design of an mtbe remediation technology evaluation

Design of an MTBE RemediationTechnology Evaluationby Ann Azadpour-Keeley and Michael J. Barcelona

AbstractThis study examines the intrinsic variability of dissolved methyl tert-butyl ether (MTBE) concentrations in ground water

during the course of a pilot-scale bioremedial technology trial in Port Hueneme, California. A pretrial natural gradient tracerexperiment using bromide was conducted in an anaerobic test section of the aquifer to characterize hydrogeology. The resultsshowed the presence of a complex velocity field in terms of vertical stratification and preferential flowpaths. The hydraulicconductivity at the test area varied by >2 orders of magnitude, and the effects of vertical stratification were made apparent bythe tracers’ detection pattern, which was predominately higher in the lower part of the aquifer. Since historically the lowerportion of the aquifer significantly influenced MTBE transport, it was emphasized by increasing the sampling frequency forMTBE and tracers during the pilot test that involved the intermittent addition of oxygen and propane into the aquifer. A sec-ond tracer experiment using bromide and deuterated MTBE (2H12-MTBE) was conducted at the onset of the technology trialand after the aquifer was made aerobic. The continuous metering of the tracer solutions into the test area was maintained for300 d. The results showed that 2H12-MTBE behaved as a conservative tracer since (1) its concentrations increased throughoutthe study approaching its designed injected level and (2) the pattern of its detection resembled that of bromide. On the otherhand, 2H12-MTBE, which was purposefully introduced into the aquifer, behaved differently from that of the existing dis-solved MTBE plume that emanated from a non–aqueous phase liquid (NAPL) source over a decade ago, thereby undergoingyears of diffusion. The data imply that a detailed understanding of the complexity of the flow field was not possible byobserving the intrinsic MTBE data alone.

IntroductionVariability in ground water measurements is funda-

mentally the result of two separate phenomena. One isintroduced by the location and construction of monitoringpoints, and the collection and analysis of discrete samples.The other is a result of the heterogeneity of the system(Leblanc et al. 1991) being sampled including both the ad-vective flow of water (Sudicky 1986) and the distributionof chemicals of concern (Reilly and Gibs 1993). Over thepast 30 years, research has been focused on developingspecific methods to minimize the impact of sources ofintroduced variability including sampling techniques, moni-toring systems, location of biologically active zones, decayrates, and predictive models to characterize contaminantfate and transport in the subsurface. The intrinsic variabil-ity of ground water measurements is a function of the sys-tem (Freeze et al. 1990; National Research Council 1990)being sampled and, therefore, cannot be reduced.

The water quality of shallow, unconfined aquifers hasbeen observed to change significantly and rather rapidly(Harter et al. 2002; Ronen et al. 1987), perhaps by as muchas 1 order of magnitude within a few hours to days. Poten-tial causes for variability in contaminant concentrationsdue to infiltration and recharge processes may pose signifi-cant challenges to the evaluation of the performance ofremedial technology demonstrations over relatively shortdistances and timeframes.

Well construction practices, sampling techniques, andlaboratory analyses are also documented sources of vari-ability in ground water quality data (Barcelona et al. 1989;Keith et al. 1983; Wilson and Rouse 1983; Nightingale andBianchi 1979), while other possible sources have beenidentified (Pettyjohn 1976). To overcome the artifact causesof spatial and temporal variability of data in ground waterstudies, investigators constructed pumps of inert materialand inflatable packers, studied the sorptive characteristicsof a variety of sampling materials, and developed low-flowsampling techniques (Puls and Barcelona 1996). As im-proved data came to light, studies turned to the use of sta-tistical methods and predictive models to reduce, or at least

Journal compilationª 2006 National GroundWater Association.No claim to US government works.

Ground Water Monitoring & Remediation 26, no. 2/ Spring 2006/pages 103–113 103

understand, the causes of spatial variability (Aguirre andHaghighi 2003; Wheater et al. 2000; Chang et al. 1999;Dou et al. 1997; Vomvoris and Gelhar 1990; Harris et al.1987; Hoeksema and Kitanidis 1985) and their influenceon ground water monitoring results (Montgomery et al. 1987).Now, there are powerful statistical analysis and decision-making tools to enable reduced sampling density andfrequency without sacrificing statistical reliability (Azizet al. 2003a, 2003b).

Fuel spills and leaking storage tanks have resulted inthe discovery of methyl tert-butyl ether (MTBE) contami-nation in shallow ambient ground water, particularly inurban areas (Squillace et al. 1996; Salanitro et al. 2000).Thus, there is considerable interest in developing passiveand active approaches for in situ treatment of MTBE. Interms of natural attenuation of fuel spills, a compellingargument can be made concerning deficiencies in earlyBTEX (i.e., benzene toluene, ethylbenzene, and xylenes)ground water monitoring findings with respect to accurateassessments of the processes involved (biodegradation,dispersion, dilution, sorption, volatilization, and stabiliza-tion) in the intrinsic remediation of more mobile and per-sistent MTBE in ground water (Amerson and Johnson2003).

The focus of this study was to design a monitoringsystem to understand site hydrogeology and quantifythe intrinsic variability of MTBE during the course of apilot-scale ground water bioremedial technology trial. Theremedial technology performance itself has been reportedseparately (Azadpour-Keeley 2002).

Field Site DescriptionThe field data were collected at the Department of

Defense National Environmental Technology Test Site atthe Naval Base Ventura County (NBVC) at Port Hueneme,California, located ~40 miles (65 km) northwest of LosAngeles. The Naval Exchange (NEX) service station wasthe source of a fuel plume that exists at the facility. Ac-cording to NEX inventory records, ~4000 gallons (15,160 L)of leaded and 6800 gallons (25,772 L) of unleaded pre-mium gasoline were lost from the fuel distribution linesbetween 1984 and 1985. Historical data from wells anddirect-push sampling indicated that a long and narrowBTEX/MTBE plume emanates from the NEX service sta-tion and extends ~5000 feet (1.5 km) from the contamina-tion source in the shallow aquifer (Figure 1). Contaminantconcentrations indicated that an NAPL source still existedand the constituents of the dissolved plume formed twooverlapping regions: one composed of both BTEX andMTBE, while the other contained only MTBE.

In general, the geology at the NBVC site consists ofunconsolidated sediments composed of sands, silts, clays,fill material, and minor amounts of gravel. The fill layerhad been applied to much of the base extending from 0 to~10 feet (3.0 m) below ground surface (bgs). Two othernaturally deposited units extend below the fill layer (Salanitroet al. 2000). A shallow unconfined aquifer is the uppermostwater bearing unit, which is comprised of three depositional

units: an upper silty sand, an underlying fine- to coarse-grained sand, and a basal clay layer at ~20 feet (~6 m) bgs.According to Kram et al. (2000), the saturated depth forthe entire plume averages ~15 feet (4.6 m) with seasonalfluctuations ranging between 1 and 2 feet (0.3 to 0.6 m).

Site CharacterizationThree zones were characterized within the contaminant

plume for redox and contaminant concentrations as well ashydraulic conductivity prior to the final selection of thedemonstration site. A GeoprobeTM (Kejr Engineering, Inc.,Salina, KS) was used to install permanent cross gradientmonitoring wells in the source zone (five wells), middlezone (two wells), and wellhead protection zone (six wells).Except for the two permanent wells in the middle zone, therest were temporary sampling points. These locations aredifferentiated by their distance from the source and areidentified as follows: the source zone is located within theimmediate vicinity of the fuel leak and characterized byhigh concentrations of MTBE and BTEX. The middle zonewas the area midway downgradient along the MTBEplume, which contained moderate concentrations of MTBE(5 to 10 parts per million [ppm]). The wellhead protectionzone was farthest downgradient along the plume and con-tained MTBE at lower concentrations than the first twozones (<1 ppm). Dissolved oxygen (D.O.) measurements

Figure 1. Port Hueneme plume map.

A. Azadpour-Keeley and M.J. Barcelona/ Ground Water Monitoring & Remediation 26, no. 2: 103–113104

indicated that anaerobic conditions prevailed throughoutthe site (D.O. < 1 mg/L). Water quality measurements wereindicative of brackish water. Outside of the BTEX zone,the ground water was characterized as having high sulfate,low nitrate, with variable concentrations of iron (II) (0.2 to12 mg/L), and a near neutral pH (6.8 to 7.2).

MTBE data were used in the final selection of ourstudy site, which was in the middle zone located ~50 feetcross gradient from the Salanitro et al. (2000) study site.Monitoring wells CBC-45 and CBC-46 represent groundwater quality conditions within the dissolved MTBE plumenear the site (Figure 1). Various parameters examined attwo locations included total organic carbon (TOC) (10.8mg/L), nitrate (<0.1 mg/L), chloride (125 mg/L), sulfate(1270 mg/L), D.O. (0.36 mg/L), and pH (6.9) at CBC-45and TOC (5.12 mg/L), nitrate (<0.1 mg/L), chloride (820mg/L), sulfate (1410 mg/L), D.O. (0.15 mg/L), and pH(6.97) at CBC-46. Based on CPT pushes, the upper siltysand unit ranged between 8 and 10 feet (2.4 to 3.0 m) thick,and the underlying sand was ~12 to 15 feet (3.6 to 4.5 m)thick. Static water levels were generally encountered atdepths between 6 and 8 feet (1.8 to 2.4 m) bgs. Assumingseasonal fluctuations ranging between 1 and 2 feet yieldeda saturated aquifer thickness of 12 to 14 feet (3.7 to 4.3 m).The porosity was between 0.30 and 0.35, and the hydraulicgradient varied from 0.001 to 0.003, with the flow ofground water generally to the southwest.

With respect to the aquifer tests conducted at theNBVC site, Kram et al. (2000) indicated that the hydraulicconductivity in the plume area varies between 10�1 and10�4 cm/s. Conductivity tests have been made using direct-push wells and drilled wells. Pumping tests, slug tests,tracer tests, and downhole flowmeter surveys have pro-vided a wide range of aquifer properties and apparent seep-age velocities.

The results of the large number of hydraulic conductiv-ity tests that have been done throughout the NEX plumeare summarized in Table 1. These data and calculated groundwater velocities given a gradient of 0.002 and porosity of0.3 show a similar degree of hydrogeologic variability atthe site as observed at various scales by a number of inves-tigators. Amerson and Johnson (2003), who conducted the

most comprehensive plume-scale flow and transport stud-ies, noted that there was considerable vertical variability inhydraulic conductivity at all study locations.

It should be noted that in more recent work, Bartlettet al. (2004) reported that hydraulic conductivities down-gradient of the middle zone averaged 2 3 10�2 cm/s witha standard deviation of 8 3 10�3. They state that variancesin the K values in individual wells were much lower thanvariances computed for K values from a number of wells inthe area. They suggested that the differences in K valuesamong wells were due more to formation heterogeneitythan well construction, installation, or the test method.

Our data from pump testing before and after the initialpretrial tracer test in the test location yielded K values thatwere within the range of values reported for the midplumeand leading edge of the plume zones. These results andthose of a flowmeter survey that was conducted at GWT-1and GWC-1 (Figure 2) on a scale of <1 m vertically com-pared well with the results of Salanitro et al. (2000) andshowed the variability of hydraulic conductivity with depthat the site. The data suggested that hydraulic conductivityincreased with depth within the upper 10 feet of the satu-rated zone with the most conductive aquifer materialsbetween ~16 and 19 feet. We anticipated, therefore, thata higher frequency of sampling may be needed in thedeeper portion of the aquifer compared to the upper por-tions to capture the dynamics of oxygen and propane injec-tion and to detect MTBE degradation during the remedialtrial.

Monitoring NetworkTwo plots were used for sampling and analysis of

ground water constituents (Figure 2). Both plots had 19staggered, fully screened tracer injection wells ~2 feet (0.6 m)apart. A series of well clusters to allow the collection ofground water samples were installed with 21-inch (0.7 m)screens at 10, 14, and 18 feet (3.0, 4.3, and 5.5 m) bgs. Thetest plot had 6 monitoring wells upgradient from the tracerinjections wells and 14 downgradient monitoring wells.The control plot had 4 upgradient and 10 downgradientmonitoring wells. Both plots were ~20 feet (~6 m) square

Table 1Hydraulic Characterization within the NEX BTEX-MTBE Plume

Hydraulic Conductivity (cm/s) Ground Water Velocity1 (m/year) Location Reference

0.076 160 Source This study0.06–0.14 60–400 Midplume Salanitro et al. (2000)0.005–0.008 11–17 Midplume Salanitro et al. (2000)0.002–0.45 90–180 Source Amerson and Johnson (2003)0.052–0.135 109–283 Midplume This study0.035–0.041 74–86 Midplume This study0.088–0.0152 32–185 Leading edge This study0.02 (60.008) 42 Leading edge Bartlett et al. (2004)

1Ground water gradient varied from 0.001 to 0.003 (m/m).Note: For this table, a value of 0.002 was used for comparison purposes. A porosity of 0.3 was used in all cases.

A. Azadpour-Keeley and M.J. Barcelona/ Ground Water Monitoring & Remediation 26, no. 2: 103–113 105

as measured from the upgradient to the last downgradientwells and the span of the injection wells. In total, therewere 140 individual sampling points including monitoringwells and the injection wells. The original monitoring planstipulated seven sampling events at every monitoring loca-tion for constituents of interest (i.e., geochemical parame-ters, the primary contaminants and likely metabolites).Realizing that the flow was stratified, the tracer test wouldbe valuable to further support data inferred by the flowme-ter survey in identification of dominant flowpaths so as tooptimize the frequency of sampling. Tracer materials,including bromide (NaBr, 352 mg of bromide ion/well/d)and fully deuterated MTBE (2H12-MTBE, 23 mg/well/d),were metered (10 mL/well/d) under the natural gradientinto each injection well so as to deliver a known and con-stant concentration to the ground water. The first bromideground water flow characterization study was conductedover a 60-d period. Three months later the 2H12-MTBEtracer study was initiated and was monitored for a period ofover 300 d. The rate of injection into the 38 tracer wellsamounts to ~0.1% of the ground water flow.

Results and Discussion

Chronology of the StudyBefore the technology trial began, a site characteriza-

tion tracer test was initiated, which included the injectionof bromide for 30 d with the system being monitored fora total of 60 d. An integral part of the technology evalua-tion included the injection of 2H12-MTBE into the aquifer,which commenced at the onset of the technology demon-stration and was monitored for a period of >300 d. A sec-ond bromide tracer test was initiated ~5 months after thestart of the technology evaluation phase of the investigationto determine if the addition of oxygen and propane alterflow characteristics.

Bromide Tracer TestThe aim of the first bromide test was primarily to

characterize flowpath continuity between the injection wellsand downgradient monitoring points, thereby obviating theneed to sample every screen location during the succeed-ing remediation evaluation. The bromide tracer test wascarried out under natural ground water flow conditions byadding 352 mg of bromide ion into each injection well ata rate of 10 mL/d. Since the average background bromidefrom the upgradient monitoring locations was 1.07 mg/L(s ¼ 0.17), a level of 2.0 mg/L was selected as the tracerbreakthrough concentration (detection limit of 0.125 mg/L).A finding of considerable importance was the distributionand frequency of bromide detections at all screens withinthe two experimental plots.

Each of the 38 injection wells were sampled 15 timesduring the 60-d period. The average bromide concentrationfor each injection well in the test and control plots duringthe predemonstration study is shown in Figures 3a and 3b.Although the average bromide concentration varied bya factor of 2 between injection wells and it was not uncom-mon for the concentration in a single well to vary by overa factor of 2, the mean bromide concentration for all injec-tion wells in both plots during this period was consistent(test plot, 51.9 mg/L [s ¼ 10.8]; control plot, 51.0 mg/L[s ¼ 10.9]). The bromide injection flow rate was monitoredregularly during the predemonstration phase of the study,and its consistency is supported by the mean and standarddeviation of bromide concentrations in the injection wells.Given this consistency, variations between wells are associ-ated with the exchange rate between aquifer and tracer wellwater, which is governed by well efficiency and variabilityin hydraulic conductivity.

During the 60-d site characterization study, 586 sam-ples were analyzed for bromide in the control plot. Ofthese, 124 were 2.0 mg/L or greater in deep screens and 19were 2.0 mg/L or greater in the middle and upper screens.

Figure 2. Test and control plots layouts.


During the same period, 672 bromide analyses were madein the test plot of which 111 were 2.0 mg/L or greater inthe deep screens and only 4 were 2.0 mg/L or greater inthe middle and upper screens. The bromide tracer behaviorprovided considerable information with respect to the ver-tical stratification of ground water flow patterns includingthe determination that flow was consistently higher in thedeeper portion of the aquifer. The results of the initial bro-mide tracer test enabled the sampling plan to be modifiedfor the remedial technology evaluation by reducing thesampling frequency from 7 to 5 at the upper screens andincreasing from 7 to 15 at the lower screens. Thus, wellswith relative high hydraulic conductivity were sampledwith higher frequency from the more permeable portion ofthe aquifer.

Since the flowmeter measurements indicated that thehydraulic conductivity at the site varies by 2 orders ofmagnitude, the movement of ground water in the upperportions of the aquifer could be expected to be ~10 to 100times slower than the lower portions; therefore, a longerbromide arrival time should have been anticipated. Thearrival time delay was addressed by increasing intrinsicMTBE sampling events (from 5 to 7) in the upper portionsof the aquifer during the subsequent tracer test of >300 d,and by scheduling at least five full sampling rounds forMTBE and 2H12-MTBE between the 6th and 15th sam-pling events. Even though budgetary and time constraintsare common while conducting pilot field testing, in

practice it is the ‘‘hard to reach areas’’ such as the upperportions of the aquifer of this trial that present a challengeto remediation. In this event, a combination of an activeremedy for the more permeable portion of the aquifer andmonitored natural attenuation for the less permeable por-tions of the aquifer may provide a suitable cleanup option.

The variability in flow between the test and controlplot, after 30 d of tracer injection, was further evidenced bybromide concentrations at the bottom screens in the firstdowngradient monitoring transect as shown in Figures 4aand 4b. The transect was ~8 feet (2.4 m) from the injectionwell curtain. The tracer arrived at the bottom screens in thefirst transect in the test plot in ~30 d and in the control plotwithin ~10 d after injection began. Tracer responses in thecontrol transect wells were fairly similar, with the excep-tion of well C24B that showed no apparent response duringthe test. It should also be noted that the bromide frontpassed the first transect in the test plot in ~55 d comparedto 30 to 40 d in the control plot.

The tracer test sampling plan design was focused ona more detailed characterization of flow rate through theexperimental plots. This was accomplished by determiningthe relative time of travel based on the initial tracer break-through time and distance between the injection wells anddowngradient observation points as shown in Table 2.Since the times provided were initial breakthrough values,

Figure 3. (a) Test plot tracer injection wells. (b) Control plottracer injection wells.

Figure 4. (a) Bromide concentration in test plot at deepscreens in first downgradient transect. (b) Bromide concentra-tion in control plot at deep screens in first downgradienttransect.


they can only be used as an estimate of velocity at the deepscreens. The ground water velocity would be expected tobe somewhat slower.

Estimated ground water velocities in Table 2 demon-strate the level of flow variability with depth at the site;those in the control plot ranged between 0.01 and 1.21 ft/d (0.01 and 0.37 m/d), while those in the test plot werebetween 0.11 and 0.94 ft/d (0.03 and 0.29 m/d). The aver-age flow velocity in the deeper formation in the controlplot was 0.48 ft/d (0.15 m/d, s ¼ 0.33) and 0.32 ft/d (0.10m/d, s ¼ 0.20) in the test plot. With respect to ground watervelocities in the upper portions, the flowmeter measure-ments, which evaluated the variability of hydraulic conduc-tivity with depth, suggested that the hydraulic conductivityvaries by 2 orders of magnitude and the natural groundwater flow at these locations was limited. Even thoughsaturated, only two monitoring wells (one in each plot) inthe upper portions provided bromide breakthrough curvesresulting in an estimated ground water velocity at thoselocations of 0.17 ft/d (0.05 m/d) during the predemonstra-tion characterization study.

Intrinsic MTBEGround water MTBE concentrations were determined

prior to and after the initial bromide tracer test in the areaof the site for background and design purposes. About 6months before the initiation of the technology evaluation,the average MTBE concentrations in the deep screens atthe four locations were relatively high including downgra-dient test plot at 5329 lg/L (s ¼ 2112), upgradient test plotat 4900 lg/L, downgradient control plot at 5989 lg/L (s ¼1852), and upgradient control plot at 4800 lg/L. However,when the remediation phase of the study began, the down-gradient test plot was at 1493 lg/L (s ¼ 1292), upgradienttest plot at 2160 lg/L (s ¼ 638), downgradient control plotat 3471 lg/L (s ¼ 1673), and upgradient control plot at4580 lg/L (s ¼ 484). With the observed variations inhydraulic conductivity and tracer breakthrough responses,as discussed previously, it would be generally expected thatthe MTBE concentrations in the plume leaving the sourcearea would also be heterogeneous. The impact of the aqui-fer heterogeneity on tracer distribution at the source area ofthe Port Hueneme plume was also suggested by others(Amerson and Johnson 2003).

With respect to vertical distribution, concentrations ofintrinsic MTBE measured during the first bromide

sampling and throughout the study were generally low atthe upper portions of the aquifer. For example, during the10-month study, the average MTBE concentration at thedeep screens in the control plot was 2238 lg/L comparedto 360 lg/L at the middle screens and 38 lg/L at the upperscreens. In the test plot, the deep screen MTBE concentra-tion averaged 597 lg/L compared to 252 lg/L at the mid-dle screens and 55 lg/L at the upper screens. Since theexperimental plots of this study were located at the lateraledge of the MTBE dissolved plume (Figure 1), low MTBEconcentrations at the upper portions of the aquifer were notunexpected.

As can be seen in Figure 5, intrinsic MTBE concen-trations in the test plot varied through the observationperiod where nodes appeared and moved downgradient. Inother areas, MTBE was very low during one samplingevent only to appear in others. The least variable observa-tion was during sampling event 7 when the MTBE concen-tration ranged from 87 to 1830 lg/L, while the mostvariable observation was sampling event 9 where theMTBE concentration ranged from 10 to 2340 lg/L. Thebehavior in the control plot was much the same. It isimportant to note, once again, that these variations occurwithin a very small area of ~20 feet (~6 m) square overa sampling period of 10 months. Since in general, the sitewas paved and devoid of rooted trees, and there was nodocumented evidence that the hydrology was altered bypumping or artificial recharge, it may be suggested that theobserved behavior was influenced by the dispersion ofMTBE during its migration from the source area and het-erogeneity in aquifer media and the flow field as one wouldexpect in nearly every plume.

Deuterated MTBE/Bromide Tracer TestThe remediation evaluation phase of the study was

conducted during a period of >300 d. The injection of2H12-MTBE was an integral part of the technology trial toaddress the anticipated variability of intrinsic MTBE. Over300 g of 2H12-MTBE were uniformly (23 mg/well/d) in-jected into the aquifer at a rate of 10mL/well/d. The use ofstable isotopes in environmental studies related to petro-leum hydrocarbons was pioneered by Thierrin et al. (1992,1993, 1995). Stable isotope–labeled tracers are valuabletools for field-scale retardation and biodegradation studiessince they avoid the limitations of using a more brute-forcemonitoring approach or regulatory approval problems

Table 2Estimated Seepage Velocities (m/d) in Control and Test Plots

First Bromide Test1 Second Bromide Test2 Deuterated MTBE3

Plot Range Average (N) Range Average (N) Range Average (N)

Control 0.01–0.37 0.15 (6) 0.036–0.37 0.11 (6) 0.036–0.056 0.05 (4)Test 0.03–0.29 0.1 (6) 0.036–0.06 0.05 (6) 0.036–0.065 0.05 (2)

1Initiated February 1, 2001—duration, 60 d.2Initiated October 29, 2001—duration, 214 d.3Initiated June 8, 2001—duration, >300 d.


associated with the use of radiolabeled isotopes. Thoughbromide has been used extensively as a ground watertracer, in recent years the simultaneous application ofbromide with an electron acceptor (Wilson et al. 2002),bromide and MTBE (Gray et al. 2002), or bromide and2H12-MTBE as a surrogate (Amerson and Johnson 2003)has been reported. These combinations have providedinsight with respect to the ground water flow as well as thecontaminants’ response to both biotic and abiotic pro-cesses, and provide the data required to quantitatively eval-uate performance and the cost of treatability studies.

In our study, due to the stability in delivering constantamounts of tracers, sampling from the injection wellscould be reduced to once per month and still adequatelydescribe the system variability. Since a total of 7 g of2H12-MTBE had previously (1998) been injected in thesource zone, several hundred feet upgradient from our site(Amerson and Johnson 2003), it became necessary to per-form background sampling on two occasions to assure thatit was not present (detection limit, 0.2 lg/L). Throughoutthe evaluation phase, at least four samples were collectedduring each of the 15 sampling events from wells

upgradient of the two experimental plots. The resultsshowed that 2H12-MTBE was not present.

Using the results from the first bromide site character-ization test, most of the sampling density in the three aqui-fer horizons could be focused on the last 6 months of thetrial (sampling events 6 to 15) to allow the arrival of trac-ers. The results indicated that at the termination of the trial(monitoring event 15), over 80% of the downgradient (con-trol and test plots) monitoring shallow, middle, and deepscreens contained 2H12-MTBE (above 0.2 lg/L). It is notedthat response to 2H12-MTBE in the deep screens in the firstcontrol plot downgradient monitoring transect was similarto that of bromide from the first tracer test, at least withrespect to the magnitude of response. As with bromide, nodetection was observed in well C24B. The tracer responsecurves for 2H12-MTBE at this location are shown for theentire project period in Figure 6.

Influence of the in Situ BiostimulationBioaugumentation Remedial Trial

It must be pointed out that during the remediation eval-uation phase, oxygen and propane were intermittently

Figure 5. MTBE concentrations (mg/L) in test plot during different sampling events.


injected into the aquifer (oxygen only in the control plot)as part of the in situ bioremediation technology trial(Azadpour-Keeley 2002). The gases were delivered to theground water only four times daily lasting 5 min (1% of thetime), while adding ~1 pound of oxygen and 0.3 pound ofpropane per day. In order to avoid potential bias due to theaddition of gases, all water level measurements and sam-pling activities occurred only after the static water levelshad become stable. The short-term addition of oxygen andpropane unquestionably altered the local permeability andhydraulic gradient briefly (Johnson et al. 2001a, 2001b), soan attempt was made to determine the extent that the natu-ral variability was affected by comparing before and aftergas addition tracer response curves. In order to make thiscomparison, the second bromide test was initiated 5months after the addition of gases began. A response curvein the control plot at a first downgradient transect beforegasses were added (Figure 4b) was compared with thatafter the addition of gases (Figure 7). In both cases, wellC23B was the most responsive while wells C21B andC22B were less responsive but behaved similarly. Also, inboth cases, well C24B continued to show no response.

Table 2 contains a summary of estimated seepagevelocities, based on breakthrough times, in the control andtest plots during the first and second bromide and 2H12-MTBE tracer tests. In the control plot, the range and aver-age seepage velocities from bromide breakthrough timeswere found to be statistically similar in both tests. The testplot seepage velocities were somewhat lower, though theywere not statistically different. The estimated velocitiesfrom the 2H12-MTBE tracer tests in both plots were lowerin general than the bromide results owing largely to highervariability in corresponding breakthrough times.

The tracer test results compared favorably with theestimated seepage velocities summarized in Table 1 froma number of studies at the NEX site. The data further sug-gest that the short-term disruption of flowpaths due to theaddition of gases could be minimized by collection ofsamples once water levels in the plots stabilized.

The results of the ground water geochemical analysesare presented in Table 3 for the control plot and Table 4for the test plot. The information comprises the averageparameter concentrations at the deep screens over the 300-dstudy period and is presented as upgradient and down-gradient locations with respect to each plot. A statisticalanalysis indicated that geochemical data were less variablethan intrinsic MTBE, 2H12-MTBE, and bromide. The rela-tive dispersion or coefficient of dispersion or variation(CV) of MTBE values at the deep screens of the controlplot over the study period was 79%, while that of the testplot was 111%. The corresponding CV for 2H12-MTBEwas 101% in the control plot and 144% in the test plot,while that of bromide was 137% in the control plot and143% in the test plot. In contrast, at the same well locationsover the same period, the CV for alkalinity and sulfate inthe control plot were 6% and 3%, respectively, while theCV for alkalinity and sulfate in the test plot were 8% and10%, respectively.

It is probable that the apparent stability in ambient andboth control and test site system geochemistry was related

to concentration magnitude, reactivity, and indeed the ori-gin of the chemical species. While bromide, MTBE, and2H12-MTBE were introduced into the ground water, themajor geochemical species, by and large, derive their val-ues from ground water interactions with the aquifer matrix.The variability in bromide, MTBE, and 2H12-MTBEconcentrations was influenced primarily by differentialinfluences on plume development, ground water flow, andgradient while that of geochemical parameters wereaffected by the homogeneity of the aquifer matrix. In com-parison to the variability of bromide and MTBE, there werenegligible geochemical differences (Table 3) between thetest and control plots over the study period. In light ofthis study’s observations, therefore, it is expected that thetracers’ behavior will not be completely analogous to thatof a contaminant plume. In that, the dissolved tracer solu-tions were metered into 38 injections wells and their be-havior was monitored within the two small experimentalplots as compared to Port Hueneme’s dissolved MTBEplume, which was originated from an NAPL source andexperienced years of transport history. Nevertheless, tracerdata presented from small-scale fieldwork may not beapplicable to the entire contaminated plume without con-siderations given to involve large-scale fieldwork, long-term monitoring, and perhaps the use of models.

ConclusionsVarious preliminary aquifer tests at different locations

throughout the plume demonstrated the presence of a com-plex velocity field in terms of vertical stratification andpreferential flowpaths. This information led to the installa-tion of the experimental plots within the middle zone of anexisting dissolved MTBE plume. Tracer experiments (bro-mide, bromide/2H12-MTBE) resulted in the collection of2161 bromide and 1442 MTBE2H12 samples, which pro-vided a detailed understanding of the aquifer heterogeneityand preferential flowpaths within the plots. Since hydro-geology controlled the distribution of tracers, the effects ofvertical stratification were made apparent by the tracers’detection pattern. With only a few exceptions, bromide was

Figure 7. Bromide concentration during second bromidetracer test in control plot at deep screens in first downgradienttransect.


detected in the lower part of the aquifer at the deep screensthat were installed at 18 feet (5.5 m) bgs. This allowedreduced sampling for MTBE and tracers at the shallow andmiddle screens during the ensuing evaluation phase of theproject period, which was in excess of 300 d. Because thehydraulic conductivity at the site varied by more than 2 or-ders of magnitude, it would be expected that the arrivaltime of the tracers within the lower conductivity regions ofthe aquifer would be longer than the duration of this study.At the time of the 15th sampling event, however, 2H12-MTBE had arrived at >80% of the shallow, middle, anddeep screens.

Since data obtained from the experimental plots indi-cated a zone of increased hydraulic conductivity existedbetween ~16 and 19 feet bgs (4.9 and 5.8 m) and that thiszone had significantly influenced MTBE transport duringthe treatment demonstration, this zone, defined by deepscreens, was emphasized with respect to the remediationevaluation with the aid of bromide and 2H12-MTBE

tracers. Data suggested that degradation was not a signifi-cant removal process for 2H12-MTBE as its average con-centration, at the bottom screens in both the control andtest plots, was approaching the injection concentration of1 mg/L at the termination of the evaluation process. Inthis context, 2H12-MTBE behavior resembled that of bro-mide in terms of higher and lower hydraulic conductivityregions.

It was anticipated that the behavior of the injected dis-solved tracer solution (2H12-MTBE and bromide) into theaquifer should behave differently from that of the existingdissolved MTBE plume that emanated from an NAPLsource over a decade ago, thereby undergoing years of dif-fusion. However, it was probable that by inducing a curtainof tracers through a series of injection wells that were capa-ble of a continuous release of bromide and 2H12-MTBEinto the aquifer under natural conditions, we could under-stand the preferential flow fields as well as the degradationpotential of 2H12-MTBE. The findings of this study showed

Table 3Water Quality Measurements in Control Plot

Upgradient Control Plot Downgradient Control Plot

Target Analytes Mean Standard Deviation High Low Mean Standard Deviation High Low

Alkalinity 533 22 558 519 502 31 544 452Ammonia-Nitrogen 0.55 0.24 0.79 <0.3 0.56 0.21 0.76 <0.3Nitrate 1 nitrite 0.13 0.14 0.29 <0.05 0.10 0.12 0.44 <0.05Phosphorus, total 0.1 0.12 0.24 <0.02 0.2 0 0.2 0.2TOC 3.3 0.1 3.4 3.2 3.5 0.4 4.1 2.8TOC dissolved 3.0 0.1 3.1 3 3.3 0.3 3.8 2.7Nitrate 0.3 0.23 0.48 <0.05 0.22 0.28 0.69 <0.05Nitrite 0.32 0.04 0.37 <0.05 0.26 0.22 0.59 <0.05Orthophosphate 0.02 0 0.02 <0.02 0.02 0 0.02 <0.02Sulfate 1132 101 1241 1041 1167 37 1217 1134

Note: Data represent average of deep screens over 15 sampling events (mg/L). The numbers above detection limit and below practical quantitation limit are reportedas less than (<).

Table 4Water Quality Measurements in Test Plot

Upgradient Test Plot Downgradient Test Plot

Target Analytes Mean Standard Deviation High Low Mean Standard Deviation High Low

Alkalinity 485 22 509 451 437 34 478 475Ammonia-nitrogen 0.89 0.07 0.97 0.82 0.57 0.28 1.08 <0.3Nitrate 1 nitrite 0.06 0.05 0.14 <0.05 0.16 0.15 0.51 <0.05Phosphorus, total 0.08 0.06 0.13 <0.02 0.02 0.0 0.02 <0.02TOC 3.3 0.2 3.6 3.1 3.6 0.3 4.4 2.8TOC dissolved 3.0 0.3 3.6 2.8 3.3 0.3 4.2 2.7Nitrate 0.05 0.0 0.05 <0.05 0.3 0.3 1.17 <0.05Nitrite 0.56 0.3 0.81 <0.05 0.1 0.1 0.59 <0.05Orthophosphate 0.02 0.0 0.02 <0.02 0.02 0.0 0.02 <0.02Sulfate 1179 118 1300 1016 1189 119 1217 1103

Note: Data represent average of deep screens over 15 sampling events (mg/L). The numbers above detection limit and below practical quantitation limit are reportedas less than (<).


that a detailed understanding of the complexity of the flowfield within the two small experimental plots was not pos-sible by observing the intrinsic MTBE data alone. Thisinformation is crucial, however, in assessing the long-termperformance of the remedial technology.

AcknowledgmentsWe are indebted to James Osgood, Dorothy Cannon,

Ernie Lory, and numerous individuals at the NBVC hostsite for considerable in-kind support; to Peter Raftery andothers at the California Water Quality Control Board fortheir technical and administrative efforts in granting theproject permits and their continued interest and support; toRandall Ross, Steve Acree, and Barth Faulkner at the R.S.Kerr Environmental Research Center of the U.S. EPA fortheir hydrological fieldwork; to Ann Vega and Steve Van-degrift at the NRMRL of the U.S. EPA for their roles inthe quality assurance and quality control; to Shaw, aGWERD contractor for system installation, sampling andlaboratory analytical services; and to SAIC, a NRMRLcontractor for assistance in the development of the projectQAPP and for contracting ALSI for laboratory analysis. Weare also grateful to Paul Johnson and two anonymous re-viewers for their careful reviews and constructive comments.

Editor’s Note: The use of brand names in peer-reviewedpapers is for identification purposes only and does not con-stitute endorsement by the authors, their employers, or theNational Ground Water Association.

ReferencesAguirre, C.G., and K. Haghighi. 2003. Stochastic modeling of

transient contaminant transport. Journal of Hydrology 276, no.1–4: 224–239.

Amerson, I., and R.L. Johnson. 2003. Natural gradient tracer testto evaluate natural attenuation of MTBE under anaerobic con-ditions. Ground Water Monitoring and Remediation 23, no. 1:54–61.

Azadpour-Keeley, A. 2002. Envirogen propane biostimulationtechnology for in-situ treatment of MTBE-contaminatedground water. EPA/600/R-02/092. Washington, D.C.: ORD.

Aziz, J.J., M. Ling, H.S. Rifai, C.J. Newell, and J.R. Gonzalez.2003a. MAROS: A decision support system for optimizingmonitoring plans. Ground Water 41, no. 3: 355–367.

Aziz, J.J., M. Ling, H.S. Rifai, C.J. Newell, and J.R. Gonzalez.2003b. Monitoring and Remediation Optimization System(MAROS) 2.0 Software Users Guide. Developed for Air ForceCenter for Environmental Excellence (AFCEE) Brooks AFB,San Antonio, Texas, 85 pp.

Barcelona, M.J., H.A. Wehrmann, M.R. Schock, M.E. Sievers,and J.R. Karny. 1989. Sampling Frequency for Ground-WaterQuality Monitoring. Las Vegas, Nevada: Environmental Moni-toring Systems Lab.

Bartlett, S.A., G.R. Robbins, J.D. Mandrick, M.J. Barcelona,W. McCall, and M. Kram. 2004. Comparison of hydraulic con-ductivity determinations in directpush and conventional wells.Naval Facilities Engineering Service Center (internal report).U.S. Navy, Port Huenene, CA.

Chang, C.M., M.W. Kemblowski, and G.E. Urroz. 1999. Transientstochastic analysis of biodegradable contaminant transport:First-order decay. Transport in Porous Media 35, no. 1: 1–14.

Dou, C., W. Woldt, M. Dahab, and I. Bogardi. 1997. Transientground-water flow simulation using a fuzzy set approach.Ground Water 35, no. 2: 205–215.

Freeze, R.A., J. Massmann, L. Smith, T. Sperling, and B. James.1990. Hydrogeological decision analysis: 1. A framework.Ground Water 28, no. 5: 738–766.

Gray, J.R., G. Lacrampe-Couloume, D. Gandhi, K.M. Scow, R.D.Wilson, and B.S. Lollar. 2002. Carbon and hydrogen isotopicfractionation during biodegradation of methyl tert-butyl ether.Environmental Science & Technology 36, no. 9: 1931–1938.

Harris, J., J.C. Loftis, and R.H. Montgomery. 1987. Statisticalmethods for characterizing ground-water quality. GroundWater 25, no. 2: 185–193.

Harter, T., H. Davis, M.C. Mathews, and R.D. Meyer. 2002. Shal-low groundwater quality on dairy farms with irrigated foragecrops. Journal of Contaminant Hydrology 55, no. 3–4: 287–315.

Hoeksema, R.J., and P.K. Kitanidis. 1985. Analysis of the spatialstructure of properties of selected aquifers. Water ResourcesResearch 21, no. 4: 563–572.

Johnson, R.L., P.C. Johnson, I.L. Amerson, T.L. Johnson, C.L.Bruce, A. Leeson, and C.M. Vogel. 2001a. Diagnostic tools forintegrated in situ air sparging pilot tests. Bioremediation Jour-nal 5, no. 4: 283–298.

Johnson, R.L., P.C. Johnson, T.L. Johnson, N.R. Thompson, and A.Leeson. 2001b. Diagnosis of in situ air sparging performanceusing transient groundwater pressure changes during startup andshutdown. Bioremediation Journal 5, no. 4: 299–320.

Keith, S.J., L.G. Wilson, H.R. Fitch, and D.M. Esposito. 1983.Sources of spatial-temporal variability in ground water qualitydata and methods of control. Ground Water Monitoring andReview 3, no. 2: 21–32.

Kram, M., D. Lorenzana, J. Michaelsen, and E. Lory. 2000. Per-formance comparison: Direct-push wells versus drilled wells.NFESC technical report TR-2120-ENV. Washington, D.C.:U.S. Navy.

Leblanc, D.R., S. Garabedian, K. Hess, L.W. Gelhar, R. Quadri,K.G. Stollenwerk, and W.W. Wood. 1991. Large-scale naturalgradient test in sand and gravel, Cape Cod, Massachusetts 1.Experimental design and observed tracer moment. WaterResources Research 27, no. 5: 895–910.

Montgomery, R.H., J.C. Loftis, and J. Harris. 1987. Statisticalcharacteristics of ground water quality variables. GroundWater 25, no. 2: 176–184.

National Research Council. 1990. Ground Water Models: Scien-tific and Regulatory Applications. Washington, D.C.: NationalAcademy Press.

Nightingale, H.I., and W.C. Bianchi. 1979. Influence of wellwater quality variability on sampling decisions and monitor-ing.Water Resources Bulletin 15, no. 5: 1394–1407.

Pettyjohn, W.A. 1976. Monitoring cyclic fluctuations in ground-water quality. Ground Water 14, no. 6: 474–480.

Puls, R.W., and M.J. Barcelona. 1996. Low-flow (minimal draw-down) sampling procedures. EPA/540/S-95/504. Washington,D.C.: ORD.

Reilly, T.E., and J. Gibs. 1993. Effects of physical and chemicalheterogeneity on water-quality samples obtained from wells.Ground Water 31, no. 5: 805–813.

Ronen, D., M. Magaritz, H. Gvirtzman, and W. Garner. 1987.Microscale chemical heterogeneity in groundwater. Hydrology92, no. 8: 173–178.

Salanitro, J.P., P.C. Johnson, G.E. Spinnler, P.M. Maner, H.L.Wisniewski, and C. Bruce. 2000. Field-scale demonstration ofenhanced MTBE bioremediation through aquifer bioaugmentationand oxygenation. Environmental Science & Technology 34,no. 19: 4152–4162.


Squillace, P.J., J.S. Zogorski, W.G. Wilber, and C.V. Price. 1996.Preliminary assessment of the occurrence and possible sourcesof MTBE in groundwater in the United States, 1993-1994.Environmental Science & Technology 30, no. 5: 1721–1730.

Sudicky, E.A. 1986. A natural gradient experiment on solutetransport in a sand aquifer: Spatial variability of hydraulicconductivity and its role in the dispersion process. WaterResources Research 22, no. 13: 2069–2082.

Thierrin, J., G.B. Davis, C. Barber, B.M. Patterson, F. Pribac,T.R. Power, and M. Lambert. 1993. Natural degradation ratesof BTEX compounds and naphthalene in a sulphate reducinggroundwater environment. Hydrological Sciences 38, no. 4:309–322.

Thierrin, J., G.B. Davis, C. Barber, and T.R. Power. 1992.Groundwater tracer test within a BTEX contaminated zone inorder to determine aquifer parameters and natural degradationrates of organic compounds. CSIRO, Division of Water Re-sources, Perth. Report No. 92/5.

Thierrin, J., G.B. Davis, and C. Barber. 1995. A ground watertracer test with deuterated compounds for monitoring in situbiodegradation and retardation of aromatic hydrocarbons.Ground Water 33, no. 3: 469–475.

Vomvoris, E.G., and L.W. Gelhar. 1990. Stochastic analysis of theconcentration variability in a three-dimensional heterogeneousaquifer.Water Resources Research 26, no. 10: 2591–2602.

Wheater, H.S., J.A. Thompkins, M. Van Leeuwen, and A.P.Butler. 2000. Uncertainty in groundwater flow and transportmodeling—A stochastic analysis of well protection zones.Hydrological Processes 14, no. 11: 2019–2029.

Wilson, L.C., and J.V. Rouse. 1983. Variations in water qualityduring initial pumping of monitoring wells. Ground WaterMonitoring & Remediation 3, no. 1: 103–109.

Wilson, R.D., D.M. MacKay, and K.M. Scow. 2002. In situMTBE biodegradation supported by diffusive oxygen release.Environmental Science & Technology 36, no. 2: 190–199.

Biographical SketchesAnn Azadpour-Keeley, corresponding author, is a Research

Microbiologist with the U.S. EPA Office of Research and Develop-ment. She is assigned to the Subsurface Remediation Branch withinthe Ground Water and Ecosystems Restoration Division of theNational Risk Management Research Laboratory. She acquiredseveral years of experience in academia and with an environmen-tal consulting company prior to joining U.S. EPA. Dr. Keeley hasprovided technical assistance to U.S. EPA’s Regional Offices andHeadquarters, states, and industry at scores of hazardous wastesites regarding monitored natural attenuation. Her research inter-ests relate to the application of innovative bioremediation tech-nologies in the restoration of ground water quality. She may bereached at the R.S. Kerr Environmental Research Center, 919Kerr Research Drive, Ada, OK 74820; [email protected].

Michael J. Barcelona is Chairperson of the Department ofChemistry at Western Michigan University. He is a leading author-ity on ground water monitoring practices and geochemistry.Dr. Barcelona’s research on ground water monitoring has spannedover 25 years and had a profound effect on the development ofmodern ground water sampling practices. He has authored morethan 100 publications in the areas of environmental geochemistry,hydrogeology, and advanced analytical methods applied to con-tamination situations. He may be reached at Department ofChemistry, Western Michigan University, 3442 Wood Hall Kalama-zoo, MI 49008; [email protected].


design of an mtbe remediation technology evaluation

Documents