SPEC

IAL

REP

OR

T

98

-4

Alan B. Crockett, Thomas F. Jenkins, Harry D. Craig, February 1998and Wayne E. Sisk

Overview of On-Site AnalyticalMethods for Explosives in Soil

Abstract: On-site methods for explosives in soil arereviewed. Current methods emphasize the detection ofTNT and RDX. Methods that have undergone signifi-cant validation fall into two categories: colorimetric-based methods and enzyme immunoassay methods.Discussions include considerations of specificity, de-

How to get copies of CRREL technical publications:

Department of Defense personnel and contractors may order reports through the Defense Technical Information Center:DTIC-BR SUITE 09448725 JOHN J KINGMAN RDFT BELVOIR VA 22060-6218Telephone 1 800 225 3842E-mail help@dtic.mil

msorders@dtic.milWWW http://www.dtic.dla.mil/

All others may order reports through the National Technical Information Service:NTIS5285 PORT ROYAL RDSPRINGFIELD VA 22161Telephone 1 703 487 4650

1 703 487 4639 (TDD for the hearing-impaired)E-mail orders@ntis.fedworld.govWWW http://www.fedworld.gov/ntis/ntishome.html

A complete list of all CRREL technical publications is available from:USACRREL (CECRL-LP)72 LYME RDHANOVER NH 03755-1290Telephone 1 603 646 4338E-mail techpubs@crrel.usace.army.mil

For information on all aspects of the Cold Regions Research and Engineering Laboratory, visit our World Wide Web site:http://www.crrel.usace.army.mil

tection limits, extraction, cost, and ease of use. A dis-cussion of the unique sampling design considerations isalso provided as well as an overview of the most com-monly employed laboratory method for analyzing explo-sives in soil. A short summary of ongoing developmentactivities is provided.

Special Report 98-4

Alan B. Crockett, Thomas F. Jenkins, Harry D. Craig, February 1998and Wayne E. Sisk

Prepared for

OFFICE OF THE CHIEF OF ENGINEERS

Approved for public release; distribution is unlimited.

Overview of On-Site AnalyticalMethods for Explosives in Soil

US Army Corpsof EngineersCold Regions Research &Engineering Laboratory

ii

PREFACE

This report was prepared by Alan B. Crockett, Idaho National Engineering andEnvironmental Laboratory, Lockheed Martin Idaho Technologies Company, IdahoFalls, Idaho; Dr. Thomas F. Jenkins, U.S. Army Cold Regions Research and Engi-neering Laboratory, Hanover, New Hampshire; Harry D. Craig, U.S. Environmen-tal Protection Agency, Region 10, Portland, Oregon; and Wayne E. Sisk, U.S. ArmyEnvironmental Center, Aberdeen Proving Ground, Maryland.

Funding for this project was provided by the EPA National Exposure ResearchLaboratorys Characterization Research Division, Ken Brown, Technology SupportCenter Director, with the assistance of the Superfund Projects Technology SupportCenter for Monitoring and Site Characterization; the U.S. Army Environmental Cen-ter; and the U.S. Army Corps of Engineers Installation Restoration Program (IRRP),Work Unit AF25-CT-006. Dr. Clem Myer was the IRRP Coordinator at the Director-ate of Research and Development, and Dr. M. John Cullinane was the ProgramManager at the U.S. Army Engineer Waterways Experiment Station.

The authors acknowledge the following individuals who provided valuable re-view comments during the production of this report: Marianne E. Walsh (CRREL),Martin H. Stutz (AEC), Karen F. Myers (WES), and Dr. Ken Brown (U.S. EPA).

This publication reflects the personal views of the authors and does not suggestor reflect the policy, practices, programs, or doctrine of the U.S. Army or Govern-ment of the United States. The contents of this report are not to be used for adver-tising or promotional purposes. Citation of brand names does not constitute anofficial endorsement or approval of the use of such commercial products.

iii

CONTENTSPage

Preface .......................................................................................................................... iiIntroduction ................................................................................................................. 1Background.................................................................................................................. 1Overview of sampling and analysis for explosives in soil ................................... 3Data quality objectives ............................................................................................... 4Unique sampling design considerations for explosives ....................................... 5

Heterogeneity problems and solutions ........................................................... 5Sample holding times and preservation procedures ..................................... 11Explosion hazards and shipping limitations .................................................. 11

Procedures for statistically comparing on-site and referenceanalytical methods ...................................................................................... 12

Precision and bias tests for measurements of relativelyhomogeneous material ............................................................................... 13

Precision and bias tests for measurements over largevalue ranges ................................................................................................. 13

Comparison to regulatory thresholds, action limits, etc ............................... 14Summary of on-site analytical methods for explosives in soil ............................ 14

Method type, analytes, and EPA method number ......................................... 15Detection limits and range ................................................................................ 15Type of results ..................................................................................................... 16Samples per batch ............................................................................................... 16Sample size .......................................................................................................... 16Sample preparation and extraction .................................................................. 16Analysis time ....................................................................................................... 17Interferences and cross-reactivity ..................................................................... 17Recommended quality assurance/quality control ........................................ 20Storage conditions and shelf life ...................................................................... 20Skill level .............................................................................................................. 20Cost ....................................................................................................................... 20Comparisons to laboratory method, SW-846 Method 8330 .......................... 20Additional considerations ................................................................................. 22Emerging methods and other literature reviewed......................................... 23

Summary of the EPA reference method for explosives compounds,Method 8330 ................................................................................................ 24

Properties of secondary explosives .................................................................. 24Soil extraction ...................................................................................................... 24RP-HPLC determination (Method 8330) ......................................................... 25Method specifications and validation (Method 8330) ................................... 25

Summary ...................................................................................................................... 25Postscript ...................................................................................................................... 26Literature cited ............................................................................................................ 26Abstract ........................................................................................................................ 31

TABLES

Table Page1. Analytical methods for commonly occurring explosives, propellants,

and impurities/degradation products .................................................... 22. Occurrence of analytes detected in soil contaminated with explosives ....... 33. Comparative data for selecting on-site analytical methods for

explosives in soil ......................................................................................... 74. Available on-site analytical methods for explosives in soil ............................ 155. On-site analytical methods for explosives in soil, percent

interference or cross-reactivity .................................................................. 196. Comparison of on-site analytical methods for TNT, RDX, and

HMX to EPA Method 8330 ........................................................................ 227. Physical and chemical properties of predominant nitroaromatics

and nitramines ............................................................................................ 24

iv

Overview of On-Site Analytical Methodsfor Explosives in Soil

ALAN B. CROCKETT, THOMAS F. JENKINS, HARRY D. CRAIG, AND WAYNE E. SISK

INTRODUCTION

The purpose of this report is to survey the cur-rent status of field sampling and on-site analyti-cal methods for detecting and quantifying second-ary explosives compounds in soils (Table 1). Thepaper also includes a brief discussion of EPAMethod 8330 (EPA 1995), the reference analyticalmethod for the determination of 14 explosives andco-contaminants in soil.

This report is divided into the following majorsections: introduction; background; an overviewof sampling and analysis for explosives in soil;data quality objectives; unique sampling designconsiderations for explosives; procedures for sta-tistically comparing on-site and reference analyti-cal methods; a summary of on-site analytical meth-ods; and a summary of the current EPA referenceanalytical method, Method 8330 (EPA 1995). Al-though some sections may be used independently,joint use of the field sampling and on-site analyti-cal methods sections is recommended to developa sampling and analytical approach that achievesproject objectives.

Many of the explosives listed in Table 1 are notspecific target compounds of screening methods,yet they may be detected by one or more screen-ing methods because of their similar chemicalstructure. Also listed are the explosives and pro-pellant compounds targeted by high performanceliquid chromatography (HPLC) methods, includ-ing EPA SW-846 Method 8330, the standardmethod required by EPA regions for laboratoryconfirmation.

BACKGROUND

Evaluating sites potentially contaminated withexplosives is necessary to carry out U.S. Depart-ment of Defense, EPA, and U.S. Department of

Energy policies on site characterization andremediation under the Superfund, Resource Con-servation and Recovery Act (RCRA), InstallationRestoration, Base Closure, and Formerly UsedDefense Site environmental programs. Facilitiesthat may be contaminated with explosives include,for example, active and former manufacturingplants, ordnance works, Army ammunition plants,Naval ordnance plants, Army depots, Naval am-munition depots, Army and Naval provinggrounds, burning grounds, artillery impact ranges,explosive ordnance disposal sites, bombingranges, firing ranges, and ordnance test and evalu-ation facilities.

Historical disposal practices from manufactur-ing, spills, ordnance demilitarization, lagoon dis-posal of explosives-contaminated wastewater, andopen burn/open detonation (OB/OD) of explo-sives sludges, waste explosives, excess propellants,and unexploded ordnance often result in soils con-tamination. Common munitions fillers and theirassociated secondary explosives include Amatol(ammonium nitrate/TNT), Baratol (barium ni-trate/TNT), Cyclonite or Hexogen (RDX),Cyclotols (RDX/TNT), Composition A-3 (RDX),Composition B (TNT/RDX), Composition C-4(RDX), Explosive D or Yellow D (AP/PA), Octogen(HMX), Octols (HMX/TNT), Pentolite (PETN/TNT), Picratol (AP/TNT), tritonal (TNT), tetrytols(tetryl/TNT), and Torpex (RDX/TNT).

Propellant compounds include DNTs andsingle-base (NC), double-base (NC/NG), andtriple-base (NC/NG/NQ) smokeless powders. Inaddition, NC is frequently spiked with other com-pounds (e.g., TNT, DNT, DNB) to increase its ex-plosive properties. AP/PA is used primarily inNaval munitions such as mines, depth charges,and medium-to-large caliber projectiles. Tetryl isused primarily as a boosting charge, and PETN isused in detonation cord.

A number of munitions facilities have high lev-

els of soil and groundwater contamination,although on-site waste disposal was discon-tinued 20 to 50 years ago. Under ambient envi-ronmental conditions, explosives are highly per-sistent in soils and groundwater, exhibiting aresistance to naturally occurring volatilization, bio-degradation, and hydrolysis. Where biodegrada-tion of TNT occurs, 2-AmDNT and 4-AmDNT arethe most commonly identified transformationproducts. Photochemical decomposition of TNTto TNB occurs in the presence of sunlight andwater, with TNB being generally resistant to fur-ther photodegradation. TNB is subject to biotrans-formation to 3,5-dinitroaniline, which has been rec-ommended as an additional target analyte in EPA

Method 8330. Picrate is a hydrolysis trans-formation product of tetryl, and is expected inenvironmental samples contaminated with tetryl.Site investigations indicate that TNT is the leastmobile of the explosives and most frequently oc-curring soil contamination problem. RDX andHMX are the most mobile explosives and presentthe largest groundwater contamination problem.TNB, DNTs, and tetryl are of intermediate mobil-ity and frequently occur as co-contaminants in soiland groundwater. Metals are co-contaminants atfacilities where munitions compounds werehandled, particularly at OB/OD sites. Field ana-lytical procedures for metals, such as x-ray fluo-rescence, may be useful in screening soils for

Table 1. Analytical methods for commonly occurring explosives, propellants, and impurities/degradation products.

Field LaboratoryAcronym Compound name method method

TNT 2,4,6-trinitrotoluene Cp, Ip NTNB 1,3,5-trinitrobenzene Cs, Is NDNB 1,3-dinitrobenzene Cs N2,4-DNT 2,4-dinitrotoluene Cp, Cs N2,6-DNT 2,6-dinitrotoluene Cs, Is NTetryl Methyl-2,4,6-trinitrophenylnitramine Cs N2-AmDNT 2-amino-4,6-dinitrotoluene N4-AmDNT 4-amino-2,6-dinitrotoluene Is NNT Nitrotoluene (3 isomers) NNB Nitrobenzene N

Nitramines Cs NRDX Hexahydro-1,3,5-trinitro-1,3,5-triazine Cp, Ip NHMX Octahydro-1,3,5,7-tetranitro-1,3,5,7-tetrazocine Cs NNQ Nitroguanidine Cs G

Nitrate esters CsNC Nitrocellulose Cs *LNG Nitroglycerin Cs *PPETN Pentaerythritol tetranitrate Cs *P

Ammonium picrate/picric acidAP/PA Ammonium 2,4,6-trinitrophenoxide/2,4,6-trinitrophenol Cp, Is A

Cp = Colorimetric field method, primary target analyte(s).Cs = Colorimetric field method, secondary target analyte(s).Ip = Immunoassay field method, primary target analyte(s).Is = Immunoassay field method, secondary target analyte(s).N = EPA SW-846, Nitroaromatics and Nitramines by HPLC, Method 8330 (EPA 1995a).P = PETN and NG (Walsh unpublished CRREL method).G = Nitroguanidine (Walsh 1989).L = Nitrocellulose (Walsh unpublished CRREL method).A = Ammonium picrate/picric acid (Thorne and Jenkins 1995a).

*The performance of a number of field methods has not been assessed utilizing approved laboratory methods. It is recommendedthat verification of the performance of any analytical method be an integral part of a sampling/analysis projects quality assuranceprogram.

2

metals in conjunction with explosives at muni-tions sites.

The frequency of occurrence of specific explo-sives in soils was assessed by Walsh et al. (1993),who compiled analytical data on soils collectedfrom 44 Army ammunition plants, arsenals, anddepots, and two explosive ordnance disposal sites.Of the 1155 samples analyzed by EPA Method8330, 319 samples (28%) contained detectablelevels of explosives. The frequency of occurrenceand the maximum concentrations detected areshown in Table 2. TNT was the most commonlyoccurring compound in contaminated samples; itwas detected in 66% of the contaminated samplesand in 80% of the samples if the two explosive ord-nance disposal sites are excluded. Overall, eitherTNT or RDX or both were detected in 72% of thesamples containing explosives residues, and 94%if the ordnance sites are excluded. Thus, by screen-ing for TNT and RDX at ammunition plants, arse-nals, and depots, 94% of the contaminated areascould be identified (80% if only TNT wasdetermined). This demonstrates the feasibility ofscreening for one or two compounds or classes ofcompounds to identify the initial extent of con-tamination at munitions sites. The two ordnancesites were predominantly contaminated withDNTs, probably from improper detonation ofwaste propellant. The table also shows that NBand NTs were not detected in these samples; how-ever, NTs are found in waste produced from themanufacture of DNT.

OVERVIEW OF SAMPLING AND ANALYSISFOR EXPLOSIVES IN SOIL

The environmental characteristics of munitionscompounds in soil indicate that they are extremelyheterogeneous in spatial distribution. Concentra-tions range from nondetectable levels (10,000 ppm)for samples collected within several feet of eachother. Some waste disposal practices, such as OB/OD, exacerbate the problem at these sites and mayresult in conditions ranging from no soil contami-nation up to solid chunks of bulk secondaryexplosives, such as TNT or RDX. Secondary ex-plosives concentrations above 10% (>100,000 ppm)in soil are also of concern from a potential reactiv-ity standpoint and may affect sample and materi-als handling processes during remediation. Anexplosives hazard safety analysis is needed formaterials handling equipment to prevent initiat-ing forces that could propagate a detonationthroughout the soil mass.

Reliance on laboratory analyses only for sitecharacterization may result in a large percentageof samples (up to 80%, depending upon the site)with nondetectable levels. The remaining samplesmay indicate concentrations within a range of fourorders of magnitude. Analyzing a small numberof samples at an off-site laboratory may result ininadequate site characterization for estimating soilquantities for remediation and may miss poten-tially reactive material. Laboratory analytical costsvary depending on required turnaround time.Typical costs for EPA Method 8330 analysis rangefrom $250 to $350 per sample for 30-day turn-around, $500 to $600 for 7-day turnaround, andapproximately $1000 per sample for 3-day turn-around, if it is available.

Because of the extremely heterogeneous distri-bution of explosives in soils, on-site analyticalmethods are a valuable, cost-effective tool to as-sess the nature and extent of contamination. Be-cause costs per sample are lower, more samplescan be analyzed and the availability of near-real-time results permits redesign of the samplingscheme while in the field. On-site screening alsofacilitates more effective use of off-site laborato-ries using more robust analytical methods. Evenif only on-site methods are used to determine thepresence or absence of contamination (i.e., all posi-tive samples are sent off-site for laboratory analy-sis), analytical costs can be reduced considerably.Because on-site methods provide near-real-timefeedback, the results of screening can be used to

Table 2. Occurrence of analytes detected insoil contaminated with explosives.

Sample with Maximumanalyte present level

Compound (%) (g/g)

NitroaromaticsTNT 66 102,000TNB 34 1790DNB 17 612,4-DNT 45 3182,6-DNT 7 4.52-AmDNT 17 3734-AmDNT 7 11Tetryl 9 1260

NitraminesRDX 27 13,900HMX 12 5700

TNT and/or RDX 72

Derived from Walsh et al. (1993).

3

focus additional sampling on areas of known con-tamination, thus possibly saving additional mo-bilization and sampling efforts. This approach hasbeen successfully used for a Superfund remedialinvestigation of an OB/OD site (Craig et al. 1993).

During site remediation, such as Superfund re-medial actions, data are needed on a near-real-timebasis to assess the progress of cleanup. On-sitemethods can be used during remediation to guideexcavation and materials handling activities andto evaluate the need for treatment on incrementalquantities of soil (EPA 1992b). Final attainment ofsoil cleanup levels should be determined by anapproved laboratory method, such as EPA Method8330. This approach was effectively used at aSuperfund remedial action for an explosives wash-out lagoon (Oresik et al. 1994, Markos et al. 1995).

DATA QUALITY OBJECTIVES

The EPA Data Quality Objectives process isdesigned to facilitate the planning of environmen-tal data collection activities by specifying the in-tended use of the data (what decision is to bemade), the decision criteria (action level), and thetolerable error rates (EPA 1994, ASTM 1996). Inte-grated use of on-site and laboratory methods forexplosives in soil helps to achieve such objectivesas determining the horizontal and vertical extentof contamination, obtaining data to conduct a riskassessment, identifying candidate wastes for treat-ability studies, identifying the volume of soil tobe remediated, determining whether soil presentsa potential detonation hazard (reactive accordingto RCRA regulations), and determining whetherremediation activities have met the cleanupcriteria.

Environmental data such as rates of occurrence,average concentrations, and coefficients of varia-tion are typically highly variable for contaminantsassociated with explosives sites. These differencesare a function of fate and transport properties,occurrence in different media, and interactionswith other chemicals, in addition to use and dis-posal practices. Information on frequency of oc-currence and coefficient of variation determinesthe number of samples required to adequatelycharacterize exposure pathways and is essentialwhen designing sampling plans. Low frequenciesof occurrence and high coefficients of variation,such as with explosives, mean that more sampleswill be required to characterize the exposure path-ways of interest. Sampling variability typically

contributes much more to total error than analyti-cal variability (EPA 1990, 1992a). Under these con-ditions, the major effort should be to reduce sam-pling variability by taking more samples using lessexpensive methods (EPA 1992a).

EPAs Guidance for Data Useability in Risk As-sessment (EPA 1992a) indicates that on-site meth-ods can produce legally defensible data if appro-priate method quality control is available and ifdocumentation is adequate. Field analyses can beused to decrease cost and turnaround time as longas supplemental data are available from an ana-lytical method capable of quantifying multipleexplosives analytes (e.g., Method 8330) (EPA1992a). Significant quality assurance oversight offield analysis is recommended to enable the datato be widely used. The accuracy (correctness ofthe concentration value and a combination of bothsystematic error [bias] and random error [preci-sion]) of on-site measurements may not be as highin the field as in fixed laboratories, but the quickerturnaround and the possibility of analyzing alarger number of samples more than compensatesfor this factor. Remedial project managers, in con-sultation with chemists and quality assurance per-sonnel, should set accuracy levels for each methodand proficiency standards for the on-site analyst.

On-site methods may be useful for analysis ofwaste treatment residues, such as incineration ash,compost, and bioslurry reactor sludges. However,on-site methods should be evaluated against labo-ratory methods on a site- and matrix-specific ba-sis because of the possibility of matrix interference.Treatability studies are used to evaluate the po-tential of different treatment technologies to de-grade target and intermediate compounds and toevaluate whether cleanup levels may be achievedfor site remediation. Treatability study waste forexplosives-contaminated soils should be of higher-than-average concentration to evaluate the effectsof heterogeneous concentrations and the poten-tial toxicity effects for processes such asbioremediation.

During remediation of soils contaminated withexplosives, it is necessary to monitor the rate ofdegradation and determine when treatment crite-ria have been met so that residues below cleanuplevels can be disposed of and additional soiltreated. Soils contaminated with explosives arecurrently being treated by incineration,composting, and solidification/stabilization(Noland et al. 1984, Turkeltaub et al. 1989, EPA1993, Craig and Sisk 1994, Miller and Anderson1995, Channell et al. 1996). Other biological treat-

4

ment systems that have been evaluated for treat-ing explosives-contaminated soils include anaero-bic bioslurry, aerobic bioslurry, white rot fungus,and landfarming (Craig et al. 1995, Sundquist etal. 1995).

UNIQUE SAMPLING DESIGN CONSIDER-ATIONS FOR EXPLOSIVES

Heterogeneity problems and solutionsThe heterogeneous distribution of explosives in

soil is often alluded to but seldom quantified. Theproblem is probably considerably greater for ex-plosives residues in soil than for most other or-ganic waste. According to available Superfund sitedata, the median coefficient of variation (CV) (stan-dard deviation divided by the mean) for volatiles,extractables, pesticides/polychlorinated biphe-nyls (PCBs), and tentatively identified compoundsin soils ranges from 0.21 to 54% for individual con-taminants (EPA 1992b). Data from 11 munitionssites show the median CV for TNT was 284%, andthe TNT CV ranged from 127% to 335% for indi-vidual sites. Comparable data for RDX show amedian CV of 137% with a range of 129% to 203%;the median CVs for 2,4-DNT, AP/PA, and PETNwere 414%, 184%, and 178%, respectively. If thenatural variability of the chemicals of potentialconcern is large (e.g., CV >30%), the major plan-ning effort should be to collect more environmen-tal samples (EPA 1992b).

Jenkins et al. (1996a, b) recently conducted stud-ies to quantify the short-range sampling variabil-ity and analytical error of soils contaminated withexplosives. Nine locations (three at each of threedifferent facilities) were sampled. At each location,seven core samples were collected from a circlewith a radius of 61 cm: one from the center andsix equally spaced around the circumference. Theindividual samples and a composite sample of theseven samples were analyzed in duplicate, on-site,using the EnSys RISc colorimetric soil test kit forTNT (on-site method) and later by Method 8330at an off-site laboratory. Results showed extremevariation in concentration at five of the nine loca-tions, with the remaining four locations showingmore modest variability. For sites with modestvariability, only a small fraction of the total errorwas because of analytical error, i.e., field samplingerror dominated total error. For the locations show-ing extreme short-range heterogeneity, samplingerror overwhelmed analytical error. Contaminantdistributions were very site-specific, dependent on

a number of variables such as waste disposal his-tory, the physical and chemical properties of thespecific explosive, and the soil type. The conclu-sion was that, to improve the quality of site char-acterization data, the major effort should be placedon the use of higher sampling densities and com-posite sampling strategies to reduce samplingerror.

In a subsequent study at an HMX-contaminatedantitank firing range, similar short-range hetero-geneity was observed (Jenkins et al. 1997). At boththe short- and mid-scale, sampling error over-whelmed analytical error. It was observed that theparticulate nature of these contaminants in near-surface soils is a major contributor to substantialspatial heterogeneity in distribution.

There are several practical approaches to reduc-ing overall error during characterization of soilscontaminated with explosives, including increas-ing the number of samples or sampling density,collecting composite samples, using a stratifiedsampling design, and reducing within-sampleheterogeneity. Because explosives have very lowvolatility, loss of analytes during field preparationof composite samples is not a major concern.

Increasing the number of samplesOne simple way to improve spatial resolution

during characterization is by collecting moresamples using a finer sampling grid, such as a 5-m grid spacing instead of a 10-m spacing. Thoughdesirable, this approach has been rejected in thepast because of the higher sampling and analyti-cal laboratory costs. When inexpensive on-siteanalytical methods are used, this approach be-comes feasible. The slightly lower accuracy asso-ciated with on-site methods is more than compen-sated for by the greater number of samples thatcan be analyzed and the resultant reduction intotal error.

Collection of composite samplesThe collection of composite samples is another

very effective means of reducing sampling error.Samples are always taken to make inferences to alarger volume of material, and a set of compositesamples from a heterogeneous population pro-vides a more precise estimate of the mean than acomparable number of discrete samples. This oc-curs because compositing is a physical processof averaging (adequate mixing and subsamplingof the composite sample are essential to mostcompositing strategies). Averages of samples havegreater precision than the individual samples.

5

Decisions based on a set of composite samples will,for practical purposes, always provide greater sta-tistical confidence than for a comparable set ofindividual samples. In the study discussed aboveby Jenkins et al. (1996a, b; 1997), the compositesamples were much more representative of eachplot than the individual samples that made up thecomposites. Using a composite sampling strategyusually allows the total number of samples ana-lyzed to be reduced, thereby reducing costs whileimproving characterization. Compositing shouldbe used only when analytical costs are significant.An American Society for Testing and Materials(ASTM) guide was developed on composite sam-pling and field subsampling (Gagner and Crockett1996, ASTM 1997).

Stratified sampling designsStratified sampling may also be effective in re-

ducing field and subsampling errors. Using his-torical data and site knowledge or results frompreliminary on-site methods, it may be possibleto identify areas in which contaminant concentra-tions are expected to be moderately heterogeneous(pond bottom) or extremely heterogeneous (opendetonation sites). Different compositing and sam-pling strategies may be used to characterize dif-ferent areas, thereby increasing the likelihood ofa more efficient characterization.

Another means of stratification is based on par-ticle size. Because explosives residues often existin a wide range of particle sizes (crystals tochunks), it is possible to sieve samples into vari-ous size fractions that may reduce heterogeneity.If large chunks of explosives are present, it maybe practical to coarse-sieve a relatively largesample (many kilograms), medium-sieve a por-tion of those fines, and subsample the fines frommedium screening as well. This would yield threesamples of different particle size and presumesthat heterogeneity increases with coarseness. Eachfraction would be analyzed separately but notnecessarily by the same method (visual screeningof the coarser fractions for chunks of explosivesmay be possible) and then could be summed toyield the concentration on a weight or area basis.In addition, aqueous disposal of explosives waste-waters, such as those found in washout lagoonsor at spill sites, often results in preferential sorp-tion to fine-grained materials, such as fines orclays, particularly for nitroaromatics.

Reducing within-sample heterogeneityThe heterogeneity of explosives in soils is fre-

quently observed during the use of on-site ana-lytical methods in which duplicate subsamples areanalyzed and differ by more than an order of mag-nitude. Grant et al. (1993) conducted a holdingtime study using field-contaminated soils thatwere air-dried, ground with a mortar and pestle,sieved, subsampled in triplicate, and analyzedusing Method 8330. Even with such sample prepa-ration, the results failed to yield satisfactory pre-cision. (The relative standard deviations [RSDs]often exceeded 25%, compared with RSDs below3% at two other sites.) Subsampling in the field ismuch more challenging because complete sampleprocessing is not feasible. However, most screen-ing procedures specify relatively small samples,typically a few grams.

To reduce within-sample heterogeneity, twomethods can be employed: either homogenizationand extraction or analysis of a larger sample. Un-less directed otherwise, an analyst should assumethat information representative of the entire con-tents of the sample container is desired. Therefore,the subsample extracted or directly analyzedshould be representative of the container. Thesmaller the volume of that subsample removed foranalysis and extraction, the more homogeneousthe entire sample should be before subsampling(e.g., a representative 0.5-g subsample is more dif-ficult to obtain than a 20-g subsample from a 250-g sample). Collecting representative 2-gsubsamples from 300 g of soil is difficult and canrequire considerable sample processing, such asdrying, grinding, and riffle splitting. Even in thelaboratory, as discussed above, obtaining repre-sentative subsamples is difficult. An ASTM guidehas been developed to help in this regard (Gagnerand Crockett 1996). Although sample-mixing pro-cedures such as sieving to disaggregate particles,mixing in plastic bags, etc., can and should be usedto prepare a sample, extracting a larger sample isperhaps the easiest method of improving repre-sentativeness. For this reason, 20 g of soil is ex-tracted for the Cold Regions Research and Engi-neering Laboratory (CRREL) method, and thesame approach may easily be used to improve re-sults with most of the on-site methods shown inTable 3. The major disadvantage of extracting thelarger sample is the larger volume of waste sol-vent and solvent-contaminated soil that needs dis-posal.

The effectiveness of proper mixing in the fieldis illustrated in the recent reports by Jenkins et al.(1996a, b; 1997). Duplicate laboratory analyses ofthe same samples, including drying, grinding,

6

Tab

le 3

. Com

par

ativ

e d

ata

for

sele

ctin

g on

-sit

e an

alyt

ical

met

hod

s fo

r ex

plo

sive

s in

soi

l*.

Met

hod

type

,So

ilA

naly

sis

tim

eM

etho

d/an

alyt

es, a

nd E

PAD

etec

tion

ran

ge a

ndsa

mpl

eSa

mpl

e pr

epar

atio

npr

edic

tion

s ra

teK

itM

etho

d N

o.ra

nge

fact

orTy

pe o

f res

ults

Sam

ples

per

bat

chsi

zean

d ex

trac

tion

(one

per

son)

CR

REL

Col

orim

etri

cTN

T: 1

to 2

2 m

g/kg

(22

)TN

T, R

DX

: Qua

ntita

tive

TNT:

Bat

ch o

r si

ngle

20 g

3 m

in s

haki

ng in

100

30 m

inut

e ex

trac

t/6

sam

ples

;TN

T, R

DX

, 2,4

-DN

T,R

DX

: 1 to

20

mg/

kg (2

0)

2,4-

DN

T: S

emiq

uant

itativ

eR

DX

: 6 to

7/b

atch

mL

acet

one;

set

tling

;TN

T: 5

min

utes

/sam

ple;

RD

X:

Am

mon

ium

pic

rate

/2,

4-D

NT:

2 to

20

mg/

kg (1

0)

AP/

PA: Q

uant

itativ

eor

sin

gle

filtr

atio

n.30

min

utes

/6 R

DX

sam

ples

;pi

cric

aci

dA

P/PA

: 1.3

to 6

9 m

g/kg

(53

)2,

4-D

NT

and

AP/

PA:

25 s

ampl

es/d

ay fo

r TN

T +

RD

X;

Sing

le o

r ba

tch

DN

T: 3

0 m

inut

es/6

sam

ples

;A

P/PA

: 15

min

utes

/sam

ple.

EnSy

s R

ISc

Col

orim

etri

cTN

T: 1

to 3

0 m

g/kg

(30

)Q

uant

itativ

eSi

ngle

10 g

Dry

1% b

ased

on

IC50 (s

ee te

xt)

Rec

omm

ende

d Q

A/Q

Csh

elf l

ife o

f kit

or

reag

ents

Skill

leve

l

CR

RE

LT

NT

= T

NT

+ T

NB

+ D

NB

+ D

NTs

+ te

tryl

;B

lank

and

cal

ibra

tion

sta

ndar

ds

Stor

e at

roo

m te

mpe

ratu

re.

Med

ium

d

etec

tion

lim

its

(ppm

); T

NB

0.5

; DN

B

10%

, and

hum

ics

inte

rfer

e w

ith

TN

T a

nd R

DX

; nit

rate

and

nit

rite

inte

rfer

e w

ith

RD

X.

2,4-

DN

T =

2,4

-DN

T +

2,6

-DN

T +

TN

T +

TN

B +

tetr

yl; h

igh

copp

er, m

oist

ure,

and

hum

ics

inte

rfer

e.A

P/PA

= r

elat

ivel

y fr

ee o

f hum

ic a

nd n

itro

arom

atic

inte

rfer

ence

s.

EnS

ys R

ISc

TN

T =

TN

T +

TN

B +

DN

B +

DN

Ts +

tetr

yl;

Met

hod

and

soi

l bla

nks

and

aSt

ore

at r

oom

tem

pera

ture

.T

NT:

Low

d

etec

tion

lim

its

(ppm

); T

NB

0.5

; DN

B 10

%, a

nd h

umic

s in

terf

ere

wit

h T

NT

and

RD

X; n

itra

te a

nd n

itri

tesh

ould

be

conf

irm

ed.

inte

rfer

e w

ith

RD

X.

USA

CE

TN

B in

terf

eres

by

rais

ing

min

imum

det

ecti

on li

mit

.B

lank

soi

l sam

ple,

and

Stor

e at

roo

m te

mpe

ratu

re.

Med

ium

calib

rati

on s

tand

ard

pre

pare

dfr

om c

lean

sit

e so

il.

EN

VIR

OL

Tetr

yl is

a n

egat

ive

inte

rfer

ence

and

the

DN

Ts a

re a

pos

itiv

e in

terf

eren

ce.

Two

calib

rati

on s

tand

ard

s;R

ecom

men

ded

sto

rage

20

CM

ediu

mbl

ank

solu

tion

;A

ccep

tabl

e ra

nge

430

Cco

ntro

l sam

ple.

Shel

f lif

e: 3

mon

ths

if k

ept

belo

w 2

5C

DT

EC

HC

ross

rea

ctiv

ity:

Sam

ples

test

ing

posi

tive

Stor

e at

roo

m te

mpe

ratu

re o

rL

owT

NT:

tetr

yl =

35%

; TN

B =

23

%; 2

-Am

DN

T =

11%

; 2,4

-DN

T =

4%

;sh

ould

be

conf

irm

ed u

sing

refr

iger

ate;

do

not f

reez

e or

exc

eed

AP/

PA u

nkno

wn

but ~

100%

at l

ower

lim

it o

f det

ecti

on.

stan

dar

d m

etho

ds.

37C

for

prol

onge

d p

erio

d. S

helf

RD

X: H

MX

= 3

%lif

e 9

mon

ths

at r

oom

tem

pera

ture

.

Idet

ekC

ross

rea

ctiv

ity:

Dup

licat

e ex

trac

tion

s;R

efri

gera

te 2

to

8C

, do

not f

reez

eM

ediu

m-h

igh,

Qua

ntix

TN

B =

47%

; tet

ryl =

6.5

%; 2

,4-D

NT

= 2

%; 4

-Am

DN

T =

2%

1 in

10

repl

icat

e;or

exc

eed

37

C. S

helf

life

9 to

12

init

ial t

rain

ing

2 sa

mpl

e w

ells

/ex

trac

t.m

onth

s. A

void

dir

ect l

ight

.re

com

men

ded

Env

iroG

ard

Cro

ss r

eact

ivit

y:Pl

ate:

Sam

ples

run

inSt

ore

4 to

8C

; do

not f

reez

e or

Plat

e: M

ediu

m-

dup

licat

e.ex

ceed

37

C. D

o no

t exp

ose

high

Plat

e: 4

-Am

DN

T =

41%

; 2,6

-DN

T =

41%

; TN

B =

7%

; 2,4

-DN

T =

2%

subs

trat

e to

dir

ect s

unlig

ht.

Tube

: Med

ium

Tube

: 2,6

-DN

T =

20%

; 4-A

mD

NT

= 1

7%; T

NB

= 3

%; 2

,4-D

NT

= 2

%Sh

elf l

ife:

Pla

te 3

to 1

4 m

onth

sTu

be 3

to 6

mon

ths

Ohm

icro

nC

ross

rea

ctiv

ity:

Dup

licat

e st

and

ard

cur

ves;

pos

itiv

eR

efri

gera

te r

eage

nts

2 to

8C

.M

ediu

m-h

igh

RaP

IDT

NB

= 6

5%; 2

,4-D

init

roan

iline

= 6

%; t

etry

l = 5

%; 2

,4-D

NT

= 4

%; 2

-Am

DN

T =

3%

;co

ntro

l sam

ple

supp

lied

. Pos

itiv

eD

o no

t fre

eze.

init

ial t

rain

ing

Ass

ayD

NB

= 2

%re

sult

s re

quir

ing

acti

on m

ay n

eed

Shel

f lif

e 3

to 1

2 m

onth

s.re

com

men

ded

conf

irm

atio

n by

ano

ther

met

hod

.

*Exp

and

ed a

nd m

odif

ied

from

EPA

199

7.

8

Tab

le 3

. Com

par

ativ

e d

ata

for

sele

ctin

g on

-sit

e an

alyt

ical

met

hod

s fo

r ex

plo

sive

s in

soi

l* (c

onti

nu

ed).

Met

hod/

Trai

ning

Cos

tsC

ompa

riso

ns to

Met

hod

8330

Oth

erD

evel

oper

Kit

avai

labi

lity

(not

incl

udin

g la

bor)

refe

renc

esre

fere

nces

info

rmat

ion

Add

itio

nal c

onsi

dera

tion

s

CR

REL

Free

vid

eo fo

r T

NT

$15/

sam

ple

plus

$15

00Br

ouill

ard

et a

l. 19

93; E

PA 1

993,

199

5aJe

nkin

s et

Dr.

Thom

as F

. Jen

kins

Larg

e w

ork

area

(2 la

rge

desk

s); r

equi

res

and

RD

X; s

ee te

xtfo

r H

ach

spec

trom

eter

.(M

etho

d 85

15),

1995

b;al

. 199

5;C

RR

ELth

e m

ost s

etup

tim

e; p

ossi

ble

TNB

inte

r-fo

r ad

dres

s.Je

nkin

s 19

90; J

enki

ns a

nd W

alsh

199

2;Th

orne

and

72 L

yme

Roa

dfe

renc

e; n

o el

ectr

icity

or

refr

iger

atio

nN

one

avai

labl

e fo

rM

arko

s et

al.

1995

; Lan

g et

al.

1990

;Je

nkin

sH

anov

er, N

H 0

3755

-129

0re

quir

ed; d

eion

ized

wat

er r

equi

red;

mus

t2,

4-D

NT,

AP/

PA.

Wal

sh a

nd Je

nkin

s 19

91;

1995

b(6

03) 6

46-4

385

asse

mbl

e m

ater

ials

; gla

ssw

are

mus

t be

Jenk

ins

et a

l. 19

96a;

Jenk

ins

and

Wal

shri

nsed

bet

wee

n an

alys

es; l

arge

r vo

lum

e19

91, 1

992;

Tho

rne

and

Jenk

ins

1995

aof

ace

tone

was

te; c

olor

indi

cativ

e of

com

poun

ds.

EnSy

sTr

aini

ng a

vaila

ble.

$21/

sam

ple

for

TNT,

EPA

199

5 (M

etho

d 85

15);

EPA

199

7;St

rate

gic

Dia

gnos

tics,

Inc.

Larg

e w

ork

area

(des

k si

ze);

pow

er s

uppl

yR

ISc

App

licab

le v

ideo

$25/

sam

ple

for

RD

X p

lus

IT 1

995;

Jenk

ins

et a

l. 19

96a,

199

6b;

375

Phea

sant

Run

requ

ired

to c

harg

e H

ach

spec

trom

eter

;on

CR

REL

met

hod

$160

/day

or

$430

/wk

Mar

kos

et a

l. 19

95; M

yers

et a

l. 19

94N

ewto

wn,

PA

189

40po

ssib

le T

NB

inte

rfer

ence

; col

or in

di-

avai

labl

e, a

ddre

ssfo

r la

b st

atio

n. L

ab s

tatio

n(8

00) 5

44-8

881

catio

n of

oth

er c

ompo

unds

; req

uire

sin

text

.co

st =

$19

50.

acet

one

and

deio

nize

d w

ater

; cuv

ette

sm

ust b

e ri

nsed

bet

wee

n an

alys

es.

Nitr

ate

and

nitr

ite in

terf

eren

ces

with

RD

Xki

t can

be

corr

ecte

d us

ing

alum

in-a

-ca

rtri

dges

from

EnS

ys.

USA

CE

Non

e av

aila

ble.

$4/s

ampl

e or

$5/

sam

ple

IT 1

995;

Med

ary

1992

Dr.

Ric

hard

Med

ary

Larg

e w

ork

area

(2 la

rge

desk

s); r

equi

res

if fi

ltere

d, p

lus

$150

0 fo

rU

.S. A

rmy

Cor

ps o

f Eng

.th

e m

ost s

etup

tim

e; p

ossi

ble

TNB

inte

r-H

ach

spec

trom

eter

.60

1 E.

12t

h St

reet

fere

nce;

no

elec

tric

ity o

r re

frig

erat

ion

Kan

sas

City

, MO

641

06re

quir

ed; m

ust a

ssem

ble

mat

eria

ls; g

lass

-(8

16) 4

26-7

882

war

e m

ust b

e ri

nsed

bet

wee

n an

alys

es.

ENV

IRO

L$3

2/sa

mpl

e fo

r TN

TM

anuf

actu

rer

only

ENV

IRO

L, In

c.Se

lf-c

onta

ined

kit;

no

addi

tiona

l sol

vent

1770

Res

earc

h W

ayre

quir

ed.

Suite

160

Nor

th L

ogan

, UT

8434

143

5-75

3-79

46

DTE

CH

2 to

4 h

ours

free

$30/

sam

ple

for

TNT

EPA

199

5 (M

etho

ds 4

050

and

4051

);C

alif

. EPA

Stra

tegi

c D

iagn

ostic

s, In

c.Sm

all w

orki

ng a

rea;

few

set

up r

equi

re-

on-s

ite tr

aini

ng.

or R

DX

plu

s $3

00 fo

rEP

A 1

997;

Haa

s an

d Si

mm

ons

1995

;19

96a

and

375

Phea

sant

Run

men

ts; n

o el

ectr

icity

or

refr

iger

atio

n re

quir

ed;

DTE

CH

TOR

(opt

iona

l)M

arko

s et

al.

1995

; Mye

rs e

t al.

1994

;19

96b

New

tow

n, P

A 1

8940

tem

pera

ture

-dep

ende

nt d

evel

opm

ent t

ime

Tean

ey a

nd H

udak

199

4(8

00) 5

44-8

881

(eff

ect c

an b

e re

duce

d by

cha

ngin

g D

TEC

HTO

Rse

ttin

g); s

igni

fica

nt a

mou

nt o

f pac

king

; rel

ativ

ely

narr

ow r

ange

; no

chec

k on

test

; eas

y to

tran

spor

t or

carr

y; k

its c

an b

e cu

stom

ized

.O

ut-o

f-ra

nge

reru

ns r

equi

re u

se o

f ano

ther

kit.

Idet

ek1

day

free

on-

site

$21/

sam

ple

for

TNT

EPA

199

7; H

aas

and

Sim

mon

s 19

95;

Idet

ek, I

nc.

Larg

e w

ork

area

(des

k); r

equi

res

setu

p ti

me,

Qua

ntix

trai

ning

.pl

us $

5880

for

lab

stat

ion

Mar

kos

et a

l. 19

9512

45 R

eam

woo

d A

ve.

elec

tric

ity, r

efri

gera

tion,

and

dei

oniz

ed w

ater

;or

$50

0/m

onth

ren

tal.

Sunn

yval

e, C

A 9

4089

requ

ires

car

eful

was

hing

of m

icro

wel

ls;

(800

) 433

-835

1re

plic

ate

run

for

each

sam

ple,

aver

age

of th

e tw

o is

the

resu

lt; le

sste

mpe

ratu

re d

epen

dent

. Out

-of-

rang

ere

runs

req

uire

use

of a

noth

er k

it.

9

*Exp

and

ed a

nd m

odif

ied

from

EPA

199

7.

Tab

le 3

. Com

par

ativ

e d

ata

for

sele

ctin

g on

-sit

e an

alyt

ical

met

hod

s fo

r ex

plo

sive

s in

soi

l* (c

onti

nu

ed).

Met

hod/

Trai

ning

Cos

tsC

ompa

riso

ns to

Met

hod

8330

Oth

erD

evel

oper

Kit

avai

labi

lity

(not

incl

udin

g la

bor)

refe

renc

esre

fere

nces

info

rmat

ion

Add

itio

nal c

onsi

dera

tion

s

Envi

ro-

Free

trai

ning

Plat

e: $

17/s

ampl

e pl

us $

4129

Haa

s an

d Si

mm

ons

1995

Cal

if. E

PASt

rate

gic

Dia

gnos

tics,

Inc.

Larg

e w

ork

area

(des

k si

ze);

requ

ires

Gar

dav

aila

ble.

for

equi

p. a

nd s

mal

l sup

plie

s.19

96c

375

Phea

sant

Run

setu

p tim

e, r

efri

gera

tion,

and

pow

er;

Tube

: $20

/sam

ple

plus

$24

09N

ewto

n, P

A 1

8940

acet

one

not s

uppl

ied.

Out

-of-

rang

efo

r eq

uip.

and

sm

all s

uppl

ies.

(800

) 544

-888

1re

runs

req

uire

use

of a

noth

er k

it.

Ohm

icro

n4

hour

s fr

ee o

n-si

te$1

3 to

$20

/sam

ple

plus

$55

00EP

A 1

997;

Haa

s an

d Si

mm

ons

1995

;C

alif

. EPA

Stra

tegi

c D

iagn

ostic

s, In

c.La

rge

wor

k ar

ea (d

esk)

; req

uire

s se

tup

time,

RaP

IDtr

aini

ng.

for

equi

p. (p

urch

ase)

or

$800

Mar

kos

et a

l. 19

95; R

ubio

et a

l. 19

9619

96d

375

Phea

sant

Run

elec

tric

ity, a

nd r

efri

gera

tion;

less

tem

pera

ture

Ass

ayfo

r fi

rst m

onth

, $40

0 ea

chN

ewto

n, P

A 1

8940

depe

nden

t; lo

w d

etec

tion

limit;

all

reag

ents

addi

tiona

l mon

th (r

enta

l).

(800

) 544

-888

1su

pplie

d; r

eage

nts

and

kit n

eed

refr

iger

atio

n.O

ut-o

f-ra

nge

reru

ns r

equi

re u

se o

f ano

ther

kit.

*Exp

and

ed a

nd m

odif

ied

from

EPA

199

7.

10

mixing, and careful subsampling, resulted in anRSD of 11%. Because this field-mixing procedurewas so effective in homogenizing the sample, thesampling and subsampling procedure is presentedhere (Jenkins et al. 1996a). Soil cores (0 to 15 cm inlength and 5.6 cm in diameter) were collected intoplastic resealable bags, and vegetation was re-moved. The sample of dry soil, a mixture of sandand gravel, was placed into 23-cm aluminum piepans and was broken up using gloved hands; largerocks were removed (sieving may work well, too).A second pie pan was used to cover the sample,which was then shaken and swirled vigorously todisperse and homogenize the soil. The sample wasthen coned and quartered, and 5-g subsampleswere removed from each quarter and compositedto form the 20-g sample for analysis. Splits of thesame sample were obtained by remixing the soiland repeating the coning and quartering.

Wilson (1992) studied sample preparation pro-cedures for homogenizing compost prior to analy-sis for explosives. Wilsons (1992) method involvesmacerating air-dried compost using a No. 4 Wileymill followed by sample splitting using a Jones-type riffle splitter. The improved method de-creased the RSD from more than 200% to 3% forTNT analyses.

Sample holding timesand preservation procedures

The EPA-specified holding time for nitroaro-matic compounds in soil is 7 days until extraction,and extracts must be analyzed within the follow-ing 40 days (EPA 1995). The specified sample pres-ervation procedure is cooling to 4C. This crite-rion was based on professional judgment ratherthan experimental data.

Two significant holding time studies have beenconducted on explosives (Maskarinec et al. 1991;Grant et al. 1993, 1995). Based on spiking clean soilswith explosives in acetonitrile, Maskarinec recom-mended the following holding times and condi-tions: for TNT, immediate freezing and 233 daysat 20C; for DNT, 107 days at 4C; for RDX, 107days at 4C; and for HMX, 52 days at 4C. Grantspiked soils with explosives dissolved in water toeliminate any acetonitrile effects and also used afield-contaminated soil. The results on spiked soilsshowed that RDX and HMX are stable for at least8 weeks when refrigerated (2C) or frozen (15C)but that significant degradation of TNT and TNBdegradation can occur within 2 hours withoutpreservation. Freezing provides adequate preser-vation of spiked 2,4-DNT for 8 weeks or longer.

The results on field-contaminated soils did notshow the rapid degradation of TNT and TNB thatwas observed in the spiked soils, and refrigera-tion appeared satisfactory. Presumably, the explo-sives still present in the field soil after many yearsof exposure are less biologically available than inthe spiked soils.

Another study (Bauer et al. 1990) has shownthat explosives in spiked, air-dried soils are stablefor a 62-day period under refrigeration. Data fromthe Grant et al. (1993) study indicate that air dry-ing of field-contaminated soils may not result insignificant losses of explosives contaminants. Ex-plosives in air-dried soils are stable at room tem-perature if they are kept in the dark.

Acetonitrile extracts of soil samples are ex-pected to be stable for at least 6 months under re-frigeration. Acetone extracts also are thought tobe stable if the extracts are stored in the dark un-der refrigeration. (Acetone enhances photo-degradation of explosives.)

Explosion hazards and shipping limitationsThe Department of Defense Explosive Safety

Board approved the two-test protocol (Zero Gapand Deflagration to Detonation Transition tests)in March 1988 for determining the explosive reac-tivity of explosives-contaminated soil. Tests onTNT and RDX in sands with varied water contentshowed that soils with 12% or more explosives aresusceptible to initiation by flame, and soils con-taining more than 15% explosives are subject toinitiation by shock (EPA 1993). Explosives exist asparticles in soil ranging in size from crystals tochunks, which can detonate if initiated. However,if the concentration of explosives is less than 12%,the reaction will not propagate. The water contentof the soil has minimal effects on reactivity. Thetest results apply to total weight percent of sec-ondary explosives such as TNT, RDX, HMX, DNT,TNB, and DNB. The tests do not apply to primaryor initiating explosives such as lead azide, leadstyphnate, and mercury fulminate. As a conser-vative limit, the EPA Regions and the U.S. ArmyEnvironmental Center consider soils containingmore than 10% secondary explosives, on a dryweight basis, to be susceptible to initiation andpropagation (EPA 1993). If chemical analyses in-dicate that a sample is below 10% explosives bydry weight, that sample is considered to benonreactive. In most cases, this eliminates the re-quirement to conduct the expensive two-test re-activity protocol.

In sampling to determine whether an explosion

11

hazard exists, a biased sampling approach mustbe adopted (Sisk 1992). Soils suspected of havinghigh concentrations of explosives should be grab-sampled and analyzed to determine whether thelevel of explosives exceeds 10%. Samples to beshipped for off-site analysis must be subsampledand analyzed on site. Explosives residues are usu-ally concentrated in the top 5 to 10 cm of soil; there-fore, deep samples must not be collected, blended,and analyzed to determine reactivity. Verticalcompositing of surficial soils with high levels ofexplosives with deeper, relatively clean materialprovides a false indication of reactivity. Soils con-taining explosives residues over the 10% level can,using proper precautions, be blended with cleanermaterial to reduce the reactivity hazard and per-mit shipment to an off-site laboratory, but the di-lution factor must be provided with the sample. Ifanalytical results indicate that explosives arepresent at a concentration of 10% or greater, thesamples must be packaged and shipped in accor-dance with applicable Department of Transporta-tion and EPA regulations for reactive hazardouswaste and Class A explosives (AEC 1994) to a labo-ratory capable of handling reactive materials.

In addition to the above information, the ArmyEnvironmental Center requires certain minimumsafety precautions, as summarized below, for fieldsampling work at sites with unknown or greaterthan 10% by weight of secondary explosivescontamination (AEC 1994). An extensive recordssearch and historical documentation review mustbe conducted regarding the contaminated area toidentify the specific explosives present, determinehow the area became contaminated, estimate theextent of contamination, and determine the periodof use. Personnel responsible for taking, packag-ing, shipping, and analyzing samples must beknowledgeable and experienced in working withexplosives. Soil samples must be taken usingnonsparking tools; wetting the sampling area withwater may be necessary. If plastic equipment isused, it must be conductive and grounded. Samplecontainers must be chemically compatible with thespecific explosive, and screw tops are prohibited.Samples are to be field screened for explosives ifpossible. Sufficient soil samples must be collectedto characterize the site in a three-dimensional ba-sis in terms of percent secondary explosives con-tamination with particular attention paid to iden-tifying hot spots, chunks of explosives, layers ofexplosives, discolorations of the soil, etc.

In screening samples for reactivity, it should beremembered that most screening procedures test

for only one analyte or class of analyte. Withoutother supporting knowledge, concluding that asoil is not reactive based upon just one analysiscould be dangerous. For assessing reactivity whenmultiple compounds are present at high levels, theCRREL and EnSys RISc colorimetric methods forTNT and RDX are more appropriate than immu-noassay test kits because colorimetric tests detecta broader range of explosives analytes. Some con-servatism in evaluating potential reactivity usingcolorimetric methods is appropriate. For example,Jenkins et al. (1996c) recommended using a limitof 7% explosives for conservatively estimating thelower limit of potential reactivity. High levels ofexplosives in soils may result in a low bias for on-site methods because of low extraction efficien-cies. Colorimetric tests of chemical compositionare used only to estimate potential reactivity. Thereare no on-site methods available to actually deter-mine explosive reactivity. Explosive reactivity is adetermination made from validated laboratoryanalyses.

PROCEDURES FOR STATISTICALLYCOMPARING ON-SITE ANDREFERENCE ANALYTICAL METHODS

When on-site methods are used, their perfor-mance needs to be evaluated; this is commonlydone by analyzing splits of some soil samples byboth the on-site method and a reference method(commonly Method 8330). The performance of theon-site method is then statistically compared tothe reference method using a variety of methods,depending upon the objective and the character-istics of the data. In most cases, measures of preci-sion and bias are determined. Precision refers tothe agreement among a set of replicate measure-ments and is commonly reported as the RSD (stan-dard deviation divided by the mean and expressedas a percent), the coefficient of variation (standarddeviation divided by the mean), or the relativepercent difference. Bias refers to systematic devia-tion from the true value.

The following discussion of statistical methodsapplies to comparisons of analytical results basedon paired sample data, e.g., soil samples are ana-lyzed by both an on-site method and a referencemethod, or soil extracts are analyzed by two dif-ferent on-site methods. Care must be taken in in-terpreting the result. For example, if splitsubsamples from a soil sample are analyzed byon-site and reference methods, the differences de-

12

tected may be caused by subsampling error(sample was not homogeneous and the splits ac-tually contained different concentrations of explo-sives), or differences in extraction efficiency (shak-ing with acetone versus ultrasonication with aceto-nitrile), rather than the analytical methods, whichmay also produce different results. However, if agroup of acetone extracts are analyzed by twodifferent methods, the subsampling and extractionerrors are minimized and any significant dif-ferences should be from the analytical methods.

Precision and bias tests for measurementsof relatively homogeneous material

When multiple splits of well-homogenized soilsamples are analyzed using different analyticalmethods, statistical procedures described inGrubbs (1973), Blackwood and Bradley (1991), andChristensen and Blackwood (1993) may be usedto compare the precision and bias of the methods.Grubbs (1973) describes a statistical approach ap-propriate for comparing the precision of two meth-ods that takes into account the high correlationbetween the measurements from each method. Anadvantage of Grubbs approach is that it providesunbiased estimates of each methods precision bypartitioning the variance of the measurement re-sults into its component parts (e.g., variancecaused by subsampling and by the analyticalmethod). Blackwood and Bradley (1991) extendGrubbs approach to a simultaneous test for equalprecision and bias of two methods. Christensenand Blackwood (1993) provide similar tests forevaluating more than two methods.

For comparisons involving bias alone, t-tests oranalysis of variance may be performed. For com-paring two methods, paired t-tests are appropri-ate for assessing relative bias (assuming normal-ity of the data, otherwise data transformations toachieve normality must be applied, or nonpara-metric tests used). A paired t-test can be used totest whether the concentration as determined byan on-site method is significantly different fromMethod 8330 or any other reference method. Forcomparing multiple methods, a randomized com-plete block analysis of variance can be used, wherethe methods are the treatments and each set of splitsamples constitutes a block.

These tests are best applied when the concen-trations of explosives are all of approximately thesame magnitude. As the variability in the sampleconcentration increases, the capability of thesetests for detecting differences in precision or biasdecreases. The variability in the true quantities in

the samples is of concern, and high variability insample results caused by poor precision ratherthan variability in the true concentration is wellhandled by these methods.

Precision and bias tests for measurements overlarge value ranges

When the concentrations of explosives cover alarge range of values, regression methods for as-sessing precision and accuracy become appropri-ate. Regression analysis is useful because it allowscharacterization of nonconstant precision and biaseffects and because the analysis used to obtainprediction intervals for new measurements (e.g.,the results of an on-site method) can be used topredict the concentration if the samples were ana-lyzed by a reference method.

In a regression analysis, the less precise on-sitemethod is generally treated as the dependent vari-able and the more precise reference analyticalmethod (e.g., SW-846 Method 8330) as the inde-pendent variable. To the extent that the relation-ship is linear and the slope differs from a value of1.0, there is an indication of a constant relative biasin the on-site method (i.e., the two methods differby a fixed percentage). Bias should be expected ifon-site methods based on wet-weight contaminantlevels are compared to laboratory methods basedon the dry weight of soil samples. Similarly, anintercept value significantly different from zeroindicates a constant absolute bias (i.e., the twomethods differ by a fixed absolute quantity). Theremay, of course, be both fixed and relative bias com-ponents present.

When uncertainty is associated with the con-centration of an explosive as measured by the ref-erence method, standard least squares regressionanalysis can produce misleading results. Standardleast squares regression assumes that the indepen-dent variable values are known exactly as in stan-dard reference material. When the on-site methodresults contain appreciable error compared to thereference method, regression and variability esti-mates are biased. This is known as an errors-in-variables problem.

Because of the errors-in-variables problem, theslope coefficient in the regression of the on-sitedata on the reference data will generally be biasedlow. Hence a standard regression test to determinewhether the slope is significantly different from 1can reject the null hypothesis even when there isin fact no difference in the true bias of the twomethods. A similar argument applies to tests ofthe intercept value being equal to zero.

13

To perform a proper errors-in-variables regres-sion requires consideration of the measurementerrors in both variables. The appropriate methodsare outlined in Mandel (1984). These methods re-quire estimating the ratio of the random error vari-ance for the on-site method to that of the refer-ence analytical method. With split sample data,suitable estimates of these ratios may generally beobtained by using variance estimates from Grubbstest or the related tests mentioned above.

If the variance ratio is not constant over therange under study, more complicated models thanthose analyzed in Mandel (1984) must be em-ployed. Alternatively, transformations of the datamight stabilize the variance ratio. Note that it isthe variance ratio, not the individual variances,that must remain constant. The ratio of variancesfor two methods with nonconstant absolute vari-ances but constant relative variances will still havea constant variance ratio.

Two other caveats about the use of regressiontechniques also are appropriate. First, standardregression methods produce bias regression pa-rameters estimation and may produce misleadinguncertainty intervals. Similarly, the interpretationof R-squared values also is affected. Second, per-forming regressions on data sets in which sampleswith concentrations below the detection limit (forone or both methods) have been eliminated mayalso result in biased regression estimates, no mat-ter which regression analysis method is used.

Comparison to regulatory thresholds,action limits, etc.

When the purpose of sampling is to make adecision based on comparison of results to a spe-cific value such as an action level for cleanup, on-site and reference analytical method results maybe compared simply on the basis of how well thetwo methods agree regarding the decision. Theappropriate statistical tests are based on the bino-mial distribution and include tests of equality ofproportions and chi-square tests comparing thesensitivity and specificity (or false positive andfalse negative rates) of the on-site method relativeto the reference analytical method. Note that anymeasure of consistency between the two methodsis affected by how close the true values in thesamples are to the action level. The closer the truevalues are to the action level, the less the two meth-ods will agree, even if they are of equal accuracy.For example, if the action level is 30 mg/kg andmost samples have levels of above 1000 mg/kg,the agreement between the on-site method and ref-

erence should be very good. If, however, the con-centration in most samples is 5 to 100 mg/kg , thetwo methods will be much more likely to disagree.This must be kept in mind when interpreting re-sults, especially when comparing across differentstudies that may have collected samples at con-siderably different analyte levels.

SUMMARY OF ON-SITE ANALYTICALMETHODS FOR EXPLOSIVES IN SOIL

There is considerable interest in field methodsfor rapidly and economically determining thepresence and concentration of secondary explo-sives in soil. Such procedures allow much greaterflexibility in mapping the extent of contamination,redesigning a sampling plan based on near-real-time data, accruing more detailed characterizationfor a fixed cost, and guiding continuous remedialefforts. Ideally, screening methods provide high-quality data on a near-real-time basis at low costand of sufficient quality to meet all intended uses,including risk assessments and final site clear-ances, without the need for more rigorous proce-dures. Although the currently available screeningprocedures may not be ideal (not capable of pro-viding compound-specific concentrations of mul-tiple compounds simultaneously), they haveproved to be very valuable during the character-ization and remediation of numerous sites. Cur-rently, available field methods that have beenevaluated against standard analytical methodsand demonstrated in the field include colorimet-ric and immunoassay methods (Table 4). Eachmethod has relative advantages and disadvan-tages, so that one method may not be optimal forall applications. To assist in the selection of one ormore screening methods for various users needs,Table 3 (modified and expanded from EPA 1997)provides information on on-site test kits for de-tecting explosives in soil. Selection criteria are dis-cussed in the following sections.

The two types of currently available on-sitemethods, colorimetric and immunoassay, are fun-damentally quite different. Both methods startwith extracting a 2- to 20-g soil sample with 6.5-to 100-mL acetone or methanol for a period of 1 to3 minutes, followed by settling and possibly fil-tration. The basic procedure in the CRREL andEnSys RISc colorimetric methods for TNT is to adda strong base to the acetone extract, producing thered-colored Janowsky anion. Absorbance is thenmeasured at 540 nanometers (nm) using a spec-

14

trophotometer. The TNT concentration is calcu-lated by comparing results to a control sample. TheU.S. Army Corps of Engineers (USACE) andENVIROL procedures use methanol for extraction,and the ENVIROL procedure uses solid phase ex-traction and liquidliquid transfer, prior to reac-tion with base and formation of a colored anion.The RDX test has more steps.

The various immunoassay methods differ con-siderably in their steps; the DTECH method forTNT is the simplest. In the DTECH kit, antibodiesspecific for TNT and closely related compoundsare linked to solid particles. The TNT moleculesin the soil extract are captured by the solid par-ticles and collected on the membrane of a cup as-sembly. A color-developing solution is added tothe cup assembly and the presence (or absence) ofTNT is determined by comparing the solution inthe assembly cup to a color card or by using thesimple field test meter. The color is inversely pro-portional to the concentration of TNT.

Method type, analytes,and EPA method number

The first criteria column in Table 3 lists the typeof soil screening method, the analytes it detects,and the EPA SW-846 draft or proposed methodnumber. A commercially available colorimetric kit,

EnSys RISc, is used to determine TNT and RDX insoil. EnSys RISc is the commercial version of theCRREL method for TNT and RDX. In addition tothe CRREL method, USACE and ENVIROL devel-oped colorimetric methods for TNT. EnSys RIScand CRREL have additional colorimetric methodsthat can also be used to determine nitramines(HMX and NQ), nitrate esters (NC, NG, andPETN), and AP/PA. ENVIROL is developing anRDX colorimetric method that should be availablein spring 1998.

Two companies, Idetek, Inc., and Strategic Di-agnostics, Inc., manufacture commercial enzymelinked immunosorbent assay (ELISA) kits to de-tect TNT in soil. Idetek, Inc., produces the Quantixkit (both a plate and tube method are available),and Strategic Diagnostics, Inc., offers DTECH,EnviroGard, and Ohmicron RaPID Assay. DTECHkits are also available for RDX. Other explosivescompounds can sometimes be detected using im-munoassay kits because of their cross reactivity(see Interferences and Cross Reactivity section).The EnviroGard TNT immunoassay kit was for-merly produced by Millipore Corp.

Detection limits and rangeThe lower detection limits of most methods are

near or below 1 ppm. The detection range of a test

Table 4. Available on-site analytical methods for explosivesin soil.

Analyte(s) Test type Developer/test kit

A. Nitroaromatics Colorimetric CRREL,* Ensys RISc1. TNT Colorimetric CRREL, Ensys RISc

Colorimetric USACE

Colorimetric ENVIROLImmunoassay DTECH

Idetek QuantixOhmicron RaPID AssayEnviroGard

2. TNB Colorimetric CRREL, EnSys RIScImmunoassay Ohmicron RaPID Assay

3. DNT Colorimetric CRREL, EnSys RISc4. Tetryl Colorimetric CRREL

B. Nitramines Colorimetric CRREL, EnSys RISc1 . RDX Colorimetric CRREL, EnSys RISc

Immunoassay DTECH2. HMX Colorimetric CRREL, EnSys RISc3. NQ Colorimetric CRREL

C. Nitrate esters Colorimetric CRREL1. NC Colorimetric CRREL2. NG Colorimetric CRREL3. PETN Colorimetric CRREL

D. AP/PA Colorimetric CRREL

*U.S. Army Cold Regions Research and Engineering Laboratory.U.S. Army Corps of Engineers, Kansas City District.

15

kit can be important: a broad range is generallymore desirable. The importance of the range de-pends on the range of concentrations expected insamples, the ability to estimate the approximateconcentration from the sample extract, the amountof effort required to dilute and rerun a sample, andthe sampling and analytical objective. Some testkits have a range factor (upper limit of range/lower limit) of just one order of magnitude (10),while other methods span two or more orders ofmagnitude (100 to 400). Because explosives con-centrations in soil may range five orders of mag-nitude (100,000), reanalyzing many out-of-rangesamples may be necessary. The DTECH immu-noassay methods require an additional test kit torun each sample dilution.

Other immunoassay methods can run dilutionsin the same analytical run, but one must preparethe dilutions without knowing whether they areneeded. The CRREL, USACE, and EnSys RISc colo-rimetric procedures for TNT provide sufficientreagent to run several dilutions at no additionalcost. For the EnSys RISc TNT kit, the color devel-oped can simply be diluted and reread in the spec-trophotometer. The procedures that the test meth-ods use for samples requiring dilution should beevaluated as part of the site-specific data qualityobjectives.

The detection range of a kit becomes much lessrelevant when the objective is to determinewhether a soil is above or below a single actionlimit; the same dilution can be used for all samples.In some cases, changing the range of a kit may bedesirable to facilitate decision-making. If a methodhas a range of 1 to 10 ppm and the contaminationlevel of concern is 30 ppm, diluting all samples(using acetone or methanol or as directed by theinstructions) by a factor of five would change thetest kit range to 5 to 50 ppm and permit decisionsto be made without additional dilutions.

Cleanup levels for explosives in soil vary con-siderably depending upon the site conditions,compounds present and their relative concentra-tion, threats to groundwater, results of risk assess-ments, remedial technology, etc. (EPA 1993). Basedon a review of data from many sites, Craig et al.(1995) suggested preliminary remediation goals of30 ppm for TNT, 50 ppm for RDX, and 5 ppm for2,4-DNT and 2,6-DNT.

Type of resultsThe type of results provided by the various

screening methods are quantitative or semi-quan-titative. The CRREL (TNT, RDX, and AP/PA),

EnSys RISc, ENVIROL, USACE, Idetek Quantix,Ohmicron RaPID Assay, and EnviroGard (Plate)kits are quantitative methods, providing a numeri-cal value. The CRREL 2,4-DNT method is consid-ered semiquantitative and provides a somewhatless accurate numerical value. The DTECH andEnviroGard (Tube) test kits are semi-quantitative (concentration range), and indicatethat the level of an analyte is within one of severalranges. For example, the DTECH TNT soil kit,without dilution, indicates a concentration withinone of the following ranges: less than 0.5, 0.5 to1.5, 1.5 to 2.5, 2.5 to 4.5, 4.5 to 6.0, and greater than6.0 ppm.

Samples per batchSeveral of the available test kits are designed to

run batches of samples or single samples or both.Using a test kit designed for analyzing a largebatch to analyze one or two samples may not bevery cost-effective or efficient. In most cases,samples may easily be batched for extraction andprocessed simultaneously.

Sample sizeThe size of the soil sample extracted contrib-

utes to the representativeness of a sample. Explo-sives residues in soil are quite heterogeneouslydistributed (Jenkins et al. 1996a, b; 1997), and asthe subsample size actually extracted decreases,heterogeneity increases. Although sample prepa-ration procedures such as drying, mixing, sieving,and splitting can reduce within-sample heteroge-neity, such procedures can be time-consuming.Based on work by Jenkins et al. (1996b, 1997), fieldcompositing and homogenization greatly improvesample representativeness. The commercial testkits use 2 to 10 g of soil, while the CRREL meth-ods extract 20 g of soil to improve the representa-tiveness of the results. For some test kits, it is pos-sible to extract a larger sample using solvent andglassware not provided in the kit, and then usingthe required volume of extract for the analyticalsteps. The smaller the sample size, the more im-portant is the homogenization of the sample be-fore subsampling, although this procedure alonecannot completely compensate for loss of repre-sentativeness due to smaller samples.

Sample preparation and extractionSoil extractions procedures for most of the

screening methods are similar, shaking 2 to 20 gof soil in 6.5 to 100 mL of solvent (acetone or metha-nol) for 1 to 3 minutes. This may be followed by

16

settling or filtration or both. One test kit(EnviroGard) specifies air drying; for the EnSysRISc colorimetric test kits, drying to less than 10%moisture is optional. For the CRREL methods,samples must contain 2 to 3% water by weight;therefore, water must be added to the extract forvery dry soils or incomplete color developmentwill occur, resulting in a false negative.

The solvent extraction times of 1 to 3 minutesused in on-site methods may result in incompleteextraction of explosives compared with the 18-hour ultrasonic bath extraction step used in EPAMethod 8330. The percent of explosives extractedis sample-specific but is generally higher for highconcentration samples, higher for sandy soils,lower for clayey soils, and lower if 1-minute ex-tractions are used relative to 3-minute extractions.For many soils, a 3-minute extraction time is ad-equate; ratios of 3-minute versus 18-hour extrac-tions of TNT and RDX using acetone or methanolrange from 66 to 109% as reported by Jenkins etal. (1996c). Jenkins recommends at least a 3-minutesolvent extraction procedure for explosives. Whenpinpointing concentrations, a short kinetic studyshould be conducted of the specific soils encoun-tered at a site (Jenkins et al. 1996c). The kineticstudy would involve analyzing an aliquot of ex-tract after 3 minutes of shaking, and again after10, 30, and 60 minutes of standing followed by an-other 3 minutes of shaking. If the concentration ofexplosives increased significantly with the longerextraction time, a longer extraction period isneeded. Jenkins et al. (1996a) found that 30-minuteextraction times worked well for clay soils at theVolunteer Army Ammunition Plant, Chattanooga,Tennessee. Where multiple analytes are of inter-est in each sample, a common extract may be usedfor both the colorimetric and immunoassay testmethods.

Analysis timeThe analysis time or throughput for the colori-

metric and immunoassay procedures ranges from3 to 11 minutes per sample for batch runs. TheEnviroGard kits specify air drying of samples(which would add considerable time), and dry-ing is optional with the EnSys RISc colorimetrickits. Cragin et al. (1985) investigated various pro-cedures for drying explosives-contaminated soils,including air, oven, desiccator, and microwavedrying. Air and desiccator drying appear to resultin only minor losses of explosives. Oven dryingof highly contaminated soil (15% TNT) at 105Cfor an unspecified period resulted in a 25% loss of

TNT; however, oven drying of less-contaminatedsamples for only 1 hour resulted in little loss ofTNT, and 30 minutes of drying was estimated tobe sufficient for analytical purposes. Microwavedrying was not recommended because of spottyheating and drying. In addition, microwave dry-ing should not be used because it may present asafety hazard and such drying degrades thermallyunstable explosives in the soil. The effective pro-duction rate depends on the number of reruns re-quired because a sample is out of the detectionrange.

Interferences and cross-reactivityOne of the major differences among the field

methods is interference for colorimetric methodsand cross-reactivity for immunoassay methods.The colorimetric methods for TNT and RDX arebroadly class sensitive; that is, they are able to de-tect the presence of the target analyte but also re-spond to many other similar compounds(nitroaromatics and nitramines/nitrate esters, re-spectively). For colorimetric methods, interferenceis defined as the positive response of the methodto secondary target analytes or co-contaminantssimilar to the primary target analyte. TheENVIROL colorimetr