Aquifers
CLAUDIO GALLO 1;2
AND GIANMARCOMANZINI
3;
1
CRS4,ZonaIndustrialeMa hiareddu,Uta,Cagliari{Italy
2
Dept.ofCivilEngineeringandGeos ien es,TUDelft,Delft,{ TheNetherlands
3
IAN-CNR,via Ferrata1,27100Pavia, {Italy
SUMMARY
Theadoptionofasuitablepumping-inje tingwellnetworkandthehumanenhan ementofthea tivity
ofsoilba teria,whosemetabolism ontributestodegradeandtransformmanypollutantsinnon-toxi
substan es,maybe ru ialinthepro essofremediationof ontaminatedsoils.Organi ontaminant
transport in a subsurfa e aquifer and its biologi al degradation kineti s is numeri ally addressed
by usinga four ontaminant spe ies model. A numeri al approa his proposed,that is based ona
ell- enter nite volume method for the system of adve tion-dispersion equations of ontaminants
with amixed-hybridniteelementmethodfor thesolutionof asingle-phaseDar y'sequation.The
ee tivenessofthemethodanditsa ura yinretainingthemainphysi alpropertiesofthe ontinuous
mathemati almodelisillustrated bysimulatingthetimeevolutionof ontaminant on entrationsin
asetofrealisti s enarios. Copyright 2000JohnWiley&Sons,Ltd.
key words: bioremediation, ontaminanttransport,mixedniteelements,nitevolumes
1. INTRODUCTION
Soil ontamination has re ently be ome a problem of major so ial on ern, be ausea wide
rangeofpollutantagentsofdierent hemi alnatureandtoxi itymaybepresentinsubsurfa e
aquifers.Pollutionsour esareeithera identalevents,likespillsandleaks,or ommonhuman
a tivities, like disposal of urban sewage, industrial wastes, and the use of pesti ides and
fertilizersinagri ulture,seeReferen es[24,25,22,20,9,8℄.The ontaminantsinasubsurfa e
aquiferaresubje tto omplexphysi aland hemi alpro esses,su hasdispersion,adve tion
bygroundwater ow, hemi alrea tionsandbiologi aldegradationduetosoilmi roorganisms.
Thegroundwater owisdes ribedbyasinglephaseDar y'sequation,whilethesubsurfa e
transport of dierent hemi al spe ies aremodeledby aset of oupled adve tion-dispersion-
rea tionequations[3℄.
The biologi al degradation depends on the mi roorganism population whose metabolism
is ae ted by the availability in soilsof substrates like organi arbon, ele tron a eptors {
oxygenandnitrogen{andnutrients[17, 19℄.
Theorgani arbonneededtosustainba teriallifeisnaturallypresentinsoil,whilenutrients
{su hasphosphates,nitrates,ammonia{maybeprovidedalsobyhumaninput,forinstan e
fertilizers. The ba terial population is normally stable be ause it dynami ally tends to an
equilibriumstateinwhi hitsgrowthrateisbalan edbyitsde ayrate.Whenthe on entration
ofnutrientsaugmentsduetoanexternalsupply, theba terialpopulationin reasesofseveral
orders of magnitude and tends to anew equilibrium state. The remarkable fa t is that the
ba terialmetaboli pro essesmayee tivelyredu ehazardousorgani pollutantstoharmless
byprodu ts,su hasCO
2 andH
2
O[1℄.Inthis ontext,aremediation strategy anbedevised
whi h relies on the enhan ement of the biodegradation a tivity, see Referen e [23℄ for a
literaturereviewonthistopi .Thebiodegradationkineti s modelsproposedinliteratureare
usually lassiedinthreedistin t lasses,respe tivelytermedfree-ba teria,mi ro olony-based
andbiolm models[16,11℄.
Thesimplestmodelsbelongtotherst lass[18℄.Theybasi allyassumethatba teriaexist
as individual parti les within the aqueous phaseor adsorbed by soil grains.No assumption
is made on the mi ros opi onguration and distribution of ba teria in soil pores, and on
the waythe organismsaregrouped togetheronthe solidporesurfa e.These latterfa ts are
onsideredirrelevantforthema ros opi des riptionofba terialpopulationgrowthandde ay.
In the se ond lass of models, ba teria do not exist as individual parti les but in small
dis rete olonies, or mi ro olonies, atta hed to the soilgrain surfa es.Growthand de ay of
the biomass ontained in mi ro olonies are formulated either by taking that the ba terial
olonydimension angrowby onsumptionoforgani substrateandele trona eptorsorby
assuming the olonydimension onstantand varying their on entration, i.e., thenumberof
oloniesperunit volume[21℄.
The main feature of the models in the third lass is that the solid parti les onstituting
theaquifermaterialare overedbyabiolm withinwhi h onsumptionofthesubstratesand
ele tron-a eptorstakespla e [26, 16,11℄. Thekeypro essesarethemassex hangebetween
bulk owandthebiolmandtheinternaldegradationoforgani substrates.
Amoredetaileddis ussionofthesimilaritiesanddieren esbetweenthesemodelsisbeyond
thes opeofthepresentwork.Werefertheinterestedreadertothedis ussioninReferen e[2℄{
seealsothebibliographytherein{whereitisshownthatunderasetofsimplifyingassumptions
the three approa hes redu e to an essentially equivalent des ription of the biodegradation
pro ess.This,however,istrueonlyforverysimple ases.
Our approa h relies on the four-spe ies model do umented in Referen es [21, 27℄ in the
ontextof the moregeneralmi ro olony-based on ept. The main feature of this model lies
in its apability of des ribing how the metabolism of subsurfa e mi robes an be enhan ed
by on urrent metabolization of oxygen, nitrogen and nutrients. From the omputational
viewpoint, it is a ompromise between the simpler free-ba teria model whi h tends to
overestimate the degradation extent, and the more a urate but also more omplex and
expensivebio-lmmodel.
Thenumeri alapproximationofthe ompletemathemati almodelisaresear hissue,and
manydierentaspe tsmakethenumeri alsimulationofabioremediationpro ess hallenging.
Wemention in thefollowinglist thetopi swefeelthe mostsigni antandthat we onsider
Treatment ofhighly heterogeneoussoil:the valueof thepermeability andierforfour
ordersofmagnitudeormorein twoadja entmesh ells.
Adve tion-dominated transport: the model spe ies on entrations in the groundwater
bulk- ow an feature strong gradient regions when sharp on entrations fronts move
throughoutthe omputationaldomain.
Non-linear ouplingee ts,evenifthesoilisasaturatedone.Thekineti softheba terial
population depends on the ontaminants whi h diuse within the mi ro olonies from
the groundwater bulk ow. It also exerts its in uen e on the bulk- ow ontaminant
on entrationsviaasetofrea tivesour etermsinthetransportequations.
These issueshavealreadybeeninvestigatedbytheauthorsin thesepreviousworks.
In[4℄wepresentedsomepreliminaryresultsonthedis retizationofthe owandtransport
equationsbyusingmixedniteelements andnite volumes.
The ouplingof the ontaminanttransport equationswithaba terialpopulationequation
and its numeri al dis retization was investigated in [12℄.In this work the model spe ies are
passivelyadve ted by a onstant velo ity and pressure elds. In order to solve the Dar y's
equation we adopted a high-order a urate mixed nite element s heme (BDM
1
). Despite
is a ura y, this approa h is not appropriate to simulate non-linear phenomena requiring a
frequentupdateofthevelo ityandpressureelds be auseofthehigh omputational ost.
Abetterapproa hfromthisviewpointisbasedonthemixed-hybrid s hemeproposedin[13℄.
Inthisworkwevalidatedthemethodonthestandardquarter-of-ve-spotsproblem,fo using
theattentiononthetreatmentofthesoilheterogeneity.
Finally,theworkpresentedinthispaper opeswiththebioremediationofaeld-sizeaquifer
thathasbeen ontaminatedbyana identalleakout.Dierentbioremediationte hniquesand
humaninterventionstrategiesarenumeri allyinvestigatedtopredi tthe lean-uptimeforan
almost ompleteremovalofpollutants.
A ordingwith ourpreviousexperien eweproposethefollowingnumeri alapproa h.The
steady groundwaterbulk owis approximatedby using thelowest-ordermixed-hybridnite
element method. This approa h yields an approximation to the steady velo ity eld that is
more a urate than the one provided by straightforward dierentiation of the onforming
nite element approximation of the pressure eld. In parti ular, we emphasize that the
mixed-hybrid nite element method ensure lo al { i.e. ell-wise { mass onservation, while
the onformingnite elementapproximationla kslo almass onservation. The ontaminant
transportequationsareapproximatedinspa ebyanunstru turedtriangle-basednitevolume
method and advan ed in time by a semi-impli it two-stage Runge-Kutta s heme. A TVD
stability onditionisimposedbyamultidimensionallimitingpro edure.Theresultings heme
is formally se ond-order a urate, onservative, and apable of apturing strong solution
gradientfrontsmovingatthe orre tphysi alpropagationspeeds.
Thespe iesintera tionsaretakenintoa ountinthefull-spe iesmodelbysolvingiteratively
theirnon-linearintera tion oupling.
Theoutlineofthepaperfollows.InSe tion2,wereviewthemathemati almodeldes ribing
single-phasebulk ow, ontaminanttransportandba terialkineti s.Thedis retizationmethod
issummarizedinSe tion3.Weaddressherethenitevolumedis retizationofthe ontaminant
transport equations aswell asthe mixed-hybridnite element approximationofthe Dar y's
phase pressure and velo ity elds. In Se tion 4 we present the results of a set of numeri al
dierentnetworksofextra tion/inje tionwells,whoserunningmodehasbeensele tedonthe
basisoftheplumelo ationandthesoilremediationstatus.The on lusionsfollowinSe tion5.
2. THEMATHEMATICALMODEL
se :mathemati al_model
2.1. TransportEquations
Transport phenomenaare mathemati allydes ribedbyasystemof N
DS
oupled adve tion-
dispersion-rea tion equations, where N
DS
is the numberof dissolved spe ies. In divergen e
form they an bewrittenas follows
R
i
C
i
t
+div(uC
i D
i (u)rC
i )=B
i
; i=1;:::;N
DS
: (1) eq:transport
The variables C
i
in equations (1) representthe bulk ow on entrationof ea h transported
spe ies;thetermsR
i
aretheretardationfa tors,whi htakeintoa ount hemi aladsorption
pro esses, the terms D
i
(u) are the velo ity-dependent dispersion tensors. The r.h.s. sour e
terms B
i
des ribe the oupling between the spe ies on entrations transported in the bulk
owand theones within mi ro olonies.Also equations(1) aresupplementedbyappropriate
boundary onditions,su h as inlet, outlet and no- ow, and initial solution statesto spe ify
theappli ationproblems.
2.2. The Dar y's Equation
Groundwater bulk ow in an heterogeneoussaturated soil is mathemati ally formulatedby
theDar y'sequation[10℄
8
<
:
u = Krp; in
divu = f; in :
(2) eq:dar y
The pressure eld is indi ated by p and the groundwater velo ity eld by u, K(x) is the
transmissivitytensor, and f(x)asour e/sinkterm. Equations(2) are ompleted by aset of
suitableboundary onditionsofNeumann/Diri hlet-type,modelizinginlet/outletandno- ow
boundary ongurations.
2.3. The Bioremediation Model
Mi ro olony-basedmodelsassumethatba teriaresideanda twithinmi ro olonies,des ribed
as a set of pat hes atta hed to soil grains [21℄. From the bulk phase, hemi al spe ies
an rea h mi ro olonies via diusive mass ex hange. Depending on the mass-transfer
oeÆ ient, on entrations within mi ro oloniesgovernthedegradationratekineti s and an
besigni antlydierentfromthosein thebulkphase.
Inthis lassofmodels,thetermB
i
ofequation(1)isexpressedintermsofadiusivemass
ex hangefrom thebulktothemi ro olonyphase,
B
i
=N
i A
(C
i
i )
; i=1;:::;N
DS
; (3) eq:B_term
where
i
isthemass-ex hange oeÆ ientbetweenbulk owandmi ro olonies,A
isthe onta t
area of one mi ro olony for the mass diusion pro ess, Æ is the thi kness of the boundary
layerbetweenbulk owandmi ro olonies,
i
isthe on entrationof the omponentiin the
mi ro oloniesandN
isthenumberofmi ro oloniesperunit volume.
Theassumptionthat thebiodegradationpro essworksessentiallyat asteady-stateregime
yieldsthefollowingform forther.h.s.termsB
i in(3):
i A
(C
i
i )
Æ
=
0;i m
N
EA
X
k =1 Y
i;k 2
4 N
i
D S
Y
j=1
j
K
j;k +
j 3
5
I k 1
b +Q
i I
i 1
b
; i=1;:::;N
DS
; (4) eq:mi ro ol
where
0;i
are the maximumrate oeÆ ients, m
is themass of ami ro olony, Y
i;k
are the
yield oeÆ ientswhi h a ount forthe stoi hiometryand eÆ ien y of degradation,K
j;k are
the half saturation onstants,andI k 1
b
are theinhibition fun tions [21℄. Inequation (4)the
symbol N
EA
denotes the number of ele tron a eptors and N i
DS
the number of dissolved
spe iesinvolvedinthedegradationofthei-thspe ies.ThetermQ
i
isnonzeroonlywhenthe
omponentiisanele trona eptor{forinstan e oxygenornitrate.Inthis ase,ittakesthe
form
Q
i
=
i
i
K
i +
i
; i=1;:::;N
EA
; (5)
where
i
is theele trona eptor oeÆ ientfor themaintainan eenergyof ba teria,andK
i
istheele trona eptorsaturation onstant.
This degradation equation states that the total amount of a ompound entering a
mi ro olony in a given interval of time is equal to the amount of spe ies that is degraded
in the same interval. The rate of degradation and onsequently the on entration within
mi ro olonies,isroughlyproportionaltothe on entrationoutsidethe olonies.ThetermsQ
i
introdu e intothe model the onsumption ofoxygendue to ba terialde omposition [21, 27℄
asarstorderde ayterm.
Ba terial kineti s is modeled by the following time-dependent dierential equation that
des ribesthemi ro olonypopulationdynami s,
1
N
(N
)
t
= ND H
X
i=1 2
4
0;i NEA
X
k =1 Y
i;k 0
N
i
D S
Y
j=1
j
K
j;k +
j 1
A 3
5
k
d
; (6) equ:ba t_growth
where k
d
is thepopulationde ay onstant,N
DH
isthe numberof dissolved hydro arbons {
organi substrates{andistheporosityofthemedium [21,27℄.
3. THENUMERICALMODEL
se :numeri al_model
3.1. The FiniteVolumeDis retizationof theTransport Equations
The numeri aldis retization in the framework of thenite volume s heme is dened onthe
samemeshT
h
()usedforthemixed-hybrids hemeofthepreviuousse tion.Theindexhisthe
maximumdiameter oftheN
T
trianglesformingthemesh, i.e.h=max
T2T
h ()
h
T
, whereh
T
isthelenghtofthelongestedgeofthetriangleT.Asusual,these triangulationsareassumed
Equations(1)arereformulatedina ell-wiseintegralformbyintegratingonea htriangular
ell T and then applying the Gauss divergen e theorem to transform the spatial divergen e
term into abalan e ofedge integral uxes. Letus introdu efor everyT 2T
h
() theve tor
U
T
,whoseelementsarethe ell-averaged on entrationsofthetransportedspe ies,
U
T j
i
= 1
jTj Z
T C
i
dT: (7)
Thesemi-dis retenitevolumeapproximationis
jTjR dU
T
dt +
X
e2(T) G
e (u
e
; e
U
T
; e
U
Te
;n
e )+
X
e2(T) H
e (u
e
; e
U
T
; e
U
Te
;n
e )
+ X
e2
0
(T) F
(b )
e
= X
q
!
T;q S
T (
e
U
T (x
T;q
)); foreveryT2T
h ();
(8) eq:FV
where the diagonalmatrixR=diag(R
1
;:::;R
N
D S
) olle tstheretardation fa tors, andfor
every ellT,
- jTjisthemeasureofitsarea,andT itsboundary;
- (T)is thesubset of itsinternal edges; these latters arethe edgesthat T shares with
an adja ent mesh ell indi ated by T
e
, so that for every e 2 (T) there exists a ell
T
e 2T
h
()su hthate=T\T
e
;
- 0
(T) is the subset of the edges of T lo ated at the boundary of the omputational
domain;thatis, foreverye2 0
(T)wehavee=T\.
The ellinterfa e uxintegralisevaluatedbyusingsuitableadve tiveanddispersivenumeri al
uxesa rosstheedgee,that are
G
e (u
e
; e
U
T
; e
U
Te
;n
e )j
i
Z
e
nuUj
i
dl; (9) equ:numeri al_fluxes
H
e (u
e
; e
U
T
; e
U
Te
;n
e )
Z
e nDj
i
(u)rU j
i
dl; (10)
andthenumeri al uxfun tion F (b )
e
atboundaryedges.
The numeri al ux ve torfun tions G
e andH
e
introdu ed in (9)depend on u
e
, whi h is
thevalueofthevelo ityelduatthemidpointoftheedgeesharedbythetrianglesT andT
e ,
and onn
e
, whi h isthenormalto eorientedoutwardfrom T andinwardintoT
e
.Theyalso
dependon e
U
T and
e
U
T
e
,whi harethepie ewisepolynomialrepresentationsofthesolutionin
T andT
e
.Thisfun tionaldependen eimpliestheusageofpointwisevaluesoftheapproximate
solution at quadraturenodes on e.These values arere onstru ted from the ellaveragesby
an interpolation pro edure at ea h time step and a multidimensional slope limiter must be
onsideredtotakeunder ontrolthenumeri alos illations,seetheappendixofReferen e[12℄.
Theintegraladve tivetermG
e
isdis retizedbyastandardupwind uxsplittingapproa h,
whiletheintegraldispersiontermH
e
,whi hinvolvesse ondderivativesinspa e,bya entral
dierentiationalgorithm.Furtherdetailsaboutthederivationandthea ura yofthismethod
Thenumeri al uxfun tion F (b )
e
at theboundaryedgee=T\dependsonthetra e
e
U
T j
e
of there onstru ted solution e
U
T
within theunique boundary triangleT, andin some
suitableformonasetofexternal dataU (b )
e .
Theintegralsour etermS
T (
e
U
T (x
T;q
))isapproximatedbyasurfa equadraturerulewith
nodesfx
T;q
gwithin thetriangleT andweightsf!
T;q g.
Thetime-mar hings hemeisobtainedbyapproximatingthetimederivativeofU
T
{whi h
appearsintherstterminthesemi-dis reteformulation(8){byrst-ordernitedieren es
dU
T (t)
dt
t=t n
U
n+1
T U
n
T
t
; (11)
where U n+1
T
and U n
T
are the ell-averaged solutions in T at times t n+1
and t n
, and t =
t n+1
t n
. This yields a full dis rete semi-impli it s heme where resultingsymmetri linear
algebrai problem is solvedby astandardKrylovsolver,su hasa pre onditioned onjugate
gradient method. Higher (se ond) order a ura y in time is attainable by a semi-impli it
Runge-Kuttamethod, builtbytwodistin tstagesof thesameform[12℄.
3.2. The Mixed-HybridDis retization ofthe Dar y's Equation
The oupledsystemofequations(2)intheunknownspanduisdis retizedbyamixed-hybrid
nite elementapproa h. Fora detailed exposition of mixed and mixed-hybridnite element
methods we refer the readerto Referen es [5, 6℄, while for the des riptionof thenumeri al
formulationadoptedin thisworkwerefertoReferen e[12℄.
In the mixed-hybrid nite element method adopted in the present work we approximate
thevelo ityeldbyusingthelowest-orderRT
0
dis ontinuouselements,whi his omposedby
two-dimensionalfun tionswhoserestri tiontoanymeshtriangleT isoftheform
uj
T
T
x
y
+
T
T
; (12)
where thereal s alar oeÆ ients
T ,
T and
T
depend onthetriangleT.Thepressureeld
is approximatedby the triangle-basedpie ewise onstant fun tions while the pressure tra e
overea h ell-interfa ebytheedge-basedpie ewise onstantones.
With respe t to Referen e [12℄, thepresentwork diers substantially in the hoi e of the
dis rete fun tional spa eused fortheapproximationofthe velo ityeld u. Weuseherethe
lowest-orderRT
0
dis ontinuouselementsinsteadoftheBDM
1
onesofReferen e[12℄,wherea
fulllineardependen eontheposition is onsidered.Noti ealsothatthe ontinuity ondition
ofthenormal omponentofthevelo ity uxisrelaxed,andaweaker onditionisimposedby
asetofsuitableLagrangemultiplierswhi happroximatesthepressuretra es.
Weexperien edin fa t that RT
0
elementsoersasatisfa torya ura ylevelataredu ed
omputational ostwithrespe ttoBDM
1
elements,seeReferen e[13℄.Theselatteronesare
formallymorea uratebutalsosigni antlymoredemandingfroma omputationalviewpoint
be ausetheyinvolvetwi ethenumberofunknownstobestoredand al ulated.
Thisissueisparti ularlyimportantbe auseinthisworkthepressureandthevelo ityelds
are iterativelyupdatedat ea h time step,see Referen e [11℄, while in thework des ribed in
Referen e [12℄ they were al ulatedonly on e at the beginning of ea h simulation and then
3.3. Rea tionSour eTermsandMi robial Population Equation
Therea tiontermsdes ribedinequations(6)are omputedbysolvingasetofnodewisenon-
linear systems via aNewton iterative method with fra tional multistep integration s heme,
see Referen e[12℄.
4. NUMERICALEXPERIMENTS
se :experiment
In this se tion we illustrate the performan e of the proposed mathemati al and numeri al
model in predi ting the ee tiveness of a human remediation intervention to redu e the
ontaminant on entrationof a pollutedaquifer. Theaquifer is hara terized by a onstant
porosity = 0:3 and a heterogeneous isotropi transmissivity, whose prin ipal values are
assumed to be onstant on ea h triangle of the omputational mesh, and dier triangle by
triangleintherangebetween10 5
and1m 2
/day,ina ordwithanequiprobabilitysto hasti
distribution.
The rst test ase that wepresentin this paper onsists in the initial soil ontamination
phaseand islabeledby T1.Thesoil ontaminationisdue tothe leakageofCy lo-Aromati -
Hydro arbons, CAH, whi h forms a plume transported by the groundwater ow eld and
spreadinthesaturatedaquifer.
Thenextthreetest ases,labeledbyT2,T3andT4,des ribesthreepossibleinterventions
fortheremediationphase.Basi ally,we onsideranetworkofpumpingwellsthatextra tthe
polluted water and onveyit to a treatment plant, where the ontaminant is removed. The
puriedwatermaybeenri hedinoxygenandnutrientstostimulatesoilba terialgrowthand
isthen re-inje tedin theaquiferviaanetworkofinje tion wells.Figure1sket hesthewater
treatmentpro edure.
All ofthewells anbesele tivelyusedeitherin inje tionorin extra tionmodeandareall
supposed to be onne tedvia pipelines to the watertreatment plant. Point \A" is also the
lo ationoftheleakingtankwhen ontaminationo urs.Theproperpositionand onguration
ofthewellshasbeen hosenbyinvestigatingtheir apabilityofinter eptingthe ontaminated
plumetransportedbythegroundwater owinasetofpreliminarysimulations.
These simulations are based on the four spe ies model proposed by Molz et al. [21℄ and
Widdowson et al. [27℄, and des ribed in [12℄. For the sake of ompleteness, we report the
modelinthenalappendix,givingalsothevaluesoftheparametersusedin thesimulations.
The ontaminantCAHistheorgani substrateSofthemodel,whiletheotherspe iesinvolved
arethedissolvedoxygeninthesoil,O,some ompounds hemi allybasedonnitrates,N,and
some ammonia-based ompounds whi h onstitute a generi nutrient supply A. The initial
on entrationsofthesespe iesfollowarandomdistribution,withvaluesintherangesreported
in TableI.
Figure 2depi tstheben hmark aseand theposition of thewellsonthe aquifer|labels
\A" through \M" | in the remediation phase. A onstant gradient of p=x = 0:04 is
superimposedonthesubsurfa ebulk oweldintheaquifer,whi histhusorientedalongthe
TableI.Mi ro olony on entrationranges(ing=m 3
)atthebeginningofthesimulation(t=0).
tab:initial_values
S
(t=0)
O
(t=0)
N
=
A (t=0)
(g/m 3
) (g/m 3
) (g/m
3
)
[0:1;1℄ [0:1;1℄ 1000
TableII.Simulationrunparameters
tab:simulation_parameters
RunLabel
Parameter
T1 T2 T3 T4
t(days) 0.5 0.2 0.05 0.05
T
max
(days) 410 2800 500 700
wells Water
Treatment
Water Enricher O 2
Recycled water Contaminated
water
Nutrients waste
Contaminant
From pumping wells
To injection
Figure 1.Te hni al s hemeofthe remediationplant installedasideofthe wellnetworkfor polluted
watertreatmentand leanwaterenri hment.
fig:plant
ondition,withanhydrauli pressureheadgivenasafun tion ofx.Forea h simulationrun,
wereportin TableII thetimesteptandthenal timeT
max
atwhi hthesimulationends
up.
During the initial pollution phase (T1), the plume of CAH spreads with an irregular,
or \ngered", front be ause of the sto hasti soil heterogeneity whi h establishes several
preferentialpaths. Figure 3 illustrates the situation at the intermediate time t = 410 days,
River
B
C
D
E G
H I F L
M A
Groundwater flow
Figure2.Planarsket hofthewelllo ations.
fig:wells
of the onning river.A steadystate solutionisrea hed at t=600days,shownin gure 3,
when the ontaminantplume doesnot spreadfurther and thetransport of the ontaminant
alongsomepreferentialpathsisthusestablished.
Three dierent possible interventionstrategies havebeen investigated. Remediation starts
after t = 410 days sin e the pollution started, that is before the ontaminant rea hes the
river onningwiththeaquifer.Wesupposethat the ontaminantsour eisremovedandthe
lean-up of the soilis performedby using anetwork of extra tion/inje tionwells. Forall of
thesimulationswewillshowthespatialdistributionoftheorgani substrate,CAH,andofthe
dissolvedoxygen,DO{thespe iesOin ourfourspe iesmodel.
InthesimulationT2we onsidertheso- alledpump-and-treatmethod.ThewellsGthrough
M of the pipeline network are a tive and extra t the ontaminated water, whi h is then
onveyedtowatertreatmentplants.
The owboundary onditionsare thesameasforthe soil ontamination phase,ex ept at
the lo ation orrespondingto the pumping wells: here,ade rease of 0.5 m in pressurehead
withrespe ttothenaturalgradient onditionisimposedinorder tomodifythe owpattern
for ontaminant re overy. As in the soil ontamination phase, just one hemi al spe ies is
onsidered inthesimulation,bynegle tingtheee tsoftheotherspe ies.
The resultof the simulation is shown in gure 5at the intermediate time t =1490 days,
that is about4yearsafter theremovalof the ontaminantsour e. Heterogeneitystill ae ts
The omputersimulationisterminated att=2500days{about7years.Itisworthnoti ing
that theremovaloftheorgani ontaminantplumehasnotyet beenfully ompleted.
The bioremediation interventions onsidered in this paper{ simulationsT3 and T4{are
essentiallybasedon thestimulation ofthe growthof thesubsurfa e ba terialpopulationby
in reasing the on entrationofoxygenand nutrientsdissolvedin soil. Weassumethat these
hemi al substan es be dire tly supplied via inje tion of \enri hed water" into the aquifer.
Part of the wells are, thus, used for extra ting the ontaminated water to be onveyed to
treatmentplants.Partoralloftheremainingwellsareusedforre-inje tingwaterwith hemi al
additivesinto thesoil. Theee tiveness of thisstrategy stronglydependson theextra tion-
inje tionoperationalmode hosenin thewell ongurationpattern, whi hisdierentforthe
twosimulationsT3andT4.
Inthe ongurationofthesimulationT3thewellsA throughD work ininje tion mode {
anoverpressureof0:5misimposed.Nutrientsandnitratesaredeliveredinex ess,sin etheir
solubility in water is mu h largerthan the one of oxygen. This latter hemi al is kept at a
onstant on entrationof20mg/l.Theremainingwellsworkinextra tionmode,atthesame
pressure onditionofthe\pump-and-treat"method.
The result of this simulation is shown at the intermediate time t = 178 days after the
removalofthe ontaminantstarted.Nosigni antimprovementsin ontaminantsremovalare
a hievedfromthis ongurationof pumpingandinje tingwellsaftert=360days.A \dead-
zone"developsin thetriangularregiondened bythewells\F-G-H",allworkinginpumping
onditions.Itisevidentby omparinggures8and6thedieren einthe ontaminantremoval
betweenthe urrentsimulationandthepre edingone,parti ularlywheredissolvedoxygenhas
beendelivered.
Inthe ongurationofthesimulationT4wellsGthroughMstillworkin extra tionmode,
wellsEandFinje tnutrientsandoxygen,whilewellsAthroughDaredismissed.gure9shows
the ontaminant and dissolved oxygen distributions at the intermediate time t = 600 days.
Weremarkthatthemajorpartoftheresidual ontaminantmassintheaquiferisremoved.A
ompleteremovalofthe ontaminantisa hievedatthenaltimet=875days,asillustrated
byFigure9.
Finally, gure 11 reports the residual ontaminant mass in the aquifer as a de reasing
fun tion oftimeandsummarizestheperforman eofthedierentremediationapproa hesT2,
T3andT4.Thisgureemphasizeshowthebioremediationstrategy anbemoreee tivethan
thesimplepump-and-treatmethod.AlthoughbothinterventionstrategiesT2andT4a hieves
analmost ompleteremovalofthe ontaminant,theremediationtime isverydierentinthe
two ases.Nevertheless,bioremediation analsobesensitivetothewell ongurationnetwork
hosenfortheintervention,asshownbytheperforman e urveT3.
4.1. ModelPerforman es
Inthisse tionwereportsomeinformationaboutthe ostsintermofCPUtimeofthe omputer
simulations.
All the simulations des ribed in the previous se tion were run on a omputational mesh
omposed of about 5000 triangles using an IBM RISC 6000/390 ma hine. The simulations
involvingthefullfour-spe iesmodelareveryexpensiveandtakeabout12hours,whilesingle-
Table 2:CPU Costs(minutes)
RunLabel
Parameter
T1 T2 T3 T4
CPU 80 90 720 720
The CPU ost is quite high in the former ase be ause all of the non-linear intera tions
among the dierent spe ies must be taken into a ount. A 44 non-linear systems must
be solved for ea h ell at ea h time step, by using a Newton iterative s heme, whi h takes
approximately40% of the total CPU osts. Moreover,the nite volume method requires a
limitedpie ewise-linearre onstru tionofea hunknowneld toensure2 nd
-order a ura yin
spa e.Thelimitingpro edureisneededtoensuremonotoni ityofre onstru tedgradientsand
to preservenon-linear stability,see Referen e [12℄ fordetails.The omputational ost ofthe
re onstru tion pro edure is also signi ant, being about35% of the total CPU osts of the
simulation.Thisin rementisevidentwhenthefullspe iesmodelis onsideredinsteadofthe
single-spe iesone.
CPU ostsare alsoin uen edbythewaythesimulationisrun.Forinstan e,in theinitial
pollutantphaseT1,the ontaminantistransportedbyasteadygroundwater oweld,whi h
is omputedonlyon e at thebeginningof therun.Instead,thesimulationT2stills involves
a single-spe ies model, but makes usage of a transient groundwater velo ity eld, whi h is
updatedevery50transport steps,thusresultinginamoreexpensive omputation.
5. CONCLUSIONS
In this work we illustrated a numeri al model to investigate the bioremediation pro ess
in heterogeneous saturated aquifers and its appli ations in devising dierent intervention
strategiesonaeld-sizes enarios.Themethodisparti ularlysuitableindete tingdead-zones
due to theheterogeneityof themedium anddependentonthe welllo ationand operational
mode.
Ourapproa hisbasedonthedis retizationofamultispe iestransportmodel oupledwitha
ba terialdegradationkineti sofMonodtype.Themi ro olonydes riptionofba teriala tivity
is onsidered.Thebulk owvelo ityisapproximatedbyamixed-hybridniteelementmethod
whilethespe iestransportequationsaredis retizedbyusingasemi-impli it ell- enternite
volumes heme.
The performan e of the method are assessed by studying the ontamination pro ess and
severalremediationstrategiesonarealisti subsurfa es enario.
A omparison of the numeri al experiments reported in this work learly illustrates the
advantageofa ombinedbiologi al-hydrauli interventionwithrespe ttothesimplehydrauli
oneinthe aseofasto hasti allyheterogeneoussoil.Theremediationtimeintheformer ase
is shown to be about half the oneof the latter ase. This fa t implies that the operational
ostsmaybesubstantiallyredu ed.
When theaquifer is stronglyheterogeneous, preferential owzones mayappear and large
quantities of ontaminant may remain isolated if the simplest pump-and-treat remediation
presentwhen remediation is enhan ed byba terial a tivity. In su h a ase, an optimal well
ongurationhasadramati impa tontheee tivenessofthehumanintervention.Forthese
reasons,itis evidentthat abetterunderstandingofhowandwheretrappingzonesappearis
riti alin devisinganee tiveremediationstrategy.
Inorder tostudy thenear-sour e ontamination zone,that isthethezone surroundingan
organi ontaminantspill,amultiphasemodelisneeded,be auseanorgani phaseappears.It
is informativeto saythatsomepreliminarywork[14, 11℄hasbeenperformedbytheauthors
todevelopasuitablenumeri alapproa htomultiphasesimulationsaswellas onsideringthe
problem of pore- loggingin biolm models. However,these topi swill betheissue of future
work.
ACKNOWLEDGEMENT
The work of Claudio Gallo has been nan ially supported by Sardinian Regional Authorities.
The authors would like to thank Dr. Fabio Bettio for his help in visualization and
Dr. Enri o Bertolazzi (University of Trento, Italy) for his areful reading of the preliminary
version of the paper and his useful suggestions. The unstru tured Delaunay grids were
generated by the mesh generator Triangle, a ode implemented by Shew hu k, see the
URL:http://almond.srv. s. mu.edu /afs / s/ proje t/q uake/ publ i /w ww/tr iang le.ht ml.
REFERENCES
Bailey:1977 1. J.E.BaileyandD.F.Ollis. Bio hemi alEngineeringFundamentals. M Graw-Hill,NewYork,1977.
Baveye:1989 2. P.BaveyeandA.Valo hi. Anevaluationofmathemati almodelsofthetransportbiologi allyrea ting
solutesinsaturatedsoilsandaquifers. WaterResour.Res.,25(6):1413{1421,1989.
Bear:1979 3. J.Bear. Hydrauli sofGroundwater. M Graw-Hill,NewYork,1979.
Bergamas hi:1995 4. L.Bergamas hi,C.Gallo,G.Manzini,C.Pani oni,and M.Putti. Amixednite-elements/TVDnite-
volumess hemeforsaturated owandtransportingroundwater. InCe hietal.,editor,FiniteElements
inFluids,pages1223{1232,Padova,Italy,1995.
Brezzi:1991 5. F.BrezziandM.Fortin. MixedandHybridFiniteElementMethods. SpringerVerlag,Berlin,1991.
Chavent:1991 6. G.ChaventandJ.E.Roberts. AUniedPhysi alPresentationofMixed,Mixed-HybridFiniteElements
andStandardFiniteDieren eApproximationfortheDeterminationofVelo itiesinWater owProblems.
Advan esinWaterResour es,14(6):329{348,1991.
Ciarlet:1980 7. P. G.Ciarlet. The niteelement method for ellipti problems. North-Holland Publishing Company,
Amsterdam,Holland,1980.
epa97heat 8. E. L. Davis. How heat an enhan e in-situ and aquifer remediation: important hemi al properties
and guidan e on hoosing the appropriate te hnique. Te hni al Report EPA/540/S-97/502, USEPA,
Te hnologyInnovationOÆ e,OÆ eofSolidWasteandEmergen yResponse,USEPA,Washington,DC,
1997.
epa96state 9. EnvironmentalProte tionAgen y.Statepoli ies on erningtheuseofinje tantsforinsitugroundwater
remediation. Te hni alReportEPA/542/S-96/001,USEPA,Te hnologyInnovationOÆ e,1996.
Freeze:1979 10. R.A.FreezeandJ.A.Cherry. Groundwater. Prenti e-Hall,NewJersey,1979.
Gallo:2000 11. C.GalloandS.M.Hassanizadeh. In uen eofbiodegradationonnapl owanddissolutioningroundwater.
InL.R.Bentleyetal.,editor,ComputationalMethodsinWaterResour esXIII,Volume1,pages129{136.
A.A.Balkema,Rotterdam,Holland,2000.
Gallo:1997 12. C. Galloand G.Manzini. Amixed nite element/nite volume approa h for solvingbiodegradation
transportingroundwater. InternationalJournal ofNumeri alMethodsinFluids,26:533{556,1998.
Gallo:1998 13. C.Galloand G.Manzini. 2-dnumeri almodelingof bioremediationinheterogeneous saturated soils.
TransportinPorousMedia,31:67{88,1998b.
Gallo-Manzini:2000 14. C.GalloandG.Manzini. Afully ouplednumeri almodelfortwo-phase owwith ontaminanttransport
and biodegradation kineti s. Te hni al report,IAN-1163,1999. to appear inComm.Int.Num.Meth.
Gallo:2001:IAN:te hni al:rep ort 15. C. Gallo and G. Manzini. Finite volume/mixed nite element analysis of pollutant transport and
bioremediationinheterogeneoussaturatedaquifers. Te hni alreport,IAN-Te h.Rep.,2001.
Hassanizadeh:1999 16. S.M.Hassanizadeh. Ups alingequationsofsolutetransportandbiodegradationinsoils. Te hni alreport,
Dept.ofCivilEngineeringandGeos ien es,TUDelft,TheNetherlands,1999.
Kelly:1996 17. W.R.Kelly,G.M. Hornberger,J. S.Herman,and A.L.Mills. Kineti s ofBTXbiodegradation and
mineralizationinbat hand olumnsystems. J.Contam.Hydrol.,23:113{132,1996.
Kindred:1989 18. J.S.KindredandM.A.Celia. Contaminanttransportandbiodegradation:2. on eptualmodelandtest
simulations. WaterResour esResear h,25(6):1149{1159,1989.
Lensing:1994 19. J.J.Lensing,M.Vogt,andB.Herrling. Modelingbiologi allymediatedredoxpro essesinthesubsurfa e.
J.Hydrol.,159:125{143,1994.
epa95material 20. D.R.M Caulou,D.G.Jewett,andS.G.Huling. Nonaqueousphaseliquids ompatibilitywithmaterials
used inwell nstru tion,sampling, and remediation. Te hni al Report EPA/540/S-95/503, US EPA,
Te hnologyInnovationOÆ e,OÆ eofSolidWasteandEmergen yResponse,USEPA,Washington,DC,
1995.
Moltz:1986 21. F.J.Molz,M.A.Widdowson,andL.D.Beneeld. Simulationofmi robialgrowthdynami s oupledto
nutrientandoxygentransportinporousmedia. WaterResour esResear h,22(8):1207{1216,1986.
insitubio:1993 22. NationalResear h Coun il. In Situ Bioremediation: When Does It Work? NationalA ademyPress,
Washington,D.C.,1993.
leanup:1994 23. National Resear h Coun il. Alternatives for Ground Water Cleanup. National A ademy Press,
Washington,D.C.,1994.
epa92gen 24. R. R. Ross. General methods for remedial operations performan e evaluations. Te hni al Report
EPA/600/R-92/002, US EPA, Te hnology Innovation OÆ e, OÆ e of Solid Waste and Emergen y
Response,USEPA,Washington,DC,1992.
epa92t e 25. H.H.Russel,J.E.Matthews,andG.W.Sewell. TCEremovalfrom ontaminatedsoilandgroundwater.
Te hni alReportEPA/540/S-92/002,USEPA,Te hnologyInnovationOÆ e,OÆ eofSolidWasteand
Emergen yResponse,USEPA,Washington,DC,1992.
Taylor:1990 26. S.W.TaylorandP.R.Jae. Substrateandbiomasstransportinaporousmedium. WaterResour.Res.,
26(9):2181{2194,1990.
Widdowson:1988 27. M.A.Widdowson,F.J.Molz,andL.D.Beneeld. Anumeri altransportmodelforoxygen-andnitrate-
basedrespirationlinkedtosubstrateandnutrientavailabilityinporousmedia. WaterResour esResear h,
24(9):1553{1565,1988.
Figure3.Contaminationphase(T1):pollutant on entrationatt=410days
s enario1a
Figure4.Contaminationphase(T1):pollutant on entrationatt=600days|steadystate s enario1b
Figure5.Pump-and-treatremediation(T2): on entrationatt=1450 days PAT1a
Figure6.Pump-and-treatremediation(T2): on entrationatt=2500 days PAT1b
Figure7.Bioremediationby1 st
well onguration(T3): on entrationatt=178days bio1a
Figure8.Bioremediationby1 st
well onguration(T3): on entrationatt=360days bio1b
Figure9.Bioremediationby2 nd
well onguration(T4): on entrationatt=600days(200daysafter
theswit h)
bio2a
Figure 10. Bioremediationby2 nd
well onguration (T4): on entration at t=875days (475days
aftertheswit h)
bio2b
Figure11. Contaminantresidualmassremovalvsremediationtime. ompeff
APPENDIX
Thekineti degradationratesof on entrationswithinmi ro oloniesin(3)forthefourspe iesmodel
usedinallofthesimulationsare
SA (C
S
S )
Æ
= m YS;O0;O
S
K
S;O +
S
O
K
O +
O
A
K
A;O +
A
+ m
Y
S;N
0;N
S
K
S;N +
S
N
K
N +
N
A
K
A;N +
A
I 1
b
(13) S_ olony
O A
(CO O)
Æ
= m
Y
O
0;O
S
KS;O+ S
O
KO+ O
A
KA;O+ A
+ Okd;O
O
K
O 0+ O
; (14) O_ olony
NA (C
N
N )
Æ
= m YN0;N
S
K
S;N +
S
N
K
N +
N
A
K
A;N +
A
I 1
b
;
+
N k
d;N
N
K
N 0+
N
I 1
b
; (15) N_ olony
AA
(CA A)
Æ
= m YA;O0;O
S
KS;O+ S
O
KO+ O
A
KA;O+ A
;
+ m YA;N0;N
S
KS;N+ S
N
KN+ N
A
KA;N + A
I 1
b
; (16) A_ olony
andthemi robialgrowth/de ayequationis
1
N
N
t
=
0;O
S
KS;O+ S
O
KO+ O
A
KA;O+ A
kd;O
+
0;N
S
K
S;N +
S
N
K
N +
N
A
K
A;N +
A
k
d;N
I 1
b
(17) eq:ba t_growth
where
-
S
=1:0310 5
[m 2
/day℄,
O
=2:1910 5
[m 2
/day℄,
N
=1:5010 5
[m 2
/day℄,
A
=1:8610 5
[m 2
/day℄ are the mass ex hange oeÆ ients for the bulk ow and the mi ro olony spe ies
on entrations;
- A =3:76810 10
[m 2
℄,isthe onta tareaofthemi ro olonyforthemassdiusionpro ess;
- Æ=5:010 4
[m℄isthethi knessoftheboundarylayerbetweenbulk owandmi ro olonies;
- m =2:8610 11
isthemi ro olonymass;
- 0;O=4:34[1/day℄and0;N =2:9[1/day℄arethespe i aerobi andanaerobi growthrates;
- YS;O=0:278andYS;N =0:5aretheheterotrophi yield oeÆ ients;
- YO =0:278 [{℄ and YN =0:5 [{℄ are the oeÆ ients for theoxygenand nitrogensynthesisof
heterotrophi biomass;
- 0=0:0402[{℄ andN =0:1[{℄ aretheoxygenandnitrogenuse- oeÆ ientsformaintenan e
energyofba teria;
- YA;O =0:122 andYA;N =0:122 are the ammonia-nitrogen oeÆ ients for produ ingbiomass
underaerobi andanaerobi onditions;
- K
S;O
=40 [g/m 3
℄,K
O
= 0:77 [g/m 3
℄, and K
A;O
= 1[g/m 3
℄are the substrate, oxygen, and
ammonia-nitrogensaturation onstantsunderaerobi onditions;
- K
S;N
=40 [g/m 3
℄,K
N
=2:6 [g/m 3
℄,andK
A;N
=1[g/m 3
℄are the substrate,nitrogen, and
ammonia-nitrogensaturation onstantsunderanaerobi onditions;
- K
O
0 =0:77 [g/m 3
℄andK
N
0=2:6[g/m 3
℄aretheoxygenandnitrogensaturation onstants;
- I 0
b
=1and I 1
b
=K
b;N
=(K
b;N +
O
) are theinhibition fun tionsof the oxygen-basedand the
nitrogen-basedrespiration,andK =0:0001[g/m 3
℄istheinhibition oeÆ ient;
- k
d;O
= 0:02 [1/day℄ and k
d;N
= 0:02 [1/day℄ are the ba terial death-per-unit-time de ay
onstantsforaerobi andanaerobi metabolism.