DRAFT
VOA Compositing Procedures
      September 1, 1994
    William A. Telliard, Chief
    Analytical Methods Staff
  Engineering and Analysis Division
  Office of Science and Technology
    U S. EPA Office of Water

-------
                                                                      VOA Compositing Procedures
SUMMARY

      This paper gives results of a VOA compositing study conducted by the U S Environmental
Protection Agency (EPA) in early 1994  In these studies, individual grab samples of real-world
effluent were collected over the course of a day  These samples were analyzed spiked or unspiked,
composited and individually, by isotope  dilution GC/MS, using Revision C of EPA Method 1624.
Pollutants spiked were the volatile organic  GC/MS fraction of the priority pollutants plus
additional compounds routinely tested for in EPA's industrial surveys. The objective was to
compare the mathematical average of the  results from analysis of the individual grab samples with
the result of the analysis of the composite sample to determine if bias occurred in the compositing
process.

      These tests showed that the mathematical average of the results of analysis of the individual
grab samples was a few percent higher, on average, than results of the analysis of composite
samples.  The cause of these slight differences is not known, and these differences are not significant
for practical purposes and would not be discernable by non-isotope dilution GC/MS methods.
BACKGROUND

Volatile Organic Compounds (VOCs)

     The Federal Water Pollution Control Act of 1972 (PL 92-500) required the Environmental
Protection Agency to control the discharge of toxic pollutants to the nation's waters.  The toxic
pollutants regulated were listed in the act (40 CFR 401 15) as 65 compounds and compound classes.
This list was refined into an initial list of 129 "priority pollutants"  and then a final priority
pollutant list of 126 individual compounds (Reference 1).

     For determination of the priority pollutants, EPA had to separate the list of 126 into groups
based on the analytical technology that could be used to measure the pollutants.  Organic pollutants
which could be determined by gas chromatography combined with mass spectrometry (GC/MS),
pollutants were further categorized into the volatile, acid, and base/neutral fractions

     The volatile fraction, also called the  "purgeable" fraction, contains those compounds that boil
below approximately 130°C and that are capable of being purged from water using a flowing gas
stream (Reference 2).   Analysis of this fraction is termed a "volatile organic analysis" (VOA) and
the compounds in this fraction are termed "volatile  organic compounds" (VOCs). Determination of
VOCs in the VOA fraction of the list of priority pollutants is the  subject of this study.
September 1, 1994

-------
 VGA Compositing Promiinef
 Pollutant Lists

      The list of VOCs in this study is given in Table 1   This ublc also lists the stable, isotopically
 labeled analog that was used for isotope dilution quantitation, whether a given analyte is a Priority
 Pollutant or other pollutant associated with the 1976 Consent Decree (Reference 1); and the
 Chemical Abstracts Service Registry Number for the pollutant and labeled analog, where available

      The list of VOCs m Table 1 is separated  into two groups  The first group contains VOCs
 that are determined by calibration of the GC/MS using authentic standards; the second group
 contains VOCs determined by reverse search for a spectral match during a given retention time
 window based on mass spectral and  retention time data contained in a mass spectral library and
 given in the method.  If a match  is found, the compound  is quantitated based on a response factor
 also given in the  method.  Although results produced by the reverse-search technique are not as
 precise  and accurate as results produced using calibration, the technique is useful for screening and
 approximation of the concentrations of VOCs in the reverse-search group and is more accurate in
 identifying compounds than a forward library search in which only the mass spectrum is tested
 against  a large mass spectral file

      In addition to the priority pollutant list of VOCs, EPA has established other lists of  regulated
 VOCs under the  Safe Drinking Water Act (SWDA) and amendments,  the Resource Conservation
 and Recovery Act (RCRA)  and amendments, and the Comprehensive  Environmental Response,
 Compensation, and Liability Act (CERCLA; Superfund) and amendments.  Although these lists are
 not identical  to the list of VOCs  in Table 1, most of  the compounds on these other lists are
 included in Table 1, and therefore the results of this study should be applicable to the VOCs on
 these other lists.
"Gases" and "Water-Soluble" Compounds

      Two groups of compounds present unique analytical problems in the determination of VOCs.
These groups are the "gases" and "water-soluble compounds" The priority pollutant "gases" consist
of chloromethane, bromomethane, chloroethane, and vinyl chloride, but any non-water-soluble
compound that boils below approximately 15° C will be lost from aqueous solutions easily. These
losses make the analysis somewhat more variable than for compounds that are not losr as readily.
Conversely, the water-soluble compounds present analytical  problems because they are not readily
purged from the water. In this study, the water-soluble priority pollutants tested were acrolein,
acrylonitrile, and 2-chloroethylvmyl ether  Other non-priority pollutant water-soluble compounds
tested were acetone,  2-butanone (MEK), p-dioxane, and diethyl ether. As will be seen from the
results presented below, these compounds produce results that are more highly variable than for the
non-gas priority pollutant VOCs.
                                                                              September 1, 1994

-------
                                                                               VOA  Compositing Procedures
Table 1
Volatile Organic Compounds Analyzed
Compounds Calibrated bv Internal Standard
Labeled Compound
Compound
Acetone
Acrolem
Acrylonitnle
Benzene
Bromodichloromethane
Bromoform
Bromomethane
Carbon tetrachlonde
Chlorobenzene
Chloroethane
2-Chloroethylvmyl ether
Chloroform
Chloromcthane
Dibromochloromethane
1,1-Dichloroethane
1 ,2-Dichloroethane
1,1-Dichloroethene
trans-l,2-Dichloroethene
1 ,2-Dichloropropane
trans-l,3-Dichloropropene
Diethyl ether
/7-Dioxane
Ethylbenzene
Methylene chloride
Methyl ethyl ketone (MEK)
1,1,2,2-Tetrachloroethane
Tetrachloroethene
Toluene
1 , 1, 1-Tnchloroethane
1 , 1 ,2-Trichloroethane
Tnchloroethene
Vinyl chloride
CAS Registry
67-64-1
107-02-8
107-13-1
71-43-2
75-27-4
75-25-2
74-83-9
56-23-5
108-90-7
75-00-3
110-75-8
67-66-3
74-87-3
124-48-1
75-34-3
107-06-2
75-35-4
156-60-5
78-87-5
10061-02-6
60-29-7
123-91-1
100-41-4
75-09-2
78-93-3
79-34-5
127-18-4
108-88-3
71-55-6
79-00-5
79-01-6
75-01-4
Analog
d6
d4
Q.
Q-,
I3C
I3C
dj
13C
QP
Q-

13C
d3
13C
dj
Q .
Q"!
Q-*
Q/
d4
d,o
dg
d,o
C!T
Q-
d,
13c2
dg
d3
13c2
I3c2
d3
CAS Registry
666-52-4
33984-05-3
53807-26-4
1076-43-3
93952-10-4
72802-81-4
1111-88-2
32488-50-9
3114-55-4
19199-91-8

31717-44-9
1111-89-3
93951-99-6
56912-77-7
17070-07-0
22280-73-5
42366-47-2
93952-08-0
93951-86-1
2679-89-2
17647-74-4
25837-05-2
1665-00-5
53389-26-7
33685-54-C
32488-49-6
2037-26-5
2747-58-2
93952-09-1
93952-00-2
6745-35-3
Priority
Pollutant
N
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
Y
N
N
Y
Y
N
Y
Y
Y
Y
Y
Y
Y
September 1, 1994

-------
 I 'O/-1 Conipositnig Procedures
 Compounds Determined bv Reverse Search
                                                                        Priority
 Compound                                 CAS Registry       •       Pollutant

 Carbon disulfide                                   75-13-0                  N
 cis-l,3-Dichloropropene                           10061-01-5                  Y
 2-Hexanone                                      591-78-6                  N
 4-Methyl-2-pentanone                              108-10-1                  N
 Tnchlorofluoromethane                             75-69-4                  N
 Vinyl acetate                                     108-05-4                  N
 w-Xylene                                         108-38-3                  N
o- and />-Xylene                                                            N


Control of Discharges

      The Engineering and Analysis Division (EAD), within the Office of Science and Technology
in EPA's Office of Water, is responsible for promulgating regulations controlling the discharge of
pollutants to surface waters of the U.S   EAD conducts surveys within various categories and sub-
categories of the regulated industry to establish the best pollutant control strategies (Reference 3).
In these surveys, EAD frequently performs sampling and analyses of wastewaters to determine the
presence and concentration of pollutants.  Although these studies focus primarily on the 126
priority pollutants (40 CFR 423 Appendix A) and the five Conventional Pollutants (40 CFR 401.16),
other "non-conventional" pollutants may also be surveyed and subsequently regulated.

      In conducting these surveys, EPA collects aqueous samples from in and around wastewater
treatment plants.  Unless treatment system characterization dictates otherwise, VOA samples are
composited to effect a savings over the costs  of analysis of individual grab samples. Normally, four
individual grab samples are collected at approximately equal time intervals over  the course of a
calendar day.  These samples are shipped under wet ice to the testing laboratory and composited in
the laboratory.  Results of these analyses are then used, in part, to develop, propose, and
promulgate effluent guidelines and standards under the appropriate industrial category at 40 CFR
Parts 403 - 499.
THEORETICAL CONSIDERATIONS AND PRIOR WORK

      VOA compositing is used extensively in EPA's data-gathering for regulation development and
is used for compliance monitoring under EPA rules  Technical literature is replete with theoretical
discussions of the effects of compositing. Book chapters on the subject by Gilbert (Reference 4) and
by Garner et al. (Reference 5) provide comprehensive discussions of the concepts behind composite
                                                                                September 1, 1994

-------
                                                                      VOA Compositing Procedures
sampling and provide cxiensi\e bibliographies referencing the iechnic.il literature on sample
compositing and statistical treatments of the compositing process

      Although the technical literature is replete with theoretical discussions of compositing, it is
remarkably silent in reports of data gathering to verify the theoretical discussions  A search of the
online databases revealed only one technical paper that presents actual results of a VOA
compositing study (Reference 6)


Variability of Individual and Composite Measurements

      Any  empirical measurement process has inherent variability, and the measurement of each
analyte in each analysis is accompanied by an analytical error.  This error is normally characterized
by replicate measurements and is expressed as the standard deviation of the concentration of these
measurements or is normalized to the concentration as the "relative standard deviation" or
"coefficient of variation."  For example, the concentration of chloroform may be determined by
purge-and-trap GC/MS with  a relative standard deviation of 10 percent

      The effect of the measurement error on the average result for a composite sample and for the
average of individual grab samples can be understood most easily if it is  assumed  that the
concentration of a pollutant is identical in all of the individual grab samples   Averaging the results
of analyses of four individual samples is equivalent to determining the concentration four separate
times.  Because the measurement error is inversely proportional to the square root of the number of
measurements, the measurement error associated with the four individual grab samples will be one-
half of the  error associated with any individual  measurement.  And because the determination of
the composite sample is an individual measurement, the error associated with the average of the
measurement of the four individual  grab samples will  be one-half of the  error associated with the
measurement of the composite sample.

      Therefore, if costs are not an important consideration, the most precise and accurate results
will be produced if the individual grab samples  are analyzed and the results averaged. Of course,
similar accuracy could  be  achieved if the compositing  process were replicated four times and the
four composites analyzed, assuming that little or no error occurred in the compositing procedure.
Pragmatic considerations of cost and time  frequently outweigh the ability to measure the grab
samples individually, so discussions  of error become moot, and the error associated with the
composite sample becomes the only measurement error that must be considered.

      This  study did not attempt to quantify or verify that the measurement error associated with
the average of the four individual composite samples was indeed one-half the measurement error
associated with the single composite sample. The objective was to compare the mathematical
average of the results from analysis of the  individual grab samples with the result of the analysis of
the composite sample to determine if  bias  occurred in the compositing process.  However, the
measurement error plays a role in this comparison. As explained  in the section on statistical
analyses below, differences in results between mathematically averaged individual grab samples and
physically composited samples are dependent on the measurement error. Smaller differences
September 1, 1994

-------
 VOA Compositing Procedures
 become statistically significant as the measurement error decreases  As will be seen from the results
 of this study, measurement errors on the order of a few percent allow discernment of differences on
 the order of a few  percent.  This high precision is attributable primarily to the isotope dilution
 quantitation technique, but also to the quality of work performed by the testing laboratory
 TYPES OF COMPOSITING

 Time Compositing

      Time compositing is the most common type of sample compositing.  Samples are collected
 from a fixed sampling point over some fixed period of time, usually one calendar day Samples can
 be collected as discrete grab samples at intervals throughout the fixed time period or continuously
 over the period.

      Transients

      The objective of sampling over time, whether the sampling is grab or continuous, is to
 attempt to capture transient compounds in the waste stream.  Capture of transients in the waste
 stream requires a knowledge of the flow characteristics of each individual stream. These
 characteristics are system volumes, flow rates, and the nature of the transient wave. If the objective
 is to capture the concentration maximum, the ideal scheme is to collect a grab sample at the apex of
 the wave   Unfortunately, this scheme  is frequently impractical. The next  best scheme is  to collect
 a sample at frequent enough intervals to assure that some fraction of the transient will be  captured.
 Although use of a continuous compositor will assure capture of the transient, the transient may be
 diluted by  the stream before and after the  passage of the wave   Therefore, if monitoring of
 transients in a waste stream is necessary to characterize treatment system operation, samples should
 be collected over the wave to model the wave.  After the wave characteristics are known,  the
 intervals for subsequent sampling can be determined.

    Treatment System Detention Times

      For treated effluents, a common  mistake made by personnel unfamiliar with treatment system
operation is to require grab samples at  intervals more frequent than the detention time of the
treatment system.  For example, if the  treatment system has a detention time of 6 hours, sampling
the effluent from the system more frequently than every few hours is unnecessary, particularly if
the samples are analyzed individually.


Spatial Compositing

      Samples from different sampling  points can be composited in an effort to save analysis costs.
If an analyte  is present in a composited sample, each sampling point can then be sampled
individually to determine the point or  points contributing to the level of the analyte  in the sample.
Spatial compositing of up to five streams is allowed at the discretion of the States under the EPA
                                                                              September 1, 1994

-------
                                                                      VOA Compositing Procedures
 drinking water regulations to reduce the total number of samples that small drinking water
 treatment system operators must analyse (40 CFR 141 24[f][14]).  However, the analytical system
 must be capable of detecting one-fifth of the maximum contaminant level (MCL) required for an
 individual sample  This requirement can usually be met by comp'osiung five 5-mL samples and
 purging a 25-mL composite, as suggested in the CFR.

 Flow or Volume Compositing

      As the name implies, flow  or volume compositing involves proportioning the sample
 according to the flow rate or volume of the stream being sampled.  The most common use of flow
 compositing is in EPA's stormwater  rules (40 CFR 122.21[g][7]; Reference 7). These rules require
 that the composited sample proportionately represent the runoff that occurs in a stormwater event.
 However, because it is impossible to  know beforehand the total volume that will be discharged
 during the event, individual grab samples must be collected at various time intervals throughout the
 event, and varying volumes from these individual grab samples must be composited to reflect the
 flow during the entire stormwater event.  The details of stormwater sampling and analysis, along
 with an example of the compositing associated with a stormwater discharge event, have been
 explained by Stanko (Reference 8).


 Problems Unique to  VOA Compositing

     The nature of volatiles, and particularly of the volatile gases, makes these analytes particularly
 susceptible to loss during any manipulation, including collection and compositing.

    Headspace During Sampling

     Losses of analytes, particularly the gases, to the headspace of a container have been
 documented by Cline and Severm (Reference 6). Therefore, it is imperative that headspace be
 eliminated during sampling and sample shipment.  In this study, the loss of volatiles was not critical
 because the objective was to compare the results of analyses of individual grab samples with the
 results of analysis of a composited sample.  So long as the loss of VOCs from the individual  grabs
 and from the grabs that  feed the  composite are the same, there is no consequence to this loss.

    Losses During Compositing

     None of the compositing procedures requires that compositing of grab samples be performed
with zero headspace, and such a system is difficult to envision.  Because such a system does not
exist, exposure of the  sample to the atmosphere can result in analyte losses.  The amount of loss can
be minimized by  keeping the sample  cold and minimizing the exposure time. In this study,  all
compositing was performed rapidly with the VOA vials chilled to 0 - 4° C.
September 1, 1994

-------
 VOA Compositing Procedures
 COMPOSIIING PROCEDURES

 Definitions

 Sample  The water collected in a sample jug from .1 speufic location at a specific time

 IndiMdual grab sample  An aliquot poured from the sample jug

 Duplicate grab sample. A second aliquot poured from the sample jug.

 Replicate grab sample.  Any aliquot poured from the sample jug.

 Composite sample.  The combination of four grab samples collected at different times on the same
                    calendar day.

 Mathematical composite:  The mathematical average of the results of four individual grab samples.


 Manual Compositing

      Two types of manual compositing procedures were tested m this study flask compositing
 and purge device compositing.  Each of these procedures is described below.  A third procedure,
 syringe compositing, is also described below but was not tested because of resource limitations.

      Flask Compositing (44 FR 69555)

      In the flask compositing procedure, a 300- to 500-mL round-bottom flask is immersed in an
 ice bath.  The individual VOA grab samples are maintained at 0 - 4°C and are slowly poured into
the round-bottom flask. The flask is swirled slowly to mix the individual grab samples.  After
 mixing, multiple aliquots of the composited sample are poured into  VOA vials and sealed for
subsequent analysis, or an aliquot can be poured into a syringe for immediate analysis.

      Purge Device Compositing (40 CFR 141.24rfiri4irv1)

      Equal volumes of individual grab samples are added to a  purge device to a total volume of 5
 or 25 mL  The sample is then analyzed.

      Svnnge Compositing (40 CFR 141.24rfiri4irivn

      In the syringe compositing procedure, equal volumes of individual grab samples at a
 temperature of 0 - 4° C are added to a 25-mL syringe while maintaining zero headspace in the
syringe.  Either the total volume in the syringe or an aliquot is subsequently analyzed.  The
disadvantage of this technique is that the individual samples must be poured carefully to attempt to
 achieve equal volumes  of each.  An  alternative procedure uses multiple 5-mL syringes that are filled
 with the individual grab samples and then injected sequentially into  the 25-mL syringe.


8                                                                             September 1, 1994

-------
                                                                      VOA Compositing Procedures
 Automated Collection and Compositing

      Two types of automated equipment are available for sample "collection and/or compositing
 These are (1) automated grab collection and (2) automated continuous collection/compositing.
 These devices are described below   Neither of these devices were tested in this study.

      Automated Grab Collection

      Automated grab collection can be accomplished using devices such as the ISCO Corp Model
 6000 automatic VOC sampler.  With this system, a small  bladder pump forces sample into a 40-mL
 VOA vial after rinsing the vial with three vial volumes to eliminate headspace.  Up to 25 samples
 can be collected at a minimum of 5-mmute and a maximum of 10-hour intervals  Samples are
 maintained at 0 - 4° C during collection.

      Automated Continuous Collection/compositing

      An automated system such as the Associated Design and Manufacturing Corp (ADM)
 automated continuous compositing system can be used to collect samples over a given sampling
 period  Samples are maintained at  0 - 4° C during collection.
SITE SELECTION AND SAMPLE COLLECTION

Sampling Sites

      Samples were collected from seven "real-world" sites, which are described in Table 2.
Information about each site was recorded in an on-site  log and included the EPA sample number,
collection date and time, descriptions of sample and sampling location, sample pH and temperature,
and preservatives used, if any.

      Sample sites were selected specifically in an attempt to find effluents that contained volatile
organics.  However, volatile organics were seldom found.
September 1, 1994

-------
 VOA Compositing Procedures
                                            Table 2
                                Description of Sites and Samples
Episode
4559
4561
4563
4573
4575
4593
4595
Industrial Category
Organic Chemicals
Organic Chemicals
Drum Reconditioning
Shore Reception
Transportation
Transportation
MSW Landfill
Sampling Point
Primary Effluent
Primary Effluent
Scrubber Water
Oily Wastewater
Separator Effluent
Separator Effluent
Leachate
PH
88
73
86
5.6
6.0
6.8
6.8
Sample Collection, Shipment, and Storage

      All samples collected at industrial or municipal sites were preserved to pH  < 2, refrigerated,
and shipped to the  laboratory under wet ice via overnight courier.  If free chlorine was present in
the sample, the sample was additionally preserved with sodium thiosulfate. Samples were stored in
the laboratory at 0  - 4° C from the time of collection until analysis. All analyses were performed
within the 14-day holding time.

      Samples were collected by passage of a portion of the flowing sample stream through a coil of
pre-cleaned polytetrafluoro-ethylene (PTFE) tubing that was immersed in a commercial picnic
cooler filled with ice   This practice reduced the temperature of the effluent to 0 - 4° C, thus
reducing the volatility of the VOCs.  The stream from the PTFE tubing was collected in a
refrigerated one-liter glass jug.

      Samples were preserved to pH < 2 in this jug and free chlorine was removed as required using
sodium thiosulfate  After preservation, samples were allocated from the 1-L jug into 40-mL VOA
vials. Eight vials were filled  from the  common jug,  thus assuring that each replicate VOA vial in
the set contained the same pollutants and concentrations as the others.  The vials were filled to
overflowing, then capped with a PTFE-faced silicone rubber septum.  After capping, each VOA vial
was inverted and inspected for an  air bubble.  If a bubble was present, the vial  was uncapped and
refilled to overflowing and re-capped until completely filled without an .ur bubble  Each vial was
10
September 1, 1994

-------
                                                                    VOA Compositing Procedures
assigned a unique sample number. Sampling times were at approximately 9 a.m., noon, 3 p.m., and
6 p.m.
LABORATORY TESTING

Sample Spiking

     All spiking solutions were prepared in the laboratory and all spiking was performed in the
laboratory.

     Seven field samples were spiked in the laboratory and analyzed in Phase II. Of these seven
samples, three were flask composited and four were purge device composited. Schematic diagrams
of the flask and purge device compositing procedures are shown in Figures 1 and 2, respectively.
For the samples that were flask composited, the grab  sample from the first sample time was
analyzed unspiked  to determine the background concentrations of VOCs present.  For the samples
that were purge device composited, the grab samples  from all four sample times were analyzed
unspiked to determine the background concentrations present.  The reason for this testing was to
determine the constancy of the background throughout the sampling period.
Tl
-™
DD
1
1
1
l
T2
B
mm
i
i
i
i
T;)
I
> <
mm
i
i
i
i
Tt
B
» 4
mm
\
\
i

                                          1
                                                         Key
                                                         llitspiked
                                                         80 m/lSpike
                             Figure 1.  Flask Compositing Scheme
 September 1, 1994


-------
 VOA Compositing Procedures
      Individual grab sample VOA vials from the four sample times were spiked at concentrations
of 20, 40, 80, and 40 /tg/L, respectively, to produce an average concentration of 45 Mg/L.  An
aliquot from these spiked VOA vials was analyzed and another aliquot was used for compositing,
thus assuring that the spike levels were identical for analyses of theJndividual and  composited
samples.
Tl T2
D U
>
D DD D HE
l
i
1
i
frf' ^T
j - "
rvi ?T
i
i
, '
TJ
•
t » «
a •• D
i
i
i
i

if?
IF
ii
                                                                it
                                                                   D
                                                                   T
                                                            llwpiked
                                                            40 nj!/L Spike
                         Figure 2. Purge Device Compositing Scheme
Analyses
     All laboratory analyses were performed at the laboratories of Pacific Analytical, Inc., in
Carlsbad, California.  A single laboratory was chosen for this work because EPA desired that
analytical variability be minimized in order to increase the probability of detecting differences
between grab and among compositing procedures.

    Calibration

         All analyses were performed by isotope dilution GC/MS using Revision C of EPA
Method 1624. Revision C is an updated version of the method promulgated for use in water
programs (40 CFR 136, Appendix A).   Revision C includes a "reverse-search" technique for
identification and quantitation of pollutants in addition to the priority pollutants. In the
promulgated version and in  Revision C of Method 1624, the Priority Pollutants and certain
\2
September 1, 1994

-------
                                                                      VOA Compositing Procedures
 additional compounds are determined using J 5-pomt calibration for quanntacion  Nominal
 calibration points are 10, 20, 50, 100, and 200 Mg/L  In addition, the list of "reverse search"
 compounds is determined from relative retention time data and response factors given in the
 method. Although the reverse-search quantitation procedure is not as accurate as the 5-pomt
 calibration,  it serves screening purposes to provide an estimate of the presence and concentration of
 pollutants over and above the priority  pollutants that may be present m environmental  samples.

      In this study, the method of quantitation was examined in relation to recovery of the VOCs
 for which the instrument was calibrated. The calibration procedures in Method  1624 require  use of
 an average relative response or a calibration curve for isotope dilution calibration based on the five
 calibration points.  However, because the analytes were spiked at known concentrations, it is  also
 possible to use the calibration point closest to each known concentration for calibration. This
 technique of using the closest calibration point  was used for calculation of all concentrations in this
 study and reduced the analytical error to less than that obtained using the average of the five
 calibration points or a calibration curve. It must be emphasized that this practice of using the
 closest calibration point should be employed only when the concentration of a pollutant in a
 sample is known to be close to the calibration point.  For samples containing unknown
concentrations, the most accurate concentration will be found using the entire 5-pomt calibration
curve

      Data Processing and Reporting

      Data were  received by the EPA Sample Control Center in the form of quantitation reports on
diskette. These data included quality control (QC) data for each analysis.  The QC data included
 recoveries for each labeled compound spiked. The QC data were tested against the QC  acceptance
criteria in the method using a modified version  of QA Formaster™ supplied by Thermo-Finnigan
Corp. Non-compliant data were resolved with  the laboratory
STATISTICAL ANALYSES AND RESULTS

Analytes Tested

     As mentioned above, data were evaluated with respect to QC requirements  Three analytes
were dropped  from further analysis due to poor quantitation.  1,1,1-tnchloroethane,
2-chloroethylvinyl ether, and trans-1-2-dichloroethcne  All other analytes met QC requirements


Background Subtraction

     The background level determined from the single, unspiked sample in each of the Flask-
composited episodes was subtracted from the result of all grab and composite samples for that
episode.  For each grab sample in the purge device composited episode, the background levels from
the sample collected at the same point and time was subtracted from the analytical result.  For each
September 1, 1994                                                                              13

-------
 VOA Compositing Procedures
 composite sample, the results ol the four individu.il backgrounds were ,i\cragcd, and the resulting
 value was subtracted from the composite results
 Outlier Screening

      For each analyte in each sample, the percent recovery relative to the spike amount was
 determined  A robust outlier screen was developed according to the following formulae

                   /./. - Ql  -  I 5 v Qrange
                   111. = Q3  -  I 5 x Orange

                           LL = Lower limn
                           UL =  Upper limn
                   wheie   Ql = 25th percenlile, value  below which fall 25% of data
                           Q3 = 75lh percenlile, value  below which fall 75% of data
                           Qrange = Q3 -  Ql
Any result where the percent recovery was below the lower limit, or above the upper limit, was
removed from further statistical analyses.
Statistical Analyses

      For each analyte in each episode, the percent recoveries in the four grab samples were
averaged, as were those of the two physical composites. The median recovery across all analytes
and episodes was calculated.  In addition, the ratio of mathematical composite recovery to physical
composite recovery was calculated for each analyte in each episode, according to the formula

                          „    _  Mean mathematical composite recovery
                                    Mean physical composite  recovery

A two-tailed t-test was performed to determine if this ratio was significantly different  from 1.0, at
the 5% level.  In addition, a two-tailed t-test using Satterthwaite's correction for unequal variances
was performed to determine if there were any differences between recoveries  in samples composited
in a flask and recoveries in samples composited in the purge device
Results

      The pollutants detected in the real-world samples were mainly the water-soluble compounds,
resulting in high analytical error and precluding discernment of differences between the
mathematically averaged results from analysis of the individual grab samples and the result from
analysis of the physically composited sample
14                                                                              September 1, 1994

-------
                                                                       VOA Compositing Procedures
      Comparison of Grab and Composite Results

          Results for these tests showed that, for the analytes for which background subtraction was
 not required and from which the gases and water-soluble compourfds were excluded, the
 mathematical average of the individual grab samples was 10 percent higher than the average of the
 two composite samples for the flask compositing procedure, and 5 percent higher for the purge
 device compositing procedure. The median recovery for the non-gas/non-water-soluble, non-
 indigenous analytes was 105.6 for the mathematical average of the grab samples and 100 3 for the
 composite samples using the flask compositing procedure and  96.9 and 87 percent, respectively, for
 the purge device compositing procedure.

      The results of the t-test for all combinations of analytes  is shown in Table 3.
                                            Table 3
                                   Results of Paired-T Tests
Composite
Location
Flask
Purge
Device
Background
Exclude
Subtract
Exclude
Subtract
Gas/H20
Soluble
No
Yes
No
Yes
No
Yes
No
Yes
N
49
19
58
23
63
28
79
36
Mean
Ratio
1.08
1.07
1.08
1.10
1.12
1.10
1 14
109
RSD
(%)
5.5
10.3
5.5
10.7
8.0
9.1
96
9.2
T
88
2.7
10.2
3.9
10.4
5.2
11.3
5.4
Prob
0001
0.015
0.001
0.001
0.001
0.001
0.001
0.001
"Background" indicates whether analytes present prior to spiking were background-subtracted or
excluded from the analysis, "Mean Ratio" is the average grab-recovery-to-composite recovery ratio,
"T" is the value of the paired T-statistic; and "Prob" is the probability that the ratio is statistically
different from 1.00
September 1, 1994
15

-------
 VOf\ Compositing Procedntes
       Flask xs Purpc Device Compositing

       Results comparing flask and purge device compositing techniques are summarized m Table 4
 As with comparisons of grab and compositing results, a paired t-statistic was used to determine the
 significance of differences between the two techniques.  As shown in Table 4, statistically significant
 differences exist between the  flask and purge device compositing techniques, with the purge device
 technique producing recoveries approximately 5 percent higher than the flask technique  As with
 comparisons of grab and compositing results, these differences are so small that they are likely
 undiscermble by other than isotope dilution quantitation and, although significant statistically,
 should not be considered significant from  an analytical chemistry perspective

                                             Table 4
                Comparison of Flask and Purge Device Compositing Recoveries
Composite
Location
Flask
Purge Device
N
72
114
Mean
Ratio
108
1.13
RSD
(%)
7.0
9.6
T
-3.52
Prob
0.0005
OVERALL CONCLUSIONS

      Mathematical averages of the results from analyses of grab samples were found to be larger
than the result from the analysis of either Flask or Purge Device composited samples, although
these differences are on the order of a few percent and would not be discernable except by isotope
dilution quantitation procedures.  In addition, the number of samples tested in this study (7) was
relatively small, even though the number of analytes per sample (40) was large.  Because the
behavior of one analyte can be expected to be correlated with that of other analytes in the sample,
it is possible that the small number of samples results in differences that would be negated or lost in
a larger study  Further, in tests of reagent water preliminary to this study of actual field samples,
composited samples yielded results that were a few percent larger than mathematically averaged
results of analyses of  grab samples; i.e., the results in reagent water tests were the opposite of the
results with field samples.  The  reasons for these differences are not known but are likely to
measuring or compositing errors, even though calibrated syringes and volumetric glassware were
used
16
September 1, 1994

-------
                                                                      VQA Compositing Procedures
 DISCUSSION
      Compositing can be useful in some situations and will result in a cost savings over the analysis
 of individual grab samples

      EPA plans to continue the use of VOA compositing in its effluent guidelines program and,
 after further studies, may promulgate compositing procedures for wastcwaters
REFERENCES

1.    Larry H  Keith and William A. Telliard, Env. Sci. & Tech. 1979, 13(4) 416 - 423.
2.    T.A. Bellar, J.J. Lichtenberg, and  R.C  Kroner, 1. Am. Water Works Assoc. 1974, 66,
      703 - 706.
3.    William A. Telliard, Marvin B. Rubin, and D.R. Rushneck, I. Chromatog. Sci  1987, 25,
      322 - 327.
4.    "Compositing and Three-Stage Sampling", from Statistical Methods for Environmental
      Monitoring. R.O. Gilbert Editor, D  Van Nostrand, New York, 1987
5.    Forest C  Garner, Mann A. Stapaman, and Llewellyn R. Williams,  "Composite Sampling for
      Environmental Monitoring", from Principles of Environmental Sampling. Lawrence H. Keith
      Editor, American Chemical Society, Washington DC, 1988.
6.    Sharon M. Clme and Blame F. Severm, Water Res. 1989, 23(4) 407 - 412
7.    EPA 833-B-92-001, 1992
8.    G.H. Stanko. Environmental Lab 1994, 6(2) 10 - 15
ACKNOWLEDGEMENTS

     The author wishes to thank Roger Litow of DynCorp Viar for assistance in collection and
tracking of the samples and statistical analysis of the results, the laboratories of Pacific Analytical,
Inc. for the isotope dilution GC/MS analyses, and Dale Rushneck of Interface, Inc. for technical
support.

     This chapter has been reviewed by the Analytical  Methods Staff of the EPA Office of Water
Mention of company names, trade names, or commercial products does not constitute endorsement
or recommendation for use.
September 1, 1994                                                                            17

-------