field demonstration of on-site analytical methods for tnt and rdx in

FIELD DEMONSTRATION OF ON-SITE ANALYTICAL

METHODS FOR TNT AND RDX IN GROUND WATERH. Craig1, G. Ferguson2, A. Markos2, A. Kusterbeck3, L. Shriver-Lake3, T. Jenkins4, and P.Thorne4, 1U.S. Environmental Protection Agency Region 10, Oregon Operations Office, 811 SW6th Avenue, Portland, OR, 97204, Phone: 503-326-3689, 2Black & Veatch Special ProjectsCorporation, 1201 Pacific Avenue, Suite 1100, Tacoma, WA, 98402-4301, Phone: 206-383-1436, 3Naval Research Laboratory, Center for Bio/Molecular Science and Engineering, Code6910, Washington, DC, 20375-5348, Phone: 202-404-6042, and 4U.S. Army Corps of Engineers,Cold Regions Research and Engineering Laboratory, Hanover, NH, 03755-1290, Phone: 603-646-4385

ABSTRACT A field demonstration was conducted to assess the performance of eightcommercially-available and emerging colorimetric, immunoassay, and biosensor on-site analyticalmethods for explosives 2,4,6-trinitrotoluene (TNT) and hexahydro-1,3,5-trinitro-1,3,5-triazine(RDX) in ground water and leachate at the Umatilla Army Depot Activity, Hermiston, Oregon andU.S. Naval Submarine Base, Bangor, Washington, Superfund sites. Ground water samples wereanalyzed by each of the on-site methods and results compared to laboratory analysis using highperformance liquid chromatography (HPLC) with EPA SW-846 Method 8330. The commercialmethods evaluated include the EnSys, Inc., TNT and RDX colorimetric test kits (EPA SW-846Methods 8515 and 8510) with a solid phase extraction (SPE) step, the DTECH/EM Science TNTand RDX immunoassay test kits (EPA SW-846 Methods 4050 and 4051), and the Ohmicron TNTimmunoassay test kit. The emerging methods tested include the antibody-based Naval ResearchLaboratory (NRL) Continuous Flow Immunosensor (CFI) for TNT and RDX, and the Fiber OpticBiosensor (FOB) for TNT. Accuracy of the on-site methods were evaluated using linear regressionanalysis and relative percent difference (RPD) comparison criteria. Over the range of conditionstested, the colorimetric methods for TNT and RDX showed the highest accuracy of thecommercially-available methods, and the NRL CFI showed the highest accuracy of the emergingmethods for TNT and RDX. The colorimetric method was selected for routine ground watermonitoring at the Umatilla site, and further field testing on the NRL CFI and FOB biosensors willcontinue at both Superfund sites. The primary use for these analytical methods would be for influentand effluent monitoring for granular activated carbon (GAC) ground water and leachate treatmentsystems, which are projected to operate for a period of 10 to 30 years.

K EYWORDS: explosives, TNT, RDX, field analytical methods, ground water

INTRODUCTION

This paper presents a comparison of eightcommercially-available and emerging on-siteanalytical methods used to determine 2,4,6-trinitrotoluene (TNT) and hexahydro-1,3,5-trinitro-1,3,5-triazine (RDX) concentrationsin ground water and leachate. Field analysiswas conducted on water collected from theExplosives Washout Lagoon area atUmatilla Army Depot Activity (Umatilla),

Hermiston, Oregon, and from Site F groundwater treatment system and Site A leachatetreatment system at the U.S. NavalSubmarine Base (SUBASE), Bangor,Washington. A description of on-siteanalytical methods employed and acomparison of results and methodperformance considerations are presented.

Field and laboratory analyses from Umatillaand SUBASE Bangor data were used to

compare the accuracy, feasibility, strengths,and weaknesses of five on-site analyticalmethods for TNT and three methods forRDX in ground water, and to determine theeffect of ground water parameters (nitrates,humics, turbidity, etc.) on these methods.The field demonstration was conducted toprovide performance data and a preliminaryeconomic evaluation for future long-termexplosives ground water remediation effortsthroughout the U.S. In addition to providingfield analytical data for general use, resultsfrom this project will be used to guideappropriate selection of methods forexplosives ground water remediationactivities at Umatilla and SUBASE Bangor.Use of field analytical methods to determineexplosive concentrations in water willsubstantially reduce the time intervalbetween sample collection and theavailability of analytical results and reducelaboratory analytical costs.

The study design, performance parameters,and economic evaluation were developed byU.S. Environmental Protection Agency(EPA) Region 10. Field analyses wereconducted by personnel from Black &Veatch Special Projects Corp. (BVSPC)under contract to U.S. EPA Region 10, theNaval Research Laboratory (NRL) Centerfor Bio/Molecular Science and Engineering(CBMSE), and the U.S. Army Corps ofEngineers Cold Regions Research andEngineering Laboratory (CRREL). CRRELpersonnel also conducted laboratoryanalyses of ground water and leachatesamples utilizing EPA SW-846 Method8330. BVSPC conducted field analysesutilizing several commercially-available on-site analytical methods. CRREL developedthe preconcentration step utilized in thecolorimetric field analytical proceduresperformed by BVSPC and was presentduring field activities at Umatilla andSUBASE Bangor to provide technical

guidance. NRL CBMSE personneldeveloped the emerging biosensors methodsfor explosives analysis and were present atUmatilla and SUBASE Bangor to evaluatethe field application of these methods [1].

BACKGROUND

Sites with explosives-contaminated groundwater or surface water exist in a number ofenvironmental settings. The objective of thisfield demonstration was to evaluate the on-site analytical methods at differingenvironmental conditions at Umatilla andSUBASE Bangor.

Table 1 shows a number of relevant wastedisposal, chemical, geological, andhydrogeological conditions of these sites.

Umatilla Army Depot Activity

In 1941 Umatilla Army Depot Activity wasestablished as an Army ordnance depot forthe storage and handling of munitions. Fromthe 1950s until 1965 Umatilla operated anonsite explosive washout plant thatprocessed munitions to remove and recoverexplosives. Flushing and draining theexplosive washout system was a standardprocedure during plant operation. Wastewater from this procedure was dischargedthrough an open metal trough into twounlined infiltration basins known as theExplosive Washout Lagoons.

The Explosive Washout Lagoons werecharacterized as a potentially hazardous sitein the initial installation assessment. In 1981an approximate 45-acre plume of RDX wasidentified in the shallow ground wateraquifer, apparently resulting from dischargesto the lagoons. Subsequent investigationsconfirmed the presence of explosives in thesoil and ground water. In the summer of1994, TNT and RDX contaminated soil wasexcavated from the lagoon area and is

currently undergoing bioremediationtreatment via composting. Ground waterremediation will be conducted through anextraction and treatment system using agranular activated carbon (GAC) treatmentunit. Treated ground water will then be re-infiltrated back into the shallow aquifer. Aportion of the re-infiltration will occurthrough subsurface soils at the originalExplosives Washout Lagoons to flushresidual explosives from the vadose zonebetween the bottom of the excavated lagoonand the top of the shallow aquifer.Additional infiltration galleries on theperimeter of the contamination plume willalso be used to enhance remediation duringthe expected 27-year operation of thetreatment system.

TNT is relatively immobile when comparedto RDX. TNT concentrations in groundwater were highest (3,000 µg/l) near thelagoons and rapidly decreased a shortdistance from the lagoons. Therefore, basedon relative mobility and monitoring wellplacement, there was no opportunity tocollect samples which contained mid-rangeTNT concentrations (50 to 500 µg/l).

Consequently, one-half order of magnitudeserial dilutions were prepared for samplesfrom two monitoring wells which containedhigh TNT concentrations. Conversely, RDXis more mobile than TNT in ground water.Near the lagoons TNT and RDXconcentrations are comparable. RDXconcentrations further from the lagoons arehigher than TNT concentrations.Consequently, the RDX contaminant plumeis relatively large in areal extent.

Umatilla ground water also containsrelatively high levels (8,000 to 40,000 µg/l)of nitrates. One of the objectives of the fieldanalytical study was to determine the effect,if any, of nitrate concentrations on themethods tested. Serial dilutions wereprepared with water collected from aUmatilla background well to avoid dilutingthe concentration of nitrates in the groundwater samples. Analyses indicated thatnitrate levels in the background well werecomparable to nitrate levels in ground watersampling wells. By using background wellwater for dilutions, the validity of theanalysis to determine the effect of nitrateson the field screening results was not

TABLE 1. SITE CONDITIONS FOR ON-SITE ANALYTICAL METHODS.

Parameter Umatilla BangorWaste disposal Lagoon Lagoon, OB/ODa

Primary analytes TNT, RDX TNT, RDXInterfere/cross react TNBb, HMXc TNBb, HMXc, AmDNTsd

Nitrates High Low - ModTurbidity Low Mod - HighHumics Low HighSurface geology Sandy LoamHydro-geology Fluvial/alluvial Glacial tillClimate Semi-arid, extreme temps Wet, moderate temps

aopen burn/open detonation (OB/OD)b1,3,5-trinitrobenzene (TNB)coctahydro-1,3,5,7-tetranitro-1,3,5,7-tetrazocine (HMX)d2-amino-4,6-dinitrotoluene and 4-amino-2,6-dinitrotoluene (AmDNTs)

compromised.

Naval Submarine Base Bangor

SUBASE Bangor is currently an activemilitary installation serving the Pacific NavalSubmarine Fleet and associated Navy ships.Site F and Site A are inactive former militarymunitions sites currently undergoingremediation. Site F is a waste water lagoonand overflow area formerly used for thedisposal of waste water generated duringordnance demilitarization at a neighboringmunitions processing facility. The site wasactively used between 1957 and 1973,during which time waste water wasdischarged from the segregation facilitydirectly into the disposal area through adrain line. After 1973, waste water wascollected in 55-gallon drums and deliveredto the SUBASE Bangor liquid wasteincinerator plant. TNT and RDX have beendetected at concentrations above risk-basedground water cleanup levels in samplescollected at the site. Six extraction wellswere installed downgradient of the lagoonsto pump contaminated ground water into atreatment system. Ground water is treatedby filtration through a GAC unit. Treatedwater is reinjected into the aquifer.

During past operations, explosive-contaminated sediments that accumulated inthe waste water lagoon at Site F wereremoved and transported to Site A forburning and disposal. Existing soil at Site Acontains levels of TNT and RDX above risk-based cleanup criteria. The contaminatedsoil has been contained in a lined basin andis currently undergoing water flushing toremove TNT and RDX. Leachate from thecontaminated soil flushing is collected andtreated in a GAC unit. At the SUBASEBangor site, individual ground water sampleswere collected directly from all six of theextraction wells at Site F. Ground water

samples were also collected from thecombined flow of all six extraction wells. Allcombined flow samples were collected fromsampling ports before and after initialparticulate filters, upstream of the GACtreatment unit. Two leachate samples werecollected at Site A. The first sample wascollected from the primary sump area andthe second sample from within the treatmentplant, after the initial particulate filter,upstream of the GAC treatment unit.One-half order of magnitude serial dilutionswere prepared for both leachate samples toobtain low, high, and mid-range contaminantconcentrations. The leachate samplecollected from the sump area containedlower TNT and RDX concentrations thananticipated which may have been due torecent rainfall. Serial dilutions utilizeduntreated leachate from Site A.

M ETHODS

Eight separate on-site analytical methodswere used to analyze original ground waterand leachate samples, serial dilutions, andstandards for TNT, RDX, 1,3,5-trinitrobenzene (TNB), and octahydro-1,3,5,7-tetranitro-1,3,5,7-tetrazocine(HMX). Three types of on-site analyticalmethods were evaluated: 1) colorimetric, 2)kit-type immunoassay, and 3) biosensors.Published or experimentally-reporteddetection limits for the methods are shown inTable 2. The target detection limits for on-site analytical methods are the EPADrinking Water Lifetime Health Advisory(HA) levels of 2 µg/l for TNT and RDX [2,3]. Laboratory analysis for comparison wasconducted using reverse phase highperformance liquid chromatography(HPLC).

Colorimetric methods measure coloredreaction products formed whennitroaromatic and nitramine compounds are

reacted with alkali or acidic solutions. Theoperator can visually determine the presenceof various compounds by the colordevelopment of the extract. The absorbanceat a specified wavelength is measured andcorrelated to the compound concentration.The CRREL-EnSys methods arecolorimetric methods.

Kit-type immunoassay and biosensormethods utilize the ability of antibodies toselectively bind to a primary target analytepresent in low concentrations in a complexmatrix. For immunoassay methods, thesample, an enzyme conjugate of TNT orRDX, and particles with antibodies specificto TNT or RDX attached are mixed. Theenzyme conjugate, and any TNT or RDX inthe sample, compete for antibody bindingsites on the particles. The presence of theprimary target analyte (TNT or RDX) isdetected by adding an enzyme substrate anda chromogen. The enzyme conjugate boundto the target compound antibody catalyzesthe conversion of the enzymesubstrate/chromogen mixture to a coloredproduct. Since the enzyme conjugate was incompetition with the primary target analytein the sample for the antibody sites, the colordeveloped is inversely proportional to theconcentration of the target compound in thesample. DTECH and Ohmicron are kit-typeimmunoassay methods.

Biosensor methods also utilize the ability of

antibodies to selectively bind to a primarytarget analyte present in a water sample.Biosensors consist of a biological recognitionelement (i.e., labeled antibodies) in contactwith a physical transducer, such as afluorimeter or a photodiode [4, 5]. The NRLContinuous Flow Immunosensor (CFI) andFiber Optic Biosensor (FOB) are biosensormethods.

Interference/cross reactivity

One of major differences betweencolorimetric and immunoassay-basedmethods is the response of the methods tosecondary target analytes. For colorimetricmethods, interference is defined as thepositive response of the method tosecondary target analytes chemically similarto the primary target analyte. Colorimetricmethods have 100% interference forcompounds within the same compound class(i.e., nitroaromatics or nitramines) andremain constant throughout theconcentration range of the method. For thecolorimetric TNT method, the primary targetanalyte is TNT and the secondary targetanalytes are other nitroaromatics such asTNB, 1,3-dintrobenzene (DNB),dinitrotoluenes (DNTs), methyl-2,4,6-trinitrophenylnitramine (tetryl), etc. For theRDX colorimetric method, the primarytarget analyte is RDX and the secondarytarget analytes are other nitramines such asHMX and nitrate esters such aspentaerythritol tetranitrate (PETN). Forimmunoassay-based methods, crossreactivity is defined as the positive responseof the method to secondary target analyteschemically similar to the primary targetanalyte. Cross-reactivity occurs when theantibody recognizes compounds that aresimilar in structure to the primary targetanalyte. Cross reactivity for kit-typeimmunoassay and biosensor methods is not100% for compounds within the same

TABLE 2. METHOD DETECTION LIMITS.

Method Analyte Conc. (µg/l)TNT RDX

CRREL/EnSys 0.9 3.8DTECH 5 5Ohmicron 0.07 --NRL CFI 20 20NRL FOB 20 --

compound class (i.e., nitroaromatics ornitramines) and is not constant throughoutthe concentration range of the methods. Inaddition, the cross reactivities for allimmunoassay-based methods are not thesame and are based on the antibodies used todevelop the specific method.

TNB and HMX were the most commonlyoccurring co-contaminants with TNT andRDX at the test sites. TNB is reported tocause 100% additive, or positiveinterference, for results in the TNTcolorimetric on-site analytical method [6].HMX is 100% additive to the colorimetricRDX on-site analytical method [7].Immunoassay-based methods exhibit TNBcross-reactivity with TNT that varies withthe concentration of TNT and TNB in thesample. DTECH reported a TNBcross-reactivity of 23% for the midpoint ofthe TNT test range. Ohmicron reported aTNB cross reactivity of 65% for themidpoint of the TNT detection range. TNB isa known interferant in both the NRLbiosensor methods; however, the percentcross reactivity has not been quantified.DNB, 2,4-dinitrotoluene (2,4-DNT), 2-amino-4,6-dinitrotoluene (2AmDNT), 4-amino-2,6-dinitrotoluene (4AmDNT), tetryl,and 2,4-dinitroaniline are also known toaffect immunoassay TNT tests; however,both the concentrations and cross-reactivityof these compounds are relatively low, andthey were not included in thecross-reactivity evaluation. DTECH reportsa HMX cross reactivity of 3% at themidpoint of the RDX test range. Highconcentrations of HMX may cross-reactwith the NRL CFI RDX method; however,this cross reactivity has not yet beenquantified.

CRREL-EnSys methods

EnSys test kits are commercially-availablecolorimetric methods developed for thedetection of TNT and RDX in soil [6, 7].Additional sample preparation is required toutilize the EnSys tests for ground watersamples. A TNT and RDX preconcentrationstep, which uses a solid phase extraction(SPE), has been developed by CRREL forthis purpose [8]. The SPE preconcentrationstep is required for the colorimetric analysisof TNT and RDX in ground water. The useof the CRREL preconcentration step inconnection with the EnSys field analyticalmethod is referred to as the CRREL-EnSysMethod.

CRREL preconcentration step

The CRREL SPE preconcentration methodpasses a 2 liter volume of sample throughtwo membranes. Jenkins, et al., found thatany TNT in the sample will be completelyretained on the top membrane [8]. After thesample passes through the membrane stack,the amount of RDX retained on the bottommembrane will be approximately equal toRDX retained on the top membrane. Theacetone extract from the top membrane willcontain RDX and TNT. Based on anevaluation performed by Jenkins, thepresence of RDX in the sample extract doesnot interfere with TNT colorimetric tests [6].The extract from the bottom membrane canbe used for accurate RDX testing because allTNT has been completely retained in the topmembrane.

In consultation with CRREL, modificationswere made to this procedure during analysisat Umatilla and SUBASE Bangor. Onedeviation at both sites involved theelimination of the glass filter from thestandard preconcentration apparatus. Theglass filter was unnecessary for mostsamples due to the overall low turbidity of

the ground water samples collected fromboth sites. If a turbid sample wasencountered, then a glass filter was placedon the preconcentration apparatus and thesample was filtered and collected in thevacuum flask. The glass filter was thenremoved and discarded. The filtered samplewas collected in a clean glass beaker and theapparatus was reassembled with the standardmembrane stack in place. The filteredsample was then processed according to theSPE preconcentration procedure. The glassfilter decreased the preconcentration timefor turbid samples by reducing membraneparticulates clogging.

An additional change to the publishedprocedures involved removing the buildup ofnitrates from the membrane stack. Highconcentrations of nitrates are present in theground water at Umatilla and nitratesinterfere with EnSys RDX analysis. Nitrateinterference can be minimized through theuse of an ion exchange resin. The resinremoves the nitrates from the extract prior toEnSys analysis. A second nitrate removaltechnique involves rinsing the membranestack with deionized water after the samplehas been filtered, but prior to extractingRDX with acetone. This technique removesresidual nitrate containing water from themembranes. BVSPC used this second nitrateremoval method. Membranes were rinsedwith 10 ml of deionized water prior toelution of the RDX with acetone.

EnSys TNT test method (EPA SW-846 Method 8515)

Extract obtained from the top filter in theCRREL preconcentration step was analyzedusing the EnSys TNT test method (EPA SW-846 Method 8515). EnSys TNT method isdesigned for soil samples; therefore,concentration calculations and detectionlimits were based on Jenkins, et al., rather

than on the EnSys TNT soil method [8]. Theinitial absorbance is multiplied by two andsubtracted from the final absorbance toeliminate background absorbance. EnSysrecommends subtracting four times theinitial absorbance to eliminate backgroundinterference. However, a factor of two wasexperimentally determined by Jenkins to besufficient [6].

In some field analytical applications, it maybe important to know more than just theconcentration of TNT in a sample. The colordevelopment of the extract in colorimetrictests can give the operator an indication ofwhat type of compounds are present in themedia being tested. In general, TNT andTNB turn the extract red, tetryl turns theextract orange, DNB turns the extractpurple, 2,4-DNT turns the extract blue, and2,6-DNT turns the extract pink after additionof the TNT indicator solution.

EnSys RDX test method (EPA SW-846 Method 8510)

Extract obtained from the bottom filter inthe CRREL preconcentration step wasanalyzed using the EnSys RDX test method(EPA SW-846 Method 8510). The EnSysRDX method is designed for soil samples;therefore, concentration calculations anddetection limits were based on Jenkins, etal., rather than on the EnSys RDX soilmethod [8].

DTECH TNT and RDX methods(EPA SW-846 Methods 4050 and4051)

DTECH TNT and RDX immunoassaymethods can be used for both soil and water.The water detection range without dilutionsis from 5 to 45 µg/l for TNT and RDX [9,10]. DTECH will provide dilutioninstructions for concentrations above this

range. DTECH results can be obtained intwo ways. When a color chart is used, thereference color from the test is matched to areference color on the chart. When thereference colors match, the sample color isthen compared to the color chart todetermine the concentration range.However, the colors printed on the charttend to have a gray tint and do noteffectively match the color of the test,creating a high degree of subjectivity.

The alternative method for obtainingDTECH results involves using theDTECHTOR. The DTECHTOR is areflective photometer which can be set toread the absorbance of the reference andsample colors. When the referenceabsorbance reaches a range from 220 to 250for TNT and a range from 320 to 350 forRDX, the test absorbance value is read. TheDTECHTOR will display the percentdifference between the reference and samplecolors, which corresponds to a specificconcentration range provided by DTECH.Utilizing the DTECHTOR eliminates thesubjectivity of the visual test. BVSPCutilized the DTECHTOR during fieldanalysis. In an effort to obtain an actualconcentration for comparison purposes,BVSPC assumed a linear correlationbetween the DTECHTOR measurement andthe associated concentration range. In thisway, the DTECHTOR value was used toderive an estimated concentration. Theestimated concentration was used todetermine the accuracy of the results.

Ohmicron TNT method

Ohmicron’s TNT RaPID Assay can be usedfor the determination of TNT in water andsoil [11]. Ohmicron does not currentlyproduce an RDX field analytical method.The method detection limit for TNT in wateris 0.07 µg/l. A control solution of 2.0 µg/l

TNT is included in the Ohmicron kit andshould be analyzed with every test batch.Four standards at 0.0 µg/l, 0.25 µg/l, 1.0µg/l, and 5.0 µg/l concentrations should beanalyzed. A replicate analysis should beconducted for all samples, standards, and thecontrol. The relationship between RaPIDAnalyzer (analyzer) absorbance values andthe concentration of TNT in samples can becalculated manually or determinedautomatically by inputting specificparameters in to the analyzer. To constructthe calibration curve manually, the meanabsorbance value for each of the standards(%B) must be calculated and divided by themean absorbance value for the zero standard(B0). The resulting value (%B/B0) should beplotted on the vertical axis of logarithmicgraph paper versus TNT concentration (µg/l)on the horizontal axis. Calculated values of%B/B0 for the samples will yield TNTconcentrations (µg/l) by using the standardscalibration curve.

The analyzer operating manual providesdetailed instructions for inputting parametersinto the instrument such that a calibrationcurve, based on the standard concentrations,can be automatically calculated and stored.Once the curve is stored in the analyzer, allreadings taken from the analyzer will beconcentrations based on interpolation fromthe curve, displayed in µg/l. The upperconcentration limit of the TNT water testrange is 5.0 µg/l. Sample concentrationsabove this limit must be diluted. Analyzerresults for diluted samples must bemultiplied by the dilution factor. Calibrationcurves were calculated automatically by theanalyzer for all Umatilla and SUBASEBangor samples.

NRL TNT and RDX Continuous-Flow Immunosensor

The NRL Continuous Flow Immunosensor(CFI) is an antibody-based biosensorcapable of detecting low molecular weightmolecules in aqueous solutions [12]. TNT orRDX antibodies are immobilized onto a solidsupport matrix that is saturated with afluorescent labeled antigen. The supportmatrix is then placed in a disposablelaboratory column. A peristaltic pump isused to pump an aqueous buffer solutionthrough the column. Sample injection isfacilitated by a sample injector line locatedupstream of the column. Samples areprepared with a buffer solution prior toinjection into the system. A fluorimeter islocated downstream of the column. When asample containing TNT or RDX isintroduced, the TNT or RDX in the samplebinds to the immobilized antibody, causingsome of the labeled antigen to be displaced.The labeled antigen is detected downstreamin the fluorimeter, resulting in a positivesignal. The intensity of the fluorimeter signalis proportional to the concentration of TNTor RDX in the sample. The fluorimeter signalis recorded and analyzed by a signalintegrator/laptop computer. All of thesolutions tested by the NRL CFI containedeither the sample or the control water. Thefluorimeter readings recorded four minutesafter injection are compared for all samples.

NRL TNT Fiber-Optic Biosensor

The Fiber-Optic Biosensor (FOB) methodfor the detection of TNT is based upon acompetitive immunoassay using afluorescent dye as the reporter molecule [13,14]. Fluorescently labeled TNB is used asthe competitor for the competitiveimmunoassay on the surface of an opticalprobe. The labeled TNB is exposed to anantibody-coated optical fiber for fourminutes, generating a specific signal that

corresponds to the 100% or reference signal.The 100% or reference signal is defined asthe signal change associated with the labeledTNB alone. Inhibition of this signal isobserved when TNT is present in a sample.The percent inhibition observed isproportional to the TNT concentration in thesample. A 100% signal value is determinedboth before and after running the sample inorder to normalize for the gradual decreasein antibody activity.

For the Umatilla and SUBASE Bangorproject, the NRL FOB was analyzed usingtwo different test solution concentrations todetermine which concentration would yieldoptimal results. The first test solutionscontained 20% sample and 80% water andbuffer solution and is referred to as fiberoptic (20%). Test solutions containing 85%sample and 15% buffer solution were alsoanalyzed and referred to as fiber optic(85%). A standard curve was constructedbased on the TNT standards. Using thisstandards curve, sample results reported aspercent inhibition were converted to ng/ml(µg/l). Accuracy for the 85% sample washigher and is reported in this paper.

Laboratory analysis (EPA SW-846Method 8330)

Splits of all samples were shipped to theCRREL laboratory for EPA SW-846Method 8330 high performance liquidchromatography (HPLC) explosives analysisutilizing the direct inject method for samplesabove 20 µg/l [15, 16]. NRL CBMSE alsoperformed Method 8330 analysis on splitsfrom the standards and some of the samples.TNT, RDX, TNB , and HMX results weredetermined by CRREL and by NRL for aportion of the data set. Standard analyticalreference materials (SARMs) of TNT, RDX,TNB, and HMX were obtained from the U.S.Army Environmental Center. Standards

contained various predeterminedconcentrations of TNT, RDX, TNB, andHMX. Standards were used to calculate aresponse factor for the CRREL-EnSysmethod, to determine if TNB and HMX actas positive interference/cross reactivity forTNT and RDX, respectively, and to providelaboratory and on-site method analyseschecks. All environmental samples were alsoanalyzed by a commercial laboratory fornitrate/nitrite using EPA Method 353.2 [17].

RESULTS

The CRREL SW-846 Method 8330 resultswere used as a baseline to evaluate theaccuracy of the TNT and RDX on-siteanalytical methods. The results from the twosites are evaluated separately because of thedifferent environmental settings at each site.Umatilla is in a relatively arid setting withminimal soil development in granular strata,while SUBASE Bangor is in a relatively wetclimate with soils containing a high organiccontent. The aquifer sampled at Umatillaconsists primarily of flood plain gravels.Ground water at the explosives washoutlagoons have high nitrate concentrations andlow turbidity. SUBASE Bangorhydrogeology consists of fluvial/glacialdeposition. Ground water and leachate fromSUBASE Bangor tend to have relatively highorganic content and higher turbidity.

Accuracy

Accuracy is a measure of bias in the testingand analyses procedures. The closer ameasured value is to the true value, the moreaccurate the measurement. For comparisonof on-site analytical results, Method 8330concentrations were considered to be the“true value” or baseline concentration. Theaccuracy of on-site analytical results was

estimated using two separate methods, linearregression analysis and relative percentdifference (RPD). The DTECH methodprovides a concentration range. To obtain anactual DTECH concentration, a linearcorrelation between the DTECHTORmeasurement and the concentration rangewas assumed. The DTECHTOR measuredvalue was used to generate an estimatedconcentration. The derived estimatedconcentration was used in linear regressionand RPD calculations.

Linear regression analysis

The first method used linear regressionanalysis of the on-site method concentrationversus the Method 8330 concentrations.Overall quality of the data can be evaluatedusing linear regression. Under idealconditions, true accuracy would have a slopeof 1.0 and a correlation coefficient (R) of1.0. A slope less than 1.0 indicates that on-site method results are less than Method8330 concentrations. A slope greater than1.0 indicates that on-site method results aregreater than Method 8330 concentrations.The closer the R value is to 1.0, the betterthe correlation is to the best fit line,indicating less scatter in the data. Linearregression parameters of slope, correlationcoefficient (R), and number of samples (N)for a zero intercept line are shown in Table3.

Relative percent difference

The second method compared the relativepercent difference (RPD) between Method8330 concentrations and the results of eachon-site analytical method. The RPD betweenMethod 8330 concentrations and on-siteanalytical concentrations was calculatedwith the following equation:

Relative Percent Difference (RPD)

( )RPD

DF DLDF DL=

−+ ×

2100 ,

where DF = field method concentration andDL = lab method concentration.

Based on this equation, the smaller theabsolute RPD, the closer the on-siteanalytical result and Method 8330concentration, the more accurate the on-siteanalytical method. RPD mean and medianresults and number of samples (N) areshown in Table 4.

CONCLUSIONS

The accuracy of the methods varieddepending on the site-specific ground water

water quality parameters. No single fieldanalytical method outperformed the othermethods in all of the comparisons. Groundwater at Umatilla tends to have high nitrateconcentrations and low turbidity. Groundwater and leachate at SUBASE Bangor tendsto have relatively high organic content andhigher turbidity.

Colorimetric TNT accuracies were similar atboth sites. Colorimetric RDX results wereslightly more accurate at SUBASE Bangorthan at Umatilla. Nitrates are known toaffect the CRREL-EnSys RDX analysis. Anitrate removal step was implemented atboth sites. However, residual nitrates mayhave remained on the membranes after thenitrate removal step for Umatilla samples.Residual nitrates which remain on the

TABLE 3. LINEAR REGRESSION ANALYSIS.

Umatilla Army Depot Activity, Oregon.TNT—Regression Parameters

Method Slope R NCRREL/EnSys 1.5 0.97 15DTECH 2.0 0.88 15NRL CFI 1.4 0.69 11NRL FOB 1.6 0.91 12

RDX—Regression ParametersMethod Slope R N

CRREL/EnSys 0.86 0.86 23DTECH 1.2 0.96 23NRL CFI 0.98 0.71 20

Naval Subase Bangor, WashingtonTNT—Regression Parameters

Method Slope R NCRREL/EnSys 1.1 0.994 9DTECH 11.0 0.994 7Ohmicron 1.1 0.98 7NRL CFI 1.4 0.82 7NRL FOB 0.89 0.64 8

RDX—Regression ParametersMethod Slope R N


TABLE 4. RELATIVE PERCENT DIFFERENCE

RESULTS.

Umatilla Army Depot Activity, OregonTNT—Relative Percent Difference

Method Mean RPD Median RPD NCRREL/EnSys 66.0 44.9 15DTECH 63.9 47.8 15NRL CFI 74.2 29.6 11NRL FOB 33.0 25.1 12

RDX—Relative Percent Difference.Method Mean RPD Median RPD N


Naval Subase Bangor, WashingtonTNT—Relative Percent Difference

Method Mean RPD Median RPD NCRREL/EnSys 58.3 63.3 9DTECH 143.0 152.5 7Ohmicron 80.4 92.4 7NRL CFI 52.3 38.3 7NRL FOB 106.6 115.6 8

RDX—Relative Percent DifferenceMethod Mean RPD Median RPD N


membrane would be extracted with the RDXand consequently would affect the accuracyof the RDX analysis. The overall accuracyof the CRREL-EnSys RDX results wasacceptable, as indicated by linear regressionanalyses and RPDs. Based on these results,the nitrate removal step is recommended,particularly for use at sites which containhigh nitrate concentrations.

Generally, immunoassay-based fieldanalytical methods for TNT and RDX weremore accurate at Umatilla than at SUBASEBangor. Due to the nature of theimmunoassay and biosensor methods, thereis a potential for interference from organicmaterial and suspended particulates. Thesematerials may compete with explosivescompounds for binding sites. The binding oforganics and suspended particulates wouldgenerally lead to biased high TNT or RDXresults for reverse coloration methods.

The majority of the TNT RPD values forboth sites were positive and the linearregression slopes were greater than 1.0,indicating that TNT on-site analytical resultstended to be higher than the Method 8330TNT results. This positive bias could be dueto TNB interference/cross reactivity. In thecombined TNT data set, the CRREL-EnSysmethod and NRL CFI had similar accuracy,followed by the NRL FOB, Ohmicron, andDTECH methods.

For both Umatilla and SUBASE Bangor,there was a relatively even distributionbetween positive and negative RDX RPDsfor the CRREL-EnSys and DTECHmethods, indicating no discernible tendencyfor the on-site analytical method to be higheror lower than the Method 8330 RDX results.The NRL RDX CFI biosensor results tendedto be lower than Method 8330 RDX results.For the combined RDX data set, theCRREL-EnSys method and NRL CFI

methods showed similar accuracy, followedby DTECH.

In general, all the on-site analytical methodsperformed better on RDX than TNT. Thiscould be due, at least in part, to theconcentrations of each compoundencountered. RDX concentrations tended tobe higher than TNT concentrations. Higherconcentrations required dilutions to getresults into the working range of the method.The results may be subsequently lessaffected by matrix interference.

PRELIMINARY ECONOMICEVALUATION

A preliminary economic evaluation for along-term ground water remediation programwas based on a typical “pump and treat”system with a number of components [18,19]. These include extraction wells pipedinto two 10,000- to 20,000-pound GAC unitsin series, and that treated ground waterwould be reinjected into the aquifer. Designflow rates for the systems are estimatedbetween 500 and 1,500 gallons per minute.Sampling would be conducted weekly to bi-weekly at two or three locations (influent,between GAC units, and effluent) in thetreatment system. This results in anestimated 50 to 150 samples per year,excluding individual extraction well samplesand quality assurance samples. Thetreatment system would be expected tooperate for 10 to 30 years at an estimatedtotal remediation cost of $5 to $6 million.

Field analytical methods have three costcomponents including per sampleconsumables, initial equipment costs, andinstrument costs. Table 5 shows thesecomponents for the eight methods evaluated.Figures 1 and 2 show the costs forconducting on-site analytical monitoring forTNT and RDX , and for TNT only,

excluding labor, for a two to ten year timeperiod, based on a 100 sample per yearaverage. For comparison, typical cost forEPA Method 8330 with 30 day turn aroundtime is from $250 to $350 per sample.

RECOMMENDATIONS

Based on the results of this study, theCRREL-EnSys colorimetric TNT and RDXmethods were selected for routine groundwater monitoring at the Umatilla site basedon several considerations. These includeaccuracy, precision, detection limits, abilityto detect total explosives (i.e., nitroaromaticsplus nitramines), ease of use, and the costper sample. In addition, the NRL biosensormethods show considerable promise forcontinued development based on a numberof factors. These include accuracy, analysistime, ease of use, lifecycle costs, substantialreduction in solvent, solid, and chemicalwastes generated during analysis, and dataintegration capabilities. Further testing of theNRL biosensors will be conducted at bothUmatilla and SUBASE Bangor in the future.

The primary use for these methods would befor influent and effluent monitoring forground water and leachate treatmentsystems, which are projected to operate for aperiod of 10 to 30 years.

Site conditions are important when choosinga on-site analytical method. Site geology andground water quality also affects whichmethods should be chosen. Theimmunoassay-based methods generallyperformed better at Umatilla than atSUBASE Bangor. SUBASE Bangor groundwater has a higher organic content thanUmatilla ground water. Organic materialmay cause non-specific binding duringanalyses which could decrease the accuracyof the immunoassay-based methods. Thecolorimetric method had similar results atboth sites, indicating minimal affect ofground water quality on accuracy. However,a nitrate removal step was implemented forthe colorimetric analyses. Factors such asthe field location, access to electricity, andwhether the instrument station will remain ina given area or be moved to various

TABLE 5. PRELIMINARY ECONOMIC EVALUATION.

Cost Items ($)—TNT and RDXMethod Consume/Sample Initial Equip. Instrument

CRREL/EnSys 76.60 550 1600a

DTECH 51.50 0 300b

NRL CFI 8.00 800 20 Kd

Cost Items ($)—TNT OnlyMethod Consume/Sample Initial Equip. Instrument

CRREL/EnSys 37.10 550 1600a

DTECH 25.75 0 300b

Ohmicron 5.80 1,305 3985c

NRL CFI 4.00 800 20 Kd

NRL FOB 2.00 800 18 Ke

aHach DR2000 SpectrophotometerbEM Science DetectorcOhmicron RPA-1 RaPID AnalyzerdResearch International CFI PrototypeeResearch International Analyte 2000 FOB

field location, access to electricity, andwhether the instrument station will remain ina given area or be moved to variouslocations should also be considered.

The user must also determine whichcompounds are of concern at the site and forthe project. Different on-site analyticalmethods have different levels ofinterference/cross-reactivity with othercompounds. The data user must determine ifthere is a need to minimize (usingimmunoassay methods) or maximize (usingcolorimetric methods) the detection of otherexplosives-related compounds. If the projectrequires RDX analysis in addition to TNTanalysis, then an on-site analytical method

that can analyze for both compounds shouldbe used. The ultimate use of the data must beknown when choosing an on-site analyticalmethod.

REFERENCES

1. U.S. EPA, Explosives in Water FieldScreening Technologies UMDA andSUBASE Bangor (draft), prepared byBlack & Veatch Special Projects Corp.,Tacoma, WA, for U.S. EnvironmentalProtection Agency Region 10, ProjectNumber: 71370, March 1996.

2. U.S. EPA, Heath Advisory for RDX,U.S. Environmental Protection Agency,

FIGURE 1. COST PER SAMPLE AND TOTALANALYTICAL COST ($)—TNT AND RDX.

FIGURE 2. COST PER SAMPLE AND TOTALANALYTICAL COST ($)—TNT ONLY.

3. U.S. EPA, Trinitrotoluene HealthAdvisory, U.S. Environmental ProtectionAgency, Office of Drinking Water,Washington, DC, 1989.

4. K.R. Rogers and L.R. Williams,Biosensors for environmentalmonitoring: A regulatory perspective,Trends in Analytical Chemistry, Elsevier,14 (1995) 289-294.

5. K.R. Rogers, Biosensors forenvironmental monitoring, Biosensorsand Bioelectronics, 10 (1995) 533-541.

6. T.F. Jenkins, Development of aSimplified Field Method for theDetermination of TNT in Soil, U.S. ArmyCorps of Engineers, Cold RegionsResearch and Engineering Laboratory,Hanover, NH, CRREL Special Report90-38, November 1990.

7. M.E. Walsh and T.F. Jenkins,Development of Field Screening Methodfor RDX in Soil, U.S. Army Corps ofEngineers, Cold Regions Research andEngineering Laboratory, Hanover, NH,CRREL Special Report 91-7, June 1991.

8. T.F. Jenkins, P.G. Thorne, and M.E.Walsh, Field Screening Method for TNTand RDX in Groundwater, U.S. ArmyCorps of Engineers, Cold RegionsResearch and Engineering Laboratory,Hanover, NH, CRREL Special Report94-14, May 1994.

9. G. Teaney, J. Melby, and J. Stave, Anovel field analytical method for TNT,Proceedings of the AmericanAssociation of Analytical Chemists,1993.

10. G.B. Teaney and R.T. Hudak,Development of an enzymeimmunoassay based field screening

system for the detection of RDX in soiland water, Proceedings of the 87thAnnual Meeting and Exhibition, Air &Waste Management Association,Cincinnati, OH, 94-RP143.05, 1994.

11. F.R. Rubio, T.S. Lawruk, A.M. Gueco,D.P. Herzog, and J.R. Fleeker,Determination of TNT in soil and waterby a magnetic particle-based enzymeimmunoassay system, Proceedings of the11th Annual Waste Testing and QualityAssurance Symposium, AmericanChemical Society, July 23-28, 1995.

12. J.P. Whelan, A.W. Kusterbeck, G.A.Wemhoff, R. Bredehorst, and F.S.Ligler, Continuous Flow Immunosensorfor Detection of Explosives, NavalResearch Laboratory, Center forBio/Molecular Science and Engineering,Washington, DC, reprinted fromAnalytical Chemistry, 65, AmericanChemical Society, 1993.

13. F.S. Ligler, J.P. Golden, L.S. Shriver-Lake, R.A. Ogert, D. Wijesuria, and G.P.Anderson, Fiber-Optic Biosensor for theDetection of Hazardous Materials, NavalResearch Laboratory, Center forBio/Molecular Science and Engineering,Washington, DC, Immunomethods 3,122-127, Academic Press, Inc., 1993.

14. L.C. Shriver-Lake, K.A. Breslin, P.T.Charles, D.W. Conrad, J.P. Golden, andF.S. Ligler, Detection of TNT in waterusing a evanescent wave fiber-opticbiosensor, Analytical Chemistry, 67:14(1995) 2431-2435.

15. U.S. EPA, Test Methods for EvaluatingSolid Waste Physical/Chemical Methods,SW-846 Method 8330, Nitroaromaticsand Nitramines by High PerformanceLiquid Chromatography (HPLC),Revision 0, September 1994.

16. S.M. Golden, C.L. Grant, and T.F.Jenkins, Evaluating of Pre-ExtractionAnalytical Holding Times forNitroaromatic and NitroamineExplosives in Water, U.S. Army Corps ofEngineers, Cold Regions Research andEngineering Laboratory, Special Report93-24, Prepared for U.S. ArmyEnvironmental Center, August 1993.

17. U.S. EPA, Methods for ChemicalAnalysis of Water and Wastes, Method353.2, Revision 2, 1978.

18. U.S. EPA, Engineering Bulletin:Granular Activated Carbon Treatment,U.S. Environmental Protection Agency,Office of Emergency and RemedialResponse, Washington, DC, and Officeof Research and Development,Cincinnati, OH, EPA/540/2-91/024,October 1991.

19. U.S. EPA, Handbook: Approaches forthe Remediation of Federal Facility SitesContaminated with Explosive orRadioactive Wastes, U.S. EnvironmentalProtection Agency, Office of Researchand Development, Cincinnati, OH,EPA/625/R-93-013, September 1993.

field demonstration of on-site analytical methods for tnt and rdx in

Documents