% PR0^0<
  SEVENTH ANNUAL
   WASTE TESTING
        AND
QUALITY ASSURANCE
    SYMPOSIUM
      JULY 8-12,1991
  GRAND HYATT WASHINGTON
    WASHINGTON, D.C.
    PROCEEDINGS
       Volume I

-------
               VOLUME
                      I
THE SYMPOSIUM IS MANAGED BY THE AMERICAN CHEMICAL SOCIETY
                   printed on recycled paper

-------
                    TABLE  OF  CONTENTS
                                         Volume I
Paper                                                                              Page
Number                                                                           Number


QUALITY ASSURANCE

 1.      Western Processing: Surface and Ground Water Monitoring During A Superfund Remediation.      I -  1
        D. Actor, Z. Naser

 2.      A Quality Assurance Program for Remedial Actions Within the USEPA ARCS Program.           1-18
        £>. M. Stainken, D. C. Griffin, K. Krishnaswami, J.C. Henningson

 3.      A National QA Standard for Environmental Programs for Hazardous Waste Management           1-26
        Activities. G. L. Johnson, N. W. Wentworth

 4.      The Impact of Calibration on Data Quality. R. G. Mealy, K. D. Johnson                        1-34

 5.      Proficiency Evaluation Sample Program for Solid Waste Analysis:  A Pilot Project.               1-47
        D. E. Kimbrough, J. Wakakuwa

 6.      Technical Data Review - Thinking Beyond Quality Control. K. D. Johnson, R. G. Mealy           1-48

 7.      Quality Assurance Strategies to Improve Project Management. T. L, Vandermark, G. F. Simes       I - 65

 8.      Bias Correction: Evaluation of Effects on Environmental Samples. M. W. Stephens,              1-66
        M. A. Paessun

 9.      Ensuring Data Authenticity in Environmental Laboratories. J. C. Worthington, R. P. Haney         1-81

10.      Establishment of Laboratory Data Deliverable Requirements for Data Validation of               1-82
        Environmental Radiological Data. D. A. Anderson

11.      An Assessment of Quality Control Requirements for the Analysis of Chlorinated Pesticides         1-83
        Using Wide Bore Capillary Columns—A Multi-Laboratory Study. /. A. Serges,
        G. L. Robertson

12.      Analysis-Specific Techniques for Estimating Precision and Accuracy Using Surrogate             1-84
        Recovery. C. B. Davis, F. C. Garner, L. C. Butler

13.      Use of Organic Data Audits in Quality Assurance Oversight of Superfund Contract               1-86
        Laboratories. E. J. Kantor, M. Abdel-Hamid

14.      Useof Inorganic Data Audits in Quality Assurance Oversight of Superfund Contract              1-87
        Laboratories. R. B. Elkins, W. R. Newberry

IS.      Improved Evaluation of Environmental Radiochemical Inorganic Solid Matrix Replicate           1-88
        Precision: Normalized Range Analysis Revisited. R. E. Gladd, J. W. Dillard

16.      Laboratory On-site Evaluations as a Tool for Assuring Data Quality. T. J. Meszaros,              1-92
        G. L. Robertson

-------
17.      Application of Bias Correction. D. Syhre                                                   1-93

18.      Matrix Spiking: From Sampling to Analysis. D. Syhre, M. Rudel, V. Venna                      1-116

19.      Land Disposal Restrictions Program Data Quality Indicators for BOAT Calculation: Past and        1-131
         Future. J. Alchowiak, L. Jones

20.      Comparison of Quality Assurance/Quality Control Requirements for Dioxin/Furan Methods.         I - 138
         D. Hooton

21.      A Study of Method Detection Limits in Elemental Solid Waste Analysis. D. E. Kimbrough,          I - 151
         J. Wakakuwa

22.      Preparation and Validation of Proficiency Evaluation Samples for Solid Waste Analysis.             I - 165
         D. E. Kimbrough, J. Wakakuwa

23.      Observation of Quality Assurance Anomalies in Superfund Activities. D. M. Stainken              1-180

24.      Functional Evaluation of Q C Samples, a Proactive Approach. D. R. Xiques, J. Allison              1-181

25.      Features of the U.S. EPA-Quality Assurance Material Bank Standards. R. A. Zweidinger,            1-186
         N.Malof

26.      Automated Data Validation—PANACEA or TOOL. G. Robertson                               1-187

27.      Building Data Quality into Environmental Data Management M. AfiOer, P. Ludvigsen              1-188

28.      A Software Approach for Totally Automating the Quality Assurance Protocol of the EPA            1-198
         Inorganic Contract Laboratory Program. C. Anderau, R. Thomas

29.      Automated Reporting of Analytical Results and Quality Control for USEPA Organic and            1-206
         Inorganic CLP Analyses. R. D. Beaty, L. A. Richardson

30.      A Customizable Graphical User-Friendly Database for GC/MS Quality Control. P. Chong,           I - 219
         J. S. Hicks, /. Janowsld, G. Klesta, C. Pochowicz

31.      Computer Assisted Technical Data Quality Evaluation. S. Hopper, J. Burnetti, M. Stock             I - 225


SAMPLING/FIELD

32.      Preparation and Stabilization of Volatile Organic Constituents of Water Samples by Off-Line        I - 243
         Purge and Trap. E. Woolfenden, J. R Ryan

33.      A Remote Water Sampler Using Solid Phase Extraction Disks. H. A. Moye, W. B. Moore            1-245

34.      Representative Sampling for the Removal Program. W. Coakley, L. Ray, G. Mallon, G. Janice        I - 262

35.      Preliminary Field and Laboratory Evaluations and Their Role in an Ecological Risk                I - 277
         Assessment for a Wetland Subject to Heavy Metal Impacts. G. Under, M. Bollman, S. Ott,
         J. Nwosu, D. Wilbom, B. Williams

36.      PAH Analyses: Rapid Screening Method for Remedial Design Program. L. Ekes, M. Hoyt,         I - 284
         G. Gleichauf, D. Hopper

37.      Evaluation of Household Dust Collection Methods for HUD National Survey of Lead in             1-285
         Homes.  J. J. Breen, K. Turman, S. R. Spurlin, B. S. Urn, S. Weitz

38.      Field Deployment of a GC/Ion Trap Mass Spectrometer for Trace Analysis of Volatile Organic       I - 296
         Compounds. C. R Leibman, D. Dogruel, E. P. Vanderveer

-------
39.      Accurate, On-Site Analysis of PCBs in Soil — A Low Cost Approach. D. Lavigne                 I - 298

40.      How Good are Field Measurements? L. Williams                                             1-311

41.      Assessment of Potential PCB Contamination Inside a Building; A Unique Multi-matrix             I -312
        Sampling Plan. W. W. Freeman

42.      Comparison of the HNU-Hanby Field Test Kit Procedure for Soil Analysis With a Modified         I - 323
        EPA SW-846 5030/8000 Procedure. /. D. Hanby, B. Towa

43.      Field Test Kit for Quantifying Organic Halogens in Water and Soil. D. Lavigne                   I - 331

-------
QUALITY ASSURANCE

-------
 WESTERN PROCESSING: SURFACE AND GROUND WATER MONITORING
                 DURING A SUPERFUND REMEDIATION

David Actor. Manager Sampling Programs, Chemical Waste Management, Inc., 150
West   137th  Street,  Riverdale,  Illinois  60627;  Zaki  Naser, Technical  and
Environmental Manager, Western Processing Project, Chemical Waste Management,
Inc., 20015 72nd Avenue South, Kent, Washington 98032

ABSTRACT

Collecting high quality,  defensible  samples  from the  environment can be  a
controversial and difficult task. This paper describes an operating ground water and
surface water sampling program that is monitoring the  progress of  a long-term
superfund remediation.

In 1983, Western  Processing was listed under CERCLA  as one of the fifty most
contaminated sites in the nation.  Written into the consent decree are requirements
for both surface and ground water monitoring. The stream that runs adjacent to the
site has intensive monitoring requirements during remediation with clearly defined
water quality objectives.  Ground water monitoring is required during  remediation
and for 30 years thereafter.

The monitoring approach includes comprehensive quality assurance/  quality control
(QA/QC) and sampler training programs.  Dedicated ground water monitoring
equipment is  utilized  to minimize introduction of  contamination  by sample
collection.  The surface stream that runs adjacent to the site is sampled with non-
dedicated equipment. A rigorous QA/QC program has been implemented to track
any possible contamination introduced during sample collection and to ensure the
integrity of every sample obtained.

The monitoring equipment and the methods employed on this project as of 1987 are
discussed  including presampling  activities,  sampling  procedures,  field  records
handling,  sample  parameters (which include priority pollutant listed  compounds,
sampling  schedule, and  analytical  parameters and  procedures,  and  QA/QC
objectives. The health and safety approach for the  environmental  monitoring
program  at Western  Processing is  discussed,  and a brief description of the
environmental cleanup is given.
                                    1-1

-------
I  INTRODUCTION

History

The   Western  Processing superfund  site  is  located  in  Kent,  Washington,
approximately 20 miles south of Seattle.  The site is presently surrounded  by
warehouse and manufacturing facilities that were built over the last decade.

From 1952 through 1961 the site was operated by the U.S. Army as an anti-aircraft
battery.  The Western Processing Company purchased the 13 acre location and
began operations in  1961  as an animal by-products and brewers yeast processor.
Operations were expanded to include the  reprocessing of pickle liquor, recovery of
heavy metals and waste solvents, neutralization of acids and caustics , electrolytic
destruction of cyanide, chemical recombination to produce zinc chloride and lead
chromate, reclamation of flue dust, metal  finishing by-products, and ferrous sulfide
in fertilizer production. In 1983, due to environmental problems associated with the
site, the U.S. Environmental Protection Agency initiated closure of the facility and
an emergency response  cleanup  action.   Following  surface  remediation and
establishment of surface water  control measures,  an  intensive  shallow soil
contamination study was conducted to determine the extent of hazardous chemical
contamination  and  provide  a baseline  for remedial  activities.   Sampling
investigations  conducted  between  1982 and  1986  have identified  over  70
contaminants in soils and 46 contaminants in  ground water samples.   From this
initial data ground water and surface water indicator chemicals were selected for
long term environmental monitoring, (Tables 1 and 2).

Geology

Western Processing is located in the Duwamish/Green River Valley flood plain and
is bounded to the west by Mill Creek and to the east by a shallow drainage ditch
(Figure 1).  Ground water is shallow, ranging from 3 to 15 feet below the ground
surface.  Underlying  soils are comprised of fill and laterally discontinuous and
unconfined lenses of sands, silts, and clays. A  discontinuous sandy and clayey silt
layer is present at about 35 feet below the ground surface. This aquitard is about
5 feet thick.  Below 40 feet the soil is generally unconfined sands and constitutes the
regional aquifer.  The regional ground water  flow  is  generally to the northwest
resulting in an upgradient  direction to the east and southeast of the  site. Shallow
ground water flow is influenced by discharge to Mill Creek at depths of 30 to  40
feet during normal to low  flow periods. Regional ground water flow is about 100
feet per year.
                                     1-2

-------
                                TABLE 1

               ANALYTICAL SUITE FOR GROUND WATER
                           (Indicator Chemicals)
Volatile Organics - All volatile organic priority pollutants

Metals (total)                  Base Neutral/Acid Extractibles
      Cadmium                     Bis (2-ethylhexyl) phthalate
      Chromium                     2,4 Dichlorophenol
      Copper                       2,4 Dimethylphenol
      Nickel                        Isopherone
      Lead                         Phenol
      Zinc
      Iron
      Manganese
      Sodium
      Calcium
      Cyanide
      3-(2-Hydroxypropyl)-5-Methyl-2-Oxazolidinone

Conventional Parameters
      Total Hardness
      Temperature (field)
      pH (field)
      Specific Conductance (field)
      Total Chlorides
      Sulfates
      Bicarbonate
      Carbonate
                                     1-3

-------
                                TABLE 2

               ANALYTICAL SUITE FOR SURFACE WATER
                           (Indicator Chemicals)
Volatile Organics
      Chloroform
      1,1-Dichloroethane
      1,1-Dichloroethene
      Ethylbenzene
      Methylene Chloride
      Tetrachloroethene
      Trans-l,2-Dichloroethene
      Cis-l,2-Dichloroethene
      1,1,1-Trichloroethane
      Trichloroethene
      Toluene

Metals (Total & Dissolved)
      Cadmium  ,
      Chromium
      Copper
      Nickel
      Lead
      Zinc
      Iron
      Manganese
      Sodium
      Calcium

Base Neutral/Acid Extractibles
      Bis (2-ethylhexyl) phthalate
      2,4 Dichlorophenol
      2,4 Dimethylphenol
      Isopherone
      Phenol
Conventional Parameters
      Temperature (field)
      pH (field)
      Specific Conductance (field)
      Dissolved Oxygen (field)
      Hardness
      Chloride
      Ammonia
      Turbidity
      Nitrate
      Phosphorous
      Total Suspended Solids
Other
      Cyanide
      3-(2-Hydroxypropyl)-5-Methyl-2-Oxazolidinone
                                    1-4

-------
      NORTH
15M17
ABCO
          15M16
          ABC
        15M15
        ABC
                         8M8
                         ABCO
                                          C-3
MONITORING LOCATIONS
     WESTERN PROCESSING
           KENT, WASHINGTON
                                                                 KEY
                                                           C-1
                                                          ABCD
                Stream Monitoring Station

                Monitoring Well Locatton

                & Description
                                                                 13M12

                                                                 ME :
                                                                        200
                              FIGURE 1
                                                               SCALE IN FEET
                                      I-5

-------
Contractor/Client/Government Interaction

The Western Processing  superfund project  may  be unique  with its proactive
approach to overall site management. Informal weekly meetings are held to inform
the Trust overseeing the cleanup, the governments (U.S. Environmental Protection
Agency, Washington State Department of Ecology, and the City of Kent) and other
regulatory parties  of  the  status of site operations.   Regulatory interaction and
cooperation to resolve project issues is very high.

II REMEDIATION

Subsurface remediation included the removal of highly contaminated soils and non-
leachable materials, installation of ground water extraction, infiltration and water
treatment systems and  a ground  water monitoring network.   A site-dedicated
laboratory was constructed to analyze both process and environmental  samples
generated by the project.

Twenty-two  thousand cubic yards  of highly contaminated  and  low permeability
materials were excavated,  and the pits backfilled with clean, high permeability fill.
The site was then graded and bermed, and a shallow (30 feet deep)  extraction well
system was installed.  Organics and metals contaminated  ground water is  pumped
from these wells by vacuum extraction and transferred to a water treatment facility
where metals are removed by precipitation/clarification and organics are removed
by air stripping and carbon adsorption. The dedicated laboratory performs analysis
on water and soils for organic and inorganic parameters. Laboratory  instruments
include three gas chromatograph/mass spectrometers for volatile and semi-volatile
analyses, an inductively coupled plasma-arc furnace  and two atomic  absorption
furnaces for metals analyses, a gas chromatograph for pesticide analysis, an ion
chromatograph for anion analysis, and a UV vis spectrometer for phenol and
cyanide.

A slurry wall was constructed to minimize lateral ground water movement into or
away from the site during ground water extraction.  This wall encircles the site and
is composed of a soil and bentonite mixture with a hydraulic conductivity of about
1X10"7 cm/sec. Its depth ranges from 40 to 45 feet so it junctions with the aquitard
zone which is present at 35 feet. The Western Processing Consent Decree stipulates
that an inward  gradient  to  the site must be maintained during ground water
remediation. Twenty-two pair of shallow and deep piezometers were  installed inside
and outside  the slurry wall to monitor the horizontal  and vertical gradient relative
to the site and to aid in management of the extraction and infiltration  systems.
                                     1-6

-------
Ill MONITORING PROGRAM

The monitoring program at Western Processing is divided into two parts:  process
monitoring, which includes discharge compliance monitoring and environmental
monitoring.  Process sampling and analysis monitors remediation progress of the
extraction field and treatment efficiency through the various treatment processes.
Environmental monitoring at Western Processing tracks relatively low levels of
contaminants in the ground water that originally migrated off-site.  Immediately
adjacent and down-gradient from the site is Mill Creek.  Performance standards for
Mill Creek water quality are specified in the Western Processing Consent  Decree.
These standards were achieved by reducing the contaminant concentrations at the
downstream sampling locations  (Figure 1) below the  applicable ambient water
quality criteria (AWQC).  The applicable AWQC are those that  were published in
the Federal Register at the  time of entry of the Consent Decree  (April 1987).
Relevant AWQC are for cadmium, chromium (hexavalent and total), copper, lead,
mercury, nickel, silver, zinc and cyanide. This performance standard was achieved
within its three year compliance period.  Sampling protocol is extremely important
for collection of representative samples and elimination of potential contamination
from sample collecting activities.  Mill Creek water is sampled monthly for ground
water source  contamination  at  one upstream location and at  two downstream
locations.  In addition, Mill  Creek sediments  are sampled semiannually at one
upstream location and three downstream locations.

Monitoring wells  have  been installed  up-gradient  of Western Processing  for
background data and down-gradient of the site to track site-related contaminant
migration.  Presently the monitoring network includes 54 wells,  11 of which were
pre-existing. The monitoring  installations are individual shallow wells and  clusters,
consisting of three or four wells with individual wells in each cluster, each screened
in a different zone.  All  newer installations are single completion wells with 10 foot
screen lengths placed in depth zones of 10 to 30 feet, 40 to 60 feet, 80 to 100 feet,
or 120  to  140  feet.  The actual  screen  interval was determined during  well
installation by both sieve analysis of the soils in the proposed  screen zone and
observations of the hydrogeologist that logged the boring. All monitoring wells were
installed using cable tool drilling equipment.  Long-term ground water  monitoring
is conducted on a  quarterly basis with special  interest  wells  being  monitored
monthly. Water levels  are being monitored monthly during the operation of the
ground  water extraction system.  The duration for  long-term ground water
monitoring is thirty years after the governments have determined that an acceptable
level of remediation has been achieved by ground water extraction and treatment.
                                     1-7

-------
Monitoring Approach

Since environmental monitoring for Western Processing will be conducted for thirty
years after completion  of remedial activities, a  comprehensive and defensible
program had  to  be developed.  To assure that the sampling was consistent,
especially due to project  duration and the long term potential for litigation
concerning the analytical data, a comprehensive Quality Assurance/Quality Control
(QA/QC) program was established.  QA/QC for the monitoring program includes:

      o    Rigorous sampling methodologies;
      o    Defensible documentation of  all  calibrations,  chain-of-custody,
            maintenance, training and sampling  procedures;
      o    Sample  contamination evaluation  to determine  if  technique or
            equipment is introducing contamination to the sample;
      o    Duplicate sampling and analyses to assess analytical precision;
      o    Round robin analyses to measure field analytical performance;
      o    Instrument calibration and performance checks to assess whether the
            field instruments are performing within acceptable parameters and;
      o    Frequent sampling audits.

Position requirements for  sampling personnel at Western Processing are high.
Physical requirements are demanding as the sampler must perform sampling tasks
in protective clothing in  a variety of weather conditions. The sampler must possess
adequate academic training to comprehend the concepts and goals of the work.
Sampler training  is essential to assure the  competency  of the sampling team in
performing their  tasks  and maintaining consistency of protocol and  technique
throughout the duration of the program. Training is required a minimum of once
a year for all samplers.  Training includes a thorough review of all relevant work
plans, discussion  of the  sampling  theory  and it's application, and supervised
demonstration of sample collection. A sampler proficiency test is administered
annually by comparing analytical results of the trainee to the trainer.  All sampler
training is documented and archived with project records.

Monitoring Equipment

The monitoring program is operated out of a 10 X 40 foot trailer that has been
modified for field operations.  The  monitoring laboratory contains both domestic
and deionized water sources  for  decontaminating equipment,  a  lab hood for
application of solvent and  acid rinses, flammable storage, equipment and  supply
storage, work  counters for instrument calibration and maintenance, and a waste
                                     1-8

-------
water tank to collect all decontamination water for treatment. Equipment common
to both the  ground water monitoring program and the  surface water monitoring
program are pH meters, specific conductivity meters and  a dedicated sampling van.
The van is the platform from which all field activities are conducted. This includes
field pH and specific conductivity analyses, field data entry and sampling equipment
and supply storage.

Monitoring Wells

Presently,  54 monitoring wells  are being sampled for  Western Processing.  To
eliminate the possibility of cross  contamination, the sampling equipment that comes
in direct contact with ground water is dedicated to each well.  The type of sampling
pump utilized for all monitoring wells on the project are submersible mechanical
pumps.  The pump is a double check valve, positive  displacement, piston pump.
Actuation of the pump  is  accomplished  from the well  head  with a portable
pneumatic motor.  The mechanical connection to the submersible pump is by a
small  diameter stainless steel pushrod. Only the pump, pushrod, and discharge pipe
contact  ground  water down  the well.   Materials  of construction of the sampling
pump are  teflon and stainless steel, which provide good chemical resistance while
maintaining reliability.  Discharge pipe is 3/4 inch schedule 80 PVC, which provides
an adequate discharge rate while providing enough room between the discharge pipe
and the well casing to allow the use  of a water level probe.  The  well head
discharge  is  a 3/4 inch PVC tee with a hose-fitting connection.  The advantage of
this system is that the well can be purged at a reasonable pumping rate (5 gal/min)
and sampled (as low as 120 ml/min) with the same pump. The compressed air
source  for the pneumatic motors is  a  trailer-mounted  air  compressor.   The
compressor produces sufficient volume to purge three wells concurrently to minimize
total sampling time.  A 1000 gallon purge water collection tank is also mounted on
this trailer.

Surface Water Monitoring

Unlike  the  ground water monitoring program, the surface water and sediment
sampling program does not employ dedicated sampling equipment. To assure high
quality  surface  water and sediment samples, rigorous QC procedures have been
incorporated to detect  sample contamination. These procedures are  described in
QA/QC objectives.

Non-dedicated equipment includes mechanical current meters for flow measurement,
a subsurface grab-type sampler for collecting stream samples, an Ekman-type dredge
                                     1-9

-------
for collecting stream sediments, a  2.4 liter pressure filter for filtering stream
samples, and stainless steel mixing  bowls  and trowels for compositing sediment
samples.

During  sample collection, current meters are used to determine stream  flow by
taking velocity measurements at different points in the stream's cross section.  Two
types of current meters are used- selection depends on the water depth. A current
meter consists of a vertical axis rotor with cups that is attached to a wading rod.
Sediment samples are collected  from Mill Creek  with  a pole-mounted Ekman
dredge.  This sediment sampler is a stainless steel box with spring loaded jaws that
are tripped after the box has been driven into the sediment. The pressure filter is
an acrylic pressure filtration unit designed for field filtration of water samples. The
filter unit is pressurized with nitrogen and is used for collecting  samples for
dissolved  metals  analysis.   A one liter grab-type  water sampler  is utilized for
collecting all water samples. Water is transferred directly from the grab sampler to
individual sample bottles.  In-situ  oxygen measurements  are observed with  a
dissolved oxygen meter.

Monitoring Methods

Pre-Sampling Activities

Before  entering  the field  and initiating  field activities, available  background
information on the monitoring station is reviewed.  This information  includes the
condition of the well or stream station and range of historical field test data (pH,
specific  conductivity, dissolved oxygen,  temperature, purge volume, etc.).  Field
equipment is checked  for  proper  operation  (i.e., the air compressor is run,
pneumatic motors are operated, current meters are assembled, the dredge is tested,
etc.). The field instruments are calibrated and results recorded in a laboratory
logbook.  Part of the calibration procedure is an instrument performance  check
(IPC/QC). Results of this IPC/QC must be within two standard deviations or the
instrument is recalibrated.  A  closing calibration check is conducted at the end of
the sampling day with the results recorded in the laboratory logbook.  All calibration
logbooks are reviewed and initialed monthly by the  site QA officer.

Pre-cleaned and quality controlled bottles  are utilized for sample collection.  To
minimize  contamination  introduced from the field, preservatives, if required, are
added by the laboratory with the exception of volatile samples. Volatile samples are
preserved hi the field immediately after sample collection.  Proper labels, chain of
custody  forms, and custody seals are assembled.
                                     MO

-------
Field Records

Field sampling records consist of the chain of custody, field parameter form, and
field logbook. The chain of custody (COC) accompanies and tracks the sample from
acquisition through analysis to final disposition. The form is designed to summarize
the contents of the shipment, dates and times of custody transfer, and signatures of
all individuals relinquishing and receiving the samples.  It includes the following
information:

      o  Project name          o  Sample number
      o  Sampler's name        o  Date/time
      o  Analysis parameters    o  Number of containers
      o  Remarks              o  Relinquished by
      o  Date/time             o  Received by

The Western Processing  COC form closely resembles the NEIC form with the
addition of analytical parameters and project name  printed on the basic form for
efficiency.

The field  parameter  form contains  information  about sampling  procedures,
equipment, conditions, and field measurements.  All field information is recorded
via a laptop computer and downloaded at the end of the day into the laboratory
database.  A hard copy is printed at that time for review, date and signature by the
field sampler.  The signed hard copy is archived for future reference.  The field
parameter form contains the following information:

      o     Sample  point - the complete sample number  which includes the
            sample location plus the laboratory ID number.
      o     Purging information - date, time, volume.
      o     Sample depth, water depth, flow.
      o     Sampling equipment information.
      o     Field measurements - pH, temperature, conductivity, dissolved oxygen,
            and ground water elevation.
      o     Field comments - weather conditions, well and dedicated equipment
            condition, sample appearance and preservatives added in the field, if
            any.

The field  logbook is a numbered controlled document that contains hand written
field notes  and data  that compliment  the  information  provided  on the field
parameter form and COC.  Each page is numbered, initialed,  and dated by the
                                    1-11

-------
sampler.   The field log book provides the additional information necessary to
respond in detail to inquiries about a sampling event, especially when conditions
require deviation from the procedures specified in the workplan.

Sample Storage and Transfer

Immediately after sample collection, the bottles are placed in an insulated shuttle
with ice  packs and transported to  the laboratory for analysis.   Samples are
transferred to the laboratory usually within 4 hours of collection, minimizing any
temperature changes that might result from shipping  to an  off-site laboratory.
Transfer of samples to the laboratory requires a properly prepared chain of custody
form. An incomplete or improperly prepared chain of custody could invalidate any
resulting data.

Analytical Procedures

Laboratory analysis  is accomplished using USEPA Methods, SW-846, Standard
Methods  for the Analysis of Water and Wastewater, and CLP Methods. Table 3
presents these methods.
                                TABLE 3
                       Analytical Laboratory Methods
Parameter

Acidity
Alkalinity
Ammonia
Anion Chromatography,
Chloride, Nitrate,
Sulphate
Cyanide
Hardness
Metals (ICP)
Arsenic (GFAA)
Antimony (GFAA)
Lead (GFAA)
Mercury (CVAA)
Mercury (CVAA)
Selenium (GFAA)
 Method
Reference

305.1
310.1
350.3
300.0
335.2
130.2
200.7
206.2
204.1
239.2
245.1
245.5
270.2
Matrix

Water
Water
Water
Water
Water
Water
W/S
W/S
W/S
W/S
w
s
W/S
                                   1-12

-------
TABLE 3 - Analytical Laboratory Methods (cont'd)

Thallium (GFAA)              279.2                    W/S
Phosphorous                    365.2                    Water
Total Suspended Solids          160.2                    Water
Turbidity                       170.1                    Water
Volatile Organics               8260                    W/S
Semi-Volatile Organics          8270                    W/S
Pesticides/PCB                 8080                    W/S

Quality Assurance and Quality Control Objectives

A strong quality  assurance/quality control (QA/QC)  program for sampling and
analysis is incorporated to determine actual environmental contamination.  QA/QC
is used to assess the sample's ability to represent its sampling location. A minimum
of 10% of the total number of samples collected are quality control samples.  These
include both sample duplicates and method blanks.

Duplicate samples are collected  in the same manner as the actual environmental
samples.  The duplicate is not a split sample, but, is collected immediately after the
original environmental sample.  Over the duration of the project each station is
being included in this process.

For  ground water  sampling,  a  field method  blank consists  of the  appropriate
sampling containers filled by the laboratory with deionized water and sent into the
field with other containers to check the quality of the sampling environment.  These
blanks are opened at the sampling station and poured into sample containers during
the sampling event and transported to the laboratory for analysis.  Volatile  blanks
are not opened in the field and serve as a trip, or transport blank to test container
quality. If contamination is present  after  analysis of the volatile blank, container
banks are  collected and analyzed  on the specific lot(s) of samples bottles used
during that event. The well would then be resampled.

For surface water sampling, field method blanks are collected  to check equipment
decontamination  procedures,  quality of the  sampling environment,  and sample
container quality. The method blank  consists of pouring laboratory deionized water
into the grab sampler and then transferring it to sample containers to be  analyzed
as a regular sample.  For filtered samples, the water  is poured into the pressure
filter and filtered as a normal investigative sample.  Volatile sample bottles are
filled in the laboratory and travel to  the field  and back without being opened.
                                    1-13

-------
For sediment sampling, the Ekman dredge, and mixing equipment interact with the
sample. Acid washed sand is placed in the dredge, emptied into the bowl and then
transferred to the jar as an actual sample would be  handled.

Monitoring Wells

Sampling Procedures

After arrival at the well  location, the conditions of the well and the immediate
surroundings are  observed and recorded. This  includes weather conditions, well
integrity, evidence of tampering or contamination, and conditions in the area that
could effect the quality of the sample (airborne contaminants, etc).  All wells are
photographed annually to document the condition of each well. Prior to sampling,
the ground water elevation is measured and the monitoring well purged. The well
is  purged so that the ground water sample collected  is representative of the
formation water  at that  point in time.   The ground water surface elevation  is
measured with an electric tape from the top of the well casing to the water level in
the casing.  Water levels are measured  to the  nearest hundredth  of a foot with
precision at +. two hundredths of a foot.  After the ground water surface elevation
has been determined,  the volume of ground water in the well  casing can be
calculated and the total purge volume determined.  The industry standard and the
approved purge amount for this project is three casing volumes. Purge volumes are
measured in the field with in-line flow meters calibrated in gallons and tenths of
gallons. Once a year pH, temperature and conductivity are measured continuously
during the purge process to verify that purging three casing volumes of ground water
is adequate to provide a representative sample. All the wells monitored to date are
stable with respect to these parameters before complete removal of the third casing
volume.

The pneumatic motor and purge water discharge line are attached to the well head
assembly and the  well pumped until the required volume has been evacuated. The
purge water discharge lines are not dedicated to each well, so the end contacting the
well head assembly  is decontaminated before and after each use.  A backflow
preventer assures that purge water cannot drain back into the well during the
pumping process.  After completion of purging, the well is allowed a reasonable
period to recharge (usually five to thirty minutes, depending on the recharge rate)
prior to sampling. During this recharge period  the dedicated small diameter 3/8
inch sample hose is attached and the motor adjusted for minimum flow operation.
Ground water samples are collected immediately after well purging and recovery.
Samples are collected over a bucket to minimize the possibility of contaminating the
                                    1-14

-------
immediate vicinity of the well. A one liter bulk sample is collected and split into
four discrete samples from which temperature, conductivity and pH are measured.

Collection of volatile organic samples involves filling the appropriate vial very slowly
with as  little  air contact as  possible.  Because the analysis requires that volatile
samples be headspace free, the vial is allowed to overflow at least 1.5 volumes. The
appropriate amount of preservative (concentrated HC1) is added and the cap is
gently replaced.  All other sample bottles are filled with a minimal amount  of air
contact. These bottles are filled as full as possible without any overflow.

Sample Parameters and Schedule

The long term ground water  monitoring network is sampled on a quarterly basis for
the constituents listed in Table 1.   Monitoring wells are sampled  in  the  same
sequence each quarter to insure that the individual sampling frequency remains one
quarter apart.  During the summer quarter, a complete priority pollutant scan is
conducted on a sample from each monitoring well in addition to those parameters
not on the priority pollutant list.

Surface Water Monitoring

The stream that lies adjacent to the site is continuously monitored for flow using a
pressure transducer located behind a weir. Resulting information is recorded by a
data logger.  Comparative flow measurement data  is manually collected at each
stream  sampling  station immediately following  sample collection.    Velocity
measurements  and stream  flow  are  determined using U.S.  Geological Survey
methods.

Sampling Procedures

Prior to sampling, weather conditions and stream height are evaluated to determine
if storm conditions (abnormally  heavy rain and high stream conditions)  exist.
Monitoring will occur if conditions are within the norm for that time of year.  This
is to develop  data representative for that period and  not storm data. The sampling
station is evaluated for  any  physical changes or conditions that could impact flow
measurement  or sample collection.  These are noted on the field parameter form.

Field measurements are conducted using  the same  protocol as for ground  water
monitoring.   A one liter bulk sample is collected with the grab sampler and the
contents split into four discrete samples from which  temperature, conductivity and
                                     1-15

-------
pH are measured.  Dissolved oxygen calibration and measurements are conducted
in-situ.  Collection of volatile organic samples involves filling the appropriate vial
from the grab sampler with as little  air contact as possible. As with ground water
samples, the analysis requires that volatile samples be headspace free. When filling,
the vial is allowed to overflow at least 1.5 volumes.  The appropriate amount of
preservative (concentrated HC1) is added and the cap is gently replaced. All other
sample bottles are filled with a minimal amount of air contact. These bottles are
filled  as full as possible without any overflow.  For  total  and dissolved metals
analysis, the sample is split into two aliquots. One aliquot is filtered in the field and
the other is left unfiltered.  This insures that laboratory analyses are performed on
one sample of water.

Sediment Sampling Procedures

Mill Creek sediment within the reach of Western Processing is generally silty with
some  sand.  The ekman dredge was selected for sediment sampling because it can
collect an adequate volume of reasonably undisturbed sample from this sediment
type.  The number of individual samples collected at each station ranges from one
to three, and is dependent upon the stream  width at each location.  Multiple
samples are  mixed to  obtain one composite sample  for each station.  Volatile
samples are  collected by taking approximately equal  aliquots from beneath the
undisturbed surface of the individual samples immediately after sample collection
and before compositing.

Sample Parameters and Schedule

Mill  Creek  water  is  sampled  monthly until the completion  of remediation.
Parameters  for analyses  are listed  in Table 2.  During  the summer quarter, a
complete priority pollutant  scan is  conducted on  each surface water sample in
addition to those parameters not on the priority pollutant  list.

IV  HEALTH & SAFETY

The environmental monitoring  program is carried out in accordance with the
approved health & safety plan for the project. Enough historical data has been
accumulated on both the monitoring wells and surface water sampling to allow
protective clothing requirements to be addressed on a well-by-well and station-by-
station basis. Minimum requirements include a working uniform with safety glasses
and disposable latex or PVC gloves.
                                    1-16

-------
V   SUMMARY

The Environmental Monitoring Program at Western Processing has been refined
with input and constructive criticism by both government and private industry from
the first workplan drafts in June 1987 until government approval of the workplans
in the summer of 1990.  The result is a monitoring program that provides defensible
and reproducible  data  that may well be tested  by the courts in the future.  A
specific monitoring history can be pulled from archives and a specific sampling event
can be recreated without reliance on the sampler's memory.

VI  ACKNOWLEDGEMENTS

The authors would like to express their appreciation to Dennis Sleeves  and  Paul
Anderson of Chemical Waste Management for their input. The authors would also
like to acknowledge the Western Processing Trust Fund, the U.S. EPA, Region X,
and the Washington State Department of Ecology for their proactive approach to
resolving  both technical  and regulatory issues  during  the Western Processing
remediation.

VII REFERENCES

U.S. EPA Region 10, Investigation of Soil and Water Contamination at Western
Processing. King County. Washington. May 1983

Landau Associates, Extraction/Infiltration Systems Management Manual. Western
Processing. July, 1990

Chemical Waste Management, Inc., Ground Water Monitoring Program Workplan.
PartB. July, 1990, Revision 1

Chemical Waste Management, Inc.,  Mill Creek & East Drain Monitoring Plan.
August, 1990,  Revision 1

Chemical Waste Management, Inc., Laboratory Quality Assurance Quality Control
Project Plan. July 1, 1990
                                    1-17

-------
                   A QUALITY ASSURANCE PROGRAM FOR

          REMEDIAL ACTIONS WITHIN THE USEPA ARCS PROGRAM

                                       by

          D.M. Stainken. D.C. Griffin, K Krishnaswami and J.C. Henningson

                               Malcolm Pirnie, Inc.
                             2 Corporate Park Drive
                          White Plains, New York 10602
ABSTRACT
The Superfund Program has continually evolved over the past decade and many of the NPL
sites have advanced to varying stages where clean-up activities are to commence.  The
USEPA has established an alternative Remedial Contracting Strategy (ARCS) whereby
contractors provide support for remedial activities (response action contracts).  Quality
Assurance (QA) activities are an integral component of the technical support provided.
These QA activities are inculcated into the RI/FS and RD/RA phases, as well as in PRP
oversight tasks. Although general QA activities are relatively defined including development
of sampling plans and QA project plans, the management and technical details remain to
be implemented.
As an ARCS contractor, we have implemented a practical QA management system with
multiple geographic locations.  This system is built around a QA Program Plan, a unique
quality assurance manual and SOPs, our laboratory and equipment facility, an audit system
and a technical director - quality management review process with review teams. Applying
this system, we have conducted audits of field events and of other contractors.  This paper
will present details of the organization of the QA program, key components and past
experience in implementing and managing the process.
INTRODUCTION
The Superfund Program has steadily grown in size and complexity over the past decade and
the number of hazardous waste sites placed on the National Priority List (NPL) has grown
accordingly.   The Superfund  Program and  its legal and  technical components  have
influenced or affected numerous Federal and State programs and industrial  practices.
During the past decade, EPA used the services of Contractors to fulfill specific tasks and
objectives.  Hazardous waste sites were and continue to be evaluated through contracts
                                      1-18

-------
termed Field Investigation Team (FIT) Contracts. Services provided survey and assess sites,
"score" for hazard ranking, and if warranted place the site on the NPL list.
The Superfund Program Contract strategy has undergone internal program management
reviews and has been changed at times to achieve EPA's goals and objectives. Consequently
acronyms for Programs, Contracts, tasks, etc., have arisen and in some cases, dropped from
use. In 1990, the EPA established a long-term contracting strategy for Superfund(l). The
Agency's objectives in developing the strategy were to analyze the long term contracting
needs of the program, and to design a portfolio of Superfund contracts to meet those needs
over the next ten years. Contract support was to be implemented for enforcement support,
regional management support, removal contract support, analytical support, preremedial and
remedial contract support.  The preremedial activities include site preliminary assessments,
inspections and "scoring" of sites for NPL consideration.  Remedial activities include a
variety of  activities necessary to actually remediate a site such as the remedial investiga-
tions/feasibility study phase (RI/FS) and  the  remedial design/remedial action phase
(RD/RA).  To conduct preremedial and remedial  activities, the Agency established
contracts termed Alternative Remedial Contracting Strategy (ARCS) contracts. As  an
additional task under ARCS, some contractors may be assigned oversight and review tasks
of PRP (Potentially Responsible Party) sponsored cleanups.
A key component of ARCS in the implementation of a Quality Assurance Program which
is a requirement within all EPA programs(2).  There are numerous technical guidance
documents (3-11) and manuals which must be  integrated into  a Superfund ARCS QA
program. The components include establishment of standard operating procedures (SOP's),
field sampling plans (FSP), quality assurance program plan (QAPP), quality assurance
project plan (QAPjP), use of the CLP program, data validation, conducting audits of the
processes (eg. field audits, data quality audits, management systems or program audits, file
audits, lab audits, etc), adherence to other "applicable and relevant agency regulations"
(ARAR's), and quality control and quality assurance activities  for remedial design and
action phases.
QA PROGRAM
As a new EPA Region n ARCS Contractor, Malcolm Pirnie, Inc. (MPT) has developed a
quality assurance program for administering and monitoring requisite Superfund Quality
Assurance (QA) activities within ARCS remedial work and site assignments. Traditionally
quality assurance at Malcolm Pirnie was a project specific activity and the responsibility of
individual project officers.  The focus was on the quality of technical work products and did
not require extensive documentation.  The ARCS program activities covered include all
phases of RI/FS and RD/RA work and includes multiple MPI Offices and the subcontrac-
tor, CH2M Hill, Inc., and functions (eg. ARCS equipment  and storage, lab, program
management, site management).  A unique aspect of the ARCS and MPI approach is the
staff structure. Unlike other EPA contracts where staff were "dedicated" to the Contract,
                                       1-19

-------
MPFs staff is not dedicated. Personnel are drawn from MPI Offices throughout Region n
and assigned to ARCS Projects on an as-need basis.  This allows a cost-effective application
of appropriate skills and talents to ARCS projects on a timely basis. This program is now
expanding to include preremedial work assignments and subcontractors.
With the initiation of the ARCS contract, a Quality Assurance Program Plan (QAPP) was
established which identified the administrative structure, oversite functions, and responsibili-
ties and process of technical and QA reviews.  A Quality Assurance Project Plan (QAPjP)
was  also  established  defining how QA functions,  activities, and  duties were  to be
implemented.  The QAPP documents the structure.  The Program Management  Office
(PMO) receives work assignments and administers the Program. Within PMO, the PMO
Quality Assurance Manager oversees the ARCS QA program. Site Managers are located
in different Offices in New York or New Jersey, depending on site location and available
resources.  The site manager is responsible  for all activities concerning a site.  These
activities include establishment of a site specific QAPjP and Field Sampling Plan (FSP) plus
additional documents,  when  needed, for the site  Work Plan.   The  site manager  is
responsible for arranging adequate technical input including technical reviews. Each site is
also  assigned  a Quality Assurance Officer (QAO) whose duties are to assist the site
manager on QA issues, and review and forward the site QAPjP and FSP to the PMO QA
Office for review.  The PMO QA Office reviews the QAPjP and FSP and may return the
documents with written review comments to  the site manager.  Figure 1 illustrates the
process of review of FSP and QAPjPs.  When these comments have been addressed or
resolved, the documents are returned to PMO QA for forwarding to EPA for approval.  A
PMO QA office provides coordination, QA reviews, QA training, and conducts and oversees
audits.  Within the MPI QA program, each site QAO is responsible for conducting field
audits of FSPs and QAPjPs. The PMO provides oversite audits of QAO's, field audits, and
management system and file audits of the ARCS Program, subcontractors, and components
within the program (i.e. ARCS  equipment storage  and staging facility,  non CLP lab
activities). The PMO QA also oversees data validation services for project data and  may
audit laboratories when non CLP laboratories are used in projects.
To manage effectively QA activities within the ARCS program, an MPI ARCS Quality
Assurance Procedures manual was established This QA Manual specifies how QA is to be
administered in the Program, the duties and responsibilities of the PMO QA and site QAO
personnel and QA standard operating procedures (SOPs) for carrying out QA functions for
the PMO QA Manager and the site QAO's.  Table 1 is The Table of Contents of the QA
Manual and lists of the QA SOP's currently in use. QA and technical details are derived
from EPA technical documents(3) whenever possible. A training session was conducted for
site QAO's and site managers, when the MPI ARCS QA manual was implemented.
Technical memoranda, reviews and audits from EPA are distributed to all site managers and
QAO's and those affecting QA are to be included in the QA manual in a Technical
Memoranda section.  In addition,  MPI has now initiated a new phase, whereby site
                                       1-20

-------
managers convene to discuss status, trends, problems and resolutions of QA and technical
issues within the ARCS program.
The MFI QA Program to date, has been relatively successful Problems noted have usually
been the result of short time leads, new personnel (EPA and MPI), normal  project
problems, technical complexity of each site, changing objectives and DQOs.  The ARCS
Program has continued to expand to include more sites, PRP oversite assignments and
preremedial work.  These tasks will require more audits and an extension of MPI's QA
Program.
MPI's audits have observed how difficult it has become to incorporate the growing number
of technical advisories, procedures, memoranda, etc, into a manageable Work Plan for a site.
In addition, the extensive paper trail necessary for all operations, field sampling and sample
booking into the CLP, and  the long lead term for booking SAS and RAS presents real
challenges for all involved. As MPI work assignments shift into RD/RA implementation
and preremedial work, so will the focus of QA activities to assure that quality data and
quality work are attained.
CONCLUSION
In conclusion, the MPI ARCS QA Program has provided a workable, cost effective program
for administering QA activities in the ARCS Program. The key components in the process
are the site QAO, the PMO QA manager and establishment and use of an MPI ARCS QA
Manual  Use of the Manual, in addition to reviews and audits, provide an effective QA
Program.
                                       1-21

-------
                                REFERENCES

1.     U.S.E.PA., 1990. Long-term Contracting Study for Superfund. Public. No. 9242.6-
      07FS.

2.     U.S. E.P A., 1984. "Policy and Program Requirements to Implement the Mandatory
      Quality Assurance Program," EPA Order 5360.1., Office of Research and Develop-
      ment

3.     CERCLA Quality Assurance Procedures Manual, Region n, U.S.E.PA.

4.     U.S.E.PA. 1990.   Quality Assurance/Quality  Control  Guidance for  Removal
      Activities. EPA Public. No. EPA/540/G-90/004.

5.     U.S.E.PA. 1991.  Preparation Aids for the development of Category  I Quality
      Assurance Project Plans. EPA Public. No. EPA/600/8-91/004.

6.     U.S.E.PA. 1991. Preparation Aids for the development of Category n Quality
      Assurance Project Plans. EPA Public. No. EPA/600/8-91-/004.

7.     U.SJ2.PA. 1991. Preparation Aids for the Development of Category n Quality
      Assurance Project Plans. EPA Public. No. EPA/600/8-91/005.

8.     U.SJLPA. 1987.  Data Quality Objectives for Remedial response activities. EPA
      Public No. EPA/540/G-87/003.

9.     U.SJLPA. 1988.  User's Guide to Contract Laboratory Program.  EPA Public No.
      EPA/540/8-89/012.

10.    U.S.E.P A. 1988.  Guidance for Conducting remedial Investigations and Feasibility
      Studies Under CERLIA.  EPA Public. No. EPA/540/G-89/004.

11.    U.S.E.PA 1990.  Guidance for Expediting Remedial Design and Remedial Action.
      EPA Public. No. EPA/540/G-90/006.
                                      1-22

-------
    Review
  Team Leader
  Site
Manager
    Review
     Team
   Members
ro
CO
Site
QAO
Operations
Manager
*
PMOQA
Manager
Complete
1

]
1 1
Yes No
J 1
r
USEPA
                            ARCS II Quality Assurance Document Review System
                                                              to
                                                              c

-------
      TABLE 1. MALCOLM PIRNIE QA PROCEDURES MANUAL - ARCS H

                           TABLE OF CONTENTS

                                                                       Page

 1.0    INTRODUCTION                                                  1-1
       1.1    Quality Assurance in Malcolm Pirnie                             1-2
       12    Quality Assurance Project Plans                                 1-3

 2.0    MALCOLM PIRNIE QA PROGRAM                                  2-1
       2.1    Individual QA Task Assignments                                 2-1

 3.0    IMPLEMENTATION/DOCUMENTATION OF QA PROCEDURES       3-1
       3.1    Standard Operating Procedures                                  3-1
             3.1.1 Preparation and Format of Standard
                    Operating Procedures                                   3-1
             3.1.2 Preparation and Format of Audit Reports                    3-2
             3.13 Numbering System of SOP's                                3-5

 4.0    SITE QA OFFICER                                                 4-1
       4.1    SOP Preparation and Review                                    4-1
       4.2    Field Sampling and Analysis Plan Preparation                      4-1
       43    QA Project Plan Preparation                                    4-1
       4.4    Data Validation                                               4-2
       45    Field Auditing                                                4-2
       4.6    Lab Auditing                                                 4-2

 5.0    PMO QA MANAGER                                               5-1
       5.1    SOP Review and Approval                                      5-1
       52    Field Sampling Plan - QA Project Plan Review                     5-1
             and Approval                                                 5-1
       53    Auditing - Site QA Officers/Field, File, Program                    5-1


                           LIST OF APPENDICES

Appendix     Description

  1          Site QA Officer Specific Tasks

  2          PMO QA Manager Specific Tasks

  3          Stylized example of a Technical SOP

  4          Standard Operating Procedures (SOPs)

  5          Technical Advisory Memos

  6          QA References

                                    1-24

-------
     TABLE 1. MALCOLM PIRNIE QA PROCEDURES MANUAL - ARCS H

                           TABT.F. OF rONTENTS
                                 (Continued)
SOP Title                                                     SOP No.
PMOQA

Procedure for the QA Technical Review of FSP-QAPjP   MP-PMOQA-001-9/90
Management Systems - QA Audit                     MP-PMOQA-002-9/90
Contents of a PMO QA File                          MP-PMOQA-003-9/90
Procedure for Filing and the Contents of a              MP-PMOQA-004-9/90
 PMO QA Site Specific File
Procedure for Documenting the                       MP-PMOQA-005-12/90
 Quality Assurance Review of FSPs and QAPjPs
Procedure for Documenting Technical Quality Reviews    MP-PMOQA-006-12/90
Procedure for QA Interaction within ARCS H Program   MP-PMOQA-007-12/90
 Between MPI and CH2M Hill
Basic Procedures for Providing Quality Assurance        MP-PMOQA-008-1/91
 to Remedial Design and Oversight of Remedial
 Design by PRP's

SQAO (Site Quality Assurance Officer)

Procedure for a Technical Systems - Field Sampling       MP-SQAO-001-9/90
 Audit
Procedure for Completing Field - CLP Paperwork        MP-SQAO-002-1/91
                                    1-25

-------
            A NATIONAL QA STANDARD FOR ENVIRONMENTAL PROGRAMS
                FOR HAZARDOUS  WASTE MANAGEMENT ACTIVITIES
Gary L. Johnson and Nancy V. Wentworth,  Quality Assurance Management Staff
(RD-680),  U.S.  Environmental  Protection  Agency,  401  M  Street  SW,
Washington, DC  20460
ABSTRACT

The clean-up  of Federally-owned facilities contaminated by  mixtures of
hazardous  chemical  and radioactive  wastes involves  critical  decisions
based on environmental data.   The  Federal Government,is currently using
several  different standards  or sets of  requirements, including  U.S.
Environmental  Protection  Agency  (EPA)  guidance  for  establishing  the
quality assurance and quality control (QA/QC) procedures for these sites.
These standards defined the criteria for  the QA activities and documenta-
tion required,  the content and format of the  documentation,  and who was
responsible for them.   Shortcomings  in  these  standards or requirements
have led to  efforts  by several Federal groups  to develop a  uniform,
consistent  standard  that  produces  the needed  type  and  quality  of
environmental data in  a  more cost-effective manner.  These  efforts are
being conducted under  the  auspices of the American Society  for Quality
Control (ASQC) and involve participation  by EPA,  the Department of Energy
(DOE),  Department of Defense (DOD), Nuclear Regulatory Commission (NRG),
and others in the contractor and regulated communities.

This paper describes  the progress which has been made toward establishing
a consensus standard and associated requirements  for use by DOE, EPA, and
others for hazardous waste management activities.  The standard proposes
two distinct but related levels for management and  technical activities,
which include,  respectively,  the organizational  structure,  policies and
procedures, and roles  and responsibilities needed to conduct  effective
site  operations, and  the  project-specific Quality Assurance  Program
activities necessary to produce the desired data  quality.  It is expected
that the "harmonization"  of QA/QC requirements will  significantly improve
the cost-effectiveness of hazardous waste management activities involving
environmental data operations, as well as provide the basis for a needed
revision and expansion of EPA guidance.
INTRODUCTION

The emergence  of hazardous  waste remediation  as  one of  the  principal
environmental issues of the 1980s has resulted in large-scale environmen-
tal sampling and analysis programs to characterize  the waste sites and to
select  and  implement  appropriate  remedies.   Quality  assurance  (QA)
programs developed  in  support  of the environmental programs have varied
widely in content and  in application.   As the  responsible authority for
implementing environmental regulations,  the U.S. Environmental Protection
Agency (EPA) has mandated a QA program for its environmental programs in
EPA Order  5360.1,C1)  issued  in  1984.    This  Order  directed that  all

                                   1-26

-------
environmental data operations conducted by or for EPA in support of Agency
decision-making develop and implement an acceptable QA program.  Prior to
the Order, Agency requirements were manifested in two documents developed
by the Quality  Assurance Management Staff  (QAMS) : QAMS-004/80,<2)  which
discussed QA Program Plans,  and QAMS-005/80,(3)  which defined QA Project
Plan requirements  for individual  data collection activities.   Both of
these documents were issued  in 1980 and became the de facto standard for
EPA QA requirements for all  environmental programs.

While the EPA Order establishes specific  authorities and responsibilities
for QA,  it  does so almost entirely within  the  context of EPA organiza-
tions.    Externalization  of  EPA QA requirements  was  accomplished largely
through the contracts regulations, found in 48CFR15,(*) and the financial
assistance regulations for grants, cooperative agreements, etc., given by
40CFR30(3) and 40CFR31(6).   Both sets of regulations  limited the require-
ments for QA essentially to QA Project Plans for individual environmental
data collection activities.  The full scope  of the EPA Order has not been
extended  to  the regulated community, and EPA guidance to implement the
Order was not officially distributed outside the Agency.

Moreover,  until recently,  QA was  not  generally  included  in  specific
rulemaking by the  agency.   Air quality regulations  have included QA and
quality  control (QC)  specifications since the  1970s, but  the  ongoing
rulemaking to incorporate Chapter One into SW-846(7) represents the first
explicit  inclusion of QA requirements  in  a regulation.   Since  neither
QAMS-004/80  nor QAMS-005/80 has been revised or  updated  since  their
initial  issue,  the public was  not aware of the  changes to  the  EPA QA
program as  it has  grown  and expanded.  Consequently, interpretations of
EPA QA requirements became varied  as  EPA programs,  such as RCRA and
Superfund,  developed  their own expanded requirements based on specific
program needs and the ten EPA Regional Offices responsible for implement-
ing the hazardous waste programs under RCRA and CERCLA developed their own
interpretations  of QAMS  QA  Program Plan and Project  Plan  guidance and
program office guidance.   In the meantime, the general public continued to
use  QAMS-004/80  and  QAMS-005/80,  and  to  interact  individually  with
Regional Offices on their interpretation.   Consequently, ten years after
issuing  QAMS-004/80  and  QAMS-005/80,  EPA  was  faced  with  multiple
interpretations  of QA requirements  documents  which do not  reflect the
current vision  of QA in the Agency.  The need  for  new QA guidance from
QAMS provided impetus for EPA to examine the experiences of the past ten
years and to develop  focused criteria upon which such guidance could be
based for use in the next ten years.

The situation  in other Federal agencies was not dramatically different
from that of EPA.   The U.S. Department of Energy (DOE)  uses NQA-1{8) as its
standard for QA requirements. The Nuclear Regulatory Commission bases its
QA program on 10CFR50, Appendix B, which is  identical  to NQA-1.  The U.S.
Department of Defense has used various MIL  Standards as well as NQA-1 in
its Installation Restoration Programs.   NQA-1  was developed for nuclear
facilities  and  its  application  to  environmental   programs  has  been
difficult to accomplish effectively.  Moreover,  different field offices in
DOE,  for  example,  like  their  EPA Regional  counterparts,  have  taken
differing interpretations of what in NQA-1 is applicable to environmental
programs.


                                   1-27

-------
Currently, DOE, DOD,. and other Federal facilities are generally responding
to multiple  sets of QA requirements while  trying  to  satisfy CERCLA and
RCRA  regulations.   Often this  has  meant preparing  two  sets of  QA
documentation,  one  to  satisfy  the  EPA  Region   (usually  QAMS-005/80
requirements and format) and one to satisfy the  "owner" agency (DOE, DOD,
etc.). which has often meant NQA-1 requirements and format.  At the least,
it has meant having to prepare one document which satisfactorily addresses
the expectations of the  EPA  Region and  the  "owner"  agency.   This has
resulted in costly and time-consuming duplication of effort.   In addition,
the perception of  inconsistent and often conflicting QA requirements has
created confusion  and frustration in the regulated community.  It became
increasingly  clear that  action  was needed to  bring  some order to the
process.   The  staggering  cost  of  clean-up  at Federal facilities has
focused considerable public pressure on all agencies involved, including
EPA, to plan and carry out  cost effective clean-ups.
THE HARMONIZATION PROJECT

In the  fall  of 1989,  an initiative was begun which would, as a minimum,
attempt to "harmonize" the varied QA requirements  into a single, uniform,
consistent set for application to environmental programs.   This effort was
conducted under the auspices of the American Society for Quality Control
(ASQC)  and  included  active participation  by EPA,  DOE,  DOD,  NRC,  and
representatives of the  contractor  community.   A Work Group and a Policy
Group were formed in the  spring  of  1990  to pursue the harmonization
effort.   The  Work Group was  composed of  experienced QA professionals
representing EPA,  DOE, and Federal contractors, who would make the initial
efforts of harmonizing current QA requirements.  The Policy Group included
senior officials at EPA,  DOE, DOD, and NRC, and senior QA consultants, who
would guide the efforts of  the Work Group.

It became  apparent very quickly  that a new, national consensus standard
would be the most effective  way to harmonize the existing QA requirements.
Moreover,  those  engaged in the  harmonization effort were  committed to
producing a standard which could  encompass the broad scope of environmen-
tal programs without  being  overly  prescriptive.   Emphasis was  placed on
defining WHAT the  requirements should be, not the HOW TO or BY WHOM.  This
recognized that a detailed, omnibus standard could not meet the needs of
all Federal agencies.  Their missions are too diverse. The Work Group and
Policy Group decided  early  to  retain as  much flexibility  as possible in
the standard and not try to prescribe format or detailed specifications.
The differing  missions  and personalities of the  organizations would be
served best by allowing  them to define  detailed requirements for their QA
programs based on the  general requirements of this  standard.  For example,
the standard would require the use of QA Project Plans,  but it would not
prescribe the  content and format of the plans.   This decision would be
left to the implementing organization.

The outline for this  proposed standard was presented in September 1990 at
the ASQC Energy Division National Conference in Tucson.(9)   Subsequently,
a standard was drafted  and is currently undergoing public review and
comment as part of the  ASQC standard-setting process.  As  part  of this
process, EPA will develop new  guidance to implement the standard across

                                   1-28

-------
Agency programs.  Included  among the  new guidance documents planned are
replacements for QAMS-004/80 and QAMS-005/80.  This guidance is expected
to be available shortly after acceptance of the formal standard by EPA.
PROPOSED NATIONAL OA STANDARD FOR ENVIRONMENTAL PROGRAMS

Environmental  data have  an important role  in decisions  involving the
protection of the public and the environment from the adverse effects of
a variety of pollutants from waste operations and discharges.  To assure
that these data are of the appropriate type and quality to support their
intended use,  a  proposed standard has been  developed for environmental
programs.

The  proposed standard  includes the  basic  requirements  for a  Quality
Assurance Program  to plan,  implement,  and assess the  effectiveness of
multimedia data  operations to  characterize  environmental  processes and
conditions and to design, construct, and operate environmental engineering
systems.  Included in this Quality  Assurance  Program are the necessary QA
and QC activities to assure that technical  and  quality specifications are
satisfied.

As noted earlier,  NQA-1  has been utilized extensively for environmental
programs by  several  Federal agencies.   Many  of the fundamental concepts
found in NQA-1 have been incorporated into this standard and, in several
cases, improved.  The standard  also reflects the current vision of EPA1 s
Quality Assurance  Program as well  as  numerous  Total Quality Management-
based concepts which have gained wide-spread acceptance.

The proposed standard provides for  two  distinct levels of requirements to
be addressed:

            the organization (or institutional) level, and

            the technical/project level

In each distinct level,  there are specific  QA elements or functions which
must be addressed.

The elements at the organization level include defining the organizational
structure, policies and procedures, and roles and responsibilities for the
activities to be performed.  These  elements define what must be addressed
in order to  establish and manage an effective  Quality Assurance Program
for planning,  implementing, and assessing effective  environmental data
operations.   The  organization  level  elements provide  a framework or
infrastructure to  enable  consistent  quality  procedures  across  similar
environmental  projects.     The  organizational  level also  defines  the
requirements  for  necessary management  functions  to support  multiple
technical activities or projects.  These include procurement of services
and items, documents and records, use of computer hardware and software,
and  operation of  analytical facilities  and  laboratories.   The  basic
requirements for these functions are assembled into Management Systems,
Part A, and may  be viewed as  an umbrella under which technical projects
                                   1-29

-------
are performed.

The technical or project level consists of two parallel parts within the
framework  defined by  the  organization level  requirements.    Each  part
describes  the  project-specific Quality  Assurance  Program  activities
necessary  to  produce the desired type  and  quality of data  --  one  part
relates to process or site  characterization  and the other part relates to
environmental engineering systems.  Dividing technical/project operations
into  two  parts  reflects  the  differences  between  requirements   for
characterizing  an  environmental process  or   condition  (Part  B)   and
requirements  for  designing,  constructing,  and  operating  environmental
engineering systems  (Part C).

Characterization  of  Environmental  Processes  and Conditions,  Part  B,
contains the basic requirements for planning, implementing, and assessing
operations  to  collect,  analyze,  and evaluate  chemical,  biological,
ecological,  or  physical data in the  environment.   This also includes
compiling, modeling, and analyzing environmental processes and conditions
by mathematical or computerized methods.  The emphasis here is on planning
and  the approach is  based  largely  on  EPA's  Data Quality  Objectives
process.(10)   The  study design completes  the planning  phase and  the
standard requires that the data operations be implemented as planned and
documented.   There is a focus on  performance-based objectives  for  the
study so  that a measure of  success  may be readily  determined.   During
assessment of the results  obtained,  it is  recognized  that  results  from
environmental data operations may not completely satisfy the performance
objectives.  However, the assessment of data usability may enable the data
or part of the data to be used provided that the data user is willing to
accept  lees  confidence in  the results and  a greater risk  in making  the
decision for which the data were needed.

Environmental Engineering Systems, Part C,  provides the basic requirements
to  ensure effective design,  construction,  and  operation of  physical
engineering systems, and their components, which remediate environmental
contamination or  remove pollutants  from  multimedia discharges.   While
these requirements were not  originally within  the scope of the proposed
standard,  it  was  recognized  that environmental  engineering  systems  may
require  rigorous   QA  activities  to  assure  their  safe  and  effective
operation.  EPA guidance on QA for engineering systems is sparse and has
limited  application to hazardous waste remediation technologies being
developed today.  The requirements for Part C were drawn largely from NQA-
1, since the principal concern driving NQA-1 was the protection of public
health  and safety from the operation of nuclear  facilities.   While  the
magnitude  of  the  potential threat  is not as great,  it is  reasonable to
suggest that  the  inadequate design,  construction, or  operation  of  some
remedial  technologies could  pose health  and environmental  concerns.
Consequently, the need to include Part C in the proposed standard became
evident.

The Basic  Requirements contained in  the  current draft of  the proposed
standard are listed in Table  I.
                                   1-30

-------
NEXT STEPS AND SUMMARY

The standard-sett ing process is under way within the ASQC.  Public comment
on the proposed standard has been  invited.   While  it is not possible to
estimate a completion date for the standard at this time, the outlook is
optimistic.

The "proof"  of the  value  of this proposed standard  lies  largely in its
acceptance  and  implementation  by  Federal  agencies.   Here again,  the
outlook is promising. By involving key senior QA officials from the major
agencies in  the  development of the proposed  standard,  many issues have
already been addressed which otherwise would have posed serious barriers
to acceptance of the standard.

The value of this standard to future hazardous waste management activities
is severalfold.   First, the QA requirements of Federal agencies would be
the same.  Some programmatic differences may  still  occur between RCRA and
CERCLA,  for  example, but  essentially the QA "playing field" would be
level.  The  consistency of  a national consensus standard also  opens new
opportunities for increased standardization in other areas, such as, field
methods,  analytical procedures,  and  data  validation  and  verification
processes.  The  standard provides a basis for  increased cooperation among
the Federal agencies conducting waste remediation activities and for the
sharing of ideas and experiences.  Given the finite  resources  available
and the  magnitude  of the clean-up job ahead, no  one can afford to re-
invent the wheel.   The cost savings  that will result from this proposed
standard are difficult to estimate, but they could be substantial.

EPA assumes  that  an acceptable national  consensus standard for  QA will
emerge from ASQC.  Preparations for adopting and implementing the standard
across  EPA  programs are  under way,   including the  development of  a
comprehensive set of new QA guidance documents.  The transition to a new
standard probably will not  happen  quickly.   In time, some programs will
likely  recognize the  added value and  benefits  of revising  their QA
programs to reflect the standard and the new guidance.  Not all programs
will have  to be  revised or  even  have  new  QA  Management  Plans (QAMP)
prepared, while some others may need more extensive changes.

This standard will help to forge new partnerships among Federal agencies
engaged  in  hazardous waste management  activities  and  to foster greater
efficiency and effectiveness in future work.
                                   1-31

-------
                               REFERENCES
(1)         EPA Order 5360.1, Program and Policy Requirements to Implement
            the Mandatory Quality  Assurance  Program.  U.S.  Environmental
            Protection Agency, 1984.

(2)         Interim Guidelines and Specifications  for Preparing Quality
            Assurance Program Documentation.   U.S.  Environmental Protec-
            tion Agency, QAMS-004/80, 1980.

(3)         Interim Guidelines and Specifications  for Preparing Quality
            Assurance  Project  Plans.    U.S.  Environmental  Protection
            Agency, QAMS-005/80, 1980.

(4)         EPA Acquisition Regulations.   U.S.  Environmental Protection
            Agency, 48 CFR 15, 1984.

(5)         Financial  Assistance   Requirements.     U.S.   Environmental
            Protection Agency, 40 CFR 30, 1983.

(6)         Financial  Assistance   Requirements.     U.S.   Environmental
            Protection Agency, 40 CFR 31, 1988.

(7)         Test Methods  for  Evaluating Solid  Waste,  Physical/Chemical
            Methods  (3rd  Ed.),  SW-846.   U.S.  Environmental  Protection
            Agency, Federal Register 55(27),  February 8, 1990.

(8)         ASME NQA-1,  Quality Assurance Program Requirements  for Nuclear
            Facilities, 1989 Edition.  ASME,  September 1989.

(9)         Blacker, Stanley M.  Harmonization of Quality Assurance First
            Focus: NQA-1  and  QAMS.   Proceedings,  17th Annual  National
            Energy Division Conference, Tucson, September 1990.

(10)         Draft  Information Guide on Data  Quality  Objectives.  U.S.
            Environmental Protection Agency,  1986.
                                   1-32

-------
                                TABLE I

                       QUALITY ASSURANCE PROGRAM
                            REQUIREMENTS FOR
                         ENVIRONMENTAL PROGRAMS
Part A.  Management Systems

      1           Management Commitment and Organization
      2           Quality Assurance Program
      3           Personnel Training and Qualification
      4           Management Assessment
      5           Procurement of Services and Items
      6           Documents and Records
      7           Use of Computer Hardware and Software
      8           Work Processes and Operations
      9           Quality Improvement
Part B. Characterization of Environmental Processes and Conditions

      1           Planning and Scoping
      2           Design of Data Collection Operations
      3           Implementation of Planned Operations
      4           Quality Assessment and Response
      5           Assessment of Data Usability
Part C. Design. Construction, and Operation of Environmental Engineering
Systems

      1           Planning
      2           Design of Environmental Engineering Systems
      3           Implementation of Engineering Systems Design
      4           Inspection and Acceptance Testing
      5           Operation of Environmental Engineering Systems
      6           Quality Assessment and Response
                                   1-33

-------
                 THE IMPACT OF CALIBRATION ON DATA QUALITY

                    Richard G. Mealy - Supervisor, Quality Assurance
                    Kim D. Johnson - Manager, Analytical Laboratory

                 Warzyn, Inc, 1 Science Court, Madison, Wisconsin 53705

ABSTRACT
The basic functional elements of the typical Quality Control (QC) program are the QC samples
used to  evaluate control of the analytical process.  The ability to meet acceptance criteria
associated with QC samples, as well as the acceptance criteria themselves, are directly influenced
by the calibration process.  Despite the critical role that calibration plays in the generation of
accurate data, it is the aspect of environmental analysis that is controlled the least, based on a
comparison of the calibration process established in the common regulatory programs.

As the requirements for litigation quality data become more restrictive, the  calibration process
will need to be re-evaluated, and  subject to more rigorous controlling measures. This paper
discusses the key aspects of the calibration process, and the strengths and weaknesses associated
with each.  Without more control, it will be  difficult to achieve data  comparability between
laboratories using the same referenced methods.

INTRODUCTION
The calibration process represents the initial controlling mechanism for the generation of quality
data, yet  there is a general lack of guidance regarding specific evaluation techniques for this
process.   One of the  drawbacks of providing  such little guidance is the potential loss of data
comparability, one of the chief Data Quality Objectives identified by the EPA.

This paper examines several critical  aspects of the calibration process,  and identifies  those
features that, if overlooked, can significantly impact the quality of the data generated. Initially, a
comparison of calibration processes, as outlined in the various regulatory programs, is presented.
In order to provide a more focused scope for this paper, the discussion is limited to the impact on
methods for the analysis of Volatile Organics, Pesticide/PCBs, and Semi-volatile Organics. The
concepts presented, however, can be extended to other organic and inorganic  analytical methods
as well.
                                           -34

-------
It is important to note that some of the issues  raised in this paper have been addressed in
regulatory programs that were not evaluated specifically for this paper.  The USATHAMA*
program, in particular,  has incorporated a  requirement that  calibration data be subject  to
statistical tests for both Zero Intercept and Lack of Fit, which serve to resolve some of the
problems associated with non-linear  data  and calibration  intercepts.   These problems  are
discussed in detail in this paper.

COMPARISON OF REGULATORY APPROACHES TO CALIBRATION
There exists a great deal of difference in the calibration protocols and requirements of the key
regulatory programs, including the 5001 and 600)2 series of EPA methods, those published in SW-
846^, and the Contract Laboratory Program*.  A summary  of calibration requirements of the
various regulatory methodologies is presented in Table 1.

In general, the 600 series of methods offers the least amount of guidance, and thus is the most
open to individual interpretation. More recent revisions to the 500 series of methods for analyses
conducted under the SDWA program introduce several new requirements that provide greater
control over the accuracy of the resultant calibration. As Table 1 indicates, wide variation exists
in the number of calibration standards required both within  and across the series of regulatory
protocols. One particularly important assumption that the 500,600, and 800$ series all share in
common relates to curve linearity.  In each of these methods, if the %RSD of response factors
associated with calibration standards is within certain criteria (10-35%), "then linearity through
the origin can be assumed." Clearly, there are widely ranging views regarding when the intercept
of a calibration curve deviates significantly from the origin. In keeping with the goal to establish
data comparability, there is a need to consider the  incorporation of a statistical technique to
provide an objective means of determining whether a particular set of data essentially has a zero
intercept.

In the event that %RSD criteria cannot be achieved, 3 of the 4 programs allow the user to simply
prepare a calibration "curve" from concentration vs. instrument response.  Unfortunately, there
are no requirements for the type of curve algorithm (linear regression, polynomial fit,  etc.)
allowed.

As cleanup criteria continue to evolve, this variability between the different regulatory  protocols
can have significant, adverse impact on the comparability of data generated by laboratories.  Due
to either regional or site specific preferences, analytical programs can be based on methodologies
from any of these programs. While each of the programs is considered to be designed to produce
quality analytical data, the differences between the calibration protocols will result in significantly
different data quality.
                                          1-35

-------
 In order to provide more control over the calibration process, each element of the process must
 be considered so that the most appropriate combination of elements is employed.  The basic
 "parts" of the calibration are shown below:

        • number of calibration levels
        • calibration algorithm
        • calibration levels
        • calibration acceptance criteria
        • effect of "curve-smoothing" routines

 The remainder of this paper focuses of detailed analysis of each of these sections. In particular,
 those aspects that potentially lead to inaccurate or biased data are discussed.  In addition, the
 areas of the methods that are open to interpretation, or require further guidance are identified.

 NUMBER OF CALIBRATION LEVELS
 Essentially,  as the number of calibration levels increases, the relative risk is reduced, as a better
 picture of the analyte's performance is obtained.  The  analytical run-time is also an important
 consideration in determining the number of levels to employ. For analyses with a relatively short
 analysis time, such as the majority of inorganic parameters, additional calibration levels do not
 represent a  burden to production. This is not the case, however, for most organic analyses, with
 routine run  times of 40 to 60 minutes. Laboratories are engaged in a constant struggle between
 quality and production. While an increased number of calibration levels would certainty improve
 the quality of the data, this is not always possible. The implications of establishing calibration
 "curves" with a minimum of data points are brought to light in the next few sections.

 CALIBRATION ALGORITHM
 While most laboratories default to the standard (least squares) method  of linear  regression
 analysis to develop a calibration algorithm, a wide  array of non-linear calibration technique
 options are  available.  These options including, polynomial fits, exponential and power curves,
 segmented fits, and even specific manufacturer options, are routinely provided as part of the
 software bundled with instrument data stations.

The most common approaches to quantitation use an average response factor (e.g., in GC/MS),
single  point quantitation (multi-component  analytes  such  as  PCBs),  and  multiple  point
calibration "curves".  Of these approaches, single point  quantitation has the greatest potential
for inaccuracy because the response from the single standard analyzed is  deemed to be
representative of the linearity of the analyte in question;  There are two sources of error in single
                                          1-36

-------
point quantitation.  First, any error in the preparation or concentration of the standard will
directly affect quantitation. In addition, if the level chosen actually represents a significantly non-
linear portion of the relationship between response and concentration, substantial bias will be
introduced.

The use of an average response factor is designed to normalize differences in response factors
over the calibration range.  The drawback to this approach is if only a single response factor
deviates significantly from the others, the bias is normalized by distributing an equivalent degree
of bias in the opposite direction over the other standards.

Non-linear data require more advanced statistical treatment. Typically, regressions of a higher
order, quadratic equations, or polynomial fits of the data are employed. The main precaution
associated with  these techniques is the  minimum number of data points required.   As the
minimum number of points required to  form a line is two, then a linear regression (1st order
polynomial fit) actually requires a minimum of 3 data points to be significant.  Similarly, with
each higher order equation, one more data point is required.  As with a simple linear regression,
the correlation coefficient must not be used as a measure of linearity. The correlation coefficient
only provides a measure of how well the data points fit the equation generated. Finally, as the
degree of non-linearity increases, the curve of a 2nd or 3rd order polynomial becomes parabolic
(Figures  IB, 2B). This results, at the upper  end of the curve, in two solutions for a  given data
point.  Unless the actual curve is carefully evaluated, the  analyst  may not even be aware that
multiple solutions are possible for a given response. The consequence associated with this type of
situation is that significantly inaccurate data could be reported.
                                                      /
Essentially, a linear regression results in the equation for a straight line, whereas polynomial fits
above the 1st order will result in the equation for a curve. The most recent versions of the 524.1,
524.2, and 525, GC/MS methods for the analysis of volatile and semi-volatile organics, specifically
allow the use of 2nd or 3rd order regression  equations if the response factor criteria cannot be
met. Figure 3 shows the curves that are associated with a linear fit as well as polynomial fits of
orders 2 through 5 for Sample Data Set #1. Note, in particular, the significant differences in the
curve fit to the data in the region between points D and E. If these higher level curves are used
for the Sample Data Set, serious inaccuracies would result  at the upper range of the curve-- the
recommended range for sample quantitation.
                                            1-37

-------
While, in specific cases, each of these statistical manipulations of calibration data can provide a
"better fit" of the calibration equation to the data, they can also have significant impact on the
quality of the data generated.  Essentially, with the number of statistical programs readily
available, an equation can be found that will provide a mathematical solution (i.e. "fit") to any
set of data.  Consequently, without a complete understanding of the actual effect on the raw data,
none of these statistical techniques should be used in the generation of data for regulatory
compliance.

CALIBRATION LEVELS
The specific levels that are selected for calibration can have a significant impact on the validity of
the calibration equation. Calibration levels should be established based on consideration of (1)
the range of  the levels, (2) the reportable detection limit, and (3)  the  linear range of the
analyte(s).  The majority of the regulatory programs reviewed provide little guidance with respect
to the range of calibration levels.  A generic statement is provided that indicates that the levels
selected should be based on the expected range of sample results.   In  some cases, the
"expected"  sample concentrations exceed  the working linear range of the detector.  In the
interest of obtaining accurate  results, it is more important to define the linear range of the
analyte and/or instrument, and dilute sample concentrations that exceed this range.

A wide calibration  range, based on only a few calibration levels, will nearly always result in a
correlation  coefficient greater than  0.995, which is frequently  used as the sole  calibration
evaluation criterion. In the example of Sample Data Set #2, the linear regression calculated
from  all 5 data points yields a correlation coefficient of  0.963.  If only the first two  and the
uppermost data points are used (10, 20, and 200) however,  the correlation coefficient is increased
to 0.999.  This is a consequence of the derivation of the correlation coefficient.

The relative difference between the concentration of the low level standard and the reportable
detection limit is critical to providing confidence in the accuracy of low level measurements. Bias
is more pronounced as the calibration curve approaches the detection limit for a particular
analyte. Consequently, if the low level standard is significantly greater than the detection limit,
then accuracy  in the proximity of the detection limit is compromised, because  linearity of
response has not been evaluated in this region. Ultimately, the detection limit itself may come
into question.   While the majority of the regulatory methods specify that the low level standard
must be prepared at a concentration "near, but above, the detection limit",  methods 524.1 and
524.2 allow  the low level standard concentration  to be as much as 10  times higher than the
detection limitl.
                                           1-38

-------
Finally, analytes have detector-specific linear ranges. In order to accurately evaluate non-linear
regions of the curve, there must not be a significant difference between the uppermost standard
(X) and the (X-l) concentration level.  The consequence of not considering this in a calibration,
is the user may fail to identify a parabolic curve. This is one of the consequences that can result
from establishing calibration levels based solely on the expected concentration range of the
samples.

Recent revisions to the 500 series of methods represent the first attempt (other than the CLP
program, where calibration levels are contractually  defined)  to provide stronger guidance
regarding the calibration range. Methods 524.1 and 524.2 require at least 3 calibration levels to
encompass a factor of 20 calibration range (i.e. 1 to 20, 10 to 200).  In addition, at least 4
standards are required to cover a range of a factor of 50, and at least 5 standards are required for
a range factor of 100.

CALIBRATION ACCEPTANCE CRITERIA
Once a calibration has been performed, there must be a set of criteria to determine if the curve is
acceptable for use in generating analytical results.  This is one of the key weaknesses in the
published regulatory methodology.  With the exception  of the CLP program,  the referenced
regulatory methods have only established acceptance criteria if the mean response factor is to be
used for quantitation. The alternative, if %RSD criteria (relative standard deviation of response
factors from the  calibration  curve) cannot be  achieved, is to  simply  generate a plot of
concentration vs. response or response factor.  This allows the generation of data without control
of data quality until the analysis of the first continuing  calibration standard, where a limited
measure of control is obtained. In addition to %RSD criteria, the CLP program has established
minimum response  factor criteria for  most analytes.   This  requirement is  associated with
confidence in the ability to detect the analyte, however, rather than in quantitation of the analyte.

As indicated in Table 1, even the  acceptance criteria  associated with continuing calibration
verification (CCV) offer little assurance of accurate quantitation.  The most  stringent CCV
acceptance criteria are found in method 502.2, which requires the analysis of a midpoint standard
to yield a response within _+_ 20% of that obtained for the same standard in the initial calibration.
In addition, this method requires the analysis of a laboratory fortified blank (LFB) per batch of
20 or fewer samples, fortified at a concentration of 20 ug/L.

For a set of data which is essentially linear, the mathematical basis of a linear regression attempts
to establish the midpoint of the curve as the point which deviates least from the linear equation.
The extent of the deviation then increases at the extremes. The deviation is absolute rather than
                                           1-39

-------
relative to concentration, which creates the greatest impact at the lower end of the curve. Due to
the magnitude of response associated with the highest calibration level, the relative effect is
minimal. In the case of strongly non-linear data, such as that in Sample Data Set #2, the point at
which the curve becomes non-linear (in this case, the upper calibration level) is central in the
minimization of deviation from the curve.  This effect is evident in Table 4, which indicates that
relatively minimal bias occurs in the upper calibration level, even considering such non-linear
data.

The relationship between bias  and concentration  has its greatest impact on  the continuing
calibration verification process (CCV). The concentration of the CCV is typically equal to the
midpoint concentration of the initial multi-point calibration.  With linear calibrations (typically
the norm), the midpoint level is associated with the least degree of bias from  the plot of the
calibration equation. Consequently, if the overall accuracy of the analysis is less than 20%, there
is a significant probability that the acceptance criteria for the CCV and the fortified laboratory
blanks can be met.

The correlation coefficient ("r") is the most commonly used statistical measure of calibration
acceptability. One longstanding  misconception is that this parameter also provides a measure of
linearity.  The correlation coefficient is a measure of the "goodness of fit" of a series of data
points.  Basically, the correlation coefficient can be viewed as a mathematical process which
determines the tightest ellipse that  defines a set of data. The more the ellipse resembles a
straight line, the higher the "r" value (to a maximum of 1.00). The more the data appear to be
randomly distributed, or the ellipse appears more as a circle, the lower the "r" value (to a
minimum of 0).  This effect is illustrated in Figure 4. Consequently, even a particular random set
of data can result in a high "r" value if the data range is such that the data can be described by a
tight ellipse.

Calibration acceptance criteria  should be  designed to evaluate  the relationship between  the
intercept of the calibration equation and  the importable detection  limit (RDL). The data in
Tables 2A and 2B, for Sample Data Sets 1 and 2 show significant negative bias at the low end of
the calibration. If, as required by most of the regulatory methods, the low level  standard is just
slightly  greater than  the actual RDL,  then the RDL would  clearly not be valid for these
calibration sets.  One requirement that should  be  imposed on calibration data is that the x-
intercept (expressed as concentration) should be no greater than 50% of the RDL. This will
serve to minimize the reporting of low level false positive results.
                                           1-40

-------
One final consideration regarding the evaluation of calibration data, is to consider the bias at
each calibration level that results from obtaining a concentration from the calibration equation
using the actual raw calibration data. The software in use today provides graphic representations
of the calibration data, but the plots are typically too small and the resolution too poor to be used
to accurately evaluate point-specific bias.

Each of the generally accepted calibration evaluation mechanisms should be considered no more
than a single data assessment tool, rather than an absolute indicator of calibration acceptability.
For example,  the correlation  coefficient, used frequently in the inorganic arena, can provide
misleading information if there is a significant range between the uppermost and  lower
calibration levels.

EFFECT OF CURVE "SMOOTHING" ROUTINES
With the advent of powerful software routines and instrument data stations, the analyst is now
provided with  a series of tools that can be used to "smooth" the fit of any curve.  While these
techniques certainly are not an element of the calibration process, their use is rapidly becoming
routine.  High-powered calibration algorithms are most often used without understanding the
mathematical  functions behind them as well as the limitations to their use and impact on a
particular data set. For this reason, these techniques are discussed here.

One of the most routine software options available is that of "forcing" the curve through the
origin. Theoretically, a blank should yield no response for a particular analyte. Due to signal-to-
noise considerations, however, this is rarely the case.  Many analysts, however, have been trained
that a curve should pass through the origin, and thus this option is selected. There are two ways
in which curves can be forced through the origin. The  first, is a simple mathematical formula
designed to result in a slope and zero-intercept. The other option is  a manual one, which is based
on the repetitive inclusion of (0,0)  data points until  the curve is eventually forced through the
origin.

Curve "weighting" techniques are often used to obtain a better fit  of the data points at either
extreme of the calibration range. Typically, the low end  of the curve is susceptible to poor fit of
the calibration equation. The most common weighting routine employed to improve the fit is a
1/X manipulation of the data.  Basically, each data point is weighted by a factor of the inverse of
the associated concentration. The result of this weighting, for the entire set of data, is a 91 point
curve vs. the original 5 point curve.  The results of this weighting are summarized in Table 3. The
table indicates that a significantly better fit is achieved at the  low end of the curve while affecting
the midpoint and upper range only minimally.
                                           1-41

-------
While each  of these techniques results in a better fit of the data  points to  the calibration
equation, they remain little more than data manipulation techniques.  In the generation of
environmental data, analysts must be trained to understand that the use of these techniques can
result in mis-interpretation of the data.

SUMMARY
In a regulatory climate that is increasingly concerned with Quality Assurance, most data quality
assessments remain reactive in that they rely on quality control information generated during the
course of analysis,  rather than prior to the analysis of environmental  samples.  The calibration
process should be viewed as the initial opportunity to assess the quality of data to be generated.
Consequently, there is a need for more structure  and guidance in the evaluation process in order
to provide analytical methods which ensure data comparability.

REFERENCES
1. U.S. EPA, 1988. Methods for the Determination of Organic Compounds in Drinkine Water.
      EPA/600/4-88,039. December, 1988.
2. U.S. EPA, 1985.  Methods for the  Organic Chemical Analysis of Municipal and Industrial
      Wastewater. 40 CFR Part 136, Appendix A. July, 1985.
3. U.S. EPA, 1990. Test Methods for Evaluating Solid Waste. Volume IB: Laboratory Manual,
      Physical/Chemical Methods. SW-846. 3rd ed., Revision 1. November, 1990.
4. U.S. EPA, 1990. Contract Laboratory Program, SOW 390.  Statement of Work for Organic
      Analyses, Multi-Media, Multi-Concentration. March, 1990.
5. USATHAMA, 1990.  United States Army Toxic and Hazardous Materials Agency. Quality
      Assurance Program. USATHAMA PAM11-41, rev. 0. January, 1990.
                                         1-42

-------
Table 1: Comparison of regulatory method requirements for various aspects of the
calibration process. |
^^^^^^^B 500 600 8000 CLP 1
lo. Standards
Low standard
Calibration
Range
nitial Calibration :
Requirement to
use mean RF
Initial Calibration:
Alternative to
RF
Cont. Calibration
Frequency
Cont. Calibration
Acceptance
Criteria
• 3 to 5
• 5 recommended
• 1 point allowable
with criteria.
• 6pre-set:525
• Near, but above
EDLs , to
o2-10xMDL
Range factor :
20 : 3 minimum
50 : 4 minimum
1 00 : 5 minimum
RSD< 1055 (502)
<2055 (508,524)
<3055 (525)
• Generate a plot of
peak height or
area response vs.
concentration.
o No acceptance
criteria.
• Daily (502, 508)
• every 8 hours :
(524, 525)
• 95 of initial
standard response :
±3098:525,524
±2095:502,508
3 minimum 15 minimum
Each analyte: near,
but above MDL.
• Expected range
of samples.
• Detector range.
RSD <1095 602,608
<3555 624,625
• Generate a plot of
peak height or
area response vs.
concentration.
• No acceptance
criteria.
Once daily
• 55 of initial
standard response :
1555:608
2055:625
• QC Check standard
analyte specific:
601 , 602, 624
Each analyte: near,
but above MDL.
o Expected range
of samples.
• Detector range.
RSD <2055
• Generate a plot of
peak height or
area response vs.
concentration.
• No acceptance
criteria.
• Once daily
• More frequently
for ECD methods.
Standard response
within ± 1 555
of initial response.
5 : general GC/MS
4 : 8 Semivolatiles
3 : Pesticides
1 : Multicomponents
Contractual
requirements
Contractual
requirements
o Generally : RSD...
<20.55K [GC/MS]
<1 0-1 555 [GC]
• No RSD criteria :
- 20 Semivolatiles
-lOVolatiles
Use mean RRF.
Every 1 2 hours
Generally :
Maximum 55D =
2595 (for most)
1-43

-------
Table2A: Sample Data Set*!
Table 2B: Sample Data Set *2
X Y RF
1 65,000 65000
2 140,000 70000
5 365,000 73000
10 680,000 68000
50 2,250,000 45000
RSDofRFs-17.3%
LSR=Y=43320X+ 110840
2nd Order- Y * -599.W2 + 75069X - 5816.7

X Y RF
10 650,000 65000
20 1,400,000 70000
50 3,650,000 73000
100 6,800,000 68000
200 9,000,000 45000
RSDofRFs = 17.3%
LSR= YM4071X+ 950580
2nd Order- Y= -238.2XA2 + 94494X - 356730
Table 3: Calculated X values for Sample Data
Set#l using both Linear regression
(LSR) and LSR weighted 1/X.
X
1
2
5
10
50
LSR
-1.1
0.7
5.9
13
49
I.SR fl/X)
0.4
2.0
6.7
13
46
Table 4: Summary of the biased obse
using different quantf afion tec
X LSR PF2 RF
1
2
5
10
50
-191%
-60%
19%
31%
-2%
6%
-2%
-1%
0%
0%
1%
9%
14%
6%
-30%
rved in Sample Data Set#l data
:hniques.
< — Single Point — >
Mid Low High
-4%
3%
7%
0%
-34%
0%
8%
12%
5%
-31%
44%
56%
62%
51%
0%
RF= Hverage Response Factor
LSR- Least Squares Linear Regression
PF 2= 2nd order Polynomial Fit

                                     1-44

-------
 3.00e+6
 2.00e+6 -
  «r
  M




  f
  at


 1 .OOe+6 -
 O.OOe+0
          y = 1.1084e+5 + 4.3320e+4x

          r= 0.995
     3.00«+6
     2.00e+6 -
     J
     1 .OOe+6 -
                                                O.OOe+0
                \i = - 5816.7 + 7.5069e+4x - 599.1 Ox"2

                r=0.999
        )     10    20    30    40    50    60

                 Concentration


         Rgure 1A: Sample Data Set#1,

                   Unear Regression Plot.
                  10    20    30    40    50    6C
                      Concentration
              Rgure IB: Sample Data Set#l,
                        2nd Order Polynomial Fit.
 1 .OOe+7
 8.00e+6 -
w 6.00e+6 -

 4.00e+6 -
 2.00e+6 -
 O.OOe+0
          r= 0.963
                 y = 9.5058e+5 + 4.4071 e+4x
                                                 1 .OOe+7
      8.00e+6 -
     w6.00e+6-


     a
     M
     «i

     K4.00e+6 H
      2.00e+6 -
                                                 O.OOe+0
                = - 3.5673e+5 + 9.4494e+4x - 238.19x*2

               r=0.999
                   100          200

                   Concentration
300
         Rgure 2A: Sample Data Set#1,
                   Linear Regression Plot.
100         200

Concentration
                                               300
               Rgure 2B:  Sample Data Set #2,
                         2nd Order Polynomial Fit.
                                                 1-45

-------
                                            Linear Regression
                                            2nd order polynomial fit
                                            3rd order polynomial fit
                                            4th order polynomial fit
                                            5th order polynomial fit
                                        ®= Data points
•2e+6
          10    20    30    40    50    60
       70
80    90    100
     figure 3: Plot of Caltoralion curves generated from various algorithms
             using data from Sample Data Set#1.
   0-
       2  '     10'       20
        Poor correlation
           r= < 0.800
10                 100
  EKcellent correlation
      r> 0.995
   figure 4: Illustration of the relationship between the range of data
           points and the correlation coefficient (r).
                                1-46

-------
             PROFICIENCY EVALUATION SAMPLE PROGRAM FOR SOLID WASTE



                              ANALYSIS:  A PILOT PROJECT








DAVID E.KIMBROUGH. PUBLIC HEALTH CHEMIST II, AND JANICE WAKAKUWA, SUPERVISING CHEMIST,



CALIFORNIA DEPARTMENT OF HEALTH SERVICES, SOUTHERN CALIFORNIA LABORATORY, 1449 W.



TEMPLE STREET, LOS ANGELES CALIFORNIA 90026-5698.








ABSTRACT



The use of Proficiency Evaluation (PE) samples for laboratory evaluation is an accepted practice lor clinical,



industrial hygiene, and drinking water chemistry laboratories. It has not been systematically applied to the



analysis of solid waste by environmental laboratories  This paper discusses the theoretical and practical



considerations involved in preparing a pilot PE sample program.  The project was designed to assess the



types and frequency of laboratory error, completeness of data  packages, and to identify logistical and



tracking problems that might occur in an ongoing PE program.








Two sets of PE samples were distributed among 319 environmental laboratories  accredited  by the



Environmental Laboratory Accreditation Program  (ELAP) for either PCB or elemental analysis. One set



consisted of five soils spiked with Arodor 1260, and the other set of five soils spiked with arsenic, cadmium,



molybdenum, selenium, and thallium.



This project is an attempt to evaluate the competence of the environmental laboratory industry providing



services in the state of California.  The data and statistical analysis for this study are presented here.
                                           I-47

-------
         TECHNICAL DATA REVIEW - THINKING BEYOND QUALITY CONTROL
                     Em P. Johnson - Manager, Analytical Laboratory
                     Richard G. Meaty • Quality Assurance Supervisor
                  Warzyn Inc, 1 Science Court, Madison, Wisconsin 53711

ABSTRACT

Considerable attention has been given to the concept of quality assurance (QA) during sampling,
laboratory testing, and  data reduction; however, few aspects of quality assurance have been
incorporated into the technical review of data that will be used to make critical project decisions.
This paper focuses on the QA aspects of technical review and presents an alternative to the
traditional multi-tiered data review process. In addition, techniques and methods used to perform
overall data assessment are described. The process includes:

     1.  Quality assurance in data processing
     2.  Technical review of data
     3.  Evaluation of trends and anomalies
     4.  Comparison of historical project information

The use of computer tools and relational databases can streamline the overall process.
INTRODUCTION

Most laboratories have implemented a multi-tier data review process as part of their internal
Quality Assurance/Quality Control  (QA/QC) program.   Typically, the review  moves up  the
management chain, beginning with activities such as simple math checks, passing into verification
of QC sample frequency and acceptance criteria (which are usually well defined) and then,
perhaps, some limited "technical" review of the data.

As environmental testing moves into the next phase of quality assurance, it is becoming evident
that analyzing increasing numbers and types of QC samples does not, in and of itself, ensure that
quality data are being reported. QC samples are routinely subject to more intense  scrutiny due to
their very nature, and the implications that they present to managers, clients, and the end users of
the data.  Simply put, errors are not being made on the QC  sample themselves, because these
samples are subjected  to very specific acceptance criteria.  Unfortunately, routine  environmental
samples rarely receive this level of effort.
                                          1-48

-------
With errors  come  legal liability.   The  analytical data generated by laboratories are used to
determine regulatory compliance of industry. The implications of not providing quality data that
can withstand the challenge of a  court of law are forcing the analytical laboratory industry to
improve its data review process.   Even data  validation  done under the Contract Laboratory
Program  does not sufficiently  provide  for review of data  in the technical sense;  it merely
contractually provides for review of specific criteria.
This paper presents an alternative to the traditional approach to data review.  Described are
specific techniques and examples that can be used to focus on the technical aspects of the data
review process, and the deductive rationale that must be applied in order to understand the "big
picture" needed for sound project decisions.
THE TECHNICAL REVIEW PROCESS

Laboratory analyses are critical in determining project direction, therefore the reliability of the
analytical data is essential. The financial implications of such decisions have prompted Warzyn to
develop an effective review of our results from a project and regulatory viewpoint.

In this alternative  approach, QA is  an integral part  of laboratory operations, rather than an
isolated entity. QA is  an interactive element in each phase of the analytical process including
technical review of data  and reports. Figure 1 illustrates this approach to QA.

Technical review is an interpretive process designed to evaluate the overall ability of the data to
satisfy the project objectives. Initially,  analytical results must be reviewed in relationship to the
other analytes  reported.  The purpose of this type of review is to attempt to identify trends,
anomalies, or interferences which can bias the overall usefulness of the data. The technical review
process begins at  the onset  of any project.  Recommendations  for testing programs and  data
quality objectives (DQO) are examples of project support services offered. This strategy moves
the laboratory  from a "black box" to a highly visible, integral part of the project team where
technical expertise is critical to:

         The selection of testing programs
         Regulatory assistance
         Development of project DQO's

The technical review process  incorporates an initial review of the testing program upon receipt of
the samples.  The reviewer evaluates  analyses requested to ensure project DQOs are being met.
                                            1-49

-------
As analytical  results  are  generated, initial  math check, QC review and supervisors' technical
release of data are performed within the operations unit of the laboratory.  Reviewers consider the
relative accuracy and precision of each anatyte when interpreting the analytical results. This may
assist in establishing  the  reliability of the  results before  proceeding with the overall project
interpretation.

The data entry process is a unique function which requires double key protocols for data entry. An
internal computer error-checking routine is employed to compare both data entries and generate
an exceptions  report.  The double key entry  greatly reduces the transcription error rates and cuts
down on  the  time required for non-technical review and the errors associated with report
preparation.

Once reports have been generated, the next phase of the technical review begins.  This review is
performed from  a technical perspective and  based on a wealth of environmental experience.
Laboratory performance  verification  (precision and assurance) is designed to determine the
quality of individual analyses.  The interpretive technical review ties all parameters together to
obtain an overall picture of the data quality.

A number of computer generated  reports  assist as  individual  parameter relationships  are
evaluated.  One of the "red herrings" that typically appear in Quality Assurance Program Plans
(QAPPs)  or Quality Assurance Project Plans (QAPjPs), is  the  discussion of historical  data
comparison as part of the review process.  While this technique is quite valuable in long term
monitoring situations, it is most often performed from memory rather than actual comparison to
previous data. Current technology allows the opportunity to store  monitoring data for future
comparison.  Advances in relational database software also provide  the capability for statistical
and trend analysis, modeling, and evaluation of regulatory compliance.

Figure 2 shows a historical comparison report which clearly presents data at a sampling  location
for the past four quarters.  With this  information the technical reviewer can trace anomalies or
request laboratory confirmation of any analyte in question. The development of  this historical
report format  has greatly increased trend analysis capabilities.  With real historical information
available, anomalies are easily identified.

Figures 3A and 3B illustrate a two stage process which shows how detailed historical reports can be
used to identify anomalous data and provide information to ascertain which specific parameters
are in question.   Initially, the  technical reviewer reviews the  summary data (Figure 3A) and
                                            1-50

-------
observes that the hardness value is significantly lower than recent historical data indicate for this
sampling point.  The alkalinity value, which closely correlates  to hardness,  does not exhibit a
similar trend.  The next step in this process is to compare these two results to the conductivity
value (Figure 3B).  The technical reviewer can now evaluate all parameters given as a whole. In
this case, the TDS and sodium values provide clues to assist in pinpointing the problem analyte,
i.e.,  hardness.  The  laboratory benchsheet for hardness is  then  reviewed  to determine if a
mathematical error was made.  In this case, all the parameters indicate that the source of the
anomaly is in the hardness value alone, which was indeed confirmed in an audit of the raw data.

Figure 3C represents a different  scenario.  In this case, there  is a significant reduction in the
hardness value from the previous quarter, and the sodium result is significantly higher than in the
previous quarter. The conductivity results, however, are not significantly different. The field log
notes would then be reviewed to determine if the sampling location might shed some light on the
hardness anomaly, such as water sampled downstream from a water softener. In this example, the
field observations confirmed that a water softener had recently been installed upstream of the
sampling point.

If no apparent error is found, the technical review section has authority to schedule an immediate
confirmation of the analyte(s) in question.  Other routine data relationships which are considered
during the review process include:

         Anion-cation balances and relationships to EC
     •   Comparison of theoretical and measured EC/TDS
                                                 /•
         Demand parameter relationships (BOD/COD/TOC ratios)
         Evaluate trace element data in terms of potential interelement interferences
         "Logical" VOC  degradation  patterns (landfill age vs. solvent breakdown product
         appearances).
         Confirm the presence/absence of common laboratory  contaminants such as:  solvents,
         phthalates, methylene chloride.
         Interpretation of data relative to detection limits and  dilutions.
         Close scrutiny of "rarely" detected anatytes.
     •   Relationship of detected analytes to potential source contamination (e.g., elevated lead
         and cadmium near highways).
                                          1-51

-------
Some projects require detailed comparison of analytes.  Figure 4 shows a detailed program which
calculates the ion balance for a monitoring well.  Upon further examination, the sum of the
parameters exceeds the measured TDS; the IDS value appears biased low. Also, by reviewing
historical data, it was determined that prior alkalinity values were approximately 250 mg/L.  With
this correction, an ion balance is achieved.

Another example of a computer database report which assists the technical review process is an
Exceedance Summary.  In Wisconsin, the DNR has "NR140" Preventive Action Limits (PALs)
for groundwater standards.  The Exceedance Summary report assimilates all historical data for a
site and compares the found values to that of the PAL and presents the comparison in a summary
table. Figure 5 is an example of this report.  In this example, the ch'ent is provided with the actual
sample results and any regulatory criteria.  Any value which exceeds a regulatory criterion level is
flagged appropriately on the report.

One of the most powerful  tools for a technical  reviewer is a database of compiled technical
information to help further their environmental knowledge. Common organic contaminants, trace
elements in natural  soils, and common inorganic interferences are a few examples of the type of
support this gives to the technical review group.  A central library is maintained, and pertinent
articles are routed throughout the QA and management staff.
TECHNICAL REVIEW - THE NEXT GENERATION

Quality is an evolutionary process. The next step in the ever-expanding quality assurance "tool-
box" is the development of an interactive database.  Rather than relying solely on practical
experience,  Warzyn  is currently  experimenting  with a user-friendly, menu-driven, software
program that integrates text and graphics in a relational database format.  The information being
electronically cataloged and cross-referenced includes the following:

         common analyte names and "aliases"
         sources of the analytes
         environmental fate
     •    chemical structures
     •    available analytical method summaries
         information regarding inclusion on various regulatory lists
         regulated levels for compliance
                                          1-52

-------
Figure 6 shows two "snapshots" of the computer screens which are available with the current
program.  These screens provide a wealth of information that can be assimilated by the technical
reviewer.  The information on sources and environmental fate can be used to evaluate whether or
not a site is likely to be contaminated  with  these analytes, and which long-term breakdown
products might be expected.  Our long  range goal is to offer strong technical support to all
locations within our firm.

We envision this program as an excellent tool for staff training as well.   Our  colleges  and
universities simply do not  prepare graduates  adequately for work in the environmental field.
However,  with modifications to this program detailed information regarding analytical techniques
can be presented in an intriguing, informative format.  The net effect would be to provide hands-
on, visual  training in critical interpretation of peak overlap, mass spectral identification, and the
impact of method interferences.  Each training module can be designed to include an interactive
"test" to help assess trainees' comprehension of technical information and concepts.
SUMMARY

Environmental decisions depend on the quality of the data used to make those decisions. There is
a growing need to look beyond quality control data and ask if the data "make sense". The aim of
the review process described in this  paper is to  improve the quality  of data generated by
incorporating QA  as an integral  part of laboratory operations.  Only with this  type of holistic
approach to data review can the "black box" aspect of environmental analysis be eliminated and
attention be focused on technical advancement with full support to project and client needs.
                                           1-53

-------
REFERENCES

"Interim Guidelines and Specifications for Preparing Quality Assurance Project Plans", QAMS-
     005-80, United States Environmental Protection Agency, EPA Document 600/4-83-004,
     1983.

U.S.  EPA, 1988 Methods  for the Determination of Organic Compounds in Drinking Water,
     EPA/600/4-88/039, December 1988.

U.S. EPA, 1979 Handbook for Analytical Quality Control in Water and Wastewater Laboratories,
     EPA-600/4-79-019 March 1979.

U.S. EPA, 1986 Test Methods for Evaluating Solid Waste. SW846 3rd Edition November 1986.

U.S. EPA Office of Solid Waste and Emergency Response.  Hazardous Waste Land Treatment.
     SW-874, April 1983, pp. 273.
                                        1-54

-------
     Quality  Commitment
  Wareyn's Quality Assurance Program
  selves as a critical link to laboratory operation,
  rather than as an ancillary function.
                   Project
               Development

           V ComDlete?.ess
                                           Sampleg
•Historical
 comparison
•Project-specific
 QC requirements
                                          Receiving!!
•Technical
 project review

                           •Technical review:
                            loc-ir. vs. COC
      •V transcription
        errors
                 • Review of test
                  selection
                  •Adherence
                   to SOPs
                  •Consultation
                   for Corrective
                   Action
                   •Blind PZ program
                       Sample
                    11 Analysis!
UllData
   Review
                        1-55
                                               Figure 1

-------
                            WARZYN ANALYTICAL LABORATORY RESULTS
                                LOCATION:  MADISON/WISCONSIN
                                                                   PROJECT NO:   30205  OORL
                                                                   CK'D:    APP'D:
                                                                   DATE ISSUED:
                                                                   PAGE:  1
                                           06/07/90   09/18/90   12/11/90   03/04/91
 Leach  MHZ       pH
                Conductivity @ 25 Deg C
                Alkalinity,  Total
                Biochemical  Oxygen Demand
                Carbonaceous BOD
                Chemical  Oxygen Demand
                Chloride
                Cyanide,  Total
                Hardness, Total
                Nitrogen, Total  Kjeldahl
                Total Suspended  Solids
6.49
5270
3610
3190
2690
4060
537
0.007
2400
89.8
198
6.26
7950
3300
4700
3600
6940
728
0.006
3560
166
520
6.68
9020
4950
5370
5490
6960
965
0.006
4160
159
128
5.63
10000
5200
6030
4840
7760
1040
<0.025
4260
158
350
Results in mg/L except elev(ft), pH (S.U.), conductivity  (umhos/cm).
                                               1-56
Figure 2

-------
                                HISTORICAL REPORT EXAMPLE

                          WARZYN ANALYTICAL LABORATORY RESULTS
                              LOCATION:  MADISON, WISCONSIN
                                       09/18/90     12/11/90

lysimeter 3    pH                            6.46         6.22
             Conductivity @ 25 Deg C       1510         1470
             Alkalinity,  Total              804          817
             Chemical  Oxygen Demand          64           48
             Chloride                         37           28
             Hardness,  Total                955          535
             Nitrogen,  Ammonia             3.06         0.37
             Nitrogen,  Nitrate + Nitrite   0.09        O.02
             Nitrogen,  Total Kjeldahl       4.1         1.64
             Phenolics, Total             0.066        0.036
             Sulfate                         38           78
             Iron                          7.58         20.4
             Manganese                     0.83         1.02
             Sodium                        20.7         14.6
             Solids, Total Dissolved       1020         1100
                                                                  PROJECT NO:  30205 OORY
                                                                  CK'D:    APP'D:
                                                                  DATE ISSUED:
                                                                  PAGE:  1
   I Historical  Hapdnessl
«•  IData indicates     I
   la low bias.        I
   I                    I
Results  in  mg/L except elev(ft), pH (S.U.), conductivity (umhos/cm).
                                            1-57
        Figure 3A

-------
                                  HISTORICAL REPORT EXAMPLE

                            WARZYN ANALYTICAL LABORATORY RESULTS
                                LOCATION:   MADISON, WISCONSIN
                                         09/18/90
Lysimeter 3   pH
                              6.46
Conductivity @ 25 Deg C       1510
Alkalinity, Total              804
Chemical Oxygen Demand          64
Chloride                        37
Hardness, Total                955
Nitrogen, Ammonia             3.06
Nitrogen, Nitrate + Nitrite   0.09
Nitrogen, Total Kjeldahl       4.1
Phenolics, Total             0.066
Sulfate                         38
Iron                          7.58
Manganese                     0.83
Sodium                        20.7
Solids, Total Dissolved       1020
12/11/90

    6.22
    1470
     817
      48
      28
     535
    0.37
   <0.02
    1.64
   0.036
      78
    20.4
    1.02
    14.6
    1100
                                                                    PROJECT NO:  30205 OORY
                                                                    CK'D:    APP'D:
                                                                    DATE  ISSUED:
                                                                    PAGE:  1
                                                                  (Alkalinity, con-
                                                                  Iductivity and TDS
                                                                  I results further
                                                                  I indicate low bias
                                                                  ion  Hardness.
Results in mg/L except elev(ft),  pH  (S.U.),  conductivity  (umhos/cm).

Benchsheets reviewed; math error  was  identified.   Confirmation  of  the Hardness
value was performed.
                                            1-58
                                                       Figure 3B

-------
                               HISTORICAL REPORT EXAMPLE

                         WARZYN ANALYTICAL LABORATORY RESULTS
                             LOCATION:  MADISON, WISCONSIN
                                      09/18/90
Lysimeter 3   pH
                              Deg C
Conductivity 


-------
  Sample Monitoring  Well Data
Based on theoretical


NOTE: Since the sum
of the parameters
enceeds measured
TDS, the TDS ualue
is biased IQUJ.


i
i NOTE: By reuietuing
historical data, it
could be determined
| that prior ualues
ujere in the range
of 250 mg/L.
This ujould result
in an ion balance.

\
pH X
Measured
7.35 *
EC ^4 2597^
TDS
435
EC, field EC
Theoretical

1404
628
%D

-46%
44%
TDS/ECm 0.17 0.24
TDS/ECt 0.31 0.45

PafaLuctcT
Alkalinity
HC03-
C03=
Chloride
Nitrate
Sulfete
Cf-ftlrrniTTi
rti ^jT"l**^ni TT'
i7>OC[^'1TTI
P^TT^^^pl^^
mg/L
^ 120
181
0
120
70
32
150
2.7
628
EC []im'hn]
203.0
asCaCO3
ssCaCOS
387.3
| 0.0
184.8
182.0
122.2
319.5
5.0
1404
"+" meq/L

3.49
2.63
6.53
0.07
12.72
is biased
high.

'-' meq/L
2.40
5.11
0.00
2.50
wmmm
10.00
Discrenancv- 77?
Ion Balance
           %Diff. =  12.0%
           Anior.s too low
           Cations OK
 Acceptance criteria= ±2.0%
Aoion £meq/L xlOO shniJd = EC
Bffected Parameter (assumed discrepancy source)
                 Subsequent impact on:
                    Theoretical             Prior Year
Alkalinity
Chloride
Nitrate
Sulfete
Calcium
Magnesium
Sodium
Potassium
^ mq/L
^ 256
277
168
250




TDS
709
724
796
758




EC
1634
1610
1598
1605




TDS/EC
0.43
0.45
0.50
0.47




Data
249
175
<0.1
109
81
26
139
7
                      1-60
                                          Figure 4

-------
                        Summary of Chapter NR 140 PAL

                  Concentration Attainments and Exceedances
Parameter
PAL    ES    SHLS 06/08/90   SWLS 09/14/90    SWLS  12/14/90
Chloride (mg/L)    125.   250.
Iron (mg/L)        0.15    0.3
                 *397.
                 *2.72
 221.
*5.46
"7.05
Motes:

(1) PAL = Preventive Action  Limit  (Chapter NR 140, Wisconsin Administrative
          Code).

     ES = Enforcement Standard  (Chapter NR 140, Wisconsin Administrative
          Code).

      * = Concentration  attains or exceeds Enforcement Standard Concentration.

(2) The concentration listed may not be actual PAL or ES exceedances depending
    on the location of the facility's Design Management Zone, Site Specific
    PAL's, etc.  What constitutes  an NR 140 exceedance is defined on a site-
    by-site basis.

(3) Does not include NR  140  Welfare parameters color and odor or organic
    health parameters.
                                     1-61
                                                                      Figure 5

-------
            Acrolein

 1. Byproduct of tobacco smoke.
 2. Thermal decomposition of fats/greases, i.e.
  restamunt kitchens.
 ....... ENVIRONMENTAL FATE ••••••••
 Under goes addition of halogens easily.
 Unstable-polymerizes rapidly in light or strong
 acid.
Polyurethane foams
Polyester resins
Intermediate for syn. glycerol
Military- poison gas mixtures
Aquatic herbicide
Warming  agent (Chloromethane
 refrigerant)
                                      I Structure ||
           Back to
            Index
                                         V       ?
                                         c=c—c
                                          I     I    I
                                         H   H  H
                                                                Other Names
 Allyl aldehyde
 Acrylaldehyde
 2-Propenal
 Agualin
  Regulatory Lists
IRCRA App.VIII
|AB1803(CA)
                                             MCL info]
                                         Analytical Methods
                                                        SOIL
                                        SW-84S: 8030: p urge* trap GC/FID
                                        SW-846:8240: purge Wrap packed col. GC/MS
                                           .............  NOTES .............
                                       IGC/M S m et h o d 6 24 is only ap prove d as a sere en
                                        for this compound. If quantitation is critical or
                                        lovHevel detection is required, the GC method is
                                        themethod of choice.
     1,2-Dibromoethane     il |stpuctupe"||

i....... ENVIRONMENTAL FATE ••••••••
 • Bio degradation occurs in 30-120 days at levels
  of 15-18 ppm
 • Degrades to Bromoethane
   Scavenger for lead in gas.
   Occurs at levels up to 0.0258
   (w/v)
   Grain and tree crop fumigant
   Waterproofing preparations
                                            H   H

                                       Br-C-C-Br
                                             I     I
                                            H   H
                                             MCL info]
                                         Analytical Methods
                                                                     Back to
                                                                      Index
                                                                Other Names
                                                             Dovfume W85
                                                             EDB
                                                             Ethyl en e di bromide
  Regulatory Lists
|SDVAList2
IRCRA Appendix VIM
                                                    WATER  —
                                     [EPA504: microerfractionGC/ECD
                                      EPA 502.1: packed P&T GC/HECD
                                      EPA 502.2: capillary P&T GC/HECD
                                     I EPA 524.2: capillary P&T GC/MS
                                                SOLID/WASTE
                                     I SW-846 8011: micrDetraction GC/ECD
                                      SW-846 8021
                                          1-62

-------
                                  TABLE 1
                 COMMON LABORATORY CONTAMINANTS
Substance
Use in Analytical Laboratory
Ll,2-Trichloro-2,2,l-Trifluoroethane
1,2,4-Trichlorobenzene
1,4-Dichlorobenzene
2.4.6-Trichlorophenol
2.4-Dichlorophenol
2-Chlorophenol
4-Bromofluorobenzene
4-Nitrophenol
Acenaphthene
Acetone
AJuminum
Arsenic
Benzene
Benzo(a)pyrene
Boron
Carbon Disulfide
Chlorobenzene
Chloroform
Chromium
Copper
Diethyl Ether
Fluoranthene
Freons (CClsF, CCbF2)

Iron
Mercury

Methylene Chloride
MTBE
Naphthalene
Pentachlorophenol
Phenol
Pyrene
Selenium
Silver
THMs

Toluene

Trichloroethene
Various Phthalates
Xylenes-
Zinc
Solvent for oil & grease/TPH extractions
Calibration compound, matrix spike
Calibration compound, matrix spike
Calibration check compound
Calibration check compound
Matrix spike compound
Surrogate compound for VGA
Matrix spike compound
Calibration compound, matrix spike
Extraction solvent
Matrix spike (high concentration)
Matrix spike (high concentration)
Matrix spike compound
Calibration check compound
Glassware
GC thermal desorp work (air)
Matrix spike compound
Extraction solvent, sample preservation
Cleaning solution/digestion reagent (COD)
Sample preservation (phenols)
Extraction solvent
Calibration check compound
Refrigerants (A'C, freezers) fire
    extinguishers
Matrix spike (high concentration)
Gas displacement/digestion reagent
    (COD &TKN) w
Extraction solvent
Solvent for many new 500 series
Petroleum distillate (pesticide spraying)
Calibration compound, matrix spike
Calibration compound, matrix spike
Matrix spike compound
Shampoo
Digestion reagent (COD)
Water supply system chlorination
by-product
Carpet glue, paints, extraction solvent.
    matrix spike, electrical tape
Matrix spike compound
Inks, plasticizers. plastics
General solvent, slide cleaning
Sample preservation, hand cream
                                     1-63

-------
                                 TABLE 2
        TRACE CHEMICAL ELEMENT CONTENT OF NATURAL SOILS
Element
Symbol
Common Range (ppm)     Average (ppm)
Aluminum
Antimony
Arsenic
Barium
Beryllium
Boron
Bromine
Cadmium
Cesium
Chlorine
Chromium
Cobalt
Copper
Fluorine
Gallium
Gold
Iodine
Lanthanum
Lead
Lithium
Magnesium
Manganese
Mercury
Molybdenum
Nickel
Radium
Rubidium
Selenium
Silver
Strontium
Thallium
Tin
Tungsten
Uranium
Vanadium
Yttrium
Zinc
Zirconium
  Al
  Sb
  As
  Ba
  Be
  B
  Br
  Cd
  Cs
  Cl
  Cr
  Co
  Cu
  F
  Ga
 Au
  I
  La
  Pb
  Li
 Mg
 Mn
 Hg
 Mo
  Ni
 Ra
 Rb
  Se
 Ag
  Sr
  Tl
  Sn
 W
  U
  V
  Y
 Zn
  Zr
  10.000-300.000
           2-10
           1-50
      100-3.000
         0.1-40
          2-100
           1-10
        0.01-0.7
         0.3-25
         20-900
         1-1.000
           1-40
          2-100
        10-4.000
        0.4-300

         0.1-40
        1-5,000
          2-200
          5-200
      600-6.000
       20-3,000
        0.01-0.3
          0.2-5
          5-500
         8x10-5
          5-500
          0.1-2
         0.01-5
       50-1,000

          2-200

          0.9-9
         20-500
         25-250
        100-300
       60-2.000
71,000

    5
  430
    6
   10
    S
    6.06
    6
  100
  100
    8
   30
  200
   30
    1
    5
   30
   10
   20
 5,000
  600
    0.03
    2
   40

   10
    0.3
    0.05
  200
    5
   10
    1
    1
  100
   50
   50
  300
Ref:  USEPA Office of Solid Waste and Emergency Response.
     Hazardous Waste Land Treatment. SW-874 (April, 1983) page 273.
                                 1-64

-------
7      QUALITY ASSURANCE STRATEGIES TO IMPROVE PROJECT MANAGEMENT
                                              by
              Tracey L. Vandennark, Maxwell Laboratories, Inc. — S-CUBED Division
                  Guy F. Simes, U.S. EPA, Risk Reduction Engineering Laboratory
            In  these times of ever-shrinking budgets,  the primary objective in effective project
     management becomes one of how to accomplish as much as possible for minimal cost.  There
     is no longer any margin for repeating work which was done incorrectly the first time, or for
     extraneous investigations which are not directly related to the objectives of the project Quality
     assurance, applied before, during, and after the inception of a  project, can vastly  decrease
     misdirected efforts and needless expenditures.

            Some of the key  aspects in the conceptual stages of any project are those involving
     identification of the types of decisions to be answered by or made as a result of the outcome of
     the project, specification of the project objectives, and identification of the uses for and quality
     of the  resulting data.  These are the  types of questions which need to be  answered before a
     quality assurance plan can be written. These concepts are inherent in the strategy embodied by
     the Data Quality Objectives (DQO) process, which, though originally developed for application
     to the array of problems associated with Superfund sites, may be applied with equal success to
     aspects of the RCRA program.

            Once a Quality Assurance Project Plan (QAPjP) has been written, a QAPjP review by an
     independent third  party  is often able to identify experimental  design flaws,  inappropriate
     experimental methods, or insufficient QA measures for either sampling or analytical procedures.
     The SITE Program has documented several examples of substantial monetary savings to projects
     which  were accomplished by employing QAPjP reviews prior to the initiation of any  sampling
     or analytical efforts.

            After the project is underway,  an audit is the primary means to ensure that the data meet
     the established project-specific QA criteria or that the project has not deviated from the quality
     assurance plan, and that overall technical  systems  are all in proper working order.  Two types
     of audits achieve these purposes: audits of data quality (ADQs) and technical systems audits
     (TSAs), respectively. These audits may be conducted by an independent third party,  or by in-
     house personnel. Regardless of who conducts the audit, however, management must place a high
     degree of commitment in responding to the findings of the audit and in rectifying any problems
     noted.

            With the incorporation of such quality assurance measures, it is possible to meet project
     goals within pre-established budgetary constraints. Without including appropriate QA procedures
     in a  project, not only will redundancy of work and cost overruns be likely, but overall project
     goals may be compromised.
                                              1-65

-------
g            BIAS CORRECTION:   EVALUATION OF EFFECTS ON ENVIRONMENTAL SAMPLES

         Marvin W. Stephens. Ph.D.. Vice President/Corporate Technical Director,
                 Michael A.  Paessun,  Environmental Regulations  Specialist

                            Wadsworth/ALERT Laboratories,  Inc.
                                  4101 Shuffel Drive,  NW
                                 North Canton,  Ohio 44720

         ABSTRACT

         September 25, 1990 began a  new era in  data evaluation.   Under the  new
         Toxicity Characteristic (TC)  rule,  as well as the Land Disposal  Restric-
         tions (LDR)  program,  a requirement for  the bias  correction  of TCLP data
         was implemented.  Since  that  deadline,  Wadsworth/ALERT Laboratories  has
         analyzed hundreds  of  samples for which bias correction  factors were
         generated and applied.  The spike  data  and the corresponding correction
         factors associated with  those samples have been  collated and evaluated.
         For  the  population of samples studied,  the  number  of  samples  having
         hazardous levels of contaminants and  those which became hazardous after
         applying the bias correction  factor was small compared to  those samples
         which were non-hazardous.

         This paper evaluates the magnitude of the corresponding bias correction
         factor for each analyte on the TC list,  the variability of the factor  for
         each  analyte,  and  the  likelihood it  will  change  a sample's  hazard
         classification.   A  comparison  of the individual correction factors  to  the
         recoveries of the respective analytes in control  samples  is presented to
         demonstrate  the possibility of  using the control sample recoveries  for
         bias determination.  And finally, analytical anomalies, especially for  the
         organic compounds,  are addressed.

         From these discussions, proposals are suggested to aid in making decisions
         for future sample analyses.   An estimate of the benefits and expenses of
         these options will  be  discussed.

         INTRODUCTION

         On March 29,  1990,  the  long  expected Toxicity  Characteristic  Leaching
         Procedure (TCLP)  was finally promulgated in the Federal Register(l).   The
         short phrase "A matrix  spike  shall be performed  for each waste and  the
         average percent recovery applied to the  waste characterization"  (Section
         10.3) sent Shockwaves throughout the  environmental laboratory industry.
         A flurry of  comments and questions prompted the EPA to issue corrections
         to the TCLP  on  June  29, 1990(2).  The  above statement was  replaced by
         another clarifying  statement,  "The bias  determined from the matrix spike
         determination shall be  used to correct the measured values."  (Section
         8.2).  Also  included was  the  formula (Equation 1) for calculating this
         corrected value (Section 8.2.4).
                                           1-66

-------
      Equation 1

                            Xc  -  100 (XjKR)

                  Where:

                  Xc    -     Corrected value
                  Xm    -     Measured value of unspiked sample
                  HR    —     Percent recovery of matrix  spiked amount
The Toxicity Characteristic (TC) for which the TCLP is to be used, became
effective for large-quantity generators of hazardous waste September 25,
1990.  Since  that date, thousands of  samples have been extracted using
this procedure and subsequently analyzed and the appropriate correction
factors calculated and applied.

Vadsworth/ALERT Laboratories  analyzed a portion of those  samples.   The
data for nearly 6000 determinations  for the toxicity characteristic have
been compiled to evaluate the effect  this requirement has had on the types
of samples this laboratory has encountered.   This paper will present data
on the magnitude and variability of the bias correction factors calculated
for these samples and  compare  it to data collected from control samples
run  in the  same TCLP  extraction fluids.    It  will also  evaluate  the
probable effect on future samples, suggest changes  in protocols, propose
the use of control sample recoveries for bias determination, and present
some interesting anomalies.

The data base for this study consisted of sample matrices from industrial
wastes,  soils,  oils,   sludges,  and waters.   All  samples  studied  had
undergone TCLP  extraction (filtration in the case of  waters),  had been
spiked  with  the analytes  on  the  TC  list, and  a correction factor
calculated following  the analysis.  Correction  factors  (CF)  as used in
this paper are  the decimal equivalents  to  the percent  recovery used in
Equation 1.  Table 1 indicates  the number of  analyses performed for each
analyte and the number of samples having a specific analyte concentration
in excess of the regulatory limit.  Of the 5871 determinations, only 0.9X
of the uncorrected results exceeded  the regulatory levels.  And it can be
readily  seen  that the  majority of  those samples  exhibiting a toxicity
characteristic are the result of metals contamination.  Not seen in this
table are another 8.31 of the results that exceeded the laboratory estab-
lished  method detection limits yet remained below regulatory levels.
(Many of  those  values were likely due  to  trace  levels  of metals in the
TCLP extraction fluid as demonstrated by the blank analysis.)  The remain-
ing 90.8% of the sample results  were  less than the method detection limit.

To determine whether an  individual correction factor will have an impact
on a given population of samples a new term was coined,  the Critical
Correction Factor (CCF) .  It is  defined in Equation 2.  The CCF allows one
to evaluate the potential impact of  bias correction on each individual
                                   1-67

-------
      Equation 2
                     CCF  -  Method Detection Limit
                                Regulatory Limit
analyte within a population of  samples  with no detectable contaminants.
It is simply the value at which a "non-detect" would be bias corrected and
thus  exceeds the  regulatory  limit.    As  the  method  detection  limit
approaches the regulatory limit, the CCF  approaches  1.   The smaller the
CCF, the  less likely that a component will  exceed the  regulatory limit
even if very poor recoveries are measured.   The  right-hand column of Table
2 presents the value of the CCF for each  analyte as  calculated for this
laboratory.  Since detection limits vary among  laboratories, the CCF will
change proportionally and must be calculated for each location.

The use  of the CCF  to  predict whether samples  within a data set will
exceed  regulatory levels  can be  demonstrated by comparing it  to the
magnitude  and variability of  measured correction  factors  (Table  2).   A
frequency distribution of the  CFs for each analyte was plotted to evaluate
its statistical behavior since a broad spectrum of matrices were included.
A normal  distribution occurred in  most cases.   Figures 1-6 show the
distributions for representative constituents from each analyte group.

For the metals, the  mean  for  the correction factors  varied from 0.86 to
1.04.  Mercury (Figure 1) had  the widest range within the data and was the
only metal where a CF was  found to be  less than the critical correction
factor.

The variation in the  averages of the pesticide correction factors was very
similar to the metals, 0.86 to 1.04. However, the standard deviations are
somewhat  larger, indicating the greater variability  within the measure-
ments expected for these analyses.  On the other hand,  the value for the
lowest calculated correction factor in  this group (endrin - Figure 2) was
nearly twenty times larger than the CCF for the compound.

The two herbicide residues had average  correction factors which were very
similar,  and the standard deviations were comparable to the pesticides.
The range within the silvex data (Figure  3) was wider  than the previous
constituents.  However, the CCF for both herbicides is so small (<0.001)
none of the results approached this level.

All  of  the volatile  organic  constituents showed  consistent  average
correction   factors, 1.01  - 1.06,  except  for methyl  ethyl ketone (MEK).
For several samples,  the correction factors for MEK (Figure 4) were in the
range of 4 - 8.  This variability caused the standard deviation to exceed
0.8.  Vinyl chloride (Figure 5)  also exhibited  a  large standard deviation
(0.47).  However, none of the calculated correction factors approached its
respective CCF since most of the recoveries were biased high.
                                   1-68

-------
Finally,  the  semivolatile  organic  compounds  had  the lowest  average
correction factors (0.53 - 0.74).  And the standard deviation for many of
the compounds were larger, which seems to be the case  for the analysis of
most semivolatile compounds.  Two of the compounds,  2,4-dinitrotoluene and
hexachlorobenzene (Figure  6), had CCFs much  larger than any of the other
compounds.  (This is due to  the fact that  the regulatory limit is only a
factor of three greater than the detection limit  [CCF - 0.308]).  In spite
of that, only  six instances  out  of the combined total of 249 determina-
tions where no contaminants  were detected exceeded the regulatory limit
when the correction factor was applied.

In all of the  above  discussions,  it has been assumed that the detection
limit  would be  bias  corrected  in  the same manner  as  any  measurable
quantities.  Since it  is  unlikely that a detection limit study would be
performed  for each  matrix  analyzed,  this  is   a  necessary assumption.
Therefore from the  above  discussions, it  can be concluded that for the
majority of matrix types  encountered in  this  laboratory the corrected
detection limits will  not  exceed regulatory  limits.   But it must also be
recognized that a few  constituents exhibit recovery characteristics that
warrant more careful review.  Of  the 288 mercury analyses,  only one sample
(0.3%) had a corrected detection limit greater than the regulatory limit.
There were  four such  occurrences  in the  116 analyses for 2,4-dinitro-
toluene  (3.4%).   And the  133  hexachlorobenzene  analyses  showed two
occurrences (1.5%).  None  of these percetages suggest a serious problem.
Several other  semivolatile compounds have relatively  low biases, but the
GCF in each case  is small  and is unlikely to be  a problem compound.  One
might expect  the MEK results to be  similarly affected,  but the bias is
toward  recoveries greater than  100%  which  tend not  to be a regulatory
concern.   Vinyl  chloride  results do  not approach the  CCF (0.05) even
though  its  very large standard  deviation accentuates the spread in the
recovery data  and makes it suspect.

Although it has been documented that some type of sample matrix effect may
exist, the  argument  for the need for bias correction must also consider
the effect  of  the TCLP buffer solution itself on the  ability to recover
each of the spiked compounds.  To evaluate this  effect, the  recovery data
for control samples  run with each batch of  samples was collated.   These
control samples  were prepared by spiking an aliquot of the TCLP buffer
blank  in the  same  manner as  the matrix spikes were  prepared.    These
control  samples  then underwent  the  same  extraction  or preparative
procedures  and analyses as  the  actual  samples.   Table 3 summarizes the
number  of  control  samples analyzed,   the  average  recovery,  standard
deviation,  and measured   recovery range for each analyte.   It should
immediately be noted  that  for both  2,4-dinitrotoluene and hexachloro-
benzene, the control sample recoveries in at  least one case were  less than
the respective CCFs.

To compare this data to the recoveries  of the matrix spikes (CFs), a graph
was  prepared  plotting the  pairs  of  mean  correction  factors  and mean
control sample recoveries for each analyte  (Figures  7 - 10).  For each
mean the range of two  standard deviations is indicated (a  95X confidence
                                   1-69

-------
interval).  A quick survey of the graphs shows the remarkable similarity
of the two sets of data.   Interestingly, the spike recoveries and control
sample recoveries for several of the compounds  previously discussed do not
stand out as being significantly different when compared in this manner.
The CCF for each analyte has also been included for comparison purposes.

Before formulating any conclusions, there are still a couple of signifi-
cant  anomalies  that should  be  discussed.    Methyl  ethyl ketone  had
recoveries that  frequently  exceeded 100X compared to  a standard purged
from deionized water.  This was true in the control samples and was even
more pronounced in the matrix spiked samples.  It is felt that this can be
explained by  the decreased solubility of the  MEK in  the  higher  ionic
strengths routinely present in the buffer.  This would then be even more
exaggerated in the extracts of many of the samples which contained soluble
salts.   Similar effects have been noted in  this  laboratory  during the
routine analysis of other water-soluble ketones and alcohols under similar
conditions.

Another  important  anomaly was found for four of the compounds from the
semivolatile fractions.   The  cresols,  2,4-dinitrotoluene, nitrobenzene,
and pyridine  showed no detectable  spike  recovery in  a small  number of
samples.   These  data were  not  included in the tables  since a  bias
correction factor  could not be calculated.   In such cases the recovery
data  could not  be used  to support or  reject  the  analytical results.
Instead, results had to be evaluated based on the generator's knowledge of
the process producing the waste sample, not on  recovery factors.  Ideally,
another method should be developed that is capable of quantitating these
compounds in the more difficult matrices.

CONCLUSIONS

Now, what can we now imply from  this data?  Can we just disregard spike
recovery data as being insignificant?   Though  it might  appear so on first
review,  the  opposite is  really true.   For  many  of  the  analytes,  the
information presented  indicates  there are potential  problem  areas that
need to be addressed.

For example,  if the sample in  question is being  characterized for the
first  time or  comes  from  a  process where  the  characteristics  are
continually changing, the  data presented suggests that the matrix needs to
be evaluated to determine the extent to which it may affect the recovery
of the designated analytes.  However, if it has been demonstrated that no
unusual sample matrix effects  are biasing the  data, there seems to be no
need to apply any correction factors.   The method bias can be represented
by  the  mean of  the control  sample  recoveries.   If a sample contains
concentrations of contaminants that approach a regulatory level,  and it is
felt  that  the bias  suggested  by the  control sample  recoveries do  not
represent this particular  matrix,  a more extensive matrix spike study for
those analytes may be considered.
                                   1-70

-------
It must be emphasized that only 0.9X of all constituents tested initially
exceeded  the  hazardous  classification limits.    When the  calculated
correction factor was applied to  the remaining data,  an additional 0.2X
exceeded  regulatory  levels,  including  the  seven samples  where  the
corrected detection  limit was elevated above  those limits.   Therefore,
nearly 99% of  all  the analyses completed by this  laboratory were still
classified as non-hazardous  following bias correction.  It can be argued
that there may be a large number of matrix types which are regulated but
are not included in this sample population.   This may be true and each of
these matrices needs to be  evaluated on its  own merit.   But as the data
base expands, it is predicted that it will become apparent that the need
to evaluate bias will be based more on the TCLP extraction fluid's effect
than on any sample matrix effects.

If the bias  correction requirements are to  remain,  it  is  proposed that
control sample  recoveries established  from  a statistically significant
database be used.  Figures 7-10 graphically  showed the differences be-
tween  the average  correction factor  and   the  average control  sample
recovery value.  In all but six cases,  this difference is less than 0.05.
Four of the cases involve pesticides where the CCF  is very small because
of very good detection limits. These larger differences  (0.08 - 0.23) are
not statistically significant and would seem  to have no  effect.   MEK's
spike  recovery  is   larger  than  the  control  sample  recovery  so  the
difference is actually negative. But both averages are well above 100% and
any bias correction would only cause the result to be adjusted downward.
Nitrobenzene (a difference of 0.11) is  the only remaining compound where
the control  sample  recoveries might be seen as adversely affecting the
data from an enforcement standpoint.  Yet the CCF for  this compound is
still very  small (0.02)  and the  actual measured  value  (or an elevated
detection limit) would have  to approach forty  times the normal detection
limit before it would be significant.  As suggested before, any concentra-
tion approaching a regulatory limit should be  evaluated on an individual
basis.

What then are the options  and any potential detrimental effects?  From the
data used in this  study,  it  seems  there  is  very little opportunity for
significant abuse.   By evaluating the bias for a specific component based
on the recoveries  from control samples as described  in this paper, the
cost of analysis can be significantly  reduced  with minimal environmental
impact.  Another option for determining bias, which has been suggested by
many, is to  use  isotope dilution techniques, thus reducing the need for
the extra  matrix spike.   Though  reducing  the cost  and  supposedly any
duplicate sample preparation problems,  this  technique is only useful for
GC/MS methods.   But  if the control sample can give similar  information,
then the same result can be  achieved with even less expense  than isotope
dilution  and can also  be used  with  all  the methods.    If there  is  a
requirement that first time  or uncharacterized samples undergo some type
of matrix  spike recovery evaluation, very  few samples which may have
seriously  biased  data will  be  overlooked and  potentially harm  the
environment.
                                   1-71

-------
SUMMARY

To approach all samples that may be regulated by the Toxicity Characteris-
tic rule using the same bias correction requirement seems to be impracti-
cal and definitely unnecessary.   Though many argue that  the  process is
statistically unsound, with which  I agree,  the fact  that  bias may exist
should be recognized.  If the  data  being generated is to be compared to a
regulatory level that has been set  based on absolute recoveries, the bias
for a  given analyte will need  to be  considered.   But  from the  data
presented,  the bias  for many matrix  types can be  represented by the
recoveries  of  laboratory control  samples without  the added  expense of
another spiked sample analysis.

REFERENCES

1) U.  S. Environmental Protection Agency, March 29, 1990.  Hazardous Waste
Management System; Identification and Listing of Hazardous Waste; Toxicity
Characteristics Revisions; Final Rule.   40 CFR Parts 261, et al.  Federal
Register 55:   11798  - 11877.

2) U.  S. Environmental Protection Agency,  June 29, 1990.  Hazardous Waste
Management System; Identification and Listing of Hazardous Waste; Toxicity
Characteristics Revisions; Final Rule.  40 CFR Parts 261,  264, 265, 268,
271, and 302.  Federal Register 55:  26986 - 26998.
                                   1-72

-------
              TABLE  1   SUMMARY OF BIAS  CORRECTION SAMPLES
                                                    No. of Analytes
                                Number of             Exceeding
Constituent                   Spiked S*»p1«»g        Regulatory Levels

Arsenic                            289                      1
Barium                             290                      2
Cadmium                            300                     16
Chromium                           313                      5
Lead                               320                     19
Mercury                            288                      3
Selenium                           286                      4
Silver                             291                      0

Chlordane                           64                      0
Endrin                              64                      0
Heptachlor (& its epoxide)          64                      0
Lindane                             64                      0
Methoxychlor                        64                      0
Toxaphene                           64                      0

2,4-D                               74                      2
2,4,5-TP (Silvex)                   65                      0

Benzene                            159                      1
Carbon tetrachloride               158                      0
Chlorobenzene                      158                      0
Chloroform                         158                      0
1,2-Dichloroethane                 157                      0
1,1-Dichloroethene                 158                      0
Methyl ethyl ketone                157                      0
Tetrachloroethene                  158                      0
Trichloroethene                    159                      1
Vinyl chloride                     159                      0

Cresols                            118                      0
1,4-Dichlorobenzene                132                      0
2,4-Dinitrotoluene                 116                      0
Hexachlorobenzene                  133                      0
Hexachlorobutadiene                132                      0
Hexachloroethane                   132                      0
Nitrobenzene                       119                      0
Pentachlorophenol                  125                      0
Pyridine                           128                      0
2,4,5-Trichlorophenol              128                      0
2.4,6-Trichlorophenol              127                      0
                                    1-73

-------
               TABLE 2  SUMMARY OF CORRECTION FACTOR DATA
Constituent

Arsenic
Barilla
Cadmium
Lead
Mercury
Selenium
Silver

Chlordane
Endrin
Heptachlor
 (& its expoxide)
Lindane
Methoxychlor
Toxaphene

2,4-D
2,4,5-TP (Silvex)

Benzene
Carbon tetrachloride
Chlorobenzene
Chloroform
1 ,2-Dichloroethane
1 , 1-Dichloroethene
Methyl ethyl ketone
Tetrachloroethene
Trichloroethene
Vinyl chloride

Cresols
1 ,4-Dichlorobenzene
2 , 4-Dini tro toluene
Hexachlorohenzene
Hexachlorobutadiene
Hexachloroethane
Nitrobenzene
Pentachlorophenol
Pyridine
2,4, 5-Tr ichlorophenol
2,4, 6-Tr ichlorophenol
Average
Correction
Factor (gR)
0.95
0.87
0.88
0.88
0.89
1.04
0.92
0.86
0.86
1.04
1.02
0.90
0.89
1.00
0.80
0.78
.04
le .02
.02
.03
.05
.01
: .32
.01
.06
.01
0.57
0.62
0.53
0.66
. 0.62
0.56
0.74
0.58
0.59
tol 0.54
10! 0.62
Standard
Deviation of
Corr. Factor
0.093
0.117
0.072
0.070
0.103
0.212
0.192
0.069
0.193
0.202
0.162
0.250
0.208
0.198
0.166
0.213
0.134
0.091
0.071
0.086
0.110
0.156
0.823
0.104
0.185
0.474
0.213
0.120
0.158
0.172
0.114
0.163
0.221
0.275
0.185
0.169
0.205


Range
0.5 - 1.3
0.5 - 1.4
0.6 - 1.4
0.55 - 1.4
0.3 - 1.5
0.08 - 1.5
3.4 - 1.5
0.45 - 1.3
0.5 - 1.38
0.45 - 1.38
0.65 - 1.4
0.36 - 1.3
0.49 - 1.25
0.54 - 1.42
0.43 - 1.23
0.15 - 1.5
0.8 - 2.0
0.7 - 1.4
0.8 - 1.2
0.8 - 1.3
0.8 - 1.4
0.3 - 1.8
0.4 - 8.2
0.8 - 1.8
0.2 - 2.0
0.4 - 5.0
0.06 - 1.14
0.1 - 1.11
0.05 - 1.1
0.11 - 1.22
0.18 - 0.93
0.06 - 1.08
0.09 - 1.36
0.08 - 1.54
0.08 - 1.06
0.015 - 0.89
0.03 - 1.1
Critical
Correction
Factor
0.100
0.001
0.100
0.020
0.020
0.100
0.300
0.020
0.017
0.025
0.013
<0.001
<0.001
0.010
<0.001
<0.001
0.010
0.010
<0.001
0.001
0.010
0.007
<0.001
0.007
0.010
0.050
<0.001
0.005
0.308
0.308
0.080
0.013
0.020
0.002
0.008
<0.001
0.020
                                  I-74

-------
             TABLE  3   SUMMARY OF CONTROL SAMPLE RECOVERIES
Constituent

Arsenic
Bariua
Cadmitui
Chroniua
Lead
Mercury
SeleniuB
Silver

Chlordane
Endrin
Heptachlor
 (& its epoxide)
Lindane
Methoxychlor
Toxaphene

2,4-D
2,4,5-TP (Silvex)

Benzene
Carbon tetrachloride
Chlorobenzene
Chloroform
1,2-Dichloroethane
1,1-Dichloroethene
Methyl ethyl ketone
Tetrachloroethene
Trichloroethene
Vinyl chloride

Cresols
1,4-Dichlorobenzene
2,4-Dinitrotoluene
Hexachlorobenzene
Hexachlorobutadiene
Hexachloroethane
Nitrobenzene
Pentachlorophenol
Pyridine
2,4,5-Trichlorophenol
2,4,6-Trichlorophenol

1
Number of
Control
Samples1
94*
92
92
92
92
89
97*
92
49
49
49
49
49
49
55
55
62
62
62
62
62
62
62
62
62
62
95
95
95
95
95
95
95
95
95
L 95
L 95
Standard Deviations
Average
Recovery
0.93
0.91
0.92
0.92
0.93
1.09
0.98
0.89
0.9C
1.12
1.07
1.13
1.04
1.07
0.83
0.73
1.00
0.98
1.01
1.01
1.03
0.99
1.18
0.99
1.02
0.96
0.62
0.61
0.48
0.69
0.60
0.59
0.85
0.52
0.62
0.57
0.68
of Control Sample
Recoveries
0.113
0.066
0.045
0.045
0.057
0.113
0.162
0.056
0.206
0.157
0.147
0.153
0.148
0.224
0.171
0.218
0.081
0.088
0.079
0.084
0.097
0.178
0.323
0.082
0.094
0.202
0.123
0.097
0.171
0.170
0.105
0.125
0.234
0.219
0.245
0.120
0.124

Range
0.10-1.19
0.73 - 1.21
0.77 - 1.06
0.73 - 1.00
0.74 - 1.30
0.80 - 1.41
0.60 - 1.42
0.56 - 1.02
0.24 - 1.32
0.74 - 1.42
0.63 - 1.45
0.75 - 1.40
0.60 - 1.35
0.53 - 1.63
0.44 - 1.34
0.30 - 1.29
0.81 - 1.22
0.68 - 1.19
0.72 - 1.19
0.79 - 1.18
0.83 - 1.29
0.30 - 2.07
0.10 - 2.17
0.78 - 1.24
0.81 - 1.72
0.47 - 1.43
0.33 - 0.99
0.37 - 0.92
0.06 - 1.12
0.20 - 1.12
0.32 - 0.91
0.28 - 0.91
0.48 - 1.72
0.07 - 1.25
0.18 - 1.82
0.28 - 1.05
0.42 - 1.29
'Represents  duplicate analyses
'Includes  both ICP and AA furnace results
                                   I-75

-------
                    Figure 1
Frequency

80r
60
4O
20
      Mercury
Correction Factors
    (288 data points)
    OOMUOO
  0   02  0.4  O.B  O.B  1   t2   14  10  18   2   3

             Correction Factor



                    Figure 2
                               Endrin
                        Correction  Factors
                              (64 data points)
              0.6  OS   1   12  14  16  18  2   3

                 Correction Factor
                       1-76

-------
                    Figure 3
  5 -
                            Silvex
                     Correction Factors
                         (65 data points)
   0   0.2  0.4  0.6  0.8   1   12  14  1.6   1.8
                 Correction Factor
                    Figure 4
Frequency
 40
Methyl Ethyl Ketone
 Correction  Factors
     (156 data points)
       0.5   0.7  0.9   11   1.3   15   17
                Correction Factor
                      1-77

-------
Frequency
 60
  10 •
      Figure 5

   Vinyl Chloride
Correction Factors
    (169 data points)
      0.2  0.4  0.6  OA   1   12  14   t6  1.8
               Correction Factor
                   Figure 6
 10 •
                   Hexachlorobenzene
                    Correction Factors
                      (133 data points)
      02  0.4  0.8  O.8   1   12  14   16  18
                Correction Factor
                      1-78

-------
Figure 7 Comparison of correction factor and control sample recovery for metals.

2.0 -|


1.5 .

03
o 1.0-
o
(D
rr
0.5 •

o.





on
II
II


x
o Correction Factor 95% Cl
• Correction Factor Average
o Control Sample 95% Cl
• Control Sample Average
x Critical Correction Factor
Q on

O « Q ||
T" JB i5 1s ° *" i°
o o 0 D
o
X
x x
— i 	 X 	 1 	 1 	 X 	 1 	 X 	 1 	 1 	 1 	 X 	 1
         Arsenic
Barium    Cadmium   Chromium
Lead
Mercury    Selenium     Silver
Figure 8 Comparison of correction factor and control sample recovery for pesticides and
herbicides.
                 o  Correction Factor 95% Cl
   2.0 T         •  Correction Factor Average
                 a  Control Sample 95% Cl
                 •  Control Sample Average
                 x  Critical Correction Factor
1.9 •


01.0.
03
OL.

0.5


0
<
_c
c





• <
i


,1


• c
]


\
I


1 <
>c



t

1
1 e> *•



<

1
>[

i


C
1

1

<
1
o
>

1


»
' _ o
<
1


1
a

<
> E


1
>
r
]


1 (

n
u


|
»


. r
j


| ^


<
i
*


r

O •" o
> c
] <





1
0 <
>C



' |


> .
]



1



D
Xi v

-X


•X 	
Chlordane Endrin Heptachlor Lindane Methoxychlor Toxaphene 2,4-D 2,4,5-TP
                             & its epoxide
                                             1-79

-------
Rgure 9 Comparison of correction factor and control sample recovery for volatile organic
compounds (VOQ.

    2.0 T
    1 5 l
    0.5-
 o  Correction Factor 95% Cl
 •  Correction Factor Average
 D  Control Sample 95% Cl
 •  Control Sample Average
 x  Critical Correction Factor


 T?     ?T     TT     1°

 i        ti     i       n
                                                      3.0
:

5C


'1
>
^


1

D

' C

»





D
O

1 O
IT «
*•
u
c



'I '
>

D
f


c


'1

1
k *
                                                         C
•1

>l

X
1 1
1 !

0>


•

<
1 .
\
]
1 <
1 1
rj

i
i

o n

l
4

D
>
O
i v i y

c
»
l


D <
4
1

?
,1



o
C
1 .
A 1
[
OQ O
**-* ' *-* • "
1 s 5 i ^
g> C jx >» CL
e .8 e °- §
x •§ |—
I ^ •»-
1
o
]

1
I

c
< >
^
•
ichlorophenol ;
£
•k
                                      
-------
9                                       ABSTRACT

                                Ensuring Data Authenticity
                              in Environmental Laboratories


           Jeffrey C. Worthineton. Director of Quality Assurance; R. Park Maney, ESQ.,
           Vice President; TechLaw,  Inc.,  12600 W. Colfax Avenue,  Suite  C-310,
           Lakewood, Colorado 80215,

     D*ta delivered to both government and public clients by environmental laboratories is no
     longer reviewed  only  for contract compliance  and acceptability  of  quality assurance
     techniques, data users are concerned that the data be "authentic".

     Investigations  into  data  authenticity  by government agencies have  sparked organized
     efforts by data users to ensure that the data delivered is exactly what it is purported to be.
     Laboratory  personnel should  seize  the  initiative  and  embrace  concerns  for  data
     authenticity as a normal part of conducting business.

     Laboratories and other generators of environmental data need to develop policies and
     procedures to  ensure the authenticity of their own data. These procedures should include
     authenticity monitoring as a function of both laboratory management and quality assurance
     staff.

     The authors present a summary of the practices that have resulted in the delivery of data
     that is not authentic and a plan for developing policies and procedures for data authenticity
     including:

           0  Education and training,

           0  Internal audits,

           0  Laboratory documentation procedures, and

           0  Automated laboratory practices.
                                             1-81

-------
•in              ESTABLISHMENT OF LABORATORY DATA DELIVERABLE REQUIREMENTS
IW                FOR  DATA  VALIDATION  OF  ENVIRONMENTAL RADIOLOGICAL DATA
      David A. Anderson,   Scientist
      Environmental  Restoration  Program
      EG&G Idaho,  Inc.
      P.O. Box 1625
      Idaho Falls, ID  83401
            ABSTRACT

                   In the world of environmental sampling and analysis,  the  importance
            of obtaining accurate, valid data has become increasingly important.  Much
            attention has been placed on the review and validation of sample data from
            CLP  organic and  inorganic  analyses  (U.S.  EPA Functional Guidelines for
            Review of Inorganic and  Organic Data).   In  the  field of nuclear waste
            management, we are seeing a significant increase in the analysis of mixed-
            waste   samples  from  waste  sites  containing   potential  radiological
            contamination.   In light  of this fact, a need has been identified  for the
            establishment  of a protocol for data quality assessment and validation of
            sample data from radiological  analyses.   The  EPA CLP program protocol
            provides a means for obtaining consistent data deliverables from different
            analytical  laboratories  performing organic  and  inorganic  analyses.  This
            allows for a reasonably equal assessment of sample data regardless of the
            laboratory doing the analyses.  The same type of protocol  is necessary to
            adequately  address  the  need for data  of  known  precision,  accuracy,
            representativeness,  comparability  and  completeness  when radiological
            analysis of environmental  samples  is performed.

                   Establishment  of  a  consistent  set  of  radiological deliverables
            provides a method of obtaining useable, accurate data even when  using more
            than one laboratory for sample analysis.  A data assessment and  validation
            program for  radiological  data  can  then  be  designed  from  the  data
            deliverable requirements.

                   At EG&G Idaho a set of forms have been designed to provide a master
            template  for  the laboratory to  format their  data deliverables.   Use of
            these  templates  insures  consistency  in the data received from different
            laboratories.  Sample results from these forms may be transmitted  to EG&G
            on  computer  diskette  or  electronically  transferred  directly   into  a
            database.  An  initial evaluation of sample results data and quality  control
            data can be performed by a computer analysis  of selected data fields in the
            forms  against an established set of criteria.  Initial flagging  of  quality
            control and sample results  data  can also be computer generated.   This
            system provides  for a substantial savings in time and money for both data
            entry  functions  and  the data  validation process.
                                            1-82

-------
4 4         AN ASSESSMENT OF QUALITY CONTROL REQUIREMENTS FOR THE ANALYSIS OF
               CHLORINATED PESTICIDES USING WIDE BORE CAPILLARY COLUMNS--
                                A MULTI-LABORATORY STUDY


     Jack A.  Serges.  Lockheed Engineering & Sciences Company,  Las Vegas,  Nevada.

     Gary L.  Robertson, U.S.  EPA Environmental Monitoring Systems Laboratory,  Las
           Vegas,  Nevada.
     Wide bore (> 0.5 mm) capillary columns are frequently used for the analysis of

     chlorinated pesticides by gas chromatography/electron capture detection

     (GC/ECD).  In March 1990, the U.S. EPA Contract Laboratory Program (CLP)

     included the use of wide-bore capillary columns in its protocols for pesticide

     analysis.  In August 1990, as part of the contracting process, single blind

     samples were distributed to 43 laboratories for pesticide analysis.


     The CLP protocols allow variability  in the analytical columns and conditions

     used for pesticide analysis, but the protocols require stringent quality

     control (QC) to enable the use of the results for a wide variety of purposes.

     The QC requirements include chromatographic resolution, compound breakdown,

     compound retention time stability, and detector linearity.


     The laboratories participating in the blind study demonstrated a wide range of

     proficiency in the use of GC/ECD.  An examination of the results of the blind

     study thus gives insight into the ruggedness of the chromatographic procedures

     under variable conditions.  This assessment presents results  of a study of the

     QC data and its impact on the analytical quality.  Recommended QC

     modifications that were derived from the study will be discussed.


     Notice:  Although the research described in this article has  been funded
     wholly or in part by the United States Environmental Protection Agency through
     contract number 68-CO-0049 to Lockheed Engineering & Sciences Company, it has
     not been subjected to Agency review  and therefore does not necessarily reflect
     the views of the Agency and no official endorsement should be inferred.

                                          1-83

-------
-j p      ANALYSIS-SPECIFIC  TECHNIQUES  FOR ESTIMATING  PRECISION  AND  ACCURACY  USING
         SURROGATE RECOVERY


         Charles B. Davis. Lockheed Engineering & Sciences  Company, Las Vegas, Nevada and
         University of Toledo, Toledo, Ohio

         Forest C. Garner, Jack A. Berges, Lockheed Engineering & Sciences Company,  Las
         Vegas, Nevada.

         Larry C. Butler,  U.S.  Environmental Protection Agency, Environmental Monitoring
         Systems Laboratory, Las Vegas, Nevada



         Statistical techniques are presented for estimating the precision and accuracy

         of gas chromatography / mass spectroscopy (GC/MS) analytical determinations by

         using associated surrogate recoveries.


         The statistical techniques employed involve variations on the so-called "using

         the regression line  in  reverse"  common in linear calibration.   An appropriate

         regression model is first identified using prior data.  In the simple case, where

         the means and variances  of analyte and surrogate recoveries are constant with

         respect to true concentration, prediction intervals for recovery are inverted to

         provide  confidence intervals for analyte concentration.  Otherwise  (if means

         and/or   variances  are   functions of  concentration),  variance  stabilizing

         transformations are first performed as needed, and then computations similar to

         those used in conditional multivariate calibration provide confidence intervals

         for analyte concentration.  In either case, the resulting confidence intervals

         provide the desired information regarding precision and accuracy.


         Examples of  the use  of the  techniques  are given,  using  data sets  based on

         quarterly blind performance evaluation studies conducted by the U.S. EPA Contract

         Laboratory Program and other  sources.   For  these data sets,  intra-analysis

         estimates of precision and accuracy obtained by the techniques presented here are
                                            1-84

-------
compared  and contrasted  with  inter-analysis estimates  available from,  for

example, matrix spike  studies.   The relationship of the improvement in intra-

analysis over inter-analysis estimates to  the variability and correlation in

analyte and surrogate recoveries is  explored with these examples.


Finally, we discuss the implications of these results regarding the usefulness

of surrogate and/or matrix spike recoveries in "correcting" analytical determina-

tions,  and suggest  statistical procedures that  might be  employed when such

"corrections" are warranted.
Notice:  Although the research described in this article has been funded wholly
or in part by the United States Environmental Protection Agency through contract
number 68-CO-0049  to  Lockheed Engineering & Sciences Company, it has not been
subjected to Agency review and therefore does not necessarily  reflect the views
of the Agency and no  official endorsement  should be  inferred.
                                    1-85

-------
•1 3                              USE OF ORGANIC DATA AUDITS IN
                 QUALITY ASSURANCE OVERSIGHT OF SUPERFUND CONTRACT LABORATORIES

                                        Edward J.  Kantor
                      U.S. EPA Environmental Monitoring Systems Laboratory
                                    Las Vegas, Nevada  89119

                                        Mahmoud S. Hamid
                            Lockheed Engineering  & Sciences Company
                                    Las Vegas, Nevada  89119
         ABSTRACT:

         Organic data audits are performed to assess the technical quality of analytical

         data and to evaluate overall laboratory performance. The technical data quality

         is assessed on the basis of the total number of problems observed in the case.

         The processes used to identify problems  in organic  analytical data range from a

         check  of the  quality control  to a  thorough  investigation of  the raw  data

         submitted with  the case will be  presented.   Besides providing  the  basis for

         determining the  technical  quality,  the  number and type of  problems  provide a

         mechanism to track data quality for the Contract Laboratory Program (CLP) , or for

         an individual laboratory,  over time.  Long-term tracking is accomplished by the

         use of an audit comment data base that contains standardized comments explaining

         common problems  found within  the data  submitted  by  CLP laboratories.   Each

         comment  represents  an individual problem,  and the frequency  of use  for the

         comments is tabulated by the data base.  Common problems observed during the past

         year  in CLP  data  packages  such  as  calibration  errors,  failure   to  submit

         deliverables, instrument contamination,  and use of incorrect  quality  control

         solutions will be presented.
         Notice:   Although the research described in this article has been supported by
         the United States Environmental Protection Agency through contract 68-CO-0049 to
         Lockheed Engineering &  Sciences  Company,  it has not been  subjected  to Agency
         review and therefore does not necessarily reflect the views  of the Agency and no
         official endorsement should be inferred.

                                            1-86

-------
4 A     USE OF INORGANIC DATA AUDITS IN QUALITY ASSURANCE OVERSIGHT OF SUFERFUND CONTRACT
1       LABORATORIES


       Robert B. Elkins. Lockheed Engineering & Sciences  Company,  Las Vegas, Nevada

       William  R.  Newberry,  U.S.   Environmental  Protection  Agency,  Environmental
       Monitoring Systems Laboratory,  Las Vegas,  Nevada


       ABSTRACT:

       Inorganic data audits are performed to assess the technical quality of analytical

       data and to evaluate overall  laboratory performance.  The technical data quality

       is assessed on  the basis of  the total number of problems observed  in the case.

       The processes used to identify  problems  in inorganic analytical data range from

       a  check of the quality control  to a thorough investigation  of the raw data

       submitted  with the  case.    Besides  providing the basis  for  determining the

       technical quality, the number and type of problems provide a mechanism to track

       data  quality  for the Contract  Laboratory Program (CLP), or for an individual

       laboratory, over time.  Long-term tracking is accomplished by the use of an audit

       comment data base that contains standardized comments explaining common problems

       found within the data submitted by CLP laboratories.  Each comment represents  an

       individual problem, and  the frequency of use  for the comments is tabulated by the

       data  base.  Common problems  observed during the past year in CLP data packages

       include   calibration  errors,  failure   to   submit  deliverables,   instrument

       contamination,  and use of incorrect quality control solutions.


       Notice:  Although the research described in this article has been supported  by
       the United States Environmental Protection Agency through contract 68-CO-0049  to
       Lockheed Engineering &  Sciences  Company,  it has not been  subjected to  Agency
       review and therefore  does not necessarily reflect the views of the Agency and  no
       official endorsement should be inferred.
                                           1-87

-------
          Improved  Evaluation  of Environmental
            Radiochemical Inorganic Solid Matrix
                             Replicate Precision:

                      Normalized Range  Analysis Revisited
                                        Robert E. Gladd
                                     James W. Dfflard, Ph.D.

                                         I.T. Corporation
                                    OAK RIDGE LABORATORY
             ABSTRACT; the Ifocmafocct Range Scai8t>c,a» BPA-6QQ/4^l» provides tfafc
                                                                                ___ r_
                                         (osuaily <2B&) a&ptettfittqaa&ryttSGQm^
             owing to dietct^wcmaptudwof mceafcubted and*cs$)ccted*steinas, tbc^opcctecf* value being basedoa
                         teof ttWatatfcoc


  The monitoring of laboratory accuracy and precision at the
IT Oak Ridge Laboratory is guided by the statistical procedures
detailed in EPA-600/4-81-004, methods which provide gener-
ally oractkal empirical point estimators of analytical perform-
ance. Accuracy evaluation is accomplished through the use of
the "Normalized Deviation" statistic, in which the  analytical
result of a spiked sample test is "normalized" to the "known"
value and "expected laboratory 1-sigma" precision; in traditional
statistical parlance, a "Z-transformation." Normalized Devia-
tion statistics (NDEVs) are computed and plotted on control
charts with a mean of zero, warning limits at +/-2.0s and control
limits of +/- 3.0s.

  Similarly, the analytical precision of unknown replicates (Le.,
samples where a "known" value from a reference standard is not
present) is assessed via the Normalized Range (NRANGE)
estimator, in which the numerical difference between replicate
results is evaluated in the context of both an "expected 1 -sigma"
precision level and an "expected range" factor.  NRANGE
statistics are computed and plotted on control charts containing
an X-Y origin of zero, an "expected range" of 1.0s, and warning
and controllimits at +3.0s and +4.0s respectively (i.e., mean, or
"expected" range plus 2 and 3 sigma). NRANGE points lying
above the +4.0$ Control Limit mandate an investigation of the
analytical data for the replicates in question to ascertain the
causes of the excessive divergence.

   These statistical tools assume the presence of a liter or kilo-
gram sample aliquot, the latter necessitated by the EPA food
matrix crosscheck. Analytical results are adjusted to their respec-
tive activities at a liter or kilogram before the statistics are com-
puted. Where the samples are constituted of low-level, low-
volume inorganic solid matrices, the generation of spurious
"outlier" statistics is a recurring phenomenon, owing principally
to the relative magnitudes of die analytically determined 1 -sigma
and the "expected 1-sigma" used by the NRANGE computa-
tion. This a-priori 1-sigma, while empirically appropriate for 1
kg. samples, imposes an unrealistic constraint on small aliquot in-
organic replicates. Clearly, in such cases a method of incorporat-
ing the analytically determined 1-sigmas must be employed to
make the NRANGE statistic reflect the true precision level of the
replicates; to the extent that these "outliers" are invalid they con-
                                               1-88

-------
INORGANIC SOLID MATRIX REPLICATE PRECISION: Normalized Range Analysis Revisited.      Pg.2
                                       'alt2
tribute to a misleading impression of laboratory precision capabili-
ties and result in unwarranted technical review of the replicate
sample data, an examination required of all "out-of-control" QC
results.

  Since every quantitative sample result is essentially a point-
csdmate of a "mean" value which approximates a "true" activity or
concentration level, it would be tempting to dismiss any "expected
sgma" constraints on replicate  results, particularly of the types
under discussion here, and apply a sort oft-test" on our experi-
mental "means" incorporating  only the  analytical 1-sigmas in
determining the acccpta-
bikyofthc range between
replicate values. Support
fcc such a  method  de-
rives from  die feet that
the relative standard de-
viation (Lc., the "percent
agma," or coefficient of
variation) accompanying
each production analysis
soot routinely evaluated
against an  "expected"
sigma; it is generally ac-
cepted mat, as activities
and/or aliquots are lesser,
sprns will tend to be pro-
portionally greater, fre-
quently approaching or
even exceeding the mag-
nitudes of the quantified
activities themselves. An
 expected sigma," while
serviceable in the main as
an objective standard of
analytical variability, is fre-
quently inappropriate in
fight of the component
measurement particulars of individual cases such as those under
review in this presentation.

  The conventional t-tcst cannot be directly applied to replicate
n&analytical results owing to the feet that "N=l" for each
sample datasct, leaving us without "degrees of freedom" to employ
iathe derivation of t-valucs. Given this problem, should we wish
to retain the mathematical simplicity of the NRANGE statistic, we
could simply replace the "expected" sigma in the formula with the
mean of the analytical percent sigmas.  Such a replacement would
jiddNRANGE statistics derived totally in the context of the error
terms of the lab  results themselves,  removing any empirically
 objective" variability standard in favor of the case-specific uncer-
tainty estimates. The virtue of such  an approach would be to
remove any potential argument over whether an "NRANGE >4"
alcuktcd in such a fashion in fact represented an "out-of-control
replicate set. Replicate results so divergent as to normalize out to
                                                          "NRANGE > 4" even after taking their own individual error
                                                          terms into account would indisputably be indicative of unaccept-
                                                          able precision and would indeed merit technical review to deter-
                                                          mine the causes of the disparity.

                                                             Alternatively, eschewing the NRANGE formula entirely, we
                                                          might statistically examine our replicates via one of two variations
                                                          on  a "Z-test" formulation, employing either the mean of the
                                                          analytical sigmas or the square root of the sum of the variances as
                                                          divisors of the replicate range, as shown in the box below.
                                                                                               Under either "z-score"
                                                                                             approach (Eq. 1 & 2 in
                                                                                             box) our "outliers" would
                                                                                             be those resulting in a z-
                                                                                             statistic > 3.0 absolute.
                                                                                             This approach makes in-
                                                                                             tuitive sense, but is a bit
                                                                                             bothersome in that any
                                                                                             utilization of a reference
                                                                                             value such  as the "ex-
                                                                         [2]
NRadi  -  NR
    acrj
                                                  eDa
                                                  epa
                                                                             [3]
                                               «s again pre-
                                   cluded, a  tactic  that
                                   contravenes an implicit
                                   assessment principle of the
                                   EPA-600 method:   the
                                   application of empirical
                                   guideposts to laboratory
                                   precision capability ac-
                                   counting.  A simple ad-
                                   justment    to    the
                                   NRANGE  statistic is
                                   therefore proposed, one
                                   that incorporates both the
                                   EPA sigma factor a/nd the
                                   mean  of the analytical
                                   sigmas    into    the
                                   NRANGE calculation, as
shown by Eq. 3 above:  a ratio of "expected" over "found."

   It should be readily apparent that where the mean lab sigma
nearly equals the "expected" sigma the NRANGE statistic will be
quite close to the value returned by the standard method. Where
the lab error coefficient is greater than the expected, the NRANGE
will be attenuated by the ratio of the two. Further, where the mean
lab sigma coefficient is smaller than the expected, the adjustment
factor wil be > 1.0, thereby expanding the NRANGE value. In
this manner the sigma-ratio adjustment factor is a double-edged
sword; if individual error terms are better than the expected, the
results had better be minimally divergent to avoid being pushed
into  "out-of-control" status.  Such a condition makes methodo-
logical sense; analytical sigmas are mathematical expressions of
our confidence in our quantitative estimates. Replicates returning
bctter-than-expected error terms and grossly disparate  "means"
are indicative of a condition warranting quality control review.
                                                       1-89

-------
  INORGANIC SOLID MATRIX REPLICATE PRECISION: Normalized Range Analysis Revisited.       Pg. 3
                           •
                           XI
                           D
                           L
                           a
   A graphical example effectively illustrates the problem posed
by the application of an expected sigma to a low-level small aliquot
solid matrix duplicate result set. Two Sr-90 results, at 3.30 and
1.68 dpm respectively, arc graphed as normal distributions (sec
box, below right), first with an assumption of a 5% CV, then
overlaid using sigmas derived in the analyses (0.59 and 0.52 1 -sig-
mas, respectively).  While
the results arc displayed as
narrow, peaked distributions
whose tails arc quite far apart
under  a 5% CV assump-
tion, when viewed in the
distributional context of the
analytically derived sigmas,
quite  another  picture
emerges;  the tails of the
distributions overlap sub-
stantially. The traditional
NRANGE statistic for this
set came in at NR= 14.73,
while   the   "corrected"
NR=3.01, and this adjusted
Normalized  Range  value
seems  appropriate;   our
replicates diverge, perhaps
more than we would prefer,
but certainly not to the ex-
tent indicated by a Normal-
ized Range of 14.73. Were
these replicate results those
of a full kilogram  vegeta-
tion matrix emanating hun-
dreds of dpm,  we would
perhaps have cause for concern at the disparate replicate values re-
turned by the lab. In the instance of 2.07 gram inorganic aliquots
evincing a few dpm, however, the range between Rl and R2 is not
all that severe, certainly not to the point implied by an NRANGE
statistic of 14.73. The relative standard deviations (the CVs) for
Rl and R2 were, respectively, 17.86% and 30.91%.

  EPA-600 expected 1-sigma % precision guidelines arc grouped
by type of analysis into four levels:  5%, 10%, 15%, and 25%. An
examination of 116 inorganic solid matrix replicate results from
our RADQC™ laboratory QC database is revealing. As might be
expected, most of the NRANGE difficulty lies with the analyses
classified in  the "5% precision" group, as the following table
illustrates:
                                                           practice, grouped by the EPA sigma classifications.  It is evident
                                                           that a 5% sigma represents an unrealistic level of precision where
                                                           small aliquot solids arc concerned. The application of the NRANGE
                                                           sigma ratio adjustment factor factor to the 116 samples investi-
                                                           gated for this research effort reduced the "NR>4.0" outliers by
                                                           75%, and the attenuated "adjusted NR" statistics  were ovcr-
                                                      RADQC SOLID MATRIX DUPLICATES
                                                             Sr-90  9 2.87 gr«m»
                                                               3.3  8.16
                                                               1.68 8.88
                                                                                                     3.3
                                                                                                     1.68  8.B2
                                                                                                             -
                                                 DPM, w/ 5%  Sigma* L Analytical  Sigm«»
       N
       68
       4
       38
       6
                       EPA 1-sigma
                         0.05
                         0.10
                         0.15
                         0.25
LAB 1-sigma
  0.125
  0.121
  0.159
  0.232
The column on the right tabulates the average CVs found in actual
                                                            whelming of a magnitude consistent with the type of graphical
                                                            evidence obtained by plotting the gaussian distributions in the
                                                            manner of the above example.

                                                              Further statistical support for the use of the NR adjustment is
                                                            seen by a correlation matrix comprised of the values obtained for
                                                            this QC data under the formulas  displayed in Eqs. 1, 2, and 3:
                                                           NR_adj
                                                            Z altl
                                                            Z alt2
                                                                           NR_adj
                                       1.000
                                       0.903
                                                                             0.925
                                                    Z altl
                              0.903
                              1.000
                                                    0.956
                                             Z alt2
0.925
0.956
                                             1.000
   The high Pcarson-R correlations among the three methods
indicate a significant agreement between the two  Z-scorc vari-
ations and the adjusted NRANGE method; we arc measuring the
same phenomenon, irrespective of algebraic method. Any of these
formulations serve to reduce the quantity of spurious QC outliers,
improving the assessment of inorganic solid matrix  precision.
                                                        1-90

-------
  INORGANIC SOLID MATRIX REPLICATE PRECISION:  Normalized Range Analysis Revisited.      Pg. 4
  A second graphic plot example is provided below to further
demonstrate the utility of the NRANGE adjustment method.
This replicate set consisted of Gross Beta analyses performed on
aliquot weights of 264 and 258 mg.  The dpm results were
calculated to be 112 +/- 23 and 134 +/- 28 (2-sigmas). The
unadjusted NR=4.38, just slightly over into the outlier realm.
The "corrected" NR=2.11, and again, the distribution plots seem
to provide visual  agreement with the numerical statistic. The
mean CV was 10.36%.
ROBERT E. GLADD is Vice President of CAMS Associates of
Knoxville, TN. He has served since 1986 as a statistical analyst and
computer applications consultant to the IT Oak Ridge Laboratory.

JAMES W. DILLARD, Ph.D. is the Technical Director of the
Oak Ridge Radioanalytical Laboratory of IT Corporation.
                                         RADQC SOLID  MATRIX DUPLICATES
                                            Qro«»  B«t« 9 £64,  268  mg.
                                112
                                134
6.6
6.7
      u
      JO
       a
       L
       a
                                                                                         -J.12--
                                                                                         134
                                      ..--11.5
                                        14
                  -
                                     DPM,  w/ 6X  •igm>« & Analytical Sigma*
   A quick look at some of the aggregate univariatc and correla-
 tion statistics for the datasct of solid matrix replicates used in this
 effort is useful.  The median uncorrcctcd NRANGE was 2.14,
 with a mean of 3.74, while the median adjusted NRANGE was
 1.48, with a mean of 2.02. The smallest aliquot was 4.8 mg., and
 the largest was 841.9 grams, with a median of 2.07 grams and
 mean of approximately 69 grams. The lab sigma precision was, as
 we might expect, inversely correlated with aliquot (R=-.25, sta-
 tistically "significant" with p=.007)  The median lab sigma
 precision was 12.17% with a mean of 14.04%.  One lesson
 flowing from these data is that perhaps those types of QC analyses
 currently classified as requiring "5% expected 1-sigma precision"
 be instead calculated using a 15% expected  precision statistic
 where the samples arc those of inorganic solid  matrices. The
 NRANGE sigma ratio adjustment factor employed here is essen-
 tially performing roughly that very sort of task in attenuating the
 Normalized Range where the sigmas arc closer to  15% in the
 laboratory production environment.
                     REFERENCES

 Kanipc, Larry G. (1977), Handbook for Analytical Quality Control
 in BjuUoanalytical Laboratories, TVA/USEPA Intcragcncy En-
 ergy-Environment Research & Development Program (EPA 600/
 7-77-088)

 Walpolc, Ronald E. & Myers, Raymond H.  (1989), Probability
 and Statistics for Engineers and Scientists, Fourth Edition, New York,
 MacMillan

 EPA 600/4-81-004, USEPA, EMSL Office of Research & Devel-
 opment, Las Vegas, NV.
          Copyright 1990 IT CORPORATION
                   All Rights Reserved
                   1550 Bear Creek Road
                   Oak Ridge, TN 37830
                                                      1-91

-------
•J g                              Laboratory On-site Evaluations
                               as  a Tool  for Assuring Data Quality
         Timothy J.  Meszaros.  Lockheed Engineering & Sciences Company,
               Las Vegas, Nevada

         Gary L. Robertson, U.S. Environmental Protection Agency,
               Environmental Monitoring Systems Laboratory-Las Vegas, Nevada


         In its role of providing quality assurance  support to the Superfund Office, the
         Environmental Protection Agency's Environmental Monitoring Systems Laboratory-Las
         Vegas has developed a program for conducting laboratory on-site evaluations ("on-
         sites").   Although developed  principally for  use  with  Superfund's  national
         Contract Laboratory  Program (CLP),  "on-sites"  have been incorporated as  an
         integral element in an overall scheme of laboratory performance evaluation for
         activities  supporting  the  Resource  Conservation and  Recovery  Act  and  the
         Department of Energy.


         The  purposes   of  the   laboratory   on-site   evaluation,   together with   the
         complementary performance evaluation sample program, are (1) to determine whether
         a given laboratory can effectively perform required analyses and (2) to identify
         laboratory problems that negatively impact performance.   As a part of the  "on-
         site," the following  elements are reviewed:  the laboratory's quality assurance
         plan;  its  standard operating procedures;  its utilization  of available space;
         adequacy   of   personnel;  instrumentation   capacity;   sources   of   possible
         contamination; and the ability to perform the required analyses.


         This presentation will describe those good  laboratory practices examined in the
         "on-site," and highlight the types of problems encountered and their potential
         impact upon data quality.


         Notice:  Although the  research  described in this article has been funded wholly
         or in part by the United States  Environmental Protection Agency through contract
         number 68-CO-0049 to Lockheed Engineering and Sciences Company, it has not been
         subjected to Agency review and  therefore does not necessarily reflect the views
         of the Agency and no  official endorsement should be  inferred.
                                           1-92

-------
17
APPLICATION OF BIAS CORRECTION
        D.Syhre.   Laboratory Supervisor,  Browning-Ferris Industries  Houston  Lab,
        5630 Guhn Road, Houston, Texas, 77040.

        ABSTRACT

        EPA has  required bias  correction of analytical data  generated  under the
        Toxicity  Characteristic Leaching Procedure  (TCLP,  SW-846,  Method  1311)
        since September of  1990. In order to determine the validity of correcting
        data  based  on  recoveries  of   matrix  spikes,  two  Standard  Reference
        Materials (SRMs) were created from real environmetal samples.  These SRM's
        were  spiked with  stable  isotopically labeled  compounds.    Results  were
        corrected for bias  by three different methods: 1)  on an individual sample
        basis per individual  compound;  2) on an individual  compound per batch of
        ten basis;  and 3)  on  a class  of compounds per individual sample basis.
        Results of uncorrected  and corrected recoveries were compared to known or
        "true"  concentrations  and  conclusions   drawn  about  the  effect  bias
        correction has on data  quality.

        INTRODUCTION

        The 1984 Hazardous  and  Solid Waste Amendments  to the Resource Conservation
        and Recovery Act   (RCRA)   directed  the  EPA  to re-examine  the Toxicity
        Characteristic (TC) portion of  the Extraction Procedure Toxicity test [1]
        and make changes necessary to better  address the leaching of  wastes and to
        regulate  additional  characteristics.  A   revision  to  the  TC  rule  was
        published  in the  Federal   Register  (55FR  11798)  which added  25  organic
        constituents, and a later  revision  [2] added quality control  measures that
        require that the  bias determined from matrix  spike  recoveries  be  used to
        correct  the  analytical data.    In  January  of  1991   EPA   extended  bias
        correction  requirements to the  Land Disposal Restriction  Rules,  and has
        proposed  including bias correction in Chapter  1  of  SW-846.   The latter
        inclusion would extend  bias correction requirements to all  RCRA testing.

        Despite the proliferation  of bias correction requirements,  little has been
        published regarding how well this  technique  works.    If one  is  to really
        know how well bias  correction works,  one  must know the true  concentration
        of  an analyte.   Merely   spiking deionized  water  does  not provide  the
        information needed  about recoveries since  one would expect uniformly good
        results from an "in control" analysis.  For this reason two SRMs made from
        real  environmental  samples were  created.   A water matrix  was  chosen for
        both  types  of matrices in order to overcome  potential  mixing problems.
        The SRM for Cases A and B  was created from 11 liters of sanitary landfill
        leachate.  SRM Case C was  created from a single  monitor well  from a closed
        remediation  site.   Method  1311  currently   requires  that  spiking  for
        correction purposes be  done on a  batch basis  (one spiked sample per twenty
        samples of a similar matrix).
                                           1-93

-------
Proposed changes  to Chapter 1 of  SW-846  would also require  spiking  on a
batch basis,  but may allow  correction by representative  compound class.
No bias correction would be required for "self-correcting" methods such as
isotope dilution.   Cases  A,  B, and C  are experiments on  the  use  of  bias
correction  in each  of  the  above ways.   In all  cases,  uncorrected  and
corrected data were compared with the true  concentrations and changes in
accuracy and uncertainty calculated.

METHODOLOGY

Case A

Eleven liters of sanitary landfill leachate from the same point source were
collected, composited, and  three  liters removed for analysis as a sample,
duplicate, and spike.  The remainder was spiked with the entire list of TC
volatile  and  semivolatile   parameters.    This  amount  is   subsequently
referred to  as  the "true"  concentration  and was the SRM  for both Case A
and Case B.  See Figure 1 for  the processing of this SRM.

The  seven liters  of  the SRM  were then  filtered in the  TCLP prescribed
manner for  volatiles and semivolatiles.  It has  been our experience that
leachate often contains a large  amount of dissolved gases which may cause
foaming during  the  volatile purge step.   Because of  this,  we routinely
dilute these samples prior  to analysis.   We had  anticipated that a five
fold  dilution of  the  SRM would  be adequate,  but the  sample foamed too
badly  and a  ten  fold  dilution  was  needed. Prior  to  each  analysis for
volatile  constituents,  the  extract (filtrate) was  diluted  ten  fold and
spiked with  isotopically labeled volatile TC  constituents.  These spikes
were used to produce spike recovery correction factors.

Prior to each of  the  7  extractions,  the semivolatile filtrate was diluted
(one to five) and spiked with the entire list of isotopically labeled TC
semivolatile  constituents.   A separatory funnel  extraction was performed
due to time  constraints.   These  spikes were also  used  for spike  recovery
correction factors.

Spike  amounts  for  volatile  and  semivolatile  parameters  varied  on  a
compound  basis.    These  levels had  been predetermined to  give   adequate
recoveries from a  leachate matrix without exceeding regulatory levels when
dilution  factors  are  considered.   Volatile  analyses were  performed via
methods 8240/8260  from  SW-846 and semivolatile analyses were performed in
accordance  with  method  8270.   Spiking  levels   corrected  for   dilution
factors are reported in Table  5.

Case B

The uncorrected  mean recoveries  of the native  compounds from Case A were
experimentally  inserted  into a  batch of  nine  other  sanitary   landfill
leachates.   However,  each of  the  nine other samples in the  batch of ten
                                    1-94

-------
had also been  spiked with the entire volatile and  semi-volatile TC list,
and recoveries calculated.  True concentrations of these nine samples were
not known.   Spike recoveries from each  of the nine samples  were used to
individually correct results of the native compounds found in the SRM.

Case C

A  simulated  monitor  well was constructed using a  five foot  section of
Corning glass pipe fitted with Teflon [3]  endcaps.  Each endcap was fitted
with a Teflon stopcock.   Total well volume was about 12.6 liters.  The top
endcap of  the  monitor  well  was  removed  and the well  filled with sample
from a single  contaminated  groundwater  monitoring well.  A volatile and a
semivolatile sample  were  removed  from the bottom of the well to determine
pre-spike  constituents  arid  concentrations.   The well  was  re-filled with
groundwater  which had been spiked with  the  following classes of volatile
parameters:     Saturated    Chlorocarbons;    Unsaturated   Chlorocarbons;
Aromatics; Bromocarbons.  The  constituents within each class are found in
Table 1.   The  endcap was replaced and the  small  remaining volume filled
via access through the  stopcock.   Top and bottom stopcocks were  connected
by Teflon tubing  to  a pump calibrated to deliver one well volume  every 5.6
hours.   The  stopcocks  were  opened  and the pump turned on  for 23 hours
during which time about  four  well volumes  were  recirculated,  enough for
adequate mixing  of the sample.  At this  point the well was considered to
contain a SRM.

The pump  was  stopped,  stopcocks closed and the top cap removed.  A three
foot top  section was added  to prevent overflow and  the well was sampled
seven times  by the  lab's field  crew.   The  volatile  samples were packed
with ice and shipped by Federal Express  to ourselves  (with  routing through
Memphis).  Just  prior  to purging,  each of  the  seven samples were spiked
with a specially  prepared surrogate mix  that contained stable  isotopically
labeled  compounds representative of  each  class  of compounds.   Recoveries
of each representative  compound were  then  used to  bias correct  all amounts
of  native compounds within that  class  found  in the  SRM.   Results  were
compared to  the true concentration in the  same way as  in Cases  A and B.

The simulated  monitor well  prepared  for  a  semivolatile experiment in the
same  manner.   The  semivolatile  classes  were:    Nitroaromatics;  Acidic
Compounds    (phenols);    Nitosoamines;    Bases;    Polynuclear   Aromatics;
Chlorinated  Hydrocarbons; and Phthalates.   Just prior  to extraction,  each
sample  was  spiked  with  lableled representative  compounds  (see Table 4).
Results were treated the same  way as  the volatiles.

Results and  Discussion

Case A

Seven  replicates  of leachate SRM   were  subjected  to normal  statistical
treatment.   The  sample  data  consisted of three sets:
                                    1-95

-------
   1) Recovery of known value.
   2) Recovery of spiked isotopes.
   3) Bias correction of SRM values with recovery of labeled compounds

The mean and  standard deviation for each set of  data  were calculated.  A
confidence interval at the 99°/, level of confidence was calculated as:

       X = mean   S = small population standard deviation
       CI = confidence interval = (S/square root of 7)*3.707
       LCL = lower control limit = X-CI
       UCL = upper control limit = X+CI

The  confidence  interval  of  a  data set  was considered  a  reflection of
uncertainty  in  the  data.   Changes  in uncertainty  that  bias  correction
produced were calculated as follows:

       '/. uncertainty = (corrected Cl-uncorrected CD* 100
                             uncorrected CI

Change  in  accuracy  is best  looked at  as a  relative indication  of the
closeness  of a  corrected  versus  uncorrected result to  a true  value.
Accuracy increases as  recovery  approaches  1007,,  but begins to decrease as
recoveries  exceed  100'/,  of  the  true  value.   Changes  in  accuracy  were
calculated as follows:

        UCX = uncorrected mean '/, recovery of true value
        CRX = corrected mean 1, recovery of true value
        ABS = absolute value
        Accuracy change = ABS(100-UCX)-ABS(CRX-100)

Table 2  summarizes the  uncertainty and accuracy of  bias  corrected data
from Case A.

Case B

While Case A was an experiment  of bias correction on an individual sample
basis. Case B was  an experiment on a  batch  basis.   Nine  leachate samples
were  spiked and recoveries  calculated.   These  recoveries,  tabulated in
Table 3,  were used to bias correct the  mean native  values found  in the
leachate SRM.  Since  the proposed  EPA guidelines for bias correction in
SW-846  Chapter   1   state  that  bias  correction  should  not be  done  if
recoveries are greater than 80%,  some  recoveries  were  not used.  Accuracy
and uncertainty results for corrected data are summarized in Table 5.

Case C

Case C was  an experiment in bias correcting an entire class of compounds
based on recovery  of  a single  representative compound from that compound
                                   1-96

-------
class on  an individual  sample basis.   Volatiles  were divided  into five
classes and  semivolatiles  into  six  classes.   Results are  summarized  in
Tables 6 and 7.

SUMMARY

Case A

Application of  bias correction  on an individual  compound per individual
sample basis  resulted  in an  overall increase of  accuracy by  about 9%.
Uncertainty also increased, however,  by  about  197,.

Case B

Application of  bias correction on an individual  compound per batch basis
resulted  in an  average  reduction in accuracy  of  1197,.   Calculation  of
change in uncertainty was  not as straight  forward since  not every data
point was used  per EPA  proposed instruction.   If batch  corrected data
for  a compound  contained  at least  3  data  points,  then the  change  in
uncertainty was calculated  as  follows:

      CIB = batch corrected confidence interval
      CII = individual sample  corrected  confidence  interval
      */, uncertainty =  (CIB  -  CII)*100
                          CII

The  average percent  increase in  uncertainty for  Case B  was 352%!  See
Figure 2.

Case C

Application  of  bias  correction  on a  class  of compounds within  an
individual  sample  was averaged on  a  volatile  and  semivolatile  basis.
Compounds which were corrected by  their  own  isotopes were not  included for
averaging  purposes.   Volatile data  increased in  percent uncertainty  by
557., but also increased in  accuracy  by only  8.  Semivolatiles  increased in
accuracy by 4 units and increased  in uncertainty by 137,.

The  above results  clearly  show  that bias   correction on  a  batch  basis
performs  very  poorly.   Case A was essentially an  isotope  dilution method.
Case C gave results that  were surprisingly similar to  Case A on an average
basis, but neither instance  showed  marked  improvment in accuracy.  Bias
correction  also failed to  correct a  glaring  case of  systematic  error in
the  TC  method.   The  average  recovery of native  hexachlorobenzene was  a
dismal 3.497..   There  was no  error  in  sample  spiking,  since  the  control
sample yielded  an excellent 907,.  These results are consistent with other
studies conducted at  this  laboratory:   hexachlorobenzene  fails to survive
the  TC  filtration step.  The mean corrected  value was only  3.877,  of the
original  concentration in the SRM.
                                    1-97

-------
Many people in  the  industry  have  lost  sight  of  the  fact  that  every
laboratory  in  the  country  using GC/MS  methods from  SW-846 are  already
employing bias  correction in a  case C  type  manner.   Every volatile  and
semi-volatile analysis goes through an extraction step and a concentration
step.  The  volatile extraction step is  usually  referred to as  the  purge
step.  The volatile extract is concentrated onto a  trap and then injected
into the  instrument.   The  semi-volatiles are  extracted  into  a  liquid
solvent, and  the extract  concentrated  to 1  mL.   Internal  standards  are
used for  quantification.   The use  of internal  standards means  that  all
data is bias corrected.

The only difference between volatile and semi-volatile analysis is in when
(or  where)  the  internal standards  are  added.  In the  volatile  analysis,
internal  standards are  added  prior  to  extraction.   The  semi-volatile
internal  standards  are  added  after  extraction.   This  is  a  critical
difference which may  partly  explain why bias  corrected data for volatile
analytes seems more consistent than semi-volatile data.

Perhaps two small changes would be in order:

     1)  Add semi-volatile internal standards prior to extraction

     2)  Change the list of internal standards for both volatiles and
         semi-volatiles.

These  changes  would   make  data  that have  already  been bias  corrected
through  use of  internal  standards more  representative  of the  sample.
Currently,  internal standards are  assigned  to an  analyte  purely on  the
basis of retention  time.   Compounds of  very  different chemical type share
the  same retention  time.   Careful examination indicates that data quality
improves  when  analyte recovery is  corrected by  compounds  which  more
closely  resemble  the analyte  in  terms  of  both  chemical and  physical
characteristics.  Recommended internal standards are:

     Volatile                              Semi-volatile
     vinyl chloride-d3                     n-nitrosodimethylamine-d6
     1»l-dichloroethene-d2                 phenol-d6
     chloroform-13C                        nitrobenzene-d5
     benzene-D6                            hexachloroethane-13C
     chlorobenzene-D5                      hexachorobenzene-13C6
     bromoform-13C                         aniline-d5
                                           di-n-butyl phthalate-d4
                                           pentachlorophenol-13C6
                                           benzo[a]pyrene-d!2
                                   1-98

-------
ACKNOWLEDGEMENTS

I would like  to express my appreciation to  the  following individuals and
their organizations for their help and participation in this study:

     M. Rudel. M. Hackfeld. M. Moore, H. Valdez, A. Valdez, R. Palomino,
     M. Noto, T. McKee, S. Mondrik, E. demons - BFI Laboratory, Houston,
     TX

     F. Thomas - Chemical Waste Management, Riverdale. IL

     M. O'Quinn - Encotec, Ann Arbor, MI

     W. Ziegler - ThermalKEM. Rock Hill. SC

     Environmental Laboratory Council, Washington. DC
REFERENCES

[1]  Method 1310 from Test Methods for Evaluating Solid Waste.
     Physical/Chemical Methods. November, 1986, Third Edition, USEPA,
     SW-846 and additions thereto.

[2]  Federal Register. Vol. 55. No.  126, Friday, June 29, 1990. pp.
     26986-98.

[3]  "Teflon" is a registered trademark of the E. I. DuPont Corporation.
DLS/slm
                                   1-99

-------
                            Figure 1
                       Landfill Leachate
                       Sample 11 liters
 Control Sample 3 liters

Analyze in duplicate

 Non-detcet levels for

   fYTfl nnn«*ittifhnt^
Duplicate Spike


good Recoveries
                                     Experimental Sample
                                      7 liters spiked with

                                      OTC Constituents
                                           SRM of Known Concentrations
7 Replicates
volatile analysis
  Spike Recovery Corrected

  Concentration in SRM based
   on value of each of 9 other

   spike replicates in batch
                                                    7 Replicates
                                                     Semivolatile Analysis
                                   Spike Recovery Corrected
                                   Concentration based on isotope
                                   spikes to SRM Sample
                                  Figure 2
         Comparison of Individual vs Batch  Based Corrections
+OUU
£ 300
O)
0
(J
Q)
DC
= 0
a
o
i_
o
c.
-300
-500
R
f^ h
rr ira
1 1
£>• •*
lot
S fl i
"v
i.
t<
E





1
1 a 1
i a
,
i 1
8 a
r l"l
9 a H _
,'li » U g
*a
M *

* - Mean
ra
V3 « Batched based rorreciions
U • Individual corrections-
Nitro-
                        Hexachloro-    O-Cresol   2.4-Dinitro-  Pyridine
              Benzene   butadiene                Toluene
                                  1-100

-------
Table 1 CASE C ORGANIC VOLATILES by COMPOUND CLASS
SATURATED CHLOROCARBONS
COMPOUND 200PPB IN DIH2Q
CHLOROMETHANE
CORRECTED RECOVERY
CHLOROETHANE
CORRECTED RECOVERY
METHYLENE CHLORIDE
CORRECTED RECOVERY
1,1-DICHLOROETHANE
CORRECTED RECOVERY
CHLOROFORM
CORRECTED RECOVERY
CARBON TETRACHLORIDE
CORRECTED RECOVERY
1,2-DICBLOROETHANE
CORRECTED RECOVERY
1 ,2-DICHLOROPROPANE
CORRECTED RECOVERY
1 ,1 ,2-TRICHLOROETHAKE
CORRECTED RECOVERY
1 ,1 2,2-TETRACHLOROETHANE
CORRECTED RECOVERY
1,1 ,1 -TRtCHLOROETHANE
CORRECTED RECOVERY
CHLOROFORM-13C
UNSATURATED CHLOROCARBON!
VINYL CHLORIDE
CORRECTED RECOVERY
CIS-1.2-DICHLOROETHENE
CORRECTED RECOVERY
TRANS-1 ,2-DICHLOROETHENE
CORRECTED RECOVERY
TRICHLOROETHENE
CORRECTED RECOVERY
CIS-1.3-DICHLOROPROPENE
CORRECTED RECOVERY
1 ,1 -DICHLOROETHENE
CORRECTED RECOVERY
TRANS-1, 3- DICHLOROPROPENE
CORRECTED RECOVERY
TETRACHLOROETHENE
CORRECTED RECOVERY
CHLOROBENZENE
CORRECTED RECOVERY
1 ,1 -DICHLOROETHENE- D2

Mean
160
183.7178
168.8571
192.9498
178.2857
204.5227
168.4286
192.8422
165.5714
189.0124
127.2857
145.4379
209
238.3924
170.8571
195.594
153.8571
1 74.961 1
1 46.4286
1 68.4957
161.142S
185.3636
175.2857
i
149.2857
172.6715
161.5714
188.4741
140.7143
165.5118
141.5714
164.9935
202.8571
236.2842
117
136.8937
, 57.54286
67.00263
97.07143
111.9346
214
250.6949
172.8571

STD
25.37059
31.90475
14.91564
1 1 .35443
26.35472
32.11179
15.24092
17.60963
14.0577
4.203377
11.57172
8.120035
30.97311
24.88816
9.546877
10.35389
22.401 1 1
12.39616
4.755949
19.57005
16.55726
27.94825
15.40254

20.5322
8.606541
1 1 .08839
18.15619
13.19993
28.75034
8.100558
1 2.47256
23.75871
29.14277
8.906926
17.96964
9.602405
1 1 .32563
27.2758
27.87805
18.81489
36.42765
21 .70583

Ct
35.54375
44.69799
20.89655
15.90736
36.92251
44.98805
21.35227
24.67078
19.6946
5.888858
16.21178
1 1 .37603
43.39278
34.86788
13.37501
14.50562
31.38356
17.3668
6.663001
27.4173
23.19643
39.15501
21 .57868

28.76526
12.05761
15.53464
25.43651
18.49287
40.27872
1 1 .34874
17.47383
33.28554
40.82852
12.47845
25.17515
13.4528
1 5.86702
38.21292
39.05667
26.35933
51.0345
30.40948

LCI
124.4562
139.0198
1 47.9606
1 77.0424
141.3632
159.5346
147.0763
168.1714
145.8768
183.1236
111.0739
134.0619
165.6072
203.5245
157.4821
181.0884
122.4736
157.5943
139.7656
141.0784
137.9464
146.2086
153.707

120.5205
160.6138
146.0368
163.0376
122.2214
125.2331
130.2227
147.5197
169.5716
195.4557
104.5216
111.7185
44.09005
51.13561
58.85851
72.87791
187.6407
199.6604
142.4477

UCI
195.5438
226.4158
189.7537
208.8571
215.2082
249.5107
189.7808
217.513
185.266
194.9013
143.4975
156.8139
252.3928
273.2602
184.2322
210.0996
185.2407
192.3279
153.0916
195.9129
184.3393
224.5186
196.8644

178.051
164.7291
177.1061
213.9106
159.2072
205.7905
152.9202
182.4674
236.1427
277.1127
129.4784
162.0688
70.99566
82.86965
1352843
150.9912
240.3593
3017294
203.2666

Mean%Re
80
9 1.85889
84.42857
96.47489
89.14286
102.2613
84.21429
96.42109
80.96402
94.5062
63.64286
72.71894
103.5679
119.1962
85.42857
97.797
76.92857
87.48054
73.21429
84.24783
60.57143
92.6818
87.64286

74.64286
86.33573
80.06513
94.23703
70.35714
82.7559
70.78571
82.49677
101.4286
118.1421
58.5
68.44684
28.77143
33.50132
48.53571
55.96729
107
125.3474
86.42857

%RSD
15.85662
17.36617
8.833287
5.884657
14.7823
15,70034
9.048894
9.131627
8.490417
2.223863
9.09114
5.583164
14.81967
10.44
, 5.587637
l~ 5.293561
14.55968
7.085093
3.247965
L 11.61457
10.2749
15.07753
8.787102

13.75363
4.984345
6.862841
9.633259
9.380659
17.37056
5.721888
7.559421
11.71204
12.33378
7.612757
13.12671
16.6874
16.90327
28.09869
24.90567
8.792004
14.53067
12.55709

CHACC
11.85889
-7.43032
12.04631
-7.33203
8.595805
-13.5244
12.2068
-15.4571
13.54218
-30.8633
9.076084
23.71317
-15.6283
.4.624747
12.36843
-20.8684
10.55197
-142663
11.03354
-3.6764
12.11037
-5.03894
-87.6429

11.69267
-6.2706
14.1719
-23.8799
12.39876
-11.9702
11.71105
16.07466
-16.7135
-23.3579
9.946844
-39.6754
4.729886
15.0344
7.431575
37.03271
-18.3474
11.77602
-86.4286

REL%UNC
25.75486
-532495
-23.8756
132.1096
21.84452
-525379
15.54175
-20.1704
-70.0991
175.2959
-29.8286
281.4405
-19.6459
-61 .6409
8.453149
116.3545
-44.6627
-61.6337
311.4857
-15.3949
68.79756
-44.8891
-100

-58.0827
28.83677
63.74059
-272979
117.8068
-71.8245
53.97156
90.48787
22.66144
-69.4369
101.749
-465632
17.94579
140.8324
2.208025
-32.51
93.61075
-40.4139
-100

-------
Table 1 (Cont.)
AROMATICS
BENZENE
CORRECTED RECOVERY
TOLUENE
CORRECTED RECOVERY
CHLOROBENZENE
CORRECTED RECOVERY
ORTHO XYLENE
CORRECTED RECOVERY
ETHYL BENZENE
CORRECTED RECOVERY
STYRENE
CORRECTED RECOVERY
JM+P1-XYLENE
CORRECTED RECOVERY
BENZENE-D6
BROMOFORMS
BROMOMETHANE
CORRECTED RECOVERY
BROMODICHLOROMETHANE
CORRECTED RECOVERY
DIBROMOCHLOROMETHANE
CORRECTED RECOVERY
BROMOFORM
CORRECTED RECOVERY
BROMOFORM -13C
GASES
CHLOROMETHANE _^
CORRECTED RECOVERY
VINYL CHLORIDE
CORRECTED RECOVERY
CHLOROETHANE
CORRECTED RECOVERY
BROMOMETHANE
CORRECTED RECOVERY
VINYLCHLORIDE-D3

1600
1849.711
101.9143
117.82
214
247.3988
109.3571
126.4244
118.5714
147.9728
116.5714
145.7193
170.8571
212.904
162.1429

167.5714
188.2825
148.8571
167.2552
111.0286
124.7512
139.4286
156.6613
167.5714

160
222.2222
149.2857
207.3413
168.8571
234.5238
167.5714
232.7381
161.1429

74.16198
85.7364
10.4391
12.06832
18.81489
21.75132
7.397715
8.552272
9.692904
19.23338
2.878492
17.90254
12.46901
23.87467
20.98752

12.35391
13.8808
20.26844
22.77352
31.21467
35.07266
6.214423
6.982497
12.92101

25.37059
35.23693
20.5322
28.51695
14.91564
20.71616
12.35391
17.15821
15.44267

103.8997
120.1152
14.62499
16.90751
26.35933
30.47321
10.36407
11.98158
13.57959
26.94564
4.032717
25.08115
17.46887
33.448
29.40316

17.30761
19.44675
28.39573
31 .90531
43.73121
49.13619
8.706298
9.782357
18.1021

35.54375
49.36632
28.76526
39.95175
20.89655
29.02298
17.30761
24.03835
21 .63492

1496.1
1729.596
87.28929
100.9125
187.6407
216.9256
98.99307
114.4429
104.9918
121.0271
112.5387
120.6382
153.3883
179.456
132.7397

150.2638
168.8358
120.4614
135.3499
67.29736
75.61501
130.7223
146.879
149.4693

124.4562
172.8559
120.5205
167.3895
147.9606
205.5008
150.2638
208.6997
139.5079

1703.9
1969.826
116.5393
134.7275
240.3593
277.8721
119.7212
138.406
132.151
174.9184
120.6041
170.8005
188.326
246.352
191.546

184.879
207.7293
177.2529
199.1605
154.7598
173.8874
148.1349
166.4437
185.6735

195.5438
271 .5885
178.051
247.293
189.7537
263.5468
184.879
256.7764
182.7778

88.88889
102.7617
47.05184
58.90999
57.1276
123.6994
51 .58356
63.21222
50.24213
73.98639
57.56614
72.85965
79.10053
106.452
81.07143

83.78571
94.14125
74.42857
83.62761
55.51429
62.3756
64.55026
78.33066
83.78571

80
111.1111
74.64286
103.6706
84.42857
117.2619
83.78571
116.369
80.57143

4.635124
4.635124
10.24302
10.24302
8.792004
8.792004
6.76473
6.76473
8.174738
12.99792
2.469294
12.28564
7.297915
11.21382
12.94385

7.372324
7.372324
13.61603
13.61603
28.11409
28.11409
4.457065
4.457065
7.710745

15.85662
15.85662
13.75363
13.75363
8.833287
8.833287
7.372324
7.372324
9.58322

8.34939
-50.1864
11.85815
-1.78239
19.17298
-24.717
1 1 .62866
-12.9701
23.74426
-16.4203
15.29352
6.240875
14.44745
-12.4766
-81.0714

10.35554
-19.7127
9.199037
-28.1133
6.861316
2.174663
13.78039
5.455056
-83.7857

8.888889
-14.246
21 .68651
-11.9008
-1.69048
1.047619
-0.15476
-3.05952
-80.5714

15.60694
-87.8242
15.60694
55.9031 1
15.60694
-65.9896
15.60694
13.33718
98.42748
-85.0339
521.9419
-30.3506
91 .47209
-12.0929
-100

12.35955
46.01783
12.35955
37.06562
12.35955
-82.2813
12.35955
85.04849
-100

38.88889
-41.731
38.88889
-47.6955
38.88889
-40.3658
38.88889
-9.99832
-100

-------
                                    TABLE  2

                    SUMMARY OF RESULTS FOP CASE A IV TABLE 1
Compound
% Uncertainty
                                           Chance In Accuracy
vinyl chloride
1 , 1-dichloroethene
chloroform
1 , 2 -dichloroethane

carbon tetrachloride
trichloroe thane
benzene

tetrachloroethene
chlorobenzene
hexachloroethane
nitrobenzene
hexachlorobutadiene
hexachlorobenzene
o-cresol
pentachl orophenol
2 , 4-dinitrophenol
pyridine
2,4, 6 -trichl orophenol
2,4, 5-trichlorophenol
m&D cresol
MEAN
+68.70
+ 1.37
+75.53
-15.41

+11.84
-41.73
+245.2

+59.89
-49.46
+58.69
- 2.97
+169.2
+ 5.99
- 9.43
-55.22
- 6.71
-69.50
-55.99
+ 7.18
-10.29
+19.3
+14.3
+23.6
+20.1
+ 0.86

+11.61
+ 6.34
+ 8.89

- 0.77
-11.41
+26.75
+ 2.60
+25.96
+ 0.38
+ 9.76
+12.65
+17.36
+34.1
- 2.84
+ 4.14
-17.00
+ 9.36



* corrected
value >100%


* corrected
value >100%














Ten parameters, or 50% had  an increase in uncertainty that ranged from +1.4% to
+245%.  Ten parameters,  or  50% had a decrease in uncertainty that ranged from -
2.97% to -69.50%.   Sixteen parameters, cr 80% had  an increase in accuracy that
ranged from +0.86 to +34.1%.  Two corrected values had increases in accuracy b_.
exceeded the maximum, or true value.  Four parameters experienced a decrease in
accuracy with a range of  -0.77% to  -17%.

Overall the  uncertainty  increased by an  average of  19%  with only  a margin?-"
improvement in accuracy  of  9.36%.
                                    1-103

-------
                     TABLE  3
SPTTCP  RFCOVERY  RESUI.TS FOR OTHER NINE SAMPLES
ANALYZED IN SAME BATCH WITH

vinyl chloride
1 , 1-dichloro
ethene
chloroform
1,2-dichloro
ethane
carbon
tetrachloride
trichloroethene
benzene
tetrachloro-
ethene
chlorobenzene
o-cresol
hexachloroethane
(n&p) -cresol
nitrobenzene
hexachloro-
butadiene
2,4, 6-trichloro-
phenol
2,4, 5-trichloro-
phenol
2,4-dinitro-
toluene
hexachlorobenzene
pentachlorophenol
pyridine
65984
95
105
100
95
80
95
100
80
95
60
77
130
75
90
115
104
92
86
114
29
66092
85
95
90
84
80
95
85
75
85
60
80
130
60
94
105
100
74
86
108
36
65058
85
80
85
75
85
90
90
85
90
50
97
99
85
138
175
90
118
88
104
28
65560
85
75
85
80
85
90
95
80
95
72
97
170
215
70
100
116
98
20
61
6
THE SPM
65068
95
95
105
110
105
95
100
95
100
62
77
130
70
58
85
88
86
54
104
10
(Case
65781
85
100
115
135
95
105
90
95
90
108
107
53
80
106
100
172
90
88
90
25
B)
64697
130
135
130
105
135
95
110
90
95
100
130
112
95
158
7
9
170
112
8
75
65238
95
105
107
120
125
120
112
100
110
58
80
110
73
96
65
60
78
68
54
95
62836
120
120
110
97
110
100
110
95
105
17
10
28
24
10
0.15
-
26
28
™
14
                     1-104

-------
                    Table 4 BIAS CORRECTION by COMPOUND CLASS
o
(fl
NKroaromatlct
ANALYTE 200PPB MW MATRIX
NITROBENZENE
Corrected % Recovery
2,4-DiNITHOTOLUENE
Corrected % Recovery
NITROBENZENE-OS
Acid Cournpounds
ANALYTE 200PPB MW MATRIX
4-NITROPHENOL
Corrected % Recovery
PENTACHLOROPHENOL
Corrected % Recovery
O-CRESOL
Corrected % Recovery
0-CRESOL-D8
NKrosoamlnes
ANALYTE 200PPB MW MATRIX
N-NITROSODI-N-PROPYLAMINE
Corrected % Recovery
N-NITROSODIMETHYLAMINE
Corrected % Recovery
N-NITHOSCOIMETHYLAMINE-D6
Base Compounds
ANALYTE 200PPB MW MATRIX
ANILINE
Corrected % Recovery
2-NITRO ANILINE
Corrected % Recovery
PYRIDINE
Corrected % Recovery
N-NITROSCOIMETHYLAMINE-D8
PNA
ANALYTE 200PPB MW MATRIX
NAPHTHALENE
Corrected % Recovery
ACENAPHTHENE
Corrected % Recovery
BENZO(A)PYRENE
Corrected % Recovery
BENZO(A)PYRENE-D12
CHLORINATED HYDROCARBONS
ANALYTE 200PPB MW MATRIX
HEXACHLOROBENZENE
Corrected % Recovery
1 ,2,4-TRICHLCflOBENZENE
Corrected % Recovery
HEXACHLOROBENZEN E - 1 3C8
PHTHALATES
ANALYTE 200PPB MW MATRIX
DIMETHYLPHTHALATE
Corrected % Recovery
DI-N-BUTYLPHTHALATE
Corrected % Recovery
Dl -N-OCTYLPHTHALATE
Corrected % Recovery
Dl -N -BUTYLPHTHALATE-D4

Mean
169.1429
216.5031
182.7143
233.8565
156.4286

Mean
168
190.4826
107.8571
122.5139
2.485714
2.793719
176.8571

Mean
159.8571
188.4448
196
229.8435
170

Mean
10328.14
12189.26
172.1429
204.1354
184.5714
214.7
170

Mean
100.7286
113.1769
51.01429
57.25054
116.5143
131.5771
178.2857

Mean
136,8714
141.4848
44.57143
47.43001
189.4286

Mean
0.142857
0.223214
25
38.17832
113.7714
174.7018
130.4714
STO
11.80838
10.37165
17.4233
19.22284
11.57378

STD
35.6376
39.44595
20.16919
24.17091
1.181
1.28363
13.10761

STD
12.07516
11.6583
30.04441
23.6537
14.14214

STD
685.0029
1255.822
31.87177
44.57137
63.58984
64.41604
14.14214

STD
4.606776
4.42547
3.174602
1.168741
44.69252
45.00046
11.70063

STD
60.66357
51.75928
2.592755
4.224737
22.48597

STD
0.377964
0.590569
7.450727
9.433954
35.66355
47.87414
20.67903
Cl
16.68501
14.53188
24.41205
26.93341
16.2182

Cl
50.21259
55.26831
28.25935
33.86622
1.854717
1.798513
18.36527

Cl
16.91869
16.33462
42.09566
33.14156
19.81476

Cl
1239.991
1759.552
44.65601
62.44861
88.11083
80.25428
19.81476

Cl
6.454622
6.200593
4.447982
1.63754
62.61936
63.05082
16.39393

Cl
84.99664
72.52071
3.632748
5.818343
31.50544

Cl
0.529572
0.827456
10.43833
13.21805
49.96874
67.07717
26.8737
LCL
152.4578
201.8712
158.3022
207.0231
140.2124

LCL
117.7874
135.2143
79.58778
68.6477
0.630997
0.895206
158.4919

LCL
142.9385
172.1101
153.8043
196.602
150.1852

LCL
8086.152
10439.73
127.4886
141.6856
95.4608
124.4457
150.1852
v
LCL
84.27395
106.9763
46.5663
55.613
55.88492
68.52625
161.8918

LCL
51.87478
68.96424
40.93866
41.51067
157.9231

LCL
-0.38671
-0.60424
14.56067
24.96027
63.80269
107.6246
101.4977
UCL
185.8279
231.0348
207.1263
260.8899
172.6448

UCL
218.2126
245.7508
136.1165
158.3601
4.140432
4.592232
195.2224

UCL
176.7758
204.7794
238.0957
263.0851
189.8148

UCL
11568.13
13958.83
216.7989
266.585
273.6823
304.8543
189.8148

UCL
107.1832
119.3775
55.46227
58.88806
181.1336
184.6278
184.6796

UCL
221.8681
214.0057
48.20416
53.34936
220.934

UCL
0.672429
1.05067
35.43933
51.39637
163.7402
241.778
159.4451
MEAN%RE
83.73408
107.1797
91.35714
115.8201
78.21428
%RSD
7.040428
4.790534
9.535818
8.216415
7.398763

MEAN WE
84
95.24131
53.82857
61.25896
1.242857
1.396858
88.42857

MEANER!
78.92857
94.22238
98
114.9718
85

MEAN*fll
106.9387
126.3127
86.07143
2.11364
92.28571
2.223028
85

MEAN Wl
38.59332
43.36278
25.13019
21.93507
58.96233
50.41267
89.14286

MEANWH
68.43571
70.74247
22.26571
23.71501
94.71429

MEANER!
0.071429
0.111607
12.5
19.08916
56.88571
87.35091
65.23571
%RSD
21.3319
20.70843
16.69991
19.72811
47.51151
45.947
7.411413

%RSD
7.55372
6.186586
15.32678
10.28674
8.318803

%RSD
8.568648
10.29423
18.51472
21.83422
34.45817
30.00281
8.318903

WSD
4.573455
3.810225
6.222966
2.041449
37.71066
34.20084
6.582854

WtSD
44.32157
36.58289
5.81708
8.807308
1 1 .87042

*flSD
284.5751
264.5751
28.80291
24.71024
31 .34667
27.40334
15.84947

CHACC
8.086178
-1.46312
-7.1772
-5.86566
-78.2143

CHACC
11.24131
-41.3127
7.328386
-60.0141
0.154002
87.03171
-88.4286

CHACC
14.29381
3.777619
-12.8718
-0.02824
-85

CHACC
-18.374
12.36415
-83.8578
90.17207
-90.0627
82.77897
-85

CHACC
4.769462
-18.2326
-3.19512
37.02726
-8.54966
36.73019
-89.1429

CHACC
2.306761
-48.4568
1.429292
70.99928
-94.7143

CHACC
0.040178
12.38839
6.58916
37.78655
30.46519
-22.1152
-65.2357

REL%UNC
-12.9046
87.98963
10.32634
-39.7915
-100

REL%UNC
10.06863
-48.8668
19.84075
-95.114
8.690042
921.1366
-100

REL%UNC
-3.45224
157.7084
-21.2708
-40.2117
-100

REL%UNC
41.90038
-97.4621
39.84593
42.69237
1.283178
-78.0456
-100

REL56UNC
-3.93562
-28.2652
-63.1647
3723.99
0.689022
-73.8888
-100

REL%UNC
-14.6781
-94.9907
62.94382
432.2455
-100

REL%UNC
56.25
1161.617
26.6178
278.034
34.23627
-56.8054
-100

-------
                 Or  APPLICATION Or EIAS  CORRECTION TC* SPJ?
                                                                     C EPA
                               OK A BATCT  BASIS ftror Table
                         True    Amount                    Bias       Change In
"propound                Value     Found   * Recovered    Corrected     Accuracy

1,1-dichloroethene        100         67        75            89          +22.0
1,2-dichloroethane        100         93        75           124          -17.0
tetrachloroethene         100         75        75           100          +25
3-cresol                 1250       1036        60          1727          -21.0
o-cresol                 1250       1036        60          1727          -21.0
o-cresol                 1250       1036        50          2072          -4E.6
o-cresol                 1250       1036        72          1439          +2.4
o-cresol                 1250       1036        C2          1671          -16.6
o-cresol                 1250       1036        58          1786          -25.8
O-cresol                 1250       1036        17          6094          -370
hexachloroethane          750        369        77           479          +14.7
hexachloroethane          750        369        77           479          +14.7
hexachloroethane          750        369        10          3690          -341
      cresol             2500       3595        53          67E3          -12E
      cresol             2500       3595        28         12639          -370
nitrobenzene              500        421        75           561          +3.6
nitrobenzene              500        421        60           762          -24.6
nitrobenzene              500        421        70           601          -4.4
nitrobenzene              500        421        73           577          +C.4
nitrofc-nzene              500        <21        24          1754          -225
hexachlorcbutadiene      125         33        70            47          +12
hexz.chlorobutadiene      125         33        58            57          +24
bexaciilorobutadiene      125         22        10           330          -50.4
2,4,C—trichli'Tophenol    500        419         7          5986          -1080
2,4,6-triChlcrophenol    500        419        65           645          -12. S
2,4,6-trichlorophenol    500        419        <5           279
2,4,5-trichlor^phencl   1250        618         9          6867         -399
2,4,5-trichi-rophencl   1250        61E        60          1030          +33.0
2,4,5-trichlcrophenol   1250        618        <5
2,4-dinitrotoluene        125        100        74           125          +12.0
2,4-dinitrotoluene        125        100        78           128          +17.6
2,4-dinitrotoluene        125        100        26           3E5         -1EE
hexachlorobenzene         125          4.4      20            22          +14.1
bexachlcrobenzene         125          4.4      54             8.15        +3.0
hexachlorobenzene         125          4.4      68             6.47        +1.7
hexachlorobenzene         125          4.4      28            15.7         +5.0
per.tachlcrophenol       12500     10350         8        129400          -510
per.tachicrophenol       12500     10350        54,        1S170          -36.2
pentachiorophenol       12500     10350        <5
pyridine                  500        226        29           942          -33.6
pyridi-e                  500        226        36           628          +25.2
pyridine                  500        226        28           8C7          -6.6
ryridine                  500        226         6          3767          -555
pyridine                  500        226        10          2260          -257
pyridine                  SCO        226        2t           504          -26.0
pyridine                  SCO        **£        75           301          +15.0
cyridine                  500        226        14          1614          -16E

Tat entire set of data had 200 applicable data points. Forty-seven recoveries fit the requirements set by EPA
guideline? for accuracy adjustment. Of these 47 bias corrections. 18, or 38%. resulted in an increase ir. accuracy
cf the r.u.—ber v.-::: respect to the true vilue averaging K.8%. Twenty-nine of the bias corrections resulted in
recoveries that reduced the accuracy of the data. This represented 62% of the corrected data. In calculating
the i\tnst chzrce in accuracy, three recoveries below 55c were not included. The overall average reduction
in ti._.-ar.- was 119%.
                                     1-106

-------
                                  TABLE 6
                SUMMARY OF RESULTS FOR CASE C SEHIVOLATIIES
COHPOUHD
                               I UUCERTAIMTY
                                                       CHANGE IK ACCURACY
nitrobenzene
2.4-DNT
4-nitrophenol
pentachloro
o-cresol
N -ni tr oso-pr opy 1
K-nitroso dinethylanine
aniline
2-nitroaniline
pyridine
naphthalene
ecenapthalene
benzo[o]pyrene
hexachloroe thane
1 ,2.4-trichlorobenzene
hex* chl or obenzene
dia«thylpthalate
di-N-butylpthalate
di -fi-octvlT>thalate
-12. e
10.33
10.07
19.84
8.69
-3.45
-21.27
41.9
22.97
1.17
-3.64
-6.82
0.69
29.2:
6.46
-14.6
HE
26.62
34.24
9.09
-7.18
11.24
7.33
0.16
14.29
-12.97
-19.37
11.63
-0.41
4,77
3.67
6.E
1.86
1.43
2.31
NR
6.59
30.46
  MEAK
                                  «13.2
                                                              «4.24
Compounds  which  were  connected  by its  own isotope  were  not used  tor
averaging purposes (see Table?).
                                    TABLE 7
                    SUMMARY OF RESULTS FDR CASE C VOLATILE
  COMPOUND
                                   UNCERTAINTY
                                                         CHANGE IK ACCURACY
chlorone thane
chloroe thane
oethylene chloride
1 ,1-dichloroe thane
chlorofont
carbontetrachloride
1 ,2-dichloroethane
1 . 1 ,2-trichloroethane
1 ,2-dichloropropane
1 . 1 . 2 . 2-tetrachloroethene
1.1.1 -trichloroe thane
vinyl chloride
cis- 1 ,2-dichloroethane
trans -1 ,2-dichloroethene
trichloroethene
cis-l ,3-dichloropropane
1 . 1 -dichloroethene
trans -1 ,3-dichloropropane
tetra chl or oe thene
chlorobenzene
benzene
toluene
chlorobenzene
ortho-xylene
ethylbcnzene
styrene
(a»p)xyl«ne
brooonethane
bromodichlorome thane
dibrooochlorome thane
brooolont
25.75
-23.88
21.84
15. 54
-70.10
-29.83
-19.64
-44.66
8.45
311.48
68.80
-58.08
63.74
117.81
53.97
22.66
101.75
17.94
2.21
93.61
15.61
15.61
15.61
15.61
98.43
523
91.47
12.35
12.36
12.36
12.36
11.86
12.05
8.60
12.21
11.72
9.08
11.7
10.55
12.37
11.03
12.11
11.69
13.45
12.40
11.71
-16.71
9.95
4.73
7.43
-16. 3c
8.1
7.38
8.7
8.0
12.45
14.57
6.12
10.36
S.20
6.86
8.62
    MEAK
                                    -55.12
  Compounds which  were  connected  by  its  own isotope  were  not used  for
  averaging purposes (see Table ?).
                                                    1-107

-------
                           Figure 1

                                 ©ff        .
Control Sample 3 liters

Analyze in duplicate

 Non-detcet levels for
Duplicate Spike


good Recoveries
                      Landfill Leachate

                      Sample 11 liters
                                   Experimental Sample

                                    7 liters spiked with

                                     OTC Constituents
                                         SRM of Known Concentrations
7 Replicates
volatile analysis
                                                  7 Replicates

                                                   Semivolatile Analysis
  Spike Recovery Corrected

  Concentration in SRM based
  on value of each of 9 other

  spike replicates in batch
                                 Spike Recovery Corrected

                                  Concentration based on isotope

                                 spikes to SRM Sample
                                Figure 2
         Comparison of Individual vs Batch Based Corrections

     +500
  £  300
  o
  o
  c
  0)

  2
  Q>
  C-
        0
     -300
                JS.
                       Mean
                       Batched based corrections
       cr\r\ I   LJ  * Individual corrections-
       QUU Nitro-
                       Hexachloro-    0-Cresol   2.4-Dinitro-   Pyridine
             Benzene   butadiene               Toluene
                                 1-108

-------
                                        TABLE 1  CASE C   ORGANIC VOLATILES BY (XMPOUND CLASS
8
SATURATED C Ml OROC ARSONS
COMPOUND 200PPB IN DIH20
CHLOROMETHANE
CORRECTED RECOVERY
ChLOROETHANE
CORRECTED RECOVERY
METHYLENE CHLORIDE
CORRECTED RECOVERY
1 1-DICHLOflOETHANE
CORRECTED RECOVERY
CHLOROFORM
CORRECTED RECOVERY
CARBON TETRACHLORIDE
CORRECTED RECOVERY
1 2-DICHLOROE1HANE
CORRECTED RECOVERY
1.2-OICHLOROPROPANE
CORRECTED RECOVERY
1,1.2-TRlCHLOROETHANE
CORRECTED RECOVERY
1,1,2 2-TETRACHLOHOETHANE
CORRECTED RECOVERY
\ 1,1 -TRICHLOROETHANE
CORRECTED RECOVERY
CHLOROFORM- 13C
UNSAT U RATED CHLOROCARBONS
VINYL CHLORIDE
CORRECTED RECOVERY
CIS- 1.2-DICHLdHOETHENE
CORRECTED RECOVERY
TRANS- 1.2-DICHLOROETHENE
CORRECTED RECOVERY
TRICHLOROETHENE
CORRECTED RECOVERY
CIS- 1,3-DICHLOROPROPENE
CORRECTED RECOVERY
1,1-DICHLOROETHENE
CORRECTED RECOVERY
THANS- 1,3-DICHLOROPROPENE
CORRECTED RECOVERY
TETRACHLOROETHENE
CORRECTED RECOVERY
CHLOROBENZENE
CORRECTED RECOVERY
1.1- UCHLOROETHENE-D2

Mean
160
183.7176
1600571
192.94 i*0
1782857
204.5227
160.42U6
1928422
165.5714
1890124
127.2057
145.4379
209
238.3924
170.8571
195594
1538571
174.9611
146.4206
1684957
161.1429
1853636
175.2657
0
149.2857
172.6715
161.5714
1884741
140.7143
165.5118
141.5714
164.9935
202.8571
236.2842
117
136.8937
57.54286
67.00263
97.07143
111.9346
214
250.6949
1 72.857 1

STO
2S37U59
31.90475
14.91564
11.25443
26.35472
32.1117'J
152409<>
17.601/63
140577
4.203377
11.57172
8.120035
30 a/a 11
24. eta 16
9.546877
10.35389
22 401 1 1
1239616
4.755949
19.57005
16.55726
27.94825
15.40254
0
205322
8.606541
11.08839
18.15619
13.19993
28.75034
8.100558
12.47256
23.75871
29.14277
8.906926
17.96964
9.602405
11.32563
27.2758
27.87005
18.81489
36.42765
21.70583

Ci
35 543/5
44.6U799
20.89655
1 59073U
3GU225I
44.9UU05
2135227
24.U7078
19.6046
5.0U0858
lli 21 176
11.37603
43.3927U
34.86788
13.37501
14.50562
31.38356
17.3668
6.663001
27.4173
23.19643
3U.15S01
21.57868
0
28.76526
12.05761
15.53464
25.43651
18.49287
40.27872
11.34874
17.47383
33.28554
40.82852
12.47845
25.1 751 5
13.4528
15.86702
38.21292
39.05667
26.35933
51.0345
30.40948

LCI
124.4562
139.0IU8
1 47.9606
177.0424
i4i.36jii
1 Si). 534 6
147.0763
160.1714
145 07C8
183.1236
111.0739
1340U1U
165.6072
203.5245
157.4821
181.0084
122.4736
167.MJ43
139.7656
141.0784
137.9464
146.2086
153.707
0
120.5205
160.6138
146.0368
163.0376
122.2214
125.2331
1302&?
147.5197
16U.5/I6
195.4557
104.5216
111.7105
44.00665
51.13561
50.85851
72.87791
187.6407
199.6604
1 42.447 /

UCI
lift 5430
2204158
10>J.75:»/
200H!./i
2I5.1'OU2
241*611)7
it9.7BOB
217.513
185.260
194.9013
(43.4975
166.8130
262.3020
273 2602
184.2322
2IU.0996
105.2407
102 3279
153.0916
195.9129
184.3393
224.5106
196.8644
0
178.051
184.7291
177.1061
2139106
159.2072
205.7905
152,8202
182.4674
23U.1427
277.1127
12U.4/04
162.0(>80
70.90506
82.06965
135.2843
150.991 2
240.3593
301.72U4
2U3.2tJl>U

Muun%ltu
00
91.8580'J
04.42057
96.4740!)
09.1 4206
102.2. -1 3
04.2142'J
96.421 09
00.96402
94.5002
C3.6420U
72.710U4
)0:i£.li7U
HU.l'Jtii!
85.42057
97.797
76.i)^'057
87.4U054
73.2142'J
04.24/03
00.57143
92 GUI B
07.64200
0
74.64206
86.33573
80.0U513
9423703
70.35714
82.7559
70.7U571
02.4iJa77
101.4286
110.1421
£8.5
68.44U04
20.7/143
33.50132
48.53571
55.06729
107
1253474
ttUMUM
1
%RSD
1S.U5U6^
17.3bU| '
0.U3-J20:'
5.U84tii>/
14.70LM
15.7UOU4
9.U400U4
9.I3IOL'/
6.490417
22^'3U03
U.09114
5&03IU4
14.01007
104-1
5.S07KJ7
S.2U35CI
14.5C000
7.0850U3
3.247&C5
11.61457
10.2749
15.07753
0.707102
ERR
13.75J1.3
4.984345
ti.86284 1
9U3259
9.360659
17.37056
5.72101)0
7.559421
ii.71204
12.33378
7.U12757
13.12671
16,6074
16.110327
28.09069
24.005(27
87U2004
1- 5301J /
t2.S!i/UU

CHACO
II.HMlU'J
-7.4U032
12.04631
~ 7.33203
0.50L.UU5
• 13.i»:M-t
12.1^068
-J54071
13.542 1H
-30 bt,J'J
0 U/LUll-l
^;j./i:ii7
• t^i.b^UU
4 62-1747
\/. JUt-l.l
- 2U U6U4
10.55197
-14.2663
11.03354
-•3.6'/.:«
12.11U3/
-.i.ajU'J-4
-a/.(J4ij
7>..64L'8(i
11.C!)L'07
-6.L70U
14.1719
-23.0799
12.39876
-11.3702
11.71105
16.07466
- 10.7 US
-23.3S/U
9.946044
-39.6754
4.7^Ml)(J
15.0344
7.431575
37.03271
-18.3474
1l.7/uUl>
- U0.40li
i.'8 1.4405
- I'J.li-lfitl
- Ul.u-lDU
U '153140
tlU.3545
41.C627
-61.6337
311.4057
-i 5.304 U
kUy'J7M)
-44.00UI
-100
ERR
-5110027
20 03677
Gi740r/J
- a/L'979
1170068
-71,0^45
53.97 IfG
90.48/07
22(iUK4
-6U4U6U
HII.74U
-4li.5(i32
17.!i-l!>79
140U324
2.200025
-•32.51
93.61075
- 411.4 130
-100

-------
TABLE 1  (continued)
AROMATICS
BENZENE
CORRECTED RECOVERY
TOLUENE
CORRECTED RECOVERY
CHLOROBEN2ENE
CORRECTED RECOVERY
ORTHOXYLENE
CORRECTED RECOVERY
ETHYL BENZENE
CORRECTED RECOVERY
STYRENE
CORRECTED RECOVERY
(MtP)-XYLENE
r*f\anc/'*Tcn acfowcov
UUnHcts 1 1 U HtUUVtnY
BENZENE-DO
BROMOFORMS
BROMOMETHANE
CORRECTED RECOVERY
BROMODICHLOROMETHANE
CORRECTED RECOVERY
DIBROMOCHLOROMETHANE
CORRECTED RECOVERY
BROMOFORM
CORRECTED RECOVERY
BROMOFORM- 13C
GASES
CHLOROMETHANE
CORRECTED RECOVERY
VINYL CHLORIDE
CORRECTED RiCOVERY
CHLOROETHANE
CORRECTED RECOVERY
BROMOMEIHANE
CORRECTED RECOVERY
VINYLCHLORIDE-D3
0
1600
1849.711
101.9143
117.82
214
247.3988
109.3571
126.4244
118.5714
147.9728
116.5714
145.7193
170.8571

cltf.WH
162.1429
0
167.5714
188.2825
148.8571
167.2552
111.0286
124.7512
139.4266
156.6613
h 167.5714
0
160
222.2222
1492857
207.3413
1688571
234.5238
167.5714
232.7381
161.1429
0
74.16198
85.7364
10.4391
12.06832
18.81489
21.75132
7.397715
8.552272
9.692904
19 23338
2.878492
17.90254
12.46901

tfJ.OrWr
20.98752
0
12.35391
13.8808
2026844
22.77352
31.21467
35.07266
6.214423
6.982497
T2.92l5i
0
25.37059
35.23693
20.5322
28.51695
14.91504
20.71616
12.35391
17.15821
15.44267
0
103.8997
120.1152
14.62499
1690751
26.35933
30.47321
1036407
11.98158
13.5?C59
26.94564
4.032717
25.08115
17.46887

33.448
29.40316
0
17.30761
19.44675
28.39573
31.90531
43.73121
49.13619
8.706298
9.782357
18.1021
0
35.54375
49.36632
28.76526
39.95175
20.89655
29.02298
17.30761
24.03835
21.63492
0
1496.1
1729.596
87.28929
100.9125
187.6407
2169256
98.99307
114.4429
1049918
121.0271
112.5387
120.6382
"153.3883

179.455
13277397
0
1502638
168.8358
120.4614
135.3499
67.29736
75.61501
130.7223
146.879
149.4693
0
124.4562
172.8559
120.5205
167.3895
147.9606
205.5008
150.2638
208.6997
139.5079
0
1703.9
1969.826
116.5393
134.7275
240.JS93
277.8/21
119.7212
138.406
132.151
1749184
120.6041
170.8005
188.326
OJA ICO

191.546
0
164879
207.7233
177.2529
199.1605
154.7598
173.8874
148.1349
166.4437
185.6735
0
(95.5438
271.5885
178.051
247.293
189.7537
263.5468
184.879
256.7764
182.7778
0
88.68889
102.7617
47.05184
50.90999
57.1276
123.6994
51.58356
63.21222
5024213
7398639
57.56614
72.859SS
79.10053
IHft 4K9

81.07143
0
83.78571
94.14125
74.42857
83.62761
55.51429
62.3756
64.55026
7U.33006
63.78571
0
80
111.1111
74.64286
103.6706
84.42857
117.2619
8378571
116.369
80.57143
ERR
4.635124
4.635124
1024302
10.24302
8.792004
8.792004
6.76473
6.76473
8.174738
12.99792
2.469294
12.20564
7.297915
n 91000

12.94365
ERR
7.372324
7.372324
13.61603
13.61G03
28.11409
28.11409
4.457065
4.457065
7.710745
ERR
15.85662
15.85662
13753G3
1375363
8.833287
8.833287
7.3723^4
7.372324
9.58322
88.88889
8.34939
-bO.IB64
11.85815
-1.78239
19.17238
-2-. 7:7
iieiooG
-121701
23.74426
-16.4203
15.29352
6.240875
14.44745
— 19 A7RR

-81.0714
83.78571
10.35554
-19.7127
9.199037
-28.1133
6.861316
2.174653
13.78039
5.455056
-8377857
80
8.868039
-14.246
21.68651
-11.9008
-1.69048
1.047619
-0.15476
-3.05952
-80.5714
ERR
15.60694
-87.6242
15.60694
55.9031 1
15.60694
-65.9896
1560694
1333718
90.42748
-850339
521.9419
-30.3506
91.47209
.19 nooo

-100
ERR
1235955
46.01783
12.3595:-
37.06562
12.35955
-82.28(3
12.35955
85.04649
-100
ERR
38.8QU89
-41.731
38.88889
-476955
38.86089
-403658
3000009
-9.99832
-100

-------
                                     TABrg  2

                    SUMMARY OF RESULTS FOR CASE A TN TABU! 1



Compound                 % Uncertainty     Chancre In Accuracy
vinyl chloride
1, 1-dichloroethene
chloroform
1 , 2-dichloroe thane

carbon tetrachloride
trichloroe thane
ber zene

tetrachloroethene
chlorobenzene
hexachl oroe thane
nitrobenzene
hexachlorobutadiene
hexachlorobenzene
o-cresol
pentachlorophenol
2 , 4-dinitrophenol
pyridine
2,4, 6-trichlorophenol
2,4, 5-trichlorophenol
n&D cresol
MEAN
+68.70
+ 1.37
+75.53
-15.41

+11.84
-41.73
+245.2

+E9.89
-49.46
+58.69
- 2.97
+169.2
+ 5.99
- 9.43
-55.22
- 6.71
-69.50
-55.99
+ 7.18
-10.29
+19.3
+14.3
+23.6
+20.1
+ 0.86

+11.61
+ 6.34
+ 8.89

- 0.77
-11.41
+26.75
+ 2.60
+25.96
+ 0.38
+ 9.76
+12.65
+17.36
+34.1
- 2.84
+ 4.14
-17.00
+ 9.36



* corrected
value >100%


* corrected
value >100%














Ten parameters, or 50%  had an increase in uncertainty that ranged from +1.4% to
+245%.  Ten parameters,  or 50% had a decrease in uncertainty that ranged from -
2.97% to -69.50%.   Sixteen parameters, cr  80%  had an increase in accuracy that
ranged from +0.86 to +34.1%.   Two  corrected values had increases in accuracy b_.
exceeded the maximum, or true value.  Four parameters experienced a decrease in
accuracy with a range of -0.77% to -17%.

Overall the  uncertainty  increased by  an  average of  19% with  only  a margin?-*
improvement in accuracy  of 9.36%.
                                    1-111

-------
                                    TABLE 3
                 SPIKE RECOVERY RESULTS  FOR OTHER NINE SAMPLES
ANALYZED IN SAME BATCH WITH

vinyl chloride
1 , 1-dichloro
ethene
chloroform
1,2-dichloro
ethane
carbon
terra chl oride
trichloroetbene
benzene
tetrachloro-
ethene
chiorobenzene
o-cresol
hexachloroethane
(afcp) -cresol
nitrobenzene
hexachloro-
butadiene
2,4, 6-trichloro-
phenol
2,4, 5-trichloro-
phenol
2,4-dinitro-
toluene
hexachlorobenzene
pentachlorophenol
65984
95
105
100
95
80
95
100
80
95
60
77
130
75
90
115
104
92
86
114
66092
85
95
90
84
80
95
85
75
85
60
80
130
60
94
105
100
74
86
108
65058
85
80
85
75
85
90
90
85
90
50
97
99
85
138
175
90
118
88
104
65560
85
75
85
80
85
90
95
80
95
72
97
170
215
70
100
116
98
20
61
THE SRM
65068
95
95
105
110
105
95
100
95
100
62
77
130
70
58
85
88
86
54
104
(Case
65781
85
100
115
135
95
105
90
95
90
108
107
53
80
106
100
172
90
88
90
B)
64697
130
135
130
105
135
95
110
90
95
100
130
112
95
158
7
9
170
112
8
65238
95
105
107
120
125
120
112
100
110
58
80
110
73
96
65
60
78
68
54
62E36
120
120
110
97
110
100
110
95
105
17
10
2?
24
10
0.15
-
26
28
™
pyridine
29
36
28
10
                                    25
                                     75
95     14
                                    1-112

-------
blAS COIIIItCllGll BrCOMI-OUtlO CLASS
                          TABLE 4 CASE C  SET-UVOLATILES BY CmPOUND CLASS
Nmctiamilici
ArjAI* 1 E 200f-P6 MlV MAlftlX
hi mo BENZENE
C.ji«ll«J 1l flKOvwy
24-UNIIHOICtLllfNE
^Mutuifu^m 	
Ai.it! Cuunipouridt ^ 	
Al.AirlE 2uiiPPB MM MAIRlX
4-l«llflOPll£NOl

HMUCmGHOPHENOl
4U>K> K 4 % fWcovtiy
O ClllSOL

O-CftESOt-08

AIMUIE 2001-Pb MW MAIHlX

CwAHiUdS R«to,«i»
M-I4IIHOSOUMCIIIYIAMINE
CX,r*4l» J V JWortiy
N- UlIHGSGUMtlllYLAMlMt-U*
U»« Compound*
Al.AiriE 2001'CB MM MAIRlX
Al.lllHE
C4"i«il»4 % llacoveiy
J-NIIHOAHIIIHE
4Zoii*i.UU V R«co**tfy
Crinl'iiE
Con.kl.J» llx.o.«i»
(1 - Ml HOSGUMl 1 IIYIAMINE - 06
I'NA
ANAirlt 200PPBUW MAlftlX
N**'II|HAIENE
CUIIIKU J Ik ftacovtiy
ACI IIAPlllHENE
CuuviU J Ik ftoc&vety

fiMiAiU J It ll«KOv«iy
III li/G(A|f Hiflii Oi;
CIMOIIINAMU IIYUIIOCAHIIOK
AUAltlt tliuCrHuAMAINX
IILXACmGIIGUFllJI Nt
C..H.II.J*. llM.Uvi.-iy
124 IlllCltl GHOIldl/ljNS 	
Cuiitii l»d % Aucuvvly
III AACIIlOUOBtllttHt- I1C6
rniiiAiAli s
[MMI 1 H>i I'lililAi Alt
CunwvUd 1L Hucnvbiy
U H bUIVIHIIIIAlAIfc
CnHvclNil Ik Racovuiy
m-M GCIYIPIIIIIAIAIE
CtKIL'tl-* 1 A MuCu.uly
UN IjUlltl'HlHAIAIt- 04

M««n
1691429
2165011
1*2/141
1564266
Miin 	
16*
1904626
10/8571
1225139
2 4b5/ 14
2/93/19
1/645/1

M«n
15965/1
168444*
196
2299435
170

Metn
10126 14
I2t»9i*
1/2.1429
204 1354
1145/14
2147
I/O

M**n
1007266
113 U6»
51 0142»
57 25054
1115141
1315/71
1/42657
Maan
136(714
jit 4B4»
445/141
47 431-01
164 4266
Mean 	
0 I4?U5/
022XM4
3* 1/t32
1117/14
174 70 II
1104/14

SID
1 1 9061*
101/165
17 4231
19 22264
"l 15/3/6
sidT"
35 *)/6
19 44595
20 16919
24 1/091
1 1(1
1 <*'Jtl
13 10/61

SID
120/516
II 6561
1004441
216517
14 14214

SID
1650029
1255 122
316/177
44 Vll/
635^194
64 41604
14 14214

SID
4 tuC/'/»
442547
3 l/4i02
1 1(6/41
44 t!HM
4500046
i i IMSI
510
to to.it/
5l 1'jUlt
4224/17
2^ 4^'b/
sin ~~
03/7SI64
0 1,'jljidi)
/ 41.0727
* 4.IM54
34 6G.I5.S
4/6/414
206)S)u3

a
16 66501
1453186
2441205
26*1341
" 16 2 162
a —
5921259
55,21*31
21 25035
31 16622
1654717
1 798511
1*36527

a
1691669
1613462
42 08568
13 14156
19114/6

Cl
1219991
1750552
44 tUiOl
6244961
6911061
9025428
19*1476

Cl
6454622
* 200591
444/962
1 63/54
626l».lfi
610'«32
I61..I!.]
Cl
64 9%b4
72 Sin /I
J632/4J
31 5IC.I.
cP" "
0 52U'j/2
u62/4:,&
I04UUI1
49»u8M
67 0/7 I/
id9/J/

ICL
152457*
201*712
15*1022
29.' 9??!
1402124
LCl
11776/4
1352141
79597/9
6664/7
0 «3UUB/
0995206
15*49 19

ICL
14293*5
I/2IIOI
1539041
196 602
1501652

ICL
9086 152
10419 73
12/4666
141665*
95410*
124 445/
1501852

ICL
94 27195
1069.761
465661
55 6ll
556UIU2
66 52625
1618911
LCI
51 U/4/U
0041. 1^4
40 >:uta
41 5IUU/
I57K21I
ici
- 0 JBb/ 1
•0104^4
14 5iAll>7
249UI//
6 3 btvt*!*
ICI/U246
1411 4/
48 2U4I8
53 3493C
220914
UCL
0 672423
1 U.',l«,/
31 4.U13
163/4(12
241/79
1594451

MtAUxn
61 /.HO*
10/.I797
91 35714
II5«20I
7*71429
Ml AN Mil
14
9524I1J
5192857
6125696
1 242857
1 3%659
4842857

MEAN Ml
79 U2657
9422238
96
1149711
IS

MEAN VIII
1068187
1263127
6007)41
211164
82285/1
2221026
85

MEAN XII
36 59132
41362/1
25 13019
21 91507
»« 8G211
50 4 Fit/
19. 14286
MI-AHXfll
(.6 41571
/O 74<>4/
21 71501
114/1429
MTAM V.IIIJ
II U7 142U
I'j uau46
t,lil>tt!>/l
HI lUjul
£5?157I

M1SD
704i>uu
4./UU534
U 515618
J2J64I5
* 7,396763
MISO
21 3119
20 70643
18C999I
19 l»ll
4751151
45947
7.4 114 lit

M1SD
75537!
6 166566
15.32671
1026674
I3IU03

MISU
6 56M4f
1029423
16.514/2
21(1422
3445817
lUIAWlll
1.116903

...*n*p
4.S7345J
3910225
6 272966
204144!!
37 7 101.6
14 200*4
I 5G2K54
"~ xnso '
4432157
36 58269
sai/ua
1 1 87U42
MISI.
2645/51
2 1.4 5/51
24 /I624
It 14l.fi/
27 4U314
15 84947

Cll ACC
906617!*
-1.46112
-7.17/2
'-/e'Sii:
" «4
II 24131
-4I.1I2/
712*381
-600141
0 15400,
670JI/1
-88.4286

/8 92857
1429JB1
1/7/618
-129/11
-0.02824
-85

9306127
-19374
1218415
-61957*
801/207
-900627
6277G9/
-15

38 5KI12
4.769462
- 16 21/li
-3.18512
37.02/2C
-« 549C6
3*71019
-89.M28
"68435/1
230C/QI
-484!>i8
/U UVJ28
-91 7143
OU/I4VJ
004,11/9
12 3UIJU
6 5UUI6
37 7UC40
aii 41.611
• 2/ 1 152
-652J!i/

HEIK.UNC
-I29UI6
6/.SU-/. 1
10 l'.'6J4
-39.7915
-iiio
Ellll
lOOtibbJ
-46*6116
U 840/1
-95.114
1 68W4;
921. 1361
-IW

EHII
-345724
157 7084
-21 2/OJ
-4U2III
-101

Lllll
41VOU1I
-97.4621
3964591
42 OH3/
1 2811/6
- 76 0456
-100

Ellll
-301562
-2*2652
-63.1847
3723 9!t
0 6U'JU*2
- 71 O'ji>5
-ibii
	 thii
-146/81
94 U'JO/
62UU02
-100
HIM
' (• 2ti
Hi,: til/
iu 1, 1/'J
'•n fnai
-5U8UI4
-KM)

-------
                                      TABU: 5
         SOMKARY O
                    APPI/TCATTOK OF BIAS  CORRECTION TO S9K  FOL1.OVTNG EPA
                               OK A BATCH BASIS t±roff Table  
-------
                                 TABU e
               SWMARY or RcsuiTr TDK CASE  c SEKIVDUTILES
COMPDUKD
nitrobenzene
2.4-DRT
t-nitrophenol
pentachloro
o-cr»»ol
N-nitroae-propyl
K-nitro>o diBcthvlaBine
aniline
2-nitroaniline
pyridine
naphthalene
acenaptfaalcne
benzol c3prr«ne
hexa chi or o* thane
1.2.<-trichlorobenz«ne
hexa chi or obenzem
diMcthylpthalate
di-K-butylpthalate
di-f-octvlcthalBX*
1. UBCERTAim
-12.9
10.33
10.07
16.64
6.69
-S.«B
-21.27
41.9
22.97
1.17
-S.64
-6.B2
0.69
29.2
6.46
-14.6
NF.
2C.62
34.24
CHANCr IK ACCURACY
9.09
-7.16
11.24
7.3S
0.16
14.29
-12.97
-19.37
11.83

4^77
£.67
6.E
1.86

2.31
HR
t.59
sr- er* not  u«»d for
                                 TABLE?
                 SUMMARY 0? RESULTS TOR CASE C VOUTIIE5
coHPtnjHr i.
chlorose thane
chloroe thane
•ethylene chloride
1 . 1 -di chloroe thene
chioroiort
carbonic tra chl ori be
1 .2-dichleroethane
1 .1 ,2-tr; chl croe thane
1 .2-eJic.%lcroproptn«
1.1 .".2-tetrachlcroethene
i.: .l-trichJ.oroet.-uJi*
vinyl ci.loraet
cis-1 .2-cickloroethane
trans-1 .2-dichloroethene
xri chloroetnene
cis-1 .c-cichioroproptn«
1 . 1 -di chloroe tnent
trans-: ,S-£:chlcropropane
tetra chloroe then*
chlorobenzene
benzene
toluene
chlorobenzene
orthc-x\lene
ethylsensene
styrent
(c-t;;-.-ien*
brocoze thane
crosoci chl croc* thtne

-------
18                        Matrix Spiking: From Sampling to Analysis

          D. Syhre, Organic Laboratory Supervisor, M. Rudel, Analyst; V. Verma, Analytical
          Laboratory Supervisor; Browning-Ferris Industries Houston Laboratory, 5630 Guhn
          Road, Houston, TX, 77040.

          Abstract

          Currently there is little knowledge about how much change in analyte concentration oc-
          curs during the steps from sampling from the sample source (e.g. a monitor well) to
          analysis. A series of experiments were performed for metal, volatile, and semivolatile
          parameters with various stages of handling from the sampling of a simulated monitor
          well to analysis; percent recoveries were determined, and conclusions drawn.

          Introduction

          Current EPA methodology using the SW-8461 procedures requires a matrix spike rela-
          tively close to die point of analysis for volatile (VOA) organics, semivolatile (SV) or-
          ganics, and metals.   Specifically, volatile matrix spikes are performed immediately
          prior to analysis, semivolatile spikes during die extraction step, and the metals spiked
          prior to digestion. Matrix spikes are performed using analytes listed in die appropriate
          sections in SW-846; it is assumed that diese analytes were chosen by EPA because they
          felt these were representative of die analytes routinely analyzed.

          Currently die purpose of these matrix spikes is to provide QA/QC for die labs perform-
          ing die analysis by demonstrating diat the analyses being performed in die batch associ-
          ated with die matrix spike are "in control" The results from die matrix spike are com-
          pared to die  appropriate control charts, and if tiiey fall within an acceptable  range die
          analyses in that batch are deemed widiin "control."  In a similar vein, die organic anal-
          yses are tracked with surrogate spiking compounds added during die extraction step.
          The function of die surrogates is to track "hi control" performance on a sample by sam-
          ple basis.

          The advent of spike correction, however, changes die role of die matrix spike.  This
          type of spike correction requires die substitution  of die particular analytes of interest
          versus die standard spike list, and replaces die "in control" aspect of die matrix spike
          with an adjustment to die data from die associated batch  of samples based on percent
          recoveries of die analytes of interest in die spike.  Part of die stated reasons for moving
          to spike correction is to provide data more representative of what is actually in die sam-
          ple source (e.g. monitor wells, leachate, etc.).

          Regardless of die appropriateness of spike correction as a method, this raises an inter-
          esting question:  how much does die sample change from die time of sampling to analy-
                                               1-116

-------
sis?   To examine this
question  in detail  we
performed  a series of
experiments  designed
to determine change in
analyte concentration at
each conceivable stop-
ping point  from sam-
pling to analysis. This
was  done  by  spiking
different  sets of repli-
cates at the possible lo-
cations, letting the sam-
ple   then    proceed
through the process as
normal, and analyzing
the samples.  The loca-
tions at which  it was
determined to spike for
the organics (see Figure
1) were:    1.  Sample
Source; 2. Water Sam-
ple (post sampling, pre-
transportation);     3.
Water Sample  (post-
transportation); 4.  Or-
ganic Toxicity Charac-
teristic (OTC) Filtered
           Sampling)
           Transportation)
                  OTC
           Extraction
                        SV
                    (Extraction
                                   VOA
Figure 1:  Organic Sample Pathway
Water Sample; 5. Extract For the metals all the
samples originated from the same source, then un-
derwent different degrees of processing along the
metals pathway (see Figure 2) and differences in
concentrations determined.  The concentrations
from  the  samples  were compared  to  spiked
amounts  and percent recovery determined.  The
recovery data from various stages was compared
to  determine trends  in  analyte  concentration
through the sample pathway.

Methodology2
For the metal constituents a spike was performed
in a simulated monitor well which yielded the
               Metal     Concentration (ppm)
                 As             2.51
                 Ba             2.51
                 Cd             1.18
                 Cr             2.51
                 Pb             1.%
                 Se             2.51
                 Ag             1.18
                 Ca             5.10
                 Fe             2.51
                 Mg             5.10
                 Mn             2.51
                 Ni             1.%
                 Zn             1.96

                  Table 1: Metals Spike
                     Concentrations
                                      1-117

-------
        S ample
                      TM
1 Preserve}
            TransportatJon}
      ( Sample)
                          TC
                                       [Transportation I
                                   •W Sample
Filter &
Preserve
                                           Filtration}
                            C Sample)
                                      f Sample)
                                                               Spike &
                                                               Preserve
                                                           Sample)
                      ^ ^  Digestion   ^ ^
                     "^"^  & Analysis, ^^~
                       Figure 2: Metals Sample Pathway

concentrations indicated in Table 1. The monitor well was constructed using a five foot
section of Corning conical glass pipe with Teflon3 endcaps, each endcap fitted with a
Teflon stopcock.  The simulated well was stood vertically in its frame,  the bottom
sealed with an endcap and partially filled with deionized (d.i.) water. The spiking solu-
tions were added, the remainder filled, and the top capped.  The remaining small vol-
ume was filled via access from the top stopcock, and a Teflon tubing was used to con-
nect the bottom stopcock to a pump calibrated to deliver one well volume (-12.58 1) in
-5.6 hours, then connected to the top stopcock. The stopcocks were opened, and the
simulated well allowed to recirculate for -23 hours.  This allowed a total  of approxi-
mately four well volumes to be recirculated, and was deemed satisfactory for adequate
mixing of the sample.
                                     M18

-------
After stopping the recirculation process the stopcocks were closed, and the teflon tub-
ing removed.  One liter of sample was drained from the bottom and analyzed for the
complete list, then the top endcap removed and the simulated well sampled.  Eight one
liter samples were then distributed as follows:
1. One sample was preserved in the manner appropriate for total metals (TM) analysis,
then analyzed for Ba, Cd, Ca, Cr, Fe, Mn, Mg, Ni, and Zn.
2. One sample was preserved as above, then shipped overnight and analyzed for Ba,
Cd, Ca, Cr, Fe, Mn, Mg, Ni, and Zn.
3, One sample was  analyzed immediately after sampling for the complete list.
4. One Sample was field filtered and preserved in the manner appropriate for dissolved
metals analysis, then analyzed for Ba, Cd, Ca, Cr, Fe, Mn, Mg, Ni, and Zn.
5. One sample was filtered and preserved as above, then shipped overnight and ana-
lyzed for Ba, Cd, Ca, Cr, Fe, Mn, Mg, Ni, and Zn.
6". One sample was shipped and analyzed  for As, Ba, Cd, Cr, Pb, Se, and Ag (the TC
metals except for Hg).
7. One sample underwent TC (Toxicity Characteristic) filtration, then was preserved
and analyzed for As, Ba, Cd, Cr, Pb, Se, and Ag.
8. One sample underwent TC filtration, then was spiked with the appropriate metals,
preserved,  and analyzed for As, Ba, Cd, Cr, Pb, Se, and Ag.
All analyses were performed in replicates of five, and the results averaged. The averag-
es were compared versus known spike amounts, and percent recovery computed.
For the semivolatile organics an experiment was performed in which the simulated
monitor well was spiked as above at a level of 200 ppb with seven sets of compounds
deemed to be representative of the priority  pollutant analytes. These sets consisted of:
1. Acids: 4-Nitrophenol, pentachlorophenol, o-cresol.
2. Bases: Aniline, 2-nitroaniline, pyridine.
3. Chlorocarbons: Hexachloroethane,  1,2,4-trichlorobenzene, hexachlorobenzene.
4. Nitroaromatics:  Nitrobenzene, 2,4-dinitrotoluene.
5. Nitrosamines: N-Nitrosodi-n-propylamine, N-nitrosodimethy?amine.
                                     1-119

-------
6. Phthalates: Dimethyl phthalate, di-n-butyl phthalate, di-n-octyl phthalate.

7. Polynuclcar Aromatics: Naphthalene, acenaphthene, benzo[a]pyrcne.

The well was allowed to rccirculate ~23 hours, then five one liter samples pulled from
the bottom of the well via the stopcock.  The top endcap was removed, and five one
liter samples taken from the top in the normal sampling method. These samples were
then sent through the normal process (see Figure 1) and analyzed.

In addition four other sets of samples were spiked and sent through the remaining steps
of the pathway:

1. Five liters spiked at 200 ppb, then shipped overnight and proceeding through extrac-
tion to analysis.

2. Five liters spiked at 200 ppb, then proceeding through extraction to analysis.

3. Five liters spiked at 200 ppb then proceeding through the TC filtration process, ex-
traction, and to analysis.

4. Five vials spiked at 200 ppm (simulating the 1000:1 concentration factor) and ana-
lyzed.

During the extraction step a set of isotopically labeled compounds consisting of the fol-
lowing were spiked:
                                                                    1^
    o-cresol-dg             pyridine-d^             hexachlorobenzene-  Cg
    nitrobenzene-d^         di-n-butyl phthalate-d4   benzo[a]pyrene-d12
    N-nitrosodimethylamine-dg
These were considered as surrogates in substitution of the standard surrogates which
are listed in SW-846, and used to track the quality of the extraction and analytical stag-
es.  Note that since these compounds are representative of the compounds of interest,
they may also serve as a matrix spike, and are added at the location stated for such.

As for the metals, percent recoveries were determined for each set of replicates.

The volatile organics were treated essentially the same as the semi-volatile organics,
except the exclusion of the extraction step. The analytes of interest were:

1. Saturated Chlorocarbons:  Chloromethane, chloroethane, methylene chloride, 1,1-
dichloroethane, chloroform, carbon tetrachloride, 1,2-dichloroethane, 1,2-dichloropro-
pane, 1,1,2-trichloroethane, 1,1,2,2-tetrachloroethane, and 1,1,1-trichloroethane.
                                      1-120

-------
2. Unsaturated Chlorocarbons:  Vinyl chloride, 1,1-dichloroethene, cis-l,2-dichloroet-
hene, trans-l,2-dichloroethene,  trichloroethene,  cis-l,3-dichloropropene, trans-1,3-
dichloropropene, chlorobenzene, and tetrachlorothene.

3. Bromocarbons:  Bromomethane,  bromodichloromethane, dibromochloromethane,
and bromoform.

4. Aromatics: Benzene, toluene, o-xylene, ethylbenzene, styrcne, and m- & p-xylenes.

5. Gases: Chloromethane, vinyl chloride, bromomethane, chloroethane; these arc com-
pounds from the above list which may be separated from the above chemical classifica-
tions in addition due to their physical properties.

Isotopically labeled surrogate/matrix spike compounds for the volatiles were:


   chloroform13C         l.l-dichloroethene-o^   vinyl chloride-d^

   bromoform- C        benzene-dg

Following the normal analytical procedures percent recoveries were calculated as be-
fore for the metals and semivolatiles.

Results and Discussion

The metals data are summarized in Tables 2 and 3. The data in the tables represent av-
erages of the percent recoveries for all five replicates, except for the TC Spike, which
indicates the recovery of the TC analytes spiked at the appropriate location.   The
%RSD (percent relative standard deviation) indicates that very little, if any, loss occurs
in any of the processes;  the only exception being a consistently slight loss during the
TC filtration stage.
Bottom/Well
As
Ba
Cd
Cr
Pb
Se
Ag
100%
99%
99%
99%
92%
100%
97%
Top/Well
100%
99%
99%
99%
93%
102%
97%
Transport
101%
100%
101%
100%
93%
100%
96%
TC Filter
94%
98%
92%
94%
86%
91%
96%
TC Spike
94%
96%
99%
98%
98%
94%
94%
%RSD
3%
2%
3%
2%
5%
5%
2%
                         Table 2: Metal Recoveries for
                            TC Analytes (except Hg)
                                      1-121

-------


Bottom/Well
Ba
Cd
Ca
Cr
Fe
Mn
Mg
Ni
Zn
92%
86%
100%
93%
98%
94%
94%
92%
92%

Top/Well
92%
88%
99%
92%
97%
94%
94%
92%
93%
Filter/
Preserve(DM)
94%
87%
102%
94%
97%
96%
95%
92%
95%

PreservefTM)
90%
86%
100%
91%
93%
92%
91%
90%
91%
                                                   Trans(DM)  Trans(TM)  %RSD
                                                      93%       92%      2%
                                                      90%       88%      2%
                                                      102%       106%     2%
                                                      94%       93%      1%
                                                      101%       98%      2%
                                                      95%       94%      1%
                                                      94%       93%      1%
                                                      93%       92%      1%
                                                      92%       91%      2%
                        Table 3: Metal Recoveries for
                              Non-TC Analytes

The volatile organic results are summarized in Tables 4 and 5 (see end of paper). Table
4 displays the recovery data for the initial attempt at performing this experiment The
overall trend displays higher recoveries for the samples drawn from the bottom of the
well via the stopcock and the samples sampled in the normal  manner than those sam-
ples  which were simply spiked and carried through their respective processes. These
results obviously represent aberrant experimental procedure, but where is the question.

The  most obvious conclusion is poor spiking methodology. The spiking methods em-
ployed in the initial experiment used four spiking solutions, each at 2000 ug/ml. The
appropriate amount was then spiked from each individual solution into the simulated
well, followed by individual spikes of the appropriate containers of d.i. water.

To determine whether the methodology was at error a  second set (three replicates  in-
stead of five) of samples was analyzed with new spiking solutions. In this set the steps
in which  samples were just spiked and run and those that were spiked, filtered, and run
were repeated with spikes being performed as previously and also in a composite man-
ner using septum capped vials  (Table 5). In addition, a single  replicate was performed
in which  a sample was spiked in the previous manner using the original set of solutions
and analyzed. Obvious degradation of the integrity of the original set had occurred, al-
though the numbers are comparable to those obtained from the  data from the spiked and
run set versus the original experiment

Although some improvement can be seen in the quality of the data in going from the
previous  spiking method to the composite/septum vial method, the improvement is not
enough by itself to explain the variance in the original data. The most likely explana-
tion is that the spiking solutions degraded during the length of the original spiking pro-
cess, causing erratic results.  Examination of the new spike data is ongoing.
                                    1-122

-------
Of interest, however, are the exceptional recoveries obtained in the samples acquired
from both sampling via the bottom stopcock and those sample from the top via the nor-
mal method. This would suggest that very little is lost during the sampling process, and
in the steps later on. Obviously these conclusions must be taken with a grain of salt,
however, due to an incomplete set of comparable data. The available data also suggests
that spiking technique is critical when performing volatile organic spikes.

The semivolatile results are summarized in Figures 3 through 9 (see end of paper). The
observed trends:

1. Acids - No excessive decrease from sampling to analysis.  Filtration provided  die
greatest reduction  in analyte concentration.  Note must be taken, however, that  d.L
water is an ideal matrix, and our experience has been that the acids  suffer greatly from
matrix interferences in actual matrices.

2. Bases - Although the bases did not suffer greatly as a group in the process from sam-
pling to analysis, it is important to note that their recoveries were highly erratic. The
best actor overall was 2-nitroaniline, which is the least basic and least water-soluble of
the three. Pyridine's erratic behavior may be attributed to the factors of basicity, solu-
bility, and volatility (i.e. during the concentration stage).

3. Chlorocarbons - A trend is obvious which displays that with increased handling from
sampling to analysis there is a reduction in analyte concentration. Of particular interest
is the total failure to recover hexachlorobenzene from the sample which underwent TC
filtration. This result concurs with previous experiments performed in this laboratory,
and begs questioning the validity of TC hexachlorobenzene results.

4. Nitroaromatics - No significant trends exist  In general, these compounds are good
actors from sampling to analysis.

5. Nitrosamines -  Overall this classification can also be considered good actors, the
only exception being the erratic behavior of N-nitrosodimethylamine. The most likely
explanation of this erratic behavior is its volatility, which could provide problems dur-
ing the concentration stage.

6. Phthalates - The obvious trends are higher recoveries as molecular weight increases,
reflective of a solubility phenomenon, and loss in analyte concentration with increased
handling, especially at the filtration stage. Of particular interest is the failure to recover
dimethyl phthalate, consistent with poor recoveries for this analyte seen previously in
studies at this laboratory.

7. Polynuclear Aromatics -  The most striking trend is the large loss of analyte concen-
tration with increased handling time  and/or length of time containerized. This trend in-
creased with molecular weight, with extremely poor recoveries being demonstrated for
benzo[a]pyrene. This is especially so during the TC Filtration stage.
                                       1-123

-------
Conclusions

Experience has shown that the matrix plays a very significant factor in analyte recover-
ies, and all results and conclusions reported herein should be taken with note that these
experiments were performed in an ideal matrix, that being d.i. water.

1. Metals - Current metals methods from sampling to analysis give data which is repre-
sentative of the actual analyte  concentration within the sample source; therefore no
need is seen for modification of procedures in involving metals samples.

2. Semivolatiles - Over the classes defined, it was seen that some classes performed ad-
equately with respect to recovery, but that on the whole increased handling showed de-
creased recoveries. Of interest to note is the erratic behavior of some analytes, defying
efforts to  determine trends for these analytes. Erratic recoveries for samples which
were measured in a replicate basis demonstrate that if spike recovery corrected values
for these analytes were used, the values must be taken with a grain of salt

It is our recommendation that the current semivolatile surrogate list, which is not repre-
sentative of typical analytes, be  changed to the more representative list below:

                           Di-n-butyl Phthalate-d4

                           Hexachloroethane-13C
                                             10
                           Hexachlorobenzene-   Cg

                           Benzo[a]pyrene-dj2
                           Aniline-d^
                           N-Nitrosodimethylamine-dg
                           Phenol-dg
                                             10
                           Pentachlorophenol-  Cg
                           Nitrobenzene-d^

We feel that this list is representative of the typical analytes  in terms of reactivity, vola-
tility, and solubility.  In addition, if a list such as this were used in place of the current
surrogate list, the recoveries determined could be used in place of the traditional matrix
spike to monitor "in control" performance.

The benefits of changing the list of spike compounds  would be to provide a more repre-
sentative list of analytes to determine laboratory  performance on a sample by sample
basis. In addition, elimination of the superfluous matrix spike, required on a minimum
one spike per twenty sample batch basis, would allow ons additional analysis per batch.
                                     1-124

-------
3. Volatiles - Obviously the data is inconclusive; the most apparent conclusion to be
drawn is that spiking methodology and handling of the  spiking solution is  a critical
step.  It may be that with proper handling there is little  loss in analyte concentration
from sampling to analysis.

Acknowl edgements

The authors wish to thank the field crew, extraction, GC/MS, and wet lab personnel for
their willingness to devote intense energy towards this project. None of this data would
have arisen without their dedication.

References

1. Method 1310 from "Test Methods for Evaluating Solid Waste, Physical/Chemical
   Methods," November, 1986, Third Edition, USEPA, SW-846 and additions
   thereto.

2. Methods for sampling and analysis obtained from SW-846.

3. Teflon is a registered trademark of the E. I. Dupont Corporation.
      0.75

        0.5

       0.25
                                                      Figure 3: Acid
                                                       Recoveries
          o
   Spk & Run
      Spk & Ext
          Filtered
             Shipped
                    Top
                    Bottom
                          Bottom

          4-Nitrophenol        66%

          Pentachlorophenol     65%

          o-Cresol            76%

          o-Cresol-d8          79%
                      4-Nitrophenol

                      Pentachlorophenol

                      o-Cresol

                      o-Cresol-d8
Top   Shipped   Filtered  Spk & Ext  Spk & Run

52%    62%     46%      70%       74%

59%    74%     51%      83%       77%

70%    71%     65%      67%       74%

78%    75%     71%      71%       85%
                                       1-125

-------
                                                                   Figure 4:  Base
                                                                     Recoveries
                                                             | Aniline
                                                            j~j 2-Nitroaniline
                                                               Pyridine
                                                               Pyridine-d5
                  Bottom
    Aniline         59%
    2-Nitroaniline    87%
    Pyridine        54%
    Pyridine-d5      40%
    Top   Shipped   Filtered   Spk & Ext   Spk & Run
42%
76%
99%
48%
52%
70%
45%
36%
48%
66%
47%
49%
38%
60%
50%
47%
70%
74%
155%
83%
                                                                 Figure 5:
                                                               Chlorocarbon
                                                                Recoveries
                                                           Hexachloroethane
                                                           1,2,4-Trichlorobenzene
                                                           Hexachlorobenzene
                                                           Hexachlorobenzene-13C6
Hexachloroethane
1,2,4-Trichlorobenzene
Hexachlorobenzene
Hexachlorobenzene-13C6
Bottom
  24%
  17%
  2%
  88%
Top
19%
14%
 3%
90%
Shipped
  51%
  62%
  42%
  87%
Filtered
  38%
  41%
  0%
  84%
Spk & Ext
   46%
   61%
   79%
   89%
Spk & Run
    80%
    79%
   94%
                                      1-126

-------
  1
0.75
  0.5
  0.25
                                                               Figure 6:
                                                            Nitroaromatics
                                                              Recoveries
                       Spk & Run
                    Spk & Ext
                 Filtered
              Shipped
                                     Top
                                  Bottom
                         Nitrobenzene
                         2,4-Dinitro toluene
                         Nitrobenzene-dS
  Nitrobenzene
  2,4-Dinitro toluene
  Nitrobcnzene-d5
    Top  Shipped
    74%    77%
    76%    84%
    72%    69%
                               Spk & Run
                                  76%
                                  71%
                                  74%
                                                                   Figure 7:
                                                                 Nitrosamine
                                                                  Recoveries
                                                        N-Nitrosodi-n-propylamine
                                                        N-Nitrosodimethylamine
                                                        N-Nitrosodimethylamine-d6
N-Nitrosodi-n-propylamine
N-Nitrosodimethylamine
N-Nitrosodimethylamine-d6
Bottom
  79%
  88%
  71%
Top
76%
101%
72%
Shipped
  72%
  95%
  66%
Filtered
  70%
  67%
  69%
Spk & Ext
   76%
   74%
   69%
Spk & Run
    74%
   114%
    94%
                                       1-127

-------
                                                                     Figure 8:
                                                                     Phthalate
                                                                    Recoveries
Spk & Run
    Spk & Ext
          Filtered
             Shipped
                      Top
                      Bottom
                               Dimethyl Phthalale
                               Di-n-butyl Phlhalate
                               Di-n-octyl Phlhalate
                               Di-n-butyl Phthalale-d4
                          Bottom    Top   Shipped   Filtered   Spk & Ext   Spk & Run
    Dimethyl Phthalate
    Di-n-butyl Phthalale
    Di-n-octyl Phthalale
    Di-n-butyl Phthalale-d4
0%
9%
9%
52%
0%
12%
7%
60%
0%
12%
37%
43%
0%
7%
2%
55%
0%
30%
59%
47%
65%
65%
66%
90%
                                                                     Figure 9:
                                                                   Polynuclear
                                                                    Anomalies
                                                                    Recoveries
 Spk & Run
    Spk & Ext
          Filtered
             Shipped
                     Top
                     Bottom
                                | Naphthalene
                                3 Acenaphthene
                                | Benzo[a]pyrene
                                I Benzo[a]pyrene-dl2
                       Bottom
    Naphthalene           32%
    Acenaphthene         17%
    Benzo[a]pyrene         2%
    Ben2o[a]pyrene-dl2    86%
Top    Shipped
31%      43%
15%      58%
 2%      39%
86%      87%
Filtered
  38%
  37%
   0%
  80%
Spk & Ext
   46%
   59%
   75%
   87%
Spk & Run
   55%
   70%
   84%
   96%
                                          1-128

-------
Chloroinethane
Vinyl Chloride
Bromomethane
Chloroethane
1.1-Dichloroethene
Methylene Chloride
1,1-Dichloroethane
cis-1.2-Dichloroethen€
Chloroforom
trans-1,2-Dichloroethene
Carbon Tetrachloride
1.2-Dichloroedune
Benzene
Trichloroethene
1 ,2-Dichloropropane
Bromodichloromethane
cis-13-Dichloropropene
Toluene
trans-1,3-Dichloropropene
1,1,2-Trichloroethane
Dibramochloromelhane
Chtorobenzene
o-Xylene
Ethylbenzene
Styrene
Broinoform
1,1^2-Trichloroethane
Tetrachloroethene
1,1,1-Trichloroethane
m- & p-Xylene
Chlorofonn-13C
l,l-Dichloroethenc-d2
Benzene-d6
Vinyl Chloride-d3
Bromoform-13C
                Table
 Bottom     Top    Shipped    Filtered   Spiked & Run
  103%     103%     27%      54%         15%
  99%     113%     36%      65%         20%
  162%     164%     64%      100%         48%
  136%     144%     57%      85%         43%
  66%     74%      43%      55%         34%
  72%     69%      37%      57%         36%
  77%     76%      55%      68%         51%
  74%     74%      48%      57%         43%
  73%     74%      54%      65%         54%
  68%     71%      44%      53%         39%
  63%     73%      52%      61%         51%
  69%     72%      51%      60%         50%
  77%     79%      55%      60%         51%
  66%     72%      56%      71%         55%
  77%     80%      55%      65%         53%
  64%     64%      42%      58%         43%
  110%     100%     102%     114%         88%
  53%     56%      50%      52%         41%
  36%     46%      33%      44%         28%
  83%     117%     51%      83%         50%
  98%     100%     54%      64%         53%
  74%     70%      65%      62%         56%
  63%     60%      54%      54%         50%
  65%     61%      82%      87%         66%
  66%     62%      58%      58%         54%
  66%     66%      56%      65%         56%
  77%     76%      57%      67%         59%
  58%     62%      68%      75%         57%
  71%     78%      58%      69%         51%
  107%     98%      117%     121%         106%
  77%     81%      75%      77%         79%
  92%     95%      82%      83%         81%
  82%     83%      82%      84%         87%
  104%     107%     99%      109%         112%
  75%     84%      80%      72%         71%
4: Initial Volatile Organic Results
                                    1-129

-------
Chloromethane
Vinyl Chloride
Bromomethane
Chloroelhane
1.1-Dichloroethene
Methylene Chloride
1,1-Dichloroethane
cis-l^-Dichloroethene
ChlorofoTom
trans-1 ,2-Dichloroethene
Carbon Tetnchtoride
1.2-Dichloroethane
Benzene
Trichloroethene
1.2-Dichloropropane
Bromodichloromethane
cis-1,3-Dichloropropene
Toluene
trans-13-Dichloropropene
1,1.2-Trichloroethane
Dibromochloromethane
Chlorobenzene
o-Xylene
Elhylbenzene
Styrene
Bromoforrn
1,1,2,2-Trichloroethane
Tetrachloroethene
1,1,1-Trichloroethane
m- & p-Xylene
Chlorofonn-13C
1.1 -Dichloroethene-d2
Benzene-d6
Vinyl Chloride-d3
Bromofoim-13C
 Old Way
 Old Sol'n
   11%
   19%
   48%
   49%
   34%
   42%
   68%
   57%
   45%
   54%
   62%
   68%
   58%
   67%
   68%
   60%
   93%
   34%
   32%
   72%
   57%
   80%
   60%
   93%
   66%
   37%
   64%
   62%
   74%
   145%
   90%
   102%
   93%
   106%
   79%
TableS:
Spike & Run
  Old Way
 New Sol'n
   108%
   126%
   157%
   146%
    65%
    57%
    56%
    75%
    50%
    79%
    76%
    64%
    78%
    79%
    81%
    75%
    103%
    42%
    25%
    54%
    65%
    88%
    66%
    101%
    72%
    36%
    59%
    57%
    84%
    161%
    90%
    97%
    89%
    91%
    74%
New Way
New Sol'n
   90%
   106%
   141%
   131%
   60%
   54%
   74%
   73%
   56%
   71%
   81%
   72%
   87%
   84%
   88%
   75%
   118%
   49%
   30%
   65%
   62%
   99%
   78%
   105%
   86%
   44%
   74%
   56%
   97%
   175%
   87%
   87%
   83%
   85%
   75%
   Spike, Filter, & Run
Old Way     New Way
New Sol'n     New Sol'n
   120%        157%
   143%        138%
   182%        135%
   168%        136%
   73%         109%
   68%         86%
   90%         105%
   85%         112%
   60%         113%
   85%         118%
   88%         110%
   79%         123%
   89%         108%
   95%         113%
   98%         114%
   87%         91%
   105%        154%
   52%         98%
   31%         44%
   74%         101%
   63%         127%
   104%        100%
   78%         97%
   104%        105%
   85%         106%
   42%         90%
   69%         88%
   66%         121%
   93%         105%
   177%        195%
   88%         97%
   100%        100%
   84%         103%
   88%         103%
   70%         79%
                                  Spike Method Comparison
                                         1-130

-------
•|Q                         LAND DISPOSAL RESTRICTIONS PROGRAM
                            DATA QUALITY INDICATORS FOR BOAT
                             CALCULATIONS:  PAST AND FUTURE
        Lisa Jones. Waste Management Division QA Coordinator, U.S. Environmental
        Protection Agency, Office  of  Solid Waste,  Waste Treatment Branch,  401 M
        Street, SW,  Washington, DC   20460; Justine  Alchowiak,  QA Coordinator,
        Versar Inc., 6850 Versar Center, Springfield, VA  22151
        ABSTRACT

        The Land Disposal Restrictions (LDR) Program is continuing to develop and
        re-examine  treatment standards  for  each of the  listed hazardous waste
        codes using data of known  quality from well-designed and well-operated
        treatment systems.   The numerical treatment  standards  are  based on the
        level of treatment  achieved,  the accuracy of the data, and the inherent
        variability  of  the treatment processes,  sampling,  and analytical data.
        Analytical data are collected by EPA or submitted by industry for the LDR
        Program  on  an  ongoing basis.   Consistent with  the quality assurance/
        quality control  (QA/QC) program mandated by the office of  Solid Waste and
        Emergency Response, the quality and usability of the data  being evaluated
        are assessed based  on the following data quality indicators:  detection
        limits, bias, precision, representativeness, and  comparability.

        EPA  OSW's  newly issued  Quality Assurance/Quality  Control Methodology
        Background   Document  (QMBD)   explains  in  detail  the  LDR  Program's
        requirements for data  quality as well  as the history behind existing LDR
        standards.   This paper discusses  these data requirements  in  order to
        present the  "whole  picture" of QA/QC in the LDR Program.

        To verify that substantial treatment has occurred, when a treatment system
        is evaluated as a candidate for BOAT, EPA examines data characterizing the
        untreated waste  and the treatment  system residuals.   Since treatment
        systems  frequently  reduce the hazardous constituents  in  many wastes to
        levels  below detection  limits,  a detection limit  is  often used as the
        numerical measure of the level of treatment that occurred.  The detection
        limit is also a factor considered in determining the applicability of the
        analytical  method for  the  waste  matrices analyzed.

        The  LDR  Program's  QA  protocols  define  bias  as  percent  recovery of
        laboratory  matrix spikes.   The results  of the matrix spikes are used to
        bias  correct the data  as well  as to  determine the  extent  of matrix
        interferences on the usability of the  data.

        Similarly, precision is defined in terms of relative percent difference of
        the matrix  spike and  matrix spike  duplicate.   The results of relative
        percent  difference  are used to  determine  the  reproducibility  of the
                                           1-131

-------
analytical  procedure and  can  provide additional  insight  into  matrix
interferences.

Representativeness is addressed through selection of appropriate sampling
locations and procedures.  For the data to be applicable for calculating
•treatment standards, the samples must be determined to be representative
of the waste and the treatment system.

Comparability is addressed through use of the same sampling and analytical
procedures.  If data from more than one study are used, the results should
be comparable.  Therefore, the samples should be collected using the same
procedures (i.e., grab or composite)  and the  analytical procedures should
be either the same or comparable for  all the  results to be applicable for
calculating treatment standards.

If  the  results  of the  data  quality  indicators meet  the  program's
objectives, then the data are defensible and can be used to evaluate the
•treatment technology to determine whether  it   is  the  best demonstrated
available  technology  and   to  calculate  concentration-based  treatment
standards based on that technology.
INTRODUCTION

Under the Land Disposal Restrictions Program,  EPA's Office of Solid Waste
(OSW) developed  and promulgated regulations for the following scheduled
wastes:

          Solvents  and Dioxins - November 7, 1986
          California Rule Wastes - July 8, 1987
          First Third Wastes - August 8, 1988
          Second Third Wastes - June 8, 1989
          Third Third Wastes - May 8, 1990

Currently, EPA-OSW's Waste  Treatment  Branch is developing standards for
newly listed wastes (i.e., wastes listed since the 1984 HSWA Amendments).
EPA-OSW  will also  re-examine  promulgated  treatment  standards  as  new
information  on  treatment   technologies  or analytical  methods  become
available to the EPA.  As discussed  in recent Advance Notices of Proposed
Rulemaking, EPA is  actively soliciting treatability information for both
groups of  wastes.   Since treatment standards  developed  and promulgated
under the LDR Program  were, and will continue to be,  based on the best
data available  for treatment systems  that  were determined to be well-
designed and well-operated.   The Quality Assurance Methodology Background
Document is intended to present  EPA's  criteria for accepting treatability
data as the basis of subsequent numerical treatment standards.

Analytical data used to develop concentrated-based treatment  standards can
come  from many  sources both  inside  and outside  EPA.    These  may  be
collected by EPA for the LDR Program or  for other EPA  programs (e.g..
Office of Water's Effluent Guidelines  Program)  or may be submitted to EPA
                                  1-132

-------
by industry, trade associations, etc., to be considered for inclusion in
the LDR  Program's  data base.  The quality  and  usability  of the data is
evaluated based on the data quality indicators—detection limit, accuracy,
precision, representativeness, and comparability.


EVALUATION OF DATA QUALITY INDICATORS

The results of the data quality indicators  are  important to determine:

          Appropriateness of the  analytical method for the constituents
          analyzed.

          Bias of the analytical procedure, especially for bias-correction
          of the data.

          Determination  of  matrix  interferences  or other  analytical
          problems.

      •   Reproducibility of the analytical results.

          Comparability of data, especially  for comparability of data sets
          submitted  from more than one source.

      •   Representativeness of samples,  especially for comparison of data
          from more  than one treatment system.

The LDR Program has established quantitative or qualitative guidelines for
each of the data quality indicators.   The guidelines are used to evaluate
the acceptability  of the data for developing treatment  standards.   EPA
OSW's Quality Assurance/Quality  Control  Methodology Background Document
(QMBD) explains in detail the LDR Program's  requirements for data quality
as well as the history of the existing LDR  treatment  standards.

Detection Limits

To verify that substantial treatment has occurred, data characterizing the
untreated waste and  the  treatment system residuals are examined.  since
the hazardous  constituents  in many  wastes are treated to non-detect
levels,  especially for destruction  technologies, a  detection  limit  is
often used to  quantify the  level of treatment  that has occurred and to
develop the treatment standards.

The detection limit  may  also be  used to evaluate the appropriateness of
the analytical  procedures.   One of the  goals of the LDR  Program was to
obtain the best data  with the lowest detection limits.  A target detection
limit of at least  1  ppra  was established for data collected specifically
for the LDR Program.   However,  lower detection  limits were achieved for
most of the data  collected.   The reported  detection limit is especially
important in evaluating data reported as non-detect.  Data may be judged
to be  unacceptable  for  calculating  treatment  standards  if  it  could  be
                                  1-133

-------
determined that an  inappropriate  analytical  procedure was used based on
the reported detection limits.  For example, in the case of metals, if the
residuals  and TCLP  extracts  were analyzed  by flame  atomic absorption
methods instead of graphite furnace or ICP methods, the detection limits
may be considered to be too high for the purpose of developing treatment
standards and lower limits could have been achieved if the  best analytical
method was used.

Bias

Matrix  spikes  are  completed  to  evaluate  whether  the  laboratory  is
performing adequately or whether there is a methodological  problem such as
the presence of an interference or a systematic laboratory  error.  For the
LDR Program,  a matrix  spike and matrix  spike duplicate are required for
the parameters of interest for treatment tests completed specifically for
the LDR Program.   A minimum recovery value of  20  percent was determined to
be  acceptable for the LDR Program.   Data  with  recovery values  below
20 percent are  deemed to be unreliable,  since the  low recovery  values
indicated the presence of matrix interferences and  may indicate difficulty
in obtaining  reproducible analytical results.   Therefore, data with low
recoveries are not used to develop treatment standards.

All data used to  develop  the treatment  standards  for the  First, Second,
and Third  Thirds  were bias-corrected.    If a matrix  spike and a  matrix
spike  duplicate  were  completed,  the  lowest  recovery  value for  the
constituent to be regulated is used.  If a spike was not completed for the
constituent to be regulated, the lowest average recovery for the subset of
constituents that are representative of the constituent class  and analyzed
by the same method (e.g., volatile organics,  base-neutral organics, acid
extractable organics, etc.) may be used.  It should be noted that, for the
Second and Third  Thirds,  data were not  adjusted if the spike recoveries
were above 100 percent.

All data used to  develop  treatment standards for  newly listed wastes or
for revising promulgated standards will also  adjust  for recovery in order
to bring the reporting concentration of the target analyte closer to the
true concentration, thus improving the overall accuracy of the value that
serves as the basis for the promulgated standard.

Precision

Precision is defined in terms of relative percent difference of the matrix
spike  and  matrix  spike  duplicate.   The results  of relative  percent
difference  (RPD)   are used  to  determine the  reproducibility  of  the
analytical procedure.  No criteria were established for an  acceptable RPD,
however, the RPD is evaluated to gain additional insight  into any matrix
interference  and  to  evaluate  the reproducibility of  the  laboratory's
performance.  Engineering  judgment is used to  evaluate the  analytical data
with RPD's exceeding 20 percent for all of the constituents spiked.
                                  1-134

-------
Representativeness

Representativeness is addressed through selection of appropriate sampling
locations and procedures.  For the data to be applicable for calculating
treatment standards, the samples must be representative of the waste and
the treatment system.   For  a specific waste code,  EPA uses data for the
most difficult to treat waste to develop the treatment standards,  i.e.,
for most cases, that would affect waste with the highest concentrations of
the BOAT constituent present.  Therefore,  to evaluate the information for
this data quality indicator  it is important to evaluate both the untreated
waste and the treatment residuals that may be land disposed.

For the treatment residuals,  the data must also be representative of the
material to be regulated.   Since organic constituents can be destroyed,
the treatment  standards are  based  on total  composition  analysis.   For
metals, treatment technologies usually remove  or  immobilize the metal,
data are  evaluated  for both  total composition  and  the  TCLP extract for
both the  untreated waste  and the  treatment residual to  determine the
leachability of the metal.

Comparabilitv

Comparability is addressed through use of the same sampling and analytical
procedures.  If data from more than one study are used, the results must
be comparable. Therefore, documentation on how the samples were collected
(i.e.,  grab or  composite)   and on  the  analytical  procedures  used  is
reviewed.

Most of the treatment standards were  developed based on grab samples
analyzed using EPA  approved methods published in SW-846.   However, this
does not  preclude the  use  of composite data  or  samples  analyzed using
other solid analytical  procedures.
BOAT CALCULATIONS FOR TREATMENT  STANDARDS

The general approach for developing treatment standards was promulgated as
part of  the November 7, 1986  Solvents  and Dioxins Rule.   Based on this
approach, EPA's  treatment  standards are based on the performance of the
"best  demonstrated   available  technology"  (BDAT)  and  are stated  as
(1) concentrations   of   hazardous  constituents  in  nonwastewater  and
wastewater residues,  (2) specific treatment technologies for the waste, or
(3) a combination of a specific treatment technology for a type of residue
and constituent  concentrations.

All valid data available to the Agency will be considered  in  establishing
the treatment standards.  All data either collected by EPA or submitted by
industry, etc.  for  a specific waste code are available to the public in
the  Administrative  Record  either  during  proposal  or   promulgation
(depending upon the data of submission)  of the rulemaking for the specific
waste code.  Whatever the information source, however, the data underlying
                                   1-135

-------
the performance  standards  must meet standards of  quality assurance and
quality control.  Therefore, information for the data quality indicators
are  evaluated as  discussed earlier.   If  insufficient  information  is
available for some of the indicators, engineering judgment may be used to
determine the adequacy  of  the  data.  If the data  for  the indicators  is
totally  nonexistent  or judged  to  be  substandard,   the data  may  be
discarded.  All  data evaluations are presented  in  either the background
document for the specific waste and/or in the Administrative Record.

Information for the data quality indicators is collected during all of the
treatment tests  conducted by EPA-OSW specifically  in  support of  the LDR
Program.  The adequacy  of the  data  is  evaluated on a case-by-case basis
and only data meeting the criteria described previously is used to develop
concentration-based treatment standards.

The final step in setting a performance standard is to define the maximum
acceptable  constituent  levels  in  treatment  residuals based  on  the
performance  of  the technologies proven to  be both  demonstrated  and
available for the waste.

All concentration data  with spike recoveries  between  20 and 100  percent
are bias-corrected.   The average treatment  value observed in all of the
acceptable data is  then multiplied by the "variability factor."   The
variability factor takes into account the fluctuations in performance that
may  result  from  inherent  mechanical  limitations  in  treatment  control
systems,  treatability  variations caused by  changing  influent  levels,
variations in procedures for collecting treated samples,  or variations in
sample analysis.

Only one major change was made  between  November 1986 and June 1990 in the
methodology used to calculate  the treatment standards.   In the solvents
and dioxins rule, the treatment standards were based on TCLP results for
both  the organic   and  inorganic constituents.    For  the  Thirds,  the
standards and all subsequent regulations were based on total composition
data for  organic constituents  and TCLP extract data only for  the metal
constituents.

If  analytical data of  adequate quality or  an appropriate  analytical
procedure are not available, the Agency set a performance standard based
on a specific treatment method.
                                  1-136

-------
SUMMATION

Data  collected by  EPA for  the LDR Program  has sufficient information
available to evaluate  all  of the data  quality indicators.  No changes in
the basic methodology are expected to be made  for future regulations.  Any
changes  in  the methodology  would be proposed and published for comment
before the  change would be implemented.  As new  data become available or
new analytical procedures are developed, EPA may re-examine promulgated
treatment  standards  to  determine  if  a  revision  in  the  standard  is
necessary.

Therefore,  data   submitted  for  potential  inclusion  in  the  future
rulemakings should  have at a minimum information on the following:

          Analytical data  for untreated waste

          Analytical data  treatment residuals (total composition and TCLP
          extracts  for all inorganic constituents)

          Analytical methods used and  any modifications

          Detection limits

          Matrix spike recoveries

          Analytical   precision   (from  matrix  spike   and  matrix  spike
          recoveries or from duplicate analysis)

          Sampling  method  (i.e.,  grab or composite)
REFERENCES

U.S. EPA.   1986.   U.S.  Environmental Protection Agency, Office of Solid
Waste and Emergency  Response.   Test Methods for Evaluating Solid Waste;
Physical/Chemical  Methods.    SW-846 (3rd  Edition), Washington,  D.C.,
November 1986.

U.S. EPA.   1987.   U.S.  Environmental Protection Agency, Office of Solid
Waste.  OSW's Generic Quality Assurance  Project  Plan for Land Disposal
Restrictions  Program  (BOAT).    March 1987.   Washington,  D.C.:    U.S.
Environmental Protection Agency.

U.S. EPA.   1991.   U.S.  Environmental Protection Agency, Office of Solid
Waste.  Draft.  Best Demonstrated Available Technology  (BOAT) Background
Document for Quality Assurance/Quality Control  Procedures and Methodology
(QMBD1.
                                  1-137

-------
20             COMPARISON OF QUALITY ASSURANCE/QUALITY CONTROL
                       REQUIREMENTS FOR DIOXIN/FURAN METHODS
         Dennis Hooton, Senior Chemist, Midwest Research Institute, 425 Volker Boulevard,
         Kansas City, Missouri 64110
         ABSTRACT

         Fundamental  areas of  quality  assurance/quality  control are compared  for several
         dioxin/furan analysis methods  which use high-resolution gas chromatography/high
         resolution mass spectrometry (HRGC/HRMS).  These methods are used for analyzing
         various environmental matrices and for testing emissions from combustion sources.
         General areas of compatability and differences are discussed, while comparative items are
         categorized into tables for EPA Methods 8290, 23, and 1613,  and also California ARB
         Method 428.
         INTRODUCTION

         Numerous analytical methods exist for the determination of polychlorinated dibenzo-p-
         dioxins (PCDDs) and polychlorinated dibenzofurans (PCDFs). Many of these methods
         have  been promulgated or  drafted under federal and  state  agencies  for use  in
         environmental programs. These methods involve diverse matrices, and some are targeted
         toward specific regulatory applications, such as testing of hazardous waste incinerators
         under TSCA, CERCLA site remediations, and for development of national effluent
         limitation guidelines.

         Although the basic  techniques of isotope dilution, gas chromatography, and mass
         spectrometry used in identifying and quantifying these compounds are similar across these
         reference methods, regulatory applications are complicated by subtle variances in quality
         assurance/quality control (QA/QC) requirements imposed by the selection of a particular
         method.  The goal of this paper is to simplify the technical review and application of
         these  methods by comparing the  various QA/QC elements.  This  information is
         categorized for some of the more commonly used  methods into general  descriptions,
         procedural checks,  calibration controls, validation criteria,  and external verification

         MJU-AVZMTCMJAJ-AP                            1
                                             1-138

-------
recommendations.  These comparisons may be helpful in formulating test plans and in
data review.
DIOXIN METHODS

For this paper, the following four methods were examined:

•     EPA  Method  8290—"Analytical  Procedures  and  Quality  Assurance  for
      Multimedia  Analysis  of  PCDDs and  PCDFs  by  High  Resolution Gas
      Chromatography/High  Resolution  Mass  Spectrometry  (HRGC/HRMS),"
      EMSL-Las Vegas draft dated May 1987.  Although this method has never been
      promulgated, it has been widely used to analyze a wide variety of matrices
      including soil, sediment, water, biota, etc. It includes matrix-specific extraction
      and analyte-specific cleanup techniques.

•     EPA  Method 23—"Determination of PCDDs  and PCDFs  from Stationary
      Sources,"  Final Rule,  Federal Register, February  13, 1991.  This method has
      been promulgated to regulate emissions from municipal waste combustors and is
      also being used for trial burns on hazardous waste incinerators.  The method
      includes descriptions of stack sampling procedures and recovery of the sampling
      train components for the PCDD/PCDF analyses.

•     ARB  Method 428—"Determination of PCDD, PCDF, and PCB Emissions from
      Stationary Sources," California Air Resources Board, March 1988.  Similar to
      Method 23,  this state-promulgated method also involves a modified Method 5
      source sampling system for collection, recovery, and analysis of source emissions
      for PCDD/PCDF.

•     EPA  Method 1613—"Tetra- through Octa- Chlorinated Dioxins and Furans by
      Isotope  Dilution HRGC/HRMS," Revision A,  October 1990.  This method was
      developed by the Industrial  Technology Division with the U.S. Environmental
      Protection Agency's  Office of  Water Regulations and Standards to provide
      regulatory test data on waters, soils, sludges, and other matrices.
MW-A\Z54*M5A.PAP
                                     1-139

-------
As indicated in Tables 1 through 6, there are many similarities and also subtle differences
between the methods.  The following sections present some general discussions in these
areas.
GENERAL AND PROCEDURAL DIFFERENCES IN METHODS

One fundamental difference between these methods is in sampling. Two of the methods
(23 and 428) include protocols for sample collection and recovery from combustion
sources, while the other two procedures  focus  on  the analysis,  with  only  limited
discussion relating to field sampling activities.

Of the sampling procedures, Method 23 and ARE  428, although very similar, do differ
in solvent recovery protocols for the sampling train. Both add labeled congeners to the
XAD adsorbent of the sampling train,  prior to  sample collection, to monitor train
performance. Method 428 also requires correction of source emission results,  if field
surrogate recoveries are below 70%.

For the analytical methods, 1613 has more restrictive criteria for data acceptance than the
draft Method 8290, primarily in the increased number of labeled congeners added prior
to extraction, in specifications for initial demonstration of precision and recovery, and in
continuing calibration verification limits.

Control blanks are required for all methods in the forms of blank sampling trains,
sampling equipment rinsates, and laboratory method blanks.

Sample holding times are only discussed in Method 1613, although the QA sections of
SW-846 provide holding time guidance for aqueous matrices which would be applicable
to Method 8290.

General and procedural comparisons for the methods are presented in Tables 1 and 2.
CALIBRATION REQUIREMENTS

As shown in Table 3, only Method 8290 requires the use of a 7-point initial calibration
curve, while the other methods require 5-point curves.  Differences occur in acceptance
                                     1-140

-------
limits for both native and labeled congeners, although none of the methods discuss the
rationale for the selection of these limits. Requirements for ion abundance ratios are the
same for all methods.  Mass spectrometer tuning and column performance checks are
standard across the methods.

Continuing calibration requirements are significantly lower  for some congeners in
Method 1613, as low  as  ±6%  difference for some congeners versus 20% to 30%
difference used by the other methods.  This is an important control parameter in that it
triggers the need to stop analyses and recalibrate the instrument.
VALIDATION CRITERIA

As presented in Table 4, qualitative determination of dioxin/furan congeners are similar
in most respects except for differences in matching requirements of relative retention
times for samples to standards.  Confirmatory analyses are required by all methods for
positive identification of 2,3,7,8-substituted congeners.

Most of the methods discuss handling of non-detect values and the estimation of detection
limits as a reportable value.  This generally involves estimation of concentrations based
on background noise or reporting values as "maximum possible  concentration"  for
chemical hits which may match standard retention times but are outside of ion ratio
criteria.

Because these methods involve trace analysis, there are discussions on reagent screening
techniques, glassware tracking, glassware cleaning, and decontamination throughout the
methods.   However,  the  methods  do not  always indicate how to  handle blank
contamination  problems other than to discard contaminated sources and  restart  the
analytical process.  Method 8290 does discuss criteria for acceptable method blanks.

Duplicate analysis of field samples (and criteria for acceptance) is only discussed in
Method 8290.  However, source testing typically involves multiple runs and samples that
are  representative of 1 test condition which may be used  for comparison.   Surrogate
recoveries across like matrices, as discussed in Method 1613, also provides a means to
evaluate analytical precision.
MM-A\Z54*M}A.PAP
                                      1-141

-------
Acceptance criteria for surrogate recoveries of labeled congeners vary significantly by
method, ranging from 25% to 150% recoveries.  The concept and use of the isotope
dilution technique for quantitation allows these tolerances without invalidating the test
data.  Poor surrogate recoveries, similar to poor detection  limits,  may be due to real
matrix effects and/or interferences, and not reflective  of  analyst performance.  As
described  in  Method 1613, the technique of adding a known amount of a labeled
compound to every sample prior to extraction allows the native compound to be corrected
for recovery because the native compound and its labeled analog exhibit similar effects
through extraction, concentration, and gas chromatography.

Table 6 lists the calibration ranges, native and labeled congeners for each method.
EXTERNAL VERIFICATION CHECKS

Because these complex methods are routinely used in regulatory-based decisions, all the
methods  specify that  an ongoing quality assurance program be implemented by the
laboratory.  This encompasses many more areas of QA/QC than are presented in the
methods, such as document control, audits, control charting, etc.  However, the methods
suggest several ways that this may be accomplished, including:  using standards traceable
to EPA reference materials or other certified standards, using EPA audit samples (spiked
XAD) for source testing, and using field-associated QC samples (field replicates, rinsates,
etc.) to demonstrate the overall quality of sampling and analysis test  data.
SUMMARY

In researching this paper, several observations were apparent:

•      Successful use of these methods require  a  comprehensive  QC system 'that
       incorporates procedural elements which are common to all the methods, including
       laboratory management, sample handling, instrument operating parameters, and
       elimination of potential sources of contamination. In addition, a QA program is
       needed to ensure that test data  for environmental programs are both internally
       consistent and comparable to other EPA data.
MW-A\Z5«W)A.PAP
                                     1-142

-------
•      Control limits  and acceptance criteria vary by method for initial/continuing
       calibrations, surrogate recoveries, and control samples.

•      Validation of test data relies on proper calibration, acceptable method blanks,
       surrogate recoveries,  qualitative identification criteria, certification of standard
       reference materials, and confirmatory analysis of suspect samples. Documentation
       of these control  parameters is essential for completing technical reviews and
       audits.

Notwithstanding the inherent method similarities and differences discussed above,  the
authors' discussions of particular technical areas vary in degree of clarity and emphasis,
so that there is complimentary  information  shared by all methods.  This diversity of
information gives the reader a better understanding  of dioxin analysis techniques,
regardless of the actual method  chosen for an environmental program.
                                        1-143

-------
                                                     TABLE 1. PCDIVPCDr METHOD COMPARISONS




METHOD 8290




METHOD 23











ARB428












METHOD 1613










Application
PCDD/PCDF in soil,
sediment, water, fly
ash, sludge, fad oil,
paper, biological
samples
PCDD/PCDF from
stationary sources










PCDD/PCDF & PCB
emissions from
stationary sources
(regulatory compliance)









PCDD/PCDF in water,
soil, and other solid
matrices








Status
EPA Draft,
May 1987



EPA final
rule,
Federal
Register,
February
1991






Adopted
March 23,
1988 by
CARB









USEPA,
Revision A,
October
1990






Sampling
protocol
No




Yes, Modified
Methods
(fitter/XAD)









Yes, modified
California
version of
Methods
(fitter/XAD)
[minimum 3 tons
® 3 h each]






No









Field spiked
surrogates
NA




Yes











Yes












NA







Preclean&
screen
reagents &
glassware
Yea




Yes











Yes












Yes, protocol
specifies








Sample train
recovery
NA




1. Acetone
2. Methylene
chloride
3. Toluene
(separate
analysis as "QA
rinse")





1. Methanol
2. Benzene
3. Methylene
chloride









NA










Analysis
• Spike samples with internal standards
• Extract aqueous with methylene chloride
• DB-5 column recommended
• Confirmation column required for
positive identification of all isomers
• Spike XAD with internal standards
• Reduce aqueous component and
combine with XAD
• Soxhlet extract with toluene
• Split sample hi half for archiving
• Column deanupi
-silica gel
—alumina
—carbon/Celite
• no add cleanup discussed
» HRGC/HRMS with DB-5 column
• Confirmation using SP-2330 on SP-2331
• Spike XAD with internal standards
(none is added to aqueous)
• Extract aqueous with methylene chloride
then add to & extract with fitter/XAD
• Extract with benzene or toluene
• Acid cleanup
• Column cleanup
-silica gd
—alumina
— carbon/Celite
• HRGC/HRMS with DB-5
• Confirmatory analysis using SP-2330 on
SP-2331
• Spike samples with 15 labeled analogs
• Labeled TCDD added after extraction to
measure efficiency of cleanup steps
• Internal standards added prior to
analysis for quantitation
• Sample cleanup by GPC, HPLC, or
column chromatography described
(column calibration required)
MRI-A\Z5470-QA.PAP

-------
                                           TABLE 2.  PROCEDURAL QA/QC



METHOD 8290



METHOD 23



ARE 428



METHOD 1613





Lab
method
blank
Yes, 1
per batch
of 24

Yes



Ongoing



1 blank
control
matrix
per batch
of 20
samples


Matrix spike/MSD
Yes, duplicate spike per
batch of 24


All XAD traps are
spiked with labeled
congeners prior to
sampling
All XAD traps are
spiked with labeled
congeners prior to
sampling
Initial demonstration of
recovery and precision
from 4 matrix spikes is
required. Also,
1 control sample per
batch is analyzed.

Holding
times
Not
discussed;
refer to
SW-846
Not
discussed


Not
discussed


1 yr @ 4°C
for samples;
40 d © 4°C
for extracts



Homogenization
of sample
Yes (% moisture
of soils)


Complete
analysis (except
for toluene "QA"
rinse)
Complete
analysis


Yes, % moisture
of soils






Field blank
Rinsate of sampling
equipment; 1
"uncontamraated" field
blank per 24
Blank train



1 blank train per three
runs


Recommended.







QC check sample
One performance
evaluation sample in
all batches

EPA audit sample
required for
regulatory basis

EPA or other
independent audit
sample

Ongoing precision
and accuracy control
samples



CO
    MRI-A\Z5470-QA.PAP

-------
                                                    TABLE 3.  CALIBRATION QA/QC

METHOD 8290
METHOD 23
ARB428
METHOD 1613
Number of
standards in
calibration curve
7-point
5-point (option for
two levels)
5-point
(replacement of
calibration
standards every
6 months)
5-point standard
solutions are
analyzed within
48 h of prepara-
tion and on a
monthly basis to
check for
degradation
Initial
calibration
precision
20% RSD
25% USD
(30% RSD for
native OCDF
and some
labeled
congeners)
15% RSD
< 20% coeffi-
cient of
variation for
calibration by
isotope
dilution;
< 35% coef-
ficient of
variation for
calculation by
internal
standard
Ion
abundance
ratios
±15% of
theory
±15% of
theory
±15% of
theory
±15% of
theory
Continuing
calibration (every
12 hr)
20% D (bracketing
RRFs used for
quantitationif RRF
is between 20-25%
D)
25% D (30% D for
OCDF and some
labeled congeners)
30% D
Varies from 6% to
20% D allowed
M5 timing
and accuracy
check
PFK (every
12 h)
PFK
PFK (every
12 h)
PFK (every
12 h)
GC column performance check
Initially and every 12 h:
• First and last elnters labeled on
chromatograms
• Presence (current switching) of both
1,2,8,9-TCDD and 1,3,4,6,8-PeCDF
Initially (daily):
• 25% valley for 2,3,7,8-TCDD
• Established RT windows for homolog series
Initial (daily):
25% valley for 2,3,7,8-TCDD
Meets ion abundance ratio criteria
Minimum S/Nof5:l
Mass correction for m/e 328 for native TCDD
Retention windows for homolog series
Initially and every 12 h:
• Mass drift correction using PFK
• Verify ion abundance ratios, minimum levels,
and signal-to-noise ratios
• Column window-defining mix
MRI-A\Z547
-------
                                                   TABLE 4. QC DATA AND ACCEPTANCE CRITERIA.



METHOD
8290







METHOD
23







ARE 428








METHOD
1613










Identification
Ions maximize within 2 s
S/N l 2.5 for both ions
RT match of -1 to +3 s of standard
Ion abundance ratio within ±15% limit
RT within homolog window
Confirmatory analysis on second column
(quantitalion is specified by column &
congener)
• No PCDPE interference
• Simultaneously (±2 s) detection of ions
• S/N * 2.5 for both ions
• RT match of ±3 s of standard
• Ion abundance ratio within ±15% limit
• RT within homolog window
• Confirmatory analysis on second column
(quantitation is specified by column &
congener)
• No PCDPE interference
• Simultaneous (±2 s) detection of ions
• S/N £ 2.5 for standards > 10 for samples
(both ions)
• Ion abundance ratio within ±15% limit
• RT within homolog window
• Confirmatory analysis on second column
(quantitation is specified by column &
congener)
• No PCDPE interference
• Ions maximize within ±2 s of one another
* S/N i 2.5 for sample extract and t 10 for a
calibration standard
• Ion abundance ratios within ±15% limits
• Relative retention time windows within column
performance mix
• Confirmatory analysis for 2,3,7, 8-congeners
using second column
• No PCDPE interference



Method blanks
• Method blank per
batch (daily, before
samples)
— internal standards
> 10:1 S/W
— background below
< 10% of target
detection limit

• No blank criteria given








• No blank criteria given








• AH materials must be
demonstrated to be
interferant-free
• Glassware tracking
recommended






Field
surrogates
NA








70-130% R
(correct test
results if
R < 70%)





60-140% R








NA









Internal
standard
recovery
40-120% R








40-130% R
(tetra-hexa)
25-130% R
(hepta-octa)
(data are still
acceptable if
S/N * 10 for
detected
PCDD/PCDF)
40-120% R








25-150% R









Duplicate
sample
analysis
<25%D
for 2,3,7,8-
congeners






Not
discussed







Field
replicates
(no criteria
given)





Not
discussed










MS/MSD
*20%D
for
23,7,8-
congeners





Not
discussed







Not
discussed







Accep-
tance
criteria
for initial
and
continuing
perfor-
mance
tests are
listed


Detection limits
Concentration
corresponding to 7.5x
the background noise






Report Theoretical
Minimum Quantifiable
Level (TMQL)" ast
TMQL (PG) = lowest
STD (pg/uL) x final
extract
volume/recovery of
internal standards

• Concentration
corresponding to
2.5x noise level or
ions outside ion
abundance criteria
as "estimated
maximum passible
concentration"

Minimal levels are
listed for water, solid,
and extract







MRI-A\Z5470-QA.PAP

-------
                          TABLES. EXTERNAL VERIFICATION
                          Reporting requirements
                                           Method recommendations
  METHOD
  8290
Results of at least one GC column
performance check first ("F") and last
("L") eluters labeled on chromatogram
Results of 2 mass resolution checks and
continuing calibration checks during a
12-h period
Toxic equivalency factors
(if needed)
•  Calibration standards traceable to
   EPA (EMSL-LV) reference
   materials
   —verification data and supporting
   records on file
   —1 to +3 s match in RT times
   —s£ 20% difference in
   concentration between standards
   and EPA reference
   —standard preparation and
   traceability records on file
  METHOD 23
Internal standard percent recoveries
Field surrogate recoveries
Analysis results of toluene QA rinse
TMQLs (detection limits)
Results for 2,3,7,8-congeners and totals
•  EPA audit sample (for regulatory
   tests)
  ARE 428
Detection limits = 2.5 x noise
"Maximum Possible Concentration" for
bits that do not meet ion abundance
criteria
Deviations from method
Totals and specific 2,3,7,8-substituted
congeners
Sample numbers, source, and chain-of-
custody records
Dates of submittal and GC/MS analysis
Raw data (mass intensities)
Ion ratios for PCDD/PCDF detected
% recoveries of internal standards
Recovery of spiked samples
Summary of calibration data
—mean RRFs
—RSDs for 5-point curve
—acceptable continuing calibration
checks for each 12 h
Traceability records  and sequence of
analysis
   Verification of 2,3,7,8-TCDD
   standards to EPA reference
   materials
   Formal QA program
   —ongoing analysis of spiked
   samples
   —records of past performance
   —ongoing screens and method
   blanks
   —field replicates for overall S&A
   precision
  METHOD
  1613
Report values to 3 significant figures
Multiple forms provided for reporting
QC and test data
Control charting required
Standards must be certified for
purity, concentration, and
authenticity
MJU-A\Z547(K!A.PAP
                                             11
                                           1-148

-------
                     TABLE 6. ANALYTE LIST AND CALIBRATION RANGES
Desig-
nation

Compound
RRF
reference
standard
Calibration range:
Method 8290
2.5-1,000
Pg/^L
Method 23
0.5-1,000
Pg/l»L
(low)
5-10,000
pg/^L
(high)
Method 428
5-10,000
pVl*L
Method 1613
05-2,000
pg/|iL
UNLABELED ANALYTES (17)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

2,3,7,8-TCDD
2,3,7,8-TCDF
1,2,3,7,8-PeCDD
1,2,3,7,8-PeCDF
2,3,4,7,8-PeCDF
1,2,3,4,7,8-HxCDD
1,2,3,6,7,8-HxCDD
1,2,3,7,8,9-HxCDD
1,2,3,4,7,8-HxCDF
1,2,3,6,7,8-HxCDF
1,2,3,7,8,9-HxCDF
2,3,4,6,7,8-HxCDF
1,2,3,4,6,7,8-HpCDD
1,2,3,4,6,7,8-HpCDF
1,23,4,7,8,9-HpCDF
OCDD
OCDF
A
B
C
D
D
E
E
E
F
F
F
F
G
H
H
I
I
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
(continued)
MRI-A\Z54TD-QA.PAP
                                            12
                                          1-149

-------
                                        TABLE6
Desig-
nation

A
B
C
D
E
F
G
H
I


AA
BB

18
19
20
21
22
23
Compound
RRF
reference
standard
Method 8290
Method 23
Method 428
Method 1613
INTERNAL STANDARDS (10)— (SPIKED ONTO SAMPLE PRIOR TO EXTRACTION)
uCir2,3,7,8-TCDD
BC,2-2,3,7,8-TCDF
BC12-l,2,3,7,8-PeCDD
BC12-l,2,3,7,8-PeCDF
BCir1^3,6,7,8-HxCDD
uC12-l,2,3,4,7,8-HxCDD
BCirl,2,3,4,6,7,8-
HpCDD
BC,2-1,2,3,4,6,7,8-
HpCDF
BC,2-OCDD
uC12-23,4,6,7,8-HxCDF
AA
AA
AA
AA
BB
BB
BB
BB
BB
BB
X
X
X
X
X
X
X
X
X
-
X
X
X
X
X
(U,3,6,7,8)»
X
X
X
-
X
X
X
X
X
(1,2,3,6,7,8)
X
(1,2,3,4,7,3,9)
X
.
X
X
X
X
X
X
X
X
X
X
RECOVERY STANDARDS (2)— (SPIKED INTO EXTRACT PRIOR TO ANAYSIS)
"CB-U.SATCDD
BCirl,2,3,7,8,9-HxCDD
-
-
X
X
X
X
X
(1,2,3,4,7,8)
X
X
SURROGATE STANDARDS (6)— (SPIKED ONTO SAMPLE PRIOR TO SAMPLING
37Cl-2,3,7,8-TCDD
BCI2-23,4,7,8-PeCDF
"Cn-M^^S-HxCDD
13C12-l^^,4,7,8-HxCDF
"0,2-1,2,3,4,7,8,9-
HpCDF
uCn-1^3,7,8,9-HxCDF
A
D
E
F
H
F
-
-
-
-
-
-
X
X
X
X
X
(alternate)
X
X
-
X
(1,2,3,4,6,7,8)
X
X"
Xs
-
x«
(1,2,3,6,7,8)
x°
X*
• Brackets indicate alternate or other congener specification by method.
b Spiked into sample after initial extraction prior to clean-up to check column
performance.
c Method 1613 uses these labeled congeners as internal standards, i.e., spiked
into sample prior to extraction.
MM-A\ZS4HW}AJ>AP
                                                13
                                               1-150

-------
                 A STUDY OF METHOD DETECTION LIMITS  IN
                     ELEMENTAL SOLID WASTE ANALYSIS
DAVID E. KIMBROUGH. PUBLIC  HEALTH CHEMIST,  JANICE WAKAKUWA, SUPERVISING
CHEMIST, CALIFORNIA  DEPARTMENT OF HEALTH SERVICES,  SOUTHERN CALIFORNIA
LABORATORY, 1449 W. TEMPLE STREET, LOS ANGELES CALIFORNIA 90026-5698.

ABSTRACT

Two types  of  data are generated  when  determining the concentrations  of
regulated  elements  in solid matrices  (e.g.,  sediments,  sludges,  soils,
spent catalysts, press cakes,  slags, powders, etc.). These are numerical
values, that indicate the amount  of analyte present, and "None Detected"
or "Less Than" values. The latter values define,  for  a given analyte, the
smallest amount  that  can be quantified.  A positive numerical value  is
usually defined within a set of precision and accuracy criteria.  A "less
than" value on an analytical report is as much a data point  as a numerical
value and  should be determined with equivalent precision and accuracy.

By far the most common method for  determining these  "less than" values is
the U.S.  E.P.A.'s Method Detection  Limit  (MDL)1>2 based on the  work  of
Glazer  et  al.3  This method  was  developed for trace analysis of organic
analytes in water matrices.   MDLs  have, however, been used extensively for
inorganic  analyses and  solid matrix analyses without examination of its
applicability.

This study attempts to assess the  applicability of the MDL method to solid
matrix analysis.  The study  compares the calculated MDLs for five analytes
in soil, arsenic, cadmium, molybdenum,  selenium,  and thallium with method
performance at concentrations above and below^the calculated MDL.  The MDL
method  is  examined both  for its  empirical  suitability  for solid waste
analysis  and  whether it  has  the proper  theoretical  tools  for  solid
matrices.

INTRODUCTION

When determining the concentration of regulated elements in solid matrices
(e.g.,  sediments,  sludges,   soils,  spent catalysts,  press  cakes, slags,
powders,   etc.)  two  types  of  data  are  generated,   numerical  values
indicating the  amount of  an analyte present and  "None Detected" or  "Less
Than" values  that let the data user estimate the smallest  amount of the
analyte that can reasonably  be quantified.  A positive numerical value is
usually defined within a set precision and accuracy.  A "less than" value
on an analytical  report  is  as  much a data point  as  a numerical value and
should  be  determined with equal precision and accuracy.  The most common
method  for determining this  "less than" value is the U.S. E.P.A.'s Method
Detection  Limit (MDL) based  on the work of Glazer et al.  This  method was
developed  for  trace analysis of organic analytes in water matrices.  MDLs
have,  however,  been used  extensively  for inorganic  and solid matrix
analyses without an examination of their applicability to the procedures.
                                   1-151

-------
This study attempts to assess the applicability of the MDL method to solid
matrix analysis.  A comparison of  the calculated MDLs for five analytes
in soil (arsenic,  cadmium, molybdenum, selenium, and thallium) with method
performance at concentrations above and below the calculated MDL.  The MDL
method is  examined both for its  empirical suitability  for solid waste
analysis and its appropriateness as a theoretical tool for estimating MDLs
in solid matrices.

METHODS

A)  The USEPA's MDL and Glazer et al. are both six step procedures.  These
procedures are designed to be used on any matrix.  The five step portion
of these  procedures that are applicable  to solid  wastes  are presented
below.

  1)   Estimate  the MDL by one of four procedures:

    a)  The  concentration that corresponds to an  instrument  signal to
    noise  ratio of 2.5 to 5.

    b)  The  concentration  value  that  corresponds  to three  times  the
    standard deviation of  replicate instrumental  measurements  for  the
    analyte  in  reagent water.

    c)  The  concentration value that corresponds to the region where there
    is significant change in sensitivity at low analyte  concentrations.

    d)  The  concentration value that corresponds  to the known instrument
    limitations.

    (This  will be referred to as the estimated MDL)

  2)   Obtain a  solid  material corresponding to  the matrix type for which
  the MDL  is to be determined.  The material must have a concentration of
  the analyte(s)  of interest at one  to five times (but not to exceed ten
  times) the estimated MDL.

  3)   Take a minimum of seven (7) aliquots of the  material and process
  each through the entire analytical procedure. Calculate the results back
  to solid phase  with  the appropriate units such  as  mg/kg or ug/g.

  4)   Calculate the standard deviation (S)  using  equation 1, as follows:
                                   x\ - £ xj
                           s
                                  Eq.  1

  5)   Calculate  the MDL by equation 2  as  follows:
                                  1-152

-------
                            MDL = t * 5

                                Eq.  2

    where  t is the  student's value approximate for  a  99% confidence
    level  for  n  -  1.   Since n is equal to seven the student's t value
    would  be 3.143.   (This will be  referred to as the calculated MDL).

6)  There  is  an optional  step which  calls  for the  preparation  of  a
material spiked exactly at the calculated MDL and repeating steps 3-5.
This will be referred to as the iterative procedure.   If the calculated
MDL results in a spiked concentration that does not allow qualitative
identification  then   repeat   steps  3-5  with  a  higher  spiked
concentration.     If   the  spiked  concentration   is   qualitatively
identifiable but the standard deviation of the seven replicates of the
spike  (Sb)is  3.05  times  greater than  the  standard  deviation  of the
calculated MDL determination  (Sa) then an other  spiked material must be
prepared.   If the  ratio of  Sa/Sb is less than 3.05 then the  MDL is
recalculated by pooling the data  following equations  3 and 4:
                                Eq.  3
                       MDL =2.681 (

                                Eq. 4

   (This will be referred to as  the pooled MDL) .

B)  Percent Inaccuracy (%I) is  defined here  as the absolute difference
between the spiked value (Xs) of a  soil  and the mean measured value  (X,,,)
divided by the spiked  value times  100.


                               Y _ y
                         %j =   e _  m  * 100


                                Eq. 5

C)  Percent  Relative  Standard  Deviation  (%RSD)  is  defined as  the
standard deviation as defined above divided by the mean value times 100.
                                 1-153

-------
                           %RSD = ^- * 100
                                  Eq.  6
EXPERIMENTAL SECTION
A)  Study Design.  To assess the applicability of the MDL method to solid
waste elemental  analysis  it will be  necessary  to determine the  MDL by
means of the five steps outlined above for a solid matrix, in this case a
loamy soil.

Three instruments were used:  a Jobin-Yvon JY-50P simultaneous Inductively
Coupled Plasma Atomic Emission Spectrometer (ICP) a Perkin-Elmer PE-5500
sequential ICP, and a Thermo Jarrel- Ash Video  12E Flame Atomic Absorption
Spectrometer (FAA) . The estimated HDL was determined by each of the four
methods for each instrument in step  one for arsenic, cadmium, molybdenum,
selenium, and  thallium  for  the  two  ICP-AESs.  The same was performed for
cadmium, molybdenum, and thallium on the FAA.

Five soils  were  spiked at several  concentrations  for  each  element (see
below) corresponding to the  estimated MDLs. Seven aliquots  of each soil
were digested and the calculated MDLs were determined for each element for
each  instrument.  When  necessary,  the  pooled MDL was  also determined.
Finally, the estimated, calculated,  and pooled  MDL values was evaluated
for both precision and  accuracy for solid waste analysis.

B)  Analytical Procedures.  The soils were digested using an aqua regia
method here designated as the SCL method4'5-6.  2.00 grams of the soil were
placed in a 150 mL Phillips beaker.   10 mL of  concentrated HC1 and 2.5 mL
of concentrated  HN03 was added  and  heated to 95° C on  a  heating block.
When there was no more  reddish-brown gas (N02) ,  the beakers  were removed
from the heating block and the digestate was  filtered through a Whatman 41
filter paper and collected in a 100  mL volumetric  flask.  The residue and
filter paper were then washed with 5 mL of hot  (95° C) concentrated HC1 and
then 20 mL of hot de- ionized water which is also  collected  into the 100 mL
flask.   The filter  paper  and  residue were  then placed  back  into  the
Phillips beaker and heated with 5 mL of concentrated HC1 until the filter
paper dissolves and then this second digestate is filtered and collected
into a second 100 mL volumetric  flask.   Both filtrates were then analyzed
and the results were mathematically combined.

Instrumental analysis methods used to analyze these materials was EPA SW
8466 methods 6010 for both the sequential and simultaneous  ICP using a 100
ug/mL standard for all elements .

Both ICPs used two background correction points for each analytical line.
                                  1-154

-------
Methods 7130, 7480, and 7840 were used for the Flame AA and a deuterium
lamp was used for all analyses for background correction.

C)  Materials.  A local soil was dried, sieved and then analyzed for of
arsenic, cadmium, molybdenum, selenium, and thallium.  The amounts found
were far below 1 ug/g.  These elements were then spiked into the soils to
produce the following concentrations:

               Sample A    Sample B     Sample C    Sample D    Sample E

Arsenic           4,000         500          50           5

Cadmium             500           50           5           -       5,000

Molybdenum           30            5           -       5,000         500

Selenium              5            -       5,000         500          50

Thallium              -        5,000         500          50           5

Here and throughout this paper the dash mark,  -, is  used to indicate that
the analyte in question  was not spiked.  Soluble salts of each element were
dissolved in water and spiked into the soils.  Additional water was added
to form a slurry  which  allowed for easy homogenization.  The soils were
then dried,  milled and  sieved,  (for a more complete  discussion  of the
preparation  of   spiked  soils,   see   "Preparation  and  Validation  of
Proficiency Evaluation  Samples"  by Kimbrough,  D.E.  and J.R.  Wakakuwa in
these proceedings.) One additional material was  prepared, designated "K",
which was made up of 100 g of the same  soil and was spiked with 20 mL of
a 100 ug/mL Cd and Tl standard.  This  material was treated exactly as the
other materials were.   Its concentration was 20 ug/g.

RESULTS

A)  Arsenic.   Table  I  lists all of the  data  for arsenic.    Using the
estimation procedures, four estimated MDLs were determined for both types
of ICP.   A wide  range  of estimates were obtained,  so three different
spiked materials with concentrations  of 500, 50, and 5  ug/g were used to
determine the calculated MDL as  well  as an unspiked background material.

For the simultaneous ICP,  using  the  500 ug/g material  the calculated MDL
is 30 ug/g.   Following the  iterative procedure then a 30 ug/g material
should be analyzed.  For this step the  50  ug/g  PE sample was used which
produced a calculated MDL of 41  and  the ratio  of  S500 to S50 is far less
than 3.05.  The pooled MDL is 29.

For the sequential ICP the 500 ug/g material yields a calculated MDL of 17
ug/g.   The  5 ug/g material  could  be used for the iterative procedure
except  the  results were not qualitatively  identifiable.   The  50  ug/g
material is qualitatively identifiable and yields a MDL of 8.4.  The ratio
                                   1-155

-------
of S500 to S50 is less than 3.05 and the pooled MDL is 10.3 ug/g.

B) Cadmium.   Three materials were used as with  arsenic,  500,  50, and 5
ug/g  as  well as  an unspiked background  material.   Table  II  lists the
estimated MDLs.

Using the 500 ug/g material to calculate  the MDL  on  the simultaneous ICP,
the value of the MDL is 17 ug/g.  If the  iterative  step is followed then
the 5 ug/g material should  be used.   The 5 ug/g sample is qualitatively
distinguishable but the ratio of S500 to S5 is far greater than 3.05.  The
same is true for 50 ug/g material.  If the 20 ug/g  material is used here
a qualitatively  identifiable signal  is  generated  which has  a standard
deviation  less  than 3.05  times  the  standard deviation  of the  5 ug/g
material.  The pooled MDL is 0.8 ug/g.

The sequential ICP analyzing the 500 ug/g material yields a calculated MDL
of 22 ug/g.   The 5  ug/g material could be used for the iterative procedure
except the  results were  not qualitatively  identifiable.   The  50 ug/g
material is qualitatively identifiable and yields a MDL of 8.4.  The ratio
of S500 to  S50 is  greater  than 3.05.   The 20 ug/g  material  produced no
qualitatively identifiable signal.

For the FAA  the 50 ug/g material  produces a calculated MDL of 4.4 ug/g.
The 5 ug/g material was qualitatively identifiable and yields  a MDL of
3.1.  The ratio of S50 to S5  is less than 3.05.   The pooled MDL is 3.0

C) Molybdenum.  Following the estimated  MDLs  on Table  III,  the 500 ug/g
material was used to determine the calculated MDL on the simultaneous ICP,
which had  a value  of 72  ug/g.   The  30  ug/g material was used for the
iterative procedure but the ratio of S50o to S30 is far greater than 3.05.
Using data  from  the 30 ug/g material,  the calculated MDL  is  8.6 ug/g.
However,  the 5 ug/g material yielded no qualitatively identifiable signal.

The sequential ICP analyzing the 500 ug/g material yields a calculated MDL
of 56 ug/g.   The 50 ug/g material is  qualitatively identifiable and yields
a MDL of 8.4.  The ratio of S500 to  S30 is greater than 3.05.   Using data
from the  30 ug/g material,  the calculated MDL  is  7.5 ug/g. However, the 5
ug/g material yielded no qualitatively identifiable signal.

For the FAA  the 30 ug/g material  produces a calculated MDL of 9.2 ug/g.
The 5 ug/g  material was qualitatively identifiable and yields  a MDL of
5.0.   The ratio of S30 to S5  is less than 3.05.   The pooled MDL is 6.7

D) Selenium.  Table IV lists the estimated MDLs  and other data.  The 500
ug/g material was  used to calculate the MDL on the simultaneous ICP, which
had a value of 29 ug/g.   The 50 ug/g material was used for the iterative
procedure producing and calculated MDL of 17  and the ratio  of  S500 to S50
is less than 3.05.

The sequential ICP analyzing the 500 ug/g material yields a calculated MDL
                                  1-156

-------
of 63 ug/g.   The  50 ug/g material is qualitatively identifiable and yields
a MDL of 30.  The ratio of S500 to S50 is less than 3.05  and the pooled HDL
is 38.

E) Thallium.   Table  V lists  the estimated MDLs.   Using  the  500 ug/g
material to  calculate  the MDL on the simultaneous ICP, the value of the
MDL is 23 ug/g.   The iterative step using 50 ug/g material produced an MDL
of 11  ug/g.  The  pooled  MDL is 14 ug/g.   This is  identical  with the
calculated MDL for the 20  ug/g material which has a S20 of less than 3.05
times  the  standard  deviation of  either  the  5 ug/g or  the  50  ug/g
materials.

The sequential ICP using the 500 ug/g material yields  a calculated MDL of
26 ug/g.  The 50 ug/g material is qualitatively identifiable and yields a
MDL of 22.  The  ratio  of  S500 to  S50 is  less  than 3.05.  The  pooled MDL is
19.  The 20  ug/g material produced not qualitatively identifiable signal.

The FAA  analyzing  the 50  ug/g material produces a  calculated MDL of 26
ug/g.  The 5 ug/g material was qualitatively identifiable and yields a MDL
of 4.5.  The ratio of  S50 to S5 is less than 3,05.  The pooled MDL is 5.0

DISCUSSION

The MDL  process as applied to  arsenic using  simultaneous  ICP produced
three estimates  30,41, and 29 ug/g which were in close agreement.  These
numbers, however, were indistinguishable from the 34.6 ug/g value  that was
obtained from the unspiked soil.  Further,  the inaccuracy found in the 50
ug/g material was quite  high despite the  good precision. For the method
used in this study,  two grams of soil digested by aqua regia and analyzed
by ICP, any results below  50 ug/g are highly inaccurate even if they were
statistically distinguishable from  zero.  This  is also true for selenium
values,  the  poor accuracy and precision obtained at  50  ug/g would make
lower  results  quite  dubious.   Cadmium and  thallium had  accuracy and
precision reasonable results at  50 ug/g but highly inaccurate results at
20 ug/g, which  is far  above the calculated and pooled MDLs.  Molybdenum
results for 30 ug/g were both accurate and precise but the calculated MDL
of 8.6 does  not appear reasonable given the fact that  the 5  ug/g material
was not qualitatively  identifiable.

The sequential ICP results for arsenic exhibit the same type of data as
that obtained from  the simultaneous  ICP.  The  three  measurements of the
estimated MDL were 17, 8.4, and 10 ug/g.   However, the 5 ug/g produced no
measurable signals on the  ICP making it indistinguishable from the blank.
This  raises serious   questions  as  to the accuracy  and  precision  of
measurements near this  point,  as  was the  case for  all   of  the  other
analytes.    Furthermore,   the  sequential  ICP  results for  thallium and
selenium were very inaccurate even at 50  ug/g so lower values were even
more doubtful.   Cadmium,  on the  other  hand, produced accurate results at
50 ug/g  but no  signal at  20  ug/g,  again far  above  the calculated and
pooled MDLs. Likewise, molybdenum results for 30 ug/g  were quite good but
                                   1-157

-------
the calculated MDL  of  7.5  ug/g is unlikely in light of the results from
the 5 ug/g material.   Again the MDL procedure generated values that can
only be described as highly inaccurate.

The FAA data were quite  different from that obtained from either of the
ICPs.   Estimated MDLs for  cadmium were  30,  5,  5,  and  5  ug/g.   The
calculated MDLs were 4.4 and 3.1 while the pooled MDL was 3.0 ug/g.  The
5 ug/g material was accurate, %I = 1.4 and the precision was acceptable,
XRSD - 19.  This was also  true for the molybdenum and the thallium. The
MDL method seems to produce values that are accurate and precise for the
FAA.

SUMMARY AND RECOMMENDATIONS

The  EPAs  method  detection  limit  procedure  is  not  adequate for  the
determination of  "less than" values for elemental  solid waste analysis
using ICP.  As the data clearly shows,  the MDL too often predicts a "less
than" value that is at  a concentration  that  either generates no signal at
all or is subject to unacceptable high inaccuracy. Interestingly enough,
precision generally was not a problem.

The EPA's MDL procedure  seems  to predict reasonably accurate values for
Flame AA determinations.  The calculated and pooled MDLs produced results
that   were  both  precise   and  accurate when using  real solid  matrix
materials.   Even for  Flame AA  however,  the amount  of work  needed  to
generate these MDLs is quite prohibitive.

The EPA MDL procedure is designed to determine the lowest concentration of
an analyte where there is a 99% confidence that the concentration is not
zero.  The determination of this value  is based entirely on the precision
of  the  method without considering  accuracy.    The  smallest  amount  of
analyte that has an acceptable  precision  (i.e.,  MDL)  is  not  the correct
question.   For the regulatory solid waste community the question should be
"what is the smallest amount of an analyte that can be quantified within
quality control limits of both precision and accuracy".

REFERENCES

1.  Appendix A,  July 1982 to Methods for Chemical Analysis of  Wastewater
EMSL-Cincinnati,  USEPA, June 1982

2.  Appendix B to Part  136 CFR 40, October 26,  1984, Federal Register Vol.
49, No.  209,  Pages 198  - 204.

3.  Glaser, J.A.,  Forest, D.L., McKee, G.D., Quave,  S.A., and Budde, W.L. ,
"Trace Analysis for  Wastewaters,"  Environmental  Science & Technology, 15,
1426, December 1981

4.  Kimbrough, D.E.  and J.R.  Wakakuwa,  "Acid Digestion  for  Sediments,
Sludges, Soils,  and Solid  Wastes.  A Proposed Alternative to  EPA  SW 846
Method 3050." ;  Environmental Science and Technology,  23,  pages 898-900,
                                  1-158

-------
July 1989.

5.  Kimbrough, D.E. and J.R. Wakakuwa, "Report of an Interlaboratory Study
of  an  Interlaboratory  Study  Comparing  EPA SW  846  Method 3050  and  an
Alternative  Method from  the  California Department  of  Health Service";
Proceedings of the Fifth Annual USEPA Symposium on Solid Waste Testing and
Quality Assurance, Washington, D.C.  July 1989.  Reprinted in Waste Testing
& Quality  Assurance:  Third Volume, ASTM STP  1075, C.E.  Tatsch,  Ed.,
American Society for Testing  and Materials, Philadelphia, 1991

6.  Kimbrough, D.E. and J.R.  Wakakuwa,  "A  Report of the Linear Ranges of
Several Acid Digestion Procedures "; Proceedings  of the Sixth Annual USEPA
Symposium on Solid Waste  Testing and Quality Assurance, Washington, D.C.
July 1990.

7.  Test  Methods  for  Evaluating Solid Wastes (EPA SW 846 Volume 1A) 3rd
Edition.  Office of Solid Waste and Emergency Response, U.S.Environmental
Protection Agency Washington, D.C., November 1986
                                   1-159

-------
                            TABLE I
                            ARSENIC
                     ESTIMATED MDLs IN /ig/9
            SIMULTANEOUS ICP
                  SEQUENTIAL ICP
     a)



     b)



     c)



     d)
268




165




5.0




5.0
 71




165




 50




 50
            CALCULATED MDL IN Jig/9 SIMULTANEOUS ICP
Spiked Value          500



Mean                  506



Calculated MDL         30



% Inaccuracy          1.2



Standard Deviation    9.7



% RSD                 1.9
             50




             78.7




             41




             57.4




             13.3




              3.5
    5




   51.4




    5.7




   902




   1.8




   3.5
34.6




 7.7








 2.3




 6.7
             CALCULATED MDL IN pg/q SEQUENTIAL ICP
Spiked Value          500



Mean                  479



Calculated MDL         17



% Inaccuracy          4.2



Standard Deviation    5.3



% RSD                 1.1
             50




             42.9




              8.4




             14.2




              2.7




              6.2
              <1
                              1-160

-------
                            TABLE  II
                            CADMIUM
                      ESTIMATED MDLs  IN
       SIMULTANEOUS ICP
                SEQUENTIAL ICP
                                                     FAA
   a)
   b)
   c)
   d)
30
115
5.0
5.0
113
120
 50
 50
 30
5.0
5.0
5.0
             CALCULATED MDLs  IN /ig/g  SIMULTANEOUS  ICP
Spiked Value           500
Mean                   518
Calculated MDL          19
% Inaccuracy           3.6
Standard Deviation     6.1
% RSD                  3.9
                    50
                    55.5
                    4.0
                    11.0
                     1.3
                     2.3
          20
          7.1
          0.9
          74.5
          0.3
          3.9
  5
 7.8
 1.1
 56.0
 0.4
  4.5
1.5
0.07
10.2
              CALCULATED MDLs  IN nq/g SEQUENTIAL ICP
Spiked Value           500
Mean                   500
Calculated MDL         22
% Inaccuracy           0.0
Standard Deviation     6.9
% RSD                  1.4
                    50
                    41.8
                    5.4
                   16.4
                    1.7
                    4.1
          20
         CALCULATED MDLs  IN  fJ,q/q  FLAME  ATOMIC ADSORPTION
Spiked Value           500
Mean                   508
Calculated MDL          15
% Inaccuracy           2.0
Standard Deviation     4.8
% RSD                  0.95
                    50
                    50.4
                    4.4
                    0.8
                    1.4
                    2.8
           5
           5.1
           3.1
           2.0
           1.0
          19.2
                                1-161

-------
                           TABLE III
                          MOLYBDENUM
                    ESTIMATED MDLs IN jitg/g
       SIMULTANEOUS ICP      SEQUENTIAL ICP          FAA






   a)        140                 163                 30



   b)        135                  30                4.5




   c)        5.0                  50                5.0



   d)        5.0                  50                5.0






           CALCULATED MDLs IN /ig/g SIMULTANEOUS  ICP






Spiked Value          500        30           5



Mean                  439        28.3



Calculated MDL         72        8.6




% Inaccuracy         12.2        5.7



Standard Deviation    23         2.73




% RSD                 5.2         9.7






            CALCULATED MDLs IN pg/g SEQUENTIAL ICP



Spiked Value          500        30         5



Mean                  473        10.8




Calculated MDL        56         7.5



% Inaccuracy          5.4       64.0        -         -




Standard Deviation    18         2.4        -         -



% RSD                 3.8        22




        CALCULATED MDLs IN /tg/g FLAME ATOMIC ADSORPTION



Spiked Value          500        30         5




Mean                  453        37.6       4.3




Calculated MDL         46        11.0       5.0



% Inaccuracy          9.4        25.3       14.0



Standard Deviation    14         3.5        1.6




% RSD                 3.2        9.2        37.0





                              1-162

-------
TABLE IV
SELENIUM
ESTIMATED MDLs IN pg/g
SIMULTANEOUS ICP
a)
b)
c)
d)
460
210
5.0
5.0
CALCULATED MDLs
Spiked Value
Mean
Calculated MDL
% Inaccuracy
Standard Deviation
% RSD
500
394
29
21.2
9.1
4.2
CALCULATED MDLs
Spiked Value
Mean
Calculated MDL
% Inaccuracy
Standard Deviation
% RSD
500
425
63
15.0
20.0
1.1
SEQUENTIAL ICP
94
120
50
50
IN /Ltg/g SIMULTANEOUS ICP
50 5
6.56 <1 <1
11
86.9
3.5
53.0
IN /Ltg/g SEQUENTIAL ICP
50 5 _
12.9 <1 <1
30.0
74.2
9.5
74.0
  1-163

-------
TABLE  V
THALLIUM
ESTIMATED MDLs IN jtig/g
SIMULTANEOUS ICP SEQUENTIAL ICP FAA
a) 120
b) 110
c) 5.0
d) 5.0
CALCULATED MDLs IN n
Spiked Value 500
Mean 452
Calculated MDL 23
% Inaccuracy 9 . 6
Standard Deviation 7.4
% RSD 0.96
CALCULATED MDLs IN
Spiked Value 500
Mean 450
Calculated MDL 26
% Inaccuracy 10 . 0
Standard Deviation 8.2
% RSD 1.5
CALCULATED MDLs IN pg/g
Spiked Value 500
Mean 446
Calculated MDL 20
% Inaccuracy 9.8
Standard Deviation 6.3
% RSD 1.4
200 25
150 10.5
50 25
50 25
g/g SIMULTANEOUS ICP
50 20 5 -
41.9 12.1
11 14 - -
16.2 39.5
3.6 4.4
8.6 36.5
Hg/g SEQUENTIAL ICP
50 20 5 -
28.6 - - -
22.0 - - -
43.8 - - -
6.9 - -
24 -
FLAME ATOMIC ADSORPTION
50 5 -
54.7 7.0
26.0 4.5
9.4 40.0
2.5 1.4
4.56 20.6
 1-164

-------
22
        PREPARATION AND VALIDATION OF PROFICIENCY
       EVALUATION SAMPLES FOR SOLID WASTE ANALYSIS

DAVID E.  KIMBROUGH.  PUBLIC  HEALTH CHEMIST II, RUSTUM CHIN,
PUBLIC HEALTH CHEMIST III, AND JANICE WAKAKUWA, SUPERVISING
'CHEMIST,  CALIFORNIA DEPARTMENT OF HEALTH SERVICES, SOUTHERN
CALIFORNIA LABORATORY,  1449  W.   TEMPLE STREET, LOS ANGELES
CALIFORNIA 90026-5698.

ABSTRACT

Development, validation,  and use of Proficiency Evaluation
(PE) samples  for solid  waste analysis has  generally been
performed by  private   firms supplying  them  on  a  self-
analysis  basis   to  private  laboratories.     This  paper
discusses preparation   and   validation  of   two   sets  of
proficiency evaluation  samples.

For the  purpose of  this study  PE  samples  are  defined as
materials used   to  evaluate  laboratory  (as  opposed  to
method, instrument,  or  analyst) performance  for  a  given
matrix specific  analysis.  Any material which  have a method
defined mean  value  and  homogeneity  can  be used.    The
easiest and  most practical  approach is  to spike  a  known
amount of analyte into a well  defined homogeneous matrix.
This not  only meets  the criterion for a PE  sample but also
gives a "true value" in addition to a mean value.

The first set  of PE samples consists  of five soils spiked
with Aroclor 1260, and  the second of five soils spiked with
arsenic,  cadmium,  molybdenum,  selenium,  and thallium.  The
samples were   first   analyzed   at   Southern  California
Laboratory  (SCL) and then sent to twenty eight laboratories
outside of California for analysis.

This validation  study was  designed  to test the  PE sample
preparation procedures   used  at   SCL.      The   data  and
statistical analysis for  this study are presented.

INTRODUCTION

The last  ten years  have seen  an explosive  growth  in the
field of  environmental  chemistry.   Concomitant  with this
growth has been  the development  of laboratory  accreditation
programs  for  environmental   laboratories.     The  federal
government does  not  accredit  environmental  laboratories.
Almost every   state  either  has  or   is   developing  an
accreditation program  for environmental  laboratories.   It
is generally   agreed  that   Proficiency  Evaluation   (PE)
samples should   be  an  integral  part  of   a  comprehensive
laboratory accreditation  program.  Significant progress has
been made in  developing PE samples  for  water  matrices,
                                   1-165

-------
there has been virtually no  progress  in the development of
solid matrix  PE  samples.   This  is  due,  in part,  to  the
newness of  programs   accrediting  laboratories  analyzing
solids and the relative  simplicity of preparing aqueous PE
samples as compared to solid matrix samples.

The California Department of Health Services (DOHS) through
its Environmental  Laboratory Accreditation  Program (ELAP)
is responsible   for   the  accreditation   of  laboratories
analyzing water,    waste  water,   and   solid  waste  doing
business in the  state.   ELAP is  mandated by California law
to distribute PE samples to laboratories they accredit.

PROFICIENCY SAMPLE THEORY

There is  significant  confusion  about  the  distinctions
between PE samples, Laboratory Control Samples (LCSs),  and
Standard References  Materials  (SRMs).    Also a   lack  of
consensus as  to  how  to  prepare these   solid  matrix  PE
samples.  A number of  issues have contributed to this lack
of consensus.   The most important  of which  is  the debate
between spiked PE  samples vs."real world" PE sample.   This
debate revolves  around the benefits of "true values" versus
mean values  or  of "real"  samples  versus  the artificial.
For the purposes of this pilot project a set of definitions
was developed for  LCSs, SRMs, and PE samples.

A Laboratory  Control  Sample  (LCS) is a material  used  by a
laboratory for quality control/quality  assurance purposes
for a method  or  set  of methods.   It  contains the analytes
of interest in concentrations within the  working range of
the method  or methods.   It  is   homogeneous,  of  the  same
matrix type as the samples and is analyzed with each batch
of samples.   The results for each batch should fall within
established control  limits.    This  data  can  be  used  to
monitor long  term  trends  and method  performance.   It  is
immaterial whether the LCS  is spiked or not,  since only a
mean value and standard  deviations  are needed  to create a
control chart and  set control limits.

A Standard Reference  Material (SRM)  is used  to  determine
the applicability,  on a particular   matrix,  of  a method,
method comparison,  instrument, or instrumental performance.
It should be  a  "real  world" sample,  homogeneous  and  must
not be spiked.   The analytes of  interest should be present
in measurable amounts .

Proficiency Evaluation  samples   are  used  to  evaluate  the
performance of the entire  laboratory  system  for  a given
analyte, not  just the  methods or   instruments.    This
includes sample   tracking,    sample   preparation,   record
keeping, method  selection,   method  application,   and  data
                           1-166

-------
reduction.  Like the  other materials, PE  sample  must have
the analytes of interest present in concentrations that are
within the  linear  range of  the methodology.   It must  be
homogeneous and   of  a  matrix that  approximates that  of
actual samples.  Laboratories analyzing solid wastes cannot
be evaluated using a reagent water PE sample.

Although laboratories use  PE samples internally  for self-
evaluation, the most important use of PE samples is as part
of a  laboratory accreditation program.   The  accrediting
agency submits  the  samples blind  to  the  laboratory.   The
laboratory's performance is evaluated based  on the results
obtained.  The  material  can also be  used  for double blind
analysis, that  is  it  can  be  submitted to  the laboratory
without the knowledge of its  personnel.   As  a result, a PE
sample should have a physical appearance that will not give
it away as  a PE sample.   For laboratories analyzing soils,
the PE sample should look like a soil.

Theoretically,  a laboratory  can be put out  of  business  if
it fails  a  PE  sample.     This leaves  the  accreditation
program with a  large  window of liability.   So in addition
to the above mentioned factors, PE sample must be legally
defensible This means PE samples must be validated prior to
distribution.

All of these needs  are best met by using a  spiked sample.
Spiking allows    for    choice    of    analytes,    their
concentrations, the  matrix,  and  can  establish  a  "true"
value.  This last point  is important in  increasing  the  PE
samples legal   defensibility.    Spiked  samples  are  also
easier and  less expensive to prepare.

EXPERIMENTAL SECTION

Experimental Design: To test this approach to PE samples, a
pilot project  was  designed.   The  project had  two  goals.
The first  was  to test the PE  sample preparation process.
The second  was  to  test  the  validation  procedure  for  PE
samples.

PE Sample Preparation

From previous experience in the preparation of soils spiked
with inorganic  analytes1'2  it was  decided  to  use water
soluble salts  of the  target  elements.    The use  of water
soluble salts means  that it will  be  easy for laboratories
to solubilize   the  target  elements  using  any  digestion
procedure.  It  is  also  easier  to make  spiking materials
using water soluble salts.  Strong oxidizers can attack the
organic component of  the soil and volatilizing it, leaving
the heavier silica and alumina portions. This increases the
                            1-167

-------
density of the  soil and  changes  the concentration  of the
spiked analytes.    The  appearance  of  the  soil  is  also
altered making  it  look more artificial. For PE  samples to
be used  as  double  blind  checks  they should  appear  as
natural as possible.

A large amount  of  a local soil was  collected,  milled, and
sieved through U.S. Standard No. 10  (2 cm2) sieves.  It was
analyzed for native amounts of sixteen elements regulated
by the  State3'   Chromium,  cobalt,   copper,  lead,  nickel,
vanadium, and zinc  were  found  to be  present in excess of 5
mg/ka.  Since,  most laboratories  use  EPA  SW  846  method
3050^ as  the digestion  procedure  antimony,  barium,  and
silver will  be   poorly solubilized.   These three  elements
were not be  used.   Beryllium  was  not used as most  of the
water soluble salts are  either unavailable commercially or
are extremely toxic.   Beryllium sulfate is relatively safe
but has a very  small mole fraction of beryllium. This would
require such  large amounts  of beryllium  sulfate  that the
soil matrix  would   be  disturbed.     This  left  arsenic,
cadmium, molybdenum, selenium,  and  thallium as  the  target
elements.  The   inorganic  samples   were  prepared  in  the
following fashion:

           Sample A Sample B Sample  C Sample D Sample E
Arsenic     4,000      500       50       5       -
Cadmium       500      50        5       -     5,000
Molybdenum      50       5             5,000      500
Selenium         5       -    5,000      500       50
Thallium         -   5,000      500       50        5

(Here and throughout this paper the dash  mark,  -,  will be
used to indicate that  the analyte in question will  not be
spiked and the  analyte is at background concentration.)

California ELAP  accredits   about   200  laboratories  for
inorganic hazardous materials  analysis.   Five  kilograms of
each sample  was  prepared  enough  to  provide  a  20  gram
aliquot of  each  sample  to the   individual  laboratories.
Spiking solutions  were  prepared  for  each of  the  salts
listed below.    These  solutions were  diluted  one  to one
hundred and   checked  against  standards   prepared  from
different stock  materials.   All of the solutions were well
within 10% of the expected value.

Salt              MF   Mass Element  Mass Salt    Cone.
As2O3             0.757   100 g      132 g       lOOg/L
3Cd(S04)'8H20     0.438   50 g       114 g       50g/L
(NH4)6M07024-4H20 0.543    5 g       9.2 g       5 g/L
H2SeO3            0.612   50 g       82 g        50g/L
T12SO*            0.808    5 g       6.19 g      5 g/L
MF = Mole Fraction
                            1-168

-------
The amount of  each salt that was to be  added to each soil
was calculated and totaled  as noted below.   The amount of
salts to be  spiked was subtracted from  the  5 kilograms of
soil.  Thus  when the  salts were  added,  the  total weight
would be 5 kg.   Due to the limited  solubility of ammonium
molybdate and  thallium sulfate, dry  salts were  added for
the 5,000 mg/kg  materials.    The appropriate  mass  of soil
was placed  in a  plastic tray  and mixed  with  enough de-
ionized water  to make  a slurry.   The  slurries are then
spiked with the amounts of  the salts as indicated below.

                 Sample A Sample B Sample C Sample D Sample
E Arsenic Trioxide  26.4g   3.30g     0.38g 0.04g
                 (200 ml)(25 ml)(250 ml 1:100)(25 ml 1:100)

Cadmium Sulfate    5.12g   0.51g     0.05g 51.2g
                   (50 ml) (50 ml 1:10) (5 mll:10) -(500 ml)

Ammonium Molybdate  0.28g     O.OSg      -      46.1g 4.6g
                   (30 ml)    (5 ml) (500ml)

Selenium Trioxide  0.04g      -      41g        4.lg 0.4lg
                 (5  ml  1:10)           (500  ml)  (50 ml) (50
mil:10)
Thallium Sulfate    -        30.7g    3.07g      0.31g 0.03g
                                   (500ml)   (50 ml)  (5 ml)

Total Mass          39 g     34 g      44 g      50 g 62 g
of Salts

This mass was removed from  each  5.00 kg batch.

Mass of Soil   4.961kg 4.966 kg  4.956 kg 4.950 kg 4.948 kG

The slurries were  then dried at  95° C with frequent mixing.
After drying,  the  materials  were  again milled  and sieved
through a U.S. Standard No. 10 sieve.

The PBC Aroclor  1260 was used to spike the same native soil
as was used to make the inorganic PE samples.  The soil was
found to be  free of PCBs,  although low level  interferences
from decomposed  vegetable   matter  was  detected.   The soil
was milled,  sieved, and autoclaved  (to kill  any bacteria
that might be present).  Approximately 150 laboratories are
certified for  PCB  analysis  by  California  ELAP.    Ten
kilograms of  each  PE  sample was  prepared  so   that each
laboratory could  be provided  with about  25  grams per PE
sample.  The five  PE samples were prepared in  the following
concentrations in mg/kg:
                            1-169

-------
               Sample F Sample G Sample H Sample I Sample J
Aroclor 1260     100      10       1.0       0.1      -

Two solutions  were prepared;  one contained  10  g Aroclor
1260 in 1 liter of hexane, while a second was made form a 1
to 100 dilution  of the first  solution.   The  Aroclor 1260
was made up  from neat PCB and checked  against EPA derived
standards and found to be within  5%  of  the expected value.
These solutions were spiked in the following fashion:

             Sample F Sample G Sample H Sample 1 Sample J
Aroclor 1260  100 ml   10 ml   100 ml    10 ml
                                (1:100)   (1:100)

The samples  were slurried with n-hexane,  homogenized,  and
dried at room temperature with periodic mixing.

Validation

A two step validation was used.  The initial validation was
performed in-house.   The inorganics were digested  seven
times using  an  aqua  regia  method5  (which  in  previous
publications was  referred to  as  the SCL  method1'2'5)  for
analysis by  simultaneous  inductively coupled plasma atomic
emission spectroscopy (ICP-M), sequential  ICP (ICP-Q),  and
Flame Atomic Absorption  Spectroscopy (FAA).   For Graphite
Furnace Atomic  Absorption Spectroscopy  (GFAA)  EPA SW-846
method 3050  was  used.   Analytical methods  include EPA SW-
846 methods 6010, 7061,  7130, 7131, 7480, 7481, 7740, 7840,
and 7841.  The organic  samples were  analyzed in duplicate
using EPA SW-846  method  3540,  3620,  and 8081.   All of the
results were within  20%  of the  expected values  and  had a
relative standard deviation of less than 30%.

The samples  were  validated  by having at  least twenty (20)
laboratories which are  not  accredited  by  California ELAP
analyze the  materials.    The  samples  would  be considered
valid if the mean  value from these laboratories was within
20% of the spiked value  and the  percent relative standard
deviation (%RSD)  is  less  than   20% for   the  two  higher
concentrations for each analyte.   It is to be expected that
the %RSD for an  analyte will  increase as the concentration
decreases, all other things being the same.  In the case of
the inorganic materials each  low  concentration analyte was
in a material  with a high  concentration analyte.   So  for
the lower concentrations inorganic analytes, the analyte is
considered validated  if  the mean  value  was within  of  20%
the spiked value and the  high  concentration analyte in the
same sample had an %RSD of less than 20%.
                            1-170

-------
RESULTS

For the inorganic samples, the majority of the laboratories
prepared the samples by an acid digestion, in most cases by
EPA SW   846   method   3050   or   a   similar  method.   Two
laboratories used  chelation  extraction  for  some  of  the
analytes.  For  the  energy  dispersive  X-ray  fluorescence
instrument (EDXRF) the  samples were ground to pass  a U.S.
Standard sieve No. 200.

The digestates and extracts of the PE samples were analyzed
on a number of  different instruments,  ICP-MS,  simultaneous
and sequential  ICPs,  FAA, GFAA,  hydride  generation  atomic
absorption (HGAA),     colorimeter,      and     fluorescence
spectrometer.   The  soils were also  analyzed  directly  by
EDXRF.  Again all of the mean values were within 20% of the
spiked values except for arsenic at 5 mg/kg.  See Table I.

The PCS results were more complex.   All of the mean values
were within 20%  of  the true value and  %  RSD was less than
30% for all except the  lowest  sample.   As can  be  seen on
Table II  the  mean  tends to  be  low and the %RSDs  high.
However, this  is  consistent   with  previous  efforts  with
solid matrix  PCB  materials.     The data was  unlike  the
inorganic materials where there was a  normal  distribution
about a  mean.    The   data  for  the  three  highest  PCB
materials, samples F, G, and H were bi-modally distributed,
with one mode at around  95% recovery and  another at 75%.

Outliers

Two types of  outliers  were identified;  even multiples and
base line interferences. Even multiples are values that are
an even factor  off of  the true value.   These are caused by
either omission  or inclusion  of dilution  factors  or bad
standards.  For  example, there  is a  Cadmium outlier for
sample C.   It is exactly 40  times the true value  and the
worksheet indicates that a 1:40 dilution  occurred.  Similar
errors seem  to  have   occurred three  other times  in  the
inorganic samples.   Bad standards  account for the five
arsenic outliers  which were  all five  to seven  times the
true value and the five high PCBs from  laboratory number 1.

Base line interferences  occurred in the  inorganic samples.
This is   caused  by   an  interference   that  raises  the
analytical background.    Without  appropriate  background
correction, the instrument reads significant amounts of the
analyte in the blank soil.  This caused elevated results in
the lowest  spiked  value.    This  can   be  seen  for  two
laboratories with   molybdenum   and   thallium.       Every
laboratory except  two,  both  with  high  reporting  limits,
measured significant  amounts   of arsenic in the  unspiked
                            1-171

-------
sample (E).

DISCUSSION

Almost all  of  the  inorganic  PE materials  were validated.
The accuracy   and   homogeneity  was   within   acceptance
criteria.  The mean values were all  well  within 20 percent
of the   spiked  values   and   the   %RSDs  for   the   high
concentration analytes  were  all  less   than  12%.     The
exception to this rule  was  arsenic  at  5  mg/kg  level  in
sample D and the  unspiked sample E.    In  almost every case
the readings  from  sample  E,  when  subtracted  from  the
results of  sample D,  produced  results  within 20  %  of the
spiked value.  From this it can be concluded that, in fact,
there was a  small amount of arsenic present  in the native
soil.  This  is not  surprising  since  arsenic  is commonly
found in soils  in the low ug/g range8.  As can  be seen on
table I the mean  value for arsenic for sample D is 13 ug/g
and 8.8 ug/g for  sample  E, the  difference  being 4.2 ug/g,
88% of the  spiked value.  Table III  shows  the results for
arsenic instrument by instrument.  As can be seen, for each
instrument type,  the difference between the mean value for
samples D and  E is about equal  to the spiked value of 5.0
ug/g.  The  ICPs obviously had  a high  spectral background
problem was well.

One point  is clear,  that if  an  analyte  is  going  to  be
spiked at   a   concentration   that   is   near   the   lower
quantitation limit,  to  properly validate  it,  an internal
standard must be  spiked  in as well at the same time as the
analyte.  The internal standard should be at a much higher
concentration and  should be  an easily analyzable analyte
such as beryllium or  cadmium.   It is  also  possible to use
background analytes, in  soils this  internal standard could
be iron  or  calcium.  For future validation  studies where
low concentration spikes are planed,  it is  suggested that a
liquid standard also be  analyzed containing these analytes
at low concentrations.

The data for the  PCBs is more  difficult  to interpret.  As
can be seen, the  data is bi-modal for  samples  F, G,  and H
while the   results   for  sample   I   is  more  normally
distributed.  Specifically, There are  nine  sets of results
clustered around the spiked value (labs 2-10) and another
eight sets  clustered  around  70% recovery.   The clustering
is not evident  in the results  for  sample  I.  The source of
this bi-modality  appears  to  be a  result of  instrument
calibration bias.     This  conclusion  is  by   means  of
elimination.  Laboratories in  both clusters  used the same
extraction equipment,   solvents,   methods,   instruments,
columns, and procedures.   The only remaining difference was
the calibration procedures.   One possibility  is the base-
                            1-172

-------
line selection procedure used during integration.

One conclusion  is  evident, the  PCB PE  samples themselves
are homogeneous.   This can be shown by  two facts:  Samples
F, G, an H all  produced results  equal  to or lover than the
spiked value with  one  exception.   Second,  the response was
generally linear  i.e.,  F,  G,  and  H  were proportionally
lower in each case.   It would be  unreasonable  to expect a
set of heterogeneous materials to respond in this fashion.

There is also a question of  accuracy.   For example, sample
F was spiked at 100  ug/g but  the mean  value is 85 ug/g.  A
similar situation  exists with samples  G and H.   Sample I,
on the other hand,  was spiked at  0.10 ug/g and had a mean
value of  0.109.   Were then,  samples  F,  G,  and  H  made
incorrectly leaving  these  materials   with  85%   of   the
expected value?    If  the  data  were  distributed  normally
about an 85% recovery this might be reasonable.  But as has
been noted, there  are  two normal distributions, one around
95% recovery and another around 75% recovery.   For samples
F, G, and H  there  are no more than  15%  of the results are
near the mean value (80 to  89%) .   Further, since  H  and I
were prepared from the exactly the  same solution and base
soil, and F and G were  made from the same stock as H and I,
it would  be  unreasonable  to  conclude that  the materials
contain a  concentration  other  than   the  spiked  value.
Rather, there are  two  distinct populations of laboratories
analyzing the same materials and getting two different sets
of results.

SUMMARY

Using spiked materials  is  a  useful and economical approach
to preparing  PE   samples   for   solid   waste  laboratory
evaluation.  The  preparation  steps  used  in  this  study
generate accurate  and  homogeneous  materials.   Difficulties
occur when  very low  concentrations  are  spiked and  this
should be  avoided  or  combined with an  internal standard.
PCB spikes are  difficult,  not due  to the preparation steps
but to  the  linearity  problems obvious  in  the  analysis
process.  More  study is needed  to identify the  source of
this problem in PCB  analysis  and so the use of PCBs spikes
for PE samples must be  used with discretion.

ACKNOWLEDGMENTS

We would like to thank  Dr. William Nilsson  and Monina Ligao
of the Southern California Laboratory  for their assistance
with the PCB work.  We would also like  to thank Dr.  Tony
Harding of Spectrace Instruments Inc.  for  his  work on the
inorganic materials and for the EDXRF spectra.
                            1-173

-------
PARTICIPATING LABORATORIES

Alaska Department of Environmental Conservation
Environmental Quality Monitoring & Laboratory Operations
Douglas Laboratory

Arizona Department of Health Services
State Laboratory Division

California Department of Health Services
Hazardous Materials Laboratory

COMPUCHEM Laboratories Inc.

Connecticut Department of Health Services
Bureau of Laboratories
Environmental Chemistry Section

Hittman Ebasco Associates Incorporated of Columbia Maryland

Hawaii Department of Health
Chemistry Branch

Indiana State Board of Health
Environmental Laboratory Division

Lower Colorado River Authority
Environmental Laboratory
Maine Department of Environmental Protection
Bureau of Administration
Division of Laboratory Services

Michigan Department of Natural Resources
Environmental Laboratory

Minnesota Department of Health
Chemical Laboratory

Mississippi State Chemical Laboratory

Ohio Environmental Protection Agency
Division of Environmental Services

Oregon Department of Environmental Quality
Laboratory Division - Portland

Pennsylvania Department of Environmental Resources
Bureau of Laboratories

Quebec Ministers de 1'Environnement
Direcion des laboratoires
Laboratoire de Montreal
                            1-174

-------
Research Triangle Institute
Center for Environmental Measurement & Quality Assurance

Salt River Project
Lab & Field Services Division
Environmental Laboratory

South Dakota Department of Health
State Health Laboratory

Tennessee Valley Authority
Environmental Chemistry, Water Quality Department

U.S. Army Corp of Engineers
Missouri River Division Laboratory

U.S. Environmental Protection Agency
Environmental Monitoring Systems Laboratory-Las Vegas
Methods Research Branch

U.S. Department of Interior
Bureau of Reclamation
Assistant Commissioner-Engineering & Research
Research & Laboratory Services Division

U.S. Geological Survey
National Water Quality Laboratory
                            1-175

-------
REFERENCES

1. Appendix A,  July 1982 to Methods  for Chemical Analysis
   of Wastewater EMSL-Cincinnati, USEPA, June 1982

2. Appendix B to Part 136 CFR 40, October 26, 1984, Federal
   Register Vol. 49, No. 209, Pages 198 - 204.

3. Glaser, J.A.,  Forest, D.L.,  McKee,  G.D.,  Quave,  S.A.,
   and Budde,  W.L.,  "Trace  Analysis   for  Wastewaters,"
   Environmental Science &_ Technology.   15,  1426,  December
   1981

4. Kimbrough, D.E.  and  J.R. Wakakuwa,  "Acid Digestion for
   Sediments, Sludges, Soils,  and  Solid Wastes.  A Proposed
   Alternative to EPA SW 846 Method 3050."  ; Environmental
   Science and Technology,  23, pages 898-900, July 1989.

5. Kimbrough,  D.E.   and J.R.   Wakakuwa,   "Report  of  an
   Interlaboratory Study   of   an   Interlaboratory   Study
   Comparing EPA  SW  846 Method  3050  and  an  Alternative
   Method from   the   California   Department  of   Health
   Service"; Proceedings   of   the   Fifth   Annual   USEPA
   Symposium on Solid Waste Testing and Quality Assurance,
   Washington, D.C.  July 1989.  Reprinted in Waste Testing
   &_ Quality Assurance:  Third Volume. ASTM  STP  1075.  C.E.
   Tatsch, Ed., American Society for Testing and Materials,
   Philidelphia, 1991

6. Kimbrough,  D.E.  and  J.R.  Wakakuwa,  "A  Report of  the
   Linear Ranges of Several  Acid  Digestion  Procedures ";
   Proceedings of the Sixth Annual USEPA Symposium on Solid
   Waste Testing and  Quality  Assurance,  Washington,  D.C.
   July 1990.

7. Test Methods  for  Evaluating Solid  Wastes (EPA SW  846
   Volume 1A)  3rd   Edition.    Office  of  Solid  Waste  and
   Emergency Response,     U.S.Environmental     Protection
   Agency:Washington,  D.C.,  November 1986

8. Arsenic; Committee on Medical and  Biological  Effects of
   Environmental Pollutants,  National  Academy of  Science,
   Washington D.C., 1979
                            1-176

-------
                                  TABLE I

                             INORGANIC RESULTS
Arsenic N = 25
True Value
Mean Value
SD
%RSD
Reporting 
-------
           TABLE II
PCS RESULTS IN MICROGRAMS/GRAM
Sample
Spiked
Value
LAB 1
LAB 2
LAB 3
LAB 4
LAB 5
LAB 6
LAB 7
LAB 8
LAB 9
LAB 10
LAB 11
LAB 12
LAB 13
LAB 14
LAB 15
LAB 16
LAB 17
LAB 18
LAB 19
LAB 20
Mean
SD
%RSD
N =

-------
                   TABLE  III
Data from Verification of PE  Samples for Arsenic
total Results
H= 25
ID
True Value
Hean Value
SD
IRSD
Reporting 
-------
00                        OBSERVATION OF QUALITY ASSURANCE
*"                          ANOMALIES IN SUPERFUND ACTIVITIES
                                                by
                                         Dr. D.M. Stainken
                                        Malcolm Pirnie, Inc.
                                       2 Corporate Park Drive
                                   White Plains, New York  10602
         The Superfund  Program consists of numerous components, programs, contracts and
         documents which have contributed to placing sites on the NPL, PRP and remedial actions.
         During this process, quality assurance activities within EPA and the States have continually
         evolved.  Accordingly, there are QA requirement for activities under the Clean Water Act,
         Safe Drinking Water Act, Clean Air Act, RCRA, and the NCP, as well as State programs
         which can affect Superfund actions as ARAR's.  Historically, the Superfund paper-trail
         process makes use of field sampling plans, QA program and project plans, and ultimately,
         Records of Decisions (ROD).  When a site is to be remediated, additional RI/FS work may
         be necessary for RD/RA design.

         A review of numerous QA documents within the Superfund process indicates that anomalies
         do occur  in the  process. As examples, RODs  have been reviewed in which numerous
         compounds were mis identified or reported as isomers based on incorrect R fits or mass
         spectra, and compounds were reported in aqueous media at values greatly exceeding solubility
         maxima. In other reviews, wrong analytes are listed on wrong lists with intermixed methods,
         or the ARAR end point MDL is mismatched with the CLP CRQL.  This paper will present
         a synopsis of QA anomalies observed in reviewing QA activities.
         PAPER.DS
                                               1-180

-------
24  FUNCTIONAL EVALUATION OF Q C SAMPLES, A PROACTIVE APPROACH

                    DONALD XIOUES AND JANICE ALLISON

                       QUALITY ASSURANCE DIVISION
                           BECHTEL NATIONAL, EMC.
                            OAK RIDGE, TENNESSEE
  Abstract

  An aggressive and systematic approach in using and evaluating Quality Control
  (QC) samples provides improved data quality and reduced costs for environmental
  sampling. Unlike the traditional approach, where evaluation of QC samples has
  typically been performed after all sample results have been received from the
  laboratory and the process of data validation/evaluation is underway, a proactive
  approach involves strategic QC sampling, rapid analytical turnaround, and
  preliminary review of QC sample results. This "just-in-time" approach allows for
  decisions to be made regarding isolation of contamination, minimization of error
  associated with data, and avoiding unnecessary sample analysis costs. This
  information provides the sampling team with necessary data to identify, isolate, and
  eliminate sources of contamination and provides a definitive means to determine if
  resampling is required while still in the field, thus avoiding remobilization costs.
  Additional savings result from eliminating unnecessary sample analysis and
  reducing the need to resample. Data quality is improved by reducing data qualifiers
  and minimizing  data gaps. The goal of this manuscript is to summarize the
  decision-making process in identifying the feasibility of the proactive approach and
  to provide a framework for that process by means of a decision-tree flowchart.

  Introduction

  Quality assurance (QA) and quality control (QC) are programmatic and systematic
  procedures to ensure that a product of known quality is produced. This "quality"  is
  defined for environmental sampling by quantitative data characteristics, accuracy,
  precision, and minimum acceptable  detection limits. The measure of how close our
  data comes to the true concentration of the contamination present at the site defines
  accuracy. Precision is a measure of how reproducible our results are. Minimum
  acceptable detection limits determine the amount of contaminant necessary to be
  present for detection. These  key elements are defined during the establishment of
  data quality objectives (DQOs) for the scope of work. Based upon the decisions
  which will be made with the sampling data and the consequences of failure, a
  minimum acceptable level will become evident. Cost effective sampling strategies
  can minimize the number of samples required to obtain the required data necessary
  for the decision making process. The result of this strategy is that as the number of
  samples is reduced, the quality of each sample data point becomes increasingly more


                                         1-181

-------
critical. If information from a sample representing a given area is deleted or
becomes unusable, a gap in the data will exist.  Depending upon the importance of
this information to the achievement of the scope of work, the data gap created by
missing or substandard data may be sufficient for resampling to be required. For
example, qualitative data may be acceptable for preliminary site investigations.  This
situation may not justify the additional costs associated with the proactive
evaluation of QC samples, requiring rapid turnaround, as qualitative (estimated)
data may be satisfactory.  On the other hand, data necessary to support closure or
permitting activities would require quantitative data.

Traditional sampling and analysis plans call for critical samples to be analyzed
under rigorous analytical standards at qualified analytical laboratories such as
laboratories participating in the Contract Laboratory Program or one of several
interlaboratory comparison programs.  Unfortunately, this process is time
consuming.  Constraints from laboratory scheduling, holding times and
documentation requirements result in a routine 45 day turnaround time for the
results from a  sampling event.  It is at this point that the data are reviewed and
evaluated.  The consequence of discovering that a sample or batches of samples  fail
to meet the minimum criteria at this point has  serious implication on both cost and
scheduling.

Minimum criteria for sampling are measured by the results of the QC samples.  If
the QC samples are discovered to  have a significant level of contamination on
review after the typical 45 day wait, it may be too late to remedy the situation at  this
point. However, if the QC samples had been evaluated earlier before the other
associated (within the same sample batch) samples were analyzed, several options
would then be available:  1) discarding the data completely, thus eliminating the
data for that point or area; 2) finding the source of the problem and eliminating  it
with subsequent resampling and reanalysis; 3) only using the data with analyte
concentrations significantly above background data (> 5X or 10X above the
concentration found in  the blanks); or 4) accepting qualitative data (instead of
quantitative), as all data less than 5X or 10X above the concentration found in the
blanks will be  qualified as estimated (J) ( EPA Functional Data Validation
Guidelines).

Conclusions

This "just-in-time" approach allows for decisions to be made regarding isolation of
contamination, minimization of error associated with data,  and avoidance of
unnecessary sample analysis costs.  This is accomplished by  reviewing the QC
sample results which are scheduled for rapid turnaround (24 to 48 hours) analysis
before other associated  samples are analyzed.  This provides initial review of QC
results and the option of cancelling associated sample analyses if it is decided that
estimated data are or will not be acceptable. It is vital that the rapid turnaround  data
be evaluated immediately so that real-time decisions can be made. For example,
sources of contamination can be identified, isolated and eliminated (see Figure 1),
                                       1-182

-------
thus preventing further contamination of QC and associated samples; and decisions
can be made to resample while still in the field, thus avoiding remobilization costs
(see Figure 2). These options may result in further cost savings by avoiding
unnecessary analyses. Scheduling delays are reduced as a result of faster evaluation
(as opposed to the traditional 45 day, standard turnaround wait).  If these decisions
are used to avoid the cost of remobilization, eliminate unnecessary analyses, and
insure that data meet DQOs, then thousands, or possibly million, of dollars in
project and analytical costs will have been saved. The likelihood of poor decision
making is reduced, and the project tasks will have been performed as efficiently and
cost-effectively as possible.
                                         1-183

-------
Contamination Source  Isolation Process
                                                       Preservation contamination.
                                                      Heminate source, evaluate new
                                                      detection limits, cancel analysis
                                                        of associated samples, and
                                                               resample.
                                       Preservation blank
                                         contaminated?
                     Held blank
                    contaminated?
  Contamination source in
  transportation/storage or
containers.  Eliminate source,
evaluate new detection limits,
 cancel analysis of associated
   samples, and resample.
Contamination from ambient
    sampling conditions.
 Qeminate source, evaluate
 new detection limits, cancel
analysis of associated samples,
       and resample.
                     Wash blank
                    contaminated?
                                                                               Contamination from
                                                                            decontamination procedures.
                                                                           Heminate source, evaluate new
                                                                           detection limits, cancel analysis
                                                                             of associated samples, and
                                                                                    resample.
              If no contamination has been
            detected in any of the field blanks
             then no evident contamination
               has been introduced by field
                       methods.
                         Figure 1
                                   1-184

-------
                                                                           Pro-active  QC Decision Tree
                            Sampling
Rapid turnaround of QC for
     critical samples
 Evaluation of QC results to
  determine if samples will
meet the required minimum
   criteria for data end-use.
Does the data meet the
 minimum usability
    standards?
                                                                                                                                                    YES_»
Analyze associated samples
00
01
                                                                                                                                NO
                                                                                                                                 1
                                                                                                                      Cancel analysis for associated
                                                                                                                         samples and direct
                                                                                                                            resampling.
                                                                                             Figure 2.

-------
25
   FEATURES OF THE U.S. EPA-QOALITY ASSURANCE MATERIAL BANK STANDARDS

Ruth A.  Zweidinger,  Chief,  Analytical Chemistry,  ManTech Environmental
Technology, Inc.,  2 Triangle Drive, Research Triangle Park, North Carolina
27709 and Nathan Malof, Research Chemist, Environmental Monitoring Systems
Laboratory, U.S.  Environmental Protection Agency, Cincinnati, Ohio 45268.

ABSTRACT

The U.S.  EPA currently provides organic solution standards  to its Contract
Laboratory Program (CLP) laboratories under the Quality Assurance Material
Bank (QAMB).  These  standards ensure  comparability between laboratories
and traceability  to  U.S. EPA  materials.   ManTech Environmental operates
the current QAMB  program over which U.S.  EPA maintains a strong quality
assurance oversight role.

The quality assurance  requirements  for QAMB organic  solutions standards
have a number of  distinct features and this presentation addresses each.
For example, one  requirement is the characterization of  the neat materials
with  purity  analyses,  including  moisture  analyses  of  hygroscopic
compounds,  and  confirmation  of the  compound identity.   An  U.S.  EPA-
supervised,  independent laboratory analyses the  purity  to  verify  the
original assay.   Furthermore, the  identity confirmation  procedures  are
designed to be unequivocal.

Each lot of standards is analyzed to verify its concentration by ManTech
Environmental and again by an U.S. EPA-supervised, independent laboratory.
Reanalysis of the standards is required periodically to verify stability,
according to a schedule optimized for each standard. Strict control limits
are placed on each of these analyses and will be presented.  The details
of  these  specific  requirements  and  their  implications on  confidence
intervals associated with the standards will be presented.

The  probable error  of each  step in  the solution  standard  production
process has been evaluated and used to assess the overall probable error.
This information, along with  the analytical method performance data for
the  quality  control,  serve  to  define the  quality  of  the  standards
available under the QAMB program.  The results of this evaluation will be
presented.
                                           1-186

-------
26
AUTOMATED DATA VALIDATION—PANACEA OR TOOL


Gary Robertson, U.S. Environmental Protection Agency, Environmental
Monitoring Systems Laboratory, Las Vegas, Nevada


ABSTRACT

     The  increased  emphasis on  the clean-up of  hazardous waste

sites and the quality  assurance (QA)  of environmental  data has

caused a substantial increase in the amount of analytical data that

must be reviewed.   Computer programs such as Computer-Aided Data

Review and Evaluation (CADRE) for the review of Contract Laboratory

Program data  and  "E-DATA"  for the review of analytical data from

emergency response teams have been developed to help deal with that

large amount  of data.   Such programs can be of great value to the

data  reviewer;  however,  the user  must be  aware  of  what these

programs can and cannot accomplish.  The types of  QA measures that

may be checked by computer such as calibrations, holding times and

analytical  sequences  will  be  described.    The  limitations  of

computerized  checking will also be discussed including the areas

of analytical methodology,  electronic data formats, chromatographic

quality and professional judgement.  The needs for standardization

and QA of the electronic data will  be  considered.


Notice:   Although the  information  discussed in this article has
been  funded wholly  or in part by the  United  States Environmental
Protection Agency,  it has  not  been  subjected to Agency review and
does  not  necessarily  reflect  the views  of  the  Agency  and no
official  endorsement should be inferred.
                                   1-187

-------
27    BUILDING DATA QUALITY INTO ENVIRONMENTAL DATA MANAGEMENT

    Mitzi Miller, Automated Compliance Systems, 673 Emory Valley Road, Oak Ridge, TN 37830; Dr.
    Philip Ludvigsen, Automated Compliance Systems, 245 Highway 22 West, Bridgewater, NJ 08807.

    ABSTRACT

    With the volume of environmental data increasing and the need to make rapid, accurate decisions,
    managing the data is essential. Data Management includes consolidating all information into
    central, consistent, accurate data bases. From these  data bases, information can  be selected,
    queried  and electronically transferred to  statistical,  graphical and  reporting packages.   By
    establishing a database with output to decision support software, accuracy and speed of decision
    making is improved. An essential part of computerizing the data is establishing a credible data base.
    This paper describes methods to improve the quality, integrity, and connectivity of information.

    INTRODUCTION

    There are many different programs requiring collection and analysis of environmental data. Among
    these are those required under the Comprehensive Environmental Response,  Compensation and
    Liability Act (CERCLA) and the Superfund Amendments and Reauthorization Act (SARA). In
    addition, the National Pollutant Discharge Elimination Systems (NPDES) requires that any effluent
    from a facility be monitored regularly according to permit requirements. Other regulations require
    monitoring of air and drinking water. No matter which program is discussed, data must be collected
    and evaluated. The problem is that if the project is large or if monitoring data is collected for
    several years, one is inundated with information. This information becomes difficult to sort and
    evaluate. While many of begun using spread sheets and  other similar tools, the amount of data can
    quickly exceed the capacity of many tools. The other problem is that the  technical staff and
    managers typically do not enjoy entering the data. Even if diskette deliverables are presented,
    someone must map and move the  data into  the existing database. Many databases have been
    established only to find that the information is inaccurate, inconsistent and difficult to retrieve. The
    purpose  of this paper is to outline issues in data management and an strategy for success. The
    information is  based on the experience which Automated Compliance Systems  has had in
    managing databases of over 3,000,000 records on over 100 projects. These projects have included
    CERCLA, RCRA, NPDES, air monitoring and other environmental projects.

    PLANNING

    Some regulations such as CERCLA, SARA and RCRA require that planning documents be
    written. Others require  that data be documented but  do not require specific planning
    documentation such as sampling plans and QA/QC Plans. Most projects do not include a data
    management plan. Millions of dollars are spent collecting and evaluating the data however, little
    thought is given to planning how the data  will be passed between parties, what information will
    be captured and passed, and who will be responsible for this process. The first recommendation
    for large projects is to outline a plan for managing the data.
                                               1-188

-------
The goal of any data management system should be to consolidate the data and allow the end user
to be able to use and evaluate the information. The primary purpose of a data management plan
is the communication of how information will  be captured, accessed, entered, and used. Data
management plans should include:

     1) Data Dictionary
     2) Data Naming Conventions
     3) Data Entry Criteria
     4) Data Consistency Filters
     5) A Traffic Control System
     6) QC Data Elements and Relationships
     7) System Design Strategy
     8) Audit Trails and Entry Serial Numbers
     9) Connectivity Requirements
     10) Tools for Compliance Screening/Data Validation
     11) Staff Responsible for Data Management

The following sections will outline issues and provide some examples of success when these areas
are addressed.

DATA DICTIONARY

Data dictionaries  often include the elements of data  to be captured and the definitions of these
elements. Some data dictionaries include the location of information in the database and diagrams
of the relationships between the data. While this is useful, the most important information is the
data elements and terms. This assures that all project members have consistent understanding of
the pieces of information to be captured.

Another critical issue in establishing a data dictionary  is that the geologist will not look at data the
same way the laboratory looks at the same piece of information. The data dictionary needs to be
reviewed by project team members with different view points. This insures that the same data
element will be understood by all parties.  A benefit of the dictionary is that redundancy in the
data system can be reduced. (1)

DATA NAMING CONVENTIONS

The planning document should include a method of naming and identifying samples. Samples are
often collected by one group, analyzed by another and evaluated by yet another group.  The key
information which is passed along is the sample identity. If all parties do not know the convention
for numbering samples mistakes can be made.

One issue with naming a sample is how much information to include in the name. Many projects
have met disaster because the convention included more fields than the sampling, lab or  end user
database would allow. If this occurs the identity may be truncated and  make matching  the data
                                            1-189

-------
with the sample number difficult. It is important that all the parties involved understand how data
is collected and entered  into the respective databases along the data processing path. As an
example if the sample collection team had understood that the laboratory computer could only
accept 12 digits for the customer sample number, then the sample number could have been
appropriately sized. Long naming conventions increase the chance of data entry errors. If the name
is exceedingly long, the laboratory and end users often have  trouble accurately entering the
number for laboratory sample tracking and for using the information. If this occurs the association
between the data and the sample may be incorrect. Keeping sample location and numbers at less
than 10-12 characters is suggested. This is especially true if bar  code  readers are not used.

ACS's experience has shown that unique pieces of information should be tracked separately and
not aggregated together.  With  the power of relational  databases, the data can still  be related
without elaborate naming conventions which encode all pieces of information. As an example, a
sampling location such as SB-12/A/10-15/3-91 may mean the soil boring (SB) was collected from
location number 12 in zone A at a depth of 10-15 feet in sampling event of May 1991. It is very
easy for data entry mistakes to be made in entering  this number into sampling and laboratory
databases. With modern relational databases, there is  no reason for this type of naming. With a
well designed system all the information could easily be tied to Soil Boring number 12. This could
be printed on chain-of-custodies and labels without being encoded.

The recommendation is that the sampling location names or numbers should be short and unique.
Other data should be related  to this point name. Information typically related to the point name
is  the sampling coordinates,  the depth of collection,  complete sampling and analytical data
including analytical data and dates of collection, receipt by all parties. Data from multiple sampling
events can be associated to a single location. Data from each event is designated by sampling and
analysis date and by the laboratory and sampling numbers used in sample identification.

A location and sample number system should be established and documented and used by  all
contractors performing work. Often the most confusion occurs when several organizations collect
samples at the same site over several years. Without an agreed upon location naming convention,
the data is difficult to connect to the correct location. Having this information available to all staff
performing work is important in maintaining accurate and traceable data.

Since  the location name and  sample  numbers are critical pieces of information, it increases
accuracy if these are passed between the field teams and laboratories via computer generated chain
of custody forms. Many projects have suffered major problems because of difficulty in transcribing
data from a hand written chain of custody to a laboratory computer tracking system. If the project
planning information is entered for early tracking, the sample names and numbers, this can  be
easily done. In addition to the forms, bar codes can be placed both on the forms and the bottles.
The bar codes should contain the sample location, sample number, depths, analysis requested and
any other pertinent information.  By putting more information on the code than the  number,
sample login in the laboratory can be quick and accurate. If changes occur while samples are being
collected, information can be hand written on the forms.  (2)
                                             1-190

-------
DATA ENTRY CRITERIA

Data entry is accomplished via manual entry, scanning or electronic transfer. Criteria should be
specified for all areas. It has been ACS's experience that scanning of documents which are not
neatly typed results in 5-10% errors requiring reentry and correction. This is particularly critical
for sampling logs, boring logs and other documents which are typically hand written. If these are
manually entered into the database, the data can be electronically searched and moved to logs and
construction diagrams. This results is typed, legible forms while entering the data only once.

If manual entry is needed, the best method is to double key the information by separate staff with
the computer making a comparison between the entries. If it is not practical for two people  to
enter the information, it should be entered at least twice by the same person followed by computer
comparison of the entries. The computer should print out differences between the entries. These
differences should be resolved prior to moving the data from temporary holding files or tables to
the final database tables.

If electronic transfer is used, typically ASCII files should be transferred. All  parties should agree
on the information to be entered and transferred and its location in the file. Two major issues are
that many organizations performing analysis do not have double key  entry processes nor do they
have electronic download from the instruments to the central database. This is especially true of
laboratories. As a result the data is rekeyed from the instrument output  to the database. This is
why so many differences are normally found between hard copy and electronic transmittals. Data
errors between these two media run from 5-10%. In some cases they are greater.  In  auditing
laboratory information processes, ACS has found as many as four transcriptions of the same data
prior to entering it into a file for electronic transfer.

As a result, ACS recommends that criteria be outlined for data entry of sampling data, laboratory
data, validation of data and other information needed for the permanent records.  These criteria
should include not only the method of transfer of data, but the way the data enters the database
initially. This  means that if a laboratory does not have electronic transfer between instruments,
double key entry may be needed. Trend analysis and other checks may be used to further assure
that data is consistent and correctly reported.

DATA CONSISTENCY FILTERS

Many problems in  using data result in inconsistencies  in the reporting. As an example, ACS
mapped data from a project. The engineering firm had observed a 20ft water level difference at the
site when ground water samples were collected. Prior to ACS involvement, much money had been
spent  on models to explain these inconsistencies. After the original well logs  and  methods of
measuring water level were discussed,  ACS determined through consistency checks that the
problem was inaccurate survey  information.
                                             1-191

-------
The consistency checks are performed via computer and via  data inventory printouts. These
consistency checks. The consistency checks include identifying:

   ' Missing sample locations
   ' Duplicate data and samples
   - Improper parameter names
   - Duplicate Test Names
   - Samples with missing data
   - Data with missing location information
   - Time traveling samples
   - Incorrect location and episode data
   - Illogical Units
   ' Illogical Qualifiers
   - Missing Detection Limits

Any data management process should look for and correct problems in these areas. While software
can help identify these issues, it takes dedicated staff to correct these problems. In order for these
corrections to occur, all the members of the project team must be easily contacted regarding these
problems. It is also helpful if these problems are identified as early as possible in the project. This
allows for correction of inconsistencies prior to use of the data.

When new data is received, a mechanism must be in place  to examine the incoming data and
compare it to the existing data in the central database. The  mechanism should  also look for
inconsistencies similar to those outlined previously in the new incoming data. The information
upload to the central database should allow only the new data to be entered. If the entire database
must  be uploaded, hours may be  wasted in data processing.  In order to facilitate these data
transfers, special utility programs need to reside on the  system  which  maintains the central
database.

DATA TRAFFIC  CONTROL SYSTEM

A major part of managing the data is knowing when deliverables are due and assuring that they
meet schedules and that the data required is delivered. This is particularly critical for sampling and
analytical data. The traffic control system varies depending on the project need. As an example,
for CERCLA/SARA projects the proposed number of samples and QC samples must be compared
to the number actually collected and analyzed. This information is used to alert project managers
of missing information and be able to assist in cost tracking. In the systems used, all the proposed
samples, matrix, methods of analysis, and QC for delivery are entered. The computer compares
incoming data to the proposed data. Printouts of differences  were  available immediately after
receipt of data. These reports must occur quickly. This allows rapid corrective action. Having the
cost information available also allows contractual issues to be quickly resolved.
                                             1-192

-------
For NPDES type of projects, not only is the laboratory data tracking important, but sampling staff
need to be alerted to collect specific samples from outfalls on specific days. In addition, the final
reporting of data should be compared to permit requirements. This comparison should include the
analytes, analyte concentration limits, and date reported. Anything exceeding permit criteria must
be flagged and managers notified immediately.

The  traffic  control system described is important since  typically, there are large amounts of
incomplete data or data which does not meet acceptance criteria. By allowing early notification of
these problems, project schedules can be met. Without using computers, tracking the large number
of samples, sample results, cost and schedule become difficult and tedious for technical staff. The
data  management system must address these needs. The method of handling these issues should
be outlined in the plan.

QC DATA ELEMENTS AND RELATIONSHIPS

Another critical item is defining the  QC samples in the planning document and a  numbering
scheme for  these samples. The critical item which is often missed is  the definition of what the
sample is and how it  is collected. Currently at least 10 definitions  exist for Field blanks and
rinsates. With this many definitions, all parties in the process should understand what the samples
are.

The  other critical information is how the QC samples are related to the actual samples. These
relationships must be captured or the QC data will not be useful. As an example if five trip blanks
were collected and shipped in 5 coolers, one must be able to relate the correct blank with the
correct cooler or to the samples in the cooler. This is typically done via the chain of custody and
the sampling date, time, and person collecting. However, a more successful method is to number
the QC samples and state which samples by number were associated with the QC samples. This
method of association should be included in the definitions in the data dictionary.

Similar issues occur in the laboratory. Many laboratories do not track batch numbers for samples.
Samples may only be related to QC by date. This leads to problems when multiple instruments are
being used for the same task and when several similar sample preparations are performed on the
same day. More problems arise when the laboratory must reanalyze a sample. Reanalysis results in
two or more batches of QC being related to a sample. If the lab system cannot assist in managing
this,  confusion as to which data should be reported may occur. All QC should be available.  This
means that QC should be associated to the samples by unique batch numbers. When reanalysis
occurs both the old and new batch number should be available with a comment on why reanalysis
was  performed. Batch numbers  will also serve to tie the instrument, method, and  person
performing  analysis to the samples. Without knowing which samples  are associated with specific
QC samples, data validation and evaluation becomes difficult.
                                             1-193

-------
SYSTEM DESIGN STRATEGY

The system must be flexible enough to handle all the sampling, project scheduling and analytical
data. The database must be able to manage large and small projects. The entry process must allow
easy addition of parameters, sampling  data points and other  information which  may not be
identified in the early planning stages. The  system must be able to handle numeric and textual
data. A design which has proven to be flexible, thinks of data as three type, sites, locations and
episodes. Site data is the area of study or the site name. This could be a building, facility, or area
name. The next type of information is associated with the location. This information is fixed in
time. It includes information such as soil lithology, sample coordinates (x,y, and z), and the name
of the  sampling location. The last division of data is the  episodic data.  This data  includes
information which changes with time. These data elements include but are  not limited to water
level, parameter names, analytical results, detection limits, units, dates sampled, dates prepared and
analyzed, methods used and other information related to the  sampling events.

Using this design strategy makes data entry flexible and easy  to add information when the need
arises. No matter how much planning occurs there  is always the need to add information during
the project.  Once data is entered in this manner, it can be extracted in many different methods
to allow one to examine data by locations, parameters, methods and other queries appropriate for
data review.

Another issue of design is the method of handling data qualifiers. These qualifiers are used to easily
indicate QC problems, detection  limits and other pieces of information needed to evaluate the
data. In many databases the qualifiers are coded as part of the result. This makes passing data to
modeling programs, and statistical programs difficult. These programs accept only numeric results.
If the qualifiers such as Us, Bs, Js are part of the result field, it is difficult to  extract the data and
move it to  user  programs.  It is  critical that qualifies be placed in separate data fields. The
recommendation is  that reporting and  usage requirements should not drive the data entry. By
allowing the computer to reformat  data rather than have data  entry dictated by the reporting
additions can be made quickly and by the user.

Other issues for data base design include the use of relational  database tools  which can be moved
to various hardware systems without rewriting software. This will allow upscale of hardware as the
database grows in size. It also allows end users to have smaller systems and to easily upload and
download between smaller and larger systems. In addition the ANSI computer standards require
use the Structured Query Language to facilitate data communication between systems. Because
of these features, ACS has had the best success using ORACLE Relational Data Management
Systems with a  design built .on a row based entry system consisting of sites, locations and episodic
data.
                                             1-194

-------
AUDIT TRAILS and ENTRY SERIALS

The computer must be able to track the entry and changing of information. An effective method
is to establish unique computer generated serial numbers for each entry session. The audit trail
should be kept, for both manual and electronic entry. The person performing entry or download,
of the data, the origin of the information, and the date entered must be kept by the computer.
The entry serial number ties  this information together. By performing this task, an audit trail is
kept on the data entry.  Changes to the database should  follow a similar strategy.

CONNECTIVITY REQUIREMENTS

In communicating data from planning to field crews, to the laboratory, and back to the data users,
a path with consistent  methods of transfer must occur.  Without documentation and a well
designed path, data will be transcribed and reentered many times. This results in errors and project
delays. The best strategy is  to outline file structures and information which must  be passed.
Typically ASCII files can be passed between organizations. However, information sometimes is not
in consistent locations within files. A data exchange language which assists in this process is a
useful tool. This exchange language  can parse  files apart even when information is not in
consistent locations.

As mentioned in the Data Consistency Filter section, as data is uploaded checking for consistent
information should be performed. These checks should be available to all parties moving data. By
doing this, early warning for missing and inconsistent information occurs.

The recommendations are to use ASCII files, to agree on  the information to be transferred and
a general format prior to beginning the project, to use a data exchange tool which will assist in
parsing files even if data moves within the file, and to provide upload and download software which
performs consistency checks on the incoming and outgoing data.

TOOLS FOR COMPLIANCE SCREENING/DATA VALIDATION

Contract Compliance Screening is  the systematic verification that the deliverables specified in the
governing contract meet delivery  and general QC  requirements.  It also  includes verifying the
: frequency of QC analysis, and limit checks. Examination of the raw data (chromatograms and
mass spectra) are not included. The process of screening generates both summary and detailed lists
of samples failing criteria with associated notes about how  they failed.

Data Validation is a systematic review of the data which includes QC frequency and limit checks.
It also includes review of the raw data, flagging of data which does not meet criteria and technical
judgement in assessing the information.

The part of screening and validation  amenable to  automation is the QC frequency and limit
checks, and much of the data flagging. Once the initial  flags are applied, some evaluation using
                                            1-195

-------
technical judgement must be done by a qualified technical person to assess the flags. By automating
much of this  process,  star! will be better utilized, screening will be improved, consistency  is
improved, and fester execution will occur. Currently, much of this process is poorly automated.

Software used for this purpose needs to allow the limits to changed to meet the objectives of the
project. It must also generate summary reports  outlining the number  of compliant and  non-
compliant items per sample. Detailed reports outlining specific non-compliant data should also be
presented. Ultimately final data with flags should be presented. This should be similar to an EPA
Contract Laboratory Protocol Form 1 with additional validation flags. The system should be able
to handle textual data so that explanations of flags and non-compliance can be related to the
numerical results.

After the screening/validation is complete, the electronic download to a  central database should
be easy and menu driven. The data from the central database should then be moveable to other
packages such as statistics and mapping packages. This will allow the central repository to maintain
upto date data. The last issue is that an audit trail on the changes should be maintained.

DATA MANAGEMENT STAFF

A key person (s) should be identified in each organization to manage data. This individual must be
a part of all planning from the initial stages. A budget should be set aside for this task. The items
above should be addressed in  the planning and continuing phases of the project. The data
manager (s) should be familiar with the types of data collected in the particular process, they should
have  a desire  to create systems which place data in the users hands, and should be willing to
communicate with the entire project team. This person or organization becomes a focal point to
manage the data traffic.

CONCLUSION

Data  management is an integral part of environmental projects. The  need  to quickly obtain and
assess data continues to  be an  important factor in successful completion of projects.  Multiple
companies and organizations typically work on these projects, the management of the information
is critical. If a central process is established for data management from the beginning, projects can
be successfully completed.  Key elements include computerized chain of custody and tracking,
establishing common terms and meanings, and establishing location naming conventions which are
documented and used by all parties. In order to maintain consistency and integrity, audit trails and
consistency checks are required. To track the information, a traffic control system is needed to
determine the planed samples versus those obtained. Screening and validation must be automated
to prevent data review bottlenecks. The data must be made available to users for decision support
in a timely manner. By considering these issues in the planning and implementation project, high
quality and timely results can be generated.
                                             1-196

-------
                                      References

1. Phyllis Koslow,"Data Dictionaries," Database Programming & Design. April 1991, 26-29.

2. Butterfield, S.; Deinse,  H.V.;  Rumford, Greg; Penalba,  Jorge; "Data Management of a
Multimedia Remedial Investigation Program," Presented at Annual Conference of the Air and
Waste Management Association, Vancouver, British Columbia, June 16-21, 1991.
                                            1-197

-------
PR    A SOFTWARE APPROACH FOR TOTALLY AUTOMATING THE QUALITY
       ASSURANCE PROTOCOL OF THE EPA INORGANIC CONTRACT
       LABORATORY PROGRAM

       Cindv Anderau. Technical  Specialist, Rob Thomas, Marketing Specialist, The Perkin-
       Elmer Corporation, 761 Main Avenue, Norwalk, Connecticut 06859-0219

       ABSTRACT

       The EPA's Inorganic Contract Laboratory Program requires  the determination of 23
       elements in a wide variety of matrices.  ICP-OES  and ICP-MS, because of their
       multielement capabilities, are ideal techniques for this type of  analysis. Although ICP-
       OES has gamed full approval by the EPA, ICP-MS, while rapidly gaining momentum, is
       still not fully approved for all the EPA Inorganic Programs. However, regardless of the
       technique chosen, the quality assurance protocol associated with a CLP analysis is very
       complex and time consuming.  Traditionally this has been a manual operation that can
       involve the rechecking, recalibrating, rerunning or even  rejecting  samples, standards,
       blanks and spikes that fall outside certain specified ranges.

       This paper will describe a software program, which shall be called "QC Expert™", that
       totally automates the operation of both ICP-OES and ICP-MS for this very tedious
       quality control protocol.  This is achieved by controlling both the instrument and the
       autosampler with the software. During an analysis, if the quality of the data is considered
       unacceptable, then pre-established procedures to restore the quality  will be undertaken.
       These will be monitored and then directed by the software.

       The major tasks for establishing the  proper criteria for this type of analysis are divided
       into two separate steps.  The first step involves setting the analytical method (standard
       concentrations, QC limits, predefined actions, etc.)  and the second step involves setting
       up the sample parameters (sample id's, weights,  and volumes, etc.). The logic behind
       this, is that once an analytical method has been set up to perform a C.L.P. analysis, it will
       not drastically change. On  the other hand, the sample information will almost definitely
       change on a regular basis.  For this reason, these two quite different tasks are separated
       and simplified to provide for a quick and easy set up for each analysis.


       INTRODUCTION

       The EPA's Inorganic Contract Laboratory program  has quality  control requirements that
       are very stringent. Laboratories participating in ths program find that these analyses can
       be extremely time-consuming on the analyst's part, and would rather not have an analyst
       spend all of his or her time monitoring the analysis of  samples, to  insure that good
       quality control of the data  is maintained.  Yet, that is  exactly what many laboratories
       must do.  Many instruments still require some user intervention to both ascertain the
       quality of the analysis and to control  it.
                                               1-198

-------
                                  Typical   OC   Protocol
                                      Calibrate
                                                KHtrumact
                                            "
                                      ctttbrttion     cucv«
                                              YES
                                      Run   Duplicate,   Spike
                                      eno   DBUton  Sample*
                                                 YES
                                             • tc.
A typical analysis requiring extensive QC is shown in the flowchart above. The protocol
shown here is very similar to that mandated by the EPA for the Contract Laboratory
Program.  Once the autosampler tray has been loaded with samples and the appropriate
sample IDs and perhaps  weights  and volumes have been entered into  the instrument
software, the analysis is started by establishing a calibration curve. The analyst must be
present at this point to verify that the calibration curve meets the requirements for the
method, for example, the correlation coefficient of the curve must be 0.9995 or better. In
some cases, the correlation coefficients must be hand calculated by the analyst.  If the
correlation coefficients do not meet the  established  limit, an action  from the analyst is
required. This action is likely to be initiation  of a recalibration.

After the calibration curve requirements have  been met, a suite of QC standards that must
be run. The result for each element in each QC standard must be within some established
limits.   If  the  results for an element or  a number of elements fail to meet the
requirements, an action by the analyst again must be  taken. A typical action would be to
recalibrate,  confirm the calibration curve and rerun  the QC standards until all of them
meet the established criteria.  Finally, samples may be run and the  analyst may take a
break from closely monitoring each  result.  However, the analyst may need  to know
when the results for an element are below the instrument detection limit, requiring that
the sample be rerun by a more sensitive technique, or when the results for an element are
above the linear range, requiring that the sample be  diluted and rerun. In both of these
                                       1-199

-------
cases, the results are not importable.  Rather than searching through the data to see where
the results  fall, the analyst may choose to continually check on the results after each
sample is run, to flag the samples that must be rerun.  In either case, after 10 samples
have been  run, the analyst must return to monitor the results  for the next set of QC
standards to  verify the continuing  quality of the results.  As you can  see from this
scenario  the analyst has little time for other tasks and the potential for errors is great.
Just imagine the  task of making sure 23  elements in  10 QC standards are within the
allowable limits.

Automation of this analysis would provide benefits to the analyst and ultimately to the
laboratory.  The greatest benefit would be that the analyst would be free to multi-task and
would not be required to make any decisions during the analysis or to physically take any
action. The analyst could be responsible for the analysis but at the same  time attend to
other challenging tasks and be assured that the quality of the data was being assessed and
controlled by the software.  The potential for human error disappears. The lab would
also be assured of the quality of the data under the close supervision of the software.

SUMMARY

This proposed automation could be met with the QC Expert software which is designed
to provide  intelligent  quality control  for ICP Emission spectrometry and ICP Mass
spectrometry. If the quality of the data is determined to be unacceptable, pre-established
procedures  to restore quality will be undertaken in real-time.

Software Structure

The strategy of this software design was to  intelligently separate the various tasks that are
required  to set up an analysis.   The instrumental parameters such as element mass or
wavelength, integration time, etc. are set up with the normal instrument software. What
we are calling the analytical parameters are established in the QC Expert software. The
analytical parameters  are divided into two parts or files.  One  is called the Analytical
Method File and contains all the information necessary to run the analysis except for the
sample information.   The other file is the Sample Description  file which  includes
information specific to the  individual samples.  The  Analytical Method file and the
Sample Description file are then combined  into an Autosampler Worksheet File, which is
basically a script for the instrument or a log of the analytical sequence. Results obtained
during the analysis may be used later to create a variety of reports and a variety of QC
charts.

Specifically the Analytical Method File contains information on the calibration standards
and the type of calibration algorithm to use. It also contains information  on all the QC
standards that will be run in the analysis and the allowable limits for them. All actions
for out of limit conditions are chosen in this file. The internal standard elements that will
be used are selected here  as well as the allowable drift limits for the internal standard
element intensities. This file also contains information on how often a QC standard will
be run during the analysis.
                                      1-200

-------
The information about the calibration and QC standards are stored in separate files. This
allows for flexibility in the selection of QC standards and calibration standards.  Once a
standard is created it may be used with any number of methods, it is not necessary to re-
enter the standard information in each Analytical Method file.  A standard file  may be
created which contains every element one ever expects to determine.  This file could then
be used with any method, but the software would only use  the information  for  the
elements in that method and ignore the others.

QC Limits

The allowable limits for several measurements are established in the Analytical  Method
file.  Each measurement that a limit has been set for may have an action associated with
it that will be  taken if any  of  the limits are exceeded.  Limits may be set  for  the
following:
              * Correlation coefficients for each element
              * QC standards
              * Intensities for the internal standards
              * Duplicates, spikes and dilutions
              * Elements in the samples

An example of when one might use the sample limit feature of the software, would be to
monitor when the concentrations of the elements in the samples are below the detection
limit or above the linear range.  In both or either case, an action could be selected.  A
message would be printed which might simply  flag the data and tell the user  that  the
sample  must be rerun  for  that particular  element   For example, the  message "Se
concentration is below the Detection Limit, RERUN by GFAA" could appear when the
concentration of Se is below the lower limit. The message "Ca concentration is beyond
the Linear Range, DILUTE and RERUN" could appear when  the concentration of Ca
exceeded the  upper limit.

Actions

An action may be chosen for every measurement for which a limit was set. This would
include: correlation coefficients,  all QC standards, upper sample limit for each element,
lower sample limit for each element, intensity drift for each internal standard element,
duplicates, spikes and dilutions. It is also possible to qualify when an action should take
place.  Rather than have an  action, say recalibration, occur when any element out of a
suite of 20 is out of limits, it is possible  to establish which  elements or how many
elements must be out of limits before an action will occur.  A recalibration takes tune and
it may not be worth the time if only one element is out of limits, especially if it is an
element that may be determined later by a more sensitive technique anyway.

There are nine  available actions  for each measurement for which you have established
limits. Two actions may be selected for each measurement. The second action becomes
the alternative action that will be taken if the first action does not  solve the problem.
This would be analagous to  a  user reruning a QC standard if  it did not meet  the
established limits.  For the QC Expert this would be action 1.  If the QC standard still
                                      1-201

-------
failed, the user would recalibrate and then rerun that standard. This would be action 2
for the QC Expert.

When an  out  of limit condition is detected  by the  software, the  selected actions are
executed.  The actions available are as follows:

              *Stop
              * Continue
              * Recalibrate and Continue
              * Recalibrate and Rerun
              * Wash for X seconds and Rerun
              * Wash for X seconds and Continue
              * Rerun current sample/standard
              * Continue from... (specify sample or standard ID)
              * User Defined Program

User Defined Program

The User Defined Program action allows the user to do virtually anything that a program
can  be  written for.   A User Defined Program is  a  batch  file containing whatever
commands are needed to accomplish the desired task. When this action is executed, the
QC Expert software is temporarily exited and the User Defined Program is run. When
the program is completed, the QC Expert software is automatically returned  to and the
analysis continues. The User Defined Program could turn on or off a switch or turn on a
dilution system or even call you at home.

Every time an  out of limit situation occurs the message in the message column would be
printed on the printout This message could be as long as 80 characters, the full width of
a page.

The  utility of  a User  Defined program as well as the  utility of establishing limits for
elemental concentrations in samples are best illustrated in a process control situation. An
example of  a  process that requires good control  is the discharge of effluent from a
factory.  Depending on  what body of water the effluent is being  discharged into, the
limits for the various heavy metals monitored vary.  In this example, the effluent is being
discharged into a trout stream thus the allowable levels of many heavy metals are very
low.  The  allowable level for the discharge of Pb, mandated by the EPA, is as low as 15
/ig/L in some states. The QC Expert software in conjunction with an ELAN 5000 could
easily be used to both monitor the level  of  Pb in the effluent as well as control the
discharge  of it. The sample limits feature of the software  could be used  to determine
when the concentration of Pb exceeds  15 jtg/L in the effluent When the level  of Pb does
exceed 15 pg/L, the action that the QC Expert would initiate  would be a User Defined
program.  This User  Defined program could control  a switch  that would  divert the
effluent to  a  holding tank where it could  be treated  to precipitate the  Pb before
discharging the effluent into the trout stream.
                                      1-202

-------
Customizing a Method

The QC Expert software has a number of features that allow one to customize a method.
This makes it easy to follow any QC protocol required.  Units are available for the
standards, the weights and volumes for the samples and the final concentrations reported
for the samples. It is possible to use different units for each element.  Several commonly
used units are available in the software and it is possible to teach the  software any other
unit desired.  The  software is smart enough  to know the various  conversion factors
needed to calculate the final sample concentrations in the units specified.

It is possible to report the sample concentrations to  a certain  number of significant
figures or a certain number of decimal places.  Each element may also be reported
differently.  The available calibration curve algorithms are selectable on a per element
basis and there are a number of linear and non-linear choices available.  The samples
have two possible labels, a 15 character batch ID and a 15 character sample ID. It is easy
to copy and increment these IDs.

Once a method has been created and all the sample IDs and weights  have been entered,
the analytical sequence is automatically created by the software  and assembled into an
Autosampler Worksheet.  The software is able to create this worksheet by combining
information on the  autosampler tray layout, the calibration and QC standards to be run
and the sample  information. This autosampler  worksheet is readily edited. It is possible
to print this worksheet according to the analytical  sequence so the user knows how the
analysis will proceed or it is possible to print the worksheet according to autosampler
positions so the user can fill the autosampler tray easily.

Running an Analysis

An example of how an analysis is run automatically by the QC  Expert is shown in the
flowchart below.  The  calibration is  initiated by the software.    The  correlation
coefficients are compared to the minimum acceptable value of  0.9995.  If any of the
correlation  coefficients are  0.9995 or  less,  the  software  directs  the  instrument to
recalibrate for those elements.  However, if a recalibration had already occurred once, the
software will simply  stop the analysis and sound an alarm to alert the user to the
problem.  If all the correlation coefficients are acceptable, the analysis will proceed and
all the QC standards are run.  After each QC is run, the results are be compared to the
acceptable limits.  If the results are unacceptable for more  than five  elements, a
recalibration occurs. If a  recalibration for this QC had already occurred, the analysis is
stopped and an  alarm is sounded to alert the user of the  problem.  If five elements or less
were out of limits, the analysis continues with  the samples.  However, the elements that
were out of limits are no longer reported  to avoid having data on the  printout that is not
useable.
                                      1-203

-------
                            AB DCMnpM «T mm Aaayttt Run tf W* OC DcpMt

                                       -M CMbrtf* ImtninMit
1
YES
F
RunQC$
                                                •te.
After each sample is run, the elemental concentrations are compared to the allowable
upper and lower limits.  If a concentration for a particular element is below the lower
limit, a message is printed, telling the user that the  concentration for that particular
element is below the detection limit and  it has to be run again by a more sensitive
technique.  If the elemental concentration is above the lower limit, it would be compared
to the upper limit to check that it does not exceed this.  If the elemental concentration
does exceed the upper limit, the autosampler would be directed to wash for 20 seconds
and then continue on with the next  sample.   This  would avoid carryover between
samples. A message would also be printed out telling the user that the concentration for
that element exceeded the linear range and the sample would have  to be diluted and
rerun.  After each sample is run, the sample count is checked.  Once 10 samples have
been run, the QC standards are run again as already described. This process continues on
in this fashion with QC standards being run inbetween  batches of 10 samples.  After the
last sample, a set of QC standards are run again with a similar protocol.

Without the automation  of the QC software,  the user would  have been required to
intervene at all of the stages, to make a decision and/or to physically take action.  The
user would also have been required to wait after each intervention to see that the problem
was solved and the analysis could continue. If the problem had not been solved, a second
action would have to be initiated by the user.  Again, the potential for errors allowing
"bad" data to slip through cannot be overstated.  With the QC software, the user has only
                                      1-204

-------
to be present to start the analysis. The user may want to be within ear shot if he wants to
know when the analysis is completed or if the analysis has stopped.

When the analysis is completed, the printout will contain all the information necessary to
track  the analysis.  Time  and date appear  for each sample. When QC standards  and
diluted samples are run, percent recoveries are calculated and printed.  When spiked
samples are run, spike recoveries are calculated and printed.  For duplicate  samples a
percent difference is printed.  All user specified messages will be printed for all out of
limit situations.

If the printout is not what  is needed as a final report, one can use the Report Generator
included which will provide a variety of reporting formats.  It is also possible  to monitor
the quality of the data over time by using the QC Charting mode of the software. In the
QC Charting mode, one can monitor any measurement of any standard and element over
time.  For example, one can plot the concentration for each element in each  QC standard
over time (an  X-bar chart) or the %RSDs for each element in each QC standard over
time.

The QC Expert software totally automates the analysis, allowing the analyst to attend to
other tasks. Actions that would normally be taken by the analyst to assure the quality of
the analysis can be executed by the software.  The laboratory can become very efficient
by having the analyst tend  to other more challenging and important tasks than "watching"
an instrument. The quality of the data is  assured by the software as it  executes the
protocol established by the user.
                                      1-205

-------
29  AUTOMATED REPORTING OF ANALYTICAL RESULTS AND QUALITY  CONTROL
             FOR USEPA ORGANIC AND INORGANIC CLP ANALYSES


     Richard D.  Beatv. President,  TELECATION ASSOCIATES, P.O.   Box
     1118,   Conifer,   Colorado  80433;  and  Leigh  A.   Richardson,
     Laboratory  Consultant,  PEAK 10 SCIENTIFIC,  P.O. Box  278,
     Conifer,  Colorado 80433


     ABSTRACT

     The   U.S.    Environmental   Protection   Agency's  Contract
     Laboratory Program (CLP)  for environmental analyses follows  a
     detailed   protocol  for   analysis,    quality  control,    and
     reporting.     The   formal   reporting   procedure  involves
     submission  of a deliverable data package,  which includes  a
     number of predefined forms and a diskette containing data in
     an  "Agency Standard"  format.   Attention to detail   in   the
     reporting  process is important, both to  assure   verifiable,
     usable  data  for  the Government,   and  to  guarantee  full
     compensation for the contractor laboratories.  Automation of
     the reporting detail is necessary,  if compliant  deliverables
     are to be generated with  regularity.

     The  QC  calculation and  reporting process  lends itself to
     automation  through  the  use of PC-based  programs for  data
     reduction.    This paper will discuss the factors  involved in
     developing and maintaining such programs for both the  Organic
     and Inorganic CLP programs.

     Automation  of the reporting process begins  with electronic
     data acquisition from the analytical instruments  being used.
     Some  beginning  efforts  at standardization  of  data   output
     format  among  instrument manufacturers has resulted   in   the
     ability  to include "standard"  data import routines   in   CLP
     report  generation  software.   For   organic  analyses,    the
     "reduced result"  file has been selected as a standard import
     structure.   For inorganic analyses,  a "comma delimited ASCII"
     import  routine has been  selected,  which is compatible  with
     some manufacturer's AA and ICP instruments.   For  both  organic
     and  inorganic  analyses,   where instrument output does   not
     match one of the "standard"  structures,   special custom  data
     file  import  routines have been devised for  automatic  data
     acquisition.

     After  instrument  data  has  been accumulated  into  the   CLP
     reporting  software,   additional required information  may be
     added by keyboard entry.    Once all  required data is present,
     software routines process data to calculate reported   results
                                   1-206

-------
according  to  EPA  approved  calibration  routines,  compare
results  to  specified limits and  conditions,  flag  results
which  lie outside these limits or otherwise meet  conditions
for  data qualification,  and finally,  generate  the  formal
printed and computer readable reports.

Until recently,  two formats for computer readable data  were
allowed.   These were known as Format A and Format  B.   Over
the  past  year,  the intention to go  to  a  single  "agency
standard"  format was announced.  The specifications for this
format,  as  it  applies to CLP  diskette  deliverables,  was
introduced  earlier this year.   This paper will discuss  the
status of automated CLP reporting software to provide "agency
standard" diskettes for organic and inorganic CLP software.

While the CLP program was devised strictly as a protocol  for
sample  analysis and reporting of data by EPA contractors  to
the EPA,  a trend has emerged for use of CLP,  or at least  a
"CLP-like"   format,   as  a  reporting  standard  for  other
environmental analyses.   In such cases,  it may be desirable
to  modify  some  of the details of the  report,  to  achieve
modified  goals.   The  modiflability of  the  CLP  reporting
software   to  address  special  reporting  needs   will   be
discussed.
INTRODUCTION

Telecation Associates  began  as  a consulting company providing
on-site training,  instrument setup  and method development for
analytical laboratories.   In 1985,  while helping a laboratory
setup  for compliance  with the  Contract Laboratory  Program1 s
inorganic  Statement  of Work 7/85, we became   aware  of  the
detailed  calculations and comparisons required to produce  a
CLP  printed   data  package.   It was quite  obvious  that  a
PC-based   software   program  designed   to   perform   the
calculations   and   the logic required to fill   in  the  forms
would  provide  an automated means  of  generating  compliant
deliverables  with  both regularity and considerable savings  of
time.

Somewhat  later a  new  inorganic Statement  of Work,  SOW  7/87
was  released.   This contract required the laboratory to  not
only  prepare  the  printed  set of  forms for  each  group   of
samples,  but also submit  the package data on a DOS  readable
diskette,  in  accordance  with one  of   two   very  specific
computer  formats.   Computer  automated reporting of analytical
results for CLP analyses had now become a  requirement.
                             1-207

-------
In  October  of  1987,  Telecation  introduced  a  commercial
software product which automated the QC calculations and  the
report and diskette generation for Inorganic CLP,  SOW  7/87.
Since then we have continued providing CLP reporting software
to comply with subsequent contracts under Statements of  Work
7/88 and 3/90, in addition to numerous official revisions and
"interpretation updates" issued by EPA for each contract.

In  the spring of 1990, Telecation expanded its CLP  software
product line to include the new Organic CLP contract for  SOW
3/90.    In  May  1990,  Telecation  began  shipping  QC  and
reporting software for all three organic protocols, including
volatiles,   semi-volatiles,   and  pesticides.    From   the
viewpoint  offered by our extensive background in  developing
software for both inorganic and organic CLP,  we will discuss
the  technical  and  economic factors which  prevail  in  the
automation  of the QC and reporting requirements  of  USEPA's
CLP  program.    We  will  also  identify  opportunities   to
substantially enhance the benefits of automation for CLP-like
applications, where the restrictions imposed by USEPA may not
be a factor.
REVIEW OF THE REQUIREMENTS

Before examining the factors affecting the development of CLP
reporting software, we will first review some of the elements
of  CLP,  which the software must  address.   The  analytical
protocols  for  CLP  describe the analysis of  set  lists  of
analytical  parameters.  In  addition  to  the  actual  field
samples,  a  series of QC samples must be analyzed.   The  QC
samples  include such things as blanks,  duplicates,  spikes,
matrix  spike  duplicates,   control  samples,   and  various
instrument  performance checks.   The analytical results  and
other  sample  related details,  as well as  the  QC,  sample
preparation,  instrument calibration and performance  checks,
are  summarized  on  printed forms, the format  of  which  is
precisely defined for each reporting protocol.   There are 14
different forms plus a cover page required for the  inorganic
data package, and 36 different forms for organics.

The CLP reporting requirements do not allow a simple transfer
of  information to a fixed format form.   All raw  analytical
results  must  first be compared to  both  actual  instrument
detection limits and to the reporting limits required by  the
contract to determine which of the values is to be  reported.
Each form has different requirements regarding the  reporting
of  data  corrected  for  sample  preparation  factors,   the
reporting  of  significant  figures,  decimals,   and  values
rounded   according   to  the  EPA  rounding   rules.    Data
                             1-208

-------
qualifiers, or "flags", summarize the analytical performance.
And,  in  some  cases the forms interact  with  one  another,
inasmuch  as the results or the flags appearing on  one  form
may  also have to appear on another form in place of,  or  in
addition to,  information which would normally appear  there.
And,  last but definitely not least,  these results must also
be  submitted on a computer readable diskette according to  a
very fixed format,  which is now entirely different than  the
format of the data appearing on any form.

In addition to the demands imposed by the contract details of
the CLP statements of work,  the practical considerations  in
making the software versatile enough to address modified  CLP
needs  must  be  considered.  When  the  Contract  Laboratory
Program  was  established,  it was originally intended  as  a
standardized   protocol   for  analysis  and   reporting   by
contractor  laboratories  who worked under  contract  to  the
USEPA.   Soon,  however, the quality control standards set by
the program began to be applied to other environmental  work,
as well.

Since legal decisions rest heavily on established  precedent,
the  analytical  and reporting protocols of CLP,  which  were
developed  specifically to provide legally  defensible  data,
were  accepted  as  a defacto  standard  for  such  analyses.
Because laboratories of all types are now being held  legally
accountable for the data they release,  CLP-type reports  are
being  requested by state environmental agencies,  industrial
accounts,  environmental  engineers,  and  practically  every
facility   that   is  submitting  data  for   any   type   of
environmental  monitoring,  especially if they are  concerned
that the data may become involved in litigation.

The  emergence  of other users of CLP  reports,  besides  the
USEPA,  has created a demand for a software product which can
automate  the  generation of a  CLP-like  reporting  package,
which  may be substantially the same as strictly  interpreted
CLP,  but  modified  in certain  details.   Even  the  USEPA,
themselves,  has issued special bid requests,  which  deviate
from  the  "Routine Analytical  Services"  contracts.   These
include   protocols  for  analysis  of   "Low   Concentration
Organics", "High Concentration Inorganics", and others.

The  market  place for CLP software is small to  begin  with,
relative to the development effort required to implement  the
complex requirements.  These special purpose modifications to
standard  CLP serve to reduce the market size  even  further.
In the case of EPA special bids,  the entire market size  may
consist  of only two or three laboratories.   Therefore,  the
only  feasible way to address the many variations of CLP  and
                            1-209

-------
"CLP-like"  applications,  is to  design  flexible  software,
which  allows  the  user to modify  certain  aspects  of  the
software's  performance.   The features to  be  modified  may
include  the  list of analytes or compounds to  be  reported,
variations  in quantitation limits (CRDL's and  CRQL's),  and
the format or file structure in which data is to be  recorded
on  the printed forms and/or the diskette  deliverable.  Some
applications  may not require all of the printed  forms,  and
others may require a modified file structure for the diskette
deliverable.   Finally,  even the  actual  analysis  details,
including  the  nature  of  quality  control  and  associated
calculations to be performed, may be different from USEPA CLP
specifications.

To  create  a software package which  automates  the  complex
requirements  of  CLP  would  be,  in  itself,   a  difficult
challenge.  To also make it flexible to address any number of
modified specifics,  while keeping the final product easy  to
use,  becomes an almost impossible task.  Combining the above
goals with the ever-changing nature of the USEPA requirements
for   CLP  makes  an  optimum  software  solution   for   all
applications extremely illusive.

In  spite  of  the  challenge  presented  above,   Telecation
provides  a  solution  in  its   "ENVIROFORMS/Inorganic"   and
"ENVIROFORMS/Organic"    software   products,    which   have
historically provided notably accurate compliance with strict
CLP  standards,  while providing the flexibility required  to
allow  the  user  to address  most  "CLP-like" 'applications,
through  the use of software utilities for modifying  output.
No  special  knowledge of a programming language  or  use  of
optional  compilers is required  to implement modified  output
through the use these utilities.


BARRIERS TO AN OPTIMUM SOFTWARE  SOLUTION; CONTINUAL CHANGES

Working with the EPA's Contract  Laboratory Program can  cause
both  the  contract  laboratories  and  the  software  vendor
considerable  frustrations.   Not  only  do  these  contracts
contain a profusion of technical and bureaucratic detail, but
also these details are in a constant state of flux.   To  the
software vendor,  once a contract is issued, the program code
necessary to implement the terms of the contract is  written,
tested,  and  documented.   But  even  as  samples  are  being
released for analysis,  revisions to the protocol are issued.
Sometimes the official revisions note errors in  calculations
or documentation in the original contract.  Other times these
amendments   are  actually   "interpretations"   intended   to
clarify  an ambiguous point,  or in cases where the  contract
                             1-210

-------
contradicts itself,  indicate which statement should prevail.
Inasmuch  as  the CLP approach is now several years  old,  it
might be expected that the number and frequency of  revisions
would  be diminishing.   But in fact,  this has not been  the
case.    Wholesale   changes  in  analytical   protocol,   QC
calculations,  and diskette file structure have led to a  new
round  of  contract problems,  the fix  to  which  frequently
creates more problems.

Each  time  revisions  are issued by the  EPA,  the  software
vendor  is required to make the corresponding changes to  the
software,  if  the product is to remain  viable.   Since  the
nature of CLP software is so interactive,  even the  smallest
change to the software in one area requires thorough  testing
to detect possible problems in other areas,  which occur as a
result  of  the change.  This process  ultimately  identifies
"bugs", which require further changes followed by a new round
of testing.  Then, the documentation for the software must be
updated to reflect these latest changes.  It should be noted,
that  revisions which the EPA might consider "minor"  may  in
fact  have a major effect on software.  The code  is  written
around  all of the details specified in the contract,  and  a
"minor"  change in procedure may have a devastating impact on
the logic used for program development.

Besides  testing  and documenting the updated  software,  the
vendor  must  also  consider the  effect  of  updating  their
current users, considering among other things, how installing
the changes will  affect  "work in progress"  at the laboratory
site.  Lastly,  all new  software and  documentation  must  be
distributed to the users.

The instability of CLP contract terms,  therefore,  sets into
motion  a series  of events,  which can only lead  to  further
changes.   Each contract change which necessitates a software
change,  requires that  new logic  and  code  be  developed,
followed  by  performance  testing,  software  bug  fixes  and
retesting,    and  issuance   of   updated   software    and
documentation.  This  is  usually followed by identification of
contract  flaws,  which  require new  contract  modifications,
which sets the whole  process in motion again.  Meanwhile,  as
this   moving  target  progresses  through  a  never    ending
evolution, the contract  laboratories and software vendors are
expected to  "conform"  to what ever the latest interpretation
happens to be.  Any deviations from  this volatile  "standard"
result   in  compensation  penalties  exacted   against  the
submitting laboratory.

Because  of  the continuous EPA revisions to the  requirements
for CLP analysis  and  reporting, most software vendors offer  a
                             1-211

-------
type  of software maintenance contract designed to  keep  the
end  user  up  to date with the latest  EPA  revision.   Some
laboratories view the price of these "Maintenance  Contracts"
as  expensive,  and indeed they are relative  to  maintenance
contract prices for most software.   But the demands outlined
above are not a part of most software.   The software  vendor
who conscientiously strives to update its clients in a timely
manner  after being notified of yet another change,  must  be
prepared to divert development,  testing,  and  documentation
resources  from other projects to CLP projects on a  moment's
notice.   If commercial software for CLP work is to  continue
to be available,  the software vendor must be compensated for
the  never-ending  development  and  untimely  diversion   of
resources.   In fact, the software vendors, like the contract
laboratories,  themselves,  realize nominal profit margin  on
CLP   work,   relative  to  other,   less  demanding   tasks.
Laboratories,  who  bear the cost for software  and  software
maintenance, should understand what is behind that cost.
BARRIERS TO AN OPTIMUM SOFTWARE SOLUTION; OTHER

The  goal of CLP software is to automate QC calculations  and
reporting in as expeditious a manner as possible. There are a
number  of  details  of the CLP  contract  which  impede  the
ultimate  achievement of this goal.   The inorganic  contract
SOW  7/85,   which  was  produced  prior  to  the  need   for
computerization,   included   some  features   which   defied
computerization.   One  such detail was  the  requirement  to
"circle"  preprinted options on the forms.   Since  computers
cannot  "circle"  choices,  this was changed for SOW 7/87  to
take advantage of the benefits which computer technology  had
to offer.  However, at least one requirement, which is almost
equally  incompatible with computer technology  still  exists
today, the EPA rounding rule.

The  EPA rounding rules call for the evaluation of  numerical
data in a way which is foreign to computers.   Computers  can
round on their own very nicely,  but the rounding required by
the   Contract  Laboratory  Program  calls  for   a   special
evaluation routine to be performed on each and every value.

Since the rounding rule is nothing new,  the software routine
has   been  developed  a  long  time  ago  to   perform   the
calculation.   But this does not mean that the rounding  rule
does not still impact CLP software in a negative way.   First
of  all,   the  necessity  to  round  by  a  special  routine
frequently  complicates  the  implementation  of  other   EPA
imposed changes.   Further,  the ever present requirement  to
evaluate  every number by a special rounding routine  impacts
                             1-212

-------
the speed of software operation.   When it is realized that a
typical sample delivery group contains thousands, if not tens
of thousands of calculated numbers, the special rounding rule
is  responsible  for  a  significant  increase  in   software
operation time.

As to the value of the rule, it has been stated that in order
to  make  the data legally defensible,  it  is  necessary  to
precisely  define the rounding conditions.   This appears  to
us,  however,  to be a flawed argument,  since it  is  highly
unlikely that the outcome of a legal action would ever  hinge
on  the  difference  in rounding of a  number  in  the  least
significant digit.   A competent attorney would see to  that.
The  USEPA  has  deemed this rule to be  valuable  for  their
purposes,  and  hence,  it is by definition  a  part  of  all
software intended for generating data for EPA use.   For  all
other  applications,  the EPA rounding rule  serves  only  to
reduce  the  automation  benefit  which  would  otherwise  be
realizable in a 1990's computer technology.

A  new  impediment  to  efficient  compliance  with  the  CLP
protocols is the "Agency Standard" diskette format.  Prior to
the  more  recent contract terms, data could be  recorded  on
diskette in one of two formats. Format A or Format B.  Format
A  has been the choice of most software vendors,  since  that
format  closely  matches  the printed form,  and  a  software
routine  which  is set up to print the hard copy  forms  can,
without  great  difficulty,  also produce the Format  A  file
structure.  By contrast, the "Agency Standard" format,  which
does  not follow the forms, and in fact contains  information
which is not contained on any form, requires all new software
code,  involving additional data manipulation, to produce the
diskette.   This incremental software development will impact
the  cost  of  future  software  and  maintenance  contracts.
Software   users  should  also  expect   increased   software
operation time,  to complete the additional data manipulation
required to create the Agency Standard diskette.

A  recent occurrence has added a new element to  the  already
uncertain  nature of CLP contract details.   This  occurrence
attacks  at the heart of every software  vendor's  confidence
that after expending the time and resources described earlier
to  develop  a product to address a new CLP  contract,  there
will be a market for that product.

When  a  new CLP contract is issued the  required  dates  for
complete  compliance  are also stipulated.   At  a  CLP  data
management  caucus   held in Raleigh,  North Carolina  in  the
spring of 1990, the  software vendors were asked  if they would
be able to respond to the proposed Statement of Work 3/90 for
                             1-213

-------
volatiles, semi-volatiles, and pesticides by a date specified
by  EPA.   It was explained to the vendors that it  would  be
important  for software to be available at that time for  use
with  Performance Evaluation samples distributed in  response
to  an  Invitation  For Bid appearing  in  Commerce  Business
Daily.   Because of this and prior commitments made to EPA to
meet the required schedules,  Telecation redirected resources
to address the new Statement of Work, and ENVIROFORMS/Organic
SOW 3/90 was ready to ship in May, 1990.  However,  EPA later
recanted   their  previously  announced   schedule,   thereby
eliminating all need for the product.

The  impact  on  software cost emanating  from  the  frequent
changes  imposed by EPA has already been discussed.   Issuing
modification  instructions and asking for  crash  development
efforts,   only  to  have  the  demand  removed   after   the
development effort has been expended,  compounds the economic
impact  of  the  changes.  The  development  effort  expended
toward a protocol which is never, in fact, implemented has to
be  absorbed in the cost of future CLP software products  and
maintenance contracts.   It should be noted that the software
vendors  affected  most  by this failure  of  EPA  to  follow
through with its announced contracts,  are those which strive
the hardest to provide prompt updates to the changes.  Such a
practice  can  only serve to discourage the  software  vendor
from reacting quickly to future USEPA changes.
OPPORTUNITIES IN CLP-LIKE PROTOCOLS

As  discussed  earlier,  there is  an  increasing  number  of
laboratories  who wish to report data in a format  resembling
CLP,   but   differing   in  certain   details.    For   such
applications,  the limitations discussed above may not apply.
It is not the authors' intention to second guess the USEPA on
the necessity for the details contained in its contracts, nor
to question the need to frequently change those details.   It
is  our purpose to identify the stumbling blocks which  stand
in the way of automating the good analytical protocol of CLP,
and  discuss  the  benefits that would  accrue  to  non-USEPA
applications, if these stumbling blocks were removed.

As  discussed earlier,  non-USEPA applications have  need  to
change  various  performance and output  features,  including
such  things  as the list of target  analytes  or  compounds,
quantitation  limits,   and  report  and  diskette   formats.
ENVIROFORMS/Inorganic  and  ENVIROFORMS/Organic  offer   this
flexibility to the user.   Further, the user may select  less
than the total CLP form set for printing. The items mentioned
above are modifiable by the user without programming, because
                             1-214

-------
the  items are controlled by user accessible data  bases,   as
opposed to program code.

Some  modifications  for  which we  have  received  requests,
involve  changes  to the analytical and  QC  procedures  from
standard CLP protocols.  Since much of the program logic  and
calculations  are  based on the standard CLP  procedures  and
calculations,  applications  which require a  departure  from
standard  CLP procedures and calculations are more  difficult
or,  in some cases,  impossible to change without changes  in
actual program code.

One such application was recently introduced by EPA,  itself.
The  statement  of work for  "Low  Concentration  Inorganics"
introduced a new technique (ICP/MS), added several additional
forms, and reported detail which was not previously required.
Since  these and other changes struck down the program  logic
assumed  for  standard  CLP,  it was impossible  to  use  the
standard   CLP   software  to   address   Low   Concentration
Inorganics.   Similarly, certain state programs,  which model
themselves  after  CLP  but  alter  certain  critical   logic
elements,  destroy the possibility of using off-the-shelf CLP
software to totally automate the response.

A potential solution exists,  which  would add the ability to
change many procedural issues,  in addition to the  currently
modifiable characteristics of the software.   This  solution,
which  is currently under investigation by Telecation,  would
make the software operation largely independent of analytical
procedure  and would put the QC calculation formulas  in  the
hands  of the user.   Such a software product  would  benefit
both the contract laboratory and contractor agency, alike, in
that it would now be technically and economically feasible to
address special contracts,  with differing analytical and  QC
requirements.   If it is easy enough for the user and/or  the
vendor  to make the changes necessary to address the  special
requirements,  then  such changes can be  economically  made,
even for low volume work.

Whether  or not such a concept is workable rests  heavily  on
the  flexibility of the contractor agency regarding  some  of
the issues which have complicated CLP in the past.   In order
to  provide the degree of user control necessary to make  the
concept  work,  the  software  code  will  have  to  be  less
regimented.  This means that some of the issues which are now
handled  by  the regimented code,  must be  sacrificed.   The
issues which cause particular problems with this concept are:
(1)  the  EPA  rounding rule, and   (2)  the  Agency  Standard
diskette.
                             1-215

-------
The  rounding  rule,   as  discussed  earlier,   offers  very
questionable benefit,  either from a standpoint of scientific
significance or legal defensibility.   However,  its presence
would  require  the  software to  execute  specific  rounding
routines,  which in turn,  would require the software to have
predefined  knowledge of the calculations,  which  eliminates
the desired goal of user-definable formulas.

The  Agency  Standard diskette format introduces  a  new  and
formidable  barrier  to  development of  a  generic  software
package for CLP-like work.   The format,  which includes such
things as special data delimiters, hexadecimal calculation of
checksums,  and unique file formats precludes any possibility
of  user configuration.   On the other hand,  a  simple  file
format  based  around  a  comma-delimited  ASCII  output   of
information contained on the forms, would make it possible to
generate a diskette under user control,  by allowing the user
to  simply indicate the sequence of fields to be  written  to
the diskette file.

It  is  our understanding that the Agency Standard  has  been
decreed  as Government policy even to those  responsible  for
generation of CLP contract terms.   Therefore,  it is assumed
that  Agency  Standard  is a nonnegotiable  point  for  USEPA
applications.   This  simply means that software  to  address
special  low  volume USEPA contracts will probably  never  be
commercially available.  Other contractors, who are not under
this constraint,  should consider the complications of Agency
Standard  very carefully,  before adopting this approach  and
promulgating  more  barriers to efficient automation  of  CLP
type work.

SUMMARY

The  technical and non-technical complications  discussed  in
this paper do not necessarily go hand in hand with the  goals
of CLP analyses.   For those who have need to develop a  good
regimented QC program accompanied by a standardized reporting
format,  we  would encourage adopting  the  sound  analytical
features of CLP, while excluding the detail which complicates
the  process without return of a corresponding benefit.   For
state  agencies which have not finalized their  requirements,
we   would   encourage  allowing  flexibility,   where   such
flexibility  does not compromise the quality of the  data  or
quality  control.   We would advise against requirements  for
special  rounding rules,  and suggest a less  rigid  diskette
format based around the printed forms.  Such flexibility will
not only reduce the cost of both analysis and automation,  it
will  also lead to a more reliable deliverable,  due  to  the
reduced complexity.
                             1-216

-------
Telecation  has been dedicated to providing  state-of-the-art
software  to  address  CLP  applications,   since  the  first
diskette  deliverable  requirement for  Inorganic  CLP.   The
current ENVIROFORMS/Inorganic and ENVIROFORMS/Organic address
detailed compliance to the USEPA's contract terms.   We  have
provided   timely   updates  to  EPA   issued   changes,   by
conscientiously  applying  development resources to  the  CLP
challenge.   As  an  historical  supplier  of  CLP  software,
Telecation  intends to continue its past policy of  providing
the  detailed  compliance and prompt  updates  necessary  for
USEPA   contract  laboratories,   consistent  with  what   is
technically possible and economically viable.

We are,  however, not content with the limitations imposed by
USEPA   details   and  the  effect  they  have   on   non-EPA
applications for CLP-type analyses.  We will, therefore, also
explore new approaches to CLP-like software, which retain the
analytical benefit of CLP,  without the cost and  operational
complexity  burdens which accompany USEPA  CLP  requirements.
Since USEPA applications will undoubtedly continue to present
obstacles  to automation,  future developments will  probably
involve  a  CLP  software product  line,  consisting  of  two
separate product categories,  one for USEPA contracts in  its
strictest   detail,   and  one  for  other,   more   flexible
applications.  This allows strict compliance to be maintained
for  our EPA contract laboratories,  while offering  expanded
capability   and  flexibility  for  those  who  are  not   so
constrained.

For  such  a  segregated software  product  line,  the  costs
associated with the changes and complexity imposed by the EPA
could  be applied to the price of the software  intended  for
USEPA  applications.   The more flexible CLP products,  would
thereby be less expensive to laboratories,  since they  would
not  carry the cost liability of the complexity and  frequent
changes imposed by the EPA program.   This would,  of course,
make the software for USEPA applications much more  expensive
than it is currently,  since there would be a smaller  market
over which to amortize the development expense.  On the other
hand, such a segregation of product application would put the
cost  where it really belongs, without penalty to  all  those
other applications, which are not so constrained.

The  authors would again like to clarify that the purpose  of
this  paper is not to provide an evaluation of the  necessity
and value of the items discussed to the USEPA's program.   It
is  impossible for us to know all of the factors  behind  the
terms  of USEPA contracts.   We are,  however,  qualified  to
identify  the effects of these terms on software  performance
and  flexibility,  and  evaluate the  enhanced  features  and
                             1-217

-------
benefits   which  could  be  made  available   for   CLP-like
applications,  where the constraints of the EPA  program  are
not a factor.
                              1-218

-------
  A CUSTOMIZABLE GRAPHICAL USER-FRIENDLY DATABASE FOR
                        GC/MS QUALITY CONTROL

Peter Chong, Programmer Analytical Systems, John Hicks, Manager Analytical Systems,
John Janowski. Supervisor Mass Spectroscopy, Chris Pochowicz Technical Writer
Analytical Systems, Chemical Waste Management, Inc. 150 West 137th Street, Riverdale,
IL 60627; Gene Klesta, Director Quality Assurance Programs, Chemical Waste
Management, Inc. 4300 West 123rd Street, Alsip, IL 60658.

Abstract

Standard methods for GC/MS analysis require a significant amount of data gathering
operations to demonstrate good quality control (QC) practices. Both volatile and semi-
volatile analysis require the use of surrogates, internal standards, matrix spikes and matrix
spike duplicates. Laboratories are required by the corporate quality assurance (QA)
program to control the analysis within fixed method specific criteria. Additionally,
laboratories are required to calculate their own control limits.

At Chemical Waste Management (CWM), we developed a customized application to
facilitate the collection, analysis and viewing of GC/MS quality control (QC) data. The
Customizable Graphical User-Friendly Database for GC/MS Quality Control was designed
according to the guidelines specified in EPA SW-846 Methods 8240 and 8270.
Compounds for the QC section use the recommended surrogate standards from the two
methods. Ranges for recovery data are taken from Table 8 of Methods 8240/8270, which
contains multi-laboratory performance-based limits for soil and aqueous samples.
Compounds selected for the duplicate and fortification analysis section were selected from
the recommended compounds listed in EPA SW-846 Method 3500. Ranges for the matrix
spike and matrix spike duplicate compounds were selected from the forms following
Chapter 1 in SW-846, which correspond to the CLP limits for these compounds. In
addition to meeting standard QC practices for collecting, analyzing and viewing GC/MS
QC data, this application allows for the creation of acceptance limits in matrices other than
those matrices allowed by the methods specified within the database.

Surrogate recovery data are compared to these fixed criteria and are used to calculate the
laboratory specific criteria. Relative percent difference and percent recovery data are
accumulated and compared to the fixed acceptance and laboratory generated criteria. QC
data can be viewed, edited and graphically displayed based upon a wide range of selection
criteria (e.g., sample id, matrix, dates, analyst, instrument, etc.). Use of this application
has already significantly increased the efficiency and effectiveness of QC practices within
the company.

Introduction

Analytical methods promulgated by the USEPA which use gas chromatograph/ mass
spectrometer instruments require a considerable amount of quality control practices. The
complexity of the matrices and the sensitivity of the instrumentation and volume of data
combine to create a situation which requires substantial use of data evaluation to show the
accuracy and precision of the analytical process. In the methods covering volatile and semi-
volatile organics, a large number of compounds of interest are included, and the number of
surrogate compounds and matrix spike compounds are significant. Moreover, the methods
require specific trend analysis procedures and specific data evaluation techniques to
demonstrate acceptable performance.
                                         1-219

-------
                                                                           Page2

A laboratory using one of the GC/MS methods must calculate the average recoveries for all
surrogate compounds added to the calibration standards, blanks, and samples. The
laboratory must compare its variability to the allowable variability in the method. The lab's
own entries should be developed and used for control from this accumulated data. The
mean percent recoveries and standard deviations need to be calculated and reviewed in a
timely manner. Additionally, the CWM QC Policy requires that the laboratories performing
GC/MS analysis report pertinent quality control data to the central quality assurance unit.
Quality Control policy also mandates that each VOA and Semi-VOA assay act as its own
QC sample with spike duplicates required every 20th sample.

A fully integrated software package was determined to be the best way to accomplish these
tasks, but no such package was commercially available. The GCMS Application, described
in this paper, was developed to fulfill these quality control requirements.

Specification

The specification for GCMS evolved over time as the developers interacted with members
of the Quality Assurance group and the laboratory. Quality Assurance required flexible
access to GCMS QC data to assure the quality of sample analysis and a means to compare
GC/MS quality across the corporation. The lab required flexibility in data input, reporting,
and graphing, plus the ability to isolate potential quality control problems of instruments,
analysts, or matrix inferences. Other requirements included the ability to selectively query
and graph surrogate data, compare our surrogate recoveries to regulatory limits, and the
capability to distribute the application inexpensively throughout the CWM laboratory
system.

Base Software

GCMS was developed in Clarion®, a comprehensive and flexible database development
language. Clarion® satisfied our database requirements and included built-in graphing tools
to allow for visual display of surrogate recoveries. Clarion® applications are compilable,
which facilitated distribution throughout the corporation.

Features

From the specifications described above, an initial prototype was built and demonstrated to
the users for immediate feedback. This process continued through several iterations until
the final application emerged. Ultimately, a "top down" versatile and responsive
screen/menu application was produced. The resulting GCMS application is a feature-rich
quality assurance quality control tool. Major features include:

       Customizable Database

       GCMS allows all CWM labs to customize the application to local site requirements.
       Laboratories incorporate instruments and GCMS personnel specific to their site,
       enabling each laboratory to isolate quality performance and quality trends before
       they become significant issues. GCMS also acts as a repository for regulatory
       surrogate recovery limits and allows for creation of new acceptance limits in
       matrices other than those listed in the regulations. GCMS allows for the
       accumulation of QC data over a 5 year period with appropriate control limits for
       each year.
                                      1-220

-------
                                                                     PageS

Data Input

Results are entered by selecting the appropriate method and pressing the [Insert]
key. Like data (analyst, instrument, matrix, etc.) are copied into the new record. As
the analyst enters QC data, the results are automatically compared to the pertinent
limits for that method and matrix. Out of control results are immediately flagged.
GCMS can also be shared on a network, allowing simultaneous access by analysts,
management and quality control staff.

Data Queries and Reports

GCMS has summary screens which show the quality performance for volatile and
semi-volatile analyses. The summary screens are queried for all data found in a
specific date range. The search can optionally be qualified for a particular analyst,
instrument, matrix or compound.  The summary screen enables the reviewer a quick
view of the quality performance for a range of samples. The screen summarizes, in
an easily read format, the sample  identification, the surrogates analyzed, the
acceptability of the surrogate (e.g., by signifying if the result is in or out of limits),
and a summary of the total performance of the assay (i.e., by listing the total
number of surrogates out of control). Since GCMS allows for qualified searches,
the data can be reviewed to identify specific quality problems in the lab.

GCMS is capable of generating two types of reports. The first report is a listing of
all surrogates and spikes entered into the application (Figure 1). The report looks
similar to the CLP format for reporting surrogate results and can be printed using all
of the filters used for the summary screens. The other report present in GCMS is
the QA/QC report which lists all surrogate parameters and gives a total of the
number of analyses entered, the number of samples, and QC calculations that
include: percent of analyses that are within QC limits, the upper and lower limits
along with a mean and die co-efficient of variation. The mean percent recovery,
mean percent error, along with the standard deviations for both spikes and
duplicates are also reported. This  report is placed in a DOS file and can be printed
or copied on disk and sent to the Quality Assurance unit.
                                       1-221

-------
                                                                       Page4
                         Chemical Want* Management,  Inc.
                   GC/MS QC Report for VOA surrogate Recovery
  Date Analyzed: 11/30/90 TO 11/30/90
  Instrument ID: *
   waste Matrix: AQU - AQUEOUS
   Analyst Name: *
              JAIDlsampla-IDI  DBF I 12D I TO8 I  BFB |T-OUT|D|
                                Report Summary
KH1
PD1
PD1
PD1
PD1
KH1
PD1

34052 TC
34147 ZHE
34148 ZHE
34196 ZHE
34197
BLANK VOA
BLANK VOA

98
110
109
119*
119*
100
96

104
113
95
117*
119*
94
100

102
93
93
102
102
92
103

101
88
86
96
91
99
96

0
0
0
2
2
0
0

N
N
N
N
N
N
N

          Glossary
      AID: Analyst ID
    T-OUT: Total t OUT of QC Limits
        D: Surrogates Diluted OUT
        *: Values outside Req. QC Limits
       -l: Represents Null Field/Data
     Surrogates
                       QC Limits
                        Low Hgh
DBF:  Dibromofluoromethane   86-118
12D:  l,2-Dichloroethane-d4  76-114
TO8:  Toluene-d8            86-110
BFB:  Bromofluorobenzene    86-115
                         Figure 1:QC Report
Graphing
One of the more useful features of GCMS is its graphical capabilities. GCMS uses
two graph formats, a QC graph and fortification/duplicate graph. The data to be
graphed can be selected by range of dates, surrogate, matrix, instrument or analyst
The QC graph plots the individual surrogate recoveries, the upper and lower
method and laboratory limits by method/matrix, and the mean of the surrogate
analysis. The duplicate/fortification graph plots the surrogate recoveries for each
duplicate, the relative percent difference of the duplicate recoveries and the method
limits (Figure 2). GCMS plots 110 individual graphs (one plot per surrogate, per
matrix). The lab limits displayed are those calculated from the beginning of the year
to date.

Each graph summarizes the limit information in the upper right hand comer of the
display. If users require more information about a datapoint, they click on the point
of interest using a mouse, and the details of the analysis are displayed in the upper
left hand comer. If users require a hard copy of the graph, they click on the print
icon to send the graph to a local printer.
                                    1-222

-------
                                                                           PageS
DOR QC GrapK for 12D:flQU HU.II« LU.IZS it
1I/Z7/J* Chmical Uasta ttarv»go»ont. Inc. HL'7« LL>77 3Z
D«t«l! 34iI3 ZHE 1 HP2 i flUU 1 KHI ! H i 124 n N • 1 8 1 3 2



p l7*:
• lse;

• , - -I

n ie*
< -



c 88^
° 7»I
U -
• 6*I


















|




















A
3









83 1









9
,'•
n
^ *""









d 1
N









y*
I
f> /
^y

















y^
>. E
v t
\
v
1
















>« ^A
\ / \
s




















A \
4 ,
1



















, a
T T
/ V




















^ '
S




















•-a a
Sj
y



















^
3
V :


















52
; •,
/ \
V










- Z8 23 38 33 41 <5 SB 35 68
u • b • r of S • • p 1 • i
                                Figure 2: QC Graph
Benefits

The following examples demonstrate how the GCMS Application makes quality problem-
solving a quick and relatively painless operation:

       Approximately 25% of the volatile surrogates in a laboratory were found to be
       running bias-high upon analysis. The problem was present on the two instruments
       used to run the volatile analyses. When the surrogate data was filtered through
       GCMS, a pattern emerged that showed that a large percentage of one analyst's
       surrogate recoveries were high. Since this analyst used a different stock solution of
       surrogates, the stock was re-analyzed and found to be incorrectly constituted.

       Samples, of a specific matrix, received from one customer were found to have low
       surrogate recoveries. Two years before the same customer submitted similar
       samples. By using GCMS to retrieve the surrogate/ matrix data for the older
       samples, it was shown that the lab had similar analysis problems.
                                          1-223

-------
                                                                           Page6

       A large group of samples in the middle of the month had low surrogate recovery.
       By using GCMS it became apparent that all of the samples were analyzed on a
       specific instrument. The instrument calibration was examined and found to contain
       a wrong (or incorrect) number. Upon correction, all surrogate recoveries were
       recalculated and found to be within range.
Conclusion

GCMS gives a laboratory with multiple instruments from different manufacturers and with
many analysts the ability to view all of the quality control data at one time. The ability to
group samples by matrix allows the laboratory to quickly and easily calculate its own
quality control limits for the many different matrices an environmental lab needs to analyze.
The GCMS application is a simplified way to satisfy these requirement, and allows the
laboratory to satisfy the method requirements in a relatively simple way. The GCMS
application also allows the laboratory to not only gather quality control data, but to also put
it to good use.  From the many sheets of paper and printouts that the GC/MS laboratory had
previously used to accumulate the quality control data, we now are able to put all of the data
into this application in a minimal amount of time. More importantly, by using GCMS we
are able to retrieve the most useful information from the inputted data. In a larger sense this
application can help gather information from a large group of labs doing GC/MS analysis to
study the affect of matrix on surrogates and to evaluate new methods as they are approved.
                                      1-224

-------
0-j         COMPUTER-ASSISTED TECHNICAL DATA QUALITY EVALUATION

       Samuel   Hopperr   Technical  Staff  Member,   James   Burnetti,
       Technical   Staff  Member,  Toxic   and  Hazardous   Materials
       Assessment  and Control,  The MITRE Corporation,  7525 Colshire
       Drive,   McLean,  Virginia    22102;  Captain  Michael  Stock,
       Systems   Analyst,   Human   Systems  Division,   Environmental
       Information  Management   Program  Office,   Brooks   Air Force
       Base, Texas  78235-5000

       ABSTRACT

       Assessing  environmental  contamination  and  designing  and
       implementing remediation  strategies are  invariably  based on
       analytical  data.  Sound  decisions for managing environmental
       assessments and cleanups  can only  be made  if  the  analytical
       data are   of  a  known  and documented quality.    Manually
       assessing  the  quality  of analytical  data  generated  in  an
       environmental study  is  a  resource intensive  endeavor,  and
       frequently  the  results  of the assessment are  not  delivered
       to the  end users  of the data.   This  paper describes  an
       automated  system developed for the Air  Force  Human Systems
       Division   Installation  Restoration  Program   (IRP)   Office
       (HSD/YAQ) to automatically evaluate the quality of technical
       data and   to  store  the  data  with  their  respective  data
       qualifiers  in an electronic database.  The data processed by
       this system are  generated as part  of  the  Air  Force  IRP
       efforts  and are stored in the  IRP Information  Management
       System   (IRPIMS).    IRPIMS currently  contains sampling  and
       analysis data for IRP efforts  at  76  Air  Force installations
       which comprise nearly one million analytical results.
                                      A
       The process of  automated  data  validation must be  considered
       as two  distinct but  interrelated activities.  One of these
       activities   is the  preparation of the  electronic  data files
       by  contractors.    The   second activity  is  the  automatic
       evaluation   of  these  data  by  the  Air Force.   The  Air Force
       has developed software  tools  that  support these activities.
       Two personal  computer tools  have  been developed  to assist
       contractors in  the preparation  of their  data submissions.
       One  of   these  tools   (the Contractor Data  Loading  Tool)
       provides the contractor with  a convenient means  of manually
       entering the  technical  data that are  to  be processed.  The
       other tool (the Contractor  Quality Checking  Tool)  enables
       the  contractor to   evaluate  the  data   integrity  of  a
       submission.   The final  software  tool developed  by HSD/YAQ
       (the Batch Loading Tool)  is  used at  Brooks Air  Force Base
       (AFB)    to   automatically  evaluate   the  submission  for
       consistency with IRPIMS  data  requirements and the technical
       correctness   of  the   data   (e.g.,   holding  times   met,
       contamination in blanks,  logic of well  completion records,
                                   1-225

-------
etc.)   Data  assessment  reviews  that  would require  staff
weeks of  effort can  be done  in  hours with  this  automated
system.   The technical  data  processed  by  this batch loading
tool are stored in  electronic form fully  qualified for all
the  technical  evaluations  that  were  automatically  done
during batch loading.

This discussion covers  the  two  automated  data  validation
activities, and the  three distinct pieces  of software that
have been developed by the Air Force to support them.

INTRODUCTION

The  analytical  data  that are  procured by the Air  Force as
part of its IRP are required to be both scientifically sound
and  legally defensible.   Procurement of technical  data that
meet  these  requirements  must be  based  on  a  structured
procurement   strategy   that   includes   clearly   defined
performance  criteria  and a  quality   assurance  review  of
received   data   for   conformance  with  these   established
criteria.   The  manual implementation  of  such a  procurement
strategy   is  very   resource  intensive/    and   a   manually
implemented review  system is  prone  to errors.  This  paper
describes  an automation  tool  that  has  been developed by the
Air  Force  to  facilitate the  implementation  of a structured
technical  data  procurement   strategy   that   is  capable  of
100  percent check  of  all  data  against  the  established
performance criteria.

The  description  of this  automated  tool will be presented in
three  sections.    The  first  section  will  describe  the
electronic  data  loading  tool  (i.e.,  the  Batch  Loading
Utility) that  is resident on the VAX 8600 at  Brooks AFB.
Included in this discussion  will  be  a description of the
organizational  process  that  supports  the operation of the
Batch Loading Utility.   This  discussion will be  followed by
a description of the various data quality  reports  that are
being produced  while  the  technical data are being  processed
by the  Batch  Loading Utility.  Finally,  a brief discussion
of   the  various  forms   of   assistance  the  Air   Force  is
providing  to  its contractors to help  them prepare suitable
electronic data submissions will be presented.

THE ELECTRONIC DATA LOADING PROCESS

The  technical data that are  generated  as  part of  Air  Force
IRP  studies cross several disciplines, and  the  electronic
collection  of  these  data  must  be   supported  by  various
technical  and administrative personnel.   Additionally,  the
electronic data  loading  process consists  of activities that
must be done  by contractors  and other activities  that must
                             1-226

-------
be done by Air Force  personnel.   The Air Force's electronic
data  loading  process  is   based  on  a  standard  operating
procedure that defines the roles and responsibilities of the
various technical and administrative staff and describes the
electronic   data   loading   review   and  decision   making
processes.    The  illustration in   figure  1  schematically
depicts  this  process.     The   review  and  decision  making
processes and the  roles  and responsibilities  of the various
staff are described below.

ROLES AND RESPONSIBILITIES

Collection  of quality  data suitable  for making  decisions
regarding  remediating  hazardous  waste  sites  is  a  team
effort.   The  roles and  responsibilities  of the various team
members include the following:

The  Remedial  Investigation/Feasibility  Study  Contractor.
The  remedial  investigation/feasibility study  contractor is
responsible   for   conducting  the   field   investigation,
performing  the   laboratory  analyses,   and  preparing  the
electronic data  submission.  The  electronic data submission
contains  a  record of  quality  assurance/ quality control
 (QA/QC)  activities  in   the field  and  in the  laboratory.
HSD/YAQ  has  provided  tools  to  assist  the   contractor in
preparing electronic  data  submissions.    Those tools are
described in  a subsequent  section of this article.

Contracts  Administrative   Branch.      Electronic    data
submissions   are   treated   exactly  like  any   other  IRP
deliverable.   The data  submissions are received by  contract
administrators,  who record the receipt of the deliverables,
and  based on the  advice  of the technical project  managers,
accept  the  deliverables from  the  contractor  or require the
contractor to resubmit  the  electronic  report  with revisions.

 Technical Project Managers.   The technical project manager
 (TPM),  with  the assistance of the  data  administrator and
hydrogeology  and  chemistry  consultants,  is responsible for
accepting or  rejecting the contractor's  data  submissions.
The  data submissions may  be  rejected for  two reasons.  If
the  data does  not  conform  to  the  standard  IRPIMS  data
 format,  they  may  be  rejected,  and the contractor may be
required to  reconstruct the electronic data  submission.  If
the  reports generated by  the  automated QA/QC tool  indicate
that the contractor did not meet  the data  quality objectives
 identified  for the project, the  contractor may be  required
to repeat field  work  and/or laboratory analyses.
                             1-227

-------
            TPMReturo
             Rejected
             Bitch Fib
                to
             Contnctcr
                   DATA FAtUNO
               ^   BATCH LOADER
                \   QC CHECKS
                              REJECTED
                               BATCH
                               FILES
 REJECTED
BATCH FILE
SUBMISSION
  DiU Admin
   Retumi
  Suhminian
   loTPM
           TPMDelmn
            Bitch File.
              to DiU
           AcninutntoT
                                       DATA FAIJJNO
                                         QC CHECKS
  D«U Admin ^
Vinullyliupcctl
  WilhWon)
  PnccMor
 FMbmuVinu
.    Check   j
                                              BATCH
                                              FILES
                                               ON
                                              FLOPPY
                                              DISK
DiU Admin
 Tniufea
Bitch Rle»
 10 VAX
 Computer
                           ELECTRONIC
                            TRANSFER
                                                       FUt File on
                                                      VAX Computer
Dm An
QCdVii
 Bitch
Lotdinf
Software
                                                                    Chemiitiytnd
                                                                    Hydroteology
                                                                    Comihuu
                                                                    AdviMTPM
                                                                         VAX BATCH
                                                                           FILES
                              CONTAMINATION
                                PREC/AOC       I
                                 REPORTS
                               ——»    DATA la
                                   W  ACCEPTED
    DATA
   DO NOT
S MEET DATA
   QUALITY
  OBJECTIVES
                                                      Temporary Tible
                                                        (Good DiU)
                                                                                               DATA
                                                                                              PASSING
                                                                                             QC CHECKS
                                                                                         DiU An
                                                                                        Appended to
                                                                                        Production
                                                                                         Diubtie
                                                                             IVooUCtlOQ
                                                                             Diubue
                                                                                                     DATA PASSING
                                                                                                      QC CHECKS
                   Figure  1.   Data  Flow  Diagram of  Batch  Loading  Process

-------
Hydrogeology  and Chemistry  Consultants.   The  hydrogeology
and chemistry consultants provide the TPM with expert advice
in interpreting the output of the data QA/QC modules.

The IRPIMS  Data Administrator.   The data administrator  is
responsible for  overseeing  the work of the  contractor that
operates the batch loading software.  The data administrator
is ultimately  responsible for maintaining the  integrity  of
the IRPIMS database.   He  maintains  the  lists of valid codes
(e.g.,  codes  for  analytes,   analytical  methods,  sampling
methods,  etc.)   that   the  batch  loading  software  uses  to
evaluate the quality of the electronic data submissions.  He
provides the TPM with expert advice regarding the quality of
a contractor's  data submission from the  database integrity
perspective   as  opposed   to  the  hydrogeology/chemistry
perspective.

The Operator  of the Electronic Data Loading Software.   The
operator  of   the  electronic  data  loading  software  is
responsible for physically loading the electronic submission
onto the  IRPIMS  computer,  running the  data quality checking
software,   distributing  the   resulting  reports   to  the
appropriate HSD/YAQ staff,  and providing technical support
in interpreting the output.   If  the TPM decides to accept
the electronic data  submission,  the operator  then appends
the electronic data submission into the production technical
database.

REVIEW AND DECISION PROCESS

As illustrated in figure I,  the  electronic  data submission
is  first  received  by  the   contracts   and  administrative
branch,  where it is logged as  received.   The submission is
then delivered to the technical project  manager  responsible
for the  IRP project.   The TPM  delivers it to  the  IRPIMS data
administrator.     An   operator  designated  by   the  data
administrator   checks   the   data  submission  for   computer
viruses,  and  then performs a  visual  inspection of the data
using a  word  processor.   If the submission fails the visual
inspection,  it  is returned  to  the TPM,  who  advises the
contracts administrative branch that the  submission  has been
rejected, and should be resubmitted.   Data  that passes the
visual  inspection is  loaded  onto the  IRPIMS computer, and
run through the  data quality  assessment software.  Data that
passes  all  of  the QA/QC  checks  are loaded into a set of
temporary  tables  that can  be appended  to  the production
database.  Data  that fail the  QA/QC checks are  loaded  into  a
set  of  temporary tables  that can be  edited  in  house  or
converted  into standard ASCII  files to  be  returned to the
contractor.    The quality  assessment  software  generates   a
series  of  reports  that are  distributed  to  the  TPM and
                            1-229

-------
chemistry and hydrogeology  consultants.   The TPM,  with the
assistance   of   the   technical   consultants   and   data
administrator,  ultimately decides whether to accept the data
submission  and  whether  the   reports   indicate  that  the
contractor  should  be required  to repeat  a portion  of the
field and/or laboratory work.

THE ELECTRONIC DATA LOADING SOFTWARE

The IRPIMS system resides on a Digital Equipment Corporation
VAX  computer,   running  the  VMS  operating  system.    The
database is managed by relational database management system
software  from   the  Oracle   Corporation.     The  quality
assessment  software  is  written  in  the  "C"  programming
language,  with  embedded  structured query language   (SQL)
statements.  The electronic data submissions are quite large
(typically  25,000  to 50,000  records),  and a  single record
may  contain  10 or more  coded  fields.   To   improve  the
performance  of  the software, many  of the  checks  to verify
that  coded  fields  contain  valid codes  are instituted in
memory  rather  than via lookups  to  tables  on  disk.    A
throughput  of approximately 40,000 records per hour has been
achieved using this algorithm.

The software evaluates the data for three different types of
quality  issues.    First,  the software  evaluates  individual
records  from the computer  scientist's  point of view.   The
software  verifies  that  key  and  required  fields  are not
blank; that numeric fields contain numbers; that date fields
contain  dates;  and that coded  values are  drawn from a  list
of  valid values.   Second,  the software evaluates the  data
from  the chemist's  and  hydrogeologist's  perspective.   For
instance,  analytical  results are evaluated for conformance
with  holding times, contamination,  precision,  and accuracy
conformance  criteria,   and  well  completion  details  are
evaluated for conformance with  specifications identified for
the project.  Finally,  the software evaluates the data  from
a  database  integrity  point  of  view.    For  instance  each
analytical  result  must  be  associated  with an  extraction
event;  each  extraction event  must  be  associated  with   a
sampling  event;  and each sampling  event must be  associated
with  a known sampling location.   Details regarding  reports
generated by the electronic data loading software are  given
below.

DATA  QUALITY REPORTS

The Batch Loading  Utility does  a  number  of quality checks on
the technical data that  it processes, and  it produces a very
comprehensive  set  of evaluation  reports.   Several  examples
of  these reports  will be  presented in  the figures of  this
                            1-230

-------
section.   A complete discussion  of  all the checks  done  by
the  Batch  Loading  Utility  and  presentation  of  all  the
reports  that  are  produced by  this  utility  is beyond  the
scope of this discussion.

The quality evaluations  of the chemistry data  generated  to
support  environmental decision making are perhaps  the most
universally  acknowledged.     These   evaluations   typically
address  the quality  concerns of contamination,  precision,
accuracy,  and  the  maintenance of  sample integrity  during
sampling  and  analysis.    The  analysis  of  environmental
samples is almost always accompanied with the analysis of QC
samples that are  used to gauge the  success  of  the sampling
and  analysis  activities.    These QC  samples  are used  to
document the precision and  accuracy  of  the  analyses and the
presence  of contamination  introduced  in the  sampling  or
analysis   process.      The   integrity   of   analytes   in
environmental samples can be  compromised if  the samples are
held for periods longer than their specified holding times.

A typical  IRPIMS  Batch  Loading Utility  holding  times report
is  shown in  figure 2.   Every  test that  is  submitted  to
IRPIMS  is  checked  for  compliance with  regulator  specified
holding  times.   Tests performed  outside the holding times
are presented in the holding times report.  These tests that
are  not in compliance  with  regulatory specified  holding
times can be rejected and  not entered into  the  database,  or
they  can  be  entered  into  the  database with flags  that
indicate the holding times have been missed.

The  IRPIMS  Batch Loading  Utility produces  a complement  of
reports  on  the many type of QC samples  that accompany field
sampling  and  subsequent  analysis.    Contamination concerns
are addressed by  a  series  of four reports that  are produced
by  the  Batch  Loading Utility.   An  example of  one of these
reports  is shown  in figure  3.   A  series  of six reports
summarize the QC  data that document  precision  concerns and
another  series  of  six  reports summarize the QC  data that
document accuracy  concerns.   These  reports  list  the normal
environmental  samples   that  are   associated   with  these
questionable QC  samples.   The  Batch Loading  Utility also
sets  contamination,  precision, and  accuracy flags  for all
normal environmental samples as passed,   failed,  or could not
determine.   Finally, two  reports summarize the  QC samples
that  accompanied  each  analytical  test  reported  in  the
electronic  submission.
                            1-231

-------
                             Submission Identification

                    Base: XXXXX  XXX XXX X xxxxx
                    Contract: 85-4533 Delivery Order: 06
                    Cont ractor: xxxxxxxxxxxxxxxxxxxxxxx
                    Submission Date: 05-MAR-91 Analysis Date: 26-MAR-91
Method:
Matrix:
SW8270
S
GC/MS for Semivolatile Organics: Capillary Colunn Technique
                                    Standards:
         SAMPLE ID
                                        DATES
                                                                     40       47
                                                               HOLDING TIMES (DAVS)
AFIID
XXXXX
XXXXX
XXXXX
XXXXX
XXXXX
XXXXX
XXXXX
XXXXX
XXXXX
XXXXX
XXXXX
XXXXX
LOG ID
TR-10F
TR-16F
TR-17F
TR-17U
TR-18F
TR-5F
TR-5U
TR-18U
TR-17W
TR-17F
TR-16U
TR-10U
MATRIX
S
S
S
S
S
S
S
S
S
S
S
S
SBD
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
0.00
SED
0.00
0.00
0.00
0.00
0.00
0.00
0.00
'o.oo
0.00
0.00
0.00
0.00
SA
N
N
D
D
N
N
N
N
N
N
N
N
SAMPLING
21-DEC-B9
21-DEC-89
21-DEC-89
21-DEC-89
21-DEC-89
21-DEC-89
21-DEC-89
21-DEC-89
21-DEC-89
21-DEC-89
21-DEC-89
21-DEC-89
EXTRACTION
19-JAN-90
19-JAN-90
19-JAN-90
30-DEC-89
19-JAN-90
19-JAN-90
30-DEC-89
30-DEC-89
30-DEC-89
19-JAN-90
30-DEC-89
30-DEC-89
ANALTSIS
22-JAN-90
22-JAN-90
22-JAN-90
16-JAN-90
22-JAN-90
22-JAN-90
16-JAN-90
16-JAN-90
16-JAN-B9
22 -JAN -90
16-JAN-90
16-JAN-90
SAMPLE TO 1
EXTRACTION
28
28
28
9
28
28
9
9
9
28
9
9
EXTRACTION
TO ANALYSIS
3
3
3
16
3
3
16
16
-1
3
16
16
SAMPLE TO
ANALYSIS
31
31
31
25
31
31
25
25
-1
31
25
25
  (The -1  in row  9 indicates the analysis data is recorded incorrectly.)
   Figure  2.   Installation Restoration Program Information
Management  System Batch  Loading  Utility Holding Times  Report
                                      1-232

-------
                                                    IRPIMS  Batch File  Loader
                                                       Chemistry Report
                                                Part 2 - Contamination Summary
                           Section 0 - Normal Environmental Samples Associated with  Contaminated Lab Blanks

                                                     Submission Identification

                                          Base:  BLTST  Batch loading test ID
                                          Contract: 99-XX99  Delivery Order: 01
                                          Contractor: H/A ERROR-N/A
                                          Submission Date: 22-MAR-91   Report Date:  Ol-APR-91
      Analysis Date: 09-NOV-88
Lab Lot Ctl I:  8809-775
Laboratory: XXXXX
N>
CO
CO
Analyte
ALDRIN
ALPHA BHC (ALPHA HEXACHLOROCYCLOHEXANE )
BETA BHC (BETA HEXACHLOROCYCLOHEXANE )
DELTA BHC (DELTA HEXACHLOROCYCLOHEXANE)
GAMMA BHC (LINDANE)
ALPHA-CHLORDAKE
GAMMA-CHLORDANE
DIBUTYLCHLORENDATE
DIBUTYLCHLORENDATE
p,p'-DDD
p.p'-DDE
p,p'-DDT
DIELDRIN
ALPHA ENDOSULFAN
BETA EHDOSULFAN
ENDOSULFAN SULFATE
ENDRIN
ENDRIN KETONE
HEPTACHLOR EPOXIDE
HEPTACHLOR
Location
00-110-P
00-110-P
00-110-P
00-110-P
00-110-P
00-110-P
00-110-P
00-110-P
LABQC
00-110-P
00-110-P
00-110-P
00-110-P
00-110-P
00-110-P
00-110-P
00-110-P
00-110-P
00-110-P
00-110-P
Laboratory
Sample ID
8809775001
8809775001
8809775001
8809775001
8809775001
8809775001
8809775001
8809775001
8809775PBLK
8809775001
8809775001
8809775001
8809775001
8809775001
8809775001
8809775001
8809775001
8809775001
8809775001
8809775001
Detection
Limit
0.0500
0.0500
0.0500
0.0500
0.0500
0.5000
0.5000
0 .0000
0.0000
0.1000
0.1000
0.1000
0.1000
0.0500
0.1000
0.1000
0.1000
0.1000
0.0500
0.0500
Actual
Result
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
108.0000
101.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
0.0000
Units
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
UG/L
                             Figure  3.   Installation  Restoration  Program Information
                                   Management System Contamination  Summary Report

-------
Another set of technical quality concerns that are addressed
by  the  IRPIMS Batch  Loading Utility  are  construction  and
maintenance   details   of   groundwater  monitoring   wells.
A schematic diagram of a typical monitoring well is given in
figure  4.    The Batch  Loading  Utility   report  shown  in
figure 5   identified   two   types   of  problems   in   the
construction  records  for monitoring  wells in  a  particular
data submission.   The first error identifies wells that have
a  total  depth  recorded that is  deeper  than the  driller
reported  in  the  hole.   Obviously,  this is a recordkeeping
error that must  be  resolved before  these  well  completion
records can  be accepted.   The other  error identified could
be much more  serious.  This  error  indicated the filter pack
length is  only 6  inches  in  length  and the screened interval
of  the  well  is 10.00 feet.   From the  schematic  diagram of
the  monitoring well  in  figure  4,  the  filter pack  should
extend  the  length  of   the   screen  to  ensure  the  proper
production  from  the  well.    This  error  could be  a  simple
recordkeeping  problem or a more serious  construction error
in the installation of these wells.

In  addition to the technical quality evaluations, the IRPIMS
Batch Loading Utility checks the submission  to insure that
all  information  required  was submitted  and that  it makes
good logical  sense.    Figure   6  cites   analytical  result
records that were submitted without  the  analyte identified
 (one instance),  without  the  analytical result stated  (three
instances),   and  no  laboratory  detection  limit  reported
 (244 instances).   Figure  7  reports nine  sample extraction
dates  reported  before  the  sample  was  taken,  34  analysis
dates reported before the  sample was  taken, and 15 analysis
dates  that  preceded the  date  the  sample was  extracted.
Three instances are also identified on this report where the
units "MG/KG" were used for water samples.

AIR  FORCE ASSISTANCE  TO CONTRACTORS

The  success  of  electronic  data  loading depends   on  the
ability   of   contractors   to   prepare   the   electronic
submissions.   The Air Force offers a  variety of assistance
to  its contractors   who  are preparing these  files.   Two
personal  computer "tools" are provided to  assist contractors
in  the  preparation  and  preliminary  evaluation  of their
submissions.   The Contractor Data  Loading Tool provides IRP
contractors with a convenient means of manually entering the
various  technical  data  collected  by  IRPIMS.   This  tool
performs  many integrity checks,  provides  complete  list of
IRPIMS  acceptable codes  online, and  offers technology that
reduces  the  number  of keystrokes   required  to  make  the
submissions.      The   Contractor  QC  Tool   provides  the
                             1-234

-------
Figure 4.  Schematic Diagram  of  an  Environmental
                 Monitoring Well
                      1-235

-------
                             File BCHUCI

                   Detailed Listing of Errors and Warnings
  Error/
  Warning
          Submission Identification

Base: XXXXX XXX XXX  X xxxxx
Contract: 85-4533 Delivery Order: 06
Contractor: xxxxxxxxxxxxxxxxxxxxxxxxx
Submission Date: 05-HAR-91 Analysis Date: 26-MAR-91

       Description of Problem with Data
Number of
Occurences
  Error       Total casing depth exceeds borehole depth (TOTDEPTH > DEPTH)
             for the following records:
             AFIID  LOCXREF
               TOTDEPTH  DEPTH
  Warning
XXXXX BG-1
XXXXX BG-8
XXXXX C-12
XXXXX C-14
XXXXX C-8
XXXXX C-9
XXXXX OG-5
XXXXX MU-2UWLB
XXXXX MW-4UULB
XXXXX HW-8HF
XXXXX MW-8UWLB
20.00
20.40
25.10
25.50
20.30
22.00
24.20
23.70
25.60
12.60
25.30
Screen length exceeds filter
following records:
AFIID LOCXREF
XXXXX BG-1
XXXXX BG-2
XXXXX BG-3
XXXXX BG-4
XXXXX BG-5
XXXXX BG-6
XXXXX BG-7
XXXXX BG-8
XXXXX C-10
XXXXX C-11
XXXXX C-12
XXXXX C-14
XXXXX C-8

SCRLENGTH
10.00
9.90
10.00
10.00
10.00
10.00
10.00
10.00
10.00
10.00
10.00
10.00
10.00
19.0
20.0
25.0
25.0
20.0
20.0
24.0
23.5
25.5
12.5
25.0
pack (SCRLENGTH > FPL) for the

FPL
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
0.50
Figure  5.   Installation Restoration  Program Information
Management  System Batch Loading Utility  Well  Completion
                     Information Error Report
                                I-236

-------
                              File BCHRES

                    Detailed Listing of Errors and Warnings


                         Submission Identification

               Base: XXXXX  XXX XXX  X xxxxx
               Contract:  85-4533 Delivery Order: 06
               Cont ractor: xxxxxxxxxxxxxxxxxxxxxxxxxxxxx
               Submission Date: 05-HAR-91 Analysis Date: 26-MAR-91
 Error/
 Warning
Description of Problem with Data
Number of
Occurences
 Error        Key field PARAMETER LABEL (PARLABEL) was left blank

 Error        Required field PARAMETER VALUE (PARVAL) was left blank

 Error        Required field LAB DETECTION LIMIT  (LABDL) was left blank

 Error        Required field EXPECTED VALUE (EXPECTED) was left blank
                                                  1

                                                  3

                                                  244

                                                  609
 Warning      Contamination in the laboratory of  field blank samples were
             detected for the following tests/analytes:

             ANMCODE

             £418.1
             SW6010
             SW7191
             SW8240
             SW8270
                                                  2
                                                  11
                                                  4
                                                  35
                                                  13
Figure  6.   Installation  Restoration  Program Information
    Management  System  Batch Loading  Utility Analytical
                         Results  Error Report
                                  1-237

-------
                              File BCHTEST


                    Detailed Listing of Errors and Warnings



                         Submission Identification


               Base: XXXXX  XXX XXX  X xxxxx
               Contract: 85-4533 Delivery Order: 06
               Contractor: XXX XXXXXXXXXXXXXXXXXX

               Submission Date: 05-MAR-91 Analysis Date: 26-MAR-91
 Error/
 Warning
Description of Problem with Data
Number of
Occurences
 Error        The extraction date must be on or after sampling date


 Error        The analysis date must be on or after sampling date


 Error        The analysis date must be on or after extraction date
                                                  9


                                                  34


                                                  15
 Error        The following soil  or tissue records did not have a value in

             the field BASIS:


             API ID LOCXREF    LOGDATE  SBD    SED    MIX SAC


             XXXXX MU-7UULB   16-NOV-89 5.00   5.00   S   N


 Error        The following UNITS OF MEASURE (UNITHEAS) are not applicable

             to a water sample:


             UNITMEAS


             MG/KG
Figure  7.    Installation Restoration Program Information
      Management System Batch  Loading Utility  Sample
                      Preparation Error  Report
                                      I-238

-------
contractors with  an  onsite means to evaluate the  format  of
their electronic  submissions.   In addition to  these tools,
the Air Force  also provides a  guidance  document,  training,
and telephone support of the contractor data loading effort.

SUMMARY

The Air Force has implemented a comprehensive technical data
evaluation process that  includes  a  high  level of automation
and results  in  the technical data being  stored in a easily
retrievable  electronic  format.    The  data processed  by the
Batch Loading Utility are  reviewed for a number of technical
quality concerns  and the  results of  this review  is stored
with each record  in the database.  The Air Force anticipates
that this  data will  eventually  be used to  supply  data  to
various   graphical   (including  geographical   information
systems)  and   modeling   tools   that   will  facilitate  IRP
remediations.

ACKNOWLEDGEMENTS

The authors  would like to  acknowledge Mr.  Stephen Shieldes
of OAO  Corporation,  the  implementation programmer who coded
the IRPIMS Batch  Loading Utility.
                             1-239

-------
SAMPLING/FIELD

-------
 Preparation and Stabilization of Volatile Organic Constituents of Water Samples by Off-Line Purge
                                         and Trap

                                            by

       Elizabeth Wodfenden, Perkin-Elmer Limited, Seer Green, Buckinghamshire, England, and
       James Ryan, The Perkin-Elmer Corporation, 761 Main Avenue, Norwalk, CT 06859-0219


Purge-and-trap gas chromatographic analysis has a 20 year history of successfully analyzing volatile
organics in water. The technique is quantitative, sensitive, and able to be automated.  Purge-and-trap is
used in all major EPA monitoring programs; RCRA, CERCLA, NPDES industrial wastewater, and drinking
water.

A conventional system is integrated, i.e. the purge vessel and sorbent trap are connected directly to the
gas chromatograph in a laboratory environment. Water samples must be collected in the field, chemically
stabilized, atmospherically sealed, and shipped to a laboratory while chilled. When received at the lab,
they must be stored at 4°C. until analyzed, and at least for CERCLA, these samples must be analyzed
within 10 days of receipt. While this existing analytical system works well, this paper will demonstrate that
it is not necessary for the purging to be done in proximity to the chromatographic analysis.

Purge-and-trap systems which incorporate an integral (on-line) thermal desorption device have been
found to suffer from several limitations. These limitations can adversely affect the practical performance of
such on-line systems.  Among the limitations are:

    -   RISK OF CARRYOVER BETWEEN SAMPLES. This can occur when a particularly high
       concentration sample is analyzed.

    -   RESTRICTED COMPATIBILITY WITH HIGH RESOLUTION CAPILLARY GC +  MS DETECTION.

    -   RESTRICTED STORAGE TIME FOR WATER SAMPLES.

    -   INCREASED CHANCE OF SAMPLE CONTAMINATION.  This can happen because of stabilizers
       added to the sample to prevent haloform formation, or from atmospheric contamination of the
       aqueous sample from improper sample seals.

One solution to overcoming these potential problems is to separate the purge-and-trap volatile chemical
collection and concentration from the desorption-chromatographic analysis. In other words, perform the
chromatography off-line from the sample concentration.

Using portable traps in combination with an automatic off-line purge unit enables water to be sampled
using conventional EPA purging methodology at field sampling stations. Once sampling is completed, the
tubes may be capped and transferred for thermal desorption GC analysis at a central laboratory facility.
This approach immediately overcomes two of the major drawbacks of conventional on-line methodology:

    -   NO RISK OF CARRYOVER BETWEEN SAMPLES.

    -   GREATLY EXTENDED MAXIMUM SAMPLE STORAGE TIMES.

    -   DISTRIBUTED FIELD SAMPLING COMBINED WITH CENTRALIZED LABORATORY ANALYSES.
                                            I-243

-------
Commercial automatic thermal desorption instruments allow multiple sample tubes or traps to be
analyzed without operator attendance. These traps, in the form of sampling tubes, are compatible with
the method detection limits specified by EPA 500 and 600 series methods, as well as those purge-and-trap
methods in the RCRA SW-846 analytical method manual. For long term storage, the sorbent tubes can be
capped with brass Swagelok ™caps and one-piece PTFE ferrules. Such tubes, spiked with benzene,
toluene and m-xylene, are available as certified standards (Ref.1), and have been shown to be stable for
up to two years of storage time.

In addition, data reported by the Netherlands Organization for Applied Scientific Research shows that
chlorinated hydrocarbons on Tenax ™ are stable for over 2 years. Multiple analyses for trichloroethylene
and tetrachloroethylene carried out over a two year period had a reproducibility with less than 10% RSD at
storage temperatures ranging from 4™ to 40 °C (Ref. 4).

                     STABILITY OF VOLATILE CHLOROALKANES ON TENAX
 Storage            Component                                      24 Month
Temperature                                Initial Mean Charge          Mean Recovery
  °C                                       ng     RSD%  #Rep       ng   %Rec.
    4°              Trichloroethylene        840    2.0%   15          856  102%
                    Tetrachloroethylene      806    1.9%   15          781   97%

    20°             Trichloroethylene        840    2.0%   15          816   97%
                    Tetrachloroethylene      806    1.9%   15          756   94%

    40°             Trichloroethylene        840    2.0%   15          842  100%
                    Tetrachloroethylene      806    1.9%   15          765   95%

Ref: TNO Division of Technology for Society: Netherlands Organization for Applied Scientific
    Research, Report No. R90/268.

The ATD-400 also overcomes another limitation of conventional procedures, i.e. incompatibility with high
resolution capillary GC, by using an optimized two-stage thermal desorption process.

REFERENCES
    1.  Certified Standard Material Reference No. CRM112. Available from the European Community
       Bureau of Reference:
              Community Bureau of Reference (BCR)
              Rue de la Loi 200
              B-1049 Brussels
              Belgium

    2.  How Efficient are Capillary Cold Traps?, J. W. Graydon and K. Grob, Chrom. 15,327,1983, pp.
       265-269.

    3.  Modified Analytical Technique for Determination of Trace Organics in Water Using Dynamic
       Headspace and Gas Chromatography-Mass Spectrometry.  Bianchi, Varney, and Phillips, J. of
       Chrom. 467 (1989), pp 111-128.

    4.  Stability of Chlorinated Hydrocarbons on Tenax, F. Lindqvist and H. Bakkeren, Netherlands Or-
       ganization for Applied Scientific Research, TNO Division of Technology for Society, Report No.
       90/268.

                                              I-244

-------
           A REMOTE WATER SAMPLER USING SOLID PHASE EXTRACTION DISKS

H. A.  Moye, Pesticide Research  Laboratory,  Food Science and Human  Nutrition
Department,  University of Florida,  Gainesville,  FL,  32611  and W.  B.  Moore,
Pesticides and Data Review Section, Bureau of Drinking and Groundvater Resources,
Florida Department of Environmental Regulation, Tallahassee, FL, 32399-2400.

ABSTRACT

In order to  reduce costs  and  inconveniences  associated with  the  sampling,
preservation, storage, shipping  and analysis of large volumes of water for the
analysis of trace levels of pesticides, we  have designed and developed a remote
water sampler employing porous extraction disks.  Commercially available disks
were studied in the  laboratory  for their  ability to  extract  the pesticides
alachlor, butachlor, ametryn, prometryn, and terbutryn from both laboratory and
groundwaters. Extraction efficiencies  were determined as functions of pesticide
type, concentration,  flow rate  through  the  disks,  disk pretreatment,  storage
temperature, and storage  interval.   Results were encouraging enough  to pursue
design and  construction  of  the remote  sampler, made possible by the  extensive
modification of an existing commercially available water sampler.  Evaluation of
the resulting sampler  using extraction  disks  showed  that it  was reliable and
accurate enough to be used in  the field.

INTRODUCTION

One of the  larger expenses in any ground  or surface water  monitoring project
entails proper collection, preservation, storage, and transportation of the water
samples to be studied.  This  is particularly true when transient events, such as
a moving  groundwater plume or a single surface contamination event, occur.  Being
able to effectively sample in a continuous  manner over a period of hours, days,
or even weeks would allow for  the observation  of such an event,  and prevent it
from escaping unnoticed.

Most approved USEPA numbered methods for the analysis of trace organics in water,
(500 and 600 series)  including  pesticides,  require the  collection  of large
volumes  of  water,  typically  one  liter.    Transporting multiple  one  liter
quantities, when multiple analytical methods are to be employed, or when sampling
over regular intervals  is to be undertaken, requires expensive transportation and
storage techniques.  And when they arrive at the laboratory, they are typically
extracted with large quanitities of  expensive and hazardous solvents,  such as
methylene chloride.  Some methods allow the water to be extracted on-site with
solid phase extractors (SPEs; 1-3), with subsequent transportation back to the
laboratory for elution and chemical analysis, usually by gas  or high performance
liquid chromatography.  These SPE devices are small cylindrical cartridges packed
with intermediate size silica particles,  usually chemically coated with an alkyl
material, such as  octyl groups.  They adsorb organics from water when it is drawn
through by pressure or vacuum, and then release the organics  to a small amount
of solvent that is passed through the  cartridge.

While on-site extraction via the  SPEs obviates several problems,  there still is
the requirement that personnel be transported, housed, maintained,  and paid if
sampling is  to be  done over time.   In order to eliminate the requirement for
collecting, storing, and transporting large volumes of water,  while at the same
time minimizing costs  for the presence of  on-site personnel,  we conceived the
idea of  creating  a remotely  located "dosimeter",  which could be  placed  at a
remote surface or  well water site,  programmed to collect  one liter samples at
various  intervals,  perform the  sampling unattended,  process  the sample water
through a solid phase extractor, direct the  processed water to waste, shut itself
off and await return of personnel.  It was our initial  intention to employ SPE
cartridges that have been commercially available.  However,  as the engineering
of the device began, it became  readily  apparent that switching a stream of water
                                      1-245

-------
into and away from an array of SPEs would be very difficult, if not impossible
to achieve,  given the pressures required to achieve reasonable flow rates through
the SPEs.   Attention turned to the more recently  developed extraction disks,
developed by the 3M Company (4,5).  These disks have proved their worth for such
an application, as we now report here.

We  evaluated  the  C8  disks  for  the  extraction  of  the  pesticides  ametryn,
prometryn, terbutryn, alachlor, and bromacil from laboratory and  three typical
Florida  groundwaters;  some evaluations were done in  the laboratory in filter
holders, and some within the "dosimeter".

DISK EXTRACTION EFFICIENCIES

EXPERIMENTAL

All studies were performed on the 4.7 cm Cs Empore extraction disks  (Analytichem,
Harbor  City, CA).   They were held by  a glass  Millipore model KGS-47TF filter
holder,  attached to  a  250  mL graduated funnel that was clamped hold the disk.
The filter holder was placed into a 1  liter suction flask  that was attached by
Tygon tubing to a Millipore vacuum pump, model 5KH10GGR28S.  This vacuum pump was
equipped with a needle valve for adjusting the vacuum present in  the flask.

Pesticide analyses were conducted on a Hewlett-Packard gas  chromatograph, model
5890, equipped with a nitrogen-phosphorus (NP) detector, and an electron capture
detector.   The instrument was capable  of automatic  sample injection and peak
integration.  Separation of the pesticides under study was accomplished on a DB-5
bonded  fused silica column, 30 meters  long x 0.25 mm ID, with a film thickness
of 0.25/jm (J and W Scientific, Folsom, CA).   Injections were made, 2 fiL, in the
splitless mode  with a 45  second  delay.  Carrier gas  was  helium  at 30 cm/sec
linear velocity.  Injector temperature, 250°C;   NP detector temperature, 300°C;
EC detector temperature, 300°;  oven  temperature programmed from 60°C  to 300°C at
40C/min.

Pesticides studied were ametryn, prometryn, terbutryn, alachlor, and bromacil,
all obtained in 99.9%  pure analytical standard  form from the  EPA Pesticide
Repository  (Research Triangle Park, NC).

All solvents were HPLC grade (Fisher Scientific or Burdick and Jackson).  Wetting
agents  (CETABS, polyethylene glycol, octanol, etc) were ACS reagent grade from
Fisher.

Since reverse phase disks, the C18 and the C8, are hydrophobic by their nature,
and require pretreatment with a water miscible solvent to allow activation of the
reverse phase material,  and to allow for water flow,  various wetting agents, neat
and dissolved in water, were examined for their abilities to wet the Empore disks
so as to allow water flow by gravity.  Efforts were made to select water miscible
alcohols and surfactants that would be involatile, so that pretreatment could be
done in the laboratory before assembly of the disks.  The disks were soaked in
the following agents, for 1 hour and 30 minutes, then they were removed, placed
in the Millipore filter holder, 100  mL of HPLC  grade water  added to the funnel,
and allowed to drain by gravity,  without vacuum from the vacuum pump.

Three procedures  were  studied  for their  ability  to  extract the  pesticides
ametryn, prometryn, and terbutryn from water.   They were designed with an eye
toward  their  implementation in  the dosimeter,  since it  was known that  the
dosimeter would place  some constraints  on how  disk  pretreatment could  be
accomplished.  We wanted to determine  what  steps  in the Analytichem procedure
could be either reduced  or eliminated,  while still  giving good  extraction
efficiencies, so that the dosimeter  design could be kept as simple as possible.
The selected extraction method was then validated on other pesticides (alachlor
                                        1-246

-------
and butachlor) during additional extraction efficiency studies at high water flow
and during volatility studies, where  losses from  disks  used  to  extract  the
pesticides from water were determined (see subsequent section).

These three extraction procedures were:

(1) Modified Analytichem (created from the procedure recommended by Analytichem
in Appendix A by omitting  the 5 mL addition of methanol  to  the water,  and by
presoaking the Empore  disks in ethyl acetate, rather than passing it  through the
disks once they were placed in the holder).
-Presoak disk in 5  mL of ethyl acetate for 6 hours; discard solvent.
-Place disk in Millipore filter apparatus.
-Apply vacuum for 5 minutes.
-Add 20 mL of methanol, apply vacuum; do not let disk go dry.
-Repeat above step  with 20 mL of HPLC grade water.
-Begin adding water sample.
-After sample is processed, pull air through disk for 5-30 minutes.
-Place test tube at tip of filter base inside flask.
-Add 10 mL ethyl acetate,  pull half of the solvent through, let stand for
approximately minute,  then pull the rest through.
-Repeat with a second 10 mL aliquot.
-Sample is ready for analysis or may be concentrated to  a  smaller volume by dry
      nitrogen at ambient temperature.

(2)  PRL Procedure  #1.
-Presoak disk in 5  mL of ethyl acetate for 6 hours; discard solvent.
-Place disk in Millipore filter apparatus.
-Wash with methanol:water (40 mL of 1:1).
-Apply vacuum.
-Add sample.
-After sample is processed pull air through disk  for 5-30  minutes to remove any
      water remaining on disk.
-Elute sample as per modified Analytichem procedure above.

(3)   PRL Procedure #2.  (As  for  PRL #1,  except  20  mL  methanol  used rather
than  methanol:water).

Since the Millipore vacuum pumps used in the laboratory are capable of pulling
water  through  the   Empore  disks  at a higher  rate  than  can  be done  in  the
dosimeter, and since high  flow rates would be desirable  when extracting 1 liter
samples, a study was done to determine what effect maximum  flow rates would have
on extraction efficiency.   Consequently,  we studied the two flows of 20 mL/min
and  80  mL/min  for  several  pesticides  in  HPLC  grade  water   at  several
concentrations.

The ability of the  Empore  disks to function at such high flow rates  (80 mL/min)
allowed for the  collection of 1 liter samples if plugging  did not result and if
"breakthrough"  did  not result.   It was conceivable  to  us that "breakthrough"
might occur for these large  volumes.  Since the  Empore disks are simply very
short and very permeable reverse phase HPLC columns, and even though 100 % water
causes organic compounds to move only slowly down these columns, there is some
volume of water that would cause  the compounds  to "breakthrough" and exit the
disks.  Whether or not this would occur within the 1 liter of water being sampled
was part of this study, as determined by measuring percent recoveries.

Consequently,  replicate one  liter  samples  were  fortified  at concentrations of
0.1,   1.0,  and  10.0  ppb   with the  pesticides  alachlor, bromacil,  ametryn,
prometryn, and terbutryn,  and the PRL #2 procedure was used for extraction and
analysis.  All  determinations were  done  in triplicate with  duplicate sample
injections on the gas chroma to graph.  Unfortunately,  some of the 0.1 ^g/L samples
were  not quantifiable due  to  method sensitivity limitations  (alachlor  and
bromacil).
                                        1-247

-------
In order for the Empore extraction disk concept to be functional in the dosimeter
in  the  field for  extended  periods approaching 30  days, it  was important to
determine whether  after  extraction,  the pesticides would remain on the disks
without unusual precautions so that they could be  transported  to  the laboratory
for analysis.  It was also important to  consider that the disks  could conceivably
experience elevated temperatures as a result of a "greenhouse" effect as they sat
in the dosimeter.

Consequently, we studied all  the  pesticides at  15 and 30 days storage at  25°C
(nominal ambient) ,  and also studied alachlor, bromacil, and prometryn at 40°C for
those intervals.   We  also  studied alachlor and bromacil  for  1 day and 10 day
intervals.  HPLC grade water was fortified at the 10.0 /*g/L level, exactly 100
mL was passed through fresh Empore extraction disks  at 80 mL/min,  and the disks
subjected to the #2 PRL procedure for extraction and analysis.   All samples  were
prepared  and analyzed  in  triplicate  with  duplicate  injections on  the  gas
chromatograph.

RESULTS AND DISCUSSION


How the disks responded to pretreatment with alcohols and surfactants  in terms
of water flow is shown in Table 1.
Table 1:Flow Through Alcohol and Surfactant Treated;Disks, Gravity.
Wetting agent
Observation
none
HeOH
80 % MeOH
50 % MeOH
30 % MeOH
Octanol
80 % Octanol
50 % Octanol
30 % Octanol
Ethylene glycol
80 % ethylene glycol
50 % ethylene glycol
30 % ethylene glycol
Acetonitrile
80 % acetonitrile
50 % acetonitrile
30 % acetonitrile
Polythelene glycol 8000 (0.5 g/lOOmL)
CETAB (0.04 g/lOOmL)
Tetrahydrofuran
80 % tetrahydrofuran
50 % tetrahydrofuran
30 % tetrahydrofuran
No flow observed
0.76 mL/min
0.7
No flow observed
Disk floated; no wetting
No flow observed
Disk floated; no wetting
No flow observed
No wetting observed
0.6  mL/min
0.6  mL/min
0.5  mL/min
No flow observed
Disk floated; no wetting
No flow observed
0.6  mL/min
0.7  mL/min
0.7  mL/min
0.6  mL/min
      These data  showed that  of the  alcohols  and surfactants  studied,  only
methanol, acetonitrile, and tetrahydrofuran would wet the disks adequately for
water flow. We then studied the  effect of  these solvents on water flow through
prewet disks  under vacuum.   Empore disks were soaked  in  each of  the three
solvents that gave flow under gravity,  and  water was passed through at a vacuum
of  either  5 or 10 pounds per square  inch (PSI).   Table 2  summarizes these
results.
                                        1-248

-------
Table 2:Flow Through  Solvent Treated Disks, Vacuum.
Solvent
Tetrahydrofuran (THF)
n
Methanol
n
Acetonitrile
n
Vacuum (PS1)
5
10
5
10
5
10
Flow (mL/mln)
54
88
50
79
59
74
      Even though both THF and acetonitrile gave slightly greater flows than did
methanol, methanol was chosen for further studies,  so that our studies could be
compared with  those  already done and with those yet to follow  by others, who
probably would use the methanol recommended by Analytichent.  Since we felt that
a 10 PSI vacuum could be achieved in the dosimeter  using the  pumps selected for
it, a flow of 79 mL/min would be adequate.  This flow would  certainly permit 1
liter samples  to be  collected within  1 hour, a  goal we set,  if acceptable
recoveries could be achieved  at this flow.  These  studies  showed that some way
of pretreating the disks within the dosimeter  on site before each water sample
was extracted would have  to be accomplished,  since 100 mL subsamples of water
would not  pass through  an unactivated  Empore disk.   This  would have  to  be
accomplished within the dosimeter by a separate delivery system for methanol.

How well the pesticides  were  recovered  from  the Empore disks using the three
pretreatment and extraction procedures is shown in Table 3.
Table 3:  Recovery of the Pesticides Ametryn,  Prometryn,  and Terbutryn from 100
mL of HPLC Grade Water Using Various Procedures and the  Empore Disks.
Pesticide
                  Procedure
Cone. (ppb)
% Rec.
%RSD
Ametryn
Prometryn
Terbutryn
Modif. Analy.
PRL #1
PRL #2
Modif. Analy.
PRL #1
PRL #2
Modif. Analy.
PRL #1
PRL #2
105
105
105
110
110
110
110
110
110
1
3
1
11°
2
1
117
68
106
65
83
95
79
75
88
6.8
12
9.3
7.2
6.1
  Number of individual determinations;  each determination injected in duplicate
on the gas chromatograph.  All flows through the filter were at approximately 18
mL per min.
b Eleven (11) of these extracted with the same Empore disk.
c Extracted with the same Empore disk.
      This data shows that there are little if no consistent differences in the
three extraction procedures used.  Consequently, since the PRL #2 procedure is
simpler than the other two, it was used throughout the remaining studies, unless
otherwise stated. Acceptable reproducibility exists for replicate determinations,
although we would like to  improve on this as the studies continue.   It also shows
that the disks can be  reused many times without suffering fracturing, erosion,
or  loss  of the  silica  particles,  all  of which  would  lead  to  diminished
                                      1-249

-------
recoveries.

  Table 4 summarizes these results for the effect of flow rate on recoveries from
the disks.


Table 4:Effect of Flow Rate Through Empore Disks on Extraction Efficiencies".

Pesticide%Recovery
                              20 mL/min                80 mL/min
Alachlor
Bromacil
Prometryn
89
36
93
96
46
92
  Average of triplicate determinations for 10 pg/L samples; average RSD less than
6%.


      These results show that extremely high flow rates can be handled with no
loss of extraction efficiency.  This ability to  rapidly process samples  in the
laboratory allowed us  to work with 1  liter samples,  giving an improvement in
method limit of detection,  and  would also allow for high sample throughput if
similar vacuum could be established within  the dosimeter.


Table 5:   Extraction Efficiencies of  Pesticides from Empore Disks Using  High
Water Sample Flows (80 mL/min)."


Pesticide               Concentration  (/ig/L)       Average % Recovery


Alachlor                10.0                                89
                         1.0                                92
                         0.1                                --b


Bromacil                10.0                                36
                         1.0                                58
                         0.1                                --b


Ametryn                 10.0                                91
                         1.0                                69
                         0.1                                89


Prometryn               10.0                                93
                         1.0                                91
                         0.1                                76


Terbutryn               10.0                                83
                         1.0                                79
                         0.1                                35


* Averages  of triplicate determinations,  with duplicate injections for each on
the gas chromatograph.  Average % RSD for triplicate determinations less than 8%.
b Unable  to quantitate  due to sensitivity problems.
                                        1-250

-------
      From the above data, it is apparent that only bromacil gives low recoveries
at all concentrations.  There is also an apparent recovery problem with terbutryn
but only at the lower concentrations. Subsequent studies on pesticide losses due
to volatility (see Table 6 below)  indicated that low recoveries  of bromacil may
be due to factors other than that of poor adsorption by the Empore disks.


Table 6:Pesticide Losses from Empore Extraction Disks with Storage at Extended
Intervals and at Elevated Temperatures.8
Pesticide Storage Interval Temperature Average % Recovery
Alachlor
Bromacil
Prometryn
Ametryn
Terbutryn
1 day
10 "
15 -
15 "
30 -
1 day
10 "
15 "
15 "
30 "
15 day
15 "
15 day
15 day
25°C
n
408C
258C
25°C
tl
It
40°C
25°C
25°C
40°C
25°C
25°C
99
87
91
98
87
95
97
64
83
78
43
30
30
4
"Average of triplicate determinations, 10 A*g/L; duplicate injections on the gas
chromatograph.   Average % RSD less than 10%.
From these data  it is obvious  that ametryn, prometryn,  and terbutryn cannot
withstand 15  day storage intervals at 25°C on the Empore disks.  Surprisingly,
both alachlor and  bromacil are  recoverable even up  to 30  days  with average
percent recoveries greater  than 78%.   Even more  surprising is that bromacil,
which had shown very poor recoveries upon immediate extraction of the disks with
ethyl acetate (see Table 5), now gives  excellent recoveries after sitting for
days after  extraction  of  the  water  samples.    Apparently,  aging  the  disk
stabilizes bromacil,  or, makes  it more amenable  to  ethyl acetate extraction.
Whether this  phenomenon would occur for other pesticides that were not studied
here is yet to be determined,  and merits investigation.

Raising the storage temperature  from 25°C to 40°C caused little additional loss
of pesticide for the three studied at the higher temperature.  This  may  be a very
important aspect for long term use of the dosimeter when temperatures may become
very high in the field.

Also of potential value is  the  fact that during  these long term  studies,  all
  pore  disks  were placed in the open in fume hoods, susceptible to an array of
                                        1-251

-------
microbes attached to air particles which would "fall out" on them.  Despite this,
none of the pesticides were significantly degraded by exposure to the atmosphere.
This is self evident from a "prima facie"  aspect when the mechanism for retention
of pesticides on the 8p irregular shaped silica Empore particle is considered.
It is well recognized that retention occurs at the alkyl  (C8) bristle which is
deep inside the pores  of the particle.  These pores  are  only 60 Angstroms in
diameter; expressed in microns (/O,  that amounts to 10"* ft.  However, microbial
cells are larger  than 2 x 10"1 p in  diameter,  making them much  too  large to
penetrate into the pores of the silica particles of the Empore disks.  Therefore,
microbial degradation  of pesticides adsorbed by Empore  disks is,  from first
principles,  impossible.   For  this reason, difficulties we faced in getting water
flow through the 0.2 ft microbial filters are of only academic concern.


EXTRACTION DISK WATER SAMPLER DESIGN.  CONSTRUCTION. AND VALIDATION

Introduction

There were several  goals that were  established  to  be  met in the  design of the
dosimeter.  The device  needed to be battery powered  and capable of operating
unattended over extended intervals of time, as long as one month.  Battery drain
should be such that as many as 24 samples, of one liter volume, could be taken
in as little as 24 hours.  Sampling  volume accuracy should be  as good as + or -
10%.  It ought to be programmable, such  that variable volumes could be sampled
at variable time intervals.  It ought  to be capable of processing more than one
sample through the  same  disk.   Only minimal redesign and modifications of an
existing inexpensive water sampler ought  to  be required.  While  still being
highly portable, the dosimeter ought to have a large enough internal volume to
accommodate at least 24 Empore disk holders.

An American Sigma Streamline water sampler (Middleport, NY), was chosen as the
nucleus  of  the disk water sampler  (See  Fig.  1).    It came  equipped  with an
"advanced programming" mode  present in  the  EPROM software that  allowed much
flexibility in the programming in regard to sample volume, intervals,  multiple
sampling, etc.  However, its main feature, which was not apparent until it became
obvious that vacuum would have to be applied to the bottom of each Empore disk
holder, was  that the motor that was used to position the sample distribution arm
over the sample bottles could be synchronized with a somewhat similar motor which
would rotate a valve for selecting which filter holder received vacuum.  This was
a critical aspect in the successful  design of the dosimeter.

Also critical was the  fact that electronic timing circuitry  was  in place and
enabled the  installation  of a switch as a means of acctuating a methanol pump at
the appropriate time for each sample.

The device as  modified will accept up to  24 holders.  During operation, 1000 mL
of water is processed through each holder, in ten  100 mL increments.   In this
report these 100 mL increments  are referred to as sub-samples.  The machine can
be programmed to vary the time between increments and also set real time start
and stop times  for  each 1000 mL  sample  as well as continuous operation once
started.

The number of 1000 mL samples can be  selected as any value up to a maximum of 24.
If the multiple stop-start method of programming is to be used and the machine
set to shut off after the  desired number of samples have been taken, holders must
be placed in the machine  in direct sequence, 1 to whatever number desired, in a
counter clockwise direction from position 1.

After completion of  the sampling program, the holders can be brought back to the
lab where the Empore  disks can be processed without removal from the holder.They
                                       1-252

-------
are processed by elution of the adsorbed pesticides from the  disk material with
a solvent suitable for the type of analysis to be used.

The device, as  described here, comes equipped with  an intake tube of 3/8 in.
internal diameter, and a weighted intake strainer for  sampling surface waters.
An interface remains to be constructed  for monitoring  wells.

Preparation of Holders


For use  in this device,  the Empore  extraction  disks are held  in a modified
disposable filter funnel (See Fig. 2).  These filter funnels are manufactured by
Micron Separations,  Inc. of  Westborough,  Mass.  (cat.#  DFN-P4SGS-S1)  and are
supplied with various types of filter media.  For  this work, all of  the supplied
filter media,  except  the  supporting membranes is discarded, making space for the
stacked disks described below.

The remaining support disk is modified by trimming it to 40 mm in diameter. The
trimmed support is then centered in the  lower half of the holder and cemented in
place with silicon rubber cement.  After this re-work  is performed, the holder
is then reassembled in the following manner:

          1st layer (bottom)  - one original support  disk  (trimmed  to 40 mm)
          2nd layer - Empore™ cat. # 1214-5002,  C8 disk (Analytichem
                     International)
          3rd layer - Whatman Multi-grade filter, #  GMF150

The two halves  of  the  holder are then  placed  together and held with moderate
downward pressure while tape is applied to assemble  them.  If the holder is to
be reused ordinary duct  tape 3/8 in. wide has proven satisfactory.   Care was
exercised while assembling the holder to limit  downward pressure to prevent the
multigrade filter being cut allowing bypass of the sample around it.


Description of System for Vetting Extraction Disks

Before passing any water through the  Empore  disc,  it must be  wet with methanol.
In this machine,  this  is accomplished  in the  following  manner.   Methanol is
placed in a 500 mL reservoir (See Fig. 3).  At  the beginning of the sampling
interval, the main pump runs in reverse for several seconds in order  to purge the
intake line to  the pump.   After that,  the methanol  pump is  turned on,   which
delivers approximately  10 mL of methanol to the main pump discharge funnel on top
of the distributor arm.   The methanol  flows by  gravity down to filter holder
number one and wets the disc.  On the following pump cycles the methanol pump is
held off until there is  counted  a total of  20  reverse run times (two for each
sub-sample interval).   Runs  in  the  forward direction are ignored due  to the
action of a diode.  After twenty run times, the  circuit is reset and is ready to
pump methanol  again for the next sample.  See Fig.  4 for a summary  of the parts
added to the Streamline  water sampler,  and Fig.  5  for a logic  diagram of the
control circuits,  both original and added.

Description of the Vacuum Pump Circuit

At the same time the methanol pump starts,  a vacuum pump is turned on to pull the
methanol through the filter,  and controlled by a four minute timer.   A vacuum
of approximately 15 in. of Hg is applied to the bottom of each individual disk
holder for drawing the  sub-sample through the disk.   This vacuum is supplied by
a small 12V powered vac  pump and distributed by a  24  port rotary  valve.   The
rotary valve is synchronized manually before the beginning  of  the program by
rotating  the  distribution  arm  and  the  valve to  position  #1  individually.
Thereafter, the  interface circuitry advances  the rotary valve in synchronism with
the distribution arm allowing  time for  the  sub-sample to be drawn  through the
                                        1-253

-------
disk before advancing the valve to the next position.

Modification to Main Pumping System

The  main pump  flow rate  is  approximately 3500  mL/min as  supplied by the
manufacturer.  Since sub-sample increments  of  only 100 mL are used, this high
flow  rate resulted in splashing and control problems.  In order to reduce these
problems the pumping rate was changed by reducing the size of the tubing used in
the pump and the suction line.  In  addition,  it  was  found that a small amount of
water was trapped in the pump due  to sagging of the tub ing, result ing in a small
carryover between sub-samples.  The tubing was shortened to eliminate this sag
and insure that the pump was blown dry during the purge cycle.

Materials were chosen for the dosimeter construction and evaluation that were of
high quality, reliable,  and  economical,  all readily obtainable from suppliers
within the United States.

Empore Filter Design Considerations

Study design called for lake water, with 3 levels of algae,  to be tested in the
dosimeter, as well  as 3 groundwater  (well)  samples with 3 levels of hardness.
Therefore, a filter holder was identified that  could accommodate multiple, or a
stacked  array of  filters.  We  erroneously believed  that some sort of microbial
filter would have  to be incorporated  in the filter holder to prevent microbes
from reaching the Empore extraction disk during water filtration.


Overflow Detection When Filter Becomes Plugged

A  convenient  way was devised  to  determine  when filter holder  overflow would
occur, and which filter had overflowed,  ostensibly  as a result of Empore filter
plugging, either by algae or other microparticulates.  This  was accomplished by
ringing  the top portion of the filter holder with  paper tape impregnated with
water soluble ink  (Sanford's Mr.  Sketch).  As the water begins to overflow the
washing  away of the  ink begins.   This feature ought  to be quite useful in that
it would prevent the analytical work from being performed in the laboratory on
samples  that plugged in the field.

Dosimeter Pump Accuracy and Reproducibility

The  accuracy  and  reproducibility  of  the   dosimeter  peristaltic  pump  were
determined by evaluations  with Fisher HPLC  grade water.  Before beginning the
study, the pump was calibrated for flow. The pump was then programmed to pump 10
sequential 100 mL samples at 12 minute intervals. Water was  collected in a 1000
mL  graduated  cylinder  and measured  after  each sample.    Of  the  10  samples
collected, 3 of them were exactly 100 mL, while 7 of them varied plus or minus
5 mL.  Total  volume delivered was 995 mL,  very well within the manufacturers
specifications for the pump.

Four more similar tests  were run, with an average delivered volume per sample of
98.9 mL,  excellent accuracy and reproducibility.

Lake Water Extraction Efficiencies

Three lake waters were  selected with varying amounts of algae as determined by
Secchi disk.  Secchi numbers (see Table 7 below) are the number of feet that a
black and white disk can be  lowered before  definition of its outline is lost.
One gallon (4 L) brown jugs were filled,  taken  to the laboratory for storage at
4°C, fortified at  10 /jg/L  with all eight pesticides,  and processed in 100 mL
subsamples through the  dosimeter to give a total of 1 L for  each sample.  These
                                       1-254

-------
studies were done on  stacked filters  that did not include the 0.2 /i microbial
filters, but did include  the GMF multigrade.   The support filter had not been
trimmed for  these.   Three filter aids were also tested  for  their ability to
filter out the algae  onto the multigrade filter:   (1) Solka Floe, James River
Corp.,  (2)  Hyflo Super Cel,  Johns-Manville  Corp.,   and  (3)  white  sea sand.
Exactly 15 mL of each were placed into the filter holders, along with several
controls (no filter aid), and the waters were run through.


Table 7:Dosimeter Extraction of Several Lake Waters  Fortified with Eight Test
Pesticides; Effect of Various  Filter Aids a

      Lake WaterSecchi Disk Reading

      Lake AliceT75~~
      Bivens Arm                    1.8
      Lake Wauberg                  2.0
      Newnans Lake                  1.8
      Red Water Lake                3.2
      Lake Lochloosa                3.5
      Riley Lake (Putnam Co.)       8.0

*Fortified at  10 A»g/L.All samples  plugged  the filter stack and flow stopped
at 300 mL of water, + or  - 50 mL, whether filter aid was present or not.


These results clearly  showed that the tested version of the Empore extraction
disks cannot be used for the extraction of 1  liter quantities of lake water, even
when a multigrade filter is present and even when 15 mL of filter aid are present
in the filter holder.

Extraction of Groundwater Samples Fortified with the Eight Test Pesticides

Three groundwater samples of various hardnesses  were  to be collected, fortified
with the eight test pesticides and extracted with the dosimeter.  Those three
chosen were:   (1)   Murphree  well  field,  City of  Gainesville,  well  #1,  (2)
shallow well at a private residence, Gainesville,  FL, and (3)  private shallow
well on Riley  Lake,   Putnam  County,  FL.   These wells ranged widely  in their
hardnesses.  For example,  the Gainesville well was the most hard, with 320 mg/L
(ppm) of bicarbonate,  followed by the Murphree  well  #1,  with 208.6 mg/L.  The
Putnam County well was  least hard, with a bicarbonate of only 10.2.

Samples were collected (8  L of each) in one gallon  brown jugs and transported to
the laboratory where they were fortified at 10 yug/L with the five test pesticides
and stored at 4°C before extraction by the dosimeter.  Exactly 1 liter samples
of each were extracted  in triplicate.   Alachlor, bromacil, ametryn, prometryn,
and terbutryn were analyzed by gas chromatography as previously described.  Table
8 summarizes the recoveries.
                                        1-255

-------
Table 8:Extraction Efficiencies of Eight Test Pesticides  Extracted from Three
Florida Groundwaters  Using the  Dosimeter Equipped with Reduced  Size (40 mm)
Support Disks."
Well
alachlor
% Recoveries
prometryn
ametryn
Murphree
Gainesville, FL
Putnam Co. , FL
109
113
100
107
110
102
108
114
106
Well
      bromacil
                                    % Recoveries
            terbutryn
Murphree
Gainesville, FL
Putnam Co., FL
       119
        91
       100
               106
               109
               100
" Averages of three separate determinations, fortifications at 10 pg/1.; RSD less
than 5%.


Although  somewhat high,  these  recoveries were  similar,  from pesticide  to
pesticide  and from well  to  well.   Relative standard deviations for triplicate
analyses averaged less than 5% for all pesticides and wells, not much larger than
what could be  expected for instrumental error.

FIELD TEST OF  DOSIMETER

The pumping  capacity and accuracy of the dosimeter in  the  field  was briefly
examined  by  placing it on  the  shore of  Riley  Lake,  Putnam County,  FL.   The
dosimeter was programmed to sample 24 one liter samples in 100 mL subsamples over
a 24 hour period, beginning at 1:06 PM on Nov. 29 and ending  at  1:06 PM on Nov.
30.  Since previous  runs  on Riley Lake water showed that the Empore filter disks
plugged at approximately 300 mL,  they were not  placed  in the disk holders for
this study.  Only the support disk and the GMF 150 multigrade disk were installed
in each of the 24 holders.   All water pumped through the disks was collected in
large carboys and transported back to the laboratory for volumetric measurement.
After all  samples were completed,  the overflow  detection tape was  examined on
each holder.   Battery voltage was measured before starting  the sampling and
immediately following completion of sampling.

No overflow  was  observed for any of the  samples, nor was  there  evidence  of
splashing  inside the dosimeter around the filter disk  holders.   Exactly 20.83
liters of water were collected, 87 % of the programmed volume.   This accuracy
falls outside  the specifications  set by the manufacturer, however,  it  may  be
explained  by  the extreme cold temperatures the dosimeter experienced  on the
morning of Nov. 30 when 408F was  measured.  Battery  dropped very little,  from
12.75 volts   to  12.48 volts,  indicating  that  much  more  sampling could  be
accomplished before recharging.

SUMMARY

We have shown  that the pesticides  alachlor, bromacil,  ametryn,  prometryn,  and
terbutryn can be  extracted from laboratory and well  waters at  trace levels using
C8 Empore  extraction  disks  and one  liter samples.  Drying  of the  disks  was
required for  the near complete recovery of bromacil. Several lake waters plugged
the disks at approximately 300 mL, whether filter aids and multigrade prefliters
were employed or not.  For  laboratory water,  extraction efficiencies decrease
little with increasing flows, maximum flow being determined by disk rupture.  The
                                        1-256

-------
pesticides alachlor and bromacil could be kept  on disks at ambient temperatures
(25°C) and elevated temperatures  (40°C) for up to 30 days without loss.  All five
pesticides could  be efficiently extracted  from  three different well  waters
varying widely in  hardness using  the dosimeter.   Pumping accuracy, in the field,
was not as good as observed in the laboratory, although it  could be measured and
used for sample volume corrections.

ACKNOWLEDGEMENTS

This work was sponsored by the Florida Department of Environmental Regulation,
under project WM 321.

REFERENCES

1.  Cochrane, W. P.; Lanouette, M.; and  Trudeau, S; J. Chrom. 243(2), 307-314
      (1982).
2.  Jones, A. S.;  and Jones,  L.  A.; J.  Agric. Food Chem. 30(5), 997-999 (1982).
3.  West,  S.  D.; and Day, E. W. Jr.; J. Assoc. Off. Anal. Chem.  64(5), 1205-1207
      (1981).
4.  Brouwer,  E. R. ; Lingeman, H. ;  and Brinkman, U. A. Th. ; Chromatographia
      29(9/10), 415-418 (1990).
5.  Hagen, D. F. ;  Markell, C. G.;  Schmitt, G.  A.; and Blevins D. D. ; Analytica
      Chimica Acta 236, 157-164  (1990).
LIST OF FIGURES

Fig. 1.  American Sigma water sampler, model 702, top cover removed.

Fig. 2.  Filter holder, MSI model DFN-P4SGS-S1.

Fig. 3.   Bottom  plate of  dosimeter with methanol  reservoir,  methanol pump,
      control circuitry and controller.

Fig. 4.  Diagram of dosimeter interconnections, dashed lines denote added parts.

Fig. 5.   Logic  diagram  of  control circuitry  for  dosimeter,  thin lines for
      original Streamline, bold lines  for  added  logic.
                                        1-257

-------
1-258

-------
   Remove Approx.
   1/32" lid on
        Bottom Surface
                                     Cover
                                     Funnel
MSI Disposable
Filter Funnel Modified
for use in Dosimeter
                                     Filter
                                     Media
                                     Base
                                     Accessory
                                     No.8 Stopper
                                     for Elution
                                     Flask

                                     Extender for
                                     use in Flask
                         I-259

-------
           Key Board
              and
            Display
          Streamline
           CUP P. C.
            Board
          Streamline
         Power Control
            P.C.B.
         Arm
         Ledex
         Motor
      i
            Auxiliary
             Battery
Dosimeter
Interface
P.C.  Board
                                             •
                                             t
                    T
Vaccum
Pump
	


Vaccum
Rotary
Value
	 1
                        Sample
                         Pump
    Methanol
      Pump
    These Connections
   Made Though Re-Hired
   Auxiliary Socket on
   Punp Control Assy.
Block Diagram of
In terconnec tions
Dashed Lines Denote
Added for Dosimeter
                              1-260

-------
             Flow  Chart far Control Crts.
  Delay  Til
 Restart Time
                         Take Sample
                            iOO mL
             Post Purge
                          Move Dist.
                         Arm  to Next
                            Pos.
                             Stop
Thin Lines

Bold Lines
Original Controls Crt.

Added Interface
                                           Move Arm
                                           to Pos.l
                                           Move Vac.
                                           Value to 1
 Move Vac.
Value to Next
   Pos.
                                            J.
                                          Reset
                                           LRi
                             1-261

-------
34
REPRESENTATIVE SAMPLING FOR THE REMOVAL PROGRAM
         William A.  Coakley,  Environmental  Response Team,  U.S.  Environmental
         Protection Agency,  Raritan Depot,  Edison,  New Jersey  08837;  Lauren Ray.
         Technical  Assistance Team,  Gregory MaiIon, Technical Assistance Team, Roy
         F.  Weston, 955  L'Enfant Plaza, Washington,  DC   20024;  Gregory  Janiec,
         Technical  Assistance  Team, Roy  F.  Weston,  Weston  Way, West  Chester,
         Pennsylvania  19380.
         ABSTRACT

         In addition to addressing catastrophic releases,  the  Removal  Program is
         involved in meeting objectives such as identifying the  threat a site poses
         to public health, welfare, and the  environment;  delineating  sources and
         extent of contamination;  evaluating  waste treatment and disposal options;
         and confirming the achievement of cleanup standards.   To help meet these
         ends, EPA assembled the U.S. EPA Committee on Representative Sampling for
         the  Removal  Program  to  develop  guidance  documents  for   collecting
         representative samples at removal sites.   The goal of a representative
         sampling plan is  to accurately depict variations in pollutant presence and
         concentration across a site  for a  given medium  (e.g., soil, groundwater).
         Five  representative  sampling guidance  documents, addressing  different
         environmental media, are in various stages  of  planning and development.
         Guidance documents for soil and air sampling are scheduled for publication
         in late FY91.  Guidance documents for biological sampling, waste sampling,
         and groundwater/surface water/sediment sampling will be completed during
         FY92.  Each document has information unique to its medium, but follows the
         overall  objectives  and  recommendations  for  effective  representative
         sampling.   The  documents address:   assessing   available  information;
         selecting an  appropriate  sampling approach (including  the selection of
         sampling locations);  properly selecting and utilizing sampling and field
         analytical  screening  equipment;  utilizing proper  sample  preparation
         techniques; incorporating suitable types and numbers of QA/QC samples; and
         interpreting and presenting  the resulting data.  Each document presents a
         case  study to  illustrate  how  a representative  sampling plan may  be
         developed to meet Removal Program objectives.   A representative sampling
         training program and an air  sampling methods database are  also  being
         developed.
         INTRODUCTION
         The EPA Removal Program

         Under  the  Comprehensive  Environmental  Response,   Compensation,   and
         Liability Act  of 1980 (CERCLA),  the  EPA may respond to a  release  (or
         threat of a release) of hazardous materials.  CERCLA authorized both long-
                                           1-262

-------
term  activities   (called   remedial   actions)   and  emergency  response
activities  (called removal actions), and  a Hazardous Waste  Trust Fund
("Superfund") to pay for them.

A removal action is a short term action  intended to stabilize or clean up
an incident or site which poses  a threat to public health or welfare or to
the environment.  Removal actions may include but are not limited to:

      •     removing and disposing of hazardous  substances;
      •     constructing  fences,  posting  signs,  and  other  security
            measures to restrict access to the  site;
      •     providing an alternative water source to local residents where
            water  supplies have become contaminated; and
      •     temporarily relocating residents.

CERCLA also  defined the duration and cost  of  removal actions.   Removal
actions  were limited  to six  months duration  and a  total  cost  of $1
million.  Exemptions would be required in cases where  the action exceeded
these time and cost limitations. In  1986 the Superfund Amendments and Re-
Authorization Act (SARA) changed the limitations to 1 year and $2 million.

In cases where imminent threats have been addressed but long  term threats
remain,  the  site   is  referred  to  EPA's Remedial  Program  for  further
assessment  and  investigation.    However,   remedial  actions are   only
conducted at sites that have been placed on EPA's National Priorities  List
(NPL).  Removal actions may be necessary at remedial sites if  an emergency
arises.   EPA has responded with multiple  removal actions  at sites  with
complex hazards.
Program Implementation

Superfund  program  implementation  is  guided  by  the  National  Oil and
Hazardous Substances Pollution Contingency Plan (NCP),  which outlines the
roles and responsibilities of the Federal agencies who respond to releases
of  hazardous substances.   The U.S. Coast  Guard has responsibility for
releases in  or near  coastal waters  and EPA  has responsibility over  those
which occur  inland.

The initial  step in a removal  action  is the discovery of  the  incident.
EPA may be notified by the National Response Center, which is operated  24
hours a day by the U.S. Coast Guard,  or be notified directly. Once EPA has
been notified, an EPA  On-Scene Coordinator (OSC) evaluates the situation.
Based on this evaluation, EPA decides whether Superfund money will be used
to  respond to the incident (if the responsible  party cannot or will not  do
so, or if state or local officials cannot or are unable to respond). EPA
notifies or  calls  for  assistance from  other agencies,  as  necessary.

Once EPA determines  that a removal  action is required,  the OSC assembles
the equipment and resources necessary  to respond.  During an initial site
                                   1-263

-------
assessment, air monitoring equipment help to determine the nature of the
on-site hazards.  After assessing on-site conditions, the OSC establishes
a  site  command post  and,  if  there is no  immediate  emergency,  begins
monitoring and sampling on-site materials or contaminants.  A variety of
media may  be  sampled,  including soil, air,  surface water,  groundwater,
waste piles, and drums or other containers.   Once sampling is completed,
the necessary equipment is mobilized to stabilize the site.  Stabilization
can include: building berms/dikes, establishing water treatment systems,
excavating contaminated soil, erecting fences, and other activities.
REPRESENTATIVE SAMPLING

Representative sampling is the degree that a sample  or a group of samples
accurately characterizes site conditions.  Representative samples reflect
the concentration of parameters  of  concern at a given time.  Analytical
results from representative samples  illustrate the variation in pollutant
presence and concentration across a site.

The U.S. EPA Committee on Representative Sampling for the Removal Program,
comprised of U.S. EPA,  state, and contractor representatives, is planning
and  developing  five  representative  sampling guidance documents,  each
addressing a different environmental medium.  Guidance documents for soil
and air sampling are scheduled  for publication  in  late  FY91.   Guidance
documents for biological sampling, waste sampling, and groundwater/surface
water/sediment sampling will be completed during FY92.  The  documents are
medium-specific, for ease of use, however, multimedia sampling is usually
necessary at  removal sites.   Each document covers  aspects unique to its
medium,  but  follows the  overall  objectives  and  recommendations  for
effective representative  sampling.    The documents  address:   assessing
available  information;   selecting   an  appropriate  sampling  approach
(including the selection  of  sampling locations); properly selecting and
utilizing sampling  and field  analytical  screening  equipment;  utilizing
proper  sample  preparation techniques; incorporating suitable  types and
numbers of QA/QC samples;  and interpreting and presenting the resulting
data.    The  air document  also addresses  analytical  techniques  and
atmospheric modeling.   The  documents address the  above  considerations
within the objectives and scope  of the Removal Program.

Each document presents  a  case study  to  illustrate  how a representative
sampling plan may be developed to meet Removal Program objectives.  The
case study illustrates the concepts  discussed in each  chapter.   For the
soil  guidance,   the case study illustrates how   "interactive"  field
analytical screening and  other  cost-effective field techniques  such as
geophysical surveys can be used to characterize a  site,  from selecting
sampling locations to confirming cleanup.
                                  1-264

-------
USE OF REPRESENTATIVE SAMPLING TO MEET REMOVAL PROGRAM OBJECTIVES

Although field conditions and removal activities vary from site to site,
the primary Removal Program sampling objectives include:

      1.    Establishing  Threat to Public  Health  or Welfare  or  to the
            Environment  -- CERC1A and  the NCP  establish  the  funding
            mechanism  and authority which  allow  the OSC  to activate a
            Federal removal action.  The OSC must establish that the site
            poses  a  threat  to public  health  or  welfare   or  to  the
            environment.   Sampling  is  often  required to document the
            hazards present on site.  The analytical data are  often needed
            quickly to activate  the removal action.

      2.    Locating and Identifying Potential Sources of Contamination --
            Sampling is conducted to identify the locations and sources of
            contamination.   The results are used  to  formulate removal
            priorities,  containment  and cleanup  strategies, and cost
            projections.

      3.    Defining  the Extent of Contamination  -- Where  appropriate,
            sampling is conducted to assess  horizontal and vertical  extent
            of  contaminant  concentrations.   The  results  are  used to
            determine the site boundaries (i.e. , extent of contamination),
            define  clean  areas,  estimate  volume  of contaminated  soil,
            establish  a  clearly defined  removal approach,  and accurately
            assess removal costs and  timeframe.

      4.    Determining  Treatment  and  Disposal Potions  --  Sampling is
            conducted  to  characterize waste and contaminated soil for in-
            situ or  other on-site treatment, or excavation  and  off-site
            treatment  or disposal.

      5.    Documenting  the  Attainment  of Cleanup Goals -- During or
            following  a  site  cleanup,  sampling is  conducted  to determine
            whether the  removal  goals or cleanup standards were achieved,
            and  to  delineate  areas  requiring   further treatment or
            excavation as appropriate.


 Development and Execution of a Sampling Plan

 The representative sampling guidance documents outline how a sampling plan
 can be  designed to meet these objectives.   The  sampling  plan  design
 consists of the following steps:

      •    Review existing historical site information,
      •    Perform a site reconnaissance visit,
      •    Evaluate  potential migration pathways,  receptors, and routes
            of exposure,
                                   1-265

-------
      •     Determine the sampling objectives,
      •     Utilize field screening techniques,
      •     Select parameters to be analyzed,
      •     Select an appropriate sampling approach, and
      •     Determine the locations to be sampled.

Unless the site is considered a classic emergency,  every effort is made to
first review relevant information concerning the site.   An historical data
review examines past and present site operations and disposal practices,
providing an overview of known and potential  site contamination and other
hazards.  Sources of  information include:  federal,  state, and local files
(e.g., prior site inspection  reports  and legal  actions);  facility maps;
blueprints; historical aerial  photography; and  interviews  with facility
owners and  operators,  current and former facility employees,  and local
residents.  ,-A  site  reconnaissance,  conducted  either  prior  to, or  in
conjunction with sampling, assesses site conditions,  evaluates areas of
potential  contamination, evaluates  potential  hazards associated  with
sampling,   and   helps  to  develop   a   sampling  plan.     During   the
reconnaissance, Removal  Program personnel observe  and photodocument the
site,  noting site  access routes,  process,  waste disposal,  and  other
potential contaminant source areas, and potential  routes for contaminant
transport off-site.

A representative sampling plan considers pollutant migration pathways,
receptors, and routes of exposure.  Pollutant migration pathways include
surface drainage, vadose zone and  groundwater transport,  air transport,
and human activity  (such as  foot or vehicle  traffic).   In urban areas,
man-made pathways,  such  as  storm and  sanitary sewers and underground
utility  lines   can  influence  contaminant  transport.   Human  receptors
include  children who can  come  into  direct  contact with  or  ingest
pollutants by  playing in a contaminated area.   Environmental receptors
include Federal- and state-designated endangered  or threatened species,
habitats  for these  species,  wetlands,  and  other Federal- and state-
designated wilderness, critical, and natural  areas.

The scope of the sampling program depends on  the Removal Program sampling
objectives previously discussed.  In order to  attain these objectives, the
quality  assurance  components  of  precision,  accuracy,   completeness,
representativeness,  and comparability are considered.

Samples are analyzed by an established and approved off-site laboratory,
or are screened or analyzed on-site using various portable direct reading
instruments. Field analytical screening equipment utilized by the Removal
Program  includes the  X-ray   fluorescence  (XRF) meter,  photoionization
detector  (PID),  flame ionization detector,  and field test kits.   Some
field  analytical  screening instruments,  such as  the  PID  and  some XRF
units,  can  be used in-situ  (without  collecting  a  sample).    Field
analytical screening methods may be utilized to narrow the possible groups
or classes of chemicals for laboratory analysis.  When used appropriately,
field screening can  cost-effectively  evaluate a large number of samples
                                  1-266

-------
for  the purpose  of  selecting a  subset  for  off-site  analysis.    The
detection limits  and accuracy of  the  screening method  is  evaluated by
sending a minimum of 10% of  the  samples to an  off-site laboratory  for
confirmation.  Field screening techniques and confirmatory sampling can be
used together to identify or delineate  the  extent of contamination and to
confirm cleanup.    Once a  contaminated area has  been  identified  with
screening  techniques,  an  appropriate  confirmatory  sampling  strategy
substantiates  the  screening results.   Used  in  tandem,  field analytical
screening and  confirmatory  sampling provide  data more representative of
the problem at the  site than off-site  laboratory analysis alone.   Field
screening is  also  utilized in the Removal  Program  for air monitoring
during removal activities and for on-site health and safety decisions.

Geophysical   techniques  such  as   ground  penetrating  radar  (GPR),
magnetometry,  electromagnetic conductivity (EM)  and resistivity surveys
may also be  conducted during  removal  actions.   Geophysical surveys, in
conjunction with  field  analytical  screening,  helps  delineate  areas of
subsurface  contamination,  locations  of  buried  drums   or tanks,  and
disturbed areas.   Geophysical data can be obtained relatively rapidly,
often without  disturbing the  site.

Locating  sampling  points  for field  screening  and  laboratory analysis
entails choosing  the most  appropriate  sampling  approach.  The sampling
objectives,  the  site  setting,   limitations in  the  sampling and  the
analytical   methods,   and   available   time  and  resources   are  all
considerations.   Representative  sampling approaches include judgmental,
random, stratified random, systematic grid, systematic random,  search and
transect methodologies.  A representative sampling plan may combine  two or
more of these approaches (as  defined  below).   Although  some  approaches
(such as judgmental and random sampling) are applicable  to a variety of
media,  it should be noted that systematic, search, and transect sampling
techniques are specific to  soil and sediment sampling.

      •     Judgmental  Sampling  - Judgmental sampling is the  subjective
            selection of sampling locations at a site, based on historical
            information, visual  inspection, and  on  best  professional
            judgment of the sampling team.  Judgmental  sampling is  most
            often used to identify the  contaminants present at the highest
            concentrations  (i.e., worst case conditions).   This is often
            the  basis   for  supporting the removal funding request.
            Judgmental  sampling has no randomization  associated with the
            sampling  strategy,   and   therefore  prevents   statistical
             interpretation  of the sampling results.

      •    Random Sampling -  Random sampling is the  arbitrary collection
            of samples within  the  area  of  concern.    Random   sample
             locations are chosen using a random selection procedure  (such
            as a  random number  table).    The  arbitrary  selection of
             sampling points requires each sampling point to be selected
             independent of  the location of all  other  points,  and results
                                   1-267

-------
in all locations within  the  area  of concern having an equal
chance of being selected.  Randomization is necessary in order
to  make  probability  or  confidence statements  about  the
sampling results.  Random  sampling  may  be performed for all
media,  however random sampling  assumes  that  the site  is
homogeneous with respect to  the parameters being monitored.
The higher the  degree  of heterogeneity,  the less the random
sampling approach will adequately characterize true conditions
at the site.   For  soil and sediment media (which are rarely
homogeneous) , other statistical sampling approaches (discussed
below)  provide  ways  to   subdivide  the  site  into  more
homogeneous areas, and may be more  appropriate than random
sampling  for  sampling soil  and sediment at  removal  sites.
Figure 1 illustrates random sampling.

Stratified Random Sampling -  Stratified  random sampling often
relies on historical  information and prior analytical results
(or field screening data) to stratify the sampling area.  Each
stratum  is more homogeneous  than the  site is as  a  whole.
Strata  can be defined based on various factors,  including:
sampling   depth,   contaminant  concentration   levels,   and
contaminant  source  areas.   Sample  locations  are  selected
within each of these  strata using random  selection procedures.
Stratified  random sampling  imparts  some control upon  the
sampling scheme (e.g., collection  of more  samples from depths
or areas having higher contaminant concentrations) but still
allows  for random sampling within  each stratum.   Different
sampling  approaches  may also  be selected  to address  the
different  strata  at  the  site.    Figure  2 illustrates  a
stratified random sampling approach for  soil where strata are
defined based  on depth.

Systematic Grid Sampling - Systematic grid sampling (of soil
and  sediment)  involves  subdividing  the area of  concern by
using a square or triangular grid  and collecting samples from
the nodes  (intersections of the grid lines).   The distance
between  sampling  locations   in   the  systematic  grid  is
determined by the size  of the area to be sampled and  the
number  of samples to  be collected.   Systematic  grid  soil
sampling  is often  used  to  meet  Removal  Program  sampling
objectives for locating and identifying potential sources of
contamination,  defining the  extent  of  contamination,  and
documenting  the  attainment  of  cleanup  goals.   Figure  3
illustrates a systematic grid sampling approach.

Systematic Random Sampling  -  Systematic random sampling (of
soil and sediment)  involves subdividing the area of concern by
using a square or triangular  grid  and collecting samples from
within   each   cell  using  random   selection   procedures.
Systematic random sampling  is a useful and flexible design for
                      1-268

-------
                       FIGURE 1 - RANDOM SAMPLING
8
to
                100-

                75-
             • -
             UJ
             uj  50-


                25-
                       I    I    I    I    I    I    I    I    I
                      25  50  75  100  125  150  175  200  225
                                     FEET
                                                    FIGURE 3 - SYSTEMATIC GRID SAMPLING
100-
75-
uj 50-
LL
25-

(j

f*~


	 	 £



r — 1
8-,


5 — j

1


h p ^

I
3!



1 i
Bl

3 r

1 '
31
- i

[\
J 	 *"
--!
fc**^*"^
D
y
                                                                        i    i    I    i    r
                                                        25   50   75  100 125  i50 175 200 225
                                                                       FEET
                   After: U.S. EPA. 1989. EPA/230/02-89-042
FIGURE 2 - STRATIFIED RANDOM SAMPLING      FIGURE 4 - SYSTEMATIC RANDOM SAMPLING
             S
             OC L
             o.
             ..
 0 •
1.5-
 3-
4.5-
 6

  i
                         STRATUM 1
                         STRATUM 2
100-

75-

50-

25-
       I        I    I    I    I    I    I    I
      25  50  75  100  125  150  175  200  225
                      FEET
                                               m
                                        LEGEND
                                   SAMPLE AREA BOUNDARY
                                   SELECTED SAMPLE LOCATION
                                   SAMPLE LOCATION
                                                                                        After: U.S. EPA. 1989, EPA/230/02-89-042

-------
            estimating the  average  pollutant concentration within each
            cell of the grid.  Also,  systematic random sampling allows for
            the isolation of cells  that may  require additional sampling
            and  analysis.    Figure  4  illustrates  a systematic  random
            sampling approach.

            Search Sampling  - Search sampling utilizes either a systematic
            grid or systematic  random sampling approach to search for hot
            spots  (i.e., areas  where  contaminants  in soil  and sediment
            exceed applicable cleanup  standards).   The number of samples
            collected and the  grid spacing  used  are determined  on the
            basis of the acceptable  error (i.e., the chance of missing a
            hot  spot).    When  conducting   search  sampling,   initial
            assumptions must be made about the size, shape, and depth of
            the hot spots to be searched for.   The smaller and/or narrower
            the hot  spots are,  the  smaller the grid spacing  (i.e. , the
            more samples) necessary to locate  them.  In  addition, the
            smaller the acceptable error of missing hot spots, the smaller
            the grid spacing must be.

            Transect Sampling  -  Transect  sampling involves establishing
            one or more transect lines across the surface of a site.  Soil
            or sediment samples are collected at regular intervals along
            the transect lines at the  surface and/or at one or more given
            depths.  The spacing between sampling points along a transect
            is   determined  by  the  length of the  transect  line  and the
            number  of samples  to be collected.   Multiple transect lines
            may be parallel or non-parallel to one another.   A primary
            benefit of transect sampling over systematic grid sampling is
            the ease of  establishing  and relocating individual transect
            lines versus an entire grid.   Transect  sampling is often used
            to  delineate the   extent  of  contamination and  to  define
            contaminant concentration gradients.   Transect sampling has
            also been used,  to  a lesser  extent,  in compositing sampling
            schemes.
Selection and Use of Sampling Equipment

The manner  in which a  sample  is collected  is  based on  the  objectives
stated in the site-specific sampling plan.  Sample collection requires an
understanding of the capabilities of the sampling equipment in obtaining
a sample which  accurately  depicts current site conditions.   The  use of
inappropriate equipment (or the incorrect use of sampling equipment) may
result in biased samples.

The mechanical  method by  which  a sampling  tool  collects a  sample may
impact  sample  representativeness.    For example,  if  the goal  is  to
determine  the  concentrations  of  contaminants  at  each  soil  horizon
interface, using  a hand auger would  be inappropriate.   Obviously, the
                                  1-270

-------
augering technique would disrupt and mix soil horizons, making the precise
horizon  interface difficult  to determine.    In addition,  all  sampling
devices used are of sufficient quality to not contribute contamination to
samples (e.g., painted surfaces which could  chip off  into the sample) and
the sampling equipment are either easily decontaminated, or cost-effective
if considered to be expendable.
Sample Preparation

Field sample  preparation includes all aspects  of  sample handling after
collection, until the  sample is  received by  the laboratory.   Sample
preparation techniques  are specific  to  the sample medium  and sampling
plan.  For soil  and  sediment  sample preparation, common techniques are:
removing  extraneous  material,  sieving,   homogenizing,  splitting,  and
compositing samples.

Proper sample preparation  and handling maintains  sample  integrity and
provides  a representative  sample  from the  total material  collected.
Improper handling can result in a sample becoming unsuitable for the type
of analysis required. For example, homogenizing, sieving, and compositing
of  soil   samples all  result   in a  loss   of  volatile  constituents  and
therefore, are inappropriate when volatile contaminants are of concern.

Homogenization is the mixing or  blending of a soil  or sediment sample in
an  attempt  to provide  uniform distribution of contaminants.   Ideally,
proper homogenization assures  that portions of  the  containerized samples
are equal or identical in composition and are representative of the total
sample collected.  Incomplete  homogenization may increase sampling error.
Quartering, as per ASTM Standard C702-87,  can be used to simultaneously
homogenize and split a sample.  Split samples are most often collected in
enforcement actions  for comparing  sample  results  obtained by  EPA with
those obtained by the potentially responsible party (PRP).  Split samples
also provide  a measure of  the sample  variability,  and  a measure of the
analytical and extraction errors.  Split  soil  and sediment samples are
commonly collected.  Splitting may also be  done, in some  cases, with water
and air samples.

Compositing  is  the  process  of physically  combining  and  homogenizing
several  individual  soil and  sediment aliquots.   Compositing  samples
provides an average concentration of contaminants over a certain number of
sampling  points,  which  reduces both the number of  required lab analyses
and the sample variability.  Compositing can  be a useful technique, but
must always be considered and implemented with caution.   Since compositing
dilutes high concentration aliquots,  the applicable  detection limits
should be reduced accordingly.   If the composite value  is to be compared
to a selected action level, then the action level must be divided by the
number of aliquots that make  up the composite  in order to determine the
appropriate detection  limit (e.g.,  if the  action level for a particular
substance  is  50 ppb,  a detection limit of  10 ppb should  be  used when
                                   1-271

-------
analyzing a 5-aliquot composite).

To help maintain  sample integrity and assure  representativeness,  it is
sometimes possible  to ship  samples  to the  laboratory directly  in the
sampling equipment.  This is the  case  for  air  sampling media.   For soil
core samples, the ends of a  Shelby tube can  be sealed with caps,  taped,
and sent to the laboratory for analysis.   To help maintain the integrity
of VOA soil  samples, soil cores can be collected using acetate sleeves and
sent in the  sleeves to  the  laboratory.  To  ensure  the integrity  of the
sample after delivery  to the laboratory,  laboratory sample preparation
procedures are made part of laboratory bid contracts.
Quality Assurance/Quality Control (OA/OC)

Quality assurance/quality control (QA/QC) measures are an integral part of
representative  sampling.   QA/QC  samples  provide  information on  the
variability  and usability of environmental  sample results.   They also
evaluate  the  degree  of  site variation,  whether  samples were  cross-
contaminated  during sampling and sample handling  procedures,  or  if  a
discrepancy in  sample results is due to laboratory handling and analysis
procedures.

In  the Removal Program, field  replicate,  collocated,  background,  and
rinsate blank samples are the most commonly collected field QA/QC samples.
Performance evaluation, matrix spike,  and matrix spike duplicate samples,
either prepared for or by the laboratory, provide additional measures of
control for  the data  generated.   QA/QC results may suggest the need for
modifying  sample   collection,   preparation,   handling,  or  analytical
procedures  if  the  resultant  data do not  meet site-specific  quality
assurance objectives.

Three  QA/QC  objectives  (QAl,  QA2, and  QA3) have  been defined by the
Removal Quality Assurance Program, based on the EPA QA requirements for
precision, accuracy, representativeness,  completeness, comparability, and
detection level.  QAl standards are used when a large amount of data are
needed quickly and relatively inexpensively, or when preliminary screening
data,  which  do not need to be  analyte  or concentration  specific,  are
useful.   QAl  requirements  are  used  with data  from  field  analytical
screening  methods   for  a   quick,   preliminary  assessment   of   site
contamination.  Examples of  QAl activities  include:  determining physical
and/or  chemical properties  of  samples;  assessing preliminary  on-site
health and safety;  determining the extent and  degree  of contamination;
assessing waste compatibility; and characterizing hazardous wastes.

QA2 verifies  analytical results.  The QA2 objective is intended to provide
a certain level of confidence for  a select portion (10% or more)  of the
preliminary data.   This objective  allows  Removal Program  personnel to
focus on specific pollutants  and concentration levels  quickly,  by using
field screening methods with laboratory verification and  quality assurance
                                  1-272

-------
for at least  10%  of the samples.  QA2 verification methods  are analyte
specific.   Examples  of QA2 activities  include:    determining  physical
and/or chemical properties of samples; defining the extent and degree of
contamination; verifying site cleanup; and verifying screening objectives
obtainable with QAl, such as pollutant identification.

QA3 assesses the accuracy of the concentration level, by determining the
analytical error as well  as  the identity of the analyte(s) of interest.
QA3  data provide  the  highest  degree of qualitative  and quantitative
accuracy of  all QA objectives  by using  rigorous  methods  of laboratory
analysis and quality assurance.  QA3  is  intended to provide a high level
of  confidence  so  that  the decisions   can  be  made  with  regard  to:
treatment; disposal; site remediation  and/or removal of pollutants; health
risk or environmental impact; cleanup verification; and pollutant source
identification.
Sources of Error

Quantifying  the  error  associated  with  a  sampling activity  can  be
difficult, but is important  in order  to  identify the possible sources of
error or variation in sampling and laboratory analysis and to limit their
effect(s) on the data.  Four potential sources of error are:

      •     Sampling design  -- Site variation (the non-uniform conditions
            which   exist  at  a  site)    include   the  identities  and
            concentration   levels  of   contaminants.     The   goal  of
            representative sampling is to  accurately  identify and define
            this  variation.    However,   error  can  be  introduced  by the
            selection of a sampling design which "misses"  site variation.
            For example, a sampling grid with relatively  large distances
            between  soil sampling points  or a biased sampling approach
            (i.e., judgmental sampling) may allow significant contaminant
            trends to go unidentified.

      •     Sampling methodology -- Error can be introduced by sampling
            methodology and sample handling  procedures, as  in cross-
            contamination  from  inappropriate use  of sample collection
            equipment, unclean sample  containers, improper decontamination
            and  shipment procedures, and other factors.   The  use of
            standard  operating procedures for collecting, handling, and
            shipping  samples  allows  for  easier identification  of the
            source(s)  of  error,  and can  limit error  associated with
            sampling methodology.  Trip blanks, field blanks, replicate
            samples, and rinsate blanks  are used to identify error due to
            sampling methodology and  sample handling  procedures.

      •     Sample  heterogeneity -- Sample heterogeneity is a potential
            source  of error, especially for soil and sediment samples.
            These  media  are  rarely homogeneous  and exhibit variable
                                   1-273

-------
            properties  with  lateral  distance  and  with depth.    This
            heterogeneity may also be present  in the  sample  container
            unless the  sample was homogenized  in the  field or  in the
            laboratory.   The laboratory uses only a small aliquot of the
            sample for analysis;  therefore thorough sample homogenization
            is  important to  assure  that  the   analytical  results  are
            representative of the sample and of the corresponding site.

      •     Analytical  procedures  --  Error  which  may  originate  in
            analytical    procedures    includes    cross-contamination,
            inefficient extraction, and inappropriate methodology.  Matrix
            spike  samples,  replicate samples,  performance  evaluation
            samples,   and associated  quality  assurance  evaluation  of
            recovery, precision,  and accuracy, can be used to distinguish
            analytical  error  from  error  introduced  during  sampling
            activities.

A  common objective  of   the  evaluation  of  soil analytical  data  is  to
delineate the extent of site contamination.   One cost-effective approach
used in  the Removal  Program is  to correlate inexpensive field screening
data with laboratory results.  When field screening techniques,  such as
XRF, are used  along with  laboratory methods  (e.g., atomic  absorption
(AA)),  a regression  equation  can be used to predict a  laboratory value
based  on the screening  results.   The model can also be used  to place
confidence limits  around predictions.  The  relationship between the two
methods  can  be described by  a  regression analysis and used  to predict
laboratory results based on field screening measurements.  The predicted
values can then be located on  a base map and contoured.   These maps can be
examined  to  evaluate  the  estimated  extent  of  contamination and  the
adequacy of the sampling program.
Data Presentation and Analysis

Data  presentation  and  analysis  techniques  can  be  used  to  compare
analytical  values,   to  evaluate  numerical  distribution  of  data,  to
determine and  illustrate the  location  of hot  spots  and the  extent of
contamination  across  a  site,  and  to assess  the  need  for removal of
contaminated soil with concentrations at or near the action level.  Data
presentation and analysis methods include:

      •     Data posting -- Data posting, the placement of sample values
            on  a site  basemap,  is  useful for  displaying the  spatial
            distribution of  sample values to visually depict  extent of
            contamination and to locate hot spots.

      •     Geologic graphics --  Geologic graphics include cross-sections
            and  fence  diagrams,  two- and three-dimensional  depictions,
            respectively, of soils  and strata to a given depth beneath the
            site.   These  types  of  graphics  are  useful  for  posting
                                  1-274

-------
            subsurface  analytical  data,  for  interpreting  subsurface
            geology and contaminant migration,  and for placing monitoring
            wells.

      •     Contour mapping -- After depicting sample values on a basemap
            with data posting, contour lines (or  isopleths) can be drawn
            at a specified contour interval.  Interpolating values between
            sample points and drawing contour lines is done manually or by
            using computer contouring software.   Contour maps are useful
            for depicting soil and groundwater contaminant concentration
            values across a site.

      •     Statistical graphics  --  The histogram  is  a statistical bar
            graph which displays the  distribution  of a data set.  A normal
            distribution of data is a bell-shaped curve, with the mean and
            median close  together about  halfway between the maximum and
            minimum values. A probability plot depicts  cumulative percent
            against the concentration of the contaminant of concern.   A
            normally distributed data set plotted as a probability plot
            would appear as a straight line.  A histogram or probability
            plot show trends and anomalies  in the data prior to conducting
            more  rigorous   forms   of  statistical  analysis.     Common
            statistical analyses  such as  the  t-test and the regression
            analysis rely on normally distributed data.  The distribution
            or spread of  the  data set is important in determining which
            statistical techniques to use.

      •     Geostatistics - - A geostatistical analysis  can be broken down
            into two phases.  First,  a model is developed that describes
            the spatial relationship  between sample locations on the basis
            of a  plot of  spatial variance versus  the  distance  between
            pairs of samples.  This plot is called a variogram.   Second,
            the spatial relationship  modeled by the variogram is used to
            compute a weighted-average interpolation of  the data.   The
            result of geostatistical mapping by data interpolation is a
            contour map that represents estimates of values across a site,
            and maps  depicting potential  error  in the  estimates.   The
            error maps are useful for deciding if additional samples are
            needed and  for  calculating best  or worst case scenarios for
            site cleanup.

The  data  interpretation  method  chosen  depends  on  project-specific
considerations,  such as  the  number  of  sampling  locations  and  their
associated range  in values.  A site depicting extremely low  soil data
values (e.g., non-detects) with significantly higher values (e.g.,  5,000
ppm) from neighboring hot spots with  little or no concentration gradient
in-between,  does  not  lend  itself  to   contouring  and  geostatistics,
specifically the development of variograms.  However, data posting would
be  useful at  such  a  site to  illustrate  hot   spot  and clean  areas.
Conversely, geostatistics and contour mapping,  as well as data posting,
                                   1-275

-------
can be  applied to site data with a wide distribution  of  values (e.g.,
depicting a "bell shaped" curve) with beneficial results.
Incorporating Representative Sampling into the Removal Program

Although  the principles  discussed  here are  utilized  in the  Removal
Program, there is no national consistency with how they are  employed.  The
first step to representative sampling consistency in the Removal Program
is the completion of guidance documents for each environmental medium.  In
order to keep the guidance documents  current while sampling methodologies
are evolving, EPA has placed the documents on a two-year update schedule.

The  second  component  of incorporating the  concept  of  representative
sampling into the Removal Program is the development of a training program
for site personnel, based on the guidance documents.  The training course
will introduce the concepts  presented in the documents, structured around
a realistic example site (an actual Superfund site).  One common site is
being  incorporated  into each  document.    This   will  facilitate  the
development of an  integrated training program addressing all media.

To enhance  the integration  of  representative  sampling into the Removal
Program, the guidance documents  discuss computer software that assist in
implementing the concepts presented  in  the documents.  This includes the
use  of  EPA's    Geo-EAS  program  (Geostatistical  Modeling  Assessment
Software) and its  application to soil sampling in the soil document, and
an evaluation of available  air  models to assist in sample design in the
air document.

Finally,  specific  tools  will  be  developed and  incorporated  into the
documents as necessary.   In  the air document, an  air sampling methods
database has been developed  to provide up-to-date information on sampling
methods and compounds that can be sampled by those methods.
SUMMARY

The U.S. EPA Committee on Representative Sampling for the Removal Program
is planning and  developing  guidance  documents to assist Removal Program
personnel in  the collection of representative samples  at removal sites.
Five  guidance   documents,  addressing  soil,  air,  biota,  waste,  and
groundwater/surface water/sediment sampling will be published in FY91 and
FY92.  Each document  addresses considerations unique  to its medium, but
follows  the  overall  objectives  and  recommendations  for  effective
representative sampling.   The  documents address:   assessing available
information;  selecting  an  appropriate  sampling  approach;  properly
selecting  and   utilizing  sampling,  field  analytical  screening,  and
geophysical equipment;  utilizing proper  sample  preparation techniques;
incorporating  suitable  types  and  numbers  of QA/QC  samples;  and
interpreting and presenting the resulting data.
                                  1-276

-------
35              PRELIMINARY FIELD AND LABORATORY EVALUATIONS
                AND THEIR ROLE IN AN ECOLOGICAL RISK ASSESSMENT
                 FOR A WETLAND SUBJECT TO HEAVY METAL IMPACTS
         Greg Linder. FIT Environmental Services, Lake Oswego, OR, Mike Bollman, Suean
         Ott, Julius Nwosu, David Wilborn, ManTech Environmental Technology, Incorporated,
         Bill Williams, US EPA Environmental Research Laboratory,  Corvallis OR.

         ABSTRACT

         An integrated laboratory/field project was initiated by Environmental Protection Agency
         [US EPA] Region 8 as part of their ecological risk assessment for the Milltown Reservoir
         Superfund  site located on the Clark Fork River in western Montana, six miles east of
         Missoula, Montana. Preliminary work supporting the ecological risk assessment included
         field studies completed at Milltown Reservoir, while laboratory work [biological testing
         and chemical analysis]  was completed  at the US  EPA  Environmental Research
         Laboratory-Corvallis  [ERL-C],  Corvallis,  Oregon.   For the wetlands evaluation at
         Milltown Reservoir, heavy metals appear the most critical contaminant which must be
         considered in the ecological assessment; those of primary interest include arsenic, zinc,
         copper, cadmium, and nickel as well as manganese and iron. Preliminary laboratory and
         field investigations evaluated the extent of contaminant and its impact on the indigenous
         wildlife and vegetation characteristic of the site.  Field work included scoping activities;
         the identification of sampling units; preliminary sampling of surface water, soil, and
         sediment; and preliminary field screening tests.  Results from the preliminary studies
         indicate  ecological effects may be subtle in expression, and future work should  focus
         upon current as well as historic sediment deposition areas in the reservoir and associated
         wetlands.

         INTRODUCTION

         Milltown Reservoir is located on the Clark Fork River in western Montana, six  miles
         east of Missoula, Montana [US EPA Region 8]. The reservoir was formed  in 1907
         following the construction of a hydroelectric facility located  on the Clark Fork  River
         immediately downstream from  its confluence with  the Blackfoot  River.    Since
         construction of the dam, a wetland habitat has been created, but because of the upstream
         mining activities on the Clark Fork River, Milltown Reservoir has accumulated a large
         volume of heavy metal-contaminated sediment.   The Milltown Reservoir wetland was
         initially  identified  under  CERCLA  in 1981 after community  well-water samples were
         found to  have  arsenic  levels that ranged from 0.22 to 0.51  mg/L;  the  EPA
         recommendation for potable water supplies suggested that arsenic not exceed 0.05  mg/L
         [Woessner, et al.  1984]. Within an ecological context, however, the impact of the
         contaminated sediments on the wetlands is unclear [Adamus and  Brandt 1990; Tiner 1984].
                                             1-277

-------
The laboratory and field investigations completed during FY90 evaluated the extent of
contaminant and its impact on the indigenous wildlife and vegetation characteristic of the
site.  During the preliminary season, field work was completed at Milltown Reservoir
and supporting laboratory work [biological testing and chemical analysis] was completed
at the  Environmental Research  Laboratory-Corvallis  [ERL-C],  Corvallis,  Oregon.
Preliminary field work included scoping activities, such as the identification of sampling
units; preliminary sampling of surface water,  soil, and sediment; and preliminary field
screening tests [NSI 1989]. In addition to field methods [e.g., earthworm, seed, and
amphibian evaluations] being used as part of the biological and ecological assessment,
complementary laboratory evaluations were completed [Greene, et al. 1989].  Routine
water quality measurements on surface water  samples and  soil  measurements [e.g.,
texture analysis] were also  collected as part of the preliminary field activities.

SYNOPSIS: PRELIMINARY FIELD AND LABORATORY ACTIVITIES

Sampling plan.  Historic information regarding Milltown suggested that a preliminary
field effort could  well  determine the extent of contamination, and  definitive field
operations planned for FY 91 could be focussed along habitat lines suggested by our
preliminary field season.  Clearly, the sedimentation problems of the reservoir would
require those matrices being evaluated in our definitive studies, but little previous work
had considered the impact  of periodic inundations upon upland habitats.  Accordingly,
the sampling plan which  guided our preliminary field season was  designed  within
topographic and  historic bounds and assured the definitive sampling and analysis plan
guiding our FY 91 field  season be developed on defensible empirical grounds. For the
preliminary field effort,  Milltown was stratified into sampling units based largely on
topography, and line transects were established on each sampling unit. Each line transect
was derived from an initial random vector, then defined along habitat gradients amenable
to each of the evaluations  being completed on  the sample units; plant and earthworm
methods were applied on the upland features, and amphibian methods were applied in
emergent zones [Schweitzer and Santolucito 1984].

Laboratory and field testing:  Vegetation evaluations. Within laboratory settings,
critical developmental stages  in plant life cycles were evaluated and  physiological
endpoints  pertinent to ecological impact were measured [Linder,  et al.  1990].  Seed
germination evaluations  were completed to evaluate soils  directly in the laboratory
without preparing eluates [Thomas and Cline 1985]; root elongation tests were completed
on site-soil eluates [Greene, et al. 1989].  On-site seed germination evaluations also
considered the germination endpoint, but were less time consuming, generally more cost-
effective and minimized  the manipulation of the site-soils, since tests were performed
directly in the field  [NSI  1989].  Data derived from  on-site testing complemented
terrestrial laboratory tests and chemical analysis of site samples. The on-site evaluations
also addressed questions regarding lab-to-field extrapolation.
                                     1-278

-------
Earthworm  testing.   Integrated  field and  laboratory  methods  using  earthworms
contributed to soil evaluations and afforded a direct test of an environmental matrix
which  may   greatly  influence  the  impact  of  soil  chemicals  on   indigenous
macroinvertebrate communities [Rhett, et §1. 1988; Marquenie, et al.  1987].  Adverse
biological or ecological effects may be expressed owing to contaminant-related effects
associated  with anthropogenic chemicals.  Or, physical alterations in  habitat  may
effectively impact terrestrial or wetland systems through direct and indirect mechanisms.
Within ecological evaluations for terrestrial and wetland habitats, then,  biological and
ecological measurements in general must assess and, if possible, distinguish between
effects  mediated by physical alterations  in  habitat and those effects mediated by
anthropogenic chemicals or contaminants associated with human activities. Additionally,
lab-to-field extrapolation bias was addressed through the integrated field  and laboratory
work completed with earthworms.

Amphibian testing.  The  integrated  laboratory and in  situ  methods using amphibians
contributed to an evaluation of the extent of contamination  as well  as lab-to-field
extrapolation error.  For the preliminary field season, in  situ evaluations were completed
at selected emergent zones at Milltown Reservoir and used field-collected eggs and  early
embryos of Rana catesbeiana [NSI  1989].  Grab samples of surface waters at Milltown
Reservoir were also collected as part of the preliminary field season  for the Milltown
Reservoir endangerment assessment; laboratory evaluations using standardized amphibian
methods [FETAX] were completed as parallel and complementary components to  these
in situ amphibian evaluations [ASTM 1991; Dawson, et al.  1988].

RESULTS AND DISCUSSION

From the preliminary field season completed at Milltown Reservoir:

 +     There are no indications that  acute effects are associated with any presumptive
       contaminant exposures on those areas surveyed and sampled during FY 90.

 +     Occasionally, biological tests  suggested that subacute or chronic effects may be
       expressed at  some locations  on the Milltown  wetlands; these  expressions of
       adverse biological effects appeared  to be associated with deposition zones where
       sediments either currently  or historically had  accumulated as a function of
       changes in the Clark Fork channel  or flow rates.

 +     Samples characterized by  adverse biological responses in  laboratory or field
       exposures were frequently  identified  by more  than one test method;  adverse
       biological responses in laboratory tests  should be evaluated for laboratory-related
       manipulation "effects," particularly when their  field analogs yielded dissimilar
       endpoints.
                                      1-279

-------
+    Potential sample heterogeneity may be minimized by stratifying future sample
      plans along topographic boundaries determined by current or historic depositional
      areas on the Clark Fork.

+    The significance of Blackfoot River input at its confluence with the Clark Fork
      could not be determined from these preliminary studies.

+    Some soil and sediment samples from the Milltown wetlands, again in current or
      historic  zones  of deposition, presented metals  concentrations which may be
      considered elevated [Beyer 1990].

+    Biological, and presumptively ecological, impacts associated with these elevated
      metal concentrations were not overtly expressed, and the biological information
      collected during  the  preliminary  field  season  should  be considered when
      interpreting these total metals concentrations.

Within an ecological context, these  in situ and on-site biological methods have proven
to be applicable to integrated field  and laboratory studies  that evaluate,  not only the
current status of the wetland, but also provide information which will contribute to future
mitigation and restoration efforts.

SUMMARY

While the work completed  in  FY 90 represents  preliminary  efforts in evaluating
ecological risks,  the  information garnered  can  focus  future work for the Milltown
Reservoir assessment.  While overt expressions of toxicity are not expressed in the
wetlands, these technical items were identified as significant considerations for discussion
in designing field and laboratory evaluations that could contribute to future Milltown
Reservoir work.

+     "Reference  areas"  must  be  defined  for  future  work  at  Milltown;  some
      comparative framework must be established, be that  "site-equivalent areas,"
      within-boundary reference locations  or an adequate historical  data  base,  for
      evaluating the  information generated in the laboratory or field.

+    Within ecological contexts, the area being sampled must be extended to include
       the entire Clark Fork Arm and other upstream areas suspected of having more
       recently deposited and potentially heavy metal-contaminated sediments.

+     Sediment evaluations  within the reservoir must  be completed to adequately
       evaluate the wetland; these evaluations should consider metals concentrations in
       sediments,  evaluations  of  sediment  toxicity  [including  evaluations  of its
       physicochemical properties], and a field survey  which would yield ecologically
                                      1-280

-------
      relevant information such as community structure.

+     Sediment evaluations should include invertebrate work [e.g., Timmermans and
      Walker 1989], and if possible, wetland plants [floating and emergent] should be
      evaluated for the  potential adverse biological effects associated with sediment
      exposures [Federal  Interagency Committee  for Wetland Delineation 1989;
      Fleming, el al 1988; Reed  1988; Walsh, gj al. 1990].

+     For more adequate site-specific evaluations regarding bioavailability of metals in
      soils  and sediments, additional  physicochemical   characterizations  may be
      advantageous; for example, to adequately evaluate vegetation responses, routine
      soil texture,  nutrient [N-P-K] and total organic carbon [TOC] analysis may be
      beneficial [Chapman and Pratt  1961; SCS 1951; Vandecaveye 1948].

+     If total metals appear inadequate for evaluating metal bioavailability, speciation
      studies may be indicated; if such studies are anticipated, site samples targeted for
      these analyses should be selected after a thorough review of the current, as well
      as historic, geochemical information.

REFERENCES

Adamus, P.R., and K. Brandt.   1990.  Impacts on  quality of inland wetlands of the
      United States: a survey of indicators, techniques, and applications of community
      level biomonitoring  data. [EPA/600/3-90/073].  U.S. Environmental Protection
      Agency, Environmental Research Laboratory, Corvallis,  Oregon, 97333.

ASTM. 1991.  Draft standard guide for conducting the frog embryo teratogenesis assay-
      Xenopus [FETAX].   American  Society  for Testing  and Materials, ASTM
       Committee E-47 on Biological Effects and Environmental Fate, Philadelphia, PA.

Beyer, W.N.  1990.  Evaluating  soil contamination.  U.S. Fish and Wildlife Service,
       Biol. Rep. 90[2].  25pp.

Chapman, H. D. and P . F. Pratt.  1961.  Methods for the analysis of soils, plants, and
       waters.  Univ. of California, Division of Agricultural Sciences.

Dawson, D.A., E.F. Stebler, S.L. Burks, and J.A.  Bantle.  1988.  Evaluation of the
       developmental toxicity of metal-contaminated sediments using short-term fathead
       minnow and frog embryo-larval assays.  Environ. Toxicol. Chem.  7[l]:27-34.

Federal Interagency Committee for Wetland Delineation.    1989. Federal manual for
       identifying  and  delineating jurisdictional  wetlands.   U.S.  Army Corps  of
       Engineers,   U.S.  Environmental Protection Agency, U.S. Fish  and  Wildlife
                                      1-281

-------
      Service, and U.S.D. A. Soil Conservation Service, Washington, D.C. Cooperative
      technical publication. 76pp.

Fleming, W.J., JJ. Momot, and M.S. Ailstock.  1988.  Bioassay for phytotoxicity of
      toxicant to sago pondweed. Chesapeake Res. Consort. Publication 129. pp. 431-
      440.

Greene, J.C., C. Bartels, W. Warren-Hicks, B. Parkhurst, G. Linder, S. Peterson, and
      W. Miller.  1989. Protocols for Short Term  Toxicity Screening of Hazardous
      Waste Sites. US EPA Research Laboratory, Corvallis, OR. EPA/600/3-88/029.

Linder,  Greg, J.C. Greene,  H. Ratsch, J. Nwosu, S. Smith, and D. Wilborn.  1990.
      Seed  germination and root elongation toxicity tests in hazardous  waste site
      evaluation: methods  development and applications.  In  Plants for Toxicity
      Assessment, ASTM STP  1091.   W. Wang, J.W. Gorsuch,  and W.R. Lower,
      Eds., American Society for Testing and Materials, Philadelphia, PA. Pp. 177-
       187.

Marquenie, J.M., J.W.  Simmers,  and S.H.  Kay. 1987.  Preliminary assessment of
      bioaccumulation of metals and organic contaminants at the Times Beach confined
      disposal site, Buffalo, N.Y. Miscellaneous Paper EL-87-6, Dept. of the Army,
      WES, Corp of Engineers, Vicksburg, MS, 67pp.

NSI Technology Services Corporation.  1989.  FY 89 Report: Initial performance
      evaluation of three bioassays modified for direct in situ testing.  Technical report
       submitted to Ecotoxicology Branch,  Ecological  Assessment Team,  US EPA,
      Environmental Research Laboratory, Corvallis, OR.

Reed, P.B.,  Jr.  1988.   National list of plant species that occur in wetlands: national
       summary.  U.S. Fish Wildl. Serv. Biol. Rep.  88(24). 244pp.

Rhett,  R.G., J.W.  Simmers and  C.R.  Lee.  1988.   Eisenia foetida  used  as  a
       biomonitoring tool to predict the potential bioaccumulation of contaminants from
       contaminated dredged material.  In Earthworms in Waste and Environmental
       Management.  Eds:  C.A.  Edwards  and  E.F. Neuhauser.  SPB  Academic
       Publishing, The Hague, The Netherlands.

Schweitzer,  G. E. and   J. A.  Santolucito. 1984.   ACS  Symposium Series 267:
       Environmental Sampling  for Hazardous Waste. American  Chemical Society,
       Washington, D.C.

Soil Conservation Service.  1951. Soil Survey Manual. U.S. Department of Agriculture
       Handbook No. 18.  Washington, D.C. 503pp.
                                     1-282

-------
Thomas, J. M. and J. E. Cline.  1985.  Modification of the Neubauer technique to assess
      toxicity of hazardous chemicals in soils.  Environ. Toxicol. Chem. 4:201-207.

Timmermans, K.R., and  P.A.  Walker.  1989.  The fate of trace metals during the
      metamorphosis of chironomids (Diptera, Chironomidae). Environ. Pollut. 62:73-
      85.

Tiner, Jr., R.W.  1984. Wetlands of the United States: Current status and recent trends.
      U.S.  Fish and Wildlife Service,  Habitat Resources,  One  Gateway  Center,
      Newton, MA. 02158.

Vandecaveye, S. C.  1948.  Biological methods of determining nutrients of soil.  In:
      H.B. Kitchen [ed.], Diagnostic  Techniques for Soil and Crops. American Potash
      Institute, Washington, D.C.

Walsh, G.E., D.E. Weber, T.L. Simon, L.K. Brashers, and J.C. Moore. 1990. Use of
      marsh plants for toxicity testing of water and sediments.  Contribution No. 694,
      Environmental Research Laboratory, Gulf Breeze, Florida.

Woessner, W.W., J.N. Moore,  C. Johns, M. Popoff, L. Sartor, and M. Sullivan.  1984.
      Final report: Arsenic source and water supply remedial action study.  Prepared
      for Solid Waste Bureau,  Montana Department of Health and Environmental
      Sciences, Helena, Montana.
                                     1-283

-------
PAH Analyses: Rapid Screening Method for Remedial Design Program
Laurie Ekes. Marilyn Hoyt, Gayle Gleichauf, David Hopper
ENSR Consulting and Engineering
35 Nagog Park
Acton, MA 01720
      The Iron Horse Park Superfund site (Billerica, MA) includes approximately 15 acres
of lagoon area.  Previous site investigations  have demonstrated  contamination by
Polycyclic Aromatic Hydrocarbons(PAHs) and petroleum hydrocarbons. The US EPA
record of decision and administrative order for the site stipulates stringent cleanup goals;
all soil with Total PAH above 1 ppm or TPH above 100 ppm must be remediated.
      As part of the pre-remedial design, iterative sampling was planned to tightly define
the spatial limits of the contamination. The program required analytical support with rapid
turn-around on high numbers of samples and action-level detection limits.
      The method developed and validated to support this program has since proven to
have wide applicability for site investigation and remediation.  Analytical protocol will be
presented with associated quality assurance/quality control data for samples from this site
and a coal gasification waste site. The method includes sample extraction by sonication
followed by direct analysis by GC/MS in the selected ion mode.  Total analysis time per
sample is under 30  minutes. Sixteen different PAH  may be identified and quantified in
samples, with individual PAH detection limits of 60 ppb. Comparison data from Method
8270 analysis for split samples will be presented.
       A statistically-significant correlation at the 95% confidence level was found between
total PAHs and Total Petroleum Hydrocarbons(TPH)  at the Ironhorse Park site.  Data for
the regression analysis, which will be presented, were used to justify reliance on the PAH
data alone for the remedial design.
                                     I-284

-------
37    EVALUATION OF HOUSEHOLD DOST COLLECTION METHODS FOR HDD NATIONAL SURVEY OF
                                     LEAD IN HOMES
        Benjamin S. Lira,  Ph.D.,  Joseph J. Bre«mT  Ph.D.,  Field Studies Branch,
        Office   of   Toxic  Substances,   U.S.   Environmental   Protection  Agency,
        Washington,  D.C.  20460; Kay Turman, B.S., Stan R. Spurlin,  Ph.D., Midwest
        Research Institute, Kansas City, Missouri 64110; Stevenson Weitz, Office
        of Policy Development and Research, U.S. Department  of  Housing and Urban
        Development, Washington,  D.C.   20410.
       ABSTRACT

       In  conjunction with the National  Survey of Lead-Based Paint in Housing
       conducted  by the Department  of Housing  and  Urban Development with the
       technical  support from  the Environmental  Protection  Agency,  it became
       necessary  to develop a method  of  collecting household dust samples for
       lead  (Pb)  analysis   from a variety of surfaces.  This  sampling technique
       needed to  be portable, simple to use, and applicable to  surfaces ranging
       from  window seats  to baseboards  to  carpeted surfaces.   In addition,
       because  the field sampling crews  would  be intrusive  into the occupied
       dwelling,  it was desirable  that adequate sample for analysis be collected
       in  5  to 10 minutes.  Three pumps fitted with  0.8 micron  membrane filters
       in  cassette holders were  evaluated at three  flowrates  (5, 20,  and 100
       1/min)/pressure heads (20, 60 and 125  mm Hg) specifications in combination
       with  four  different  nozzle designs.

       The results indicated the higher f lowrate sampling pumps produced not only
       better collection efficiency in a shorter period of time,  but also were
       more reproducible in their collection efficiency.  The final system design
       adopted provides  for better than 80% collection efficiency from a 4 sq.
       ft. area in  less  than 5 minutes.   The system  weighs  less than 10 Ibs and
       is  usable  with a minimum of training.
       INTRODUCTION

       While developing the design for  the HUD National Survey, HUD carried out
       a  pretest survey of  several housing  units in three  counties  of North
       Carolina  (Boyle et al.  1989).  HUD provided the data from this survey to
       EPA's Office  of Toxic Substances to help select appropriate methods of
       sample  collection  and  analysis.   The analytical  methods  used, quality
       control samples, analytical results, and  numbers of samples collected as
       part of the design were reviewed.
                                          1-285

-------
In the review of the data generated in HDD's pretest survey, it was noted
that  no data  were presented  on lead  in household dust.   However,  a
procedure was  presented  in  the survey design  for sampling and analyzing
household dust (RTI Survey Design 1988).   It was determined dust sampling
had been carried out,  but that none  of  the samples  collected indicated
lead above the method  detection limit.    It was  decided to evaluate the
pretest survey techniques to determine  why the data  were not consistent
with  previously observed results of lead in  almost  all  dust  samples
(Bornschein  et  al.  1986).     Studies by  Bomschein  and  Clark  at  the
University of  Cincinnati showed that  the majority of  indoor dust had at
least 100 Mg/g of lead.  These concentration levels were predicted to be
detectable by the AAS protocol, if sufficient sample mass was collected.
Dust sample weights were not determined in the pretest survey to verify
adequate sample  mass had been collected.   The lack of measurable lead
levels suggests the sampling method did not collect a sufficient amount of
sample  for  analysis.   Therefore,  a limited evaluation  of  the  sampling
technique was undertaken. The  objective was to formulate a protocol to be
used in the HUD National Survey of Lead in Homes.

Experimental Methods

The first step of the  evaluation was  to select a vacuum system adequate
for dust collection from carpets, windowsills,  and floors.   (The criteria
for selection of methods include:  (1) less than 5 min to sample  4  ft2 of
sample  area;  (2)  sampling  apparatus   weighs  less   than  10  Ib;  and
(3) simplicity of operation.)   The  major problem with  carrying  out  a
systematic method comparison of indoor dust sampling  techniques is the
reproducible   generation  of   representative  dust  samples  on  various
substrates.  Dust is a complex mixture of organic  and inorganic particles
of various shapes and sizes.   Because no representative "standard" dusts
were available due to time constraints and lack of commercial availabil-
ity, the dust  used throughout this evaluation  was composited from dust
collected in vacuum cleaner bags at a personal residence and from an MRI
floor vacuum.  Staff sieved the material to remove all particles greater
than 250 microns and any extraneous carpet fibers.  This produced a fine
dust  that  appeared  to settle  in a  pattern  similar  to that found  in
households.

Carpet  represented  a  unique  problem   in carrying  out  a  systematic
evaluation.  Staff used the  percentage of weight of a representative dust
sample as one of  the criteria to evaluate collection efficiency.  However,
when vacuuming carpet,  the vacuum picks up a large number of carpet fibers
along with  the dust.   This prevents an accurate determination  of the
percentage of  dust recovered from carpeted surfaces.   For the  National
Survey  in which the Pb in  dust was determined as micrograms of Pb per
square  foot of sampled  surface,  this  does not  represent a major concern
unless the carpet fibers contained significant amounts of Pb.  However, in
trying  to assess the efficiency of a particular  collection system, the
weight of the fibers collected could significantly bias the results in a
positive direction.
                                  1-286

-------
1.    Method of Application to Surfaces

The method of application of dust samples to the surface areas needed to
produce a  fairly even  distribution of a known  amount of material.   A
fairly  even distribution  of  dust  over the  surface would represent  a
typical field sampling situation.  Static effects  and non-retrievable dust
in cracks  or crevices  in the natural  dust settling patterns  are  more
difficult to assess and were  beyond the scope of this investigation.  A
child's toy flour sifter (Kitchen Play) was  used  to  evenly apply the dust
to the  tested surfaces.   Evaluation of the  holdup of the dust  in the
sifter showed that greater than 90%  of the dust reached the surface of the
sample area (Table 1).  The ability to reprodueibly deliver greater than
90% of the dust over  a dust burden  range of 60 to 229 mg was considered
adequate to proceed with this testing.  The dust scattered fairly evenly
over the templated area.
                   Table 1.  Dust Recovery for Sifter
  Weight of                Weight of
  dust added             dust determined
to sifter (mg)           on surface (mg)               % recovery
112
59.6
229
78.2

102
57.0
220
70.8

91.1
95.6
96.1
90.5
Avg = 93.3 ± 2




.8%
2.    Evaluation of Vacuum Collectors

Three vacuum pumps were evaluated, each with four different sample nozzle
configurations.  Table 2 lists the three pumps along with their pertinent
specifications.  The maximum  flow rate  for  each  pump was determined with
a clean sampling  cassette containing  both filter  and pad attached to a
4-ft  section of  3/8-in  heavy-wall Tygon  tubing.    The  flow  rate  was
measured using an NBS traceable anemometer.

Figure 1 illustrates the  nozzle designs evaluated.   Nozzle A, the design
used in the HUD pretest study, attaches to  a small  sampling tube that is
                                   1-287

-------
connected to the small sample port on the cassette.  Nozzles B and C are
designs constructed at MRI for evaluation.  The fourth option, Nozzle D,
is the open end of the cassette used directly as a sampling port.  Each of
these  nozzle designs was evaluated using each  pump operating  at its
maximum  flow  rate.    Nozzle  designs  B  and  C  evolved  during  the
investigation  of  sampling  methods to  meet  the  desired  sampling  time
constraints and were not evaluated simultaneously with Nozzles A and D.

The efficiency of each dust collection method was evaluated by determining
the weight of dust collected from a surface spiked with a known weight of
household dust.    In  determining the amount  of dust collected  in each
evaluation, a problem is the weight change in the  cassette  due to changes
in the amount of moisture on the filter.  This problem would be especially
severe at high flow rates which could chill  the  filter with subsequent
condensation.   A preliminary evaluation,  sampling just air  for  5 min,
showed very  little change « 5  mg)  in a 2-L/min  (SKC  pump)  system and
larger changes  (>  15 mg)  in the higher  flow systems (i.e.,  Cast pump).
This problem was minimized by drying the  cassettes at 80°C, removing them
from  the  oven  and  immediately  sampling,  redrying   for 5  min,  and
immediately weighing after removal.   Weight gains  associated with air
sampling were less than 2 mg on cassettes after this treatment.
          Table 2.  Pumps Evaluated for Oust Collection
                           Flow rate     Pressure head
                         specification   specification     Weight
Manufacturer    Model       (L/min)         (mm Hg)         (Ib)
SKC              101-M          5             25            1.2

Gast           302-100        100            125           10.8

Fisher          A-20           20             60            7.2
                                   1-288

-------
                                          To Pump
rvj
CD
CO
                               A
                          Tubing 4 Flange
  JL
           To Pump
To Pump
                                                               4"
      B
Square Cut Teflon
                                                                                                        To Pump
                                                           2.r
                 D
          Open End Cassette
                                             Figure 1.  Nozzles evaluated In dust collection study.

-------
Dust collection was  determined on three different surfaces:   (1)  vinyl
tile;  (2)  construction-grade plywood  (unpainted);  and  (3)  a piece  of
enamel-painted, construction-grade plywood.   Table  3  shows the  test
matrix. Attempts at evaluating the collection efficiency  from carpet were
not successful due to continuing interferences from carpet fibers.  These
fibers  did not  allow  for  accurate  dust  weight  determinations,  and
therefore resulted in artificially high recoveries.

Tables 4 and 5 show  recovery results of the sampling evaluations.   Each
determination was  made  in duplicate,  and  average value reported.   The
procedure used to collect the dust involved 50% overlapping passes made on
the surface from left to right and then top to bottom.  Table 6 includes
the approximate time to vacuum a 1-ft2 area for each nozzle type and pump.
The time difference to cover the specified area was significant, and this
impacted the final selection between the systems.  The early data on the
time required to vacuum resulted in fabricating a larger vacuum nozzle for
dust sampling (Nozzles B and C).

In evaluating the  data in Tables 4 and 5, it is clear the  larger flow rate
pump is necessary to achieve high levels of dust recovery.   This finding
is contrary to reports by other researchers  (Bornschein et al. 1985), who
have previously reported >  80% recovery using the SKC  5-L/min pumps*  This
difference may be the result of sampling technique, time, or type of dust
(i.e., particle size) sampled.  Our finding, however, is consistent with
the results reported in the HUD pretest survey.  After  considering the
sampling efficiency (amount collected on filter)  and sampling time, Nozzle
C and  the  Gast pump operated at full  capacity  were  selected to collect
household dust in the HUD National Survey of Lead in Homes.  Table 7 shows
a summary recovery (%) performance of Gast pump/nozzle  combination system.

It is clear significant  research efforts in this  area  of  dust  sampling is
warranted.  Our efforts and decisions were based on an immediate need to
meet the short-term deadlines required for the HUD national survey.  EPA
continues to investigate improved  methods  for household dust collection
studies (Wilson et al. 1991).
      Bornschein, R.L.; Succop,  P.A.;  Krafft,  K.M.; Clark, C.S.; Peace,
      B.; Hammond, P.B.  (1987b) Exterior surface dust lead,  interior house
      dust lead and childhood lead exposure in  an urban environment.  In:
      Hemphill, D.D., ed. Trace substances in environmental health - XX:
      proceedings of University of Missouri's 20th annual conference; June
      1986; Columbia, MO.  Columbia, MO:  University of Missouri; pp. 322-
      332.
                                  1-290

-------
2.    Boyle, K.E.; Gutknecht, W.F.} Neefus, J.D.; Stutts, E.S.; Williams,
      E.E.j Williams, S.R.; "Design A National Survey of Lead-Based Paint
      in Housing - Pretest  Report".  HUD Contract No.  HC5796 (August 3,
      1989).

3.    Wilson, N.K.;  Lewis, R.G.;  Fortmann,  R., et al.;  "Evaluation of
      Methods for Measurement of Exposure of Young Children to Lead in
      Homes,"  to  be  presented at  International  Society  of  Exposure
      Analysis, Annual Meeting, Atlanta, Georgia, November 1991.
                                    1-291

-------
            Table 3.  Test Matrix  for Vacuum Dust  Collection
Pump
(operating flow)
SKC (2 L/min)
Fisher (20 L/min)
Cast (100 L/min)
Vinyl
2a
2a
2a
Surface type
unpainted "olvwood
2a
2a
2a
Enamel plvwood
2a
2a
2a
aNumber of replicate collections carried out  at each  sample  weight
 (50 mg and 150 ing).
                             1-292

-------
           Table 4.  Recovery of Surface Dust Applied at 50 ± 10 mg/ft2
Pump and nozzle
SKC (2 L/min) + A
SKC (2 L/min) + B
SKC (2 L/min) + C
SKC (2 L/min) + D
Gast (100 L/min) + A
Cast (100 L/min) + B
Gast (100 L/min) + C
Gast (100 L/min) + D
Fisher (20 L/min) + A
Fisher (20 L/min) + B
Fisher (20 L/min) + C
Fisher (20 L/min) + D
Vinyl
6.5(23.0)
7.5( 6.7)
10,5(14.0)
3.5(100)
45.5(14)
73.5(7.5)
90.0(2.2)
36.0(16.6)
12.0(0)
21.5(7.0)
27.0(7.4)
16.0(6.2)
Unpainted plywood
2.0(100)
3.0(33.3)
12.0(25.0)
19.5(79.5)
46.5(3.2)
85.5(6.4)
92.0(5.4)
25.5(37.2)
17.0(29.4)
28.0(10.7)
34.5(10.1)
15.5(100)
Enamel-painted
plywood
7.5(20)
3.0( 0.0)
5.5( 9.0)
7.5(6.7)
53.5(43.9)
93.0(13.4)
60.0(46.7)
30.5(1.6)
35.0(67.9)
26.5(17.0)
51.0(11.8)
17.0(0)
*RPD = Relative percent deviation  |x-x|  x 100
                                          1-293

-------
Table 5.  Recovery of Surface Dust Applied at  150 + 25 mg/ft2
                              Average Recovery  (%). x  (RPD)
Pump and nozzle
SKC (2 L/min) + A
SRC (2 L/min) + B
SKC (2 L/min) + C
SKC (2 L/min) + D
Cast (100 L/min) + A
Cast (100 L/min) + B
Cast (100 L/min) + C
Cast (100 L/min) + D
Fisher (20 L/min) + A
Fisher (20 L/min) + B
Fisher (20 L/min) + C
Fisher (20 L/min) + D
Vinyl
19.0(10.5)
2.0(100)
6.0(50)
16.5(30.3)
41.0(87.8)
69.5(20.9)
85.0(1.2)
35.0(14.2)
22.5(28.9)
43.0(9.3)
68.0(48.5)
17.0(0)
Unpainted plywood
10.0(20.0)
4.5(55.5)
7.0(71.4)
6.0(16.7)
38.5(13.0)
82.5(4.2)
94.5(2.6)
35.0(2.9)
16.0(6.3)
42.0(11.9)
50.5(16.8)
10.5(100)
Enamel-painted
plywood
21.5(25.6)
3.5(100)
8.0(12.5)
14.0(7.1)
37.5(4.0)
77.5(3.2)
98.5(8.6)
25.5(37.2)
21.0(4.8)
29.0(44.8)
53.5(12.1)
23.5(17.5)
                               1-294

-------
                       Table 6.  Time to Vacuum Surface
                     Nozzle
                 Time to cover area
                   of 4 ft2 (min)a
                       A

                       B

                       C

                       D
                         24

                          5

                          5

                         32
                     aTime to execute 50% overlapping passes as
                      specified in the text.  Visual inspection
                      of collection efficiency not used for this
                      evaluation.
Table 7.  Summary Recovery (%) Performance of Selected Pump/Nozzle Combinations
          (Gast(100 1/min) + C)
Dust Loading
Vinyl
                                        Average Recovery (%). x (RPD)
Unpainted plywood
Enamel-painted
   plywood
50 ± 10 mg/ft2
150 ± 25 mg/ft2
90.0(2.2)
85.0(1.2)
92.0(5.4)
94.5(2.6)
60.0(46.7)
98.5(8.6)
                                         1-295

-------
38            FIELD DEPLOYMENT OF A GC/ION TRAP MASS SPECTROMETER FOR TRACE
                         ANALYSIS OF VOLATILE ORGANIC COMPOUNDS

         Chris  P.  Leibman.  David  Dogruel,  Health  and  Environmental  Chemistry
         Group, HSE-9, Eric P. Vanderveer, Instrumentation Group, MEE-3 Los Alamos
         National Laboratory, M/S K-484, Los Alamos, New Mexico, 87545
         ABSTRACT

         Field analytical support can directly impact the expense of environmental
         clean-up  by reducing the cost-per-analysis.  Cost  for sample packaging,
         shipment,   receiving and  management  are eliminated 1f  analyses  are
         performed on site.  Field analytical  support improves the chances that
         schedules and monetary constraints  associated  with  remedial  activities
         are met.
         To  reduce  the cost associated  with environmental  clean-up  we  have
         developed  a purge  and  trap/GC/Ion  Trap  Detector (ITD)  at Los Alamos
         National Laboratory for the identification and quantification of  volatile
         organic compounds  present  at  chemical waste sites.  A custom purge  and
         trap/GC  sampling  system was integrated with  a  modified ITD to  achieve
         instrument  operation consistent with field  activities.
         The instrumentation and associated  methods  parallel those outlined  in
         method 8260,  SW-846.  Qualitative and quantitative  analysis for the  68
         target compounds and the associated internal  standards and surrogates  is
         completed  in an automated sequence  that  1s  executed every 25  minutes.
         Sample purging,   analysis,  data  reduction,   and  preliminary report
         generation  proceeds  automatically.   The Instrument can be  operated in a

                                          1-296

-------
continuous mode, pausing only for sample loading and data file
specification.   All data  are archived on  floppy disk  for  subsequent
review by a skilled analyst.  Part-per-trillion detection limits can be
attained for many compounds from either 5 gram soil  or 5 milimter water
samples.

The  GC/ITD  1s  being  deployed  in  a mobile  laboratory which  has  been
designed  to  support  volatile  organic  analysis.    The  use  of  the
transportable GC/ITD for  support  of environmental surveillance and the
characterization/clean-up  of  hazardous  waste sites is being evaluated.
We will discuss  field activities completed to date and the evolution of
field  operation plans  and  field documentation.   Additionally, we will
discuss the quality control we have Implemented for field analysis using
the GC/ITD.  Results obtained from blind quality control  samples will be
presented.

One  purpose  of  the  presentation will be to examine problems encountered
with field analyses using  the GC/ITD and to discuss any actions taken to
address  those problems.   Notably,  we will discuss how small quantities
of water  introduced into the ITD from the purge and trap  sampling system
negatively  impact quantitation and the steps we have  taken to mitigate
those  problems.
                                   1-297

-------
OQ                        ACCURATE, CN-SITE ANALYSIS OF PCBs
                             IN SOIL — A LOW COSH APPROACH

                        Deborah Lavicme. Quality Control Manager
                                   Dexsil Corporation
                                 One Hamden Park Drive
                               Hamden,  Connecticut  06517

         ABSTRACT

         Polychlorinated Biphenyls  (PCBs) are very  stable materials of low
         flammability used as insulating materials  in electrical capacitors
         and transformers,  plasticizers  in waxes, in  paper manufacturing, and
         for a variety  of other industrial purposes.

         There are many PCS transformers and capacitors still in service
         throughout the United States today.  The Environmental Protection
         Agency   estimates that there are 121,000 (askarel) PCB transformers,
         20 million PCB contaminated mineral oil  transformers and 2.8 million
         large PCB capacitors currently  in use.   A  certain percentage of this
         equipment will leak, fail  or rupture and spill PCB into the
         environment  each year (1).

         Because of equipment leakage and widespread  industrial dumping, PCBs
         have  appeared  as ubiquitous contaminants of  soil and water.  Chemical
         analysis for PCBs has been almost exclusively performed by gas
         chromatography.  Other analytical techniques such as nuclear magnetic
         resonance (NMR) and liquid chromatcgraphy with UV detection are
         alternative  methods for PCB analysis but can only be successfully
         applied where  the suspected concentration level  of PCB is greater
         than  1000 ppm.

         A new instrumental method has been  developed to  analyze for PCB
         content using  electrochemical methodology  and a  chloride specific
         electrode to measure quantitatively the  amount of chloride.  The
         instrument converts the chloride concentration into a PCB equivalent
         amount  of PCB  in an oil or soil sample and gives a direct readout in
         parts per million of PCB.   The  preparation steps involve extracting
         the PCBs from  the soil (not necessary  for  oil samples) and reacting
         the sample with a sodium reagent to transform the PCBs into chloride
         which can be subsequently quantified by  the  instrument.  Oil samples
         take  about 5 minutes to prepare and soils  about  10 minutes.  One
         operator can complete about 150 oil tests  or 100 soil tests in an
         eight hour day.

         Although this  paper will concentrate on  the  results of soil samples
         obtained from  a Superfund  site  analyzed  electrochemically and by gas
         chromatcgraphy, it demonstrates the accuracy and economic advantage
         of employing the electrochemical procedure in analyzing both oil and
         soil  samples.
                                          1-298

-------
PCBs were first formulated as far back as 1881.  Although they were
known to exist in the late 1800s, manufacturing on a aammercial  scale
did not start until 1929.  It was not until 1977 when all U.S. PCB
production was halted.

In the late 1960s, PCBs were recognized as a potential environmental
problem, which was probably due to the unregulated maintenance and
handling of PCBr-containing equipment.  A series of studies has been
done to identify and quantify the distribution of PCBs in the U.S.
The overall distribution is shown in figure (1).

The wide use of PCBs was due to their non-flammable characteristics
as well as their thermal and chemical stability, low vapor pressures
at atmospheric temperature and high dielectric constants.  Although
the use of PCBs has been banned in most applications, they are still
being used in vacuum pumps and gas-transmission turbines.  PCBs have
been used as plasticizers in synthetic resins, in hydraulic fluids,
adhesives, heat transformer systems, lubricants, cutting oils and in
many other applications.

The EPA currently recommends two PCB specific methods of analysis;
the GC/MS Method 680 for quantitating PCB isomer class totals and the
GC/ECD Method 8080 for quantitating Aroclors.  Over the past decade,
the use of these instrumental methods has increased dramatically and
it is the purpose of this paper to provide an example of one type of
non-specific analysis of PCBs where simple inexpensive chemical
procedures can in fact, under certain circumstances, be a preferable
alternative to chromatographic methods.

The examples chosen in this paper are the analyses of PCBs in
transformer oil and soil.  The tests involve measurements of PCB
concentration down to a few parts per million where, as a result of
extensive legislation, inaccurate results would likely evoke
expensive litigation and heavy fines.  The different methodology and
apparatus will be described, the accuracy and precision of each
method discussed, and the costs of each analysis reported.


METHOD FOR THE ELECTROCHEMICAL DETERMINATION OF PCB IN OILS AND SOIL

This procedure utilizes sodium metal to remove chlorine from any PCB
present in the sample.  The concentration of chloride contained in
the final aqueous extract can be determined electrometrically by
means of a chloride specific electrode.  By immersing a chloride
specific electrode in the aqueous extract and measuring the EMF
produced, the chloride concentration and thus, the PCB content can be
estimated.  The chloride concentration is exponentially related to
the electrode EMF and thus with a suitable electronic circuit design
the results can be presented digitally in ppm of the selected PCB on
an appropriate meter.
                                  1-299

-------
This is a non-specific method,  testing for the presence of chlorine
in the sample being examined.   As a result, other chlorinated
compounds will cause a false positive result because the analysis
method reads all chlorinated compounds as PCB.  False negative
results should no occur, however, because if no  chlorine is present,
PCBs cannot be present.
SAMPLE PREPARATION

I/Oil Samples
0.2 ml of a solution of naphthalene in diglyme is added to 5 ml of
oil sample.  To this mixture is added 0.4 ml of a dispersion of
metallic sodium in mineral oil and the mixture shaken for one
minute.  5 ml of buffer is then added to neutralize the excess sodium
and to adjust the pH to 2.0 to ensure the pH of the mixture is within
the operating range of the electrode.  5 ml of the aqueous layer is
then carefully decanted into a suitable vessel.

2/Soil Samples
10 g of the sample of soil is extracted by shaking for one minute
with 12 ml of solvent containing 2 ml of distilled water in 10 ml of
an immiscible hydrocarbon.  The soil is then allowed to settle and
the supernatant liquid filtered through a column containing Florisil
to remove any moisture and inorganic chloride.  5 ml of the dry
filtrate is then treated with 0.2 ml of a solution containing
naphthalene in diglyme, followed by 0.4 ml of a dispersion of
metallic sodium in mineral oil and shaken for 1 minute.  5 ml  of
buffer solution is then added and the aqueous layer allowed to
separate.  5 ml of the aqueous layer is then decanted into a suitable
vessel.
ANALYTICAL METHOD

The measuring instrument (Dexsil 12000™, Hamden, CI)  is fitted
with temperature compensation as the output of the chloride specific
electrode varies with temperature.  Initially the temperature
compensation adjustment is set to the sample/electrode temperature.
The electronic measuring device is then calibrated employing a
solution containing chloride equivalent to 50 ppm.  The electrode is
immersed in 5 ml of the calibration solution and appropriate
adjustments made to the calibration control to provide an output on
the digital meter of 50 ppm of chloride.

After rinsing and drying, the chloride specific electrode is immersed
into the 5 ml sample, gently stirred for 5 seconds and allowed to
stand for 30 seconds.  The concentration of PCB in ppm is then read
directly from the digital output meter.  The dynamic range of this
analytical procedure is from 5 to 2000 ppm.  The precision varies
with the concentration.  At concentrations between 50 and 2000 ppm,
it is +/" 10% •  Between 5 and 50 ppm it is about +/- 2 ppm.
                                 1-300

-------
ANALYTICAL TESTS, RESULTS AND DISCUSSION

Oil Samples
In general, PCS specific methods are more accurate than the
non-specific methods, but they are also more expensive, more lengthy
to run, and less portable.  The 12000™ PCS analyzer provides
accurate analysis of PCS concentration in oil by testing for the
total amount of chlorine that is present in the sample.

The PCS concentration is calculated from the chloride concentration
using a conversion factor based on the Aroclor present in the
sample.  If the specific Aroclor is not known, then the most
conservative estimate- results from assuming that the PCS present is
Aroclor 1242.  Aroclor 1242 contains the lowest percentage of
chlorine of the commercially produced FOB mixtures.

The 1260 setting is used when a sample contains Aroclor 1260, but not
the associated trichlorobenzene.

The Askarel setting is used for samples that contain Aroclor 1260 and
associated trichlorobenzene.  Askarel accounts for the majority of
contaminated transformer oil samples and therefore this setting will
usually supply the most accurate results; however, if a 1242
contaminated sample is tested on the askarel setting, a false
negative will result if the sample contains between 50 and 120 ppm.

Tables  (1) and  (2) show comparison results of transformer oils
contaminated with 1242 and 1260  (as Askarel) respectively, analyzed
by the PCB specific GC method versus the L2000  .  The GC method
used to analyze the transformer oils in this study is EPA
600/4-81-045.

It is seen that accurate and precise results are obtained over a wide
concentration range of PCBs and although false positives can cause
unnecessary secondary testing, this method can be very economical
when used on transformer oil, which contains few sources of chlorine
other than PCS.  Used crankcase and cutting oils, however, always
contain some chlorinated paraffins and almost always give false
positive results with non-specific testing.  More expensive gas
chromatographic analysis is required when testing for regulated
levels of PCS in these matrices

Soil Samples
The EPA Spill Cleanup Policy stipulates that a PCB spill, once
detected, must be cleaned up within 48 hours. (3)  The EPA mandates
that cleanup actions are taken in this short time frame in order to
minimize the risk of human and environmental exposure to the spilled
PCB.  In addition to the many  PCB Superfund sites, there are still
many other PCB spill sites that have not made the National Priorities
list that still must be cleaned up.
                                  1-301

-------
One of the most time consuming steps in laboratory soil analysis is
the drying time.  When a soil sample is received for GC analysis by
ASTM D3304, the sample is dried for 24 hours.   The sample is then
weighed and placed in a soxhlet extractor and allowed to cycle for 8
hours.  The sample must be completely dry, since the extraction
solvent (visually hexane or isooctane) is immiscible with water.
Extraction of a wet sample would yield a low result since the  solvent
cannot fully interact with the soil to extract the PCBs.  Typically,
90% of soil samples received for laboratory analysis by GC require
drying prior to extraction.  With a 48 hour cleanup policy,
twenty-four hours of drying time could be a substantial set-back.
The content of the spilled material must ideally be determinedat
once and the cleanup procedures begun immediately.  The 12000™
allows the operator to respond immediately and to make a quick
evaluation of the concentration of PCB at the site.  At an excavation
site where soil analysis is being performed, the decision can be made
immediately if more soil needs to be removed or if the excavation has
been carried far enough.

The results of soils obtained from a Super-fund site and analyzed by
GC and the 12000™ are compared in Table  (3).  Since gas
chromatography can quantitate each Aroclor present, the GC results
are presented for each Aroclor aqboally detected in the soil
samples.  The corresponding 12000™ results for that particular
sample are seen on the same line.  These results are listed according
to each setting available to the analyst.  The 12000™ does not
have the capability to quantitate each Aroclor; instead, all the
chloride present is interpreted according to the Aroclor setting
being used.  For samples contaminated with an unknown Aroclor, the
prudent analyst would use the 1242 conversion to provide the most
conservative estimate.

Using the 12000™ as a screening method, the samples are evaluated
according to column 4 interpreting chloride as 1242.  For the ten
samples analyzed, samples 2, 3, 4 and 6 would be considered as below
the Code of Federal Regulations limit of 10 ppm set by the EPA.
Since this is a site remediation, the results would indicate that
these areas can be considered "clean" and would not need further
treatment.  If active clean-up were underway, these samples would
indicate that the excavation has gone far enough in that area.

The remaining samples indicate that there is still possible
contamination above the 10 pm level.  This would result in further
excavation being required to reach safe levels.  If active excavation
is not underway then the samples can be further analyzed to determine
the specific Aroclor content.  Whether the samples are further
analyzed or excavation is continued based on the 1242 estimate will
depend on the cost consideration of waiting for lab results while
paying for an idle excavation team and remediation equipment,  or
excavating excess material while the crew and equipment are still on
site.
                                  1-302

-------
Frxm the GC analysis it was determined that only two of the six
"positives" were "false positives" in that the total chlorine
indicated an equivalent of PCS above the regulatory 10 ppm limit
whereas GC analysis of those samples showed an actual level below 10
ppm.

The problem of contamination with chlorinated solvents is exemplif ied
by sample 1 where the 12000™ result is considerably higher than
the GC results.  This high reading is again an over estimation of the
PCB present and would result in a conservative action being taken
such as retesting using GC or further excavation.

To make a systematic comparison of the GC results which quantify each
Aroclor separately, to the 12300™ results, an equivalent amount of
a single Aroclor must be calculated from the sum of all Aroclors
detected.  For the results given in this paner Aroclor 1242 was
chosen as the basis for equatincr the 12000™ results with the GC
results.  The equivalent 12000™ reading, which converts the
chloride concentration to PCB using a single Aroclor conversion
factor, can then be calculated.  The direct conversion of ppm 1260 by
GC to its equivalent in ppm 1242 is based on the percent chlorine
difference of 1242, 42%, versus 1260, 60%, according to the equation:

              12000 equivalent ppm 1242  = (X) (60/42)
              where:  X = ppm 1260 by GC
                      60/42 = ratio of percentage chlorine

For example, the GC results for the first soil sample shown in Table
(3) of 11.59 ppm 1242 and 2.24 ppm 1260 should theoretically read
14.79 on the 12000's 1242 setting.  The value of 14.79 is attained by
converting the GC 1260 value to 1242 according to the equation above,
and adding it to the GC value for 1242.   The actual reading on the
12000 1242 setting was 25.0 ppm, which is significantly higher than
the theoretical prediction.  The false high reading can probably be
attributed to other chlorinated compounds being present in the soil
that the GC does not detect.  Nevertheless, from a regulatory point
of view a false positive is preferable.  A more realisitc and
expected result is seen from the results for the seventh soil
analysis shown in Table (3), and the once again a theoretical
concentration of 1242 can be predicted from the conversion equation.
The GC result for that sample was 92.66 ppm 1242 and 15.08 ppm 1260.
15.08 ppm 1260 converts to 21.54 ppm 1242, which when added to 92.66
ppm 1242 gives a theoretical projection of 114.2 ppm 1242 as the
12000 result.  The actual 1242 result given by the 12000 was 122.7,
which is within the +/- 10% accuracy level accepted for GC analysis.

Table  (4) shows a comparison of results from soil samples obtained
from a PCB spill site.

Like the oil samples, soil sample concentration of PCBs are also
based on the detection of chlorine; however, it is only chlorine
present from an organic source that would cause a false positive, as
seen in the first example above, rather than an inorganic source such
as road salt or sea salt.  Some possible sources of chlorine
contamination are pesticides and solvents.

                                  1-303

-------
One benefit to the laboratory personnel analyzing soils is that using
the 12000™ first to screen PCB content allows the GC chemist to
make an accurate dilution right away.   The appropriate dilution is to
1 ppm and one chromatographic analysis is approximately 40 minutes
long.  The analysis time can certainly add up with trial-and-error
dilutions being made, especially if there are many samples waiting to
be analyzed.  Kicwing the right dilution also prevents overloading
the column with PCB contamination.

The 12000™ system can analyze to fewer than 5 ppm in oil and soil,
can be used in the field by non-technical personnel,  and requires
less than 10 minutes to run an analysis.  These attributes make the
instrument an excellent alternative to gas chromatographic analysis,
especially for soil samples.

Although this new technique does not replace gas chromatography,  it
can significantly reduce the number of samples requiring GC  analysis,
and therefore allow a greater amount of samples to be run at a lower
cost.
                                  1-304

-------
REFERENCES

1./Environmental Progress and Challenges;  EPA's Update.  United States
Environmental Protection Agency.  EPA-230-07-033.  August 1988.

2,/Finch, S.R.,  lavigne, D.A.,  Scott, R.P.W.  "One Example Where
Qiromatography May Not Necessarily Be the Best Analytical Method."
Journal of Chramatographic Science.  July 1990.  pp. 351-356.

3./40 CFR 761.125.  Office of the Federal Register.  Rev. July 1, 1989.

4./PCS Equipment, Operations and Management Reference Manual.  SCS
Engineers, Inc.
                                   1-305

-------
                      FIGURE 1



            U.S. Distribution of PCBs (4)



Presently in use             750 million pounds     60%



In landfills and dumps       290 million pounds     23%



Released to environment      150 million pounds     12%



Destroyed                     55 million pounds      5%



Total production           1,245 million pounds     100%
                        1-306

-------
                                  TABLE 1


         RESULTS OF GC ANALYSIS OF PCBs  (1242) IN TRANSFORMER OIL
                                    VS
                         RESULTS OF L2000 ANALYSIS
 Standard      Results from GC Analysis
fppm 1242)      	fppm 1242)	
Results from 12000 Analysis
	(ppm 1242)	
0 None Detected (< 2 ppm)
None Detected (< 2 ppm)
None Detected (< 2 ppm)


10 10.0
10.8
10.4
MEAN 10.4
STD. DEV. 0.4
50 51.6
52.3
50.3
MEAN 51.4
STD. DEV. 1.0
100 96.8
95.8
94.2
MEAN 95.6
STD. DEV. 1.3
500 474.0
482.2
497.0
MEAN 484.4
STD. DEV. 11.7
0.6
0.9
1.5
MEAN 1.0
STD. DEV. 0.4
9.7
9.3
9.7
MEAN 9.6
STD. DEV. 0.2
50.7
46.2
51.4
MEAN 49.4
STD. DEV. 2.8
104.9
95.2
95.4
MEAN 98.5
STD. DEV. 5.5
522.0
492.0
470.0
MEAN 494.0
STD. DEV. 26.1
                                  1-307

-------
                                  TABLE 2

                  COMPARISON OF RESUIUS FROM THE ANALYSES
            OF OIL SAMPLES OCNTAINING AROCLOR 1260 (ASKAREL A) :
                        GAS C3JROMATOGRAPHY VS L2000
Standard
fppm 1260)
GC Analysis Results
    fppm 1260)
12000 Analysis Results
	fppm 1260)	
10
         MEAN
         STD.DEV.
     9.482
     9.241
     9.186
     9.303
     0.129
MEAN
STD.DEV.
 9.2
 9.5
10.6
 9.8
 0.6
50







MEAN
STD.DEV.
50.923
48.409
51.883
50.405
1.465



MEAN
STD.DEV.
53.7
48.6
50.8
51.0
2.1
250               233.911
                  232.007
                  230.215
         MEAN     232.044
         STD.DEV.   1.509
                                     255
                                     262
                                     261
                            MEAN     259
                            STD.DEV.   3.8
500               493.232
                  486.400
                  472.423
         MEAN     484.018
         STD.DEV.   8.661
                                     530
                                     519
                                     510
                            MEAN     520
                            STD.DEV.  10.0
                                1-308

-------
                 TABLE 3

COMPARISON OF SUEERFUND SITE SOIL ANALYSES:
   GAS CHROMATOGRAHK VS L2000 READINGS

1242
11.59 ppm
0.32 ppn

0.33 pptl
5.00 ppm
0.77 ppm
92.66 ppm
7.18 ppm
7.87 ppm

GC RESULTS
1254 1260
2.24 ppm
0.25 ppm
2.64 ppm 1.78 ppm
0.20 ppm
2.53 ppm
0.80 ppm 0.35 ppm
15.08 ppm
1.54 ppm 0.08 ppm
3.25 ppm 0.30 ppm
9.43 ppm
L2000 RESULTS (read as)
1242
25.0 ppm
0.9 ppm
7.9 ppm
2.8 ppm
10.6 ppm
7.5 ppm
122.7 ppm
11.5 ppm
13.0 ppm
16.2 ppm
1260
17.5 ppm
0.6 ppm
5.5 ppm
2.1 ppm
7.5 ppm
5.3 ppm
85.8 ppm
8.1 ppm
9.2 ppm
11.4 ppm
ASKAREL
10.6 ppm
0.4 ppm
3.3 ppm
1.4 ppm
4.6 ppm
3.2 ppm
51.7 ppm
4.9 ppm
5.6 ppm
6.9 ppm
                   1-309

-------
                  TABLE 4

COMPARISON OF PCB SPILL SITE SOIL ANALYSES:
        GAS CHRCMATOGRAPHY vs L2000
GC RESULTS
1242 1254
.30 ppm
.10 ppm
.97 ppm
.38 ppm
.68 ppm



1260
6.09 ppm
41.59 ppm
0.40 ppm
0.05 ppm
6.67 ppm
4.42 ppm
206.0 ppm
1699.0 ppm
                                 12000 RESULTS (read as)
                            1242        1260        ASKAREL

                            10.8 ppm    7.5 ppm     4.5 ppn

                            62.5 ppm   43.8 ppn    26.4 ppm

                             5.7 ppm    4.0 ppn     2.4 ppm

                             6.1 ppm    4.3 ppm     2.6 ppm

                            14.8 ppm   10.3 ppm     6.2 ppm

                             7.3 ppm    5.1 ppm     3.1 ppm

                           404.0 ppm  281.0 ppm   167.5 ppm

                           >2000 ppm ,1642.0 ppm   996.0 ppm
                    1-310

-------
40               HOW GOOD ARE FIELD MEASUREMENTS?

                   Llewellyn R.  Williams, Director
         Quality Assurance and Methods Development  Division
                   EPA - EMSL - Las Vegas, Nevada


                              Abstract


       Quick!   Cheap!   High throughput!   We have all heard these
  words associated with field measurement methods.  But what does
  the record show with respect to the application of these new
  technologies and the quality/  acceptability, and usability of
  data produced in the field.  Are the touted advantages of field
  screening methods and field analytical methods being fully
  realized to measure,  monitor,  or characterize waste sites or
  waste streams?  Highlights will be presented from the Second
  International Symposium on Field Screening Methods that was
  conducted earlier this year.   Technologies presented ranged from
  simple chemical and immunochemical test kits to highly
  sophisticated fieldable instrumentation for analysis of toxic
  metals and organic chemicals in all environmental media.  Case
  studies indicate the current utility of several key technologies
  for monitoring and site characterization.  In addition,
  information will be furnished on field measurement technologies
  recently demonstrated under the Superfunds Innovative Technology
  Evaluation program.   Some institutional inertia appears to hinder
  the broader acceptance of field-produced data.  One thing remains
  clear? the new field technologies do not, nor should they be
  expected to, replace operator skill and judgement in generating
  environmental data.   But they do constitute a battery of new and
  available tools thai can improve the confidence of decisions
  based upon such data.
                                    1-311

-------
A-\     ASSESSMENT OF POTENTIAL PCB CONTAMINATION INSIDE A BUILDING;
                    A UNIQUE MULTI-MATRIX SAMPLING PLAN


       William W. Freeman, Principal Scientist, Roy F. Weston, Inc.
       Weston Way, West Chester, Pa. 19380

       ABSTRACT

       Characterization  of  potential  PCB  contamination inside a
       large, active market and restaurant area was required.  There
       was the possibility for PCBs to have entered the facility as
       a result of construct ion/demolition activities taking place in
       the area above the market.

       This  case history  describes  a unique  assessment approach,
       including a multi-matrix sampling procedure. A representative
       number of samples had  to be collected from this facility which
       occupies  approximately  175,000 ft.2 in area.   The sampling
       also had to be performed in a practical manner, with the least
       possible disruption of routine daily activities.

       A visual inspection and reconnaissance of the market was first
       conducted in order  to  identify entry points for potential PCB-
       containing materials  such as  dust,  water and  debris.   Six
       different  matrices were identified  for sampling,  including
       air, dust, water and sediments.  Wipe samples were also taken
       from non-porous  surfaces  such  as  counter tops and fixtures.
       Destructive  (chip)  samples  were  taken  from  porous  solid
       surfaces such as wood and insulation  materials.  Composite
       samples were taken from  some areas.   Quality Control samples,
       including items such as  duplicates and field blanks were also
       taken.

       The sampling plan is discussed  in detail, including equipment
       used,  statistics,  and  the  selection  of  random and  biased
       sample locations.   Analytical  procedures  are also reviewed,
       including extraction techniques and quantitation limits.
                                   1-312

-------
1.   INTRODUCTION

Roy F. Weston, Inc. was retained by the Department of Health
of a large Eastern United States city to collect and analyze
environmental samples from within a large urban food market.
The purpose of this investigation was to assess the extent of,
or  confirm the absence  of,  polychlorinated  biphenyl  (PCB)
contamination.

Demolition  activities,  including PCB removal, had  been on-
going in the open shed which is located above the market.  The
shed is an elevated area with a roof, but  is open at one end.
The market is  approximately 175,000  ft2  in area,   and  the
platform of the shed is  supported by steel columns within the
market area.  The  ceiling of  the market  is suspended from the
shed steel  structure, and  consists of  a wood layer, covered
with roofing  paper and  sheet metal.   The interstitial space
between the market's wooden ceiling and the underside of the
shed structure houses a combined support-beam and stormwater
drainage system.

The market is primarily a varied food market with a major area
devoted to restaurants of different types  to serve the people
from the office  buildings surrounding the market.  Concern had
previously centered around the potential for PCB contamination
entering the  market  by way  of leaks through  or  around the
ceiling structure  from the demolition activities on-going in
the  shed  above.    This concern was  accentuated  when  the
extremely heavy rains of one day deluged the shed and resulted
in severely heavy  leaks into the market below.

This  report  summarizes  the  assessment   carried  out  and
environmental  samples  collected from  surfaces  within  the
market.  Field  sampling  activities were conducted by Roy F.
Weston,  Inc.  personnel  on  a  Sunday,  while the  market  was
closed.  All  sampling was conducted in Level "D" personnel
protection requirements except during dust sampling when Level
11C" protection was used.

It  should  be noted  that this  sampling effort  included  55
samples from a variety of matrices (e.g., sediments, solids,
air, dust,  wipes  of surfaces,  water).   Although this  is a
relatively  small number  given the size of the market,  it is
considered  representative  of conditions at that  time within
the market  regarding assessing potential PCB contamination.
Roughly one-half of  the  samples were  biased toward areas of
higher probability for contamination (such  as areas of leaks
or visible staining)  and one-half were random throughout the
active, occupied areas of the market.
                             1-313

-------
2.   SCOPE OF WORK

The sampling team first performed a visual inspection of the
market in order to identify  potential  entry points for air,
dust, water, oil and/or debris  from the shed above into the
market and delineate  initial  sampling locations.  These entry
points, and  potentially  affected areas  below these points,
were candidates for sample collection.  The visual inspection
to identify potential entry  points  consisted of looking for
areas with distinct discoloration/staining, watermarks, rust,
damaged roof structure, clogged  drainages, or penetrations in
the roof.  The visual inspection was supplemented with data
from  previous  reports   of  a   testing   effort  designed  to
determine PCS  contamination  within  the  interstitial space.
Sampling  locations were  also identified  for the  biased and
unbiased samples.

The battery limits of this study were  from the floor of the
market to ceiling level and within the four walls.

The following  types  of  samples were collected  by the field
sampling personnel:

     1.   Wipes  -  From  non-porous  surfaces such  as metal
          poles/beams,  counter   tops and market  furniture,
          fixtures, food handling/preparation equipment, and
          floor drains.

     2.   Destructive -  From porous hard surfaces  such as
          wood, pipe insulation and rusted pillars.

     3.   Sediment - Standing sediment/solids from drainage
          outfalls, or "catch" samples from plastic or metal
          covers over stalls.

     4.   Water -  Drippings  from plastic  catch areas,  roof
          leaks, drains,  or standing water.

     5.   Dust  -  Primarily   from  floor sweepings   and  in
          corners.

     6.   Air - Continuous air samplers  to sample the ambient
          air in the market.

The actual sampling activities took place three days after the
initial reconnaissance visit.

3.   SAMPLING STATISTICS

A total of 55 field samples  (and 14 quality control samples)
were collected.  Table 1  gives  the breakdown of the various
types of samples collected.  All the samples were preserved on
                             1-314

-------
                         Table 1

   Summary of Environmental Samples Taken at the Market
Type of Samples
        Number of Samples for PCB Analysis
                 Test  Control  Duplicate  Field Blank  Total
Wipes

Destructive

Dust

Sediment/solids

Water
  a)  Unfiltered
  b)  Filtered

Air

Totals
>7 2
3
7
6
4
4
4 1
2
1
1
1
1
1
2
—
—
1
1
33
4
8
8
5
5
6
55
                                      69
                            1-315

-------
ice  and  transported  to the  Weston  Analytics  Laboratory
following completion of sampling activities.  The individual
sampling procedures are discussed in subsequent sections.

4.   SAMPLING PROCEDURES

4.1  wipe
Wipe sampling was conducted on 27 non-porous surfaces within
the market.   Samples were collected from a variety of surfaces
such as counter tops,  eating tables, freezers, steel pillars,
cooking range hoods,  glass cases,  and others.   The sampling
locations were spread out over the entire market in order to
get representative  coverage.    See Appendix A  for complete
procedure .

The wipes were divided into 2 categories:

  •   Biased - discrete
  •   Random - composite or discrete

The biased samples were collected as discrete samples.  These
samples were collected where visual observation and review of
previous reports suggested a potential for PCB contamination.

The  random  samples  were collected   at  various  locations
distributed within the building.  These samples were collected
either as discrete or composite samples.  The discrete samples
were collected on structures such as pillars, food cabinets,
and range hoods.  Composite samples were generally collected
on larger surfaces such as long counter tops.

The composite samples  were collected by taking three separate
hexane  soaked gauze  pads and  wiping each  pad in  various
locations over a given surface and then collecting them in one
sample bottle.   The sampling  area  for composite samples was
three times as large as that for the discrete samples  (300 cm2
vs. 100 cm 2) .  The analytical results were then adjusted to
consistent units of 100 cm2 for all  samples to facilitate data
comparison .

A  total  of six  additional  wipe  samples were  collected  as
quality control  (QC)  samples.   Of  these,  two were  controls,
two were duplicates, one equipment  blank and one field blank.

The control samples were collected from surface areas within
the market  which  suggested  the least likelihood of  being
contaminated  (such as  inside  a closed food cabinet.)    The
equipment blank  sample was collected  by wiping the aluminum
foil  covered template  with  the  hexane soaked  gauze  and
analyzing for PCBs.   The  field blank  sample consisted  of a
laboratory  prepared  wipe  sample  gauze  pad  in the  sample
                            1-316

-------
container taken to the market and returned for analysis along
with the other samples.

The samples were placed in 250 ml wide-mouth glass jars (soil
sample type jars) and preserved at 4°C by ice.

4.2  DESTRUCTIVE  (CHIP) SAMPLING

Three destructive samples were taken of hard porous surfaces.
Three different  porous mediums  were selected.   They were a
piece of  wood,  rust  from a  pillar and wrapping  from pipe
insulation.

All the samples  selected were biased samples from different
areas of  the  market.    The  rust sample  from a  pillar  was
collected  near the wood  ceiling.    The  insulation wrapping
sample was collected from a drain pipe which comes down from
the shed.   The wood sample was collected from a temporary wall
which was  erected adjacent to  a pillar in a corner of  the
market.   A duplicate sample of the wood was taken at the same
location.

Samples were collected using a decontaminated stainless steel
trowel or  chisel.  Samples were collected  in a 250 ml wide-
mouth jar with teflon-lined lid and preserved at 4°C using ice
placed  in  an  ice  chest.    See Appendix  B  for  complete
procedure.

4.3  DUST SAMPLING

The team collected eight dust samples including one duplicate.
Dust samples  were collected from, a  variety  of  surfaces  and
locations.  Two samples were collected from floor sweepings of
two aisles, one sample from on top of the men's room roof, one
from the louvers of an  air exchange unit, one from a wall fan,
one from the  cold storage room screen, and the last one from
a pipe near the roof directly across from an air conditioning
unit.   All   samples  were  preserved on  ice as  previously
described.

4.4  SEDIMENT/SOLID SAMPLING

A  total of  six  sediment  samples  were  collected  from  the
market.   Additionally, one duplicate sample and a field blank
were also  taken.  Four  sediment samples  were  collected in
separate locations from the plastic suspended from the ceiling
to catch water and solids which  had  dripped or fallen in from
the  shed  above.    These  plastic  sheets  are  predominantly
located along the perimeter of the market.

Except in  one area,  no water leaks  were visually evident on
the day of the sampling.   The sediment and water collected on
                             1-317

-------
the  plastic  was  potentially an  accumulation  from  prior
infiltration.  It should be noted that during the initial site
visit these  plastic  sheets were filled with  water and some
were overflowing due to the  severe rain storm.   However, by
the day on which the sampling took place, much  of the standing
water was absent.  Moist sediment and some water remained.

Except for one sample which was from  a floor drain, the other
samples were  from  solids/sediments which apparently dropped
from the ceiling level.  All  samples  were preserved on ice as
previously described.

4.5  WATER SAMPLING

Water samples were  collected in areas  where there was stagnant
or standing water.   These areas included canopies of some of
the stores and the suspended plastic surrounding some of the
roof leak areas.  The water samples were collected in 950 ml
amber glass jars with teflon-lined caps  and preserved at 4"C.

Eight  water  samples  were taken from  four  locations   (two
samples per location) plus two duplicates (one  location).  For
each location, one of the  two samples was filtered and then
analyzed  for PCBs  while the  other  sample was  not filtered
prior to  analysis.  This protocol was used to assess if the
PCBs in the sample, if any, were potentially associated with
the water phase or the suspended sediment/solid phase.

The  water samples collected were from roof  leakage which
either collected onto the plastic sheeting beneath the leak or
from a bucket under the leak or from a trough at roof level.

4.6  AIR SAMPLING

Five  air  samples were   taken  at   locations  inside  and
immediately outside the market building.   Four samples were
collected at  inside  locations, two from diagonally opposite
corners and two samples within the active space of the market.
One sample was drawn from a location outside the building to
serve as  a  control sample.  One blank  sample tube was also
analyzed as a field blank.

Air samples were drawn over an eight-hour period to determine
time  weighted  average  concentrations  and   for  ease  of
comparison  to OSHA  Permissible  Exposure  Limits  (PEL)  and
Threshold Limit Values  (TLV)  as set by the ACGIH.  Collection
of the analyte was accomplished  as outlined in the National
Institute of  Occupational  Safety and Health  (NIOSH)  Method
5503 as modified by Versar, Inc.  to provide for the sampling
of a greater volume  of  air and to provide a  lower limit of
detection  of  the  analyte.    The   analytical  methods  and
detection limits are discussed in detail in Appendix C.  The
                             1-318

-------
control sample was  collected  from outside the building near
the entrance to the market.  This  was accomplished by running
the source tube outside while  the  sampler  remained inside the
market.

5.   RESULTS AND CONCLUSIONS

The analytical results are confidential to the client, but, in
summary, none of the active areas  of the market were found to
contain PCBs at levels above the analytical detection limits.
This includes  the  air and wipe samples  from the restaurant
areas.   Trace amounts  of  PCBs were found  in a few  of the
sediment and chip samples from isolated perimeter areas of the
facility.  These appear to be the result of an accumulation of
residues from various small leaks over a period of time, and
can be removed by routine maintenance operations.

6.   SUMMARY

This survey  of the extent  of potential PCB contamination in
the air, dust, sediments, etc.  and on various porous and non-
porous surfaces within the market was completed within a week.
This included one day for reconnaissance and identification of
sample  locations,   one  day of  sampling,  and production of
validated analytical result within 72 hours by the laboratory.

The  sampling  scheme  represented,  both  statistically  and
logistically,  a good survey  of various  matrices within the
market.    The  results  indicating  a  lack  of  detectable
quantities of  PCBs  in the active  market and restaurant areas
was  important  to  the continued safe  operation  of  these
facilities.
                             1-319

-------
                         APPENDIX A

                 Procedure  for Wipe  Sampling


1    Prior to field activities,  3"x3" gauze  pads are soxhlet-
extracted in  the laboratory with hexane and placed  in the
laboratory-cleaned  glass  sample containers  equipped  with
teflon-lined caps.

2    Bring dedicated, prepared  gauze pads  (secured in glass
containers)   to  sample  site.    Select  appropriate  sample
location  and  area.    Photograph  area  to be  sampled,  if
necessary.

3    Measure  area to  be  wiped or  use dedicated aluminum
template to mark area.   Generally, a 100 cm2 area is sampled;
however, a smaller or larger area may be wiped,  depending on
the degree of  cleanliness  encountered  in the  field.   Record
size of area to be sampled.

4    Put on a clean pair of surgical gloves.

5    Hold gauze pad with clean glove and initially wipe sample
area in a horizontal direction  using a forward  and backward
motion.  Wipe sample area a second time with a clean portion
of the same gauze pad in a  vertical  direction using a forward
and backward motion.

6    After wiping, replace the  gauze pad in the appropriate
laboratory-prepared container and secure the teflon-lined lid
on the sample container.

7    Duplicate wipe samples will be  taken in an area directly
adjacent to the original sample location.

8    Attach the  sample  label with the  sample identification
number  and  other appropriate  sample   information.    Apply
custody seals and place in a plastic self-sealing bag.

9    Record all pertinent information in the site log and, if
appropriate, on the site map,  and complete the sample analysis
request form and chain-of-custody record.

10   Follow the sample documentation, packaging, shipment, and
chain-of-custody procedures.
                            1-320

-------
                         APPENDIX B

                       Procedure for
                  Destructive/Chip  Sampling


1    Select  appropriate  sample  location  and  record/mark
location and area.  Photograph area to be sampled.

2    Put on a clean pair of surgical gloves.

3    Using a decontaminated chisel and hammer,  proceed to chip
material to  a  depth of less  than  2 cm, taking  care not to
scatter pieces outside the marked area.  Clean, dedicated, or
decontaminated  aluminum pans or dust pans  may be  used to
shield the  area to prevent pieces  from  scattering.   Record
area and depth of sample.

4    Using a dedicated brush and dust pan or tweezers, collect
the sample and transfer to an appropriate laboratory-cleaned
container and secure the teflon-lined lid on the container.

5    Duplicate  samples will  be taken  by  homogenizing  the
sample material by mixing in an  aluminum container.   The
duplicate portion of  the  sample  will be taken from the same
container as the original sample.

6    Record all pertinent information in the site log and, if
appropriate, on the site map, and complete the  sample  analysis
request form and chain-of-custody record.

7    Follow the sample documentation, packaging, shipment, and
chain-of-custody procedures.
                            1-321

-------
                         APPENDIX C

                Analytical Methods for PCBs

The WESTON Analytical Laboratory used the  following methods
for analysis of PCB samples.  Method references are to EPA SW-
846 (Test Methods for Evaluating Solid Wastes).

I   Solid  and Water Analysis
    •  Analytical  Method  - EPA Method  8080
      This is a  gas  chromatographic  method for  analysis of
      PCBs in various matrices.  Prior to use of this method,
      appropriate sample extraction techniques,  as described
      below,  are  employed.

    •  Extraction  Methods
      Solids: Method  3540, Soxhlet Extraction
      Water:   Method  3520, Liquid/Liquid Extraction

    •  Detection Limits
      Both the extraction methods and the  analytical methods
      referenced  from  above will,  in  most  instances, yield a
      detection  limit of  0.1 ppm with acceptable accuracy.
      Some matrices, such as  those with a high organic and/or
      bituminous   content,   can  present   interferences  to
      achieving this  low detection limit.

      Appropriate "cleanup" procedures, such as the florisil
      method   (3620)    are   employed   to   eliminate   the
      interferences,  if  necessary.   However,  there may be
      instances,  such  as PCBs on oil-based paint surfaces or
      PCBs on  tar-based  roofing  materials,  where  precise
      analysis  down to  0.1 ppm is not possible.   Detection
      limits in these  cases can be in the  0.5 ppm range.

      Detection limits for  wipe samples are in the 0.2 to 1.0
      ug/wipe (100 cm2) range.

II  Air Sample Analysis

    •  Method: NIOSH 5503

      This is  the standard   NIOSH method  for  analysis  of
      Florisil sorbent samples.  The normal working range for
      this procedure yields results with a detection limit of
      10  ug/m3.   There is a modification employed by Versar,
      Inc., in New York,  in  which a larger Florisil sample
      tube and larger  air volume  are employed.    This  can
      generate accurate  results in the 0.1  to 1  ug/m3 range.

Ill   Quality control  (QC)

    All standard EPA and/or Contract Laboratory Program (CLP)
    laboratory and field QC protocols are  followed by WESTON
    including use of  blank  samples  and  replicate  samples.
    These  QC samples  are an  integral part  of the analytical
    scheme and are important in substantiating validity of the
    analytical data.  All sample data and QC data is reviewed
    and validated prior  to  issuing  a  report.

                           1-322

-------
42       COMPARISON  OF THE HNU-HANBY FIELD TEST KIT PROCEDURE
              FOR SOIL ANALYSIS WITH A MODIFIED EPA SW-846
                           5030/8000 PROCEDURE


     John D.  Hanby.  Technical Director,  HNU-Hanby;  Bruno  Towa,
     Chemist,  Hanby Analytical Laboratories;  HNU  Systems,  Inc.,  160
     Charlemont Street, Newton Highlands, Massachusetts  02161-9987


     ABSTRACT

     A sample generation procedure for  the  preparation  of gasoline
     in soil standards  was  developed to produce  homogeneous  blends
     of consistent concentration and stability.  Three sets of these
     standards prepared  at different concentrations were  analyzed
     utilizing the  HNU-Hanby  Field test procedure for  analysis  of
     soils and a modified EPA SW-846, 5030/8000 GC procedure.   The
     relatively high  degree of consistency  of the  concentrations
     within  each set of the  gasoline/soil  standards allowed a
     statistically meaningful comparison between the accuracy of  the
     two methods.
     INTRODUCTION

     Concurrent with  the development of  methodologies  for  the
     chemical analysis of environmental samples  has,  perforce,  been
     the  search for  representative  analytical standards.    This
     principle, of course,  is  "sine qua non" to  accurate  chemical
     analysis, but probably in  no case  is it more de rigueur than in
     the  analysis  of  soils.   Soil  matrices  vary from  relatively
     simple  configurations,  e.g.  quartz  sand, to  those hellish
     quagmires we call "toxic waste  sites".

     This investigation focuses on a  problem of sufficient magnitude
     to be  of large  scale  environmental  concern  yet  still  retain
     tractability in  an  analytical  sense.   Underground  fuel  tanks
     are  generally bedded in sandy soil, primarily, because of
     structural concerns for tank and associated line integrity.  A
     procedure was developed at HNU-Hanby Environmental Laboratories
     which facilitated the preparation, packaging, and analysis of
     gasoline contaminated sand samples which could  produce  fairly
     large numbers (100-200) of homogeneous blends of the mixture at
     different concentrations.    Three  different  sets  of  the
     gasoline/sand standards were prepared and subjected to analysis
     using the HNU-Hanby  field test kit  extraction/colorimetric
     procedure and modified EPA SW-846  5030/8000 GC methods.
                                   1-323

-------
PROCEDURE

An  alluvial  sand from a  Houston  nursery,  typical  "river  bank
sand", was  obtained in  sufficient  quantity to  allow for  the
preparation of  a large number of samples  for  this as well  as
subsequent investigations.  First experiments with this soil in
attempts to produce homogeneous  mixtures of fuel  contaminated
aliquots  were  frustratingly  unsuccessful.    Microscopic
examination of  the  soil revealed that  it typically  contained
relatively large  "chunks"  of clay  interspersed  with the  more
abundant quartz sand.  The obvious remedy  for  this  problem was
to  wash the  more water-dispersive  (and  much  more  organically
sorbent)   clays  out  of  the  soil matrix.    This  washing  was
performed with  warm laboratory tap water  on  approximately  25
kilograms of  the soil.   Subsequent drying  of batches of  the
soil was carried out, rather tediously, in a laboratory vacuum
oven at approximately 80  C and 29" Hg.  A corollary  effect of
this vigorous approach to the removal  of  the  clay turned out to
be  the sterilization  evidenced  by the  fact  that  subsequent
washings of the  soil  in de-ionized  water produced no biota on
filter samples cultured on M-F endo  broth media plates.

Thus treated, a  2 Kg.  batch of the sand was introduced into a
gallon, glass screw-capped jar and placed  on a small  ball  mill
roller device.  A 25 ml  solution  of super  unleaded gasoline in
methanol of approximately  10%  was prepared in  a 60 cc  plastic
syringe fitted  with  a  26  guage needle.    This solution  was
sprayed through a small hole, previously drilled in the cap, as
the bottle turned on the roller device.   The mixture was turned
for several hours with occasional hand shaking of the  jar and
intermittent  tapping  with a  small  wooden mallet to  dislodge
sand which accumulated on the sides  of the jar.  Observation of
this decreasing tendency of the sand to adhere  to the glass was
used  as  the  indication of an  appropriate  end  point for  the
procedure.  That is, the mixture was  tumbled for approximately
one hour after cessation of appreciable  adherence  of  the
mixture.   The jar was  then removed  from the ball  mill  roller,
and, using a  specially prepared  dispenser rack, the  sand was
very rapidly transferred to  1  dram  screw cap  vials which  were
immediately capped  and placed  in refrigeration  at  4   C.    the
dispenser rack  was  designed to hold  24  small plastic  funnels
positioned over  24  of  the 1 dram  glass  vials  so that  rapid
removal and  capping  of the  full vials  was  facilitated.    A
vibrator  was  attached  to  the filling  rack to  promote  rapid
funneling and settling of the sand into the vials.

Three  separate  concentrations  of  the  gasoline/sand  mixtures
were  prepared,  designed to  give  final concentrations  of
approximately 500, 200 and 50 mg/kg.  Twenty samples  from each
set (approximately 100 in each set), were randomly selected for
                               1-324

-------
analysis  by  each of the  two  methods.   The  average  weight of
sample  in the 1 dram  vials  was 6 grams.   Actual weights for
each sample were utilized in the calculations of  concentration
(mg gasoline/kg sand).

The method  utilized with the  HNU-Hanby  field test  kit  is as
follows:

1.  Empty sample vial into the 50 ml  beaker.

2.  Immediately snap one 10 ml solvent  ampoule from the kit and
    empty into the beaker.

3. Stir the sample/ solvent mixture  with  a  spatula  for  three
    minutes.

4.  Pour solvent from the beaker into one  of  the kit  test tubes
    up to the mark (4.2 ml).

5.  Add contents of one of the catalyst vials from the kit into
    the test tube.

6. Shake  test tube  vigorously for three minutes and observe
    developed color in the catalyst at  the bottom  of  the tube.

Comparisons of  the  color with  standard   gasoline in  soil
photographs supplied with the  kit  as well as with photographs
made  specifically  to  provide  matches  with  the  standard soil
concentrations obtained in this  investigation were facilitated
by juxtaposition of  the two sets ,of pictures with the actual
test tube results.  Apparent differences in hue can be seen in
the color photographs accompanying this  paper.   These hue
differences can  largely be accounted  for because of variable
composition in the make up of the gasoline used in the original
kit  supplied  photographs  and the gasoline  used in  this
investigation.   Use of black  and  white  photography minimizes
the bias  that these hue  differences can cause.   A quotation
from MIT Professor of  Physics Phillip Morrison's book The Ring
of Truth  is  appropriate  here,  "...spectral photos  in  black
white...bear the full information of  the spectrum." *'

The GC methods utilized in this investigation are adapted from
the EPA SW-846 manual, 5030/8000 and from a  study conducted by
the Midwest Research Institute for the U.S.  Environmental
Protection Agency's Office of Underground  Storage  Tanks.    The
method  employed  the  following procedures  and  instrumental
parameters:
                              1-325

-------
1.  40  ml VOC vials  were  placed on a  top loading balance  and
    tared.

2.  A sample  soil  vial was  emptied into  the  voc vial.    The
    weight of sample was recorded.

3.  Immediately, de-ionized water was added to carefully bring
    the meniscus of  water to  slightly  above  the  voc vial top.
    Another weighing was recorded after this step to determine
    the weight (volume) of water added.

4.  The vials  were  placed  in  a Fisher  Scientific  Bransonic
    ultrasonic water bath for  15 minutes,  then removed to  and
    Eberbach shaker for 30 minutes on high speed.

5.  The  voc  vials were  then  analyzed  on an  HNU  Model  421  gas
    chromatograph equipped  with a photoionization detector
    connected in series with  a flame ionization detector.  5 ml
    of water  was poured  from  the  voc  vial  into a  gas tight
    syringe.  This  was then connected  to an O.I. Corporation
    Model 4460A  sample concentrator.    The  sample was   sparged
    with helium for 11 minutes at 25 C.

The trap was desorbed at 180  C for 4 minutes  through a heated
transfer line to the GC.  GC  conditions were:   25  meter Nordion
fused silica .53 mm column with a 1.0 micrometer MB 30 coating.
The temperature  program employed was:   initial temperature 45
C, 2 minutes, then 10  C/minute to 100  C, hold for 3 minutes.
Data was sent  to  a Spectra  Physics  4270 integrator.   GC
parameters  were initiated  via  an  IBM  PS/2 Model   70   386
channeled through the 4270.

A standard solution for calibration  of  the GC runs was prepared
composed of the gasoline plus added concentrations of 2-methyl
pentane and 1,2,4-trimethylbenzene.   The GC standard curve  was
prepared by analysis of various concentrations of the gasoline
in methanol standards which were injected through the front of
the 5.0 ml sample syringe using a 10 or 100 ul syringe.  Using
the  protocol  established in  the MRI   study  peaks  considered
typical of the gasoline range  organics  elute inclusive of  and
between  the 2-methyl  pentane and  the  1,2,4-trimethylbenzene
peaks.    Total  area counts from  the FID as  integrated  by  the
SP4270  were used  in  this  investigation.   Further  studies
utilizing  these  same  standards  are  planned which  will
incorporate the  data from the  photoionization detector which
will correspond  to  the SW-846  5030/8020  method.    A separate
series  of GC data  charts indicate relevant  quality  control
methods for blanks,  surrogates and spiked  samples.
                              1-326

-------
All  stock solutions,  standard  dilutions,  and prepared  soil
standards were stored at 2 -4  C in a walk-in refrigerator.   A
primary  consideration  in  the  design  of  this  study  was  the
determination of the temporal stability of the soil standards.
The period involved  in this  investigation was approximately  8
weeks.   That is, soil  samples  were analyzed by  both  methods
(FTK and GC)  over a period of some  56 days.

Figure  1 is  a  photograph  of   the HNU-Hanby  Field  Test  Kit
utilized  in  the investigation.    The components  pertinent  to
this study are  the  50  ml beaker, the  10  ml  solvent  ampoules,
the marked test tubes and the 1  dram vials of catalyst.   Figure
2 is  a chromatogram typical of  the flame ionization  detector
peaks described in the  report.


SUMMARY;

The  utility of  this method of preparation of consistently
uniform samples of  gasoline  in  soil  is  realized  in  the
relatively small variability of the data  obtained  by both the
HNU-Hanby Field Test Kit  as  well as the EPA purge and  trap/GC
methods of analysis.   Correlation between the two methods
themselves, in this visible comparison manner, is also evident.
The gasoline range organic method of analysis as documented in
the  MRI study  utilizing  FID chromatographic  detection  gives
consistent comparison  to  the test kit  results  even though  GC
method  is  essentially  a total organic  integration whereas the
test  kit is based  on  the  colorimetric determination of  the
total aromatic components of  the gasoline.

The  utilization  of photoionization detection being much  more
responsive  to   the  aromatic   components  in  gasoline  will
doubtless be  even  more   closely correlative  with   this
extraction/colorimetric  technique.   Investigations of  this
correlation  are already  underway  as well  as  more  elaborate
techniques  involving   the   utilization   of   reflectance
spectrophotometry which,  of  course, will provide a means  of
obtaining data with  the technique  which will not be dependent
on visual observation of relative color intensity.  A  study is
also in  process which will add a third dimension of analytical
measurement, i.e.,  the  utilization  of  a  head space vapor
technique.
                             1-327

-------


-------
CHflNNEL fl     R29.RflW
               INJECT  05/28/91  16:42:28
   PffT
                        Solvent Peak
           4. 88
              4. 83
                                                      .62
                                     2-Methylpentane
                                     1,2,4-Trimethylbenzene
          0       10. 08
  1.fl DfiTfi CflPTURED TO:  \LflBNETNR29.RflM
 TPHflfl004
FILE 1.
PEflKtt
1
--,
C-
3
4
5
6
7
8
3
10
11
12
13
14
15
METHOD
FIRE FIX
31.
7.
1.
0.
0.
0.
0.
6.
0.
0.
0.
0.
0.
0.
0.
59
61
027
645
205,
655
17
17
848
303
323
044
214
195
001
0.
RT
0.
2.
J.
4.
4.
5.
5.
6.
6.
7.
7.
8.
8.
9.
10.
RUN 33
flREfl
62228452893
91
47

83
F,
87
24
72
31
82
27
67
44
08
21308942
2876678
1306063
573927
1834958
476975
17275023
2373753
847648
305771
122859
593116
544697
2528
BC
02
02'
02
02
02
02
02
02
02'
02
02
02
02
02
03
 TOTHL
                  95/28/31 16:42:28     CH= "fl"  PS= 1.
                           INDEX  43
100.
280001331
                                      Gasoline Range Peaks
  1. fl Sending Report...Done

   Figure  2   Typical Chromatogram of  Gasoline  Range Organics
                               1-329

-------
ACKNOWLEDGMENTS:

The authors are grateful to the U.S.  EPA Office of  Underground  Storage
Tanks and  to the Midwest  Research  Institute,  particularly Ms.  Linda
McConnell, for inviting this laboratory's participation in the  round-
robin  study  designed  to produce  a  technically  defensible Total
Petroleum Hydrocarbon method of analysis of  environmental  samples.


REFERENCES;

1.  Morrison,  Phillip  and Phylis,  1987,  The  Ring of  Truth,   Random
House, Inc., New York, N.Y.,  p. 227.

2.  Midwest Research Institute Draft Report,  Nov.,  1990, Evaluation of
proposed analytical methods to determine total  petroleum  hydrocarbons
in soil and groundwater, MRI, Falls  Church,  VA.

3.   U.S.  Environmental Protection  Agency,  SW-846 Test  Methods  for
Evaluating Solid Waste, 3rd Edition.
                            1-330

-------
43                   FIELD TEST KIT FOR QOANITiraNG ORGANIC HALOGENS
                                      IN WATER AND SOIL

                          Deborah Lavigne, Quality Control Manager
                                     Dexsil Corporation
                                    One Haraden Park Drive
                                 Hamden, Connecticut  06517

         ABSTRACT

         In a continuing data-gathering program, the EPA monitors organic chemicals
         in the waters of the United States.   The list of monitored chemicals
         includes both aliphatic and aromatic hydrocarbons,  pesticides,  industrial
         chemicals, plasticizers,  and solvents.  Many of these materials are
         halcgenated, produced by chlorination of water during purification
         processes, through industrial and municipal runoff, natural sources,  and
         sewage purification practices.

         Chlorine is a contaminant often found in oils, soils, sludges,  and organic
         liquids found at hazardous waste sites.  Controlling wastewater discharges
         and landfilling of chlorinated compounds have become priority issues for
         EPA since the passage of the Hazardous and Solid Waste Amendments  of 1984.

         In response to toxicological and environmental concerns of trihalomethanes
         and other halcgenated compounds present in water and soil, a quick,
         accurate, easy to use, portable field test kit has been developed  for
         quantifying organic halogens.  The analytical procedure requires an
         extraction with a suitable solvent, followed by colorimetric chemistry to
         quanitfy the organic halogens present.

         This paper will detail field and laboratory results, limits of
         detection, matrix effects, and cost analysis.


         INTRODUCTION

         EPA regulation 40 CFR 261 establishes that any used or waste oil
         containing greater than 1000 ppm organic chloride may have to be
         classified as a hazardous waste.   Chlorinated solvents are the primary
         contaminants found in waste oils and oily wastes.

         Currently available instrumental methods of chlorine analysis
         (microcoulometric titration, X-ray fluorescence spectometry, oxygen bomb
         combustion and gas cnromatography) are time consuming and must be
         performed in a laboratory by trained technicians.  Foreseeing the
         additional testing that would be required under the new regulations, the
         EPA Region II contracted Dexsil Corporation to develop a field-portable
         test kit that could be used by untrained personnel.  The result was two
         small, disposable test kits that require less than five minutes to
         determine chloride contamination in waste oil.  The first method is a
         go/no-go test, indicating over or under 1000 ppm chloride.  The second
         method is a quantitative analysis giving an amount of contamination
         between 200 - 4000 ppm.
                                          1-331

-------
These test kits were evaluated Toy Research Triangle Institute (Raleigh,
NC) for EPA and were found to be acceptable methodology for chlorine
detection.  As a result, the kits were assigned EPA method 9077,  to be
published in the upcoming SW-846 manual.   Interest has since increased in
a test kit that would work on oil containing large quantities of water
(oily waste) and, in light of the current regulations pertaining to
leaking underground storage tanks, it would be useful to have a kit that
would detect total organic halogens in soil.  Two field-portable test
procedures have been developed which address these issues of halogens in
wastewater, oily waste, and soils.

The different methodology and apparatus will be described, the accuracy
and precision of each method discussed and the costs of each method
reported.
USED OIL CONTAMINATION

How do chlorinated solvents contaminate used oil?  Chlorinated solvents
are not ingredients of crankcase oil, but are indirectly introduced
through careless management practices, such as pouring used degreasing and
cleaning solvents into used oil storage drums.  The most common solvents
found in waste oils are dichlorodifluoromethane, trichlorotrifluoroethane,
1,1,1-trichloroetnane, trichloroethylene, and tetrachloroethylene (1).
Levels of contamination range from 100 ppm to thousands of ppm.  The
possible presence of chlorinated solvents can be flagged by checking total
chlorine, an indicator of the potentially hazardous chlorinated substances
present.

The EPA estimates that over 350 million gallons or about 30 percent of all
used oil is landfilled or dumped annually.  Approximately 160 million
gallons comes from the "do-it-yourself" oil changers, who typically
dispose of their oil by dumping it on the ground, in sewers, or in
waterways, or by placing it with the household trash destined for a
landfill that has not been lined to protect against soil and groundwater
contamination.  The remaining 190 million gallons is dumped or landfilled
by automotive shops and industrial facilities.  (2)
OILY WASTE SOURCES

Sources of oily waste include bilge and ballast, rain runoff, washings
from cleaning vehicles and tanks, and cutting oils.  All of these
materials are predominantly water, containing from 0.1 to five or ten
percent oil.

Bilge oil is a mixture of fuel oil, lubricating oil, and hydraulic oil
dispersed in sea water along with dirt, rust, and bacterial sludge.
Ballast oil composition depends on what is carried in the ballast tanks
when the ship is not in ballast, usually fuel oil, crude oil, or petroleum
products.  The oil will usually exist as free oil droplets in the
seawater, or as a sheen on the water surface.
                                 1-332

-------
Bain runoff that carries oil from contaminated areas often cannot be
legally discharged to storm sewers.  Trucks and fuel storage tanks  are
cleaned with water containing detergents.  This produces oily water
containing solids, emulsions, free oil, dissolved oil, and detergents.
Metalworking fluids are used for both lubrication and cooling in various
machinery processes such as cutting and grinding.  Oily waste resulting
from used oil mismanagement causes damage to streams, ground water, lakes,
and the oceans.  For instance, the Coast Guard estimates that sewage
treatment plants discharge twice as much oil into coastal waters as do
tanker accidents - 15 million gallons per year versus 7.5 million gallons
from accidents.  A major source of this pollution is dumping of oil by
do-it-yourselfers into storm drains and sewers.  A startling example of
this has occurred in the Seattle area, where more than 40 percent of the
water quality trouble calls received are related to used oil and other
wastes dumped down storm drains, contaminating water bodies (3).
ENVIRONMENTAL IMPACT

Many contaminated sites containing oily wastes and oily waste sludges are
now being cleaned up under authority of Superfund.  The Superfund
regulations affect the handling of oil wastes in the areas of spills and
accidental releases, leaky storage tanks, and abandoned storage
facilities.  Oils from abandoned storage facilities fall into one of three
catagories:  Abandoned tank pumpings, abandoned drummed oils, or sludge
pit residues (4).

The composition of the oils in each of these catagories can vary
significantly from site to site.  Over time, the oils in tanks and drums
absorb material from the walls of the container.  This process is
exacerbated by corrosion due to seasonal temperature variations, rain,
mechanical abrasion, and the like.  The oils are usually significantly
diluted by water infiltration.  In order to fall under Superfund
jurisdiction the sites must present a danger to the public or the
environment.  Thus the emphasis is on the quick and inexpensive analysis
and disposal of the materials, rather than on recycling and reuse (5).
Ideally, hazardous waste determinations, whenever possible, should be
carried out in the field to quickly identify the extent and magnitude of
the contamination.  The advantages of alternative simple chemical tests
have been foreseen by the EPA and some procedures have, in the face of
alternative instrumental methods, been examined and subsequently become
EPA approved.
A CHEMICAL METHOD FOR THE DETERMINATION OF ORGANIC HALOGENS IN WASTEWATER,
OILY WASTES AND SOILS

This procedure requires  an  extraction with a suitable hydrocarbon
solvent.   Covalently bonded halogens present in the hydrocarbon solvent
are  then stripped from their solvent backbones by sodium metal according
to the Wurtz reaction:

                        2Na + 2R-X 	> 2NaX + R-R
                                    1-333

-------
Any halogens that are present (now in ionic form)  are extracted into an
aqueous buffer, to which is added a color reagent to quantitate resulting
chloride.  A solution of mercuric nitrate is added dropwise until a color
change from yellow to purple is realized, and the concentration (in ppm)
is read directly off the dropper.
ANALYTICAL MEIHDD

I/Method for Samples Containing Water
10 ml of the liquid sample is extracted by shaking for one minute with 10
g of an Immiscible hydrocarbon and 0.5 g of a (granular)  emulsion breaking
material.  The sample is allowed to settle until it has separated into
distinct phases (about three minutes).  Approximately one-third of the top
layer is dispensed into a vial containing a drying agent which will  remove
any moisture and inorganic chloride.  The vial is shaken and the drying
agent is allowed to settle.  0.34 g of the dried solvent is then treated
with 1.5 ml of a solution of naphthalene in ethyl diglyme followed by 0.4
ml of organic dispersion and metallic sodium, and shaken for 1 minute.   7
ml of buffer solution is then added and the aqueous layeris separated and
combined with 0.5 ml of a solution of s-diphenyl carbazone in alcohol.   A
solution of mercuric nitrate is added dropwise from a 1 ml microburette.
When a true purple color is realized,  the test is stopped and the chloride
     ntration of the original oil/water or wastewater sample is read
directly off the microburette.

2/Method for Soil Samples
10 g of the soil sample is extracted by shaking for one minute with 12 ml
of a mixture that contains 2 ml of distilled water and 10 ml of  an
immiscible hydrocarbon.  The soil is then allowed to settle and  the
supernatant liquid filtered through a column containing florisil to remove
any moisture and inorganic chloride.  0.34 g of the dry filtrate is then
treated with 1.5 ml of a solution of naphthalene in ethyl diglyme followed
by 0.4 ml of organic dispersion and metallic sodium, and shaken  for 1
minute.  7 ml of buffer solution is then added and the aqueous layer is
separated and combined with 0.5 ml of a solution of s-diphenyl carbazone
in alcohol.  A solution of mercuric nitrate is added dropwise from a 1 ml
microburette.  When a true purple color is realized, the test is stopped
and the chloride concentration of the original soil sample is read
directly off the microburette.

ANALYTICAL TESTS, RESULTS AND DISCUSSION

The samples chosen were both laboratory mixtures and Superfund samples
containing a range of 125 ppm to 6500 ppm chloride.  The procedures
employed are the same as those previously described except a packed kit
was used (Hydrodor-Q   , Dexsil, Bamden CT).  All reactions with this
kit are carried out in sealed plastic tubes and all reagents are contained
in crushable glass tubes to obviate any need to handle the reagents. This
is advisable, as some of the reagents are hazardous to handle in the
normal manner.  The results obtained from the laboratory samples are shown
in table (1) and table (2), and the results from the Superfund samples are
shown in table (3).  All three tables include results from the
microooulometric titration (EPA method 9076) of the same samples.


                                   1-334

-------
It is seen that the results f rom both the test kit and the
micxocoulometric titration of the samples agree very reasonably.   It is
also clearly demonstrated that no interference occurs in the presence of
inorganic chloride.  laboratory soil sarrples were also tested in the same
manner using an analytical kit (Dexsil, Hamden CT).  This is a similar
type of kit to the one used for liquids, but also provides a simple
balance for weighing out the soil.  The procedures previously described
were used and the results obtained for wet and dry soils are shown in
table (4) and the results for wet and dry sands are shown in table (5).
Mcrocoulometric titration results of the same samples are shown in each
table and it is seen that agreement is good between the two methods.

The cost of each kit is $10-13 and no capital investment in instruments is
needed.  The kits can readily and easily be used in the field and little
skill is needed.  The test takes about ten minutes.  With increasing
testing requirements, laboratory fees and laboratory turn-around times,
the field-portable chemical test with colorimetric end-point would be the
first choice for a suspect site or container, prior to laboratory
analysis.
                                   1-335

-------
REFERENCES

I/Guide to Oil Waste Management Alternatives, Final Report, p. 4-15,
Energy and Environmental Research. Corporation, Irvine, CA, April, 1988.

2/Nolan, J.J., Harris, C., and Cavanaugh, P., Used Oil;  Disposal Options.
Management Practices and Potential liability, 2nd Ed., p. 12, Government
institutes, Inc., Rockville, MD, 1989.
3/How to Set UP a
)il, EPA Eept. No.
530-SW-89-039A, p. 1, U.S. EPA, Washington, D.C., May 1989.
                                   1-336

-------
TABIE 1

COMPARISON OF lABORATORY PREPARED SAME>IEANALYSES:
mCBDCOUKmiKLC TITRATION VS HYDRCCLOR™

                                           Micrxxxulometric
Sample                  Hydroclor™        Titration
2000 ppm d"~ as         2000 ppm           1980 ppm
C12C2C12 k*             250° PP"11           246° PP"1
1% oil in pond H20

2000 ppm Cl~ in         2250 ppm           2250 ppn
previous matrix         2275 ppm           2210 ppm
+ dirt

1000 ppm Cl"" as         900 ppm            760 ppm
CgH3Cl3 in              1050 ppm           980 ppm
1% oil in pond H2O

1000 ppm Cl~ in         850 ppm            849 ppm
previous matrix         900 ppm            897 ppm
+ dirt

1000 ppm CT~ as         900 ppm            996 ppm
CHC13 in 1% oil         975 ppm            959 ppm
in pond H20 +
4000 ppm Cl" as  NaCl

1000 ppm Cl" in          1000  ppm          936 ppm
previous matrix          900 ppm            871 ppm
+  dirt
                                   1-337

-------
TABLE 2
COMPARISON OF IABORATORY PREPARED ANTIFREEZE SAMPLE ANALYSES:
mGRDOOOICMETRIC THRAHON ys HraRDdQR™
Microcxsulametric
Matrix
Tetrachloro-
ethylene in
antifreeze/HjjO
Same

Same

Trichloro-
ethylene in
antifreeze/H20
Same

Same

1,2-Dichloro-
ethane in
antifreeze/H2O
Same

Same

1,2, 4-Trichloro-
benzene in
antifreeze/H20
Same

Same

Chloroform in
antifreeze/H20
Same

Same

Sample
2740 ppn
2670 ppn

1230 ppn
1140 ppn
481 ppn
444 tXXCL
3000 ppn
3000 ppn

1200 ppn
1200 ppn
451 ppn
462 com
2950 ppn
2800 ppn

1400 ppn
1490 ppn
697 ppn
711 con
3260 ppn


1400 ppn
1640 ppn
812 ppn
791 ron
3090 ppn
2930 ppn
1300 ppn
1310 ppn
728 ppn
718 ppm
Titration
2690 ppn
2760 ppn

1280 ppn
1280 ppn
535 ppn
548 ppm
2810 ppn
2800 ppn

1120 ppn
1160 ppn
509 ppn
521 ppm
2820 ppn
2800 ppn

1370 ppn
1410 ppn
693 ppn
671 ppm
2880 ppn
2940 ppn

1510 ppn
1620 ppn
857 ppn
856 ppm
2930 ppn
2930 ppn
1410 ppn
1440 ppn
732 ppn
730 ppm
HydroClor™
2900 ppn
2850 ppm

1200 ppn
1350 ppn
500 ppn
500 ppm
3000 ppm
3100 ppn

1200 ppm
1250 ppn
600 ppm
600 ppm
3300 ppn
3300 ppn

1550 ppn
1600 ppn
800 ppn
800 ppn
2800 ppn
2800 ppn

1500 ppm
1500 ppn
800 ppn
825 ppm
2900 ppn
2800 ppn
1400 ppn
1350 ppn
800 ppn
725 DOT
                                     1-338

-------
TABLE 3

CCMPARISON OF LIQUID SUEERFUND SAMPLE ANALYSES:
                 TITRATION VS IKDRDCLOR111
Sample
TX - 563 ppm
TOX - 242 ppm

TX - 604 ppm
TOX - 315 ppn

TX - 2260 ppm
TOX - 1400 ppm

TX - 1910 ppm
TOX - 1690 ppm

TX - 6420 ppm
TOX - 5690 ppm

TX - 4940 ppm
TOX - 3980 ppm

TX - 1560 ppm
TOX - 712 ppm
Mica^coulometric
Titration
230 ppm
242 ppm

417 ppm
396 ppm

1187 ppm
1425 ppm

1539 ppm
1518 ppm

5750 ppm
5900 ppm

3270 ppm
3870 ppm

774 ppm
748 ppm
HvdroClor™
200 ppm
200 ppm

300 ppm
350 ppm

1350 ppm
1400 ppm

1600 ppm
1700 ppm

5800 ppm
5600 ppm

3600 ppm
3400 ppm

900 ppm
800 ppm
                                 1-339

-------
TABLE 4

COMPARISON OF LABORATORY PREPARED SOIL SAMPLE ANALYSES:
MCKOOOUICMETRIC ITERATION VS SOIL FIELD TEST KIT
Sample
500 ppn Cl~
in dry soil
600 ppm Cl~
in dry soil
700 ppn Cl~
in dry soil
800 ppn Cl~
in dry soil
900 ppn d~
in dry soil
1000 ppn d"
in dry soil
1500 ppn d~
in dry soil
500 ppn Cl~
in wet soil
600 ppm Cl~
in wet soil
700 ppm Cl~
in viet soil
800 ppn d~
in wet soil
900 ppn d~
in wet soil
1000 ppn d"~
in wet soil
1500 ppn d~
in wet soil
2000 ppn d~
in wet soil
Soil Kit
600 ppm
500 ppn
650 ppm
650 ppm
850 ppm
650 ppm
800 ppn
800 ppn
950 ppm
900 ppm
1000 ppn
950 ppn
1500 ppn
1450 ppm
500 ppn
450 ppm
700 ppn
650 ppn
750 ppn
800 ppn
800 ppm
800 ppm
900 ppn
950 ppn
1100 ppn
1000 ppn
1600 ppm
1600 ppm
2050 ppm
2000 ppm
MiCTOcoulametric
Titration
515 ppm
509 ppn
635 ppm
624 ppm
700 ppn
727 ppm
784 ppn
790 ppn
931 ppm
948 ppn
960 ppm
979 ppn
1450 ppn
1490 ppm
558 ppn
595 ppm
689 ppm
719 ppm
654 ppm
677 ppm
861 ppm
883 ppm
960 ppm
946 ppm
1070 ppm
1080 ppn
1520 ppm
1520 ppm
1860 ppm
1910 ppm
                                  1-340

-------
TABLE 5

COMPARISON OF IABORATORY PREPARED SAND SAMPLE ANALYSES:
MKRDCCUKMETRIC TTTRATICN VS SOIL FIELD TEST KIT
Sample
300 ppa Cl""
in wet sand

400 ppn Cl~
in wet sand

500 ppa Cl~
in wet sand

500 ppa Cl"
in dry sand

600 ppai Cl~
in wet sand

700 ppa Cl~
in wet sand

1000 ppa Cl"
in dry sand

1186 ppa Cl"
in dry sand

1200 ppo Cl"
in dry sand

1500 ppa Cl"
in dry sand

2000 ppa Cl"
in dry sand
Soil Kit
350 ppa
300 ppn

400 ppa
450 ppn

500 ppa
550 ppn

400 ppa
575 ppa
650 ppa

775 ppa
1050 ppa
1050 ppa

1200 ppa
1250 ppa

1200 ppa
1500 ppn
1550 ppa

1800 ppa
Microcx3Ulc3rnetric
Titration
312 ppa
315 ppa

421 ppa
429 ppn

452 ppa
457 ppn

533 ppa
528 ppa

633 ppn
632 ppn

823 ppn
812 ppn

1110 ppn
1220 ppn
1200 ppn
1200 ppn

1570 ppn
1510 ppn

1880 ppn
                                1-341

-------