof the
temperatureof the lake inflow.
3.2.
Field
made 10-20
cm below the surface and 10-20
cm above the bed
at each site in the upper forebay shownon Figure 2b using
Onset temperatureloggerswith 0.2øCresolution.In 1997 the
temperatureloggersat the outlet of the wetlandwere repeatedly stolen, so the following year, the measurementswere
repeatedwith additionalloggersdeployedat the outlet and in
the shallowvegetatedzone. Lake water temperatureswere
measuredin 1997 at 1-1.5 m depth intervalswithin the surface
mixed layer and thermocline(1.8-10.9 m depth) at the locations shownon Figure 2a usingthermistor chainswith 0.1øC
resolution[Frickerand Nepf, 2000]. Lake intrusiondepthwith
and without the wetlandswas estimatedby matchingthe averageof the near-surfaceand near-bedwetlandchanneltemperature and the river bed temperature,respectively,to the lake
thermistordata. Surfaceintrusionswithin the top 2 m in the
lake
were
a)
not resolved
because
no thermistor
2939
Upper
Forebay
,,.J t,•Aberjon
a
.ive
Lower
(•0.053i•••
'•-•
Observations
Water temperatureswere monitoredevery5 min at different
locationsin the Upper Mystic Lake systemfrom July to November in 1997 and 1998. Temperature measurementswere
IN NATURAL LITI'ORAL WETLAND
Førebay
•1•
••'•
• /
Upper
Mysti•
Lake
/
•
b)
• Coonta•
was located
there. Changesin suspendedsedimentlevels did not significantly affect the water densityand were neglectedin the cal-
IDead
Zone
•* ' '
culations.
Wind speed and direction were measured 10 m above
ground at the southernend of the lake at 10 min intervalsin
1997 and 1998 (Figure 2a). Hourly Aberjona River flow was
measuredby the U.S. GeologicalSurveyapproximately800 m
upstreamof the inlet to the upper forebay. In addition, air
temperature,relative humidity,and solar radiationwere monitored every5 min during 1997 (Figure 2a). Cloud coverwas
measured at 3 hour intervals at the Boston Logan Airport
located 15 km east of the studysite.
3.3.
Dead-Zone Model Application
3.3.1.
Model
formulation.
The dead-zone
model was de-
' 0.8
0
lower, and more vegetatedthan the northern dead zone. The
circulationregime in the wetland is describedby the areal
ratio, q = Ac/(A c + Aa), and the nondimensionallateral
exchangecoefficienta* - AQ/Q r, where AQ is the total
lateral exchangerate betweenthe channeland dead zonesand
Qr is the inflow flow rate. Neglectinglongitudinaldispersion,
the simplifiedgoverningequationsfor the depth-averaged
wa-
tertemperature
in thechannel,
•c(X, t), anddeadzone,•a(x,
t), are
O?c a*u
o-7+u = L
OTa
a*u
q
Ot
L
1- q
0
- ?c)
+pCdqc' (5)
rO +
0
pCpHa'
(6)
The boundaryconditions
at the inlet (x = 0) andoutlet(x = L)
are
100
200
300
Figure 2. Monitoring program in the Upper Mystic Lake
system,Winchester,Massachusetts.
(a) Site overviewand surface areas.(b) Detailed map of the upper forebaywith mean
water depths during the 1998 monitoring period. Thermistor
locationsare denotedby T, the anemometeris denotedby W,
and the weather stationis denotedby S.
scribedin detail byAndrad6ttirand Nepf [2000]. In short,the
wetlandareais dividedinto two zones,a flowzone(or channel)
with cross-sectional
area A c and a stationarydead zone with
cross-sectional
areaA a- As shownon Figure 2b, while the river
traversesthe wetlandchannel(shadedarea), it exchanges
continuouslywith two stationarydead zoneson either side of the
channel. Note that the southern dead zone is smaller, shal-
metres
_
_
Tc(O,t) - To(t),
ø?c
I =0
Ox
x=L
'
(7)
(8)
Here u = Q•(t)/Ac is the averagespeedin the channel,L is
the length of the wetland, To is the wetland inflow tempera-
ture,45is the netsurface
heatfluxper surfacearea,andCp is
the specificheat of water.
The parametersq and a* are site-dependentand are typically evaluatedby conductingdye experiments[e.g., Bencala
and Walters,1983]. For preliminary designand assessment,
however, it is useful to derive expressionsbased upon the
inflow and site geometry alone. First, consider the riverdominatedcirculationregime,when the river flowsthroughthe
wetland like a jet (Figure lb). The laboratoryexperimentsof
Chu and Baines[1989] suggestthat the jet spreadslinearly
from its initial width Wo, with spreadingangle/3 •-- 0.1, such
that the channelwidth is Wc = Wo + 213x.This yields the
mean wetland
areal ratio of
q=
W13L
• •/3 W'
Wo
+
) Hc
L
(9)
2942
ANDRADOTTIR AND NEPF:THERMAL MEDIATION IN NATURAL LITTORAL WETLAND
ticular,duringthe initialstageof the storm(J.D. 305.7-306)
the water temperaturein the channelincreasesabruptly,
a)
whereasthe water temperaturein the dead zone risesmore
gradually.
Thisisconsistent
withshortcircuiting
(Figurelb) in
whicha largeportionof the riveradvects
rapidlythroughthe
wetlandchannelwhile the remaindermixesslowlywith the
deadzones.The timelagbetweenthe temperature
increasein
the riverandwetlandchannel(1 hour)matchesthe advective
timescale
estimated
for the riverjet [ChuandBaines,1989].
24
T
(øC)
20
Also, note that the water column in the flow zone is well mixed
duringthe risinglimb of the hydrograph,
as expectedwhen
riverinertiadominates
buoyancy
andwind (Fi -2 and 12 <
16
220
230
240
250
260
270
Julian Days
October
November
b)
1998
16
-
0.5). However,this one-layerflow structurebreaksdown on
therecedinglimbshortlyafterthe onsetof a southerly
windat
J.D. 306.0.The colderwaterthat appearsat the channelbedis
likelyadvectedfrom the northerndeadzoneby the subsurface
returnflow associated
with the windsetup.
Next considerthe Auguststormillustratedon Figure5b. In
_
a) 14 ....
I
, , , ....
November
19'97'
T
(øC)
12
13
_
T
, ....
20
.........
ß
T
.......
'.........
""'
""?-----(_
..R
,,T• ,, ,,
15
12
:
T
•
10QR
11
(øC)
I
270
280
290
300
310
320
(m3/s)
10.
.,._.>.•
_/,
O
_
5
Julian Days
8
Figure 4. Watertemperatures
in thewetlandchannelTc and
,
,
,
305
,
I
-•',
305.5
the northernandsoutherndeadzones,Ta• andTa2, respectively,during(a) early and (b) late fall 1998.The bold line
b)
depictsthe near-surface
temperatures,
andthe lightlinesdepictnear-bedtemperatures.
Cross-hatched
areasdenotelarge
stormoccurrences
whenQr > 1 m3/s.Wetlandstratification
21and diurnalsurfacetemperaturefluctuations
decreaseprogressing
intothefall because
of convective
coolingandincreas- T 20
,
,
I
I
306
306.5
0
31•7
Julian Days
20
22:.:'_;'•.'
.... '''' Adg-•t'l,•9'
1
'
_.,
T• -',
ing winds. Observationresolutionis _+0.2øC.
10 to 1000. Under these conditions the wetland can exhibit a
complex,transient,three-dimensionalflow behavior. Since A
rangesfrom about30 to 300 (Table1), vegetationdragis also
expectedto play an importantrole in distributingthe river
laterally[DePaoli,1999].Dye experiments
showthat the integratedeffectof all theseprocesses
overtheresidence
time(>3
weeks)is that the wetlandbehavesas a partiallywell mixed
reactor[Andrad6ttir,
1997].Thissupports
the useof a simple
one-dimensional
(l-D) modelsuchasthe dead-zonemodelfor
low-flowconditions,
eventhoughthe instantaneous
circulation
may be far from 1-D.
4.2.2. Storms. The cross-hatched
areason Figure4 show
that the diurnal thermal fluctuations are reduced or erased
15
10 Q•
(øC)
19- "'".,
,:"""T•
(m3/s)
18
i
17
•
233
C)17
16
T 15
(øC)14
233.5
234
234.5
5
0
235
JulianDays
'
'
'
, '
' ' I ' ' '
October
1998
, ' , , , , , , 120
/:::;:::..
1
R ''":"
'%'•""
10 Q•
(m3/s)
13
duringstorms(e.g.,J.D. 229,265,and282)asthewetlandheat
12
budgetismorestrongly
influenced
bytheriverheatinputthan
281.4
281.6
281.8
282
282.2
282.4
atmospheric
fluxes.Other importantstormfeaturesare demJulian
Days
onstratedin the threelargestormson Figure5. First,consider
thelargeststormin 1997,portrayed
onFigure5a.As istypical Figure 5. River and wetland water temperaturesat x =
for late fall storms,the river is a sourceof warm water to the
0.4L duringlargestormsin (a) November1997,(b) August
wetland.Unlikelow-flowconditions
thewatertemperatures
in 1997,and (c) October1998.Bold linesdepictnear-surface
the channelanddeadzonedeviatefromoneanother.In par- temperatures,
andlightlinesdepictnear-bedtemperatures.
ANDRAD(STTIR
AND NEPF: THERMAL MEDIATION
IN NATURAL LITTORAL WETLAND
the beginningof the stormthe river is a sourceof coldwater to
a) 26
the wetland,the reverseof the previousstorm.Shortcircuiting
is again observedas the channel temperature drops more
24
abruptlythan the dead zone temperaturearound J.D. 234.0.
Furthermore,the water columniswell mixedon the risinglimb
T
of the stormbut becomesstratifiedon the recedinglimb. Un22
like the previousstorm, this transition cannot be easily ex(øC)
plained based upon the available data, becauseit does not
coincidewith wind setupnor a significantincreasein buoyancy
forcing. After J.D. 234.0 the colder river flows through the
20
wetland as an underflow until around J.D. 234.4, when the
river becomes warmer
than the wetland
water.
The river then
18
switchesto an overflow that appearsto be short-circuiting,
indicatedby the suddendrop in the near-surfacetemperature
in the channel
J.D.
that is not mirrored
' I ' I ' I .' I '
ß
z = -0.2m.'"
'•,, •
'
i
I1'/
,
--
,
simulated
'• and
Ta
.... observed
TI
,
240
I
I
242
I
,
244
I
I
246
248
250
Julian Days
b) 26
Finally, considerthe largeststormin 1998, depictedon Figure 5c.The wetlandthermalresponseisverycomplicated,both
'
I
'
I
'
I
'
I
'
T atx=L
c
24
because this event consists of a series of storms and because it
occurson windydays(peakwindspeed7 m/s).In the following
we focussolelyon the secondstorm (J.D. 282.2), when the
river iswarmerthan the wetlandwater and the wind is blowing
from the north, producinglessinterferencewith the jet structure than in the first storm (J.D. 281.6). The temperature
recordson the risinglimb of this storm showthe warm river
pulsemovingthroughthe wetlandasan overflow,reachingthe
channel station more rapidly than both the northern and
southerndeadzones.This is in agreementwith the conceptual
pictureof shortcircuitingportrayedon Figure 2b in whichthe
x = 0.4L
,
in the dead zone around
234.7.
2943
T
22
(øC)
18
240
,
I
242
I
244
246
Julian Days
248
250
channel cuts through the middle of the wetland. The river
continuesto short-circuiton the recedinglimb (J.D. >282.2), Figure 6. Low-flowmodel simulations.(a) Comparisonbeas seenby the fact that the water in the channelis significantly tween dead-zone model simulations and field measurements at
warmer than that in the northern dead zone. Furthermore, in x = 0.4L. (b) Wetland outflowtemperaturepredictedby the
contrastto the stormson Figures5a and 5b, the flow is strat- dead-zone and stirred reactor models.
ified on the risinglimb of the stormbut is well mixed on the
recedinglimb.
To summarize,although the temperature records during convectivecoolingcanbe explainedby the fact that the model
stormeventsare quite complex,they consistentlydemonstrate does not accountfor wind sheltering.During this low-flow
that the river short-circuits
throughthe wetland (Figure lb). period the residencetime in the wetlandis so long, r w •- 8,
Our observationsthus support the hypothesisthat wetland that the upperforebayapproaches
the limit of functioninglike
circulationchangesduringstormsfrom partiallywell mixedto a stationarywater body [Andrad6ttirand Nepf, 2000]. At this
short-circuiting.
In addition,the upperforebayexhibitsvertical limit the thermal budget is predominantlygovernedby the
structureboth duringlow flowsand storms.This suggests
that surfaceheat flux and not the river heat input, and the detailed
the use of a depth-averageddead-zonemodel is at times an circulationregime in the wetland is not so important in preoversimplification,
especiallyduringstorms.This will be inves- dicting thermal mediation. This is demonstratedby the fact
tigated in greater detail in section4.3.
that both the stirredreactorand the dead-zonemodelspredict
very similarwetlandoutflowtemperatures(Figure 6b). Note,
4.3.
Dead-Zone
Model Simulations
however,the wetlandcirculationis still importantfor predictTo further validatethe applicationof the dead-zonemodel, ing chemicalor solid removal which is unaffectedby atmowe now consider how well the model simulates the thermal
sphericexchange[Thackstonet al., 1987].
4.3.2. Storms. Figure 7 summarizesmodel simulations
responsewithin the upper forebayduringboth low flowsand
storms.
during the three large stormsin 1997 and 1998. Recall from
4.3.1. Low flows. Figure 6 illustratesmodel simulations section4.2 that thesestormsall exhibit short circuitingeither
and observationsduring a 10 day period in 1997, when the as a one layer-flowor as an overflow/underflow.
An important
AberjonaRiverflowratewas0.1 m3/s.Sincethe dead-zone function of the model with respectto lake transport is to
model is l-D, we comparethe simulateddepth-averagedtem- predictthe temperatureof this short-circuiting
flow. Sincethe
peratures(solidlines)to the observednear-surfaceand near- model is l-D, this may involve simulatingonly the surface
bedwater temperatures(dottedlines),whichdefinethe upper wetland temperatureduring overflows(or bed temperatures
and lower limit for an observeddepth-averagedtemperature. duringunderflows).First, considerthe 1997Novemberstorm,
Figure 6a showsthat the model simulateswell low-flowcondi- whichhasthe strongestsignatureof shortcircuiting.Figure 7a
tions, producingvaluesthat lie within the observedtempera- showsthat the dead-zonemodel predictswell the propagation
ture range,excepton windydays(J.D. 246-248) when simu- of the warm short-circuitingriver flow to the middle of the
lations are colder than observations.This overpredictionof wetland(x - 0.4L), evenafterthe flowis no longer1-D (J.D.
2944
ANDRADOTTIR AND NEPF: THERMAL MEDIATION IN NATURAL LITTORAL WETLAND
14
T at x = 0.4 L
T at x = 0.4 L
½
•....
ß
13
12
T=
at
x=0.4
L
T
z = -0.2m
11
(øC)
...,••
8
--simulation
•
r-•.z
/
13
J J ,
,
I
,
•
,
,
I
,
,
I
I
" "-
•
I
I
I
-"
observation
I
23
Td•atx = 0.4L
T
= -1.3mj
I 5m
/'" observation
',..•.,
','....
, , il•'z
, I =•-,',
, I , '"
, , øbservatiøn/
, I , , , ,/
C
.,
t s,mu,ai,11
z = -0.2m
lO
,,,
0.2
T
ß'.
12
22
11
21
dl
at x = 0.4 L
T atx=
L
c
,. 0.3øC
Z = -0.
(øC)
10
z•-0.2m
..:.";'
ß
19
_ --
simulation
z = -1.3m
observation
8
13
•, ,,
18
'
'
'
'
I
'
'
I
12
T
11
22
T•
at
x=L •
T=
at
x=Lt
21 -
T atx=
L
c
_
_
(øC)
lO
• •
dead-zone
model
19- __ dead_zone
model
--
stirred reactor
18
305
305.5
306
306.5
307
233
- - stirred
reactor
....
I , , , , I , , , , I , , , ,
233.5
234
234.5
•
I
I
235
,
281
dead-zone
model
stirred reactor
,
,
,
I
281.5
,
,
,
,
I
,
282
,
,
,
I
,
282.5
Julian Days
Julian Days
Julian Days
a)
b)
c)
,
,
•
283
Figure 7. Stormsimulations
for the (a) 1997November,(b) 1997August,and (c) 1998Octoberstorms.
Dead-zone
modelsimulations
(solidlines)arecompared
withobservations
(dottedlines)andstirredreactor
simulations
(dashedlines)at differentlocations
withinthewetland.The horizontalbarsindicatewhentheflow
is jet-dominated,
i.e., Fi -2 < 2 and12 < 0.5.
>306). The model,however,slightlyunderpredicts
the water
temperaturein the northerndead zone, probablybecauseit
does not accountfor lateral exchangegeneratedby winddrivencirculation.For the 1997 Auguststormportrayedon
Figure 7b, the model again simulateswell the channeland
dead-zonetemperatures
in the middleof the wetland(x =
0.4L). Similarly,Figure 7c showsthat the dead-zonemodel
doesa goodjob predictingthe movementof the thermalfront
alongthewetlandchannelbothatx - 0.4L andx = L during
reactormodelsimulations,
givenat the bottomon Figure7,
highlightsthe strengthof the dead-zonemodel relativeto the
stirred reactor model. The dead-zonemodel capturesthe
sharprises/drops
in water temperatureassociated
with short
circuiting,whereasthe stirredreactormodel predictsa diffused front because it assumes that the river mixes instanta-
neouslythroughoutthe wetland.This diffusionof frontalsignaturescan affectthe predictionof lake intrusiondepth.This
can be seenby comparingthe model simulationsto the mean
temperaturein the lakeepilimnionTt., ason Figure7a.Notice
thermistorprobesis 0.2øC, and the differencebetween the thatbecauseof thelargerreceiving
volumeof thelake,thelake
observedand simulatedwater temperatures(0.2-0.3øC)is water temperaturedoesnot follow the samesharptemperathereforenot significant.
Thesethree simulations
thusdemon- ture increaseobservedin the wetland.During this 1997Nostratethat the l-D, quasi-steady,
dead-zonemodelgivesreli- vemberstormthe stirredreactormodelpredictsa 4 hourdelay
ablepredictionsof shortcircuitingand the transitionbetween in surfaceintrusions
(Tx=z. > Tz.) duringthe peakflowrates.
shortcircuitingand laterallywell mixedflow duringstorms, Sincethe highestconcentrations
typicallyoccurduring the
evenin systems
suchasthe Upper MysticLakeforebays
which risinglimbof a storm,the stirredreactormodelwill underpreexhibitverticalstructure(i.e., non-l-D behavior).
dict the contaminant/nutrient
transportto the lake surface,
Finally,the comparisonbetweenthe dead-zoneand stirred whichcan haveimportantimplicationsfor humanhealthand
the 1998 October storm. Recall that the resolution of the
ANDRADOTTIR
AND NEPF: THERMAL MEDIATION
IN NATURAL LITTORAL WETLAND
2945
reservoirmanagementaswill be discussed
in greater detail in
section 4.4.
4.4. Lake Intrusion Dynamics and Water Quality
Wetlandthermalmediationis an importantphysicalprocess
in part becauseit can alter lake intrusiondynamicswhich, in
turn, impactdownstreamwater quality.Figure 8 describesthe
expectedintrusiondepth in the Upper Mystic Lake with and
without the upper forebay,basedon the water temperature
-4
z -6
(m)-8
measurements in 1997, as discussedin section 3.2. The sea-
sonaland synoptictrendspresentedin Figure 8a showthat in
the absence of the wetland, the colder river water would
plungewhen enteringthe lake throughoutthe fall exceptfor a
few daysof short-termheating(e.g., J.D. 284 and 305-310).
However,with the littoralwetlandpresent,episodesof surface
intrusionare prolongedand occur during 28 daysor 40% of
the monitoringperiod(e.g.,J.D. 256-265, 279-291, and 306312). Therefore substantially
more river-bornefluxesare delivered
into the surface water
of the lake with
-10
F -- w,wetland
".'
\,•,.! ,..• :.
-12
250 260 270•' •80
.
b)-2
290 300 - 310 320
Julian Days
.......
' ....
' ....
1997
the wetland
present,which can have a high impact on the nutrient and
chemicalbudgetsof the lake. A close-upon diurnaltimescales
(Figure 8b) showsthat the intrusiondepth can alsovary significantlyover the courseof a day, especiallyduring synoptic
heatingperiodswhen the inflow insertsinto the lake epilimnion. During these times a small temperaturevariation can
lead to a large changein intrusiondepth becauseof the low
thermalgradientin the lake surfacelayer.In the Upper Mystic
Lake thesevariationscan be as large as 4-6 m. Overall, the
temperaturemeasurementsin the Upper Mystic Lake system
suggestthat the presenceof a littoral wetland shiftsthe timescaleof lake intrusion depth variability from predominantly
seasonal(no wetland)to synopticanddiurnaltimescales(with
wetland).A similarresultwaspredictedby the linearizeddeadzonemodel[Andrad6ttirandNepf, 2000].The shiftin intrusion
depth variabilitycan impactlake water quality.For example,
nutrientsand contaminantsmay be more homogeneously
distributedover depth as the river inflow coversa largervertical
domainin the lake. In addition,lateral and verticaltransport
processes
in the lake canbe enhanced.Both of thesewill affect
the managementof reservoiruse and withdrawals.
The intrusion depth during stormsis of particular importance for lake water quality. Becauseof surfacerunoff and
remobilization of channel sediments, river nutrient and con-
taminant concentrationsusually increase during storms. For
example,stormsaccountfor 30-60% of the total annualfluxes
of heavymetalssuchasarsenic,chromium,iron, and copperto
the Upper Mystic Lake [Solo-Gabriele,1995]. The arrowson
Figure 8a depict the storm occurrencesin the Upper Mystic
lake systemduring fall 1997. As expectedfrom the previous
discussion,the presenceof a wetland does not significantly
modifythe lake intrusiondepth duringstormsbecauseof the
low effectivethermal capacityassociatedwith the high flow
rates. Figure 8a also showsthat sevenout of nine of the late
summerand fall stormsoccurduringperiodsof synopticcooling and produceintrusionsin the thermocline.Thesemeasurements thus suggestthat elevatedsurfaceconcentrationshap-
-4
,,,
i i
(m)
-•
.... w/oweftand
-•0 • •
"'"
w/weftand
',:',,. ,,
-12
280
285
290
295
300
Julian Days
Figure 8. Intrusion depth with and without a wetland in
1997.Wetland thermalmediationincreases(a) synopticand
(b) diurnalintrusiondepthvariabili•. Boldarrowsdepictlarge
(Q• > 1 m3/s)stormoccurrences,
andlightarrows
depictsmall
(Q• < 1 m3/s)stormoccurrences.
Unccrtain•is •0.8 m for
surface intrusions
cline.
determine
and •0.3
the downstream
m for intrusions
into the thcrmo-
fate of land-borne
nutrients
and
contaminants.Measurementsin the Upper MysticLake system
in Massachusettsduring summer and fall 1997-1998 demonstrate that thermal
mediation
occurs in a natural littoral
wet-
land located in a small-to-medium-sized watershed, where sun
shadingproducesa different water temperaturefor the river
than the wetland and lake. In addition, thermal mediation is
affected by flow conditions.During low flows the wetland
raisesthe temperatureof the lake inflow by approximately
1-3øC,whichis sufficientto changethe lake intrusiondepthby
1-6 m and shift the timescaleof intrusiondepth variability
from seasonal
(no wetland)to synopticanddiurnal(withwetland). In contrast,duringstormsalmostno thermalmediation
occurs.These observationsare both in agreementwith the
theorypresentedby Andrad6ttirand Nepf [2000].The change
in thermal responseassociatedwith flow conditionscan parpen infrequentlyduringsummerand fall in the Upper Mystic tially be explainedby a shiftin wetlandcirculation:During low
flows the river inflow fills the entire wetland volume, and the
Lake.
5.
Conclusions
Thermal mediationin littoral wetlandsis an importantprocessthat can alter the intrusiondepth in lakes and thus can
wetland behaveslike a partially well mixed reactor. During
storms,however, the temperature records demonstratethat
the river short-circuitsacrossthe wetland, and thus only a
fractionof the wetlandvolumeis availableto producethermal
mediation.A simplel-D, quasi-steady,
dead-zonemodel can
2946
ANDRAD•TYIR AND NEPF:THERMAL MEDIATION IN NATURAL LITTORAL WETLAND
give reliablepredictionsof thermalmediationduringboth
mountainpool-and-riffle
stream:A transientstoragemodel,Water
Resour.Res.,19(3), 718-724, 1983.
regimesandpredictswell the transitionbetweenshortcircuitBras,R. L., Hydrology,
an Introduction
to Hydrologic
Science,
Addisoning and laterallywell mixedflow.
Wesley-Longman,
Reading,Mass.,1990.
Carmack,
E. C., R. C. Wiegand,R. J. Daley,C. B. J. Gray,S.Jasper,
Notation
Ac, Aa
cross-sectional
area of wetland channeland dead
zone, respectively.
A r river cross-sectional
area at wetland entrance.
Fi-C•
friction
factor.
inverseinternal
Froudenumber,equalto
pigilIUm.
g accelerationof gravity.
H meanwetlandwater depth.
H c, H a meanwater depth in wetlandchanneland dead
zone, respectively.
K
surface heat transfer coefficient.
L wetland length.
AQ lateral exchangeflow rate.
Qr river flow rate.
r thermalcapacity,
equalto 7/theat.
J nominal
residence
time,equaltoHWL/Qr.
theat thermalinertia, equalto H/K.
Tc, Ta wetlandchanneland dead-zonetemperature,
respectively.
TL, T•
u
Uo
q
W, Wc
Wo
a*
lake andriver temperature,respectively.
meanwetlandchannelspeed,equalto Qr/Ac.
river inflowspeed,equalto Qr/Ar.
channelarealratio, equalto A c/(A c + A a).
meanwetlandandchannelwidth,respectively.
initialchannelwidth(at wetlandentrance).
lateralexchange
coefficient,equalto AQ/Qr.
/• linear spreadingcoefficient.
qb surface heat flux.
A dragparameter,
equalto• CfL/H.
p water density.
Ap densityanomalybetweenriver andwetland.
•w mean wind shear stress.
1• windparameter,
equalto •wL/(pU•H).
and C. H. Pharo,Mechanisms
influencingthe circulationand distributionof water massin a mediumresidence-time
lake, Limnol.
Oceanogr.,
31(2), 249-265, 1986.
Chen,C., Flowresistance
in broadshallowgrassed
channels,
J. Hydraul.Div.Am. Soc.Civ.Eng.,102(HY3), 307-322,1976.
Chu,V. H., andW. D. Baines,Entrainment
by buoyantjet between
confined
walls,J. Hydraul.Eng.,115(4),475-492,1989.
DePaoli,L. L., Numericalmodellingof wetlandhydrodynamics,
Master'sthesis,Mass.Inst. of Technol.,Cambridge,1999.
Dunn, C., F. Lopez, and M. Garcia, Mean flow and turbulencein a
laboratory
channel
withsimulated
vegetation,
Hydraul.
Eng.Ser.,51,
Univ. of II1., Urbana, 1996.
Fisher,H. B., E. J. List,R. C. Y. Koh,J. Imberger,andN.H. Brooks,
Mixingin Inlandand CoastalWaters,
Academic,
SanDiego,Calif.,
1979.
Fricker,P. D., andH. M. Nepf, Bathymetery,
stratification,
andinternal seichestructure,
J. Geophys.
Res.,105(6),14,237-14,251,
2000.
Gu, R., F. N. Luck,andH. G. Stefan,Water qualitystratification
in
shallow
wastewater
stabilization
ponds,WaterResour.
Bull.,32(4),
831-844, 1996.
Jirka,G. H., and M. Watanabe,Steady-state
estimation
of cooling
pondperformance,
J. Hydraul.Div.Am. Soc.Civ.Eng.,106(HY6),
1116-1123, 1980.
Kadlec,R. H., Detention
andmixingin freewaterwetlands,
Ecol.Eng.,
3, 345-380, 1994.
Kadlec,R. H., andR. L. Knight,TreatmentWetlands,
CRC Press,Boca
Raton, Fla., 1996.
MetropolitanCouncil,Lake McCarrons
wetlandtreatmentsystemPhaseIII studyreport,Publ. 32-97-026,City of Roseville,Minn.,
1997.
Mitsch,W. J., andJ. G. Gosselink,Wetlands,
Van NostrandReinhold,
New York, 1993.
Petryk,S.,andG. Bosmajian
III, Analysis
of flowthrough
vegetation,
J. Hydraul.Div.Am. Soc.Civ.Eng.,101(HY7), 871-884,1975.
Ryan,P. J., andD. R. F. Harleman,An analytical
andexperimental
studyof transient
cooling
pondbehaviour,
Tech.Rep.161,RalphM.
Parsons
Lab.for WaterRes.andHydrodyn.,Mass.Inst.of Technol.,
Cambridge,1973.
Shih,S.F., andG. S.Rahi,Seasonal
variations
ofManning's
roughness
coefficient
in a subtropical
marsh,Trans.ASAE,25(1), 116-119,
1982.
Sinokrot,B. A., and H. G. Stefan,Streamtemperature
dynamics:
Measurements
andmodeling,
WaterResour.
Res.,29(7),2299-2312,
1993.
H., Metal transportin the AberjonaRiver system:
Acknowledgments.Thisworkwasfundedby the NationalInstitute Solo-Gabriele,
Monitoring,modelling,andmechanisms.
Ph.D. thesis,Mass.Inst.of
of EnvironmentalHealth Sciences,SuperfundBasicResearchProgram,grant P42-ES04675.Specialthanksgo to the Winchesterboat
clubfor theirinterestin ourresearch
andfor allowingusto launchour
boatsfromtheirdocksandto the Medfordboatclubfor providinga
pole for our anemometer.The authorswould also like to thank Eric
Adamsfor his insightand feedbackon thiswork and Laura DePaoli
Technol.,Cambridge,1995.
Tchobanoglous,
G., Constructed
wetlands
andaquaticplantsystems:
Research,design,operationalandmonitoringissues,
in Constructed
Wetlands
for WaterQualityImprovement,
editedby G. A. Moshiri,
A. F. Lewis, New York, 1993.
for her commentson the manuscript.
Thackston,E. L., F. D. ShieldsJr., and P. R. Schroeder,Residence
References
Wu, J., and I. K. Tsanis,Pollutanttransportand residence
time in a
timedistributions
of shallow
basins,
J. Environ.
Eng.,113(6),13191332, 1987.
Andrad6ttir,
H. 0., Circulation
andmixing
intheupper
forebay
ofthe
distorted
scalemodelandnumerical
model,J. Hydraul.
Res.,32(4),
583-598, 1994.
Mystic Lake system,Winchester,Massachusetts,
Master's thesis,
Mass.Inst. of Technol.,Cambridge,1997.
H. 0. Andrad6ttir
andH. M. Nepf,RalphM. Parsons
Laboratory,
Andrad6ttir,
H. O., andH. M. Nepf,Thermal
mediation
bylittoral Departmentof Civil andEnvironmentalEngineering,
Massachusetts
wetlandsand impacton lake intrusiondepth,WaterResour.Res., Instituteof Technology,
Cambridge,MA 02139. (hoa@mit.edu;
36(3), 725-735,2000.
(hmnepf@mit.edu)
Aurilio,A. C., R. P. Mason,andH. F. Hemond,Speciation
andfateof
arsenicin three lakes of the AberjonaWatershed,Environ.Sci.
Technol.,28, 577-585, 1994.
(ReceivedOctober25, 1999;revisedMay 26, 2000;
Bencala,K. E., andR. A. Walters,Simulation
of solutetransport
in a acceptedJune30, 2000.)
Download