Investigating cross-contamination of aquifers

Paul M. Santi · John E. McCray · Jamie L. Martens

Abstract Shallow aquifers can cross-contaminate deeperaquifers through penetration of an intervening aquitard,via sandy intervals in the aquitard, along well casings,across long well screens, or around aquitard pinchouts.Distinguishing among these potential pathways requiresthe use of evaluation tools that may support or eliminatecertain pathways. These tools include groundwater gra-dient and aquitard penetration calculations, aquiferpumping test interpretation techniques, chemical con-centration plots and statistical methods, hydraulic con-ductivity-based travel time calculations, pumping-con-centration tests, methods for evaluating leaky wells, andmethods for evaluating aquitard continuity. Based onanalysis of several of these techniques at three sites ex-periencing aquifer cross-contamination, the authors con-clude that calculation of flow rates for various pathwaysis the single most useful method to confirm or refutespecific pathways. Evaluation of leaky wells and aquitardcontinuity generally must be coupled with other tools toadequately test specific pathways. While fingerprinting,statistical analysis or ratio analysis of contaminants fromvarious sources and receptors was not completed for anyof the evaluated sites, the authors believe that thesetechniques have strong potential for confirming or elim-inating pathways. Future research in this area is sug-gested. Furthermore, the interpretation of pumping-con-

centration tests is not well developed and needs furtherassessment.

R�sum� Les aquif�res de surface peuvent contaminer lesaquif�res plus profonds � travers un aquitard, un inter-section sableuse dans l’aquitard, le long des forages et deleur cr�pines, ou de la terminaison biseaut�e des aqui-tards. Distinguer et s�lectionner ces diff�rents modes decontamination n�cessite l’utilisation d’outils d’�valuation.Ces outils incluent le gradient d’�coulement des eauxsouterraines et le calcul de p�n�tration des aquitards, lestechniques d’interpr�tation des essais de pompage, lesm�thodes d’interpr�tation chimiques et de traitementstatistique, les calculs des temps de transferts sur base dela conductivit�, les tests de pompage coupl� aux analysesde concentration, les m�thodes pour �valuer l’infiltrationdirecte alimentant les forages, et les m�thodes permettantd’�valuer la continuit� des aquitards. Bas� sur l’analysede plusieurs de ces techniques sur trois sites exp�rimen-tales pr�sentant des contaminations transversale, les au-teurs concluent que le calcul des rapports d’�coulementest la m�thode la plus simple pour confirmer ou refuserles voies sp�cifiques. Alors que le tra�age, les analysesstatistiques et les analyses de la contamination provenantde diff�rentes sources et r�cepteurs n’�tait pas compl�tespour la plus part des diff�rents sites �valu�s n’�taient pascompl�ments certains, les auteurs croient que ces tech-niques ont un potentiel fort pour confirmer ou �liminerces techniques. Des recherches futures dans cette zone estsugg�r�e. Par ailleurs, l’interpr�tation des tests de pom-page coupl�s � l’analyse des concentrations n’est pasd�velopp�e et n�cessite de meilleurs estimations.

Resumen Acu�feros someros pueden ocasionar contami-naci�n transversal de acu�feros profundos mediante lapenetraci�n de un acuitardo intermedio, a trav�s de inter-valos arenosos en el acuitardo, a lo largo del revestimientode pozos, o en las inmediaciones de lentes de acuitardos.Para distinguir entre estas trayectorias potenciales se re-quiere el uso de herramientas de evaluaci�n que puedenapoyar o eliminar ciertas trayectorias. Estas herramientasincluyen c�lculos de gradientes de agua subterr�nea ypenetraci�n de acuitardos, t�cnicas de interpretaci�n depruebas de bombeo de acu�feros, diagramas de concen-traci�n qu�mica y m�todos estad�sticos, estimaciones deconductividad hidr�ulica en base al tiempo de viaje,

Received: 2 November 2004 / Accepted: 26 November 2004Published online: 26 January 2005

� Springer-Verlag 2005

P. M. Santi ())Department of Geology and Geological Engineering,Colorado School of Mines,Golden, CO, 80401, USAe-mail: psanti@mines.eduTel.: +1-303-2733108Fax: +1-303-2733859

J. E. McCrayDivision of Environmental Science and Engineering,Colorado School of Mines,Golden, CO, 80401, USA

J. L. MartensURS Corporation,Suite 100, 10975 El Monte, Overland Park, KS, 66211, USA

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

pruebas de concentraci�n de bombeo, m�todos para eva-luar pozos con fugas, y m�todos para evaluar la continui-dad de acuitardos. Bas�ndose en el an�lisis de varias deestas t�cnicas en tres sitios que experimentan contamina-ci�n transversal de acu�feros, los autores concluyen que laestimaci�n de ritmos de flujo para varias trayectorias es elfflnico m�todo m�s ffltil para confirmar o rechazar trayec-torias de flujo espec�ficas. La evaluaci�n de pozos confugas y continuidad de acuitardos generalmente debe estaracompaada con otras herramientas para probar adecua-damente trayectorias espec�ficas. Aunque no se complet�para ninguno de los sitios evaluados t�cnicas de huellas,an�lisis estad�stico o an�lisis de relaciones de contami-nantes de varias fuentes y receptores, los autores creen queestas t�cnicas tienen fuerte potencial en la confirmaci�n oeliminaci�n de trayectorias. Se sugiere investigaci�n futuraen esta �rea. Adem�s, la interpretaci�n de pruebas deconcentraci�n de bombeo no se ha desarrollado bien ynecesita evaluaci�n posterior.

Keywords Confining units · Contamination · Generalhydrogeology · Groundwater flow · Groundwaterprotection

Introduction

A common community drinking-water supply is a lowerdrinking-water aquifer separated from an upper uncon-fined aquifer by a low-permeability aquitard. When thedrinking-water aquifer is discovered to be contaminated,the chemical pathway is usually via a discontinuity in theaquitard, which allows communication with a contami-nated upper aquifer. These discontinuities typically fallinto three categories: natural breaks or “holes” in the

aquitard, man-made “holes” through the aquitard (e.g.,from a well bore), or complete pinchout of the aquitard.For aquifer restoration, as well as for assignment of lia-bility, it is important to identify which of these pathwaysis the source of contamination.

The purpose of this paper is to assemble and demon-strate some tools available for analysis of aquifer cross-contamination, with the goal of identifying the likely flowpaths of contaminants. Each analysis tool will be de-scribed in detail and the potential use of each tool forpinpointing a specific flow path will be summarized. Fi-nally, three case histories will be presented using the toolsto interpret site hydrogeology.

Potential cross-contamination paths

When groundwater contamination of the lower aquifer ina multi-aquifer system occurs, three mechanisms forcross-contamination are generally proposed. Thesemechanisms are shown schematically on Fig. 1:

1. Penetration of an aquitard: in this case, contaminantshave traveled through the aquitard separating the twounits. Specific chemicals may alter the hydraulicproperties of an aquitard to accelerate this process.

2. Discontinuous aquitard: minor discontinuities inaquitards, such as gradational sandy zones or coarsechannel deposits, can create significant avenues forcontaminant migration.

3. Seepage along well casings: seals along well casings,particularly in older wells, may be imperfect.

As less common, yet equally important mechanism forlower aquifer contamination has also been documented:

Fig. 1 Aquifer cross-contamination pathways through an intervening aquitard. In this case there is a downward gradient from the upperaquifer to the lower aquifer

52

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

4. Aquitard pinchout and flow reversal: in this case,natural aquitard pinchout allows mingling of contam-inated upper aquifer groundwater with clean loweraquifer groundwater. Pumping from lower aquifers forwater supply accelerates or reverses natural ground-water gradients, which exacerbates cross-contamina-tion and causes contaminant plume migration into ar-eas that may not have been anticipated. This mecha-nism is shown schematically on Fig. 2, where con-tamination from the upper aquifer may come fromupgradient, downgradient, or crossgradient of aquitardpinchouts.

An example of Mechanism 4 is described in Bethune et al.(1996), who analyzed contamination of a water supplylake induced by reversal of groundwater flow throughheavy pumping. They were able to distinguish betweentwo possible sources of lake contamination resulting fromthe change in groundwater flow by using principles ofinorganic and organic chemistry, piezometric cross-sec-tions, and capture-zone modeling.

Mechanism 4 is also supported by data in this paperfrom sites at Fort Ord and Merced, California, and from asite in the central U.S. Methods of evaluating sources ofcross-contamination of aquifers are reviewed, followed byapplication of some of these methods to interpret con-taminant sources at each site.

Potential evaluation methods

Numerous evaluation methods are available to confirmcontamination sources and flowpaths. The application ofthese methods is summarized below, and the followingsection provides guidance on the use of each method toconfirm suspected cross-contamination mechanisms.

Groundwater gradients and aquitard penetrationOne of the strongest arguments against contamination of alower aquifer from an upper aquifer would be demon-stration of an upward vertical groundwater gradient. Anotable exception to this is the downward movement ofdense non-aqueous phase liquids (DNAPLs), where thedensity gradient works against groundwater flow. Thehigher density of a DNAPL pool, compared to freshwater, may be represented as an increased equivalent head(Lusczynski 1961) as represented by the followingequation:

hdnapl¼ hw � rdnapl=rw

� �ð1Þ

where hdnapl=equivalent head of the DNAPL pool in theupper aquifer, hw=water head in the upper aquifer,rdnapl=density or unit weight of the DNAPL compound,rw=density or unit weight of water. If the groundwatergradient is weak, downward flow will be produced if the

Fig. 2 Examples of aquifer cross-contamination pathways aroundaquitard pinchouts. In all three directions shown, contaminantmovement depends on favorable local gradients in the upper aquifer

and gradients toward the contaminated wells in the lower aquifer.Each of these flow pathways is documented in a case study later inthe paper

53

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

DNAPL head exceeds the capillary resistance of theaquitard:

hdnapl¼ 2 cosg� �

snw=reffgh ¼ hcap ð2Þwhere hcap=capillary head developed in the clay, g=wet-ting angle (often assumed to be 0�, so cosg=1),snw=NAPL-water interfacial tension (may be obtainedfrom Mercer and Cohen 1990), reff=effective pore radiusof aquitard (may be estimated by d50/8, see Mercer andCohen 1990).

The capillary resistance is usually negligible in sands.If the groundwater gradient is strong, downward flow willbe produced only if the DNAPL head exceeds both thecapillary head and the groundwater gradient head:

hdnapl¼ 2 cosg� �

snw=reffgh ¼ hl ð3Þwhere hl=water head in the lower aquifer.

The effects of the equivalent head are shown in Fig. 3.In Fig. 3A, the head in the lower aquifer, hl+hcap, isgreater than the head in the upper aquifer, hu. The re-sultant upward gradient between the two aquifers inhibitsdownward migration of contaminants.

In Fig. 3B, the head in the lower aquifer is now lessthan that in the upper aquifer, resulting in a downwardgroundwater gradient and potential downward migrationof contaminants.

The measured heads in Fig. 3C are the same favorablearrangement as in Fig. 3A, with the lower aquifer showinghigher head than the upper aquifer. However, the highdensity DNAPL pool in the upper aquifer results in anequivalent head higher than that in the lower aquifer. Thenet gradient is then downward, as in Fig. 3B. The traveltime of the DNAPL through the aquitard also depends onthe hydraulic conductivity of the aquitard, which may bevery slow.

The previous approach (using Eqs. 1 and 3) may alsobe applied to dense aqueous phase liquids (DAPLs).

However, because DAPLs are miscible with water,snw=0, and there is no capillary tension associated with anaquitard.

Pumping testsSignificant transfer of contaminants from an upper to alower aquifer will manifest itself as leakage to the loweraquifer during a pumping test. A schematic example ofthese effects is shown on Fig. 4, using a pumping well,two wells in the upper aquifer (U1 and U2) and two wellsin the lower aquifer (L1 and L2).

In Fig. 4A, no leakage through the aquitard occurs. InFig. 4B, the entire aquitard is leaky, and both the pumpingand lower aquifer observation wells show slightly lessdrawdown than that predicted, because additional water issupplied through the aquitard (Neuman and Witherspoon1969, 1972). In Fig. 4C, leakage through the aquitard isconcentrated in one area, such as where the aquitardgrades sandy, or in areas where it thins, pinches out, or isfissured.

Chemical concentrationsComingling of water from different sources has tradi-tionally been represented graphically. Figure 5 is an ide-alized example using a Stiff plot, but for site-specificcontaminants, which will be referred to as “indicator”chemicals, rather than the ionic species normally plotted.In this example groundwater at Well A is contaminatedwith chlorinated solvents (primarily TCE and PCE, andtheir degradation products). The contaminants move to-wards Wells C and E, whose Stiff plots show slight di-lution of the contaminants. Ideal indicator chemicalsshould be non-reactive and conservative, so that theirconcentrations and concentration ratios may be assumedto be entirely a function of original concentration modi-fied by dilution or mixing. Mixing, dilution, and reactionof chemicals can be modeled with groundwater chemistrycomputer models such as PHREEQC.

Fig. 3A–C Example showing the effects of vertical ground-water gradients and equivalent head of DNAPL pools on migration ofcontaminants into a lower aquifer

54

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

Several examples of this type of analysis may be foundin the technical literature. For instance, Mariner et al.(1997) distinguish between two possible sources of arse-nic contamination in sediments in a waterway by notingcorrelations of arsenic concentrations with other metalsconcentrations (Pb, Cu, and Zn) that are similar to thearsenic/metals correlations for one source of contamina-tion but not the other.

Fingerprinting groups of chemicals is useful to distin-guish between multiple sources of contaminants or multi-ple contaminated aquifers. For example, Douglas andMcMillen (1996) developed a detailed testing and inter-

pretive strategy to fingerprint petroleum hydrocarbons insoil. Zemo et al. (1995) use characteristic GC/FID traces tofingerprint multi-compound petroleum hydrocarbons, in-cluding gasoline, diesel fuel, and bunker and motor oil.Hydrocarbons that have “weathered” due to volatilizationor biodegradation may also be fingerprinted and distin-guished from their parent sources. Powers et al. (1997) usefour-axis star plots (a type of radar graph) to representseveral types of weathered and unweathered hydrocarbons.

Statistical clustering techniques for classification ofwater samples into hydrochemical facies (Back 1966) todefine groundwater flow paths have been used in the past

Fig. 4A–C Example showing the effects of leakage on aquifer pumping test results

Fig. 5 Idealized Stiff plot usingindicator chemicals instead ofcations and anions. In this case,Well A is contaminated, and theeffects of this contamination,with a degree of dilution isshown progressively in Wells Cand E. Lower aquifer Wells Band D are unaffected

55

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

for inorganic parameters (Jaquet et al. 1975; Alther 1979;Williams 1982; Farnham et al. 2000). For instance, G�leret al. (2002) found that statistical clustering techniques forinorganic water-chemistry parameters significantly im-proved on traditional graphical fingerprinting of watermasses. They concluded that statistical clustering is notonly more objective and robust than typical visual plot-ting methods, but can also incorporate more data into theanalysis.

Hydraulic conductivity argumentsCalculation of travel time and solute concentrations de-pends strongly on hydraulic conductivity. Figure 6 is anexample case where the expected travel time is calculatedfrom an upper aquifer source, around an aquitard pin-chout, to a lower aquifer well (using data from the Mid-western U.S. site described in more detail later in thepaper). As is sometimes the case, the flow direction in thelower aquifer has been reversed because of long-termpumping from the water supply well (examples of thissituation are discussed in the “Case Studies” section).

Using Darcy’s Law and the values given in Fig. 6, theexpected travel time around the aquitard pinchout is162 days or roughly 0.5 years. The expected travel timefor water to penetrate the aquitard is 536 days or roughly1.5 years. In this case, the two pathways could be dis-tinguished based on travel time alone. Diffusive effects,which are not included in this calculation, could signifi-cantly increase contaminant transport and shorten traveltime.

If the contaminant is a DNAPL, such as TCE, the ef-fective hydraulic conductivity of the aquitard may actu-ally be higher or lower, depending on the fluid densityand viscosity, as well as the relative volumes of DNAPLand water in the pore spaces:

K ¼ k � krel � r � g

mð4Þ

where K=hydraulic conductivity of the aquitard for agiven fluid (which may or may not be fresh water),krel=permeability adjustment factor caused by presence of

NAPL (less than one), r=density of the fluid, g=acceler-ation of gravity, m=viscosity of the fluid. Values for r andm for various DNAPLs can be obtained from Mercer andCohen (1990).

Based on this equation, the hydraulic conductivityexperienced by a DNAPL, Kdnapl, may be shown to be

Kdnapl¼ Kwater � rdnapl=rwater

� �� mwater=mdnapl

� �� krel

ð5ÞFor the example shown on Fig. 6, the travel time

through the aquitard is expected to be accelerated, pro-vided the fraction of pore space filled by DNAPL is high,because TCE is more dense and less viscous than waterand will therefore travel more easily through the aquitard.

As an example of the influence of the density, vis-cosity, and relative permeability of a DNAPL on its rateof flow, consider a plume of TCE (1.46 times as denseand 0.57 times as viscous as water at 20 �C). Using rel-ative permeability numbers for TCE from Lin et al.(1982), the following values for TCE are calculated:

% TCE saturation krel for TCE KTCE

100 1.0 2.6 Kwater95 (estimated) 0.9 2.3 Kwater60 0.4 1.0 Kwater40 0.2 0.5 Kwater

These calculations show that TCE will move fasterthan water for TCE saturations in excess of 60% andslower than water for saturations less than 60%.

The distinction between free-phase DNAPL and lowerlevels of saturation is an important one. Calculation ofhydraulic conductivity at intermediate saturations or formulti-phase flow is more complicated than for free-phasefluids, and examples of these calculations may be foundin Faust (1985), Abriola and Pinder (1985), Kueper andFrind (1991), Oolman et al. (1995), McCray and Falta(1997), and Fetter (1999).

Fig. 6 Information required for travel time calculation to evaluate potential contaminant flow paths. Data for this example is from theMid-Western U.S. site analyzed later in the paper

56

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

Changes in concentration of contaminantsduring pumpingA pumping-concentration test may also be used to dis-tinguish between possible contaminant sources. Such atest is performed by sampling aquifer water and testingchemical concentrations before pumping, and then sam-pling and testing at specified intervals after pumpingbegins. Possible test interpretations are suggested for anidealized case in Fig. 7, where a pumping well is in adifferent aquifer from the main source of contaminantsand the primary contaminant plume.

Locke (1994) notes the importance of stratification ofconcentrations of ion species within an aquifer, andevaluates the changes in concentrations with pumping forboth the pumping well and an observation well. He con-cludes that the greatest concentration changes occurred inor directly adjacent to the screened interval, and thesmallest changes occurred at the base of the aquifer. Someion species increased with pumping, some decreased, andsome did not show a trend in either direction.

For long periods of pumping, greatly in excess of thatshown on Fig. 7, a concentration decrease is expected,regardless of the contaminant travel path. This is the re-sult typically observed in pump-and-treat programs forgroundwater remediation, where chemical concentrationsare dominated by rate-limited dissolution of NAPLsources, rate-limited diffusion out of contaminated clayzones, or preferential removal of more soluble compo-nents in NAPL mixtures.

Evaluation of leaky wellsOld wells or wells for which installation information isincomplete are common scapegoats for aquifer cross-contamination because of improper backfill or screening

across multiple water-bearing zones. Fryberger and Tinlin(1984) estimate that the United States alone has nearlytwo million abandoned wells. Examples of groundwatercontamination caused by abandoned wells are given inGass et al. (1977), and examples of methods of detectionare given in Javandel et al. (1988) and Aller (1984). La-combe et al. (1995) present fluid flow and solute transportalgorithms, as well as groundwater modeling results forleaky boreholes in the same idealized setting discussed inthis paper: an upper unconfined aquifer, a middle aqui-tard, and a lower confined aquifer.

Improper backfill and screened intervals of wells maybe identified with logs or completion diagrams or byusing video logs of wells. Evaluation of open-hole sec-tions can be done by video or by using Borehole ImageProcessing (BIPS). The Acoustic Borehole Televiewerhas similar capabilities, and is effective through muddywater, since it relies on sonic waves rather than lightwaves (Welenco 1995). Casing leakage and upward ordownward water flow may be detected using temperature,flowmeter or hydrophysical (conductivity) logs.

A geophysical Cement Bond Log (CBL) may be usedto gauge the quality of the bonding of cement grout to theoutside of a steel or PVC well casing (Welenco 1995).The CBL relies on the difference in sonic velocities be-tween cement and void space or unconsolidated formationmaterials and it will also detect casing perforations as alow amplitude signal (Welenco 1995).

Tracer tests can be used to evaluate leakage throughannular backfill at the ground surface or interconnectionbetween aquifers via unknown screened intervals. Forinstance, Meiri (1989) describes the use of a sodiumbromide tracer to confirm leakage from an upper uncon-fined aquifer through a clay aquitard, and into a lower

Fig. 7 Idealized results from a pumping-concentration test

57

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

confined aquifer via a poorly constructed well completedin the lower aquifer.

If the appropriate wells are available, pumping testscan evaluate the “leaky” condition induced by leakagealong the well casing or through multiple screened in-tervals. The ideal arrangement would be to have at leastone shallow aquifer well and one deep aquifer wellflanking the suspect well. If leakage is detected, then themagnitude and rapidity with which it becomes apparentmay be used to distinguish between the three possiblecases: (1) natural leakage through the aquitard (lowleakage rates, delayed occurrence), (2) leakage along thewell annulus (moderate leakage rates, moderately rapidoccurrence), or (3) leakage through a well screen thatextends through both aquifers (high leakage rates, rapidoccurrence).

Aquitard continuityAquitards may be ineffective where they are thin, pin-ched-out, intermittently absent, or graded with coarsermaterial. Tools for detecting these features might includeexploratory borings, cone penetrometer studies, geo-physical surveys, and geologic analysis of depositionalenvironments. Examples of the use of geophysical toolsare presented below.

Lindenberg (1997) evaluates the use of Ground Pene-trating Radar (GPR) to assess aquitard continuity at aTCE contaminated site. She used 50 MHz and 200 MHzfrequencies, and achieved good resolution to depths of upto 16 m, but with imperfect correlation to geology. Sheconcludes that more sophisticated migration and filteringprocesses will improve the geologic interpretation andcorrelation, and lower frequency antennas will providedeeper penetration, albeit with less resolution.

Cardimona et al. (1998) use both shallow seismic re-flection and GPR to image the sand-clay interface at thebase of a shallow unconfined aquifer (10–14 m depth).They confirmed their interpretations with cone pene-trometer soundings. They note that although it is chal-lenging to obtain good seismic reflection data at suchshallow depths, correlation with GPR data strengthens theinterpretations. However, they also warn that it is un-common to get good data from one site for seismicmethods, which favor wet, clayey soils, and GPR meth-ods, which favor dry, sandy soils (Clement et al. 1997).Neither data set provided a good image of the water table,probably a result of a large capillary fringe above the truewater table.

Moore (1991) describes the use of electrical resistivitytechniques to develop an isopach map of a shallow clayaquitard. The generated map was calibrated against in-formation from 168 boreholes.

Matching evaluation methods to paths

Table 1 summarizes the evaluation methods and the travelpaths each method may support or disprove. Many ofthese methods are non-unique; that is, a specific result

may support more than one potential travel path, althoughseveral others may be eliminated. For this reason, severalmethods should be used in conjunction to identify themost likely travel path.

Case studies

Three case studies are presented to demonstrate the ap-plication of the evaluation tools using typical data col-lected during remedial investigations. All three sites sharecommon geologic characteristics: at each location, ashallow unconfined aquifer is contaminated and is sus-pected of contributing chemicals to a deeper confinedaquifer. The intervening aquitard pinches out within thesite boundaries, so proper mitigation of both aquifersdepends on establishing whether and how contaminationpenetrated the aquitard, or whether contamination trav-eled around the aquitard and in which direction.

Fort Ord Landfills, Fort Ord, California

BackgroundThree landfills comprise about 100 acres within thenorthern portion of the former U.S. Army Fort Ord, nearMonterey, California, as shown on Fig. 8. The mainlandfill was operated as a Class III facility from 1960until 1987 and received household and commercial refuseand a small amount of chemical wastes. The primaryconcern for the site investigation centered on the presenceof volatile organic compounds (VOCs) in soil andgroundwater (Dames & Moore 1991). VOCs were de-tected in lower aquifer wells in 1985, so the travel timefrom the landfills, if they were indeed the source, was lessthan 25 years.

Geologic and hydrogeologic factorsThe Salinas Valley aquiclude separating the upper andlower aquifers pinches out at the western edge of the site,so the two aquifers are interconnected locally. Ground-water flow in the upper aquifer is to the west andgroundwater flow in the lower aquifer is to the east. Theopposite flow direction in the lower aquifer is probablydue to gradient reversal by pumping from the lower andother deeper aquifers from municipal wells to the east.

Investigation methodsA drilling program was completed to install upper andlower aquifer wells. One well was continuously coredthrough the aquitard and lower aquifer to assess theconsistency of the aquitard and the extensiveness of claystringers in the lower aquifer. Gamma, sonic, SP, andresistivity logging was completed in lower aquifer bor-ings. Aquifer parameters were measured through labora-tory hydraulic conductivity tests of the aquitard clay, slugtests in the upper and lower aquifer, and pumping tests inthe upper and lower aquifer to assist in design of ex-traction wells. A time-domain electromagnetic (TDEM)

58

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

Tab

le1

Pot

enti

alev

alua

tion

met

hods

Eva

luat

ion

met

hod

Des

crip

tion

Sig

nifi

cant

resu

lts

Tra

vel

path

supp

orte

d

Par

alle

lin

depe

n-de

ntpl

umes

Dis

con-

tinu

ous

(pin

ches

out/

sand

y)

Pen

e-tr

atio

nW

ell

annu

lus

Wel

lsc

reen

Dow

n-gr

adie

ntpi

n-ch

out

Cro

ss-

grad

ient

pinc

hout

1.G

roun

dwat

ergr

adie

nts

Exa

min

atio

nof

the

grad

ient

acro

ssth

eaq

uita

rd(u

sing

equi

vale

ntgr

adie

nts

inth

epr

esen

ceof

DN

APL

s)

A.

Upw

ard

4

B.

Dow

nwar

d4

44

4

C.

DN

AP

L-i

nduc

eddo

wnw

ard

44

44

2.P

umpi

ngte

sts

Exa

min

atio

nof

the

resp

onse

ofw

ater

leve

lsin

the

aqui

fers

due

topu

mpi

ngA

.O

bser

vati

onw

ells

inup

per

aqui

fer

show

nodr

awdo

wn

and

wel

lsin

low

eraq

uife

rsh

owdr

aw-

dow

ncu

rves

foll

owin

gT

heis

curv

e.N

ole

akag

e

4

B.

Lar

gest

draw

dow

nin

low

eraq

uife

rne

arpu

mp-

ing

wel

l,sl

ight

lyle

ssdr

awdo

wn

than

pred

icte

d,up

per

aqui

fer

show

sun

conf

ined

resp

onse

.E

ntir

eaq

uita

rdle

aky

4

C.

Loc

aliz

edef

fect

sof

leak

age

inbo

thaq

uife

rs.

Pin

chou

t4

D.

Low

leak

age

rate

san

dde

laye

doc

curr

ence

.N

atur

alle

akag

eth

roug

haq

uita

rd4

4

E.

Mod

erat

ele

akag

era

tes

and

mod

erat

ely

rapi

doc

curr

ence

.L

eaka

geal

ong

annu

lus

4

F.

Hig

hle

akag

era

tes

and

rapi

doc

curr

ence

.L

eaka

geth

roug

hw

ell

scre

en4

3.C

hem

ical

conc

entr

atio

nE

xam

inat

ion

ofge

nera

lw

ater

qual

ity

rati

osan

din

dica

tor

chem

ical

sus

ing

Pip

eran

dS

tiff

diag

ram

s

A.

Dis

sim

ilar

rati

os4

B.

Sim

ilar

rati

osat

dilu

ted

conc

entr

atio

ns4

44

44

4

4.H

ydra

ulic

cond

ucti

vity

Cal

cula

tion

oftr

avel

tim

ean

dso

lute

conc

entr

atio

nsde

pend

ent

onhy

drau

lic

cond

ucti

vity

A.

The

tim

efo

rtr

ansm

issi

onm

ust

corr

elat

ew

ith

the

dete

ctio

nof

the

grou

ndw

ater

cont

ami-

nati

on.

Tim

e=d*

n e/(

K*i

)

44

4

5.G

ener

algr

ound

wat

erfl

owdi

rect

ions

A.

Bot

haq

uife

rsfl

owin

gin

sam

edi

rect

ion

wit

hno

reve

rsal

offl

ow4

B.

Upp

eraq

uife

rfl

ows

tow

ard

dow

ngra

dien

tpi

ncho

ut,

low

eraq

uife

rfl

ows

away

4

C.

Upp

eraq

uife

rha

sa

com

pone

ntof

flow

tow

ards

ala

tera

laq

uita

rdpi

ncho

ut,

low

eraq

uife

rha

sa

com

pone

ntof

flow

away

4

6.C

hang

esin

conc

entr

atio

ndu

ring

pum

ping

Com

pari

son

ofch

emic

alco

ncen

trat

ions

inth

ew

ater

sam

pled

and

test

edbe

fore

pum

ping

and

then

sam

pled

and

test

edat

spec

ific

inte

rval

s

A.

Con

cent

rati

onof

cont

amin

ants

isco

nsta

ntdu

ring

pum

ping

44

4

B.

Con

cent

rati

onin

crea

ses

over

tim

e4

C.

Con

cent

rati

onde

crea

ses

over

tim

e4

7.E

valu

atio

nof

leak

yw

ells

Inve

stig

atio

nof

area

tolo

cate

aban

done

dor

leak

yw

ells

.M

etho

dsin

clud

efl

uid

inje

ctio

n,B

oreh

ole

Imag

ing

Pro

cess

ing,

Aco

usti

cB

oreh

ole

Tel

evie

wer

,ge

ophy

si-

cal

cem

ent

bond

log,

tem

pera

ture

and

flow

met

erlo

gs,

and

trac

erst

udie

s

A.

Hyd

rolo

gic

pres

sure

s4

B.

Poo

rba

ckfi

llor

long

scre

ened

inte

rval

4

C.

BIP

S/v

ideo

/aco

usti

cbo

reho

lete

levi

ewer

44

D.

Cem

ent

bond

log

44

E.

Flo

wm

eter

/hyd

roph

ysic

s4

8.A

quit

ard

cont

inui

tyE

xam

inat

ion

ofaq

uita

rdto

loca

tepi

nch-

outs

,th

inse

ctio

ns,

inte

rmit

tent

abse

nces

,an

dco

arse

mat

eria

l.T

echn

ique

sin

clud

eex

plor

ator

ybo

ring

s,co

nepe

netr

omet

erst

udie

s,ge

ophy

sica

lsu

rvey

s,an

dge

olog

icde

posi

tion

alan

alys

is

A.

Aqu

itar

dap

pear

sco

ntin

uous

wit

hno

gaps

,sa

ndle

nses

,or

pinc

hout

s4

B.

Sig

nifi

cant

gaps

orsa

ndy

zone

sde

tect

edw

ithi

naq

uita

rdbo

dy4

C.

Com

plet

epi

ncho

utof

aqui

tard

dete

cted

44

59

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

60

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

geophysical investigation was completed to delineateaquitard pinchout.

Potential travel pathsBased on the site history and local stratigraphy deter-mined from the site investigation, three potential travelpaths of contamination are identified:

1. A discontinuous aquitard could allow mingling ofwater from both aquifers and potential VOC transferwhere the aquitard is missing.

2. VOCs could penetrate the aquitard, particularly in ar-eas where it thins.

3. The downgradient merging of plumes at the aquitardpinchout could allow VOCs to enter the lower aquiferfrom the upper aquifer.

Likely travel pathTable 2 summarizes the analysis of potential travel paths.Drilling and sampling of the Salinas Valley Aquicludeindicates that it is on the order of 50 feet (15 m) thick andcontains few sand lenses in the vicinity of the landfills.Calculated travel times for contaminants to penetrate theaquiclude are on the order of thousands of years. Based onthese indications of the nature of the aquiclude, the po-tential travel pathways through the aquiclude (based ondiscontinuities or penetration) are eliminated from furtherconsideration.

A downgradient pinchout of the aquiclude was delin-eated using boring and geophysical data, and groundwaterflow directions and calculated travel times for contami-nant transport around the pinchout are reasonable. Whilenot irrefutable, these lines of evidence support the con-clusion that the lower aquifer was contaminated by travelaround the pinchout. Evidence of aquitard pinchout andgroundwater flow directions implies that parallel inde-pendent plumes in each aquifer are unlikely.

Fig. 8 Maps and cross-sections from the Fort Ord, California site

Table 2 Summary of evaluation methods for Fort Ord site

Evaluationmethod

Description Discontinu-ous (pinchesout/sandy)a

Pene-tration

Down-gradientpinchout

Summary of analysis

1. Groundwatergradients

D. Downward 6 6 Downward gradient induces downward movementof contaminants through the aquitard. Analysis under#4 below shows that the rate is too slow to be aprobable source of contamination. Continuous coringof aquitard in vicinity of landfills indicates thick claywith few sand lenses

2. Pumpingtests

? ? Pumping test in lower aquifer did not show significantleakage. Two upper aquifer wells monitored duringpumping of lower aquifer did not show clear responseto pumping (one well showed no response, the secondshowed steady decline, continuing for 30 h afterpumping stopped)

4. Hydraulicconductivity

B. The timefor transmissioncalculation

6 6 4 Calculated travel time (advection only) for each unit:Upper aquifer, landfills to pinchout=2–11 yearsAquitard, through 50 foot thickness=17 103–22 103

yearsLower aquifer, pinchout to center of plume=400–1,700yearsContamination was detected within 25 years, settingan upper limit on total travel timeThese results imply that travel through the aquitardis unlikely. Travel from the landfills to the pinchoutis rapid, but the low gradient in the lower aquiferslows movement. It is strongly possible that chemicalscrossed the ragged pinchout and then spread in thelower aquifer by diffusion

5. Generalgroundwaterflow directions

B. Upper aquiferflows towarddowngradientpinchout, loweraquifer flowsaway

4 Groundwater levels measured on two different occasionsshow case B

8. Aquitardcontinuity

C. Completepinchout ofaquitard detected

4 Downgradient pinchout of aquitard identified in threeTDEM lines and five well borings

a ?, test is inconclusive for identifying contaminant travel path; 6, test assists in eliminating specific travel path; 4, test supports a specifictravel path

61

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

Mid-Western U.S. site

BackgroundThis site is an industrial manufacturing plant coveringover 100 acres (40 hectares), shown on Fig. 9. Approxi-mately 15 years after beginning operations, Trichloro-ethylene (TCE) was detected in a municipal water well inthe lower aquifer southeast of the site.

Two events at the site may have contributed togroundwater contamination. First, an improperly linedsurface impoundment was constructed on the southeastcorner of the site. The impoundment accepted wastewaterfor the first eight years of operation. Next, a spill occurredon the north side of the plant, eight years after beginningoperations, resulting in the loss of several thousand gal-lons of TCE. TCE was detected in the upper aquifer in thevicinity of both of these sources, and at several locationsin the lower aquifer. A distinction between contaminationresulting from the surface impoundment and the spill wasimportant for cleanup and liability reasons.

Geologic and hydrogeologic factorsThe clay confining unit separating the upper and loweraquifers pinches out to the north and east of the site,allowing commingling of water from both aquifers. Theclay contains local sand lenses that affect its ability toretard vertical movement of groundwater from the upperto lower aquifers. Previous studies indicate significantvertical leakage through the unit. Supporting evidenceincludes the following:

– shallow cone of depression in the lower aquifer com-pared to the pumping rate,

– digital flow model that requires a 20% leakage factorto simulate observed water levels in the lower aquifer(however, this leakage may come from a nearbystream, located beyond where the two aquifers merge),

– difference in hydraulic head between the two aquifersindicating a downward driving force, and

– anomalies in the normal geothermal field as indicatedby geophysical temperature logs.

Investigation methodsOver 80 borings were drilled at the site, 37 of which wereconverted to monitoring wells. Wells were monitoredquarterly to obtain water levels and samples for analysis.A constant-rate aquifer pumping test was performed in thelower aquifer to evaluate hydraulic parameters, includingpossible leakage through the clay. Downhole gammalogging was used in several borings to characterize thesubsurface, and a TDEM geophysical survey was com-pleted to delineate aquitard pinchout.

Potential travel pathsFour potential routes of cross-contamination were iden-tified:

1. Contamination traveled toward the north in the upperaquifer, in the direction of greatest groundwater ve-locity. At the aquitard pinchout, the contaminationentered the lower aquifer and then traveled southeasttoward the city well. This route is supported by de-tections of contamination in lower aquifer wellsnorthwest of the site, but refuted by clean loweraquifer wells northwest of the city well.

2. Contamination traveled to the northeast from the pondarea to the northeastern aquitard pinchout, and thenentered the lower aquifer. This pathway is supportedby detection of contamination in a shallow well nearthe pinchout. No wells were installed in the loweraquifer between the northeastern pinchout and the citywell.

3. Contamination leaked through the confining unit intothe lower aquifer.

4. Contamination traveled down from the upper to thelower aquifer via well casing in the city well. At thetime this data was collected, the city had not providedrecords of well construction for their well.

Likely travel pathTable 3 summarizes the analysis of potential travel paths.Using measured groundwater gradients and hydraulicconductivity values, the fastest travel path for contami-nation (1.5 years) is from the impoundment around theaquitard pinchout to the west (crossgradient). Travel fromeither source around downgradient or crossgradient pin-chouts is also possible, although slightly slower (4.5–6.3 years). Based on the available evidence, it is notpossible to conclusively identify which source contami-nated the city wells, although the impoundment couldhave contaminated the wells sooner than the spill.

Travel through the aquitard is significantly slower (32–74 years, depending on the location of penetration), evenif a free-phase TCE is assumed to have ponded on top ofthe aquitard (minimum time of 14–29 years, depending onTCE saturation). For this reason, supported by drillingobservations of continuous clay throughout the aquitard,penetration of the aquitard by TCE is not considered alikely pathway. Evidence of aquitard pinchout andgroundwater flow directions implies that parallel inde-pendent plumes in each aquifer are unlikely.

City of Merced, California

BackgroundThe City of Merced, California identified a regionalTetrachloroethylene (PCE) plume in several of theirdrinking water wells in 1987, shown on Fig. 10. Severalsources were identified, mostly consisting of dry-cleaningestablishments (five inactive establishments and six activeones), which had been in operation between 10 and50 years preceding the discovery of PCE in lowerdrinking water aquifers.

62

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

Fig. 9 Maps from the Mid-Western U.S. site

63

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

Geologic and hydrogeologic factorsThree contaminated aquifers have been identified in theregion: a “shallow,” unconfined aquifer, an “intermedi-ate” semiconfined to unconfined aquifer, and a deeper“confined” aquifer (Walker 1989). The shallow and in-termediate aquifers are separated by the “shallow clay”and the intermediate and confined aquifers are separatedby the “Corcoran clay.” Both clay units pinch out in thewest part of the city, and the Corcoran clay has beennoted to contain several sandy zones in this area (Elliott

1984). The intermediate and confined aquifers have agroundwater flow gradient to the east and the shallowaquifer has a local gradient to the west in the vicinity ofthe pinchouts (Elliott 1984; Walker 1989). The reversedgradient of the shallow aquifer in this area is probably dueto recharge of the intermediate and confined aquifers inthis location by shallow aquifer water. Downward gradi-ents have been measured across both clay units (Walker1989).

Table 3 Summary of evaluation methods for Mid-West U.S. site

Evaluationmethod

Description Discon-tinuous(pinchesout/sandy)a

Pene-tration

Down-gradientpin-chout

Cross-gradientpin-chout

Summary of analysis

1. Groundwatergradients

E. Downward 6 6 Downward gradient induces downward movementof contaminants through the aquitard. Analysisunder #4 below shows that the rate of penetrationthrough the aquitard is not fast enough for therequired time frame (15 years for the impoundment,7 years for the spill). Drilling indicates continuousclay with few sand lenses

F. DNAPL-induceddownward

2. Pumpingtests

? ? Results are difficult to interpret (barometricpressure was not recorded, but could accountfor much of the noted water level fluctuation).Leakage into lower aquifer is indicated, butsource of leakage is unclear (could be fromupper aquifer or from nearby stream. Author’sinterpretation is that leakage is inflow from stream)

4. Hydraulicconductivity

C. The timefor transmissioncalculation

6 6 4 4 Calculated travel time (advection only) for eachpathway, including travel through the lower aquiferto the city well:downgradient path, from impoundment=5.4–5.8 yearsdowngradient path, from spill=5.3–6.3 yearscrossgradient path, from impoundment=1.5 yearscrossgradient path, from spill=4.5 yearsthrough aquitard, near impoundment=74 years(29 years if free-phase TCE ponded aboveaquitard)through aquitard, near spill=32 years (14 yearsif free-phase TCE ponded above aquitard)Contamination was detected within 15 years,setting an upper limit on travel timeThese results imply that travel through theaquitard is unlikely. Travel via downgradient orcrossgradient pathways is possible, with travelfrom the impoundment crossgradient being theshortest and fastest pathway

5. Generalgroundwaterflow directions

B. Upper aquiferflows towarddowngradientpinchout, loweraquifer flowsaway

4 4 Groundwater levels measured on two differentoccasions show case B

7. Evaluationof leaky wells

B. Record of poorbackfill

4 Hearsay evidence suggests poor backfill aroundcity well. Well construction logs had not beenreleased by city at time of evaluation

8. Aquitardcontinuity

C. Completepinchout ofaquitard detected

4 4 Downgradient and crossgradient pinchoutof aquitard identified in six well borings andsuggested as a possibility by the TDEM survey

a ?, test is inconclusive for identifying contaminant travel path; 6, test assists in eliminating specific travel path; 4, test supports a specifictravel path

64

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

Fig. 10 Maps and cross-sections from the City of Merced, California site

65

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

Investigation methodsThe region has been extensively characterized by indi-vidual site investigations, periodic sampling of city wells,and several regional and city-scale studies by the CentralValley Regional Water Quality Control Board (CVR-WQCB; Walker 1988, 1989; Izzo 1989) and the U.S.Geological Survey (Elliott 1984). These studies includecity well logs, hydrostratigraphic interpretations, inven-tories of impacted wells, soil gas surveys, and pumpingconcentration tests (Luhdorff & Scalmanini 1987).

Potential travel pathsFive potential paths were identified for PCE to migratefrom the shallow aquifer to the intermediate and confinedaquifers:

1. The sandy intervals of the Corcoran clay allowed PCEmigration into the confined aquifer.

2. Downward gradients allowed PCE to penetrate theclay units separating the aquifers.

3. PCE migrated down the annuluses of older watersupply wells that did not have a surface sanitary seal orimpermeable backfill.

Table 4 Summary of evaluation methods for City of Merced site

Evaluationmethod

Description Discon-tinuous(pinchesout/sandy)a

Pene-tration

Wellannulus

Wellscreen

Upgra-dientpinchout

Summary of analysis

1. Groundwatergradients

G. Downward ? 4 Downward gradient induces downwardmovement of contaminants through theaquitard. Analysis under #4 below showsthat the rate of penetration through theaquitard is fast enough for the required timeframe (approximately 20 years) for theshallow clay, but not fast enough for theCorcoran clay

4. Hydraulicconductivity

D. The timefor transmissioncalculation

4 4 Calculated travel time (advection only)from source area in upper aquifer to plumein lower aquifer:through shallow clay aquitard=15 yearsthrough Corcoran clay aquitard=300 yearsupgradient path, from source to intermediateof confined aquifer=3–25 years (dependingon source location and specific aquiferconditions nearby)Contamination was detected within 20 years,setting an upper limit on travel time. Theseresults imply that travel through the aquitardor around the aquitard are both possible

5. Generalgroundwaterflow directions

D. Upper aquiferflows towardupgradient (in aregional sense)pinchout, and theintermediate andconfined aquifersflow away

4 Flow direction information is from studiesby the CVRWQCB and the USGS

6. Changes inconcentrationduringpumping

C. Concentrationdecreases over time

? 4 Two different pumping-concentration testsshow decrease in PCE levels over time(a drop from 20 to 1 ppb over 24 h in thefirst test and a drop from 53 to 23 ppb over2 h in the second test)

7. Evaluationof leaky wells

B. Record of poorbackfill or screenedinterval selection

4 4 The Central Valley RWQCB reportsimproperly abandoned wells in the vicinityRecords of various city and private wellsindicate screens across both the intermediateand confined aquifers and across theCorcoran clay. City wells did not havesanitary seals to the surface

8. Aquitardcontinuity

B. Significant gapsor sandy zones

4 4 4 Records of various city and private wellsindicate substantial sandy zones in theCorcoran clay and complete upgradientpinchout of both the shallow clay andCorcoran clay

C. Completepinchout of aquitarddetected

a ?, test is inconclusive for identifying contaminant travel path; 6, test assists in eliminating specific travel path; 4, test supports a specifictravel path

66

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

4. PCE migrated between aquifers through wells thatwere screened across more than one aquifer.

5. PCE migrated from the shallow aquifer to the inter-mediate and confined aquifers where all three are incontact west of the pinchouts of the clays that normallyseparate the aquifers.

Likely travel pathTable 4 summarizes the analysis of potential travel paths.Sandy zones identified in the Corcoran clay validate thepotential of PCE penetrating the clay. Similarly, thedownward groundwater gradients, used to calculate pen-etration rates, confirm that PCE could penetrate theShallow clay within 15 years. Stratigraphic informationconfirming the pinchout of both clay layers, coupled withflow direction information in all aquifers and calculationsof travel time show that PCE could have entered the loweraquifers from a flow path around the upgradient pinchoutsof the clays. Finally, well construction records and in-formation on well completions available in the technicalliterature indicate the PCE could travel down well annu-luses or across aquifers via well screens.

In summary, none of the pathways could be eliminat-ed, and quite likely all the pathways have contributed tosome degree, considering the large number of sources, thesize of the site, and the length of time involved. It wasconcluded that the best method to distinguish between thepathways would be a careful fingerprinting, statisticalanalysis or ratio analysis of contaminants from eachsource and from locations within each aquifer. Pumpingtests did not prove to be effective in evaluating sourcesbecause there are too many interfering factors, includingmultiple water supply wells that would be in operationduring the test and a site area too large to be evaluatedwith a single pumping well.

Conclusions

A number of tools to detect and evaluate pathways ofaquifer cross-contamination were described, and thenapplied at three sites. Because the investigations for allthree sites had been completed in the past, there was noopportunity to tailor the investigative elements to bestidentify pathways of cross-contamination. Nevertheless,several observations can be made on the quality of eachtool:

1. In the absence of specially-designed investigative el-ements, calculations of water (and therefore contami-nant) flow rates using hydraulic conductivity, gradient,and aquitard continuity information provided thestrongest arguments for or against certain flow paths.

2. Evidence of aquitard pinchout or discontinuity (sandyintervals) strongly suggests certain flow paths, butmust be coupled with other analysis tools to conclu-sively eliminate or confirm paths.

3. Similarly, evidence of leaky wells must be used inconjunction with other tests to establish their influenceon cross-contamination.

4. Aquifer pumping tests are useful only if confoundingfactors (such as atmospheric pressure changes, leakagedue to streams, and leakage from multiple locations)can be identified and accounted for in the interpreta-tion.

While fingerprinting, statistical analysis or ratio analysisof contaminants from various sources and receptors wasnot completed for any of the evaluated sites, the authorsbelieve that these techniques have strong potential forconfirming or eliminating pathways. Future research inthis area is suggested. Furthermore, the interpretation ofpumping-concentration tests is not well defined and needsfurther assessment.

Acknowledgments Brian Aubry (currently of Geologica, Inc.) andAllison Remple developed the conceptual groundwater flow modelat the Ft. Ord site. Julio Guerra of the City of Merced generouslyshared technical information on their regional PCE problem.

References

Abriola LM, Pinder GF (1985) A multiphase approach to themodeling of porous media contamination by organic com-pounds. 2. Numerical simulation. Water Resources Res 16:19–26

Aller L (1984) Methods for determining the location of abandonedwells, EPA-600/2–83/123. US Environmental ProtectionAgency, Ada, Oklahoma

Alther GA (1979) A simplified statistical sequence applied toroutine water quality analysis: a case history. Ground Water17:556–561

Back W (1966) Hydrochemical facies and groundwater flow pat-terns in northern part of Atlantic coastal plain. Washington,DC, US Geol Surv Prof Pap 498-A

Bethune DN, Farvolden RN, Ryan MC, Guzman AL (1996) In-dustrial contamination of a municipal water-supply lake byinduced reversal of ground-water flow, Managua, Nicaragua.Ground Water 34:699–708

Cardimona S, Clement WP, Kadinsky-Cade K (1998) Seismic re-flection and ground penetrating radar imaging of a shallowaquifer. Geophysics 63:1310–1317

Clement WP, Cardimona S, Endres AL, Kadinsky-Cade K (1997)Site characterization at the groundwater remediation field lab-oratory. Lead Edge Nov:1617–1621

Dames & Moore (1991) Remedial investigation report, Fort Ordlandfills, Fort Ord, California, for U.S. Army Corps of Engi-neers. Dames & Moore, Consult Rep

Douglas GS, McMillen S (1996) The stability and utility of diag-nostic ratio hydrocarbon fingerprinting for soils contaminatedwith petroleum hydrocarbons, AAPG/SEPM Annu Meet Abstr5:38–39

Elliott AL (1984) Ground-water conditions and shallow test-wellinformation in the eastern half of Merced County, California,1977–82. US Geol Surv Water Resources Inv Rep 83-4081

Farnham IM, Stetzenbach KJ, Singh AK, Johannesson KH (2000)Deciphering groundwater flow systems in Oasis Valley,Nevada, using trace element geochemistry, multivariate statis-tics, and geographical information system. Math Geol 32:943–968

Faust CR (1985) Transport of immiscible fluids within and belowthe unsaturated zone: a numerical model. Water Resources Res21:587–597

67

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8

Fetter CW (1999) Contaminant hydrogeology, 2nd edn. Prentice-Hall, Upper Saddle River, New Jersey

Fryberger JS, Tinlin RM (1984) Pollution potential from injectionwells via abandoned wells. In: Fairchild DM (ed) Proc 1st NatlConf Abandoned Wells: Problems and Solutions. Environ-mental and Ground Water Institute, University of Oklahoma,Norman, Oklahoma, pp 84–117

Gass TE, Lehr JH, Heiss HW (1977) Impact of abandoned wells onground water. US Environmental Protection Agency, Wash-ington, DC, Rep EPA/600/3-77-095

G�ler C, Thyne GD, McCray JE, Turner AK (2002) Evaluation ofthe graphical and multivariate statistical methods used forclassification of water-chemistry data. Hydrogeol J 10:455–474

Izzo VJ (1989) Dry cleaners, a major source of PCE in the groundwater. Central Valley Regional Water Quality Control BoardRep 19 September 1989

Jaquet JM, Froidevoux R, Verned JP (1975) Comparison of auto-matic classification methods applied to lake geochemicalsamples. Math Geol 7:237–265

Javandel I, Tsang CF, Witherspoon PA, Morganwalp D (1988)Hydrologic detection of abandoned wells near proposed injec-tion wells for hazardous waste disposal. Water Resources Res24:261–270

Kueper BH, Frind EO (1991) Two-phase flow in heterogeneousporous media. 2. Model application. Water Resources Res27:1059–1070

Lacombe S, Sudicky EA, Frape SK, Unger AJA (1995) Influence ofleaky boreholes on cross-formational groundwater flow andcontaminant transport. Water Resources Res 31:1871–1882

Lin C, Pinder GF, Wood EF (1982) Water resources program,Princeton University, Princeton, New Jersey, Rep 83-WR-2

Lindenberg KA (1997) Down the drain: a ground penetrating radar(GPR) survey of a trichloroethylene (TCE)-contaminated site inCamben County, New Jersey. BSc Thesis, Princeton Univer-sity, Princeton, New Jersey

Locke RA (1994) Immediate pumping effects on inorganic waterchemistry and nitrate stratification in the Iowa River alluvialaquifer near Marengo, Iowa. MSc Thesis, University of Iowa,Iowa City, Iowa

Luhdorff & Scalmanini (1987) Analysis of wells 3A, 3B, and 5,City of Merced, California. Luhdorff & Scalmanini ConsultRep

Lusczynski NJ (1961) Head and flow in ground water of variabledensity. J Geophys Res 66:4247–4256

Mariner PE, Holzmer FJ, Jackson RE, White JE, Wahl AS, WolfFG (1997) Fingerprinting arsenic contamination in the sedi-ments of the Hylebos Waterway, Commencement Bay super-fund site, Tacoma, Washington. Environ Eng Geosci 3:359–368

McCray JE, Falta RW (1997) Numerical simulation of air spargingfor remediation of NAPL contamination. Ground Water 35:99–110

Meiri D (1989) A tracer test for detecting cross contaminationalong a monitoring well column. Ground Water Monitor Rev9:78–81

Mercer JW, Cohen RM (1990) A review of immiscible fluids in thesubsurface: properties, models, remediation, and characteriza-tion. J Contam Hydrol 6:107–163

Moore DJ (1991) The delineation of a surficial aquitard in CalvertCity, Kentucky using electrical resistivity and borehole infor-mation. MSc Thesis, University of Kentucky, Lexington,Kentucky

Neuman SP, Witherspoon PA (1969) Theory of flow in a confinedtwo-aquifer system. Water Resources Res 5:803–816

Neuman SP, Witherspoon PA (1972) Field determination of thehydraulic properties of leaky multiple-aquifer systems. WaterResources Res 8:1284–1298

Oolman T, Godard ST, Pope GA, Jin M, Kirchner K (1995)DNAPL flow behavior in a contaminated aquifer. GroundWater Monitor Rev Fall, pp 125–137

Powers SE, Villaume JF, Ripp JA (1997) Multivariate analysis toimprove understanding of NAPL pollutant sources. GroundWater Monitor Remed 17:130–140

Walker S (1988) Soil gas survey. Merced PCE Inv, Central ValleyRegional Water Quality Control Board Memo, 8 August 81988, file AB1803

Walker S (1989) Well investigation program assessment report,City of Merced wells 3A, 3B, 5, 2A, and 2B. Central ValleyRegional Water Quality Control Board Rep, 9 June 1989

Welenco (1995) Water and environmental geophysical well logs,7th edn. 1. Technical information and data. Welenco, Inc.,Bakersfield, California

Williams RE (1982) Statistical identification of hydraulic connec-tions between the surface of a mountain and internal mineral-ized zones. Ground Water 20:466–478

Zemo DA, Bruya JE, Graf TE (1995) The application of petroleumhydrocarbon fingerprint characterization in site investigationand remediation. Ground Water Monitor Remed 15:147–156

68

Hydrogeol J (2006) 14:51–68 DOI 10.1007/s10040-004-0403-8