wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 2

Avai lab le a t www.sc iencedi rec t .com

journa l homepage : www.e lsev ie r . com/ loca te /wat res

Colloid transport in porous media: Impact of hyper-salinesolutions

Einat Magal a,b, Noam Weisbrod a,*, Yoseph Yechieli b, Sharon L. Walker c,Alexander Yakirevich a

aDepartment of Environmental Hydrology & Microbiology, Zuckerberg Institute for Water Research, Balustein Institutes for Desert Studies,

Ben Gurion University of the Negev, Sede Boqer 84990, IsraelbGeological Survey of Israel, Jerusalem 95501, IsraelcDepartment of Chemical and Environmental Engineering, University of California, Riverside, Riverside, CA 92521, USA

a r t i c l e i n f o

Article history:

Received 14 January 2011

Received in revised form

12 April 2011

Accepted 14 April 2011

Available online 22 April 2011

Keywords:

Colloids

Brine

Breakthrough curve

Salinity

Microspheres

* Corresponding author. Tel.: þ972 8 6596903E-mail address: weisbrod@bgu.ac.il (N. W

0043-1354/$ e see front matter ª 2011 Elsevdoi:10.1016/j.watres.2011.04.021

a b s t r a c t

The transport of colloids suspended in natural saline solutions with a wide range of ionic

strengths, up to that of Dead Sea brines (100.9 M) was explored. Migration of microspheres

through saturated sand columns of different sizes was studied in laboratory experiments

and simulated with mathematical models.

Colloid transport was found to be related to the solution salinity as expected. The

relative concentration of colloids at the columns outlet decreased (after 2e3 pore volumes)

as the solution ionic strength increased until a critical value was reached (ionic

strength > 10�1.8 M) and then remained constant above this level of salinity.

The colloids were found to be mobile even in the extremely saline brines of the Dead

Sea. At such high ionic strength no energetic barrier to colloid attachment was presumed

to exist and colloid deposition was expected to be a favorable process. However, even at

these salinity levels, colloid attachment was not complete and the transport ofw30% of the

colloids through the 30-cm long columns was detected.

To further explore the deposition of colloids on sand surfaces in Dead Sea brines,

transport was studied using 7-cm long columns through which hundreds of pore volumes

were introduced. The resulting breakthrough curves exhibited a bimodal shape whereby

the relative concentration (C/C0) of colloids at the outlet rose to a value of 0.8, and it

remained relatively constant (for the w18 pore volumes during which the colloid

suspension was flushed through the column) and then the relative concentration increased

to a value of one. The bimodal nature of the breakthrough suggests different rates of colloid

attachment. Colloid transport processes were successfully modeled using the limited

entrapment model, which assumes that the colloid attachment rate is dependent on the

concentration of the attached colloids. Application of this model provided confirmation of

the colloid aggregation and their accelerated attachment during transport through soil in

high salinity solution.

ª 2011 Elsevier Ltd. All rights reserved.

; fax: þ972 8 6596909.eisbrod).ier Ltd. All rights reserved.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 23522

1. Introduction the presence of energy barriers (low IS) are slow or reaction-

Colloid transport in groundwater has been studied extensively

in the past years andmajor findings have been summarized in

several review papers (e.g., Bradford and Torkzaban, 2008.;

Kretzschmar et al., 1999; McDowell-Boyer et al., 1986; Ryan and

Elimelech, 1996; Sen and Khilar, 2006). The interaction energy

between colloids and collectors is a key issue in retention

processes of colloids in porous media. DLVO theory describes

the interactionenergybetweensurfacesas a sumof theelectric

double layer and London-van der Waals forces (Kretzschmar

et al., 1999; Ryan and Elimelech, 1996). The ionic strength (IS)

of the suspending solution has amajor impact on the potential

for interaction of colloids with surfaces. In response to an

increase in the IS, the diffusive layer compresses enabling the

attractive London-van der Waals forces to play a more domi-

nate role, and in certain circumstances these forces may

overcome electrostatic repulsion and result in an attractive

energy interaction (McDowell-Boyer et al., 1986). For solutions

composed of divalent and trivalent ions, the increase in colloid

attachment can occur at substantial lower IS compared to

mono-valent ions (Ryan and Elimelech, 1996). A DLVO inter-

action energy profile is constructed as the total interaction

energy as a function of separation distance between a colloid

andsediment grain surfaceor betweencolloids.A typicalDLVO

energy profile (low IS) of like-charged colloid and sediment

grains (or colloids to colloids) is characterized by a deep

attractive well (primary minimum) at a small separation

distance, a sizable energy barrier, and a shallow attractivewell

(secondaryminimum) at a larger distance (Shen et al., 2008). As

the IS increases, there is a point at which the energy barrier to

particleemedia interaction and attachment is eliminated

basedon the surfaces and solutionchemistry (McDowell-Boyer

et al., 1986). Deposition rates of colloids on sediment grains in

0.0001 0.001

Compere et al., 2001

Ko and Elimelech, 2000

Nocito-Gobel and Tobiason, 1996

Johnson et al., 2007

Kretzschmar et al., 1997 (CaCl2)

Kretzschmar et al., 1997 (NaCl)

Kuhnen et al., 2000

Bradford et al., 2007

Saiers and Ryan, 2006

Tufenkji and Elimelech, 2005

Grolimund et al., 2001 (CaCl2)

Grolimund et al., 2001 (NaCl)

Current Study

Fig. 1 e The ionic strengths of the solutions that were used in p

transport was investigated in solutions of a single, mono-valen

natural saline solutions of DSW containing a variety of salts we

saline than that of all other studies were used. (For interpretation

referred to the web version of this article.)

limited; whereas at high IS, as the repulsion energy barrier

disappears every collision results in attachment and deposi-

tion kinetics is fast and transport-limited (Elimelech et al.,

1995). The relative deposition (colloidecollector) and aggrega-

tion (colloidecolloid) rate constants follow very similar trends,

featuring fast (favorable) and slow (unfavorable) regimes at

high and low salt concentrations, respectively (Grolimund

et al., 2001; Kretzschmar et al., 1997).

Extensive research in recent years has focused on the

improvement of the filtration theory by incorporating

processes such as straining (entrapment of colloids within

small pores and within pore-space constrictions associated

with grainegrain contacts) (Bradford et al., 2003, 2006a,b, 2007;

Shen et al., 2008; Torkzaban et al., 2008a), hydrodynamic

effects (inhibition of particle deposition at high flow rates)

(Ahfir et al., 2007; Bradford et al., 2006a). Research has also

sought to account for the distribution in the interaction

energies between particles (Tufenkji and Elimelech, 2005).

Somework has also referred to the role of retained particles in

the dynamics of colloid deposition in porous media (Ko and

Elimelech, 2000; Kretzschmar et al., 1997; Kretzschmar and

Sticher, 1997; Liu et al., 1995; Song and Elimelech, 1993).

Particle transport is typically studied on the column scale

usingwell-defined, simplemodel systems of uniformparticles

and sediment, low IS solutions, and a single type of electrolyte

(Bradford et al., 2007; Compere et al., 2001; Grolimund et al.,

2001; Kretzschmar et al., 1997; Nocito-Gobel and Tobiason,

1996). Particle release from packed columns has also been

extensively studied in relatively high salinity environments

(Blume et al., 2002, 2005; Khilar and Fogler, 1984; Khilar et al.,

1983; Roy and Dzombak, 1996). Colloid detachment is a rather

complicated process, but in general it is related to the

decrease of solution salinity below a threshold value at which

0.01 0.1 1 10

IS (molar)

revious work on colloid transport. In these studies colloid

ce salt, unless specified otherwise. In this study (blue bar)

re used. Also high salinities on an order of magnitude more

of the references to color in this figure legend, the reader is

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 2 3523

particles are released from the sediment (Blume et al., 2005;

Khilar and Fogler, 1984; Roy and Dzombak, 1996; Fig. 1).

Single-salt solutions typically used for exploring colloid

deposition in porous media (Bradford et al., 2007; Compere

et al., 2001; Grolimund et al., 2001; Johnson et al., 2007; Ko

and Elimelech, 2000; Kretzschmar et al., 1997; Kuhnen et al.,

2000; Nocito-Gobel and Tobiason, 1996; Saiers and Ryan,

2006; Tufenkji and Elimelech, 2005) do not adequately mimic

the complex nature of groundwater containing a mixture of

salts. Moreover, previous studies have focused mainly on low

(up to w1 M) IS solutions. The current work studies the

transport of colloid in the naturally extreme saline solution of

the Dead Sea water (DSW, Table 1) with an IS of 8.5 M.

Furthermore, studying colloid transport in a wide range of

salinities and in a complex mixture of salts has been achieved

by diluting the DSW. To the best of our knowledge the afore-

mentioned conditions (wide range of salinities of natural and

complex solutions) have not been examined before.

The importance of colloid transport in DSW, beyond purely

scientific interest, is its potential contribution to the creation

of sinkholes around the Dead Sea coast. Numerous collapsed

holes have developed along the Dead Sea shore causing

a safety risk to people and infrastructure. It is well accepted

that these sinkholes are generated as a result of subsurface

dissolution of salt layers (Yechieli, 2006). This process is

governed by the eastward and downward movement of the

DSWefresh groundwater interface in response to the drop in

the Dead Sea water level (Abelson et al., 2003; Yechieli, 2006).

Consequently, the subsurface salt layers have been exposed

to relatively fresh groundwater. In that dynamic environ-

ment, where the groundwater salinity changes over a time-

scale of months (Kiro et al., 2008), particle release and

migration from the sediments was suggested as a supporting

mechanism for the creation of the sinkholes (Arkin and Gilat,

2000). This is the motivation for this current work, which is

aimed at exploring colloid transport in high salinity DSW

using column experiments (on two scales). With experi-

mental data and complementary mathematical modeling, the

goal was to identify the potential for colloid transport in the

complex solution of the DSW and, for the first time, compare

between results from single-salt experiments to those of the

natural system.

2. Materials and methods

2.1. Solution chemistry

Dead Sea Water (DSW) is of marine-evaporitic origin

with ionic strength of 100.9 M (ten times more concentrated

Table 1 e Chemical analysis of undiluted Dead Sea water (DSWstrength, and total dissolved solids (TDS).

Na K Ca Mg Sr Cl

mM/L

1416 189 873 3716 8 6065

than ocean water). DSW is typically Ca-chloridic brines,

displaying the equivalent concentration relationship of

Ca > (SO4 þHCO3), Na/Cl < 0.86 and low SO4/Cl ratios relative

to those of sea water due to precipitation of salts and several

watererock interactions (Starinsky, 1974). In all the experi-

ments, DSW taken from the En Gedi shore was used after

filtration through a 0.45 mm filter. The DSW filtering proce-

dure has been tested intensively in different experimental

procedures including under the microscope and the filtered

DSW did not contain any natural colloids. For certain exper-

iments, noted below, DSW was diluted with deionized water

(DI) water as follows: DSW/2 (i.e. 50% concentration of the

DSW), DSW/5 (20% DSW), DSW/10 (10% DSW), DSW/100 (1%

DSW), DSW/1000 (0.1% DSW), and DSW/5000 (0.02% DSW).

The dilution process of the DSW increases the pH of the

solution (Amit and Bentor, 1971) and the pH of DSW solutions

that have been diluted by 100 (DSW/100) increases to 7.6

compared to pH w5.4 of undiluted DSW solution (Table 1).

The increase of pH through the dilution of the DSW is an

innate chemical property of saline solutions that carriesmore

than 1.2e1.5% of total dissolved solids and is suspected to be

connected to increased dissociation of bicarbonates salts

(Amit and Bentor, 1971).

2.2. Colloids

Fluorescent Carboxylate-Modified Latex (CML) microspheres

of 1 mmdiameter and excitation/emissionwavelengths of 505/

515 or 540/560 were used for visualization (FluoSpheres�,

Molecular probe�, Eugene, Oregon). CML have often been used

as model colloids due to their spherical and well-defined size

and the ease in detection at low concentrations (Shani et al.,

2008; Zvikelsky and Weisbrod, 2006). In these experiments,

CML concentration of 3.64 � 107 colloids per ml (equivalent to

2 ppm) was used. The concentration was analyzed using

fluorescence spectrophotometry (Cary Eclipse Fluorescence

Spectrophotometer, Varian�, Palo Alto, CA). Fluorescence

intensity is slightly influenced by the solution IS, as was also

demonstrated for fluorescent dye solutions (Magal et al., 2008);

therefore, in each experiment, the concentration of colloids

was determined utilizing linear calibration curves developed

for the same ionic strength and composition as the test

solution used (Magal, 2011).

2.2.1. Conservative solute tracerAs a reference for the colloid transport, SodiumNaphthionate

(Naph) was used as a conservative solute tracer. A previous

study carried out on DSW solutions showed that Naph, with

excitation/emission of 320/420, behaves as a conservative

tracer even in highly saline brines (Magal et al., 2008).

) including the concentration of key salts, total ionic

SO4 Br Ionicstrength (M)

Totaldissolvedsolids (g/L)

8 70 8.5 324

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 23524

Synchronous analysis proved a lack of spectrographic distur-

bances between the Naph and the CML used. The Naph was

purchased as a dry powder (Fluka Chemika) andwas dissolved

in DI. In the experiment, solutions of different IS were

prepared with the Naph tracer concentration of 150 ppb

(6 � 10�10 M). The solute tracer was analyzed by fluorescence

spectrophotometry (Cary Eclipse Fluorescence Spectropho-

tometer, Varian�, Palo Alto, CA). The tracer concentration was

determined by transforming fluorescence results to concen-

tration using calibration curves in solutions of similar

compositions and salinities.

2.3. Porous media

Natural sand originating from the Hatira Formation (Creta-

ceous) exposed in theMachteshRamon (Israel)wasused as the

granular porous media. The sand is commercially distributed

by Machteshim LTD (Mizpe Ramon, Israel) after sieving to an

average diameter of 0.8 mm and uniformity coefficient of 1.18

(d60/d10). It is composed of highly pure quartz sand (more than

99% SiO2), typically with <400e500 ppm of metal impurities

(mostly iron, aluminum and titanium oxides). Sand pretreat-

ment and purification procedures included washing with

dilute, heated solutions of HCl, washing and rinsing with DI

water, and drying in the oven at 50 �C (Shani et al., 2008).

Following the pretreatment, the quartz grains were packed in

columns of two sizes, one a 7 cm-long polypropylene column

with a 1.4 cm inner diameter, and the other a 30 cm-long

Plexiglas� column with a 5.5 cm inner diameter. The columns

weredry-packedand the calculatedporosity of the sandwasof

37e38%.

2.4. Colloid stability experiment

To assess the stability of the CML colloids used in the column

experiment with respect to settling or floating, an experiment

was conducted as a function of the density of solution and

under hydrostatic conditions. Solutions of 2 ppm colloid

concentration (3.64 � 107 colloids per ml) were prepared in

DSW and DSW diluted to DSW/2, DSW/5, DSW/10, DSW/100

and DSW/1000 with DI. One L of the colloid suspension was

placed in a beaker with a 10 cm depth of the water column.

The beaker was strongly agitated to ensure a uniform colloid

suspension prior to the initiation of the hydrostatic condi-

tions. Next, samples of 2 mL were taken periodically (after 4,

14, 20, 30, 40, 55, 70, 95, 120, 145, 175, 205, 225 and 280 min)

from the solution at three depths (<0.2 cm from the surface,

2.5 cm and 5 cm from the surface) in order to determine the

extent of vertical transport of the CML colloids in the beaker.

2.5. Colloid transport experiments

2.5.1. 30 cm columnsSolutions used in the long-columnexperimentswereDSWand

diluted DSW solutions (DSW/2, DSW/5, DSW/10, DSW/50,

DSW/100, DSW/500, DSW/1000 and DSW/5000). At least two

repetitions of the column experiments were made for each

type of solution. The columns were set-up vertically, dry-

packed homogenously, and then first saturated from the

bottom with CO2 for 20 min. The column inlet at the bottom

was connected to a three-point valve leading to two reservoirs,

one containing the colloid suspension (colloids and conserva-

tive solute in DSW) and the other containing a colloid and

tracer-free solution of DSW. The solution with colloids and

tracer was continuously stirred in an opaque reservoir vessel,

preventing light from entering the solution. Both solutions

(with and without tracer/colloids) were pumped at a constant

rateof 1� 0.05mL/minbyaperistaltic pump (GilsonMinipuls 3,

Gilson�, Middleton, Wisconsin). Prior to the experiment, the

column was flushed by the background solution (according to

the salinity used in each experiment) with a solution volume

identical tow3porevolumes (PV).Next, the tracer solutionwas

diverted into the column andwas injected for 10e12 h (2.2e2.3

PVs) before the experiment was terminated. A fraction

collector (Spectra/Chrom� CF-1, Spectrum Laboratories,

Houston, TX or Gilson FC 203B, Middleton, Wisconsin) was

connected to the column outlet. Samples of 5 mL were

collected from the column outflow at different frequencies

over the course of the experiment every 15 min for the first

3.5 h, every 5min for another 2.5 h, and then every 15min until

the end of the experiment. The 5mL sampleswere collected in

10mL opaque glass vials andwere stored at 4 �C until analysis,

typically within 24 h from the end of the experiment. The flow

rate in the columnwasmeasured periodically by weighing the

effluent solution over a known time interval and accounting

for solution density. The experiments were conducted at room

temperature (25 � 2 �C) and under dark conditions in order to

prevent tracer photo-degradation.

2.5.2. 7 cm columnsExperiments were conducted for longer time periods (at least

100 h) in 7 cm columns. In these column experiments two

solutions were used: DSW/5000 (IS ¼ 10�2.8 M) and slightly

diluted DSW (90% DSW and 10% distilled water IS ¼ 100.9 M).

A slight dilution of the DSW solutionwas necessary to prevent

the precipitation of salts during the long experimental period.

At least two repetitions were made for each solution chem-

istry condition tested. Apart from the columns’ size differ-

ence, the experimental procedure was conducted under the

same conditions as those of the 30 cm column experiments

described in Section 2.5.1. The procedure deviated only in the

following ways: prior to the experiment, the column was

flushed by the background solution for 7e8 PVs (versus the 3

PV of rinsing for the 30 cm-long column experiment) and the

tracer solutions were applied for 100 h, (several thousand PVs)

in contrast to w2 PV in the 30 cm-long column. Effluent

collection was conducted as follows: once every 2 min during

the first hour (2 mL aliquots), once every 5 min in the second

hour (volume of 5 mL), and once every 9 min in the third hour

(volume of 9 mL). Subsequently, the sampling frequency was

decreased to once every 36 min for 8 h, followed by samples

collected once every 100 min till the experiment terminated.

2.6. Mathematical models

Two types of models were used in order to simulate the exper-

imental results that are described below. The HYDRUS-1D

model (Simunek et al., 2008) was used for simulating the

results from the 30 cm-length column experiments. The

experimental results on two scales (7 and 30 cm experiments)

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 2 3525

were modeled using the more sophisticated model of limited

entrapment (Pachepsky et al., 2006) assuming bimodal deposi-

tion coefficients of the colloids (as will be elaborated on below).

2.6.1. HYDRUS-1D simulationsThe HYDRUS-1D computer code (Simunek et al., 2008), that

simulates water, colloid, and solute movement in one-

dimensional saturated porous media, was used for modeling

the experimental results. HYDRUS-1D numerically solves the

Fickian-based advectionedispersion equation with a non-

linear equilibrium and kinetic reactions. The code is coupled

to a non-linear least-square optimization routine to find

model parameters by fitting a simulated BTC to experimental

BTC data. The transport of colloids through the sand columns

was formulated as (Simunek et al., 2008):

vCvt

þ rb

ne

vSvt

¼ aLvv2Cvx2

� vvCvx

(1)

where C is the colloid concentration in the solution (Nc/L, Nc

denotes number of colloids), S is the concentration of deposited

colloids inthecolumn(Nc/kg), t is thetime (min),x is thedistance

(cm), aL is the dispersivity (cm), rb is the soil bulk density (kg/

dm3), ne is effective porosity, and v is thefluid velocity (cm/min).

The mass balance for the colloids attached to the solid

phase is given by (Simunek et al., 2008):

rbvSvt

¼ nekattjsC� rbkdetS (2)

where katt is the colloid deposition coefficient (1/min), kdet is

a first-order detachment rate coefficient (1/min) and js is

a dimensionless function for deposited colloids. In order to

simulate reduction in the attachment coefficient due to filling

sorption sites on the sediment grains, js decreases with

increasing adsorbed colloid mass as follows (Simunek et al.,

2008):

js ¼Smax � SSmax

(3)

where Smax is the maximum concentration of deposited

colloids (on the solid phase, Nc/kg). It should be noted that the

impact of Smax is mainly in advanced stages of the experiment

after an injection of >2e3 PV of particle solution. The impact

of Smax value is elaborated on below.

2.6.2. Limited entrapment model simulationsThe limited entrapment model (Pachepsky et al., 2006) was

used for simulation of the experiments conducted in the 7 cm

column. The model was developed for simulating transport of

bio-colloid (bacteria) with bimodal breakthrough curves. The

main assumption of the model (Pachepsky et al., 2006) is the

existence of a limited attachment capacity of the sediments

such that no bio-colloid attachment occurs after this capacity

is reached. According to the model equations, the bio-colloid

attachment rate is assumed to be a function of the number

of previously attached bio-colloids. However, the attachment

rate drops again as capacity for further attachment is excee-

ded. Consequently, the first maximum (or plateau) observed

in the BTCs is consistent with the initial slow attachment rate

(Pachepsky et al., 2006), while the second maximum (or

plateau) is consistent with achieving attachment capacity.

The mathematical model describes the bio-colloid trans-

port according to Equation (1) with themass balance equation

for the bio-colloid attached to the solid phase given by:

rbvSvt

¼ neðka1 þ ka2SaÞjsC� rbkdetS (4)

where ka1 [min�1] is the initial attachment rate that is appli-

cable at early stages of transport when transport-related

attachment has occurred, ka2 is the parameter responsible

for the increase of the attachment rate with the increase in

attached bacteria amount; a is a parameter which accounts

for the acceleration of the attachment with the increase in the

trapped amount of particles. The trapping rate grows linearly

with the trapped amount when a ¼ 1. The trapping rate

accelerates with the growth of S when a> 1. The trapping rate

shows some retardation when a < 1 (Pachepsky et al., 2006).

Note, that the original Pachepsky et al. (2006) model has been

modified slightly by introducing a dimensionless colloid

retention function js similar to that in Equation (2). Thus, if

ka2 ¼ 0 the limited entrapment model is similar to that of the

HYDRUS-1D. The least-square optimization was carried out to

calculate the colloid transport parameters by fitting themodel

to the experimental results (inverse solution).

3. Results and discussion

3.1. Colloid stability

In diluted DSW experiments (DSW/5000, IS ¼ 10�2.8 M), the

concentration of colloids in the beaker was relatively uniform

and stable over the 5 h duration of the experiment (with

a coefficient of variance <3% for colloid concentration along

the beaker). The visual appearance of colloids uniformity

suspended in solution implies that colloids in the water

column during the course of the experiment are quite stable

and that gravitational settling is insignificant. The concen-

tration of colloids was less uniform with time in the DSW

experiment (IS ¼ 100.9 M). Specifically, an increase in colloid

concentration (w20e30% higher value in comparison to the

initial concentration) was observed at a depth of 0.2 cm after

3 h. A similar increase of colloid concentration was observed

following 2 more hours, 2.5 cm below the water surface. It

should be noted that at the same time, colloid concentration

at the 5 cm depth remained quite constant across the entire

5 h experiment (with a coefficient of variance of 4.5% with

time for colloid concentration with no distinctive trend).

Upward transport of the colloids (density of 1.05 g/cm3) in the

relatively dense solution of DSW (1.25 g/cm3) may occur;

however, upward colloid movement could only be detected

after relatively long time periods (3e5 h), suggesting any

additional buoyancy of the colloids occurs at a relatively slow

pace. This observation is supported by the low Stokes velocity

(0.01 cm/h) for 1 mm colloids in DSW solution.

Toconfirmwhether thebuoyancyof thecolloids inthehigh IS

solution contributed to transport in the column, BTCs of exper-

iments conducted with both upward and downward vertical

flows were compared. Quite similar BTCs were measured in

these experiments (data not shown); hence, it was concluded

that this mechanism is negligible in the dense DSW solution.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 23526

3.2. Colloid transport experiments

3.2.1. Transport in 30-cm columnsBTCs for colloids and the conservative tracer in solutions

ranging across three orders of magnitude of IS are presented

in Fig. 2. In the diluted solution, the colloids arrived at the

outlet w0.2 PV (45 min) before the conservative tracer

(IS ¼ 10�2.8 M or DSW/5000, Fig. 2a). The extent of early arrival

of the colloids is reduced with the increase of IS, and in the

concentrated DSW solution (IS ¼ 100.9 M, Fig. 2e) colloids and

the conservative tracer arrived simultaneously. Early arrival

of the colloids was detected in many studies (e.g., Bradford

et al., 2003, 2006a; Kretzschmar et al., 1997; Mishurov et al.,

2008; Nocito-Gobel and Tobiason, 1996; Shani et al., 2008) as

was also the lack of correlation between preferential transport

of colloids and the increase of IS (Harter et al., 2000; Nocito-

Gobel and Tobiason, 1996). The transport velocity of colloids

is known to be higher than that of the average water velocity

inmany cases. This occurs because not all pores are accessible

for the colloids, which are excluded from the smaller pore

throats (Harter et al., 2000). Subsequently, smaller effective

pore volumes are available for colloid transport as compared

to solution transport (Bradford et al., 2003, 2006a; Kretzschmar

et al., 1997; Nocito-Gobel and Tobiason, 1996). Physical

straining is unlikely considering the sand grain and colloid

sizes (Bradford et al., 2006a). The early arrival of the colloids is

more pronounced at low IS, due to the expansion of the

electrical double layer at the solid surface (Harter et al., 2000).

The shape of the BTCs in the concentrated DSW experi-

ment is irregular, with the relative concentration (C/C0, C0

denotes concentration of colloids in the inlet solution) fluc-

tuating between 0.18 and 0.38 (Fig. 2e). This seemingly

Fig. 2 e BTCs of colloids and the conservative tracer in different s

IS [ 10L1.8 M, (c) DSW/100, IS [ 10L1.1 M, (d) DSW/10, IS [ 10

colloids BTCs.

unstable phenomenon has not been observed in the diluted

DSW experiments and is therefore considered to be related to

the high salinity. The observed irregular shape of the BTCmay

be due to sporadic release of varying amounts of colloid

aggregates in the column. The colloids in the concentrated

DSW are likely to aggregate, since there is no energy barrier to

colloid interaction at this high IS and every collision should

result in attachment (Grolimund et al., 2001). Therefore,

colloid aggregates may be presented in the injected suspen-

sion, as well as at the outlet due to creation of additional

aggregates occurring in the column.

An attemptwasmade to determine the colloid or aggregate

size by a few techniques (dynamic light scattering and visu-

alization under the microscope) with no conclusive results.

Under the light microscope colloid aggregates were found in

DSW solutions, while at the same time in the highly diluted

solution of artificial rain water (Zvikelsky andWeisbrod, 2006)

no aggregates has been inspected. This is attributed to the

difficulties resulting from the complexity of dealing with

suspensions in concentrated solutions such as DSW brines.

Other routine methods, such as settling in a column or laser

diffraction, could not be implemented for DSW solutions

without major and inherent adjustments, which are subjects

for future research.

Comparing the shape of the colloid BTCs at the various IS

conditions (Fig. 2) reveals that the breakthrough value (C/C0)

after w1.5e2 PV reaches a maximum of 0.6 for IS < 10�1.8 M

(DSW/500), while it isw0.3 for IS� 10�1.1 M (DSW/100) (Fig. 2f).

The corresponding colloid deposition rates were calculated

based on filtration theory (e.g., Yao et al., 1971; Logan et al.,

1995; Tufenkji and Elimelech, 2004) using the system phys-

ical parameters (such as collector diameter, column length,

olutions of DSW: (a) DSW/5000, IS [ 10L2.8 M, (b) DSW/500,L0.1 M, (e) DSW, IS [ 100.9 M, and (f) a compilation of all

Table 2 e Colloid transport parameters calculated byfitting the HYDRUS-1D model for results obtained by the30 cm column experiments.

IS (M) Exp.Num.

ne aLcm

Smax katt(min�1)

kdet(min�1)

R2

8.5 1 0.35 0.05 153 0.006 0 0.69

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 2 3527

porosity, fluid approach velocity). Here, the collector contact

efficiency was assessed accounting also for the fluid and

particles properties and the breakthrough curve plateau C/C0

(Tufenkji and Elimelech, 2004). Fig. 3 presents calculated

deposition rate coefficients as a function of the solution IS up

to IS ¼ 100.23 M (DSW/5). The deposition rate of colloids was

w0.002 m�1 for solutions up to IS of 10�1.8 M (DSW/500) and

increased to w0.005 min�1 for IS of 10�1.1 M (DSW/100) and

higher (Fig. 3). Such an increase in colloid deposition rate with

solution IS was expected according to the DLVO theory as the

greater IS reduces the energy barrier for deposition of colloids

(McDowell-Boyer et al., 1986).

Previous experiments demonstrated the existence of an IS

threshold value between the reaction-limited and transport-

limited deposition regimes, the critical deposition concen-

tration (CDC) (Grolimund et al., 2001; Kretzschmar et al., 1997).

The CDC is in the range of 10�1 M for NaCl solutions and

10�2e10�3 M for CaCl2 solutions (Grolimund et al., 2001;

Kretzschmar et al., 1997). Evaluation of the CDC in DSW is

complicated since it is composed of substantial cation

concentrations beyond simple Ca2þ and Naþ (Table 1). The

CDC defined by the concentration of the monovalent salts in

DSW is between 10�2.5 and 10�1.8 M (e.g., for solutions between

DSW/500 and DSW/100) and is two orders of magnitude lower

than the CDC of an artificial NaCl solution. Similarly, the CDC

defined by the concentration of the DSW divalent ions is

between 10�2.3 and 10�1.6 M and is on the same order of

magnitude as the artificial CaCl2 solution.

Grolimund et al. (2001) measured the CDC of colloids in

mixed solutions of NaCl andCaCl2, at different ratios. The CDC

of a mixed solution of Ca/Na with a molar ratio of 1.4 (similar

to the ratio of mono to divalent cations in DSW) was

w10�2.2 M, somewhat lower than the CDC of 10�1.9 M and

10�1.2 M in DSW/500 and DSW/100 suspensions respectively.

The similarity between the CDC values of a simple mixture of

salts and the complex composition of DSW indicates the role

of cation composition on the attachment of colloids. This

merits further study using natural (and complex) solutions

Fig. 3 e Average deposition rate of the colloids (dots) and

standard deviations (bars) at different salinities as

a function of the solution’s ionic strength in the 30 cm

columns.

with varied cation compositions. Ocean water has a high

salinity and a substantially different ionic composition with

a di- tomono-valent cation ratio of 0.13 (compared to the ratio

of 1.4 in DSW).

3.2.2. HYDRUS-1D simulation results for the 30-cm columnsThe above results from the 30 cm-long column experiments

were inversely simulated successfully using the HYDRUS-1D

model, as evidenced by the correlation coefficients (R2 of

0.69e1.00, Table 2). The colloid transport parameters derived

from the simulation are presented in Table 2. The fitted values

of the effective porosity was slightly lower than the measured

porosity (0.36e0.37 and 0.37e0.39, respectively), possibly due

to colloid size exclusion and the resulting phenomenon that

colloids can only migrate through and access a fraction of the

total voids. Colloid dispersivity was found to be on the same

order of magnitude as determined for the experiments (Table

2). The colloid attachment coefficient (katt) at high IS was

higher than at low IS, with the detachment coefficient (kdet)

being at least an order of magnitude smaller than the

attachment coefficient determined for the same solution

conditions (Table 2). Colloid attachment is therefore consid-

ered an irreversible process, as was indicated in other studies

(Compere et al., 2001; Johnson et al., 2007).

HYDRUS-1D model simulations were highly insensitive in

determining the maximum normalized concentration of

deposited colloids, Smax ¼ Smax=N0c (N0c is the number of

colloids in a unit volume of the influent colloid suspension

2 0.35 0.05 72 0.006 0 0.75

3a 0.36 0.05 6.1 0.009 0 0.87

4a 0.37 0.18 13.3 0.008 0 0.90

4.2 1 0.35 0.25 76 0.007 0 0.93

2 0.35 0.07 74 0.007 0 0.86

1.7 1 0.35 0.09 364 0.005 0 0.96

2 0.35 0.21 183 0.005 0 0.96

0.8 1 0.35 0.1 9 0.005 0 0.96

2 0.35 0.2 99 0.005 0 0.94

0.2 1 0.37 0.1 4 0.008 0.0003 0.97

2 0.35 0.1 55 0.007 0.0002 0.95

0.1 1 0.35 0.14 21 0.007 0.0002 0.98

2 0.35 0.11 11 0.006 0.0003 0.99

0.02 1 0.37 0.27 5 0.002 0 1.00

2 0.36 0.25 497 0.002 0 1.00

0.008 1 0.36 0.26 4 0.003 0 0.99

2 0.37 0.26 2 0.002 0.0007 0.99

0.002 1 0.36 0.26 1 0.002 0 1.00

2 0.36 0.27 30 0.002 0.0002 0.99

a Results of downward diversion of inlet solution experiment.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 23528

(Bradford et al., 2009)), as was demonstrated by two substan-

tially different resulting fitted values for two repetitions of the

same experiment (IS¼ 0.02 M or DSW/500, Table 2). In order to

understand the significance of Smax on colloid transport, the

HYDRUS-1D model was used for forward simulations of the

colloid BTCs. Simulations were conducted using the transport

parameters obtained from fitting the DSW experimental

results in the 30 cm-long column (aL¼ 0.2 cm, katt¼0.008min�1,

kdet¼ 5� 10�7 min�1) andwere repeatedwith several values of

Smax (0.02, 0.2, 1 and 10). The simulations were also conducted

for the 7 cm-long column, as described further below. The

forwardsimulations (Fig. 4) indicated that at lowSmax (0.02), the

BTC shape is similar to that of the conservative tracer (Fig. 4);

whereas, for higher Smax values, after the first increase to C/C0

of 0.6, the relative concentration gradually increases and

approaches a value of one. The value of Smax is related to the

time at which the C/C0 reaches a value of 1; for example, for

greater Smax value a longer elution time is needed to achieve C/

C0 ¼ 1 (Fig. 4). The forward simulations demonstrated the Smax

value influences the shape of BTC, especially at the advanced

stages of the experiment after eluting colloid suspension of

tens PV. The evaluation of Smax is highly important for under-

standing long term changes in colloid transport, especially for

the evaluation of these processes in natural environments in

time-scales of years rather than hours (or days) in the labora-

tory. Consequently, to accurately determine Smax the experi-

ment should be significantly longer. It was decided, therefore,

to repeat part of the experiments, using a small diameter

column of 7 cm length, to reduce the elution time of one pore

volume from hours (in the 30 cm-long column) to minutes (in

the 7 cm-long column). This way experiments could be con-

ducted over 100 s of PV over a reasonable period of time. These

experiments are discussed below.

3.2.3. Transport experiments in 7-cm columnsThe short column (7 cm) experiments were conducted at two

extreme salinity levels, low IS DSW (diluted by 5000, Fig. 5a)

and a solution of highly concentrated DSW (Fig. 5b).

Fig. 4 e HYDRUS-1D forward simulations of colloid

transport in DSW solution in a 7 cm-long column at

different Smax values (lines for values of 0.02e10) and

experimental results of colloid transport (circles).

Comparing longer duration experimental in the 7 cm

columns (Fig. 5a,b) and the shorter duration experiments on

the 30 cm columns (Fig. 5c,d) reveals some notable differ-

ences. First, the relative concentration of colloids eluted from

the 7 cm column after injection of 2e3 pore volumes was

greater than that from the 30 cm columns (C/C0 of 1 and 0.5

in the dilute experiment and 0.8 and 0.2 in the DSW experi-

ment, for 7 and 30 cm columns, respectively). The value of

C/C0 was confirmed to be directly related to the length of the

column or to the pore volume of the column (Brown and

Abramson, 2006). Another important trend was in the DSW

experiments in the 7 cm column, after the first initial

increase of the eluted colloids (C/C0) there was another

increase approaching a value of one observed after w18 PVs.

This two-step increase in C/C0 values could not be simulated

by the HYDRUS-1D, which predicted for an identical experi-

mental set-up a first initial breakthrough at C/C0 ¼ 0.6

followed by a more gradual increase of the C/C0 to the value

of one (Fig. 4). All the attempts to model these results with

the HYDRUS-1D model resulted in poor correlation coeffi-

cients. This bimodal behavior was observed in five repetitions

of the DSW experiments in the 7 cm columns (not presented).

Similar bimodal BTCs were reported previously for a few

types of bio-colloids: Cryptosporidium (Harter et al., 2000),

viruses (Jin et al., 1997) and bacteria (Pachepsky et al., 2006,

Tufenkji et al., 2003).

A bimodal shape of the colloid BTCs suggests a complex

deposition process occurring in two stages. At the beginning

of the experiment the fast increase of C/C0 up to 0.8 results

from a low initial rate of colloidal deposition. As the colloid

suspension continues to be injected (to approximately 18 PVs),

the deposition rate increases further, leading to flattening of

the BTC or even to a slight decrease of the relative concen-

tration (C/C0). Subsequently, the deposition rate decreases

and the eluted C/C0 increases again and approaches a value of

one, possibly when the entrapment capacity has been excee-

ded. The exact point of transition between these stages is

unknown. Jin et al. (1997) speculated that the second increase

of virus concentration (followed the achievement of an initial

steady-state value for w10 PV) reflects complete filling of the

available retention space on the sediment surfaces. The

available retention space for colloid deposition, however, is

not the entire surface of the sediment grains. The surface area

that is occupied by a monolayer of the colloids (deposited

during the 18 PV of solution injected in to the column) is less

than 0.1% of the total sediment grain surface area (10�3m2 and

3.5 m2, cumulative surface of colloids and surface area of the

sand, respectively). Jin et al. (1997) speculated that the active

surfaces for the bio-colloid (virus) deposition are correlated

with patches of positively charged sites on the edges of the

sediment grain that occupy a small fraction of the sediment

surface. Yet, in the high salinities of the DSW, these impurities

would be insignificant due to the overall favorable conditions

for colloid attachment. Also, Torkzaban et al. (2008b)

demonstrated the existence of hydrodynamically discon-

nected regions in which the majority of colloid deposition

takes place. A bimodal breakthrough curve may result, alter-

natively, from elution of colloid aggregates of two sizes that

are retarded differently by the column due to size exclusion

(Tufenkji et al., 2003).

Fig. 5 e Comparison between the results of colloid transport and model simulations of 7 cm-long and 30 cm-long columns:

(a) diluted solution of DSW/5000, 7 cm column, and (b) DSW solution, 7 cm column; (c) diluted solution of DSW/1000; 30 cm

column and (d) DSW solution, 30 cm column. The limited entrapment model simulation result is sketched as lines:

continuous line for the conservative tracer and dashed line for the colloids.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 2 3529

3.2.4. Modeling colloid BTCs using a limited entrapmentmodelThe conventional model (Equations (1) and (2)) did not provide

fair agreement for the observed BTC’s of colloids in solution

with high salinity after eluting more than 10 pore volumes.

Therefore, the limited entrapment model was applied to

perform inverse simulations of the BTCs obtained with 7 cm-

long columns. A good fit was obtained between simulated and

observed concentrations (Fig. 5), as evidenced by the high

correlation coefficients (R2 of 0.87e0.95, Table 3). Simulations

were carried out assuming the acceleration parameter, a, is

equal to 1. This was based on the sensitivity analysis

(Pachepsky et al., 2006) and the shape of the observed colloid

BTCs. Following the results of model sensitivity, as the value

of a decreases, less acceleration in colloid attachment occurs

and an almost unimodal BTC is seenwith a¼ 0.5. For a close to

1, a first peak of concentration appears, and for a > 1

Table 3 e Colloid transport parameters calculated by fitting thelong column experiments.

Ionicstrength (M)

Tracers

aL(cm)

ne R2 aL(cm)

ne

8.5 2.1 0.39 0.99 2.1 0.33

1.4 0.37 0.99 1.4 0.35

0.002 1.1 0.37 0.97 1.1 0.32

a pronounced bimodality of the BTC is observed (Pachepsky

et al., 2006). The shape of colloid BTCs observed in the

experiment for the 7 cm column with DSW is most similar to

those calculatedwhen a¼ 1. Therefore, we assumed the linear

growth in colloid deposition ratewith an increasing amount of

attached colloids. The prescribed a value was also used to

improve the uniqueness of the inverse solution. Thus, the

limited entrapment model includes one additional parameter

(ka2), which accounts for the enhanced deposition rate with

the increase of the deposited particles concentration.

Model parameters found by best fit are presented in Table 3.

Large values of calculated dispersivity could be related to the

mixing in a relatively large volume of solution sample (2 mL)

compared to PV (w4.3mL). The values of the detachment (kdet)

and the initial attachment (ka1) rates in the experiments with

DSWare an order ofmagnitude larger than those found for the

longer column, while Smax is smaller for the 7 cm column

limited entrapment model for results obtained by the 7 cm-

Colloids

kdet(min�1)

ka1(min�1)

ka2(min�1)

Smax R2

0.0088 0.027 0.444 0.54 0.85

0.0015 0.028 0.363 0.35 0.89

0.0000 0.028 0 18.01 0.91

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 23530

(Table 3). In the experiment with DSW/5000 it was determined

that ka1 ¼ 0.028 min�1, which is similar to the values found in

the experiments with DSW. Such high values of this param-

eter are explained by the substantially greater flow velocity in

the 7 cm-long column experiments (0.74 cm/min) in compar-

ison to the velocity in the 30 cm-long column experiments

(0.046 cm/min). The theoretical value of ka1 ¼ 0.0157 min�1

(calculated using the filtration theory for DSW/5000) is almost

half. Bradford et al. (2006a) noted that the predicted attach-

ment rate may significantly underestimate deposition with

increasing size of the sand or colloids, thus suggesting that the

collector contact efficiency is not a complete descriptor of

attachment and another mechanism may be involved in the

colloid deposition.

The values of ka2 are responsible for the increasing depo-

sition rate with time and are presented in Table 3. They were

estimated from the colloid breakthrough curves using a non-

linear optimization procedure. Constant or increased colloid

deposition rates with time at high ionic strength are observed

when attached colloids can act as additional collectors for the

attachment leading to the formation of multilayer films on

matrix surfaces (e.g., filter ripening) (Rajagopalan and Chu,

1982; Liu et al., 1995; Kuhnen et al., 2000). Kuhnen et al.

(2000) developed a model that accounts for the additional

deposition rate of colloid particles to previously deposited

colloids attached to matrix surfaces. They demonstrated that

at moderately high IS values (up to 10�1 M), the model accu-

rately describes multilayer deposition assuming constant rate

of particleeparticle deposition. Unlike the model of Kuhnen

et al. (2000), Equation (4) accounts for increase in the overall

deposition rate with time due to an increase in concentration

of the deposited colloids and, as a result, an increase in the

effective surface area available for colloid capture. Modeling

results demonstrate that at very high IS values the ripening

mechanism could be important.

4. Conclusions

Colloid transport was found to decrease with greater solution

IS, in natural solutions with salinities ranging widely from the

hyper-saline brines of the Dead Sea to diluted solutions

similar to fresh groundwater. Colloid transport on a column

scale was found to be non-negligible even in the brines of

DSW, and 30e90% of the colloids are eluted from the 30 and

7 cm-long columns, respectively. The current experimental

results were found to be comparable to previously conducted,

similar experiments using artificial solutions of a single salt

(Grolimund et al., 2001; Kretzschmar et al., 1997). Like the

single-salt solution studies, a threshold value of’ IS was found,

above which the colloid deposition rate is constant (achieved

at the critical deposition concentration, CDC). It was found

that comparing the CDC of natural (and complex) solutions of

the DSW to an artificial single-salt solution is possible only by

accounting for the sum of the mono- or di-valent cations in

solution. The values are generally in agreement, although the

ratio between the mono- and di-valent ions in the solution

should be taken into account due to its major impact on the

CDC of the solution (Grolimund et al., 2001).

In the 7 cm-long column experiments that were long

enough to enable achieving a steady-state (C/C0 ¼ 1), colloid

transport in high salinities of DSW was found to be complex

and composed of two stagese low initial attachment followed

by an even lower stage of deposition rate. The bimodal shape

of the colloid breakthrough curve was found to be quite

similar to previously reported breakthrough curves of bio-

colloids in certain situations. The HYDRUS-1D model could

only simulate the colloid transport experimental results

before the achievement of C/C0 ¼ 1, but failed in simulating

the bimodal BTC of the colloids in DSW. These experimental

results were successfully simulated only by using the limited

entrapment model (Pachepsky et al., 2006) developed for the

transport of bacteria through a porous media.

The current work shows that colloid deposition involves

a two-step process in high salinities and that the explora-

tion of colloid migration using a single-salt solution cannot

truly represent natural conditions. Further work is needed

in order to understand the role of the solution salt compo-

sition on the transport of colloids, especially the role of

the balance between mono- and di-valent cations in the

solution.

r e f e r e n c e s

Abelson, M., Baer, G., Shtivelman, V., Wachs, D., Raz, E., Crouvi, O.,Kurzon, I., Yechieli, Y., 2003. Collapse-sinkholes and radarinterferometry reveal neotectonics concealed within the DeadSea basin. Geophysical Research Letters 30 (10), 1545. doi:10.1029/2003GL017103.

Ahfir, N.-D., Wang, H., Benamar, A., Alem, A., Massei, N.,Dupont, J.-P., 2007. Transport and deposition of suspendedparticles in saturated porous media: hydrodynamic effect.Hydrogeology Journal 15, 659e668. doi:10.1007/s10040-006-0131-3.

Amit, O., Bentor, Y.K., 1971. pH-dilution curves of saline waters.Chemical Geology 7, 307e313.

Arkin, Y., Gilat, A., 2000. Dead Sea sinkholes e an ever-developinghazard. Environmental Geology 39, 711e722. doi:10.1007/s002540050485.

Blume, T., Weisbrod, N., Selker, J.S., 2002. Permeability changes inlayered sediments: impact of particle release. Ground Water40, 466e474. doi:10.1111/j.1745-6584.2002.tb02530.x.

Blume, T., Weisbrod, N., Selker, J.S., 2005. On the critical saltconcentrations for particle detachment in homogeneous sandand heterogeneous Hanford sediments. Geoderma 124,121e132. doi:10.1016/, Geoderma.2004.04.007.

Bradford, S.A., Simunek, J., Bettahar, M., Van Genuchten, M.T.,Yates, S.R., 2003. Modeling colloid attachment, straining, andexclusion in saturated porous media. Environmental Scienceand Technology 37, 2242e2250. doi:10.1021/es025899u.

Bradford, S.A., Simunek, J., Bettahar, M., Van Genuchten, M.T.,Yates, S.R., 2006a. Significance of straining in colloiddeposition: evidence and implications. Water ResourcesResearch 42. doi:10.1029/2005WR004791, pp. W12S15.1eW12S15.16.

Bradford, S.A., Simunek, J., Walker, S.L., 2006b. Transport andstraining of E. coli O157:H7 in saturated porous media. WaterResources Research 42, W12S12. doi:10.1029/2005 WR004805.

Bradford, S.A., Torkzaban, S., 2008. Colloid transport andretention in unsaturated porous media: a review of interface-,collector-, and pore-scale processes and models. Vadose ZoneJournal 7, 667e681. doi:10.2136/vzj2007.0092.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 2 3531

Bradford, S.A., Torkzaban, S., Walker, S.L., 2007. Coupling ofphysical and chemical mechanisms of colloid straining insaturated porous media. Water Research 41, 3012e3024. doi:10.1016/j.watres.2007.03.030.

Bradford, S.A., Torkzaban, S., Leij, F., �Simunek, J., vanGenuchten, M.T., 2009. Modeling the coupled effects of porespace geometry and velocity on colloid transport andretention. Water Resources Research 45, W02414. doi:10.1029/2008WR007096.

Brown, D.G., Abramson, A., 2006. Collision efficiency distributionof a bacterial suspension flowing through porous media andimplications for field-scale transport. Water Research 40,1591e1598. doi:10.1016/j.watres.2006.02.016.

Compere, F., Porel, G., Delay, F., 2001. Transport and retention ofclay particles in saturated porous media. Influence of ionicstrength and pore velocity. Journal of Contaminant Hydrology49, 1e21. doi:10.1016/S0169-7722(00)00184-4.

Elimelech, M., Gregory, J., Jia, X., Williams, R.A., 1995. ParticleDeposition and Aggregation e Measurement, Modeling andSimulation. Elsevier, 441 p.

Grolimund, D., Elimelech, M., Borkovec, M., 2001. Aggregation anddeposition kinetics of mobile colloidal particles in naturalporous media. Colloids and Surfaces A: Physicochemical andEngineering Aspects 191, 179e188. doi:10.1016/S0927-7757(01)00773-7.

Harter, T., Wagner, S., Atwill, E.R., 2000. Colloid transport andfiltration of Cryptosporidium parvum in sandy soils and aquifersediments. Environmental Science and Technology 34, 62e70.doi:10.1021/es990132w.

Jin, Y., Yates, M.V., Thompson, S.S., Jury, W.A., 1997. Sorption ofviruses during flow through saturated sand columns.Environmental Science and Technology 31, 548e555. doi:10.1021/es9604323.

Johnson, W.P., Li, X., Assemi, S., 2007. Deposition and re-entrainment dynamics of microbes and non-biologicalcolloids during non-perturbed transport in porous media inthe presence of an energy barrier to deposition. Advances inWater Resources, Biological Processes in Porous Media: Fromthe Pore Scale to the Field 30, 1432e1454. doi:10.1016/j.advwatres .2006.05.020.

Khilar, K.C., Fogler, H.S., 1984. The existence of a critical saltconcentrationforparticle release. JournalofColloidandInterfaceScience 101, 214e224. doi:10.1016/0021-9797(84)90021-3.

Khilar, K.C., Fogler, H.S., Ahluwalia, J.S., 1983. Sandstone watersensitivity: existence of a critical rate of salinity decrease forparticle capture. Chemical Engineering Science 38, 789e800.

Kiro, Y., Yechieli, Y., Lyakhovsky, V., Shalev, E., Starinsky, A.,2008. Time response of the water table and saltwatertransition zone to a base level drop. Water Resources Research44. doi:10.1029/2007WR006752.

Ko, C.-H., Elimelech, M., 2000. The “Shadow Effect” in colloidtransport and deposition dynamics in granular porous media.Measurements and Mechanisms: Environmental Science andTechnology 34, 3681e3689. doi:10.1021/es0009323.

Kretzschmar, R., Barmettler, K., Grolimund, D., Yan, Y.-D.,Borkovec, M., Sticher, H., 1997. Experimental determination ofcolloid deposition rates and collision efficiencies in naturalporous media. Water Resources Research 33, 1129e1137.

Kretzschmar, R., Borkovec, M., Grolimund, D., Elimelech, M., 1999.Mobile subsurface colloids and their role in contaminanttransport. Advances in Agronomy 66, 121e193.

Kretzschmar, R., Sticher, H., 1997. Transport of humic-coated ironoxide colloids in a sandy soil: Influence of Ca2þ and tracemetals. Environmental Science and Technology 31,3497e3504.

Kuhnen, F., Barmettler, K., Bhattacharjee, S., Elimelech, M.,Kretzschmar, R., 2000. Transport of iron oxide colloids inpacked quartz sand media. Monolayer and Multilayer

Deposition: Journal of Colloid and Interface Science 231,32e41. doi:10.1006/jcis.2000.7097.

Liu, D., Johnson, P.R., Elimelech, M., 1995. Colloid depositiondynamics in flow through porous media: role of electrolyteconcentration. Environmental Science and Technology 29,2963e2973. doi:10.1021/es00012a012.

Logan, B.E., Jewett, D.G., Arnold, R.G., Bouwer, E.J., O’Melia, C.R.,1995. Clarification of clean-bed filtration models. Journal ofEnvironmental Engineering 21, 869e873.

Magal, E., Weisbrod, N., Yakirevich, A., Yechieli, Y., 2008. The useof fluorescent dyes as tracers in highly saline groundwater.Journal of Hydrology 358, 124e133. doi:10.1016/j.jhydrol. 2008.05.035.

Magal, E., 2011, Flow and Transport Processes in an ExtremelySaline Aquifers, Example from the Dead Sea Area, Ph.D. thesis.Ben-Gurion University of the Negev, Beer-Sheva, Israel.

McDowell-Boyer, L.M., Hunt, J.R., Sitar, N., 1986. Particle transportthrough porous media. Water Resources Research 22,1901e1921.

Mishurov, M., Yakirevich, A., Weisbrod, N., 2008. Colloid transportin a heterogeneous partially saturated sand column.Environmental Science and Technology 42, 1066e1071. doi:10.1021/es071780f.

Nocito-Gobel, J., Tobiason, E., 1996. Effect of ionic strength oncolloid deposition and release. Colloid and Surfaces 107,223e231. doi:10.1016/0927-7757(95)03340-8.

Pachepsky, Y.A., Devin, B.A., Polyanskaya, L.M., Shelton, D.R.,Shein, E.V., Guber, A.K., 2006. Limited entrapment model tosimulate the breakthrough of Arthrobacter and Aquaspirillum insoil columns. International Agrophysics 20, 207e218.

Rajagopalan, R., Chu, R.Q., 1982. Dynamics of adsorption ofcolloidal particles in packed beds. Journal of Colloid andInterface Science 86 (2), 299e317.

Roy, S.B., Dzombak, D.A., 1996. Colloid release and transportprocesses in natural and model porous media. Colloid andSurfaces 107, 245e262. doi:10.1016/0927-7757(95)03367-X.

Ryan, J.N., Elimelech, M., 1996. Colloid mobilization and transportin groundwater. Colloids and Surfaces A: Physicochemical andEngineering Aspects 107, 1e56. doi:10.1016/0927-7757(95)03384-X.

Saiers, J.E., Ryan, J.N., 2006. Introduction to special section oncolloid transport in subsurface environments. WaterResources Research 42, W12S01. doi:10.1029/2006WR005620.

Sen, K.T., Khilar, K.C., 2006. Review on subsurface colloids andcolloid-associated contaminant transport in saturated porousmedia. Advances in Colloid and Interface Science 119, 71e96.doi:10.1016/j.cis.2005.09.001.

Shani, C., Weisbrod, N., Yakirevich, A., 2008. Colloid transportthrough saturated sand columns: influence of physical andchemical surface properties on deposition. Colloids andSurfaces A: Physicochemical and Engineering Aspects 316,142e150. doi:10.1016/j.colsurfa. 2007. 08.047.

Shen, C., Huang, Y., Li, B., Jin, Y., 2008. Effects of solutionchemistry on straining of colloids in porous media underunfavorable conditions. Water Resources Research 44,W05419. doi:10.1029/2007WR006580.

Simunek, J., Sejna, M., Saito, H., Sakai, M., Van Genuchten, M.T.,2008. The HYDRUS-1D Software Package for Simulating theOne-dimensional Movement of Water, Heat, and MultipleSolutes in Variably-saturated Media. University of California-Riverside Research Reports, Riverside, California, p. 281.

Song, L., Elimelech, M., 1993. Dynamics of colloid deposition inporous media: modeling the role of retained particles. Colloidsand Surfaces A: Physicochemical and Engineering Aspects 73,49e62.

Starinsky, A. (1974) Relationship between Ca-chloride brines andsedimentary rocks in Israel, Ph.D. thesis. (in Hebrew, Englishsummary), Hebrew University, Jerusalem.

wat e r r e s e a r c h 4 5 ( 2 0 1 1 ) 3 5 2 1e3 5 3 23532

Torkzaban, S., Bradford, S.A., van Genuchten, M.T., Walker, S.L.,2008a. Colloid transport in unsaturated porous media: the roleof water content and ionic strength on particle straining.Journal of Contaminant Hydrology 96, 113e127. doi:10.1016/j.jconhyd.2007.10.006.

Torkzaban, S., Tazehkand, S.S., Walker, S.L., Bradford, S.A., 2008b.Transport and fate of bacteria in porous media: coupledeffects of chemical conditions and pore space geometry.Water Resources Research 44, W04403. doi:10.1029/2007WR006541.

Tufenkji, N., Redman, J.A., Elimelech, M., 2003. Interpretingdeposition patterns of microbial particles in laboratory-scalecolumn experiments. Environmental Science and Technology37 (3), 616e623. doi:10.1021/es025871i.

Tufenkji, N., Elimelech, M., 2004. Correlation equation forpredicting single-collector efficiency in physicochemical

filtration in saturated porous media. Environmental Scienceand Technology 38, 529e536. doi:10.1021/es034049r.

Tufenkji, N., Elimelech, M., 2005. Breakdown of colloid filtrationtheory: role of the secondary energy minimum and surfacecharge heterogeneities. Langmuir 21, 841e852. doi:10.1021/la048102g.

Yao, K.M., Habibian, M.T., O’Melia, C.R., 1971. Water and wastewater filtration concepts and applications. EnvironmentalScience and Technology 5, 1105e1112.

Yechieli, Y., 2006. The response of the groundwater system tochanges in the Dead Sea level. In: Enzel, Y., Agnon, A., Stein, M.(Eds.),NewFrontiers inDeadSeaPaleoenvironmental Research.Geological Society of America, Special Paper, pp. 113e126.

Zvikelsky, O., Weisbrod, N., 2006. Impact of particle size on colloidtransport in discrete fractures. Water Resources Research 42,W12S08. doi:10.1029/2006WR004873.