nitrogen transformation and transport modeling in ...clemept/publsihed_pdf/...ecological modelling...

Ecological Modelling 192 (2006) 143–159

Nitrogen transformation and transport modelingin groundwater aquifers

Mee-Sun Leea, Kang-Kun Leea,∗, Yunjung Hyuna,T. Prabhakar Clementb, David Hamiltonc

a School of Earth and Environmental Science, Seoul National University, Seoul 151-742, Republic of Koreab Department of Civil Engineering, Auburn University, AL 36830, USA

c Department of Biological Sciences, University of Waikato, Private Bag 3105, New Zealand

Received 13 September 2004; received in revised form 29 June 2005; accepted 6 July 2005Available online 19 September 2005

Abstract

Nitrogen pollution in urban and rural groundwater is a common problem and it poses a major threat to groundwater-baseddrinking water supplies. In this study, a kinetic model is developed to study nitrification–denitrification reactions in groundwateraquifers. A new reaction module for the reactive transport in three-dimensions (RT3D) code is developed and tested to describethe fate and transport of nitrogen species, dissolved oxygen (DO), dissolved organic carbon (DOC) and biomass. The proposedmodel is verified against some published numerical results and analytical solutions. The model is later used to study the field-scalenitrogen transformations at a cattle feedlot site within the Vasse Research Station, located south of Busselton in Western Australia.M eactionsi©

K

1

fots

nnttheandza-ngesncethe

n-alth

0

odeling results compare favorably with the field data. The developed model describing nitrification and denitrification rs a useful framework for simulating the fate and transport of nitrogen species in groundwater aquifers.

2005 Elsevier B.V. All rights reserved.

eywords: Nitrogen model; Transformation; Transport; Nitrification; Denitrification; RT3D

. Introduction

Nitrogen, originating from agricultural sites, animaleedlots, septic tanks and other waste disposal sites, isne of the most common contaminants in groundwa-

er. In soil-groundwater systems, nitrogen species con-ist of ammonium–nitrogen (NH4–N), nitrite–nitrogen

∗ Corresponding author. Tel.: +82 2 880 8161; fax: +82 2 871 3269.E-mail address: [email protected] (K.-K. Lee).

(NO2–N), nitrate–nitrogen (NO3–N), organic nitrogeand nitrogen gas (N2). The predominant form preseis determined by the environmental conditions ofwater body particularly pH, temperature, oxygenmicroorganism activity coupled with the mineralition rates of labile organic nitrogen. Seasonal chacan also be a key control of the speciation balaregardless of the total nitrogen concentration ofwater body (Burt et al., 1993). Excessive nitrate cocentration in groundwater is a significant public he

304-3800/$ – see front matter © 2005 Elsevier B.V. All rights reserved.doi:10.1016/j.ecolmodel.2005.07.013

144 M.-S. Lee et al. / Ecological Modelling 192 (2006) 143–159

problem. The threat of methemoglobinemia causingsevere oxygen deprivation, especially in children, iswell documented and is also reflected in the U.S. drink-ing water standard of 10 mg/l as nitrate–nitrogen (EPA,1980). There have been many studies on the adversehealth effects of nitrate and nitrite in drinking water(Bosch et al., 1950; Shuval and Gruener, 1977; NAS,1978; White, 1983; Dorsch et al., 1984). There are otherenvironmental consequences that arise due to excessnitrogen being leached into the groundwater. Excessnitrate can contribute to eutrophication and can also betoxic to some aquatic organisms.

In many cases, the study of nitrogen transportis complicated by the presence of various nitrogenspecies and the transformations that can occur in thesaturated zone due to ambient microbial processes.The objectives of this paper are to develop a nitro-gen transport and transformation model for saturatedgroundwater systems and assess its performance byapplying it to a field site contaminated by nitrogen.Details of the mathematical model that describes thenitrification–denitrification coupled processes are pre-sented. The model is coded as a reaction modulewithin the public-domain reactive transport code reac-tive transport in three-dimensions (RT3D;Clementet al., 1998).

In order to verify the proposed model, data regard-ing concentrations of ammonia, nitrate and dissolvedoxygen (DO) transported in saturated porous mediacoupled with nitrification and denitrification processesare taken from the literature (Widdowson et al., 1988;Kindred and Celia, 1989) and are used in the valida-tion step. Further, model calibration and validations areperformed to reproduce the contamination distributionpatterns observed at a field site in Australia.

2. Conceptual model: nitrogen transformations

Fig. 1outlines the major concepts of nitrogen trans-formations associated with the soil-groundwater envi-ronment in the vicinity of a feedlot. The bulk of thenitrogen in wastewater at the feedlot is in the form ofaqueous ammonium (NH4+). Most sediment and soilcolloid surfaces are negatively charged, giving themthe ability to act as cation exchangers. The ammoniumanion can therefore be immobilized geochemically byadsorption to aquifer sediments. Otherwise, ammoniamay be rapidly oxidized to nitrite (NO2−) in the pres-ence of oxygen by the autotrophic ammonia-oxidizingbacteria (the major genera isNitrosomonas). This is thefirst step in the process of nitrification, which is a two-

nitrog
Fig. 1. Conceptual model for en cycle in soil and groundwater.

M.-S. Lee et al. / Ecological Modelling 192 (2006) 143–159 145

stage oxidation process mediated by the autotrophicammonia-oxidizing bacteria (the nitrosifiers). In thesecond step, the autotrophic nitrite-oxidizing bacteria(the true nitrifying bacteria) oxidize nitrite to nitrate.The ammonia-oxidizing bacteria are effective in con-verting ammonium to nitrite. Nitrite is a rather unsta-ble nitrogen species, which will generally be reducedor oxidized. The nitrite-oxidizing bacteria (the majorgenera isNitrobacter) then oxidize nitrite to nitrate(NO3

−). Nitrate is quite soluble in water and is not sig-nificantly adsorbed by clay-rich soils because it is ananion. Nitrate represents the stable end product of thenitrification process (Behnke, 1975). Usually, nitrifica-tion occurs mostly in the aerobic unsaturated zone. It iscommon knowledge that nitrification is an oxic process.However, recent investigations have shown that at leastammonia oxidizers are able to oxidize ammonia underanoxic conditions (Schmidt and Bock, 1997, 1998).Moreover, a new planctomycete has been described thatis capable of anoxic ammonia oxidation by a mech-anism not yet fully understood (Strous et al., 1999).DeSimone and Howes (1998)suggested that nitrifica-tion in the suboxic saturated zone, along with oxidationof the residual organic carbon, may have buffered oxy-gen in the plume.

Below the groundwater table, nitrate is reduced tonitrogen gas by denitrification. This process occursprimarily in anoxic conditions and involves het-erotrophic bacteria (the major genera isParacoccus,Pseudomonas). Additionally, once the concentrationo tura-t one( nt oxicd l.(

3

ict-i lvedo nd-w tudyu ort ins olvest riber o-

bile species in three-dimensional saturated ground-water systems (Clement, 1997). RT3D requires thegroundwater flow code a MODular three-dimensionalfinite-difference groundwater FLOW model (MOD-FLOW;McDonald and Harbaugh, 1988) for computingspatial and temporal variations in groundwater headdistribution. The RT3D code uses the operator–splitstrategy, which allows definitions of different reactionmodels through a plug-in user-defined reaction mod-ule (Clement, 1997; Clement et al., 1998, 2000). Inthis study, a kinetic model is developed to analyzethe nitrification–denitrification reaction in groundwa-ter systems. Then a new reaction module for the RT3Dis developed to describe the fate and transport of nitro-gen species, DOC, DO and biomass in the groundwaterenvironment systems.

Studies on nitrogen transformation associated withfirst-order kinetics can be found inChowdary et al.(2004), Clement (2001), Senzia et al. (2002)andVanclooster et al. (1995). Senzia et al. (2002)presenteda dynamic rational model for nitrogen transformation,and investigated the main mechanisms for nitrogenremoval in primary facultative ponds. In their work,denitrification was modeled using first-order kinet-ics while nitrification was modeled based on Monodkinetics. Studies on reactive transport in the satu-rated/unsaturated zone with multiple-Monod expres-sions for processes converting ammonium to molecularnitrogen can be found inEssaid et al. (1995), Kindredand Celia (1989), Kinzelbach et al. (1991), ClementeM y( lti-c ifera portp thec ullM m-p ther thefi owni

3

onsd olid-p atu-

f nitrogen gas in the groundwater reaches saion, it tends to migrate out of the saturated zKorom, 1992). However, although denitrificatioakes place preferably under anoxic conditions,enitrification has been described byRobertson et a1989).

. Model development

In this section, a mathematical model for predng the fate and transport of nitrogen species, dissorganic carbon (DOC), DO and biomass in the grouater environment systems is presented. This sses the RT3D code to analyze the nitrogen transpaturated groundwater systems. The RT3D code she coupled partial differential equations that desceactive transport of multiple mobile and/or imm

t al. (1997), MacQuarrie and Sudicky (2001)andacQuarrie et al. (2001). MacQuarrie and Sudick

2001) proposed a theoretical development of muomponent transport simulation in a shallow aqund applied their model to a field nitrogen transroblem. The model proposed in this study addedapability of computing nitrite–nitrogen using a fonod kinetic model. Thus, it was possible to coare the computed nitrite–nitrogen concentration ineaction process from ammonium to nitrate witheld-measured nitrite–nitrogen concentration as shn Fig. 8(c and d).

.1. Solute transport

In RT3D, the general macroscopic equatiescribing the fate and transport of aqueous- and shase species, respectively, in multi-dimensional s


rated porous media are represented as (Clement, 1997):

∂Ck

∂t= ∂

∂xi

(Dij

∂Ck

∂xj

)− ∂

∂xi

(viCk) + qs

φCsk ± rk,

k = 1, 2, . . . , m (1a)

dCim

dt= ±rim, im = 1, 2, . . . , (n–m) (1b)

where n is the total number of species,m the totalnumber of aqueous-phase (mobile) species (whencen–m is the total number of solid-phase or immobilespecies),Ck the aqueous-phase concentration of thekthspecies (ML−3), Cim the solid-phase concentration oftheimth species (MM−1),Dij the hydrodynamic disper-sion coefficient (L2T−1),vi the pore velocity (LT−1),qsthe volumetric flux of water per unit volume of aquiferthat represents sources and sinks (T−1), φ the poros-ity, Csk the concentration of source/sink (ML−3), rkrepresents all aqueous-phase reaction rate terms thatdescribe the mass of the species removed or producedper unit volume per unit time (ML3T−1) andrim repre-sents all solid-phase reaction rate terms (MM−1T−1).

3.2. Biogeochemical reactions

The multiple-Monod kinetics (Molz et al., 1986;Widdowson et al., 1988; Kinzelbach et al., 1991;Lindstrom, 1992; Chen et al., 1992; Essaid et al., 1995;Mu pro-c ratee ctionp thatt tra-te ple-M ssa

r

w ionp te

utilization rate for reaction processp (T−1), Xm thebiomass concentration of the populationm responsiblefor the reaction (ML−3), C1,C2, . . ., Cn are the aqueousspecies concentrations (ML−3) andK

p1 , K

p2 , . . . , K

pn

are the half-saturation constants for the respectivespecies (ML−3). The competitive inhibition model isused to represent the inhibition of a secondary sub-strate when the primary substrate is still present (Essaidet al., 1995). Ib(Xm) andInc(CI ) are referred to as thebiomass and non-competitive inhibition terms, respec-tively (Lehninger, 1975; Parkin and Speece, 1982;Widdowson et al., 1988; Essaid et al., 1995) and aregiven by (Kindred and Celia, 1989; Chen et al., 1992):

Ib(Xm) =[1 + Xm

kbm

]−1

(3a)

Inc(CI ) =[1 + CI

kCI

]−1

(3b)

wherekbmis the biomass inhibition constant (ML−3),

kCI the inhibition constant of the inhibiting substances(ML−3) and CI is the aqueous concentration of theinhibiting substances (ML−3).

The total reaction rate for speciesk is equal to thesum of the rates for all processes in which the speciesk is involved, hence

rk =∑p

ypk rp (4)

w sf rys ofs ionp

byt oxy-g her tioni a,1

w iamt stant(

acQuarrie and Sudicky, 1997; Lu et al., 1999) aresed for describing the biodegradation reactionesses that involve several solutes. The kineticquations are used to describe all of the major rearocesses in the RT3D. The formulation assumes

he biodegradation reaction is limited by the concenion of all substances involved in the reaction (Essaidt al., 1995). Based on these assumptions, the multionod kinetics for biodegradation reaction procepre given by:

p = µpmaxXmIb(Xm)Inc(CI )

[C1

Kp1 + C1

]

×[

C2

Kp2 + C2

]· · ·

[Cn

Kpn + Cn

](2)

hererp is the substrate utilization rate for reactrocessp (ML−3T−1), µ

pmax the maximum substra

hereypk is the specific uptake coefficient of speciek

or reaction processp. When the solute is the primaubstrate,yp

k is equal to 1. Otherwise, it is the ratioolutek relative to the primary substrate for reactrocessp.

Microbial metabolism is assumed to be limitedhe existence of either organic carbon, dissolveden or nitrogen (ammonium or nitrite or nitrate). Tate of biomass growth for each microbial populas expressed as (Molz et al., 1986; Kindred and Celi989):

dXm

dt= Ymrp − Xmdm (5)

hereYm is the microbial yield coefficient for bacterwhen mediating reaction processp (M/M) anddm is

he specific biomass death or maintenance rate con1/T).


3.3. Nitrogen transformations and transport

Ammonium oxidation to nitrite can be written as(Reddy and Patrick, 1975):

r1 : NH4+ + 3

2O2(aq)→ NO2− + H2O + 2H+ (6)

Nitrite oxidation to nitrate can be written as (Reddyand Patrick, 1975):

r2 : NO2− + 1

2O2(aq)→ NO3− (7)

The kinetic equations for nitrification have the fol-lowing form.

r1 = µnit1maxX1

[kb1

kb1 + X1

] [NH4

KNH4 + NH4

]

×[

O2

KO2 + O2

](8)

r2 = µnit2maxX2

[kb2

kb2 + X2

] [NO2

KNO2 + NO2

]

×[

O2

KO2 + O2

](9)

where r1 is the substrate utilization rate by ammo-nium oxidation process 1 (mg/(l day)),r2 the sub-strate utilization rate by nitrite oxidation process 2(mg/(l day)),µnit1 the maximum substrate utilizationr iscm byn y),X ia-o eaK

N -o n-s

biala latiles ,t sim-p bed ally.W ac-

tion for denitrification is:

r3 : CH2O + 45NO3

− → 25N2 + 4

5HCO3−

+ 15CO2(aq)+ 1

5H2O (10)

In general, denitrifying enzyme inhibition isassumed to be reversible and non-competitive in nature(Widdowson et al., 1988). The kinetic equations fordenitrification have the following form.

r3 = µdenitmaxX3

[kb3

kb3 + X3

] [kO2I

kO2I + O2

]

×[

CH2O

KCH2O + CH2O

] [NO3

KNO3 + NO3

](11)

wherer3 is the substrate utilization rate by denitrifica-tion (mg/(l day)),µdenit

max the maximum substrate utiliza-tion rate for denitrification, whereby nitrate–nitrogenis reduced to nitrogen gas (l/day);KCH2O, KNO3 thehalf-saturation constants for CH2O and NO3

−(mg/l),X3 the heterotrophic biomass concentration (mg/l) andkb3 andkO2I are the heterotrophic biomass and oxygeninhibition constant (mg/l).

The above reaction kinetics are combined with thecoupled transport equations in the saturated zone asfollows:

RNH4

∂ [NH4] = ∂(

Dij

∂ [NH4])

− ∂(vi[NH4])

R

R

maxate for nitrification, whereby ammonium–nitrogenonverted to nitrite–nitrogen (l/day),µnit2

max the maxi-um substrate utilization rate for nitrification, whereitrite–nitrogen is converted to nitrate–nitrogen (l/da1 the concentration of autotrophic ammonxidizing biomass (mg/l),X2 the concentration of thutotrophic nitrite-oxidizing biomass (mg/l);KNH4,NO2 andKO2 the half-saturation constants for NH4

+,O2

− and O2 (mg/l) andkb1 andkb2 are the ammoniaxidizing and nitrite-oxidizing biomass inhibition cotants (mg/l).

Dissolved organic carbon is the source of micrond nitrate reaction, and is assumed to be non-voubject to sorption (Jardine et al., 1992). In this paperhe organic compounds are represented by thelified chemical formula CH2O, and assumed toegraded by heterotrophic microorganisms aerobichen coupled with oxidation of DOC, the overall re

∂t ∂xi ∂xj ∂xi

+ qs

φ[NH4]s − r1 (12a)

NO2

∂[NO2]

∂t= ∂

∂xi

(Dij

∂[NO2]

∂xj

)− ∂(vi[NO2])

∂xi

+ qs

φ[NO2]s + yNO2/NH4r

1 − r2

(12b)

NO3

∂[NO3]

∂t= ∂

∂xi

(Dij

∂[NO3]

∂xj

)− ∂(vi[NO3])

∂xi

+qs

φ[NO3]s + yNO3/NO2r

2 − r3

(12c)


RN2

∂[N2]

∂t= ∂

∂xi

(Dij

∂[N2]

∂xj

)− ∂(vi[N2])

∂xi

+ qs

φ[N2]s + yN2/NO3r

3 (12d)

where [NH4], [NO2], [NO3] and [N2] represent con-centrations of ammonium–nitrogen, nitrite–nitrogen,nitrate–nitrogen and nitrogen gas (mg/l), respectively,RNH4, RNO2, RNO3 and RN2 are the retardation fac-tors for the various species andyNO2/NH4, yNO3/NO2

and yN2/NO3 are the ratios of secondary substrateto primary substrate consumed, as determined fromstoichiometry (mg/mg). The RT3D code selected tosolve this mathematical model uses an operator–splittechnique to represent kinetic reactions. Using theoperator–split method, the biochemical reaction kinet-ics included in the nitrogen transport equations can beseparated and expressed as a set of ordinary differentialequations:

d[NH4]

dt= − r1

RNH4

(13a)

d[NO2]

dt= yNO2/NH4r

1 − r2

RNO2

(13b)

d[NO3]

dt= yNO3/NO2r

2 − r3

RNO3

(13c)

odelt

3o

ro-c ort theb gra-d oxi-d ion.T ena

r

The kinetic equations for DOC oxidation have thefollowing form.

r4 = µoxidmaxX3

[kb3

kb3 + X3

] [CH2O

KCH2O + CH2O

]

×[

O2

KO2 + O2

](15)

wherer4 is the substrate utilization rate by denitrifica-tion (mg/(l day)),µoxid

max the maximum substrate utiliza-tion rate for organic carbon oxidation (l/day).

The fate and transport of DOC in a multi-dimensional saturated porous media can be written as:

RCH2O∂[CH2O]

∂t= ∂

∂xi

(Dij

∂[CH2O]

∂xj

)

− ∂(vi[CH2O])

∂xi

+ qs

φ[CH2O]s − r4 (16)

where [CH2O] represents the concentration of DOC(mg/l) andRCH2O is the retardation factor of DOC.

After using the operator–split method, the biochem-ical reaction kinetics included in the DOC transportequations can be expressed as:

d[CH2O]

dt= − r4

RCH2O(17)

n am writ-t

R

w /l),R

y b-s inedf

em-i gen

d[N2]

dt= yN2/NO3r

3

RN2

(13d)

These reaction kinetic equations were used to mhe coupled denitrification processes.

.4. Dissolved organic carbon and dissolvedxygen transport

The nitrogen losses through denitrification pesses require a suitable carbon substrate to suppiological activity in the saturated zone. The biodeation of organic compounds is represented as anation reaction mediated by a microbial populathe stoichiometry of DOC oxidation can be writts:

4 : CH2O + O2(aq)→ CO2(aq)+ H2O (14)

The fate and transport of dissolved oxygen iulti-dimensional saturated porous media can be

en as:

O2

∂[O2]

∂t= ∂

∂xi

(Dij

∂[O2]

∂xj

)− ∂(vi[O2])

∂xi

+ qs

φ[O2]s − yO2/NH4r

1 − yO2/NO2r2

− yO2/CH2Or4 (18)

here [O2] represent concentrations of oxygen (mgO2 the retardation factor for oxygen;yO2/NH4,O2/NO2 andyO2/CH2O are the ratios of secondary sutrate to primary substrate consumed, as determrom stoichiometry (mg/mg).

After using the operator–split method, the biochcal reaction kinetics included in the dissolved oxy


transport equations can be expressed as:

d[O2]

dt= −yO2/NH4r

1 − yO2/NO2r2 − yO2/CH2Or4

RO2(19)

3.5. Microbial reactions

In this work, the microbial population was concep-tualized as having three types of microbial biomassthat involve autotrophic ammonia-oxidizing bacteria,autotrophic nitrite-oxidizing bacteria and heterotrophicbacteria. The equations for the growth rate of thebiomass have the following form:

dX1

dt= Y1r

1 − X1d1 (20a)

dX2

dt= Y2r

2 − X2d2 (20b)

dX3

dt= Y3(r3 + r4) − X3d3 (20c)

whereY1, Y2 and Y3 are the microbial yield coeffi-cients for the autotrophic ammonia-oxidizing bacteria,autotrophic nitrite-oxidizing bacteria and heterotrophicbacteria (mg biomass/mg substrate), respectively, andd1, d2 andd3 are death or maintenance rate constants(l/day) of the autotrophic ammonia-oxidizing bacteria,autotrophic nitrite-oxidizing bacteria and heterotrophicbiomass. The fate and transport model presented abovei ecies,a T3Ds imu-l ns.T ithint ,1

4

telyd itro-g anicc one-d

them

Table 1Summary of aquifer hydrology, geometry and transport parametersused for verification cases (cases 1 and 2)

Parameter Value

Hydraulic conductivity (cm/day) 832Porosity 0.44Soil bulk density (mg/l) 1.56× 106

Longitudinal dispersivity (cm) 0.5Average linear velocity (cm/day) 10Linear partitioning coefficient of

DOC (l/mg),KdCH2O

2.93× 10−8

Linear partitioning coefficient ofammonium (l/mg),KdNH4

35.2× 10−8

multiple electron acceptor respiration (oxygen-basedand nitrate-based) limited decay of a carbonaceoussubstrate is considered. The test case assumes one-dimensional transport in a hypothetical laboratory-scale column (Widdowson et al., 1988). The col-umn length is 50 cm, and the node spacing is 0.5 cm.Hydrodynamic properties and numerical parametersare given inTable 1. In the second simulation, theinitial substrate, oxygen, nitrate and biomass con-centrations are assumed to be 1.0, 2.0 and 5.0 mg/l,and 3.64× 10−3 mg/cm3, respectively, and to be uni-form across the domain. The substrate, oxygen, nitrateand biomass concentrations at the left boundary ofthe domain are fixed at 20.0, 2.0 and 5.0 mg/l, and3.64× 10−3 mg/cm3, respectively. The concentrationgradient for all solutes is fixed at zero for the rightboundary condition. The Monod kinetics reactionparameters used for this case are listed inTable 2. Theresults are similar to the first case except in that thiscase does not include biomass growth. The simulationtime is 8 days and the time step size 0.08 days. Simu-lation results indicated that the model conserved masswell. The overall mass balance error at the end of the8 days is about 0.15%. The concentration profiles fort = 4 and 8 days are plotted inFig. 2.

The second simulation shows ammonification basedon the governing equations presented previously. Theinitial substrate, oxygen, ammonia and biomass con-centrations are assumed to be 1.0, 2.0, 6.0 and0.565 mg/l, respectively, and to be uniform across thed massc arefi Thec for

ncludes six aqueous species and three attached spnd the reaction model was implemented into the Roftware as a reaction package to perform field sations of nitrogen and microbial species distributiohe details of the computational procedure used w

he RT3D code are discussed elsewhere (Clement et al.998, 2000).

. Model verification

To verify that the proposed model adequaescribes the fate and transport behavior of the nen compounds, dissolved oxygen, dissolved orgarbon and biomass, as it undergoes reaction,imensional verification examples are presented.

The verification simulations are compared withodeling results ofWiddowson et al. (1988)where

omain. The substrate, oxygen, ammonia and biooncentrations at the left boundary of the domainxed at 10.0, 2.0, 6.0 and 0.565 mg/l, respectively.oncentration gradient for all solute is fixed at zero


Table 2Summary of Monod kinetics reaction parameters for case 1

Parameter Value

µnit1max (day) 10

µnit2max (day) 10

µdenitmax (day) 40

µoxidmax (day) 30

KNH4 (mg/l) 1KNO2 (mg/l) 1.8KNO3 (mg/l) 2.6KO2 (mg/l) 0.77KCH2O (mg/l) 40kO2I (mg/l) 0.01kb1 (mg/l) 1kb2 (mg/l) 1kb3 (mg/l) 0.5yNO2/NH4 2.5504yNO3/NO2 1.3478yN2/NO3 0.4518yO2/NH4 1.7739yO2/NO2 0.6955yO2/CH2O 1.0657Y1 0.45Y2 0.45Y3 0.5d1 (day) 0.02d2 (day) 0.02d3 (day) 0.02

the right boundary condition. Hydrodynamic parame-ters and Monod kinetics reaction parameters are givenin Tables 1 and 3. Simulated concentration profiles fordissolved oxygen, ammonia and substrate are presentedin Fig. 3. The simulation time is 12 days and the time-step size 0.12 days. The overall mass balance error atthe end of the simulation period is about 0.14%. Resultsindicated that the model conserved mass well. In bothcases the model results compared favorably.

5. Simulation of nitrogen transport at a fieldsite

5.1. Field site description

The Vasse Research Station is operated by the West-ern Australian Department of Agriculture and is locatedsouthwest of Western Australia (Fig. 4). The VasseResearch Station is located within a coastal plain thatis flat and gently undulating. A superficial aquiferis developed in the Geographe Bay Catchment that

Fig. 2. Numerical simulation results for case 1 in mg/l: (a) at 4 daysand (b) at 8 days.

contains the Vasse Research Station. The superficialaquifer comprises a clay and silt layer overlain by aveneer of a sand layer. The water table in the uncon-fined sand aquifer closely parallels the land surface.

Fig. 3. Numerical simulation results for case 2 in mg/l.


Table 3Summary of Monod kinetics reaction parameters for case 2

Parameter Value

µnit1max (day) 1

µnit2max (day) 1

µdenitmax (day) 40

µoxidmax (day) 40

KNH4 (mg/l) 1KNO2 (mg/l) 1.8KNO3 (mg/l) 2.6KO2 (mg/l) 0.77KCH2O (mg/l) 40kO2I (mg/l) 0.01kb1 (mg/l) 1kb2 (mg/l) 1kb3 (mg/l) 0.5yNO2/NH4 2.5504yNO3/NO2 1.3478yN2/NO3 0.4518yO2/NH4 1.7739yO2/NO2 0.6955yO2/CH2O 1.0657Y1 0.45Y2 0.45Y3 0.5d1 (day) 0.02d2 (day) 0.02d3 (day) 0.02

Fig. 4. Location of Vasse Research Station within the GeographeBay Catchment (Fahrner, 2002).

Groundwater recharge is mainly by direct infiltrationof rainfall. The rainfall of the Busselton region for thelast 100 years ranges from 1100 mm/year in the westto 700 mm/year in the east (Fahrner, 2002).

The Research Station contains a cattle feedlot,which has operated intermittently since its construc-tion in 2000.Fahrner (2002)provided a comprehen-sive review of the nitrogen contamination at this siteand provided a detailed dataset for characterizing theperformance of a groundwater bioremediation trench(GBT) used for controlling the nitrogen problem. InApril 2001, the GBT, which was filled with a mixture ofsand and sawdust, was constructed by the Water Corpo-ration of Western Australia. Sawdust GBT assumes thatnitrate contaminants can be bioremediated by provid-ing excess carbon. Nitrate concentrations in groundwa-ter monitoring bores at the feedlot site ranged from 2.1to 180 mg/l nitrate–nitrogen prior to the constructionof the trench. Average nitrogen loading to the trench isabout 1767 kg N/year based on mean nitrate–nitrogenconcentrations of 63 mg/l down-gradient from the feed-lot. The GBT was constructed between the feedlot anda local groundwater discharge drain. A trench (170 mlong and 1.5 m wide) was excavated through sand toa clay layer approximately 1.5–2 m deep. The exca-vated soil was mixed with 160 m3 of sawdust using alarge excavator (ratio of soil to sawdust 30:70). Thesawdust was fresh, untreatedPinus radiata and gen-erally <5 mm in diameter. The soil/sawdust mix wasthen returned to the trench to form the groundwaterb pril2 estt llelt thed edi-a secto nch( resw resd theb -t m-p 001a tio-n d tor res,c nchb sam-

ioremediation trench. The trench was created in A001 when the groundwater table was at its low

o minimize construction difficulties. Three pararansects of groundwater bores were installed, toepth of the clay hardpan, parallel to the bioremtion trench. In the simulation domain, one tranf 15 bores was located up-gradient of the treparallel to the GBT), another transect of 15 boithin the trench and the final transect of 14 boown-gradient from the trench. The locations ofores are shown inFig. 5 with hydraulic head con

ours. All groundwater monitoring bores were saled monthly for nutrient analysis between May 2nd April 2002. The monitoring program was raalized after 12 months of data had been collecteeduce monitoring costs. A reduced selection of bohosen to best represent the function of the treased on previous groundwater analysis data, was


Fig. 5. Simulated groundwater flow field (a) and cross-sectional view of the shallow aquifer (b) (adapted fromFahrner, 2002).

pled for nutrient analyses from May to September2002.

5.2. Groundwater flow simulation

Two-dimensional numerical simulation is carriedout using MODFLOW. The model domain for thepresent study is a 160 m× 28 m section with 20 m thickand the grid consists of 80 grids in thex-direction, 70grids in they-direction and 1 grid in thez-direction.Grid spacing is 2 m, and 0.4 m in thex- and in they-direction, respectively.

Specified head boundaries were used along theup and down borders of the model grid to simulatethe groundwater flow pattern measured at the site.Details of the aquifer hydrology, geometry and trans-port parameters are given inTable 4.

The aquifer comprises sand overlying clay with thesand layer ranging in thickness from 1.89 to 1.26 m.The aquifer is deepest in the northeast (1.89 m) andbecomes thinner to the southwest (1.26 m).

Fig. 5 shows the calibrated groundwater flowfield. The numerical simulation result of the flowshows that flow is in a northeast–southwest direc-tion and groundwater velocities are in the rangeof about 0.28–0.71 m/day. These values agree wellwith the field-determined values of 0.3–0.6 m/day.Flux of groundwater through the trench was inthe range of 18,615–37,230 m3/year. Fig. 6 repre-sents a comparison between the observed and com-puted hydraulic head distribution. The mean error,mean absolute error and root mean squared (rms)error between observed and computed values ofheads at these wells are 0.001, 0.095 and 0.164 m,


Table 4Summary of aquifer hydrology, geometry and transport parametersused for the Vasse Research Station simulation

Parameter Value

Hydraulic conductivity (m/day)Out trench 7Within trench 21

Porosity,φ 0.3Soil bulk density (mg/l) 1.6× 106

DispersivityLongitudinal (m) 5Ratio of transverse to longitudinal 1.5

Linear partitioning coefficient ofDOC (l/mg),KdCH2O

a1.5× 10−6

Linear partitioning coefficient ofammonium (l/mg),KdNH4

b0.34× 10−6

a Jardine et al. (1992).b Ceazan et al. (1989).

respectively. The model error is within a reasonablelevel.

5.3. Reactive transport simulation

The reactive transport simulation was run with atime step of 0.5 days to a total time of 1000 days.In this study, six biogeochemical reaction zones wereassumed to be present at the site. The reaction zoneapproach used in this work is based on the largeamount of site-specific data. The Vasse Research Sta-tion site is contaminated with nitrogenous compoundsfrom the manure disposed at the feedlot.Fig. 7shows

Fig. 6. Correlation of observed vs. computed heads in 44 observationwells.

six biogeochemical reaction zones discretized at thesite.

The source area was divided into five zones based onthe field data: zone 1 to zone 5. The other area besidesthe source zones is zone 6 and it covers most of themodel domain down-gradient from the source zone.Concentrations of nitrogen compounds and DOC ofthe source zones increase from zone 1 to zone 3 asthe peak and then they decrease toward zone 5. Thatis, zone 3 has the highest concentrations and zone 1and zone 5 have relatively lower concentrations while

Fig. 7. Biogeochemical reaction zones at the Vasse Research Station.


zone 2 and zone 4 have concentrations in-between thehigher and lower values. In the source zones, the initialNH4

+, NO2−, NO3

− and DOC were specified basedon field observations from monitoring wells installedinside the zones. Zone 3 has the highest concentra-tions of NH4

+ (60 mg/l), NO3− (110 mg/l) and DOC

(50 mg/l). Zone 1 and zone 5, respectively, are locatedat the upper left and the upper right corners of the simu-lation domain and have relatively lower concentrationsof NH4

+ (20 and 10 mg/l), NO3− (20 and 70 mg/l) andDOC (30 and 20 mg/l). Zone 6 consists of initiallyoxygenated uncontaminated groundwater. Therefore,

Table 5Summary of Monod kinetics reaction parameters for the VasseResearch Station simulation

Parameter Value Reference(s)

µnit1max (day) 1 Regression equation ofLegget and

Iskandar (1981)µnit2

max (day) 3.5 Regression equation ofLegget andIskandar (1981)

µdenitmax (day) 10 Henze et al. (1987), Kinzelbach et al.

(1991)µoxid

max (day) 10 Henze et al. (1987), Baek et al. (1989),Kinzelbach et al. (1991)

KNH4 (mg/l) 0.1 Regression equation ofLegget andIskandar (1981)

KNO2 (mg/l) 0.3 Assumed valueKNO3 (mg/l) 0.5 Henze et al. (1987), Kinzelbach et al.

(1991)KO2 (mg/l) 0.77 Legget and Iskandar (1981), Henze

et al. (1987)K l.

k

k

k

k

y

y

y

y

y

y

Y )

Y

Y

ddd

in this zone the initial NH4+, NO2−, NO3

− and DOCconcentrations were specified as zero, as backgroundconcentration in the aquifer was below the detectionlimit. The initial DO concentration was specified basedon field observations made outside the feedlot contam-inant plume. The initial ammonia-oxidizing bacteria,nitrite-oxidizing bacteria and heterotrophic bacteriawere uniformly set at 0.001, 0.001 and 0.1 mg/l, respec-tively, over the entire domain.

Boundary conditions were specified as zero concen-tration gradient for all constituents at each side of themodel grid except for the source zone. In the sourcezone, concentrations for all model constituents wereset equal to the initial condition and held constant overtime.

The composition of feedlot waste is obtained fromthe monitoring data. It is assumed that the effluent iscompletely devoid of dissolved oxygen and molecularnitrogen (N2). In the model, molecular nitrogen is usedonly as an indicator of the occurrence of denitrifica-tion. This assumption of no molecular nitrogen in thewastewater does not affect the other reactive species(MacQuarrie and Sudicky, 1997).

Monod kinetics reaction parameters are given inTable 5. Most of the biogeochemical kinetic param-eters are obtained from the literature, and thus areindependent input data for the reactive transport simu-lation.

Trial and error adjustment of the initial microbialpopulation density, maximum substrate utilization ratea ratet tiond delc con-c

TT g)

CC

CH2O (mg/l) 6 Henze et al. (1987), Kinzelbach et a(1991)

O2I (mg/l) 0.01 Kindred and Celia (1989)

b1 (mg/l) 1 Assumed value

b2 (mg/l) 1 Assumed value

b3 (mg/l) 0.5 Assumed value

NO2/NH4 2.5504 Calculated value

NO3/NO2 1.3478 Calculated value

N2/NO3 0.4518 Calculated value

O2/NH4 1.7739 Calculated value

O2/NO2 0.6955 Calculated value

O2/CH2O 1.0657 Calculated value

1 0.5 Henze et al. (1987), Baek et al. (1989,Kinzelbach et al. (1991)

2 0.5 Henze et al. (1987), Baek et al. (1989),Kinzelbach et al. (1991)

3 0.5 Henze et al. (1987), Baek et al. (1989),Kinzelbach et al. (1991)

1 (day) 0.02 Assumed value



DMM

T 1

ND

nd initial DOC concentration was required to calibhe reactive transport model to the solute concentraistribution collected after trench installation. Moalibration was evaluated by comparing simulatedentrations with field data.

able 6he overall mass balance error at the end of 1000 days (unit, m

In Out

onstant concentration 9708300.616 −3921652.470onstant head 0.000 −1744763.843ecay of biodegradation 7075829.946 −8000389.843ass storage (solute) 305656.398 −686047.675ass storage (adsorbed) 6084.335 −1168731.700

otal 17095871.295 −15521585.53

et (In− Out) 1574285.764iscrepancy (%) 0.092


The observed concentration distributions of fourreactive species are shown inFig. 8(a–h). These field-measured concentration contours are compared withthe model-predicted concentration contours for NH4

+,NO2

−, NO3− and DO. Comparison inFig. 8shows that

the calibrated model is able to reproduce the generalpatterns of NH4+, NO2

−, NO3− and DO concentration

plumes. However, field-measured nitrite and nitrateconcentrations were slightly higher than the model-predicted values.

Fig. 8(a–d) show that an extensive nitrite plume ofthe highest concentration coincides with the edges ofthe NH4

+ plume corresponding to the nitrification ofammonium to nitrite. Nitrification requires dissolvedoxygen, so autotrophic bacteria increased from an ini-

tial value of 0.001 mg/l in the western portion of thedomain. Consequently, nitrite concentration increasednear the western portion of the domain. Results indi-cated that the model conserved mass well. The overallmass balance error at the end of 1000 days of simulationis about 0.09% (seeTable 6).

In Fig. 9(a–c), the computed isolines for N2, DOCand biomass concentrations 426 days after the begin-ning of the remediation are shown. Ammonium massin the system increases nearly linearly and laterstarts to decrease as a result of increases in theammonia-oxidizing bacteria. Finally, the ammoniummass remains constant, which indicates that steadystate conditions have been achieved. Due to low initialnitrifying biomass, the generation of nitrate is delayed.

Fnmc

ig. 8. Comparison of field-measured ammonium concentration (itrite–nitrogen concentration (c) and model-predicted nitrite–nitrogenodel-predicted nitrate–nitrogen concentration (f); field-measured di

oncentration (h). Units of contours are mg/l.

a) and model-predicted ammonium concentration (b); field-measuredconcentration (d); field-measured nitrate–nitrogen concentration (e) and

ssolved oxygen concentration (g) and model-predicted dissolved oxygen


Fig. 8. (Continued ).

After this short lag time, the nitrate concentrationincreases. But after 431 days, the nitrate concentra-tion was reduced gradually, and was less than 10 mg/laround the trench at 468 days. Reductions in nitrateare considered to be from microbial denitrification.This is supported by a high correlation between nitrateand molecular nitrogen concentrations where molecu-lar nitrogen release increases from the system.

The denitrifying bacteria that mediate the deni-trification process require an organic carbon sourceto metabolise nitrate. The possible sources of thisorganic carbon are dissolved organic carbon originat-ing from the manure in the feedlot, and solid organiccarbon, supplied to the system in the form of sawdust.Fig. 9(b) shows that denitrifying bacteria increased inand around trench and feedlot. Generally, the GBTusing sawdust assumes that organic carbon is depletedin the environment in which the nitrate plume is located,

thus providing a source of carbon, sawdust is expectedto enhance microbial denitrification. The simulationand field results show that the carbon supplied fromcattle manure is not a sufficient source of carbon formicrobial denitrification and hence the addition of acarbon source, such as sawdust, is required for deni-trification. If the quantity of carbon within the manurewere sufficient to cause complete reduction of nitrate,then the formation of a plume with high nitrate concen-trations would not be expected. Since the fundamentalreason this study has been conducted is due to thenitrate plume, this suggests that either the carbon frommanure is not sufficient for complete denitrificationor that the site conditions were not suitable for den-itrification. It is suggested that the installation of theGBT would provide carbon when otherwise the car-bon source from manure was limiting the denitrificationprocess.


Fig. 9. Model-predicted concentrations of molecular nitrogen (a), dissolved organic carbon and heterotrophic bacteria (b) and ammonia-oxidizingbacteria and nitrite-oxidizing bacteria (c). Units of contours are mg/l.

6. Summary and conclusions

In this study, a mathematical model was developedto describe transformations and transport of nitrogencompounds in a saturated groundwater aquifer. Themodel was coded as a reaction module within theRT3D framework. The model was verified by compar-ing simulation results obtained using the code againstproblems previously published in the literature. Sev-eral sets of one-dimensional simulations have been runto check the self-consistency, feasibility and physicalreasonableness of the model before the field applica-tion. These simulations show that the model providesphysically reasonable results. The developed nitrogentransformations and transport model were then usedto simulate the behavior of the major reactive nitro-gen species at the Vasse Research Station, Busselton,Western Australia. The model was used to reproduce

the field data for ammonia, nitrite, nitrate and dis-solved oxygen concentrations. The model simulationresults for the major reactive species are comparedwith data from the site to apply the model in a predic-tive mode. The results show close agreement betweenthe simulated and measured concentrations for nitro-gen compound and dissolved oxygen. It is concludedthat the developed model is a rigorous, practical anduseful forecasting tool, which can be used to simulatethe fate and transport of nitrogen species in ground-water systems. It can also be used to design remedialsystems such as bioremediation trenches.

Acknowledgments

This study was financially supported by a grant(code: 3-4-2) from the Sustainable Water Resources


Research Center of 21st Century Frontier ResearchProgram, Korean Ministry of Science and Technology.AEBRC at POSTECH partially supported this study.The field data for this research was taken from thehonors thesis of Ms. Sabina Fahrner (Fahrner, 2002).We would like to thank Mr. Brett Jago and Mr. BenO’Grady of Perth Water Corporation for their supportwith the honors field project.

References

Baek, N.H., Clesceri, L.S., Clesceri, N.L., 1989. Modeling ofenhanced biodegradation in unsaturated soil zone. J. Environ.Eng. 115, 150–172.

Behnke, J., 1975. A summary of the biogeochemistry of nitrogencompounds in groundwater. J. Hydrol. 27, 155–167.

Bosch, H.M., Rosenfield, A.B., Huston, R., Shipman, H.R., Wood-ward, F.L., 1950. Methemoglobinemia and Minnesota well sup-plies. J AWWA 42, 161–170.

Burt, T.P., Heathwaite, A.L., Trudgill, S.T., 1993. Nitrate: Process,Patterns and Management. Wiley, New York.

Ceazan, M.L., Thurman, E.M., Smith, R.L., 1989. Retardation ofammonium and potassium transport through a contaminated sandand gravel aquifer: the role of cation exchange. Environ. Sci.Technol. 23, 1402–1408.

Chen, Y.M., Abriola, L.M., Alvarez, P.J.J., Anid, P.J., Vogel, T.M.,1992. Modeling transport and biodegradation of benzene andtoluene in sandy aquifer material: comparisons with experimentalmeasurements. Water Resour. Res. 28 (7), 1833–1847.

Chowdary, V.M., Rao, N.H., Sarma, P.B.S., 2004. A coupled soilwater and nitrogen balance model for flooded rice fields in India.Agric. Ecosyst. Environ. 103, 425–441.

C tingaterNL-

C sportater

C , C.,nds:site.

C ersen,nderon-

C elingound

D ans-rge:271–

Dorsch, M.M., Scragg, R.K.R., McMichael, A.J., Baghurst, P.A.,Dyer, K.F., 1984. Congenital malformations and maternal drink-ing water supply in rural south Australia: a case-control study.Am. J. Epidemiol. 119 (4), 473–486.

Essaid, H.I., Bekins, B.A., Godsy, E.M., Warren, E., Baedecker,M.J., Cozzarelli, I.M., 1995. Simulation of aerobic and anaerobicbiodegradation processes at a crude oil spill site. Water Resour.Res. 31 (12), 3309–3327.

Fahrner, S., 2002. Groundwater nitrate removal using a bioreme-diation trench, Honors thesis, Centre for Water Research, TheUniversity of Western Australia.

Henze, M., Grady, C.P.L., Gujer, W., Marais, G.V.R., Matsuo, T.,1987. A general model for single wastewater treatment systems.Wat. Res. 21, 505–515.

Jardine, P.M., Dunnuvant, F.M., Selim, H.M., McCarthy, J.F., 1992.Comparison of models for describing the transport of dissolvedorganic carbon in aquifer column. Soil Sci. Soc. Am. J. 56,393–401.

Kindred, J.S., Celia, M.A., 1989. Contaminant transport andbiodegradation 2. Conceptual model and test simulations. WaterResour. Res. 25 (6), 1149–1159.

Kinzelbach, W., Schaafer, W., Herzer, J., 1991. Numerical model-ing of natural and enhanced denitrification processes in aquifers.Water Resour. Res. 27 (6), 1123–1135.

Korom, S.F., 1992. Natural denitrification in the saturated zone: areview. Water Resour. Res. 28 (6), 1657–1668.

Legget, D.C., Iskandar, I.K., 1981. Evaluation of a nitrificationmodel. In: Iskandar, I.K. (Ed.), Modeling Wastewater Renova-tion: Land Treatment. Wiley, New York, pp. 313–358.

Lehninger, A.L., 1975. Biochemistry. Worth, New York.Lindstrom, F.T., 1992. A mathematical model for the one-

dimensional transport and fate of oxygen and substrate in awater-saturated sorbing homogeneous porous medium. WaterResour. Res. 28 (9), 2499–2511.

Lu, G., Clement, T.P., Zheng, C., Wiedemeier, T.H., 1999. Naturaleld-

M im-llownces,

M im-llowe. J.

M lti-car-

o a

M ee-aterrvey,

M iongen207–

lement, T.P., 1997. RT3D-A modular computer code for simulareactive multi-species transport in 3-dimensional groundwaquifers, Pacific Northwest National Laboratory Report, PN11720 (http://bioprocess.pnl.gov/rt3d.htm).

lement, T.P., 2001. Generalized solution to multispecies tranequations coupled with a first-order reaction network. WResour. Res. 37 (1), 157–163.

lement, T.P., Johnson, C.D., Sun, Y., Klecka, G.M., Bartlett2000. Natural attenuation of chlorinated ethene compoumodel development and field-scale application at the DoverJ. Contam. Hydrol. 42, 113–140.

lement, T.P., Peyton, B.M., Skeen, R.S., Jennings, D.A., PetJ.N., 1997. Microbial growth and transport in porous media udenitrification conditions: experiments and simulations. J. Ctam. Hydrol. 24, 269–285.

lement, T.P., Sun, Y., Hooker, B.S., Pertersen, J.N., 1998. Modmulti-species reactive transport in groundwater aquifers. GrWater Monit. Rem. 18 (2), 79–92.

eSimone, L.A., Howes, B.L., 1998. Nitrogen transport and trformations in a shallow aquifer receiving wastewater dischaa mass balance approach. Water Resour. Res. 34 (2),285.

attenuation of BTEX compounds: model development and fiscale application. Ground Water 37 (5), 707–717.

acQuarrie, K.T.B., Sudicky, E.A., 1997. Multicomponent sulation of wastewater-derived nitrogen and carbon in shaunconfined aquifers, PhD thesis, Department of Earth ScieUniversity of Waterloo, Waterloo, Ontario.

acQuarrie, K.T.B., Sudicky, E.A., 2001. Multicomponent sulation of wastewater-derived nitrogen and carbon in shaunconfined aquifers. I. Model formulation and performancContam. Hydrol. 47, 53–84.

acQuarrie, K.T.B., Sudicky, E.A., Robertson, W.D., 2001. Mucomponent simulation of wastewater-derived nitrogen andbon in shallow unconfined aquifers. II. Model application tfield site. J. Contam. Hydrol. 47, 85–104.

cDonald, M.D., Harbaugh, A.W., 1988. A modular thrdimensional finite difference flow model, Techniques in WResources Investigations of the U.S. Geological SuBook 6.

olz, F.J., Widdowson, M.A., Benefield, L.D., 1986. Simulatof microbial growth dynamics coupled to nutrient and oxytransport in porous media. Water Resour. Res. 22 (8), 11216.


National Academy of Sciences (NAS), 1978. Nitrates: an environ-mental assessment, a report prepared by the panel on nitratesof the Coordinating Comm. Sci. Tech. Assess. Environ. Pollut.Washington, DC.

Parkin, G.F., Speece, R.E., 1982. Modeling toxicity in methane fer-mentation systems. J. Env. Eng. Div. 108, 515–531.

Reddy, K.R., Patrick, W.H., 1975. Effect of alternate aerobic andanaerobic conditions on redox potential, organic matter decom-position and nitrogen loss in a flooded soil. Soil Biol. Biochem.7, 87–94.

Robertson, L.A., Cornelisse, R., De Vos, P., Hadioetomo, R., Kue-nen, J.G., 1989. Aerobic denitrification in various heterotrophicnitrifiers. Antonie van Leeuwenhoek 56, 289–299.

Schmidt, I., Bock, E., 1997. Anaerobic ammonia oxidation withnitrogen dioxide byNitrosomonas eutropha. Arch. Microbiol.167, 106–111.

Schmidt, I., Bock, E., 1998. Anaerobic ammonia oxidation by cell-free extracts ofNitrosomonas eutropha. Antonie van Leeuwen-hoek 73, 271–278.

Senzia, M.A., Mayo, A.W., Mbwette, T.S.A., Katima, J.H.Y.,Jørgensen, S.E., 2002. Modelling nitrogen transformation and

removal in primary facultative ponds. Ecol. Model. 154,207–215.

Shuval, H., Gruener, N., 1977. Infant methemoglobinemia and otherhealth effects of nitrates in drinking water. Prog. Water Technol.8 (4/5), 183–193.

Strous, M., Fuerst, J.A., Kramer, E.H.M., Logemann, S., Muyzer,G., van de Pas-Schoonen, K.T., Webb, R., Kuenen, J.G., Jetten,M.S.M., 1999. Missing lithotroph identified as new plancto-mycete. Nature 400, 446–449.

U.S. EPA, 1980. Nitrogen-ammonia/nitrite/nitrate, water qualitystandards criteria summaries. GPO: 1980-341-082/107. Wash-ington, DC.

Vanclooster, M., Viaene, P., Diels, J., Feyen, J., 1995. A deterministicevaluation analysis applied to an integrated soil-crop model. Ecol.Model. 81, 183–195.

White, R.J., 1983. Nitrate in British waters. J. Int. Water SupplyAssoc. (AQUA) 2, 51–57.

Widdowson, M.A., Molz, F.J., Benefield, L.D., 1988. A numericaltransport model for oxygen- and nitrate-based respiration linkedto substrate and nutrient availability in porous media. WaterResour. Res. 24 (9), 1553–1565.

nitrogen transformation and transport modeling in ...clemept/publsihed_pdf/...ecological modelling...

Documents