Flow and Reactivity Effects on Dissolved Organic Matter Transport in Soil ColumnsHarald Weigand and Kai U. Totsche*

ABSTRACTDissolved organic matter (DOM) plays a prominent role in the

transport of contaminants in porous media. As DOM has to be consid-ered as a reactive component, flow regime and sorbent reactivityshould affect overall DOM transport in an important way. We focusedon DOM transport in unsaturated column experiments using quartzsand (QS) and goethite-coated quartz sand (GS). Rate constrictionsto DOM sorption were investigated by varying the volumetric flowrate, while extent and reversibility of sorption were studied in consecu-tive adsorption and desorption steps. In the QS, DOM retentionwas low and unaffected by changes in flow rate. Desorption-stepbreakthrough curves (BTCs) and mass balances show full reversibilityof the sorption process. However, DOM retention in GS was signifi-cant and sensitive to flow variation, indicative of nonequilibrium sorp-tion. At lower flow rates, DOM breakthrough exhibited a change incurvature (shoulder) due to the superimposition of two BTCs repre-senting reactive and nonreactive DOM fractions. Transport was suc-cessfully modeled assuming these two fractions governed overallDOM mobility. At higher flow rates, the ETC shoulder vanished dueto reduced contact time between the DOM and the solid phase (rate-limited sorption). Sorption of DOM on GS is accompanied by amarked rise in effluent pH, indicative of a ligand-exchange mecha-nism. Recovery of DOM during desorption was incomplete due toeither partially irreversible sorption or strongly rate-limited desorp-tion. Increased DOM mobility in the consecutive adsorption stepresulted from partial blocking of sorption sites by the initial pulseof DOM.

DISSOLVED ORGANIC MATTER plays a prominent rolein the understanding of transport and reaction pro-

cesses in soils. Dissolved organic matter is known to beinvolved in the processes of podsolization (Dawson etal., 1978; Buurmann, 1985), soil acidification (Brownand Sposito, 1991), and mineral weathering (Heyes andMoore, 1992). Recently, increasing interest has arisenin understanding the role of DOM as a transport-facili-tating agent of nutrients (Quails et al., 1991) and con-taminants (McCarthy and Zachara, 1989; Knabner etal., 1996).

Dissolved organic matter can be conceived as a con-tinuum of substances of biotic origin, which are partiallyor fully degraded and transformed (Quails et al., 1991;Guggenberger et al., 1994). The compounds involvedshow a wide range of ionization constant (p/Q andpoint of zero charge (PZC) values, molecular sizes, andfunctional groups (Gu et al., 1994). The mobility ofDOM in soils is governed by both the chemistry ofthe bulk liquid and the composition of the soil mineralphase, where the former determines DOM's dissolutionproperties (Schlautmann and Morgan, 1994) and thelatter controls the extent of DOM adsorption. Besidesedges of layer silicates and quartz grains (Hiemstra and

Soil Physics Div., Univ. of Bayreuth, Germany. Received 21 July 1997.* Corresponding author (totsche@uni-bayreuth.de).

Published in Soil Sci. Soc. Am. J. 62:1268-1274 (1998).

Van Riemsdijk, 1990), metal oxides and hydroxides areknown as effective soil sorbents for DOM (Tipping,1981; Davis and Gloor, 1981; Murphy et al., 1992).

For these, DOM is a reactive component of the soilsolution and affects contaminant mobility twofold: en-hancement due to the formation of a mobile associate(Johnson and Amy, 1995) or reduction due to eitherincreased sorption capacity of the solid phase or cosorp-tion of DOM-associated contaminants (Murphy and Za-chara, 1995, Totsche et al., 1997).

For a given mineralogical composition, DOM sorp-tion will be modified by the flow process. Especiallyunder the conditions of varying pore water velocities,one may expect rate-limited sorption behavior (Brus-seau and Rao, 1989) to affect the mobility of DOM.Most transport experiments involving DOM have beenperformed under saturated conditions. The results ob-tained are therefore limited to saturated porous media,e.g., aquifers and sediments. However, soils of the ter-restrial environment are predominantly unsaturated.The degree of saturation affects the flow regime andthereby possible rate constraints to sorption process.This is enhanced by the restricted accessibility of sorp-tion sites under the prevailing moisture conditions ofsoils. To account for these effects, our experiments wereperformed under unsaturated flow conditions.

Our study addressed the influence that flow regimeand solid-phase reactivity have on DOM mobility in theunsaturated zone. Special consideration was given tofeatures of sorption nonideality, such as sorption irre-versibility, isotherm nonlinearity, rate-limited sorption,sorption hysteresis, and the effect of the composite na-ture of DOM.

MATERIALS AND METHODSMaterials

Quartz sand (AKW, Amberg, Germany) and goethite-coated quartz sand (provided by the Dep. of Soil Science andPlant Nutrition, University of Wageningen, Wageningen, theNetherlands) were used as model porous media. PedogenicFe and Al (hydr)oxides were determined as the dithionite-citrate-bicarbonate-extractable fraction of metal oxides ac-cording to the method introduced by Mehra and Jackson(1960). Porosity was calculated from bulk densities and therespective substance densities. Table 1 lists the physicochemi-cal features of the solid phases. A DOM stock solution wasprepared by extracting forest-floor organic matter on a 1:10solid/liquid ratio. The forest-floor material was sampled atthe Waldstein experimental site, northeast Bavaria, Germany.The DOM was subject to operational fractionation according

Abbreviations: ETC, breakthrough curve; DOM, dissolved organicmatter; GS, goethite-coated quartz sand; LEM, linear equilibriummodel; PV, pore volume; QS, quartz sand; TSTR, two-site, two-region model.

1268

WEIGAND & TOTSCHE: DISSOLVED ORGANIC MATTER TRANSPORT 1269

Table 1. Physicochemical properties of quartz sand and goethite-coated quartz sand.

Grain size distribution, %2-0.63 mm0.63-0.2 mm0.2-0.063 mmbulk density, g/cm3

PorositypH (CaCUtFe, kg kg-'§Al, kg kg-'§

Quartz sand

IS85ND1.60.386.2

3.0 x 10'54.0 X ID'5

Goethite-coatedquartz sand

NDf33671.60.405.7

3.5 X 10~<ND

t ND = not detectable.t Extraction ratio soil/0.01 M CaCl2:1:2.5.§ Dithionite-citrate-bicarbonate extractable.

to Leenheer (1981), revealing =62% hydrophobic moietiesand =32% hydrophilic moieties.

To protect the percolation solutions from microbial trans-formation, AgNO3 was added at a concentration of 2 x 10~6

M. The extract was filtered twice through SUPUR-102 0.45|xm membrane filters (Gelman Sciences, Ann Arbor, MI) andthen diluted to an inflow concentration of =4 X 1CT3 M on acarbon basis. As a conservative tracer, 10~3 M Nad was addedto the adsorption-step solutions (Table 2). All solutions wereprepared with deionized and degassed water. All chemicalswere Merck nanograde.

Experimental MethodsSoil column experiments were performed using a computer-

controlled repacked soil column system (7 cm in height, 9 cmin diameter, Fig. 1). A stainless steel porous plate (Krebssoge,Germany) was used as the bottom capping of the soil column,with suction applied by means of a hanging water column.For continuous feed of the percolation solution to the sprin-kling unit and transport of column effluent to the fractioncollector, a peristaltic pump (Gilson Minipuls 3, Gilson Co.,Worthington, OH) was used with Technicon Tygon R3607tubes (i.d. 1.49 mm). A fraction collector (Foxy, Isco Inc.,Lincoln, NE) was used to separate distinct effluent fractionsat time intervals corresponding to about 0.3 pore volumes(PVs). Prior to the experiments, soil columns were saturatedwith water from bottom to top at low flow rates (3 mL/h) toprevent air entrapment and guarantee a uniform flow domain.Steady-state unsaturated flow conditions were establishedwhile irrigating the system with a tracer-free solution afterthe suction had been applied. Based on the soil moisturecharacteristic of the materials (data not shown), suction headsof 25 and 55 cm were used in the QS and GS columns, respec-tively. This procedure guaranteed similar and time-constantwater contents in the experiments (Table 3). As macroscopicpore water velocity is given by the ratio of Darcy velocityand water contents, variations were achieved by changing thevolumetric flow rates. After the effluent flow rates provedconstant with time, consecutive adsorption and desorptionsteps were carried out. Constant feed of the percolation solu-tions was maintained until breakthrough was completed (typi-

Table 2. Composition of percolation solutions.

Dissolved organic C, mol C/LNaCl, mol/LAgNO3, mol/LKCIO4, mol/LPH

Solution A:adsorption

tracer4.2 X 10 3

io-3

2 X 10-'10 '4.5

Solution B:stationaritydesorption

<0.2 x 10-"<10~5

2 X 10-'io-3

5.7

Fig. 1. Schematic representation of the experimental setup.

cally 1200 cm3 for the QS and 2300 cm3 for the GS systems).Column Peclet numbers as obtained from tracer BTCs indi-cated a moderately convection-dominated flow regime. Exper-imental conditions are given in Table 3.

Analytical MethodsAnalysis of effluent fractions comprised the measurement

of nonpurgeable organic C, Cl~, pH, and electrolytic conduc-tivity. Nonpurgeable organic C was measured on a ShimadzuTC 5050 TOC analyzer (repeated analysis; Shimadzu Corp.,Tokyo) accepting a standard deviation of 2%. Chloride andpH were electrochemically determined with ion-sensitive elec-trodes (Cl~: Model 96-17, Colora, Ulm, Germany; pH: ModelLOT 405-60-S7/9848 combination electrode, Ingold, Stein-bach/Taunus, Germany). Electrolytic conductivity was mea-sured using an Ingold 980-K19/120 conductivity cell.

EvaluationAll breakthrough data are given as reduced variables. The

number of PVs eluted is given as the ratio of volume eluted andthe volumetric water content of the soil column (gravimetricaldetermination at the end of the experiments). Reduced con-centrations were calculated from the ratio of effluent andinfluent concentrations. Mass balances of resident organic Cwere obtained by subtracting the integrated effluent and influ-ent masses. Analogously to the reduced concentrations, resi-dent masses were scaled to theoretical resident mass presentin the aqueous phase in the absence of adsorption. All ETCdata were subject to error analysis using Taylor's theory oferror propagation.

Breakthrough curves were analyzed using two differentmethods of numerical analysis: Parameters of the advection-dispersion equation were estimated using the FORTRANcode CXTFIT (Parker and van Genuchten, 1984). Both the

Table 3. Experimental parameters of dissolved organic matterbreakthrough curves in quartz sand (WS) and goethite-coatedquartz sand (GS) columns.

Parameter!L, cmA, cm1

*„,, cmPb, g/cm3

0,, m m J

PV, cm3

v cm/s

QS slow7.2

63.6225

1.620.1655

75.611.50 x 10~3

QS fast7.18

63.6225

1.610.1639

75.072.99 X 10-3

GS slow7.2

63.6255

1.560.1477

67.661.62 X 10-3

GS fast7.26

63.6255

1.590.1961

90.62.42 x 10'3

t L = length of column; A = cross-sectional area of soil column; ^tfm =pressure head; pb = bulk density, 0, = volumetric water content; PV =pore volume; v = pore water velocity (v = qlf>,).

1270 SOIL SCI. SOC. AM. J., VOL. 62, SEPTEMBER-OCTOBER 1998

10.8

0.6

0.4

0.2 .6OO

0.8

0.6

0.4

0.2

0

OAof

<AA

o crA DOM

TA o crA DOM

0A

pore volumeFig. 2. Tracer and dissolved organic matter (DOM) breakthrough in

quartz sand at (a) higher and (b) lower pore water velocity. Theindex i reflect the fact that the two different substances, DOM andtracer, are normalized to the respective inflow concentration levels,which are not the same (» = 1 for DOM; i = 2 for Cl~).

linear equilibrium model (LEM) and the two-site, two-regionmodel (TSTR) approach were used. The first assumes a singletype of equilibrium sorption site. In this case, fitted parametersare the dispersion coefficient D and the retardation coefficientR. The TSTR approach allows chemical or physical nonequi-librium to affect the shape of the ETC. It considers either thecoexistence of bulk solid fractions providing equilibrium andnonequilibrium sorption sites or the presence of mobile andimmobile water regions. Additional parameters are thereforea measure of these fractions or regions F and a first-orderrate constant a. Longitudinal dispersivity was set equal forCr and DOM. The effect of the composite nature of DOMand of isotherm nonlinearity on the shape of BTCs was ana-lyzed in simulations with the numerical code CARRY(Totsche et al., 1996). This program operates with a bulk solid

10 20 30 40 50

pore volumeFig. 3. Tracer and dissolved organic matter (DOM) breakthrough in

goethite-coated quartz sand at (a) higher and (b) lower pore watervelocity. The index i reflect the fact that the two different sub-stances, DOM and tracer, are normalized to the respective inflowconcentration levels, which are not the same (i = 1 for DOM; / =2 for Cr).

oO

o"

0.8

0.6

0.4

0.2T DOM measured quartz sand— DOM measured goethite sand

— DOM calculated, linear equilibrium—DOM calculated, two-site, two-region

10 20 30

pore volumeFig. 4. Adsorption step breakthrough curves in quartz sand and goe-

thite-coated quartz sand columns at lower pore water velocity usingmeasured and calculated data.

phase composed of different and specific sorption sites, eachcharacterized in terms of bulk density fractions. Different reac-tive solutes may coexist within the liquid phase. The followingscenario was assumed to hold true for the transport of DOMin the GS columns: (i) bifractional composition of DOM (Fl,F2) with different affinities toward the porous medium (40and 60% nonreactive and reactive DOM, respectively); (ii)specific sorption site density for DOM (/^OM) is given as thebulk density fraction occupied by Fe and Al (hydr)oxides; (iii)sorption of reactive DOM follows a nonlinear Freundlich-type isotherm, which was normalized to contents of Fe andAl (hydr)oxides (Kfe<M); (iv) sorption of reactive DOM is sub-ject to rate constrictions by introducing a first-order rate con-stant (kr) to account for the influence of flow velocity onDOM retention.

Prior to the application of this scenario, the validity ofassumption (ii) was tested by comparison of simulated BTCsand fits obtained by the LEM of CXTFIT.

RESULTS AND DISCUSSIONFigure 2 shows the adsorption and desorption BTCs

of Cr and DOM in QS columns at lower (Fig. 2a) andhigher (Fig. 2b) pore water velocity. In comparison tothe tracer, DOM was slightly retarded. Almost 100%of the inflow concentration level was reached within sixPVs. Variation of pore water velocity had a negligibleeffect on DOM mobility in the QS columns. DesorptionBTCs, initiated after 10 PVs, showed a decline in DOMeffluent concentrations to zero concentration after 15PVs. Adsorption and desorption steps were thus com-pleted in equal time intervals. A nonsingular DOMsorption characteristic is suggested by differences in theareas between DOM and Cl~ adsorption and desorptionBTCs. However, this is not the result of a partiallyirriversible DOM sorption, but of a higher retention ofCl~ during the desorption step (Table 4). The overallbreakthrough of DOM in QS is governed by (i) thelow adsorption capacity of the porous medium given bytraces of Al (hydr)oxide, which leads to only a slightreduction in DOM mobility compared with the tracer,(ii) the absence of rate limitations during DOM adsorp-tion rendering contact-time invariance of breakthrough(i.e., pore water velocity invariance), and (iii) the action

WEIOAND & TOTSCHE: DISSOLVED ORGANIC MATTER TRANSPORT 1271

Table 4. Results of fits to breakthrough curve data by the linear equilibrium and the two-site, two-region approaches for quartz sand(QS) and goethite-coated quartz sand (GS) columns.

BTC step

Adsorption

Desorption

Adsorption

Adsorption

Parameter!

RCI(95% CI)D, cmVs(95% CI)Pe^DOM(95% CI)KI DOM, cm'/gRo(95% CI)D, cm2/s(95% CI)PeROOM(95% CI)Ka DOM, cmVg

Ra(95% CI)D, cmVs(95% CI)PeROOM(95% CI)KI DOM, cmVg

R(95% CI)KM, cm'/gKa, cnrVgP(95% CI)Fci>(95% CI)a,s-'

QS slow

1.21(1.18-1.25)5.2 x 10-"

(4.2 X 10 -'- 6.8 X 10-")14.51.72

(1.63-1.82)7.0 X 10-!

1.65(1.60-1.70)3.8 X 10~4

(2.4 X 10~4-5.2 X 10~4)29.11.90

(1.85-1.94)9.0 X 10~2

QS fast

Linear equilibrium model1.17

(1.14-1.20)1.5 X 10~3

(1.1 X 10-'-1.9 X 10 3)14.31.48

(1.36-1.61)5.0 X 10~2

1.17(1.14-1.20)1.3 X 10-3

(1.0 X 10~3-1.6 X 10~3)15.61.12

(1.02-1.22)2.0 X 10~2

1.21(1.19-1.23)1.5 X 10^3

(1.0 X 10~3-1.6 X 10 3)13.91.20

(1.15-1.26)3.0 X 10~2

Two-site, two-region model

GS slow

1.60(1.54-1.67)2.2 X 10~3

(1.7 X «T3-2.6 X 10-3)5.34.5

(4.19-4.82)3.3 X 10-'

1.07(1.03-1.11)1.2 X 10~3

(9.5 X 10~4-1.5 X 10'3)9.31.10

(0.98-1.13)5.0 x 10~3

1.50(1.49-1.52)1.6 X 10~3

(1.0 X 10~3-1.6 X 10'3)7.21.70

(1.48-1.93)6.0 X 10~!

5.76(5.56-5.97)4.5 X 10-'3.4 X 10 '

0.38(0.35-0.41)

0.250.61

(0.56-0.67)3.3 X 10-s

GS fast

1.01(0.97-1.05)1.1 X 10~3

(7.3 x lO-'-l.S X15.23.12

(2.98-3.27)2.7 X 10~'

1.06(1.03-1.09)1.2 X 10~3

(9.4 X 10 4-1.5 X15.11.21

(1.18-1.25)3.0 x 10~2

1.10(1.08-1.13)1.1 x 10~3

(1.0 x 10~3-1.6 x15.81.27

(1.18-1.36)3.0 X 10~2

3.60(3.52-3.68)3.2 x 10-12.7 X 10-'

0.37(0.35-0.40)

0.151.37

(1.71-1.59)2.1 X 10~4

10~3)

io-4)

10 3)

t D = dispersion coefficient (assumed to be constant for tracer and dissolved organic matter (DOM)); Pe = Peclet number Pe = vLID; R = retardationcoefficient; Ka = slope of linear isotherm for DOM and Cl~. Ka = (R — 1) 6/pb; Ka, and A"d, = slope of linear isotherm corresponding to spontaneousand rate-limited sorption, respectively; F = fraction of bulk density for spontaneous sorption (site or region), F = (|3(0 + Pb/fd) - 0)/(Opi,Kd); a = first-order rate parameter for rate-limited sorption, a = <o/(l — P)v/(fl£); CI = confidence interval.

of a reversible sorption characteristic that is also corrob-orated by mass balances.

These results contrast severely with the findings forthe GS columns (Fig. 3). Compared with the tracer,DOM BTCs of the first adsorption steps were stronglyretarded. This holds true for both the higher (Fig. 3a)and the lower pore water velocity (Fig. 3b). However,sorption was more pronounced at the lower pore watervelocity, showing completion of the first adsorptionBTCs after 35 PVs. Compared with the marked retarda-tion in the adsorption step, the desorption step yieldeda fast decline to low concentrations. This indicates theaction of an adsorption-desorption hysteresis. Similarbehavior has been reported for the adsorption of DOMon Fe oxide powder in batch experiments by Gu et al.(1994, 1995). A direct experimental approach to sorp-tion hysteresis in flow systems is the performance ofmultiple adsorption pulses (Gu et al., 1996). This wasconsidered by performing a second adsorption step inthe GS columns (Fig. 3). In contrast to the first adsorp-tion step, a tracer-like behavior of DOM was observedat the higher pore water velocity. As sorption hysteresisinduces incomplete removal of DOM within the timeof the desorption step, saturation of accessible sorptionsites in the second adsorption step was reached faster.

Mobility of DOM was thereby increased. For the lowerpore water velocity, additional sorption is indicated bytailing of the BTC after 55 PVs. This suggests nonequi-librium sorption even at reduced flow velocity.

Figure 4 shows the observed QS and GS BTCs atlower pore water velocities and the fitted BTCs as de-scribed by the LEM and the TSTR. For the QS, a fitby the LEM shows acceptable agreement with measureddata and yields a retardation coefficient for DOM equiv-alent to Cl~ (Table 4). Slight deviations between mea-sured and fitted BTCs, however, are observable in thetailing parts of the adsorption step. Deviations betweenmeasured and calculated data are expressed better inthe GS BTC. Again, the model fails to meet the tailingpart and predicts a faster breakthrough at low concen-tration levels. Contrasting this, the fit of GS data to theTSTR approach yields good agreement. According tothe model-derived process parameter F, only 25% oftotal DOM removal from solution occurred due to aninstantaneous adsorption (Table 4). It is important tonote that extended tailing of the BTC can also resultfrom nonlinear sorption following, e.g., a Freundlich-type isotherm with an exponent >1 (Biirgiesser et al.,1993). Therefore, the good agreement between the ob-served shape of a BTC and the fit to the TSTR does

1272 SOIL SCI. SOC. AM. J., VOL. 62, SEPTEMBER-OCTOBER 1998

Table 5. Input parameters for the simulation of dissolved organicmatter DOM breathrough curves using Kte

Parameter)

L0, m3 m~3

q, cm/spb, g/cm3

/?<>", g/cm3

X, cmKs,,,̂ d« cnrYg Fe, Al

Goethite-coatedquartz sand, slow

7.20.1477

2.40 X 10-"1.56

5.62 x 10-"1.44

940

Quartz sand, slow

7.20.1655

2.48 X 10-"1.626.81~3

0.34940

t L = length of column; 6 = water content; q = Darcian f lux; /°OM =fraction of bulk density occupied by sorbents for DOM; /™m = pb (g/cm3) X sesquioxide content (kg/kg), \ = dispersivity; KFe,Ai (cm3/g Fe,Al) = partition coefficient obtained from fit to linear equilibrium modelscaled to sesquioxide contents, 7fFe,AI = (R - 1) 0//5OM.

not necessarily mean that the underlying process is ofa kinetic nature. Nevertheless, the response of the DOMbreakthrough to variations in pore water velocity is anindependent experimental observation that is stronglycorroborated by the agreement with the TSTR model.

To test whether the differences in overall retardationfor the QS and GS columns can be attributed to differ-ences in pedogenic Fe and Al (hydr)oxides only, wecarried out simulations based on the results of the fitsto the LEM (Table 5). The linear isotherm was scaledto the presumed DOM-specific sorbent fraction of thematrix (/^OM: metal (hydr)oxide contents times bulkdensity) and resulted in a K?^ AI of 940 cm3/g Fe and Al.Consequently, in our simulations, only/^OM was consid-ered a sorbent for DOM. As expected, this scaling pro-cedure was invariant to DOM breakthrough in GS (Fig.5). However, the close agreement with the fit to theQS ETC indicates that indeed the metal (hydr)oxidecontent is an appropriate estimate of the affinity ofDOM toward porous media.

The development of effluent pH along with the DOMis presented in Fig. 6. Although stationary flow condi-tions were established at pH 5.7 (Table 2), both systemsare characterized by initial effluent pH values around6.3. We attribute this buffering to the protonation ofgoethite surface functional groups: =Fe-OH2

+ ^=Fe-OH° + H+; intrinsic ionization constant (p&j,1") =

QUARTZ SAND:T——TCARRY; adsorption onto oxide

T——TCXT4; adsorption onto whole soil

GOETHITE SAND :•——•CARRY; adsorption onto oxide•——•CXT4; adsorption onto whole soil

10 15 20

pore volumeFig. 5. Effect of the parameter A'Fc M (derived from the linear equilib-

rium model) on the simulation of quartz sand and goethite-coatedquartz sand breakthrough curves.

0.8

o 0.6O

o"0.4

0.2

— pH goethite sand fastO DOM goethite sand fast

— pH goethite sand stow— DOM goethite sand slow

6.5

6 i.

5.5

pore volumeFig. 6. Breakthrough curves of dissolved organic matter (DOM) and

effluent pH in goethite-coated quartz sand at varied pore watervelocity. The pH values vrere subject to smoothing by movingaverage.

6.2 (Sigg and Stumm, 1980). After the start of percola-tion with DOM, additional acidity was introduced. How-ever, the first three PVs of both BTCs were accompa-nied by a further increase in effluent pH values.Electrostatic interactions, induced by changes in ionicstrength, have been shown to alter effluent pH values(Meussen et al., 1996). In our experiments, this effectcan be ruled out because ionic strength, as calculatedfrom electrolytic conductivity (Griffin and Jurinak,1973), is only doubled during breakthrough of DOM.We therefore interpret the rise in pH values as a conse-quence of the ligand-exchange sorption mechanism asreported by several researchers from batch experiments(Tipping, 1981; Davis, 1982; Ochs et al., 1994). A hypo-thetical reaction equation for the exchange between acarboxylate functional group and a surface hydroxide-ligand is given by: Fe-OH° + R-COO~ ^Fe-OOC-R + OH~ (Murphy and Zachara, 1995). Inagreement with the slower breakthrough (stronger ad-sorption), hydroxide release at lower pore water velocityis more pronounced. In the course of both experiments,

Table 6. Input parameters for the simulation of dissolved organicmatter (DOM) breakthrough curves through goethite-coatedquartz-sand using DOM fractions with different reactivities.

Parameterf

L, cm8, m3 m~3

q, cm/sPb, g/cm3

y$OM, g/cm3

A, cmKm cnrVgK,2, crn'o/gfPvk «-'Rrtp, "Fraction 1Fraction 2

Goethite-coated Goethite-coatedquartz sand, slow quartz sand, fast

7.20.1477

2.40 X 10~4

1.565.62 X 10 4

0.455.0

3.2 X 102

0.855 X 10-3

40% DOM60% DOM

7.20.1961

4.75 x 10-"1.56

5.62 x 10-4

0.45

t L = length of column; 6 = water content; q = Darcian flux; pb = bulkdensity;/*'"' = fraction of bulk density occupied by sorbents for DOM;KVi = slope of linear isotherm describing the adsorption of Fraction 1;Kr2 = affinity parameter of Freundlich-type isotherm describing thesorption of Fraction 2; p,. - power of (non)linear nonequilibrium iso-therm describing the adsorption of Fraction 2; krlf = first-order rateparameter for the adsorption of Fraction 2.

WEIGAND & TOTSCHE: DISSOLVED ORGANIC MATTER TRANSPORT 1273

0.8

50.606

0.4

0.2 • DOM measured——— overall DOM simulated— • • • reactive fraction simulated—it— nonreactive fraction simulated

10 15

pv

0.8

°0.4

0.2 O DOM observed•—— overall DOM simulated• " • • reactive fraction simulated•— non-reactive fraction simulated

10 15

pvFig. 7. Measured and independently simulated breakthrough curves

(BTCs) for dissolved organic matter (DOM) for the (a) fast and(b) slow transport experiments. The formation of a shoulder bysuperimposing the BTCs of nonreactive and reactive DOM frac-tions is sensitive to pore water velocity.

the pH dropped again, reflecting the lower influent pHvalues relative to the initially buffered pH of the po-rous medium.

A closer look at the ETC shape at lower pore watervelocity reveals a plateau between two and three PVs.This shoulder separates the initially steep rise in effluentconcentrations from the tailing in the ETC. Shoulderformation is a common phenomenon associated withthe transport of DOM (Liu and Amy, 1993; Gu et al.,1996). Experimental investigations into this phenome-non have indicated a fractionation of DOM during thetransport process into subcomponents according to dif-ferences in reactivity of molecular size classes (Gu etal., 1994, McCarthy et al., 1996) or the hydrophilic-hydrophobic fractions of DOM (Jardine et al., 1989;McCarthy et al., 1993). Explicit modeling of the compos-ite reactivity of DOM, however, has so far not been per-formed.

To numerically describe the breakthrough of DOMindependently from measured BTCs (parameters givenin Table 6), we propose the shoulder formation to bethe result of the superimposition of BTCs of two distinctDOM fractions comprising compounds with similar

™o

AA

>~V4**** • goethite sand fast

A goethite sand slow

• quartz sand fast

10 20 30 40 50

pore volumeFig. 8. Reduced resident mass as obtained from balancing inflow and

outflow of dissolved organic matter. M/M0 = 1 represents the levelof reduced resident mass for completed infiltration in the absenceof adsorption.

sorption behavior (Totsche et al., 1997). The first frac-tion is assumed to be nonreactive, while the secondfraction follows a rate-limited sorption characteristic.Results are given in Fig. 7. A good qualitative agreementwith the experimental information can be observed: su-perimposing BTCs of subcomponents leads to the for-mation of a plateau at lower pore water velocity. Athigher pore water velocity, however, due to the rateconstrictions on the sorption of reactive DOM, a clearseparation is hindered.

Figure 8 shows the normalized mass balance for resi-dent organic C of the two model sorbents. The weakaffinity of DOM toward QS is evident along with the fullreversibility of DOM sorption indicated by the completedecline of resident C in the desorption step. Comparedwith this, reduced resident C levels for the GS at higherpore water velocity were almost doubled at the endof the first adsorption step. This finding reflects thedynamics of DOM in the field, which leads to subsoilhorizons depleted or enriched in organic matter. Theevaluation of batch and column experiments in the senseof the underlying KKtM should therefore be promisingin the assessment of DOM mobility in different porousmedia. The effect of a sorption hysteresis in GS is evi-dent from the level of resident mass present after thedesorption step. Because of discontinuities in DOM dis-charge under natural conditions, the sorption hysteresisobserved in GS can be considered a contribution to thestability and efficient buildup of C-enriched horizons.As discussed above, hysteresis leads to a faster stockupto sorption capacity in the consecutive adsorption step.Thus, for a given mineralogical composition, it promotesthe downward migration of C enrichment. Addressingthe different mass levels in the GS experiments, theeffect of a more reactive substrate (QS fast vs. GS fast)is almost equaled by the variation of pore water velocity(GS fast vs. GS slow), again showing the importance ofa rate-limited adsorption process for this material. Asa consequence, with increasing pore water velocity, dataobtained from equilibrium isotherms would underesti-mate DOM concentrations in pore water. The apparent

1274 SOIL SCI. SOC. AM. J., VOL. 62, SEPTEMBER-OCTOBER 1998

effect of chemical nonequilibrium should be consideredin addition to reduced accessibility of sorption sites inaggregated soils. As to the significance of DOM mobilityin the transport of xenobiotics, hysteresis and rate limi-tations of the sorption process counteract each other.While hysteresis helps in stabilizing the sorptive soilpool, rate-limited adsorption at a given flow regimepromotes the export of cotransported substances fromthe system together with possible contamination ofgroundwater. The absence of an overall DOM mobilityas outlined by the GS experiments indicates that therecan be no general statement on whether transport ofcontaminants will be enhanced or reduced in the pres-ence of DOM. Future research should consequently ad-dress the behavior of DOM subcomponents and theirrole in contaminant binding and transport.

ACKNOWLEDGMENTSWe wish to thank Dr. Phil Jardine (Oak Ridge, TN) for

his valuable comments and for fruitful discussion. This workwas financially supported by the Deutsche Forschungsgem-einschaft under Contract no. To 184/3-2.