hols

Johnted Staf Geos

Article history:Received 23 July 2010Received in revised form 29 November 2010

Journal of Contaminant Hydrology 123 (2011) 1119

Contents lists available at ScienceDirect

Journal of Contami

e lse1. Introduction and motivation

Correct interpretation of concentration data from tracerexperiments conducted at eld sites is critical for under-standing solute transport processes in the subsurface.Intraborehole ow and resulting groundwater sampling biashave been previously recognized to occur in fully-screenedwells, and thus such wells were not recommended forgroundwater monitoring (Reilly et al., 1989; Lacombe et al.,

collected from fully-screened wells, and the characteristics ofsolute and groundwater age distributions due to intrabore-hole ow (e.g., Reilly et al., 1989; Church and Granato, 1996;Zinn and Konikow, 2007; Shalev et al., 2009; Mayo, 2010).However, if correctly interpreted, the water samples collectedfrom fully-screened wells can provide valuable informationabout groundwater ow and water quality distribution (e.g.,Sukop, 2000; Izbicki et al., 2005; Clark et al., 2006; Konikowand Hornberger, 2006; Landon et al., 2010), and often they1995; Church and Granato, 1996; Elci et al.fully-screened wells are commonly used dsite-specic knowledge of hydrogeologic cohigh cost of nested (multi-level) well instastudies have focused on illustrating the

Corresponding author. Tel.: +1 205 348 0579.E-mail address: czheng@ua.edu (C. Zheng).

0169-7722/$ see front matter 2010 Elsevier B.V.doi:10.1016/j.jconhyd.2010.12.001 2010 Elsevier B.V. All rights reserved.Solute transport modelingAquifer heterogeneitya b s t r a c t

Correct interpretation of tracer test data is critical for understanding transport processes in thesubsurface. This task can be greatly complicated by the presence of intraborehole flows in ahighly dynamic flow environment. At a new tracer test site (Hanford IFRC) a dynamic flow fieldcreated by changes in the stage of the adjacent Columbia River, coupled with a heterogeneoushydraulic conductivity distribution, leads to considerable variations in vertical hydraulicgradients. These variations, in turn, create intraborehole flows in fully-screened (6.5 m)observation wells with frequently alternating upward and downward movement. Thisphenomenon, in conjunction with a highly permeable aquifer formation and small horizontalhydraulic gradients, makes modeling analysis and model calibration a formidable challenge.Groundwater head data alone were insufcient to dene the ow model boundary conditions,and the movement of the tracer was highly sensitive to the dynamics of the ow eld. Thisstudy shows that model calibration can be signicantly improved by explicitly considering (a)dynamic ow model boundary conditions and (b) intraborehole ow. The ndings from thisstudy underscore the difculties in interpreting tracer tests and understanding solute transportunder highly dynamic ow conditions.Accepted 1 December 2010Available online 16 December 2010

Keywords:Hanford IFRC siteIntraborehole owDynamic owa r t i c l e i n f oImportance of considering intraboreunder highly dynamic ow condition

Rui Ma a,b, Chunmiao Zheng a,, Matt Tonkin c,a Department of Geological Sciences, University of Alabama, Tuscaloosa, AL, Unib MOE Laboratory of Biogeology and Environmental Geology, China University oc S.S. Papadopulos & Associates, Inc., Bethesda, MD, United Statesd Pacic Northwest National Laboratory, Richland, WA, United States

j ourna l homepage: www., 2001). However,ue to the lack ofnditions and thellations. Previousbias in samples

All rights reserved.e ow in solute transport modeling

M. Zachara d

tesciences, Wuhan, China

nant Hydrology

v ie r.com/ locate / jconhydare the only data available. However, there is little work toshow how intraborehole ow at a specic eld site affects thecalibration of a solute transport model developed to interpretthe tracer data collected from fully-screened wells.

The Hanford Integrated Field Research Challenge (IFRC)site provided a useful location to address the intraboreholeow issue. The IFRC site is located in the U.S. Department ofEnergy (DOE) Hanford 300 Area along the Columbia River in

southeastern Washington State (Fig. 1). This site has beencharacterized in great detail in previous studies and amixtureof fully-screened and depth-discrete observation wells werepreviously installed for monitoring tracer tests, with theformer being in greater number. Pleistocene-age HanfordFormation gravels and sands overlie Pliocene-age RingoldFormation ne-grained sediment and gravels in the uncon-ned aquifer at the IFRC site (Fig. 2). The total thickness of theHanford Formation beneath the IFRC site is approximately14 m (Newcomer et al., 2010) and the water table varies from8 to 11 m below the ground surface during different seasonscaused by the changes in the Columbia River stage. Thehydraulic conductivity of Hanford Formation was reported tobe greater than 2000 m/d (Peterson et al., 2005; Zachara,et al., 2005; Williams et al., 2007) and recent constant-rateaquifer pumping tests indicate that the average hydraulic

conductivity of the Hanford Formation is on the order of7000 m/d (Rockhold et al., 2010). Electromagnetic BoreholeFlowmeter (EBF) tests conducted within the Hanford Forma-tion at the IFRC site have indicated that this part of the aquiferis somewhat stratied, with the upper and lower portionsbeing more permeable than the middle portion, althoughthere is considerable variability between the hydraulicconductivity proles for different wells (Rockhold et al.2010). This stratication is subtle, and was not recognizedduring geologic and geophysical logging performed duringwell installation. The typical values of hydraulic conductivityfor Ringold ne-grained sediments is 1 m/d and those forRingold gravels are on the order of 40120 m/d (Williamset al., 2008).

The aquifer at the IFRC site is in hydraulic communicationwith the Columbia River along its easternmargin. As a result of

in as a

12 R. Ma et al. / Journal of Contaminant Hydrology 123 (2011) 1119Fig. 1. Plan view of the Hanford IFRC site showing the model doma rectangle in green color (a) and IFRC experimental well eld (b).,

for, complicates the interpretation of solute concentrations in

Three-well clusters (9 wells total) allow monitoring of

the wells both within and near the IFRC site were shown in

he IFRC site and Columbia River (the location of cross section is shown in Fig. 1 and

13R. Ma et al. / Journal of Contaminant Hydrology 123 (2011) 1119water samples extracted from these observation wells.In this study, a bromide anion (Br) tracer experiment

conducted in March 2009 at the IFRC site was numericallysimulated. The purpose of these simulations was to under-stand the principal physical transport processes prior toundertaking reactive tracer tests. The objectives of this studywere to a) quantify intraborehole ow induced by thestratied aquifer and dynamic groundwater ow eld; b)document the effect of intraborehole ow reversal on soluteconcentrations in samples collected from fully-screenedwells; c) illustrate the challenges in interpreting a tracerbreakthrough curve in a highly heterogeneous and dynamicow eld; and d) demonstrate the importance in consideringintraborehole ow in model calibration.

2. Field tracer experiment

A conservative tracer test was conducted fromMarch 13 toMarch 24, 2009 at the IFRC site. In this test, an aqueous solutionchilled to a temperature of 9.5 C with a Br concentration ofaverage 95 mg/L was injected into the upper 6.5 m of theunconned aquifer in the Hanford Formation, starting at 10:55am on March 13, at a ow rate of 16.3 m3/h through well 399-2-9 located in the NW corner of the IFRC site. The test lastedupstream hydroelectric dams the Columbia River experiencesfrequent (i.e., hourly) uctuations in river stage, in addition toseasonal variations. The hydraulic conditions imposed by theColumbia River lead to highly dynamic groundwater owconditions (in terms of ow direction and magnitude of thehydraulic gradient) at the IFRC site. The dynamic nature of theow eld, combined with the stratied hydraulic conductivitydistribution,may result in substantial verticalowwithin fully-screened observation wells, which, unless properly accounted

Fig. 2. Hydrogeologic cross section across the Hanford 300 Area intersecting tthe cross section is modied from Ma and Zheng, 2010).for 9.45 h. Groundwater samples were periodically collectedfor Br analysis from 35 wells (excluding well 399-2-25)covering the IFRC site over an 11-day period (Fig. 1b) followinguid injection.

The areal footprint of the IFRC site intersects with theextent of a larger uranium plume at 300 Area. The plumeresides within the upper 67 m of the unconned aquifer,within the highly permeable Hanford formation sediments. Awell-eld of 36 wells was installed at the IFRC site for cross-hole geophysical characterization, water level monitoring andsampling (Fig. 1b) (Bjornstad and Vermeul, 2008). Themonitoring wells were completed in the Hanford Formationexcept for well 399-2-25 which was not used for this test.Fig. 4(a).Water levels in theHanford aquifer respond rapidly tothe uctuations of the Columbia River stage. The frequency andamplitude of water-level uctuation in wells attenuated withincreasing distance from the river. Groundwater ow direc-tions, the magnitudes of hydraulic gradients determined usingthe classic three-point approach (Silliman and Frost, 1998)were both dynamic over the duration of the injectionexperiment as shown in Fig. 4(b). The three distinct periodsof groundwater ow direction and hydraulic gradient magni-tudedirectly affect tracermigration as shown in Fig. 4(c) duringthe tracer experiment.

In Period #1 (3/133/17/2009) a relatively low river stage,especially from 3/15 to 3/17, caused groundwater owdiscrete 1 m vertical zones representing upper, middle, andlower elevations of the plume, while other 26 wells(excluding well 399-2-25) are fully screened through the 67 m plume interval. All well diameters are 4 inches (Bjornstadand Vermeul, 2008).

From the tracer response in the northern multi-level wellcluster (Fig. 3), it can be seen that tracer arrival and elutionsignicantly lagged in the middle zone (blue, 399-2-28), andthe earliest arrival occurs in the deeper zone (red, 399-2-27),closely followed by that in the shallow zone (green, 399-2-26),which is consistent with EBF hydraulic test results (Rockholdet al., 2010). This conrmed the presence of a functionallystratied system.

3. Dynamic characteristics of the ow eld

The stage for the Columbia River and hydrographs forFig. 3. Br breakthrough curves at a well cluster: 399-2-26 (shallowelevation, green), 399-2-28 (middle elevation, blue), and 399-2-27(deepelevation, red).

14 R. Ma et al. / Journal of Contaminant Hydrology 123 (2011) 1119toward the south-southeast (azimuth less than 180o) of thesite and resulted in higher ow magnitudes. This, in turn,caused the Br concentration to peak in well 399-2-15located to the southeast of the injection well, and to be lowerin well 399-2-14 located to the south-southeast of theinjection well. In Period #2 (3/173/21/2009) a higher andrelatively cyclic period of river stage changed the groundwa-ter ow direction from south-southeast to south-southwest(azimuth greater than 180) and resulted in a reducedgradient magnitude. Correspondingly, the Br concentrationrose in well 399-2-14 to the south-southeast and dropped inwell 399-2-15 to the southeast as the plume's location shiftedin response to the groundwater ow direction. In addition,the concentration uctuated in both wells in response to thesignicant variations in river stage and groundwater owdirection. Finally, during Period #3 (3/213/25/2009), arelatively consistent river stage led to fairly stable ground-water ow directions.

The relationship between the changing river stage,groundwater ow direction, magnitude of hydraulic gradi-ents, and tracer concentrations was evident within eachperiod (Fig. 4). River stage uctuations cause rapid changes inthe groundwater ow direction and magnitude of thehydraulic gradient, which, in turn, could lead to dynamicchanges in hydraulic heads and gradients within the upper,middle and lower portions of the aquifer.

Fully-screened wells that intercept multiple, variablypermeable intervals could act as conduits between these

Fig. 4. (a) River stage uctuations and well hydrographs; (b) Groundwater ow ditracer experiment; (c) Br concentration breakthrough curves at wells 399-2-14 apoint, respectively). The shading in the gures illustrates three distinct periods ofuctuations.intervals. But potential ow through fully-screened wells isnot evident from Fig. 4 or the preceding discussion. That willbecome clear only when we examine ows into and out offully-screened wells at different intervals as discussedsubsequently. The dynamic changes in groundwater owdirections and rates, occurring within a stratied aquifer, leadto dynamic changes in the vertical head distributionthroughout the layered sequence.

4. Numerical modeling

4.1. Model setup

A three-dimensional groundwater ow model was devel-oped with MODFLOW (Harbaugh et al., 2000) to simulatesaturated owwithin themodel domain shown in Fig. 1a. Themodel domain was discretized into 60 columns (x-direction),61 rows (y-direction) and 27 layers (z-direction), covering ahorizontal distance 120 m in the x-direction, 122 m in they-direction, and a vertical thickness of 22.5 m. Boundaryconditions were dened for the sides of model domain(looking in plan view) using the Time-Varying Constant-Head (CHD) Package. The bottom of the grid was repre-sented as a no-ow boundary. The uppermost layer wasalso treated as no-ow boundary since vertical groundwa-ter recharge is negligible in this semi-arid environment.Zachara et al. (2005) and Yabusaki et al. (2008) determinedthat an hourly temporal discretization is the coarsest

rection and magnitude of hydraulic gradient at the IFRC site during the Br

nd 399-2-15 (located to the south-southeast and southeast of the injectiongroundwater ow and tracer migration resulting from Columbia River stage

15R. Ma et al. / Journal of Contaminant Hydrology 123 (2011) 1119frequency required to adequately represent the lateralboundary conditions due to the rapid uctuations in riverstage. Measured hourly water level data from the existingwells within and outside the IFRC site (Fig. 1a) were used tointerpolate time-varying head values at the four lateralmodel boundaries. Three-dimensional distribution of hy-draulic conductivity and porosity were initially set based ondata from the constant-rate and EBF tests (Rockhold et al.,2010) and were used as tting parameters in the subse-quent calibration.

Bromide transport was simulated with the MT3DMS code(Zheng andWang, 1999; Zheng, 2010). Initially, a longitudinaldispersivity (L) of 1 m was used for the Hanford Formationand 0.5 m was assigned to Ringold unit. The horizontal andvertical transverse dispersivities were assumed to be 10% and1% of the longitudinal dispersivity in all geological units,respectively, based on previous studies (Ma et al., 2010).Dispersivities were also used as tting parameters in thesubsequent calibration.

Both manual adjustment and inverse modeling software(PEST (Doherty, 2002)) were employed for both ow andtransport model calibration to optimize the heterogeneousparameter distribution of the aquifer (mainly the hydraulicconductivity). The pilot point method was used to parame-terize the hydraulic conductivity at predened locations fromwhich the hydraulic conductivity distribution for the entiremodel layer was interpolated using kriging (Doherty, 2003).Although the model domain was extended beyond the IFRCsite to minimize the effects of boundary conditions, calibra-tion was only attempted where detailed hydraulic conduc-tivity measurements were made within the IFRC site. Anaverage hydraulic conductivity value was assigned whendetailed hydraulic conductivity data were not available.

4.2. Method for accounting for intraborehole ows

The Multi-node Well (MNW) Package for ow (Halfordand Hanson, 2002) was used to assess intraborehole verticalow in the fully-screened wells. TheMNWpackage calculatesa single head inside the fully-screened wellbore which cutsthrough several model layers:

hwellhn = AQn + BQn + CQpn 1

where

hwell is head inside well borehole [L];hn is head in aquifer at elevation (layer) index n [L];A is linear aquifer-loss coefcient [TL2];B is linear well-loss coefcient [TL2];C is nonlinear well-loss coefcient [TpL(3p1)];p is power of the nonlinear discharge component of

well loss [dimensionless]; andQn is ow between well bore and aquifer at elevation

(layer) index n [L3T].

When the head in the aquifer at elevation n is higher thanthe wellbore head, the ow is toward the wellbore from theaquifer. Conversely, if the head in the aquifer at elevation n islower than the wellbore head, the ow is away from thewellbore into the aquifer. Thus, if hydraulic gradients existacross different model layers that are intersected by the wellscreen, the well can act as a conduit for intraborehole ow.

In MT3DMS, a solute transport companion to MODFLOW,the following equation was used to account for intraboreholesolute mixing in response to ow exchanges between thewellbore and aquifer (Zheng, 2010):

Ctavg =Qtw

Ctw +

N

n=1Qtn

Ctn

Qtw

+

N

n=1Qtn

2

where

Cavgt is the ux-averaged composite concentration in-

side the wellbore at time t [ML3];Qwt is the total prescribed ow into the multi-node well

(i.e., source) at time t [L3T1];Cwt is the concentration of the injected source Qwt at

time t [ML3];Qnt is the ow rate at elevation (layer) index n of the

aquifer discharging into the well at time t [L3T1];Cnt is the concentration associated with the ow rate

Qnt , i.e., the concentration in the aquifer at elevation

(layer) index n at time t [ML3]; andN is the total number of model layers intercepted by

the well screen.

The ow and transport capabilities of theMNWPackage asdescribed by Eqs. (1) and (2) have been successfullybenchmarked against analytical and numerical solutions byNeville and Zhang (2010).

5. Results and discussion

5.1. Occurrence of intraborehole vertical ows

Simulation results revealed the potential for complexintraborehole ow movement inside the fully-screenedobservation wells. As an example, Figs. 5 and 6 show thevariation in the ow rates between the aquifer and well boreat different depths in injection well 399-2-9 and three otherselected wells at seven selected times. A positive value of theow rate indicates that the groundwater ows from thewellbore to the aquifer, while a negative value indicates owfrom the aquifer to the wellbore. If the ow rate is negative inthe lower part of the well bore and positive on the upper partof the wellbore, then the groundwater discharges into thewellbore in the lower part, moves upward along the borehole,and returns to the aquifer, resulting in vertical intraboreholeow. The direction of intraborehole ow is downward if theow rate between the aquifer and borehole is positive in thelower part of the borehole and negative in the upper part. Themagnitude and direction of intraborehole ow in these wellsvaried substantially with depth, and also with time in theobservation wells, reecting the complicated and highlydynamic nature of intraborehole ow. The simulationindicated that vertical ows may change direction daily oreven more frequently depending on river stage changes.Direct eld measurements of intraborehole ows by EBF in

the IFRC well-eld have conrmed the existence of thesebehaviors (Newcomer et al., 2010). A sensitivity analysis (not

to preferential sampling of waters from separate aquifer zoneswith different aqueous composition and transport properties.

Fig. 5. Horizontal ows between the aquifer and wellbore inside the injection well 399-2-9 at different elevations calculated with the MODFLOWMulti-node WellPackage (a positive ow rate indicates that water ows from wellbore to aquifer while a negative value indicates ow from the aquifer to the wellbore).

16 R. Ma et al. / Journal of Contaminant Hydrology 123 (2011) 1119reported here) shows that the calculated intraborehole owsvary from well to well and are affected by the aquiferheterogeneities in both horizontal and vertical directions, butthe vertical heterogeneity is the primary controlling factor.

When the river water intrudes and retreats, groundwaterin some parts of the aquifer would be preferentially affectedby the river stage change due to stratied aquifer heteroge-neity. Local water levels and gradients at different depthswould behave differently. Penetrating, fully-screened wellsrespond to these gradient differences via intraborehole ow.Abrupt changes in solute concentration can arise in fully-screened observation well waters, depending on aquifersolute stratication and other factors. Water samples mayconsequently represent temporal conditions at a specicaquifer depth, rather than a homogenizedmixture of multipledepths (Vermeul et al., 2010).

Thus, there are two explanations for the concentrationuctuations in the observed breakthrough curves: (a) frequentchanges in the horizontal groundwater ow and tracermigration direction; and (b) intraborehole vertical ow thatalternates between upward and downward directions leadingFig. 6. Horizontal ows between aquifer and wellbore inside three selected observaWell Package (conventions and legends are the same as those in Fig. 5).5.2. Importance of considering intraborehole ows in transportmodeling

Fig. 7 compares the simulated Br tracer breakthroughcurves obtainedwithout (red) andwith (green) considerationof intraborehole ow at six selectedwell locations. Agreementbetween simulated (green lines) and observed concentrations(open dots) was markedly improved by explicitly simulatingintraborehole ow. The RMS (root mean squares) error of 939unweighted residuals in 35 wells was reduced from 7.0 mg/Lin the model that does not account for intraborehole ow to5.1 mg/L in the model that accounts for intraborehole ow. Itis noteworthy that the simulated concentrations with theMNW Package for well 399-2-9 (injection well) remainedconstant at 95 mg/L during the injection period, which ismuch more consistent with the observed data. There was aperiod of 22 min during which water without Br wasinadvertently injected into the aquifer during the experiment.The simulated breakthrough curve of 399-2-9 with the MNWPackage captured this occurrence correctly (Fig. 7).tion wells at different elevations calculated with the MODFLOW Multi-node

17R. Ma et al. / Journal of Contaminant Hydrology 123 (2011) 11195.3. Challenges in model calibration under highly dynamic owconditions

The unusually high hydraulic conductivity values averaging7000 m/d in the Hanford Formation implied that over smalldistances the measured head differences, and hence owdirections and magnitudes, were unreliable because of theminusculeheaddifference. Inaddition, themeasurementerrorsstemming from survey (e.g., well elevations), injection timingand duration, and well design (e.g., head loss), were cumula-tively signicant under such highly dynamic and permeableaquifer conditions. Tracer concentrations varied strongly inresponse to the direction and magnitude of the ow eldcontrolled by the Columbia River (Fig. 4). Thus, obtainingaccurate ow boundary conditions and resulting ambienthydraulic gradients is critical for model calibration.

Both manual trial-and-error adjustment and automatedoptimization based on the PEST code were applied to calibratethe hydraulic conductivity distribution by matching thesimulated hydraulic heads and solute concentrations againstthe observed values. We experimented with different interpo-lationmethods and selected differentwells near the IFRC site toobtain the specied-headboundary conditions for the IFRC site-scale owmodel. Different boundary conditions were found toproduce equally good matches between the observed andcalculated head distributions, but drastically different tracer

Fig. 7. Comparison of observed and calculated concentrations with and withplume movements. Thus the boundary condition is a crucialaspect ofmodel calibration for this site, and theuseof headdataalone is insufcient to constrain model calibration.

Tracer data were used to further constrain the interpolationof boundary conditions and the estimation of the hydraulicconductivity distribution, based on the comparison of observedand simulated Br tracer concentrations at different wells andobservation times. For the fully-screened wells, the observedconcentrations were obtained via the MNW option whichaccounts for intraborehole ow and solute mixing. Thehydraulic conductivity distribution was xed, and then theboundary condition and ambient ow gradient were adjusteduntil the general direction of simulated plume movement wasin agreement with observation. The inverse modeling proce-dure was then applied to adjust the hydraulic conductivitydistribution and improve the match between simulated andobserved Br concentrations.

During the inverse modeling process, the importance ofhead dataweighting was reduced in the PEST code to focus onan objective function comprised of head, gradient azimuth,and concentration. The calibration results were evaluated bycomparing the simulated and observed Br concentrations at20 observation wells at six selected times. There was anoverall agreement between the observed and measuredbreakthrough curves in the 20 observation wells, especiallyin terms of arrival times and general trends. As noted above,

out the Multi-node Well Package, respectively, at six well locations.

Biased monitoring of fresh watersalt water mixing zone in coastal

18 R. Ma et al. / Journal of Contaminant Hydrology 123 (2011) 1119the simulated concentrations were corrected by the Multi-node Well Package to account for the intraborehole verticalows. The model calibration processes indicated that thegroundwater ow gradient was equally or more importantthan the aquifer heterogeneity under the highly dynamic owconditions that existed. Thus, great care must be taken toensure that the boundary conditions for the IFRC site-scalemodel are properly constructed and calibrated for successfulanalysis and understanding of the tracer tests.

6. Conclusions

An unusually permeable aquifer formation (hydraulicconductivity ~7000 m/day) and a highly dynamic ow eld,driven by changes in the stage of the Columbia River, posedmajor challenges to model calibration at the Hanford IFRCsite. Groundwater head data alone were of little value inmodel calibration. Hydraulic gradients and tracer concentra-tion data must be explicitly considered to constrain andimprove the estimation of the hydraulic conductivity distri-bution and determination of the boundary conditions.

The high hydraulic conductivity and frequent aquiferriver interaction caused very dynamic ow conditions at theIFRC site. The model calibration process indicated thatobtaining accurate ow boundary conditions and resultinghydraulic gradients is critical for robust calibration. Ground-water ow gradients were equally or more important thanthe hydraulic conductivity distribution in determining thetracer breakthrough curves and plume migration under thehighly dynamic ow conditions that existed at the HanfordIFRC site.

Through the analysis and modeling of a non-reactivetracer experiment, signicant vertical hydraulic gradientsinduced by frequent oscillations in Columbia River stage werefound to exist in the aquifer. The resulting vertical headgradients caused signicant intraborehole ows with alter-nating upward and downward movements in the fully-screened observation wells. Intraborehole ow complicatesthe interpretation of solute concentrationsmeasured in watersamples from these wells. The agreement between model-simulated and observed tracer concentrations from the fully-screened observation wells improved markedly when theeffect of the intraborehole ow was taken into considerationin the ow and solute transport modeling. Thus, theintraborehole vertical ow must be considered in samplingand modeling analysis for the experimental and interpreta-tion results to be more meaningful.

The work described in this paper is part of a larger eldcampaign at the Hanford IFRC site that has been underwaysince 2007. Understanding and dealing with intraboreholeow is key to many other aspects of the eld researchprogram. The dynamic rivergroundwater system at theHanford IFRC site is unique and provides a compelling casestudy that underscores the importance of considering owdynamics and resulting intraborehole ows in tracer testanalysis and model calibration.

Acknowledgments

This research was supported by the Integrated FieldScaleSubsurface Research Challenge (IFRC) Project of the U.S.aquifers. Ground Water 47 (1), 4956.Silliman, S.E., Frost, C., 1998. Monitoring hydraulic gradient using three-point

estimator. J. Environ. Eng. 124, 517523.Sukop, M.C., 2000. Estimation of vertical concentration proles from existing

wells. Ground Water 38 (6), 836841.Department of Energy (DOE).We are grateful to Keith Halfordand two anonymous reviewerswhose constructive commentshave led to signicant improvement of this paper.

References

Bjornstad, B.N., Vermeul, V.R., 2008. Drilling, Sampling, andWell-InstallationPlan for the IFC Well Field, 300 Area, Report PNNL-17512. PacicNorthwest National Laboratory, Richland, Wash.

Church, P.E., Granato, G.E., 1996. Bias in ground-water data caused by well-bore ow in long-screen wells. Ground Water 34 (2), 262273.

Clark, B.R., Landon, M.K., Kauffman, L.J., Hornberger, G.Z., 2006. Simulation ofsolute movement through wellbores to characterize public supply wellcontaminant vulnerability in the High Plains Aquifer, York, Nebraska.Proc.MODFLOWandMore 2006:Managing Ground-Water Systems,May2224, 2006. International GroundWaterModeling Center, Golden, Colo.

Doherty, J., 2002. Manual for PEST, 5th edition. Watermark NumericalComputing, Brisbane, Australia. Downloadable from www.sspa.com/pest.

Doherty, J., 2003. Ground water model calibration using pilot points andregularization. Ground Water 41 (2), 170177.

Elci, A., Molz, F.J., Waldrop, W.R., 2001. Implications of observed and simulatedambient ow in monitoring wells. GroundWater 39 (6), 853862.

Halford, K.J., Hanson, R.T., 2002. User guide for the drawdown-limited, multi-node well (MNW) package for the U.S. Geological Survey's modularthree-dimensional nite-difference ground-water ow model, versionsMODFLOW-96 and MODFLOW-2000: U.S. Geological Survey Open-FileReport 02-293. 33 pp., Sacramento, California.

Harbaugh, A.W., Banta, E.R., Hill, M.C., McDonald, M.G., 2000. MODFLOW-2000, the U.S. Geological Survey modular ground-water model userguide to modularization concepts and the Ground-Water Flow Process.U.S. Geological Survey Open-File Report 00-92. 121 pp.

Izbicki, J.A., Christensen, A.H., Newhouse, M.W., Smith, G.A., Hanson, R.T.,2005. Temporal changes in the vertical distribution of ow and chloridein deep wells. Ground Water 43 (4), 531544.

Konikow, L.F., Hornberger, G.Z., 2006. Modeling effects of multi-node wellson solute transport. Ground Water 44 (5), 648660.

Lacombe, S., Sudicky, E.A., Frape, S.K., Unger, A.J.A., 1995. Inuence of leakyboreholes on cross-formational groundwater ow and contaminanttransport. Water Resour. Res. 31, 18711882.

Landon, M.K., Jurgens, B.C., Katz, B.G., Eberts, S.M., Burow, K.R., Crandall, C.A.,2010. Depth-dependent sampling to identify short-circuit pathways topublic-supply wells in multiple aquifer settings in the United States.Hydrogeol. J. 18, 577593.

Ma, R., Zheng, C., 2010. Effects of density and viscosity in modeling heat as agroundwater tracer. Ground Water 48 (3), 380389.

Ma, R., Zheng, C., Prommer, H., Greskowiak, J., Liu, C., Zachara, J., Rockhold, M.,2010. A eld-scale reactive transport model for U(VI) migrationinuenced by coupled multirate mass transfer and surface complexationreactions. Water Resour. Res. 46, W05509 doi:10.1029/2009WR008168.

Mayo, A.L., 2010. Ambient well-bore mixing, aquifer cross-contamination,pumping stress, and water quality from long-screened wells: what issampled and what is not? Hydrogeol. J. doi:10.1007/s10040-009-0568-2.

Neville, C.J., Zhang, J., 2010. Benchmark analysis of solute transport withmultiaquifer wells. Ground Water 48 (6), 884891.

Newcomer, D.R., Bjornstad, B.N., Vermeul, V.R., 2010. Vertical wellbore owmonitoring for assessing spatial and temporal ow relationships with adynamic river boundary. GroundWater Monit. Rem. doi:10.1111/j1745-6592.2010.01304.x.

Peterson, R.E., Freeman, E.J., Murray, C.J., Thorne, P.D., Truex,M.J., Vermeul, V.R.,Williams, M.D., Yabusaki, S.B., Zachara, J.M., Lindberg, J.L., McDonald, J.P.,2005. Contaminants of potential concern in the 300-FF-5 Operable Unit:Expanded annual groundwater report for FY2004, PNNL-15127. PacicNorthwest National Laboratory, Richland, Wash.

Reilly, T.E., Franke, O.L., Bennett, G.D., 1989. Bias in groundwater samplescaused by wellbore ow. J. Hydraul. Eng. 115, 270276.

Rockhold, M., Vermeul, V., Mackley, R., Fritz, B., Mendoza, D., Newcomer, E.,Newcomer, D., Murray, C., Zachara, J., 2010. Hydrogeologic character-ization of the Hanford 300 Area Integrated Field Research Challenge Siteand numerical modeling of the rst aquifer tracer test. Submitted toGround Water.

Shalev, E., Lazar, A., Wollman, S., Kington, S., Yechieli, Y., Gvirtzman, H., 2009.

Vermeul, V.R.,McKinley, J.P., Newcomer, D.R.,Mackley, R.D., Zachara, J.M., 2010.River-induced ow dynamics in long-screen wells and impact on aqueoussamples. GroundWater. doi:10.1111/j.1745-6584.2010.00769.x.

Williams, B.A., Brown, C.F., Um, W., Nimmons, M.J., Peterson, R.E., Bjornstad,B.N., Lanigan, D.C., Serne, R.J., Spane, F.A., Rockhold, M.L., 2007. Limitedeld investigation report for uranium contamination in the 300 Area,300-FF-5 Operable Unit, Hanford site, PNNL-16435. Pacic NorthwestNational Laboratory, Richland, Wash.

Williams, M.D., Rockhold, M.L., Thorne, P.D., Chen, Y., 2008. Three-DimensionalGroundwaterModelsof the 300Areaat theHanford Site,WashingtonState,PNNL-17708, Pacic Northwest National Laboratory, Richland, Wash.

Yabusaki, S.B., Fang, Y.,Waichler, S.R., 2008. Building conceptualmodels ofeld-scale uranium reactive transport in a dynamic vadose zone-aquifer-riversystem. Water Resour. Res. 44, W12403. doi:10.1029/2007WR006617.

Zachara, J.M., Davis, J.A., Liu, C., McKinley, J.P., Qafoku, N.P., Wellman, D.M.,Yabusaki, S.B., 2005. Uranium geochemistry in vadose zone and aquifersediments from the 300 Area uranium plume, PNNL-15121. PacicNorthwest National Laboratory, Richland, Wash.

Zheng, C., 2010. MT3DMS v5.3 supplemental user's guide. Department ofGeological Sciences, University of Alabama, Tuscaloosa, Alabama.

Zheng, C., Wang, P.P., 1999. MT3DMS: documentation and user's guide. U.S.Army Engineer Research and Development Center Contract ReportSERDP-99-1, Vicksburg, MS. 202 pp. http://hydro.geo.ua.edu/mt3d/.

Zinn, B.A., Konikow, L.F., 2007. Effects of intraborehole ow on groundwaterage distribution. Hydrogeol. J. 15, 633643.

19R. Ma et al. / Journal of Contaminant Hydrology 123 (2011) 1119

Importance of considering intraborehole flow in solute transport modeling under highly dynamic flow conditionsIntroduction and motivationField tracer experimentDynamic characteristics of the flow fieldNumerical modelingModel setupMethod for accounting for intraborehole flows

Results and discussionOccurrence of intraborehole vertical flowsImportance of considering intraborehole flows in transport modelingChallenges in model calibration under highly dynamic flow conditions

ConclusionsAcknowledgmentsReferences