Agency
Proceedings Of The
Fourteenth Annual EPA
Conference On Analysis Of
Pollutants In The Environment
May 8 & 9,1991
-------
UNITED STATES ENVIRONMENTAL PROTECTION AGENCY
OFFICE OF WATER
OFFICE OF SCIENCE AND TECHNOLOGY
ENGINEERING AND ANALYSIS DIVISION
FOURTEENTH ANNUAL EPA CONFERENCE ON
ANALYSIS OF POLLUTANTS IN THE ENVIRONMENT
MAY 8 & 9, 1991
-------
FOREWORD
The Fourteenth Annual EPA Conference on Analysis of
Pollutants in the Environment was a resounding success. The
Conference was attended by over 300 scientists from
industry, environmental laboratories, and state, local, and
federal government agencies. The Conference provided the
attendees with the opportunity to discuss the latest
developments in analytical methodologies for the
determination of pollutants in the environment.
These proceedings document 24 technical and policy
presentations on subjects ranging from advanced sample
preparation and data reduction technigues for GC/MS analysis
to EPA's efforts towards analytical methods integration and
the implementation of good automated laboratory practices.
We would like to thank Jan Sears of ERCE for
coordinating the conference, Harry McCarty of Viar for his
assistance in arranging the technical program and all the
others who helped make the Fourteenth Annual Conference a
success. We are looking forward to your attendance at the
Fifteenth Annual EPA Conference in May of 1992.
W. A. Telliard
-------
FOURTEENTH ANNUAL EPA CONFERENCE ON
ANALYSIS OF POLLUTANTS IN THE ENVIRONMENT
Office of Water
Office of Science and Technology
Engineering and Analysis Division
May 8-9, 1991
Norfolk, Virginia
TABLE OF CONTENTS - 1
Page
Wednesday, May 8, 1991
Welcome and Introduction 1
William A. Telliard
Director, Analytical Methods Staff
USEPA, Office of Science and Technology
Keynote Address 2
Tudor Davies
Director, Office of Science and Technology
USEPA
Determination of Total Petroleum Hydrocarbons
by Capillary Gas Chromatography 16
Ileana Rhodes
Shell Development Co.
A Screening Method for Total
Polynuclear Aromatics 76
Richard Beach
Hydro-Systems, Inc.
Fingerprinting of Petroleum Hydrocarbons
in Water 104
Greg Douglas
Battelle Ocean Sciences
A Novel Approach to the Extraction and Analysis of
Chlorophenoxy-Acid Herbicides in Soil and Water 148
Gary Jackson
Support Systems
-------
FOURTEENTH ANNUAL EPA CONFERENCE ON
ANALYSIS OF POLLUTANTS IN THE ENVIRONMENT
TABLE OF CONTENTS - 2
Page
Advanced Techniques for the Measurement of
Acidic Herbicides and Disinfection
By-Products in Aqueous Samples 164
Jimmie Hodgeson
USEPA, EMSL-Ci
Evaluation of the HP5971-A GC/MS Equipped
with a Temperature-Programmable
On-Column Injector 195
Robert Beimer
S-Cubed Laboratories
Comparison of Historical Data for Complex
Samples to Method Criteria for EPA
Methods 1624C and 1624C 221
Margaret St. Germain
MR I
Heated Purge and Trap GC/MS Analysis of
Water Soluble Compounds and Alcohols
by Method 1624C 235
Lance Steere
S-Cubed Laboratories
Improved GC/MS TIC Spectral Assignments through
Mass Chromatograra Peak Centroid Analysis 261
Bruce Colby
Pacific Analytical Laboratory
Results of the ITD Reverse Search
Compound Study 294
William Eckel
Viar and Company
In-Situ Acetylation Analysis of Chlorinated
Phenolics in Pulp and Paper Industry Wastewaters:
Further Investigations and Refinements 315
Larry LaFleur
National Council of the Paper Industry for
Air and Stream Improvement
-------
FOURTEENTH ANNUAL EPA CONFERENCE ON
ANALYSIS OF POLLUTANTS IN THE ENVIRONMENT
TABLE OF CONTENTS - 3
Page
In-Situ Acetylation Analysis of Chlorinated
Phenolics in Pulp and Paper Industry Wastewaters:
Further Investigations and Refinements - Part II 349
Peggy Knight
USEPA, Region X
Thursday, May 9, 1991
The Environmental Monitoring Methods Council
(EMMC) Approach to Methods Integration
and Laboratory Certification 370
Ramona Trovato
Director, Ground Water Protection Division
USEPA, Office of Water
EMMC Methods Integration Objectives and
Superfund/RCRA Methods Integration Successes 378
Joan Fisk
USEPA, Office of Solid Waste and
Emergency Response
The EPA's Analytical Methods for Organic Compounds
in Water and Wastewater: The Next Generation 404
Ron Kites
Indiana University
Automation of the BOD Test for Wastewater 452
W. A. Michalik
Shell Oil Co.
Evaluation of SFE for Extraction of
Environmental Contaminants 467
Merlin K. L. Bicking
Twin City Testing Laboratories
Liquid Chromatography/Mass Spectrometry: An
Emerging Technology for Nonvolatile Compounds 479
William Budde
USEPA, EMSL-Ci
-------
FOURTEENTH ANNUAL EPA CONFERENCE ON
ANALYSIS OF POLLUTANTS IN THE ENVIRONMENT
TABLE OF CONTENTS - 4
Page
A Simple Rule for Judging Compliance
Using Highly Censored Samples 518
Ian Hau
University of Wisconsin-Madison
Assessment of Analytical Variability of
Dioxins and Furans in Environmental Samples
from Bleached Pulp Mills 545
Henry Kahn
USEPA, Office of Science and Technology
Statistical Analysis of Analyte Concentrations
in Municipal Sewage Sludge with Multiple
Detection Limits 574
Charles E. White
USEPA, Office of Science and Technology
Good Automated Laboratory Practices 597
Rick Johnson
USEPA, OIRM
Performance Evaluation Study of Environmental
Analytical Contract Laboratories 648
George Stanko
Shell Development Co.
Interlaboratory Quality Control Approach to
Achieving Pre-specified Data Quality Criteria 717
Fred Haeberer
USEPA, OAMS
Closing Remarks 741
List of Speakers 742
List of Attendees 745
-------
PROCEEDINGS
MR. TELLIARD: Good morning.
Welcome to the 14th Annual Norfolk Analytical Meeting. I'd like
to welcome you. My name is Bill Telliard. I am with EPA and I
am here to help you.
For those people who are new to the meeting, there are
some rules that we abide by. There is no physical abuse to your
neighbor; oral is always acceptable. In so doing, we'd like to
have you during the proceedings, if you have questions of the
speakers, come to the microphones that are around the room and
state your name and your organization and ask your question. If
for some reason you don't do that, these two women over here will
physically abuse you.
The agenda is kind of a full one and we're going to try
to stay on time for a change, which means that at the breaks when
you go out and get your strawberry and coffee, if you would
kindly get your tushes back in here so that we can keep the
papers moving...
I'll have some more announcements later on, but I'd
like to get the show on the road.
Our first speaker this morning is Tudor Davies.
Dr. Davies has been with the agency for about 20 years,
unable to find gainful employment. Tudor has, in another life,
been in charge of the Marine and Oceans Program prior to coming
over the Office of Water just about a month ago or three weeks.
So, he is going to give you an insight into some of the things
that the Office of Water is going to be looking at and talk a
little bit about the reorganization that has just occurred in the
Office of Water.
-------
DR. DAVIES: You're a large audience,
and you sort of disappear off into the distance, so I'm a little
bit intimidated—by the size of the audience as well as by looking
at the program, which I saw last night....
Bill is a little irreverent in most things that he does,
so I was bracing myself for his introduction—which turned out to
not be too bad.
I was thinking last night about what I was going to say
today....Yesterday I came from a meeting with sewage authorities,
with a different sort of orientation—but at least I knew what they
wanted me to talk about. Last night, I was wondering about Bill
Telliard's motives for asking me to come here, and I thought about
the worst things first: having looked at the program and knowing
my analytical chemistry is 20 years behind me or more, I started to
think perhaps Bill was out to intimidate me by the science here, to
show me how much I didn't know. But then after awhile I ascribed
a better motive to him, because he knows that in the Office of
Science and Technology, which is where we both reside now, we do a
lot of regulatory decision-making. He understands some of the
leaps of faith we have to make, safety factors that we impose, cind
I began to think that what he was trying to show me was that—at
least in the analytical chemistry area—the words, precision,
improved quality, etc. , all of these have meaning. . .and that we eire
at least moving progressively to improve the science, in order to
reduce the minimum detection levels and even to automate the
analytical methods and make them cheaper. I think Bill's motive
today was to tell me that, at least in this area, we have good
science and very capable people. I hope that was the message that
he's giving me today.
I've been on the job in the Office of Science and
Technology for a couple of weeks, and I'm going through a whole
series of briefings on what goes on. And so I think the most
appropriate thing to talk about today is to tell you a little bit
about what we do—about why we reorganized and what we reorganized
in the Office of Water at EPA. That's the thing that's on my mind
-------
at the moment and I can be a little coherent about that.
What I'd also like to do is perhaps to tell you some of
the things that we're doing in the Office of Science and Technology
that are relevant to you and discuss some of those issues that
might be of interest to you.
EPA has an administrator who's been on board two years,
and he is looking for a change of agenda within the Agency. He
sees that we've had success—particularly in the Office of Water
over the last 20 years—in applying a technology-based approach to
water-pollution control, and we've done this very much through
command-and-control type operations. And what he's looking for—
and I have some sympathy with this—is that we should be moving
beyond command-and-control and dealing with issues such as cross-
media pollution. Realizing that many programs have now been
delegated from EPA to the states, many programs have been delegated
from the states to the cities, and most pre-treatment and toxic
reduction programs are done by the cities managing industrial
input, we are looking at pollution prevention as much as control,
and we should be taking advantage of any innovation that's out
there. So we should be looking at some degree of flexibility in
our systems.
Further, the administrator talks about taking advantage
of the best science that we have in decision-making. In trying to
move beyond command-and-control, trying to think ecologically as
much as in the past we've focussed on human health, we have to
start talking about geographic targeting, about focussing on
specific environmental areas that are susceptible to management and
control, and perhaps move beyond the basic command-and-control
programs and do some new and innovative things. In that light, we
reorganized the Office of Water.
[First slide. The Office of Water used to be seven
offices. We've now focussed into four areas. (1) We have an
Office of Waste Water Enforcement and Compliance that takes the old
Enforcement, Permitting, and Municipal Control Office and "gloms"
them all together so that the result is a spectrum that they like
-------
4
to say "runs from white hat to black hat." (The old office
director of that area said that perhaps the motto for that office
should be, "I want to go out and crush someone.") (2) Then we have
the Office of Science and Technology, which we'll talk about in a
couple of minutes. (3) We put together some of the programs
developed over that last couple of years on wetlands, the coastal
area, storm water management, and things of that sort into the
Office of Wetlands, Oceans, and Watersheds. That office will have
a watershed function in the future. (4) Finally, we've put the
drinking water program together with the ground water program.
We have three major statutes that we work with in the
Office of Water. (1) We have the Clean Water Act, a major statute
which is currently being reviewed for reauthorization on the Hill,
and there's a great deal of activity within the Agency working with
the Hill on the reauthorization. (2) We have the Safe Drinking
Water Act. You probably saw that we were in the press yesterday on
the lead rule and lead activities in the Agency, particularly for
drinking water. (3) And then, finally, we have the "ocean dumping
act," the Marine Protection, Research, and Sanctuaries Act, which
has been very much narrowed over the last couple of years as we've
gone away from permitting the dumping of industrial waste and
sewage sludge into the ocean, so that now the only thing that we
permit to be disposed of into the ocean is dredged material. We'll
talk about sediments and some other issues in a moment.
[Second slide.] Again, the Office of Science and
Technology...we've divided into three major groups. The first of
those groups is the one that you are probably most familiar with,
having had to suffer through with Telliard for the last 14 years:
the Engineering and Analysis Division, which largely carries the
responsibility for the Effluent Guidelines Program. This is the
program that develops technology-based guidelines for specific
categories of industry and is very much involved with your
analytical program in terms of defining effluent, looking at the
technologies available for managing effluent, and particularly,
looking at the economics of that management and the significance of
-------
regulating those industries.
BAD has perhaps more statutory and judicial deadlines
than I like to think about. In fact, I think the whole office has
about one judicial deadline per person to go around, which means
that sets your agenda pretty closely.
One of the things I'd like to reflect on as we talk about
the Effluent Guidelines Program is that what Bill Reilly has talked
about is perhaps a "kinder and gentler" EPA—and that we're going
to start moving away from command-and-control somewhat. In talking
to EMSL yesterday, I kept on emphasizing that we were looking to
flexibility and change. . .and the guy behind me on the program stood
up and started to talk about all the things that we had done in
enforcement, how we had fined this city $3 million and hit this
industry for something else...and someone in the audience stood up
and said, "You know, you have a very mixed message that you're
giving here. On the one hand, you're talking about flexibility,
and on the other hand, the Agency still has this command-and-
control mentality."
I think we have a real problem with communicating in the
Agency, and there's a little story I want to tell you, because I
think we all have problems with communicating. I think you have a
problem in dealing with biologists and regulators in terms of
conveying chemistry to them and the depth and understanding of your
techniques.
We often talk and hear very different things. I'd like
to tell you a story that emphasizes that. It's a little bit off
the point, but it adds a little light relief to the proceedings,
and there is no religious connotation to this story at all. In
fact, it was told by my minister about three or four weeks ago, and
I thought it was very appropriate to the subject of communication.
In the Middle Ages, there was a Pope in Rome who was
very, very flexible in terms of his beliefs, and the rest of the
clergy in Rome was very, very conservative and wanted to get the
Jews out of Rome. And the clergy pushed on this Pope, and they
said, "You know, you've got to get them out."
-------
And the Pope said, "Well, let's do it this way. We'll
have a single combat between the Jews and the Catholics here to
decide who will leave, and I will be part of the single combat, I
will be the Catholic's champion, and they can choose their
champion. To make the contest difficult, it will be a debate—and
because I have such an enormous reputation as a debater, we'll do
it in mime."
The Jewish community was very upset about this, but of
course this was their chance to stay in Rome. And so they asked
the most distinguished Rabbis if they would debate the Pope, and
the Rabbis didn't want to do it. It eventually came down to a guy
low in the structure who was a bit of the clown. He had to take on
the Pope.
On the day of the debate, the Pope showed up in all of
his finery with the clergy. And the little Jewish guy showed up,
and he wasn't very well dressed, he looked sort of out of place
among these people.
The Pope began in mime...I hope you can see me. He sort
of took his arm...and just made a broad stroke, like this...and
that was the opening shot of the debate.
His opponent looked at him for a while and did that. . .put
his finger out and sort of shoved it down toward the ground, like
that.
The Pope thought for a while and he made the next
step...he moved three fingers toward his opponent, like that.
And his opponent stopped and thought about it and he just
showed him one finger back, like that.
And then the Pope thought for a little while longer and
then turned aside and pulled out the bread and the wine and showed
it to his opponent. And his opponent looked at him for a while and
took an apple out of his pocket and bit into the apple.
And the Pope said, "Oh, that's fantastic!" He said, "You
have won, you can stay in Rome."
And so everybody left and all of the Cardinals, etc.,
said to the Pope, "Well, what happened? Interpret the debate for
-------
us."
So the Pope said, "Well, I started off by saying, 'God
made the heaven and the earth.' He came back and he did that,
which meant to me that He also made man on the earth. And then I
said, 'The Trinity,' and he again came back to me and said, 'He
made man in his image. ' And then I took out the sacrament and
showed it to him, and he turned back to me and showed me the apple
and he said, 'Man is human. He was tempted.' You know, that was
wonderful. He had countered everything I said, so the Jews can
stay."
So there was enormous celebration in the Jewish
community, as you'd expect, and finally they got their champion
aside and said, "Well, what happened?"
So he said, "Well, this is the way the debate went. He
did this, which meant to me, 'All of you get out of town by
nightfall.' And I said back to him, 'Not one of us is leaving.'
Then he countered again and he said, 'You have three hours to get
out of town.' And again I said, 'Not one of us is leaving.' So he
said I had convinced him at that point. And so what we did then
was he took out his lunch and I took out my lunch and we sat down
and had lunch."
Think about how many conversations you've had that sort
of worked like that.
Anyway, let me tell you a little bit about the rest of
the office and the great difficulty we_ have in communicating some
of the decisions that we make on regulatory matters.
We decided, when we put this office together, that we
would try to pull some of the scientific people out of some of the
other offices within Water, to try to consolidate and get a
critical mass of scientific issues and people that we were dealing
with. We thought that we would do that by having a risk assessment
group outside of the Effluent Guidelines Program, the risk
assessment group that you see in the middle there, and a risk
management group.
The risk assessment group is charged with developing
-------
8
health criteria for both drinking water and surface water
regulation. If you'll look at that group, you'll see that we first
analyze surface water (which we do very well) and then worry about
the effect of the contaminants we find in the surface water on
human health and the environment. Most of the work that we do in
that area is based on translation from animal studies and aquatic
toxicity studies that we do—so our ability to precisely measure
chemicals in water and other media is really not matched by the
science we have in looking at human and ecological risk assessment.
We impose many safety factors, and we're often translating
information from things like rat studies, trying to understand the
impact of toxicity and perhaps cancer on human beings. So think
about that as you consider the human health criteria that we've
developed.
Often, one of the things that we're concerned about when
we're setting drinking water standards is the reproducibility of
the analysis method for water—and you're very helpful to us there.
But again, there are those uncertainties in the risk assessment
part of the program.
When we come to the ecologic criteria, we're interested
in water quality. And we have a set of toxicity measurements emd
bio-concentration measurements that we make, particularly on a
various file of animals, to come up with risk levels that are then
incorporated into standards to protect the safety of surface water.
One of the areas that we're getting into is sediment
criteria. When we look around the country, we see many, many areas
in which the historical discharges, and perhaps the current
discharges, are contaminating surface sediments. And that
contamination doesn't stay in one place but recycles into the water
column. There is exchange between the sediment medium and the
biota so that we have a great deal of concern, particularly for
those persistent toxic chemicals which bio-accumulate in humans.
I'll talk in a moment about the Great Lakes, which is the site
perhaps the furthest along in determining where those areas of
sediment contamination are and in trying to deal with strategies to
-------
remediate the problem.
The other part of the risk assessment area is the sewage
sludge program. We have proposed a multi-media regulation for
sewage sludge. It deals with placement of the material in certain
environments, incineration of sludge, land filling, land spreading,
etc. I think we've broken some new ground there in risk
assessment. I know that Bill was very involved with the sewage
sludge studies in terms of characterizing the toxicants in sewage
sludge around the country...and we've had to work with many, many
people across the Agency and deal with multiple statutes, as well
as the Clean Water Act, in developing this regulation for sewage
sludge, which will be proposed (we hope) next year. It will have
multiple monitoring and assessment requirements what I'm sure some
of you will be involved with. So we're dealing with human health
in considering water, drinking water, and sewage sludge, on the
assessment side.
On the risk management side, we're dealing with setting
standards, particularly for surface water. We have a proposal that
will be coming out shortly that will require all states to have
standards for the materials that we've developed criteria for, and
we see a further proposal coming out of the Great Lakes area, a
proposal which is supported by all of the Great Lakes states, to
set further, more specific water quality standards to protect the
Great Lakes environment. The Great Lakes states are perhaps at the
cutting edge in developing new techniques in that area. Those of
you that are interested in development and implementation of water
quality standards should watch the Great Lakes program to see what
they are doing.
We are trying to develop standards for surface water,
we're trying to develop standards for ecosystems, and we're
developing standards for things like wetlands and sediments.
It's particularly interesting, I think, that we're
developing standards and criteria for sediments, an area where we
have a difficult medium to deal with and a whole series of
different purposes at work.
-------
10
We're looking, one, at preventing the toxic material from
getting out into the environment in a concentration that can cause
sediment contamination, and so we need criteria for our control
programs for sources, particularly point sources. We have so much
contamination out in the environment in Superfund sites and in
contaminated sediments that we need criteria that will allow us to
define the clean-up level. That's another purpose.
Then we have the problem of dealing with material that we
want to move for navigation, material that is in the navigation
channels and is contaminated: the dredge material program. In
this program, we're trying to deal with sediments as a total
medium. We're trying to deal with the whole toxicity, the whole
potential bio-accumulation of toxics from those sediments, so we
have a series of methods. And there is some controversy within the
Agency about the methods that we use.
For non-polar organic chemicals, we're coming out with a
predictive method for sediment criteria that's dependent upon
partition coefficients and upon the organic carbon content of the
sediment. We would look at these parameters and then determine
concentrations for the particular chemicals. We're also looking at
developing sediment criteria using a calculation based upon acid
volatile sulfides within the sediment.
There's another group that says that what we should be
doing is biological effects testing on those sediments, to see
whether there is in fact a direct effect on benthic animals and,
further, whether there is the possibility of any bio-concentration
into the environment from those sediments...and they say that
should be the regulatory framework.
Then there are a whole range of intermediate positions.
We are concerned about whether this medium is binding chemicals,, so
that the direct chemical analysis doesn't reflect the biological
availability of the chemicals.
This is an interesting debate, and we're having a great
deal of interaction with the scientific community on this subject.
It will be an interesting regulatory issue to follow, and I'm sure
-------
11
that you will be called upon to develop very specific methods in
this area.
Associated with the sediment issue is an issue related
to fish advisories. Most of the contaminated sediments contain
these bio-cumulative persistent chemicals. The thing that we're
particularly concerned about is that these chemicals bio-accumulate
and get into the fish and shellfish, and then we're called upon to
issue fish advisories to the states to protect human health. There
isn't a good procedure out there at the moment. There are FDA
measures for interstate transfer of shellfish that are not very
good, and there are methods that EPA has used that are again not
very good.
We've recently had a symposium where the American
Fisheries Society and the federal agencies got together with the
states. Now we have a strategy that's working to develop better
risk assessment and to look at fish consumption rates, and that's
a critical component of any risk assessment. There is debate over
whether subsistence fishermen eat a great deal more fish than
people who are just recreational fishermen and what the normal
population eats. Finally, we have to decide what level of
protection to establish in setting a fish advisory.
We do not have a consistent methodology for looking at
fish. Some people look at whole fish and some look at the edible
portion, and the risk assessment varies accordingly. We also need
a good standardized QA/QC program for sampling and analyzing fish
tissue. We need to have a clearing house, so that all the states
are dealing with these risk issues in a compatible way.
And then we come back to communication. . . .We need a good
risk communication strategy so that we don't frighten people, but
we protect public health.
One of the final things that we're doing in this new
office is trying to develop better load allocation methodologies.
We can measure it. We can perhaps understand the environmental
effects and the human health effects of chemicals in the
environment. But what we have to do is develop a regulatory scheme
-------
12
that allocates loads to specific sources, which in the future may
include non-point as well as point sources. However, in most of
the country, we are still doing dilution calculations, which is a
pretty unsophisticated way to deal with controlling the release of
contaminants into the environment. So that's an area that I think
our office will focus upon.
I think that at this point I'll draw your attention
finally to the Great Lakes, where we'll be dealing with these
issues in a much more pro-active way. The Great Lakes community,
along with Canada, is very concerned about the quality of the
environment. We've had much success in dealing with nutrient
issues. Bob Booth and I were talking over coffee this morning
about activities that we'd had going on the Great Lakes 20 yecirs
ago during the International Field Year, and now we're back ageiin
to a very high focus on the Great Lakes. The eutrophication
problems are largely solved, but the toxics problems in the
environment are certainly not. We have significant inputs from the
atmosphere. We have contamination in sediments and we have the
normal land-based sources.
We have to develop a better standard methodology. We
have to build sediment remediation methods that are economically
acceptable, and those remediation methods are very slow in coming.
But focus on the Great Lakes—I think that's where we'll be making
a lot of new progress. And that, I think, is consistent with
Reilly's intention of trying to work with local communities, work
with the states, trying to develop political will for environmental
control, and focusing on a geographic area that people can
associate with. It's sort of difficult for the public to talk
about technology-based standards for industry and for
municipalities. I think people are more interested in their water
body and protecting that in any way they can.
So we're going to think multi-media. Hopefully, we're
going to use better science. I hope we can count on you to help
with that, and I think we're going to try to be a kinder, gentler
EPA, if that's possible. Thank you.
-------
13
QUESTION AND ANSWER SESSION
MR. TELLIARD: Are there any
questions for Tudor? Silence.
In the back of the room during the break you'll find an
organization chart for the Office of Water and a phone listing so
you can call your favorite Office of Water person.
-------
14
*
Q
i
i —
<
z
o
LU
[*'
>
O
o
o
LU
1 —
Q
^~^7
<;
O
z
UJ
O
CO
1 1
o
LU
O
u_
LL
'o
CJ
^"^1
fl
(U
O
rj
•i— 1
O
CO
o
O
1 1
tT"j H"^
CO CO
C/)
P S
*r^ tUQ
^ \ ^^<
1
Si"u
tj)
O PH
O ^
fll
^ ^
CO
p , CQ
•a S
|||
ca'S.u
CH ^ C
^ 1
C
_o
M "rt S
"g o|j
^
QJ
^0
1
1
1
"rt C
(-< O
a
tf) PQ
-------
14-A
o
*
o
\ —
I
M
~7
i i O ** •*-»(,_,
LJ — «. 4-> JJ rt *3
t • 113 fH „. C U CD
| i 13 VH TJ C -£3 +j
^i COpcU ^rt'Blgcn
u Sag §| 1*1
«|S ^fg |||
>s c3 sa^ §BP^
b*5 co 3 g !:£)«
I
J OHO^WCQ,§K
1 >H
-t->
O -M ^
cu ^
.y ^ g
^H T"} "a
0 §•§
2Q
o
^H tuO^^ ^>
° S r| "3
1 5 2 K 5
^^ *t3 SQ
|c S£ §•§
^2 § § t«e3
^ tj <" « C T3
•O J5 o -^ -g C
§-g &§ -25
O ,V C "-1 TH W
K P* M Pi
0 w P
1 1 1
1 1
CO* -- ^
-a ^
Co r^
^^H »•« *^™^
-M w co
^ § ^
t^^ O
8° -a
O j
v y
S~ ~~* ° 73 >:
^ g Q c?
K/ i; !rl x3
o -S P^
1^ s SI
OS H *— *
CO _, "H
s
t
o
ex
3 4J
CO tO
•£H O
S c
A r~(
V ii
H
5 So,.
<• h &
•§ « s
« o ^
S5 73 "«3 ^
c
§ Q 0
CO "H ">
•a S S
B
S1
-------
15
MR. TELLIARD: Our next session is
going to deal with hydrocarbon analysis and our first speaker is
from Shell.
Ileana Rhodes and I met two children ago, she said, I
had nothing to do with that. At that time we were looking into
trying to measure some brines out of oil wells for laughs and she
in her other job was to keep George Stanko in line. She's been
fairly successful in that.
Ileana, do you want to come up?
-------
16
DETERMINATION OF TOTAL PETROLEUM HYDROCARBONS
BY CAPILLARY GAS CHROMATOGRAPHY
Gasoline to Diesel Range Total Petroleum Hydrocarbons (TPH) and
Approximate Boiling Point Distribution
I. A. L. Rhodes, R. Z. Olvera, J. A. Leon and E. M. Hinojosa
SHELL DEVELOPMENT COMPANY
HOUSTON, TEXAS
ABSTRACT
Assessment and remediation of soil contamination by petroleum products
requires the identification of the type and extent of contamination.
There are several analytical procedures that are used to obtain this kind
of information. The term "total petroleum hydrocarbons" (TPH) is used to
describe extent of contamination but the actual value determined is
method dependent and, thus, must be defined by the method used. All
procedures have limitations and care must be exercised in interpretation
of data. None of the methods available provide information on boiling
point distribution of the contaminants and limited information on product
type.
A procedure was developed for the determination of product type, gasoline
to diesel range TPH, and its boiling point distribution in soil. This
procedure involves extraction of the soil followed by analysis of the
extract using gas chromatography with flame ionization detection.
-------
17
DETERMINATION OF TOTAL PETROLEUM HYDROCARBONS
BY CAPILLARY GAS CHROMATOGRAPHY
Gasoline to Diesel Range Total Petroleum Hydrocarbons (TPH) and
Approximate Boiling Point Distribution
Ileana A. L. Rhodes, Ramon Z. Olvera,
John A. Leon and Emiliano M. Hinojosa
Shell Development Company
Environmental Analysis Department
Westhollow Research Center
Houston, Texas
INTRODUCTION
Assessment and remediation of soil contamination by petroleum products
requires the identification of the type and the extent of contamination.
There are several analytical procedures that are used to obtain this kind
of information. The term "total petroleum hydrocarbons" (TPH) is used to
describe the extent of contamination but the actual value determined is
method dependent and thus must be defined by the method used. One of the
most commonly used procedures is modified EPA Method 418.1 which is an
indicator method that provides information on Freon extractable petroleum
hydrocarbons measured using infrared spectrophotometry1. Other methods
are based on extraction of the soil followed by gas chromatographic
analysis of the soil extract using direct injection, headspace or purge
and trap analyses with determination of selected components or sums of
components2"5. All of these procedures have different advantages and
limitations,thus care must be exercised in interpretation of data.
Most of the chromatographic procedures require two types of analysis.
Volatiles (ie. gasoline range) are determined by extraction of the
samples followed by analysis of the extract using purge and trap
techniques. Semivolatiles (heavier than gasoline range) are determined
using a different extraction procedure followed by concentration of the
extract and then analysis of the concentrate using a direct injection
approach. These methods are necessary for determination of low levels of
TPH contamination (<100 ppm). However, it is often not necessary to
determine TPH concentrations below 100 ppm. Cleanup standards in about
half the states are at 100 ppm TPH and above6. Results from several
analyses must be combined to obtain limited information on product type.
The concentration steps are time consuming and are often not necessary
since the concentrate may need to be diluted for analysis.
The method described in this paper was developed to meet the following
requirements: 1) identify the type of contamination (gasoline range,
diesel range, mixtures, crudes, etc.), 2) quantitate TPH from gasoline to
-------
18
diesel range, 3) quantitate selected target analytes, and 4) determine
approximate boiling point distribution of the material present in the
soil to assist in selection of appropriate remediation technology. It was
also desirable to develop a method with minimal sample handling and simple
calibration/instrumentation techniques.
None of the methods currently available can provide information that
satisfies all of the stated goals. EPA Method 418.1 cannot provide
information on the type of hydrocarbon contamination, light boiling point
components are easily lost during sample handling and calibration is only
accurate if samples contains about 30% aromatics1. Headspace procedures
are very sensitive but are biased towards the light ends. To fulfill the
stated requirements, a method was developed using methylene chloride or
methanol for extraction of the soil followed by analysis of the extracts
using gas chromatography with flame ionization detection (GC/FID). These
solvents were chosen because they are relatively low boiling, and are
commonly used in EPA methods and other proposed methods5.
The extracts are analyzed using capillary gas chromatography. Separation
is done using a high resolution fused silica capillary column with bonded
methyl silicone phase. This is a non-polar stationary phase in which
separation essentially takes place by differences in boiling points of
the components in a mixture. The areas under all peaks that elute after
the extraction solvent are summed for both samples and calibration
standards. Calibration solutions can be prepared of either gasoline or
diesel range TPH in the same solvent as the extraction solvent.
Alternatively, calibration can be done with a mixture of selected
gasoline and/or diesel components. Individual target analytes such as
benzene, toluene, and xylenes (BTX) can be identified and quantitated if
desired.
This method takes advantage of the fact that the flame ionization
detector response is essentially the same for all hydrocarbons as
indicated in Table 1 where the response factors of an abbreviated list of
hydrocarbons present in all commercial gasolines are tabulated and
normalized with respect to n-heptane. Only methyl-t-butyl ether (MTBE) is
significantly different. For heavier material, similar data is available
in the literature (C14-C32 alkanes and C10-C22 polycyclic aromatic
hydrocarbons)7.
Approximate boiling point distribution is obtained by normalization of
the cumulative areas of peaks between retention times of elution of
compounds of known boiling points. The normal hydrocarbon of a homologous
series has the highest boiling point for its carbon number and thus
elutes last. The approximate boiling point distribution plots provide
information on weathering of the material.
The method described in this paper has been applied for gasoline to
diesel range (gasoline range only, diesel range only as well as mixtures)
TPH concentrations from 50 to 10,000 ppm. Soil moisture (<10%) does not
appear to have a significant effect on extraction efficiency. For samples
with relatively high moisture, it is recommended that the sample be mixed
with sodium sulfate prior to extraction if methylene chloride is used for
extraction. Drilling muds fall in this category.
-------
19
EXPERIMENTAL PROCEDURE
Sample preparation
The method involves weighing 10-20 g of sample in a vial with a Teflon
lined cap. Sodium sulfate may be added for samples with moisture levels
above 10%. The extraction solvent, 10-20 ml of methylene chloride or
methanol, is added to the vial. Solvent purity is essential (99+?
purity). A series of extraction steps involve mixing for 1 minute using a
vortex mixer and shaking with a horizontal shaker or a wrist action
shaker for at least 1-4 hours. The samples can be centrifuged if
necessary. The extracts can then be directly transferred to autosampler
vials and analyzed by gas chromatography using the parameters listed in
Table 2. Typical chromatograms are shown in Figures 1-4. Figure 5 shows
the chromatogram of a synthetic standard of selected gasoline to diesel
range components.
Preparation of Standards
Calibration standards - Standards are prepared using gasoline and/or
diesel of any grade or source or a blend of selected hydrocarbons in the
concentration range expected in the sample. These standards are prepared
by weighing the required amount of gasoline and/or diesel or of selected
hydrocarbons and diluting by volume with the extraction solvent. Typical
concentration ranges are 50 to 10,000 iig/mL. Calibration curves or
average response factors can be used. It is desirable to have standards
in a similar concentration range as the samples. Typical calibration
plots using a regular gasoline, diesel, 1:1 gasoline/diesel and a
synthetic standard of selected gasoline/diesel range components are shown
in Figure 6-9.
Boiling point distribution reference standard - A solution
containing approximately 200 ppm each of n-hexane through eicosane and
pentacosane is used for determination of the retention times
corresponding to the different boiling point fractions. Table 3 lists the
boiling points of each of these n-alkanes and the retention times
observed when the instrumental parameters specified in Table 2 are used.
The boiling point distribution can be used to assess whether or not a
site is amenable to soil venting. For that reason, detailed boiling point
distribution is usually done only for the gasoline range (up to C12) and
only a few markers are used for heavier materials. The heavy materials
are usually described by the carbon number range rather than by boiling
point range.
Instrumental procedure
The instrumental parameters used for the analysis of the soil extracts
and standards are listed in Table 2. Any data system capable of grouping
and summing selected peak areas can be used. A VG Multichrom data system
was used throughout this study for collection of data and computation of
results.
-------
20
The method, as described in Table 2, is limited to determination of
hydrocarbons up to "C25. This cutoff was chosen for several practical
reasons. Column phase bleed at oven temperatures above 280°C results in a
rise in baseline. This rise in baseline makes integration of unresolved
peaks (such as in diesel range material) quite difficult and blank
baseline subtraction is often necessary. This added computation is not
always appropriate primarily due to changes in the electronic zero at the
beginning of each run. A final temperature of 280°C minimizes baseline
rise while still allowing elution of components up to and over C25.
Another reason for selecting a cutoff of C25 is that alkanes above C25
are not sufficiently soluble in methylene chloride or methanol and thus
would not be effectively extracted from the soil. Carbon disulfide can be
used to extract heavier hydrocarbons.
Determination of TPH
The areas of all peaks detected that elute after the extraction solvent
peak and up to the retention time where pentacosane elutes are summed.
The report includes both total area sums of all peaks detected up to and
including pentacosane and area sums of peaks eluting between the n-alkane
markers for both standards and samples. The total area sums are used for
determination of the gasoline to diesel range TPH in samples. The
information on area sums between selected markers is used to generate an
approximate boiling point distribution plot for a normalized sample.
Calibration can be done using gasoline and/or diesel or synthetic
mixtures of selected gasoline to diesel components. The total area sum
for samples to be quantitated must be within the calibration range.
It is often desirable to determine what portion of the TPH present in
samples is due to gasoline range and what portion is due to a heavier
product such as diesel range material. Both of these products overlap to
some extent. A recommended cutoff is CIO. Generally, gasoline is 5-15%
above CIO and diesel can be ~20% below CIO. This is obviously a
compromise.
Calculation of Approximate Boiling Point Distribution
The approximate boiling point distribution is calculated by
normalization of sums of peak areas of portions of the chromatograms
eluting between preselected retention times as indicated in Table 3.
These retention times correspond to known boiling points selected as
references. The chromatographic column used in this method is essentially
a boiling point non-polar column and compound separation is achieved by
boiling point differences. A homologous series of n-alkanes is used as
approximate boiling point references. The cumulative boiling point
distribution is graphically displayed by plotting the cumulative area
percents versus boiling points of the n-alkanes. The plots are similar to
those obtained from simulated distillation or true boiling point gas
chromatographic analyses. Figure 10 includes several approximate boiling
point distribution plots.
-------
21
RESULTS AND DISCUSSION
A single analysis can be used to determine product type, TPH, target
analytes and approximate boiling point distribution.
Product Type Identification
Product type can simply be determined by visual inspection of the
chromatograms. The "fingerprints" of gasoline, diesel and mixtures of
these two petroleum hydrocarbons ranges are shown in Figures 1-4. The
chromatogram can get more complicated if crude oil, jet range material or
other refined products are also present. Nevertheless, it may still be
possible to determine that the contamination is due to some sort of fuel
oil. Industrial solvents can interfere in the analysis, however, the
chromatographic fingerprints would be noticeably different.
Determination of TPH
Preliminary experiments indicated that methanol is somewhat less
efficient in extraction of diesel range material than methylene chloride.
Thus all subsequent spiking experiments were done using methylene
chloride as extraction solvent.
The method was tested by spiking known amounts of regular gasoline,
diesel oil and mixtures of gasoline and diesel oil in soil. Typical
chromatograms are shown in Figures 1-3. As previously stated and shown in
Table 1, the method takes advantage of the fact that the response of
flame ionization detector is essentially the same for all hydocarbons (on
a weight basis) and based primarily on effective carbon number. It is
therefore not essential that calibration be performed using material
similar to the material in the samples. For example, any gasoline,
diesel, synthetic mixture or single hydrocarbon can be used for
calibration and calculation of TPH in samples with any type of petroleum
hydrocarbon contamination. This is essentially true. However, because
products such as gasoline or diesel are composed of more than 300
individual components, at low concentration of total product, many of the
individual components are simply too small to be detected and cannot
contribute to the total signal detected and thus linearity falls off.
Conversely, when synthetic standards are used, typically no more than
10-20 components are used and thus the TPH is distributed among a few
peaks which can be all detected for all concentrations of the standards
above the stated practical quantitation limits. The use of synthetic
standards always results in underestimation of the TPH present in the
samples.
In addition, by using extraction solvents that are in the gasoline range
(
-------
22
Figures 6-9 show the calibration plots using different types of
materials. These calibrations were used to quantitate results tabulated
in in Table 4 which summarizes the results obtained for the soil spiking
studies. Soils were spiked with gasoline, gasoline plus diesel mixtures
and diesel range materials. Quantitation was done using calibration
standards of gasoline range only, gasoline plus diesel range only, diesel
range only and synthetic standards. The extracts were analyzed each using
two different instruments. Evaluation of the results compiled in Table 4
indicates that it is acceptable to use any type of standard. The overall
accuracy average is "90% and the overall accuracy average percent limits
are 70 - 110%. As expected, the least accurate results for the lower
concentrations were obtained when synthetic mixtures were used for
calibration where the slope is the highest.
Determination of Selected Target Analytes
Selected components are indicated in Figures 1-4 and can be measured
individually if desired. The most practical approach is to simply use the
same calibration as that used for total TPH using the area of the target
analyte in a given sample. In this study, target analytes were not
determined since spiked samples were used.
Approximate Boiling Point Distribution
Approximate boiling point distribution similar to those obtained from
simulated distillation or true boiling point types of analysis can be
obtained using this procedure which uses the retention times of n-alkanes
as markers for determination of boiling point distribution of the
contamination. Figure 10 shows the cumulative boiling point distribution
of super and regular grades fresh gasolines. The approximate boiling
point distributions of regular gasoline, diesel and a mixture of gasoline
and diesel spiked onto soils are also included. This type of information
can be used to determine not only product type(s) but also to
characterize the state of the material, such as severity of weathering
and relative concentrations of mixed range materials. A soil sample from
a service station was analyzed using this method and the boiling point
distribution of the contamination is also included in Figure 10. It is
evident that the source of contamination in this soil is gasoline range
material but it is extremely weathered since there are essentially no
components that boil below 125°C.
Typically, a detailed boiling point distribution is needed only for the
gasoline range so as to obtain information in assisting selection of
suitable remediation technology (for example, soil venting). The driving
force for development of this method was the generation of boiling point
distribution information to assist in determination of what service
stations with soil contaminations can be remediated effectively using
soil venting. Beyond the gasoline range (above C12), it is not as
important to detail the boiling range of the material but it is more
practical to simply define the carbon number range to categorize the type
of material (for example, jet, diesel, motor oil, etc.)
-------
23
The method described in this paper only allows the characterization of
material beyond C6 since any material lighter that C6 is obscured by the
extraction solvent (methylene chloride or methanol). As much as 25% of a
fresh gasoline can be below C6. When a sample is known to contain only
gasoline range material, an alternate method is recommended to properly
characterize the approximate boiling point distribution of gasoline range
contamination. This alternate method involves extraction of the soil with
tetradecane (elutes beyond the gasoline range) which allows the
estimation of boiling point distribution from Cl to C128.
METHOD LIMITATIONS
As with any gas chromatographic procedure using non-selective flame
ionization detection, interferences are possible from coelution of
gasoline components with other soil contaminants of other sources.
Potentially, any compound with similar boiling point and polarity as the
hydrocarbons of gasoline to diesel range may have retention times within
the range of interest and may result in overestimation of the TPH
concentration. For example, volatile industrial solvents, cleaners, and
naturally occurring compounds not of petroleum origin may interfere with
this analysis. It is often possible to assess the presence of solvents
and cleaners since the characteristic fingerprint of gasoline, kerosene,
diesel and heavier materials is altered.
Decisions must be made by the analyst in determination of cutoff points
for quantitation of different product ranges when contamination is caused
by a combination of sources. For example, if soils are contaminated with
gasoline range and diesel range materials, there is an area of overlap
where certain components are common to both types of petroleum fractions.
A compromise cutoff for mixtures of gasoline with diesel fuel range
material is CIO. There is no appropriate cutoff for a mixture of jet fuel
or kerosene and diesel fuel since there is a great deal of overlap. Crude
oil contamination also contains a wide range of materials. In cases where
mixed products are present, it is perhaps best not to quantitate how much
is due to what type of product but to simply quantitate TPH.
In order to minimize quantitation problems due to column bleed, the
method is best suited for analysis of materials up to diesel range.
Heavier materials can be detected but not quantitated effectively.
SUMMARY AND RECOMMENDATIONS
A high resolution gas chromatographic method was developed for the
determination of gasoline to diesel range total petroleum hydrocarbons
which involves extraction of the soil with methylene chloride or
methanol. The method can be used for obtaining information on the boiling
point distribution of the contamination and, if so required, on
individual components of interest. The method has been applied to the
analyses of spiked soil samples, soils from service stations and
marketing distribution terminals, and drilling muds.
-------
24
A limitation of the method is that the practical quantitation limit is
50-100 ppm. However, these levels are compatible with cleanup standards
for many states6. A practical approach to the determination of TPH and
characterization of the material present is to follow the procedure
described in this paper and outlined in Figure 11 where a flow chart is
outlined and the decisions are based on data quality objectives. The
approach proposed involves extraction and analysis of soil samples for
characterization of TPH as described here. If TPH is above 50-100 ppm
(ie. if peaks are detected with appropriate fingerprints), then
information on product type, TPH, selected target analytes and
approximate boiling point distribution can be obtained from a single
simple extraction and analysis. If no peaks are detected, and if
information is needed below 50-100 ppm TPH, then the samples can be
analyzed using other more sensitive methods.
-------
25
REFERENCES
1- EPA Method 418.1, "Methods for the Chemical Analysis of Water and
Wastes," EPA-600/4-79-020, U.S. Environmental Protection Agency, 1979.
2- Calabrese, E. J., and Kostecki, P. T., Petroleum Contaminated Soils,
Vol.2, "Remediation Techniques, Environmental Fate, Risk Assessment,
Analytical Methodologies," Chapter III, Analytical and Environmental
Fate, Lewis Publishers, Inc., Chelsea, Michigan, 1989.
3- Roe, V. D., Lacy, M. J., Stuart, J. D., and Robbins, G.A., "Manual
Headspace Method to Analyze for the Volatile Aromatics of Gasoline in
Groundwater and Soil Samples," Analytical Chemistry, 61., 2584-2585,
1989.
4- Leaking Underground Fuel Tanks Field Manual: Guidelines for Site
Assessment, Cleanup, and Underground Storage Tank Closure, State of
California, May 1988.
5- Parr, J. L., Walter, G., and Hoffman M., "Sampling and Analysis of
Soils for Gasoline Range Organics," Presented at the West Coast
Conference on Hydrocarbon Contaminated Soils and Groundwaters, Newport
Beach, California, February 21, 1990.
6- Bell, C, E., "State-by-State Summary of Cleanup Standards: A Review of
the New Wave of State-Level Rules," Soils, Nov-Dec, 1991.
7- Tong, H. Y., Karasek, F. W., "Flame lonization Detector Response
Factors for Compounds Classes in Quantitative Analysis of Complex Organic
Mixtures," Analytical Chemistry, 56, 2124-2128, 1984.
8- Rhodes, Ileana A. L., Olvera, Ramon Z., Leon, John A., "Determination
of Gasoline Range Total Petroleum Hydrocarbons (TPH) and Approximate
Boiling Point Distribution in Soil By Gas Chromatography," poster
presentation at Fifth Annual Conference Hydrocarbon Contaminated Soils,
Amherst, MA, September 1990.
-------
26
QUESTION AND ANSWER SESSION
MR. CALLAMORE: My name is Martin
Callamore, City of Tacoma. We have done some extraction work
doing TPHs in sewage sludge and what we found is that the FID has
a big humpogram that makes quantification very difficult. We'll
get the classic distribution of your diesel, but it's riding on
top of a very large hump that will vary. Apparently, the
extraction is taking out some things other than TPH also.
Doing the silica gel cleanup to remove the carboxylic
acids hasn't really helped.
MS. RHODES: I can't hear you very
well...sorry.
MR. TELLIARD: That's better.
MR. CALLAMORE: Okay, sewage
sludge TPHs...cleaning up with silica gel to remove the
carboxylic acids doesn't seem to remove all the biogenic material
so we have a real severe problem with interference in the
chromatogram trying to determine TPH. The traditional, normal
hydrocarbon distribution rides on top of a very large hump which
makes quantification very difficult. I was wondering if you had
ever run into that sort of thing before.
MS. RHODES: I couldn't hear your
whole question. I don't know if this is working or not.
I couldn't hear your whole question, but essentially
what you've got is you've got all of that biological material to
deal with as well and, I don't know, you might try some GPC-type
techniques to try to clean it up a little bit better. We usually
don't deal with that kind of material, but I know what you're
doing because we have experienced that in some other cases...not
for TPH measurements. You have to go through a whole lot of
cleaning to get rid of the bio mass. Sorry.
MR. PRONGER: Greg Pronger,
National Environmental Testing. Have you experimented with any
solvents that would move your solvent front out in front of the
pattern of gasolines such as maybe carbon disulfide or any of those?
-------
27
MS. RHODES: I tried carbon
disulfide. I work with carbon disulfide quite a bit and carbon
disulfide gives you a peak right on the same spot where methylene
chloride does. Carbon disulfide, however, is a better solvent
for heavier hydrocarbons. For example, one of the other reasons
why I stopped at C25 is because you can't dissolve C30 in
methylene chloride. CS2will work just as well, but it also would
interfere in the front end of the chromatogram.
MR. PRONGER: Are you saying it's
got an interferent? Is it the grade of carbon disulfide or it...
MS. RHODES: No, it gives you
response right on the spot. The FID will respond to CS2when it's
in a solvent amount right on the same spot with methylene
chloride within a few...like a half a minute or so. It will come
out on the same spot. But it's a good solvent as well; it can be
used. It just stinks so much.
MR. PRONGER: There are some
disadvantages of the solvent.
MR. TELLIARD: Thank you.
-------
TABLE 1: RELATIVE RESPONSE FACTORS OF SELECTED
GASOLINE RANGE COMPONENTS USING GC-FID.
(NORMALIZED WITH RESPECT TO n-HEPTANE)
Methyl-t-butyl ether 0.70
n-Butane 1.00
i-Pentane 1.00
2-Methylbutene-1 0.96
n-Pentane 1.00
Cyclopentane 0.96
2-Methylpentane 1.00
n-Hexane 1.00
Methylcyclopentane 0.97
2,4-Dimethylpentane 1.00
Benzene 0.92
Cyclohexane 0.99
Cyclohexene 0.98
2-Methylhexane 1.00
3-Methylhexane 1.00
t-1,3-DimethylcycIopentane 1.00
t-1,2-Dimethylcyclopentane 1.00
3-Ethylpentane 1.00
2,2,4-TrimethylDentane 1.00
n-Heptane 1.00
Methylcyclohexane 0.98
Ethylcyclopentane 0.98
2,4-Dimethylhexane 1.00
2,3,4-Trimethylpentane 1.00
Toluene 0.93
2-Methylheptane 1.00
3-Methylheptane 1.00
t-1,3- & c- 1,4-Dimethylcyclohexane 0.99
n-Octane 1.00
n-Propylcyclopentane 0.99
Ethylbenzene 0.96
m-Xylene 0.96
p-Xylene 0.96
o-Xylene + 3-Methyloctane 0.98
n-Nonane 1.00
i-Propylbenzene 0.98
2,6-Dimethyloctane + n-Propylbenzene 0.98
1-Methyl-4-ethylbenzene 0.98
1,3,5 -Trimethylbenzene 0.98
1 -Methyl-2-ethylbenzene 0.98
4-Methylnonane 1.00
t-Butylbenzene + 1,2,4-Trimethylben. 0.99
n-Decane + 1,2,3-Trimethylbenzene 0.98
Indan 1.00
1,2,3,5-Trimethylbenzene 0.99
Naphthalene 0.96
n-Dodecane 1.00
AVERAGE RESPONSE FACTOR O.98
-------
29
TABLE 2: INSTRUMENTAL PARAMETERS
Gas Chromatograph: Hewlett-Packard 5880 or 5890
Column: Quadrex MS-007, fused silica
capillary column 25 m X 0.25 mm
ID, 1.0 urn film thickness methyl
silicon
Carrier gas: Helium, 15 Psig.
Make-up gas: Nitrogen, 30 ml/min.
Split Ratio: 30:1 (minimum)
Sample size: 1-5 uL
Injector: 325°C
Detector: Flame ionization, 350°C
Column program: 40°C hold for 4 min, program at
10°C/min to 280°C.
Hold for 15 min at 280°C.
Data System-. VG Multichrom
-------
30
TABLE 3:
RETENTION TIMES AND BOILING POINTS OF n-ALKANES
FOR DETERMINATION OF BOILING POINT DISTRIBUTION
OF GASOLINE TO DIESEL RANGE TPH IN SOIL USING
DESIGNATED INSTRUMENTAL PARAMETERS
,x--
GASOLINE
RANGE
V
^~
BP
°C
'36
69
98
126
151
174
196
216
~236
253
270
287
302
316
329
343
402
Retention
Time (min)
2.15
4.09
6.85
9.55
11.93
14.03
15.92
17.65
19.26
20.76
22.18
23.51
24.77
25.98
27.1 1
28.20
35.99
Alkane
Marker
n-C5
n-C6
n-C7
n-C8
n-C9
n-C10
n-C1 1
n-C12
n-C13
n-C14
n-C15
n-C16
n-C17
n-C18
n-C19
n-C20
n-C25
-------
31
CO
Q
DC
< Q
QZ
Z <
£ CO
CO tO Q
— 1 LU ^
0?c5
co-iz .
_ O < a
Q co i- LU
LU < CO |_
v o , <
— -1 o
b?™
CO ~ LU Z
£&5*
O LU rfc <
"•si-
CO UJ CC Q
b^S*
ID * Z Q
co!=»- o o
CC LU —
< co z >-
^ h- ± LU
-^ -I -J I
^ 3 Oh-
±r co co z
-^ LU < s.
CO DC O CO
% m
^f
LU
j
CQ
O m
O>^ «o^-ooio<-OOO)O>
Ul 2
"> ^~
< a
ZQ2
>- h— 0.
(0 COO.
|j Q 2
_ >- OL
a co a.
QQ 2
* 1- 0-
O co a.
COQ 2
< i- a.
W co a.
i-
ui uj 2
C9UO.
OC-Jfl.
^ ^
h- >
I-
OOOO>KOOOOCOO
O^O>O>O>N>^*CM
o a> co co
CM »-•*•»
cocot^-ooooo
« o 10 co
*- CM •* T
•^ CO O O
a »- o N.
«- CO t^
CM •»!•
u ui in uj
__
o E z z z
3
Q UJ
OQ-
cc>-
Q. 1-
-I _l -I _J
O 0 0 0
CO CO CO CO
3 CD CD CD
fv*-T-O>COOOO>OOO>
OCOCOKIOOOO
lOIOOKOOK^-CD
T- !•- «- •* O)
>- CM CO CO
^^CO^OOOO
•*CMOO'*'«-CMtDO)
^- CO h- O>
T- »- CM CM
•-aocMoooo
KCOCOOOCOCOCO
r- i- CD CO IA h>
•^ CM CO •»
CO'-OOOOOOO
•VIOCMKlOOOCMi-
»- T- 1- *- O
CM CM •* ^t
OOK'-lOOOOO
CO(OO)OOOOU5O'-
00
UJ •
co
O-J
u uj
< cc
o
cr
UJ
a.
UJ
a
cc
UJ
o
cc
o
o
cc
UJ
a)
x
a.
<
en
o
O
O
QC
X
o
CO
<
o
_l
111
UJ
_l
z
0
JI
LU
co
UJ
Q
.. -f
UJ 111
22 ..
*" """ ^~ 1
"OOw
-coco"
2«LH
_i o a Q
LU
cc
LU
u_
u_
Q
CM
o
z
ff\
Q
UJ
N
co
t-
o
<
DC
H
X
UJ
-------
32
o
CO
o
LU
LT)
r\j
Q.
CO
M
u>
o
c
E
O
E
i
5
IT)
O
QC
o
j- co
X ^
LU —
LL LU
O Z
LU
QC
D
O
LL.
I I
I I
J I
O
CM
O
o
o
CO
o
r^-
o
<>
o
in
o
Xi i sueau]
-------
33
o
in
•o
1
00
LJ
in
(M
(II
0)
«
E
O
in
o
o
o
LO
11 iiii
! I I
o
fM
O
o-
o
oo
o
CO
G
III
Q.
CO
O
DC
O
X O
LU CO
<
CC
CO
LU
P
5 CL
2 Q.
O CM
c\i
LU
DC
g
u.
(/\UI) XllSUBJU]
-------
34
in
-------
35
o
'o
!O
~!m
itn
o
m
io
o
o
01
=
O
«
o
|o
"I in
a.
o
CXJ
J L
O
O
J L
J L
O
CO
o
to
_J L
o
ru
LLJ
CO
LLJ
UJ Q
-J UJ
a. DC
"5 UJ
CO
o
CO
LU
LU
in
01
o
o
CVJ
'
1
o
in
n
0)
jj
3
C
t-(
^e
E
-^H
t-
0
DC
U.
h-
0
DC
1-
X
UJ
Li_
O
2
^^
UJ
DC
j—
X
111
Q
Z
<
UJ
Z
o
CO
^*»
cr
o
O
S
CD
I O
o o
cr.
LLJ
UJ
o
Z
<
DC
111
DC
(AW)
-------
36
I
M J
m
z
01
. N .
mW ^ < <
^"uj m GUJW
^Z -"llJm ^^O
sslss^»ss|3
z 2 m > uj H P p < oc o i^
[U UJ J x j uj o. 0 0 t- g Z
N3> , >-5UJoilJllJ_UJ
Z _J I g X - I , Q 1- UJ OL
UJOH+ iDCi' i i i i
(Dl-maoi-ccccc 0
x : -
.. p °Q..,^O^-OIO
.. .. OQ7x'2^^'-^ojw
OlI-lU o.OHOU0000
^
C-J
•
o
X X
e o J
Q. -^
oo —
(-J ^
H- — '
fs»
0 -J
O
LU
X
h-
z
>-
CO
5
^ «
« n
2 o-
J o
o
1 £
f—
LL
O
oc
o
^^
^M
O
oc
X
in
LU
cc
0
LL
(A"") Xnsueiu]
-------
37
0)
c
o
01
0)
cn
o a.
cn f—
X 03
CL CD
H- CL
CL
Q.
c
o
cr
o
o
o
OJ
o
o
CM
O
-O
O
o
f\J
O
CO
O
^r
CM
U
03
L
~0 O
03 O
3
0)
>
C.
D
LJ
UJ
o o
o oo
o —
N-)
C
m
c
o
(0
C
e
t_
Q)
ij
Q)
fNJ
O
IV. o
-O O
O hO
CM -O
L
O
L
0
o t_
03
«- TJ
u_ c
CD 0)
O w
cj tn
r\i
+
LU in
o
o
o
o
I
0) 0
CU 0
L L
cn cn
Q) 0
TJ T3
C " Tl
OCOC
LJ — r\i
CO
D
DC
<
a
z
£
CO
LU
z
_l
o
CO
<
o
o
z
CO
D
Z Q-
O Q-
H °
< s
DC q
OQ O
_j V
< o
o m
CO
LU
DC
D
CD
LL
(sAn) e
-------
38
10
o
<_> tn
•4-
ui eg
CK +
-O UI
~O O- K1
o in ex
ooo
. o o
i — d
0) Q)
0) 0)
<-. L L
C CD CD
CD Oj 0)
« "0 T3
C " "O
O W C
c_) — <"M
Q_
Q_
O
O
O
o"
i
O
10
CO
Q
QC
CO
—I
LU
CO
LLJ
Q
O
CO
Z)
O
cc
m
o
LJJ
CC
o
LL
(sAn)
-------
39
o
NJ
0)
cn
0)
0)
c
OJ
cn
_i
O CL
CD I—
CL
Q.
•o «
I
cr
o
o
o
OJ
a
o
o
L O
T> O
(0
rxj
+
LU
ID
O
O —
c
o
ro
c
E
L
(D
00
o o
L
O
L.
L
0)
M- -o
«- C
(D (D
O w
K1
+
UJ —.
03 +
00 LL)
r\J O 03
hO O —
S3 "O —
o r^ o
. r^. o
I 03 O
c
m
u
(fl
c
o
0) 0)
(D (1)
L L.
cn cn
o -> T)
0) C
— (M
Q
DC
<
D
1
CO
X
LU
CO
LU
Q
LU
O
CO
<
O
O
z
CO
D
O
Q.
"Z. CL
g CL
b °
< a
oc o
CD o"
^ S
O to
00
LU
OC
ID
O
O
O
O
•o
o
O
oo
-J-
o
-J-
r\j
-------
40
o
o
CO
Ul
tn
er
o Q.
en i—
re to
Q- Q)
I— Q.
CL
O.
C
O
cr
I
o
Q
O
O
-J-
O
o
f\l
o
-o
o-
o
fM
O
DO
O
-J-
fM
u
o
o
o
TJ O
to o
D
a —
C
o
CM
•t-
LJ
-O
O
O
CO
CO
C
£
L
M-
(_ 0)
D O
(— ) (_>
L
O
L
C
0)
TJ
L.
OJ
~Q
C
fl)
l_f
cn
•t-
LL) OJ
O +
ui LJJ
CO N~>
LO
co
O CO
r\j
o o
og
I
C
(X)
i_>
01
C
o
o
o
— o
(D 0)
0) 0)
L L
•D cn
D 0)
a
cr
<
o
I
CO
O
h:
UJ
a)
CD
•ZL
55
13
6
z
o
a.
a.
< o
CE O
CD o
i 5
O 10
LJJ
tr
:D
a
L
-------
41
LU
H
CO
2
O
cc
LL
O
CO
UJ
O
CO
<
O
CL
13
CO
UJ
O
CO
<
O
ci
UJ
cc
CO
O
a.
z
O
H
ID
CD
DC
I-
co
H
Z
O
Q_
O
CD
LU
X
O
cc
CL
CL
III
DC
:D
O
ooooooooooo
-------
42
0
w
CO
5
Q <
O «•
s w
O
C/3
X
IX
E-
Q
i
a,
w
u
to
0
w
a
O
-------
43
w
CD 2
* O
W
CO
W
s
8
w
2
O
a
g-
O
O
w
ac
05
O
04
X
u
o tf
« CD
O
CO
o s
CO
9 « s
o u o
K > u
« O ^
CSi
CO
o
"->
o
-------
44
WHAT IS "TPH"?
THE TERM "TOTAL PETROLEUM HYDROCARBONS" IS
USED TO DESCRIBE THE EXTENT OF CONTAMINATION
IN WATER, SOIL AND WASTE. HOWEVER, THE ACTUAL
VALUE DETERMINED IS METHOD DEPENDENT AND THUS
IT MUST BE DEFINED BY THE METHOD USED
WHAT ELSE CAN BE MEASURED AS "TPH"?
ANY OTHER ORGANIC COMPOUND (CLEANING FLUIDS,
SOLVENTS, POLAR COMPOUNDS, ETC)
WHAT IS NOT "TPH"?
• IT IS NOT TOTAL SINCE HEAVY HYDROCARBONS ARE
NOT ALWAYS EXTRACTED, VOLATILES CAN BE LOST
• SOME METHODS NEGLECT AROMATICS
• SUMS OF ONLY SELECTED COMPONENTS IN SOME
CASES
PROBLEM
WIDE ARRAY OF METHODS THAT PROVIDE DATA
OF VARYING AND QUESTIONABLE UTILITY
-------
45
WHAT ARE THE METHODS AVAILABLE FOR
TPH DETERMINATION
1-MOST METHODS INVOLVE SOME SORT OF
EXTRACTION PROCEDURE FOLLOWED BY
ANALYSIS USING:
• GRAVIMETRY
• INFRARED SPECTROPHOTOMETRY
• GAS CHROMATOGRAPHY MEASURING
-SELECTED COMPONENTS DETERMINATION
OR
-SUMS OF ALL COMPONENTS IN A GIVEN RANGE
2-HEADSPACE ANALYSIS USING GAS CHROMATOGRAPHY
(WITH AND WITHOUT EXTRACTION OF THE SAMPLE)
-------
46
CO
CO
PQ
Q
£ g
So
m E-1
••^ CjJ
W ^
co E-«
o
t-
w
oo
-------
47
TOTAL PETROLEUM HYDROCARBONS (TPH)
SOLVENT EXTRACTION/GAS CHROMATOGRAPHIC METHODS
• SAMPLE IS EXTRACTED WITH A SOLVENT
• EXTRACT IS INTRODUCED INTO A GAS CHROMATOGRAPH
EITHER BY DIRECT INJECTION OR BY PURGE AND TRAP
TECHNIQUES (THE LATTER IS ONLY APPLICABLE FOR
GASOLINE RANGE ORGANICS)
• THE CHROMATOGRAPHIC COLUMN SEPARATES
COMPONENTS IN THE SAMPLE
• THE COMPONENTS ARE DETECTED PRIMARILY BY A
FLAME IONIZATION DETECTOR WHICH RESPONDS TO
ALL CARBON-HYDROGEN CONTAINING COMPOUNDS
(THERE ARE OTHER DETECTORS THAT CAN BE USED,
SUCH AS PHOTOIONIZATION DETECTORS AND MASS
SPECTROMETERS, HOWEVER INCOMPLETE INFORMATION
IS USUALLY PROVIDED)
• TOTAL AREA OF CHROMATOGRAM IS INTEGRATED AND
QUANTIFIED BY COMPARISON WITH STANDARDS
-------
48
TPH USING GAS CHROMATOGRAPHY
GAS CHROMATOGRAPHIC METHODS CURRENTLY USED
INVOLVE:
• DETERMINATION OF GASOLINE RANGE MATERIAL
-Extraction
-Purge and Trap or Headspace Analysis
• DETERMINATION OF HEAVIER THAN GASOLINE RANGE
MATERIAL
-Extraction
-Concentration
-Analysis of concentrated extract
ADVANTAGES
• DETECTION LIMITS IN THE LOW PPMs
• SIMILAR TO EPA METHODS / BASED ON EPA METHODS
DISADVANTAGES
• TWO METHODS ARE REQUIRED
• INTENSIVE SAMPLE PREPARATION
• SCREENING NECESSARY
• LIMITED/SEGMENTED INFORMATION ON PRODUCT TYPE
WHAT ARE SIGNIFICANT TPH CONCENTRATIONS???
-------
49
Q
E-
co
(X
w§
O
°
o
u
w
Q
i
>
O
W OS,
E-J^
5<
O
CO
CO
>-
PQ
W
g
CO
ffi
a,
a,
CO
PU
O
-------
50
H
DC
O
Q.
CL
ID
CO
Q
Otf
DC
DC
O
LL
LU
Q.
O
UJ
>
LJJ
Q
Q
O
I
H
LLI
2
—I
<
O
H
co
0
o
••->
0
CO
h-
2
LLI
v^
2
LJJ
DC
Z)
O
LU
DC
0~
"o
CO
CO
O)
0
O)
c
CO
^
c
o
-l-«
cO
c
"E
rrt
vw
-»-«
C T
O 0
0 tt
0
^^~
O fli"
IU
T-«
0 ?
MX ^
a >~
>s 0
^^« ^^
•»->
0 "0
C CO
— CD
E •-
~ T3
i_ *^
0
^_j ^
t« •*-•
S-£
•
CO
T3
^^
0
••-•
CO
C
E
CO
-t->
C
o
o
^^
rt
vv
_co
'o
CO
c
•^
X
CL
1 —
1
**—
O
C
0
-»-•
CO
c
E
k_
0
*®
Q
•
X
HI
H
OQ
co
0
-*-•
^
CO
C
CO
•+-•
0
O)
u.
CO
•*-*
T3
0
•*-j
O
0
0
(ft
ws
>*—
o
c
o
•+-•
CO
c
E
v_
0
0
Q
•
M—
o
c
O
•*->
0
_0
0
CO
c >*
.= 0)
"O "^.
lj
o o
*- 0
+-•
c
0 C
~ o
3 ^
^3 CO
• ^M *«••
^ "2
w 2
S E
2
•»-•
.E
CO
T3
O
*-•
0
E
X
CL
O
C
0
«
O
CO
•a
0
o c
= o
o
*• c
=1
31
0 1-
o o
2c
CL""
_ CO
0
.22
O m
-------
51
O 0)
71
E
o
^^
^^
CO
>>
CO
o o
0)
C .ii
"a
u CD
O)
.c i
o 2
CL
CO
O
C
CO
o «j
c +-
CO C
.C » CO
CD X
E CD
E-::
3
O Q_
CM
O .5
•a
•a
CD
ex
ZJ
o
CO
o
^
CO
CO
CO
o
£ I
ZJ CJ
I i
T- *
-C
CD
H— -5
i_ CO
o
>
CB
*- p
CO
CO
CD
4-1
CO
o
"c
O
CO
•a
c
CO
CO
- o
CO
o
"c
o
CO
*~ >._•«>» v*
* >- 3
Q
o
CO
CO
CO Q. —
CO
CO
CO
CD
o
CD
C
-
**~
CD
O)
ZJ
•*—
i_
•+••
c
CD
O
_-
CO
•^
CD
Q.
E
CO
CO
o
+^
ZJ
CO
o
"o
CO
£
CD
'o
c
o
L.
0
Q.
CD
«*-
co
c
CO
^
H
c
CD
E
ZJ
|T
CO
c
-a
CD
"co
c
O)
'co
CD
•Q
O)
c
'co
ZJ
a
LL
O
0
CO
o
CO
i_
s^ CO
g»
CD O
-^ CO
CO »-
cz co
< a
co
73
CO
C
CO
M
JZ
i
c
o
CO
T"
CO
a
E
o
o
>»
JD
Q.
E
co
CO
."•
X
Q_
h-
0
5
ZJ
o
CO
O
CO
CD
>
ZJ
o
c
o
'^'Z
-Q"?
•~ c
CO ^
O kZ
_ E
o o
0
^ o
CO *-
0 •-
o co
CO u
**" >
CD D
CO O
c
0 c
a. o
CO • —
2 co
*- k.
O) ^"^
.Elo
CO 0
ZJ
»*-
i_ 0
CD
x: CD
r to
CD ID
-------
52
INSTRUMENTAL PARAMETERS
Gas Chromatograph: Hewlett-Packard 5880 or 5890
Column-. Quadrex MS-007, fused silica
capillary column 25 m X 0.25 mm
ID, 1.0 gm film thickness methyl
silicon
Carrier gas: Helium, 15 Psig.
Make-up gas: Nitrogen, 30 ml/min.
Split Ratio: 30:1 (minimum)
Sample size: 1-5 |jL
Injector: 325°C
Detector: Flame ionization, 350°C
Column program: 40°C hold for 4 min, program at
10°C/min to 280°C.
Hold for 15 min at 280°C.
Data System: VG Multichrom
-------
o
l/i
LJLJ
o
d
O
CO
<
O
in
rs»
0.
O.
O
O
CO
CM
o
in
O
co
Q
LU
*
Q.
CO
2
O
DC
LL
O
x
HI
u.
O
GC
O
S
2
O
GC
O
d
O
03
O
•o
o
in
^1I3U•1U1
-------
54
in
•o
o
o
J
1
in
fM
o
«M
u
C
O
•
in
o
•
o
LU
CO
LU
O
2
CL
CL
O
O
N.
CM
O
CO
Q
HI
^
QL
CO
2
O
GC
U_
O
<
GC
K
X
LU
U_
O
in
QC
CD
O
I I
J I
J I
1
J I
J I
o
o-'
o
CO
-o
Xaisueiu]
o
\r>
0 _i
CC d
X O
O CO
-------
55
in
(N
o
o
I I I
I I I
I I I
I I I
I I I
I I I
i IT)
ro
O
to
Lf>
(M
c
E
o
E
O
•
in
o
(M
o
o*
o
CO
o
•o
o
in
O
O
Xjisueiuj
HI
O
CO
<
O
CL
a.
o
o
C\J
CM
O
CO
Q
LLJ
^
Q_
CO
^
O
DC
U.
H
O
<
cc
h-
X
LU
< i^
§ -•
o S
< y
2 Q
O co
O Q_
-------
56
o
'o
o
5
10
"in
' 01
i «
10 c
• • **
O E
(U *—
>
5 o
=—*
o
ooooo oo
- CC
Z LU
II
O uj
O o
UJ Z
-J <
Q. OC
CO
CO Q
O
at
H- UJ
X 2
111 UJ
CC
I-
X
a.
O
o o
QC CO
I <
O O
ru
(Aiu)
-------
57
CO
Q
QC
Q
I
CO
o
DC
o
z
o
£
z
2
cc
UJ
H
LU
Q
X
Q_
»*—
O
o
o
v_
3
O
CO
u.
O
O
•g
L«
O
-
0
CO
O
•5
-o
c.
03
_^
O
o
X. *-
o co
O
w E
w o
^-
w
^ o
+-•
03 _D
+-•
03
x:
Hh^
O
O)
c
03
L_
0
JC
H-*
E
CO
c
o
•+-•
rrt
Z
O
<
*>v
z
c
••~
LO
CM
O
C
O
•«""•
E
Q.
a
o
o
o
T"~
1
O
o
o
•+«•
D
0
CO
T3
03
-o
T3
C
03
CO
T3
C
13
O
a.
£
o
o
o
O)
c
^""
"o
JD
•o
C
03
CO
4—1
c
a.
~' ^ 03 . .^ c
O Q. ••- *r o
> CO gj C 03 o
c 03 2: o £ c
03 0>GL > £ o
?o»ss:
co 3 o - )
-a •*- Q. 03 03 c
*- x E *- "- rt
03 •- = *- 0 «-
T3 E J< X C
c ° o o
H 05 _- O)-C
h: ~ <» ~ o *- ^
^- CO
o
"° - c -o
<»
-a c -a
o
c "JS> 2
o -
S I
8
co
DC
LLJ
H
LLI
Z
o
I-
3
go
DC
H
co
Q
O
£L
O
Z
03 03 o O co c 03 —;
(•) O) CO CO OJ O CO ^
O
• CO
CO 03 ¥ •£
O*- .*- -C O
( 03 CO •*- -^
c o .52 o °
O) C .C •*-
c oj •*- CO O
•- -c o o
c •*-• **•" C T
« g ° I S
c c o
0 - Q. S 2
o o £ .2
"O O S ^ O
•- ^ co Q .2
03 ._ •»- *- C
"O >- 03 O 03 O
C ^ *" £ '-^
o3 ^: o fl, *- 3
•*- t5 N ® o .a
CO >>. -C- «*- ._
— -^ C *-
«c g o •:«
Q- 5 03 oj a, .E
2 ® o « w °
A" E £ -o a Q.
-------
58
B: BENZENE
T: TOLUENE
, 1
o
~T
^J
in 1
r-j 1
UJ " ± "
Z
LU
m N ^ UJ
2 Z Z Z
^co uj < <
UUJ 03 OujOT
y z -JLUMJ mzO
-^_UjO^f1 _
t"-izz<5z<^<
m>tui-H:p i >2LUOLULU_LU
5 E X ~ I V Q H LU Q.
I— + i OC i p i i i i
Luaot-ccccc 0
.. t- 03 .. •• 6 ^ 6 to
mfx'SK^-'-wN
UJ aoi-OuOOOO
LJ J
fi 3
X ><
s O _^^_____^ ^
« -^
5 -
-------
59
Q)
C
o
(fl
cn
_i
O CL
~r (D
CL Q)
H- CL
r-o
u +
LU
w O O
CD O OO
L. O —
T> O ro
(0 O CM
D
a — —
c
o
C
e
L c.
0) O
*-* c_
0) L.
"D 0
0)
c
D
(M
o
o-
NO
0)
0)
L.
(D
T)
C
(TJ
LJ LJ tn
LU tM
^f +
o LU in
O
O
o
c
(0
0)
Q)
L L.
cn cn
Q) 0)
w "D ~D
cn
c ^ TJ
owe
(_) — 00
CO
Q
DC
CO
HI
o
CO
O
O
CO
CQ O
_i V
< o
o **>
(s/\nj eejy
-------
60
o
0)
VI
0)
O CL
CO K-
rt)
0)
CL
Q.
Q.
O
tx.
§
o
fM
O
o
rg
O
O
O
O
00
O
-r
r\j
O
(0
(. O
T3 O
03 O
LU
O CO
O
(M
a — —
c
o
(TJ
ID
E
L
0)
"
0)
T3
0)
o
i
t-
O
C-
C.
0)
(_
(D
TO
C
(0
(_)(_) CD
LJ CM
O +
-O LU
^D -3" hO
O LO O
N"> -O —i
O O O
. o o
— CD
L U
c cn cn
C *-* ~O
O cn c
<_) — P\J
a.
a.
o
o
o
o"
i
o
If)
ai
a
oc
CO
_J
LLJ
CO
LU
O
CD
CO
3
o
DC
m
o
(£-••01*) (sAn)
-------
61
CD
cn
CD
Q)
C
O
CO
(0
en
o CL
co H-
IX CD
Q. CD
h- Q.
U
i-> O
CO O
C O
T3 O
c\j
LT)
CO
00
O O
O — -O
C
O
w
u-
Q)
>
C_
D
(_)
0)
C
E
(_
0)
O
ID
*4_
0
V4—
V4—
0)
0
(_>
L
O
/
\_
(_
a>
ID
L
m
TJ
C
CO
«-»
in
LU —
CO +
CO UJ
f\J O CO
ro O —
-o -o —
o r^- o
. r^ o
i
C
(TJ
CO O
0) 0
CD 0)
(_ C_
01 cn
CD CD
T) TS
C w -D
owe
LJ — C\J
Q
DC
co
X
UJ
co
UJ
Q
HI
o
CO
o
Z
co
Z)
o
Z Q.
g Q-
\- o
< o
DC O
CD o"
li 7
< o
O 10
(£•••01*) (sAn) e
-------
62
o
o
o
o
tO
UJ
to
d
CD
O CL
CO I—
Q. 0)
i— CL
o
o
o
8 9=
o
cr
o
o
o
8
o
r
fNJ
u
0)
O
O
_ o
•o o
(0 O
u
a —
c
o
LO
GO
O CO
PxJ
I
C
(T)
CM
LU
O
-O
OO
._
u—
0)
>
C
D
(D
C
£
L
0>
j_»
Q)
T)
t~i_
O
*o_
V4—
0)
o
c
o
L
L
0)
-a
L
(0
•Q
C
(Q
w
tO
ro
-t-
LU CM
O +
LO UJ
OO N1 OJ
(M
O O
O
O
— O
01 0)
0) Q)
1- L
cn cn
0) 0)
" "D -D
C u
O U)
a
a:
<
9
I
CO
o
H
UJ
oo
O
:i
GO
ID
O
-I Q_
O Q.
H O
< O
cn o
EE o
< 5
a 10
-------
63
p
•
0°*
cn^#>nnirt^-fnf^K. r^r-»-i«^encD h^»-ror^CMf^»— Oi K.rflH
< 3E o>oococoo>coo>o> coo>a>coo> o>a>oococococoo>i co I
LU 2
> °-
«t Q.
j
ZQ2
•
_J
^^
__J ^ 1
mooot^cooocoo oco CM^^CJ)!
cM't-Tr'v '-CMCOCO CMCMinmi
,
i
1
cocos>ooooo ^^co^oooo o>ooomoooo
co^^co ^coh*o) ooh^^in
'-'-COCO '-'-CMCM »- »- ^- ^ I
/ Y} ^— '
mil i,, O *£
^^^* _ _ in- rt f^ CM IO ^9 CO C3 CO O '— O) ^3 CM CD CD O CD CO CM CO ^ CD CD CO O
f^^ ^1 ^*l (Q rt j f** to CM CM ^™ CO W5 CO f^ CO CO CO CO CO CO CO ^ CO ^" O) O) CO ^ CO
^L ^^ T-»-cM^h»v T-»-o>«o>of^ ^- coomo
I CO 1
Q. -1
H0 oof
l.l_ (0 C3COQ.
°Q
CM CM TT lO »~CMCO^ CMCOinh-
UJ'-OOOOOO CO'-OOOOOOO OCMWCOOOOO
T-T-mo*o- LU ~" KOO"J ^
5^ coo^
0 i
^f _ o ^ *"
^— - ' Z •• LU
> : LU - O
^ a' cr
ID Z— LU
CO £012
COZJO.
OC_I a.
££
1
O -~ Q.
f-cooo OOKOO CMTOO _ — %
IO *- O !-• U> »- CO » lOOt^CO1 &rf
i-COt^ '-CM'- i- K CD U* _ ^
CM'* CM-V CM(0 £;>LU
LU >
LUQ >
CgOC O
; rf *( ^
/r Q CC
rT -y *~i
j >< o
i -^^ ^
H ILULULULLIOTOTOTOT i^iu,
o zzzZuji^j^jm rf>"j
Qtu'oooo °99? mgud °^^<
OO.COCOCOCO COCOCOCO 2 S S m : O-1LU
OC5- <<<< <<<< !fi!ii!fi!ff OLU >
cLi-'cscocsa ooao QQQQ -
z
co
H
0
cr
t-
X
LU
-------
64
03
o
h-
gg
DC
CO
Q
O
a.
o
O
CQ
LLJ
x
O
DC
CL
CL
•o
0
CO
13
C
E
ZJ
0
O
0
'^1
a
03
0)
o
•*-»
03
O
i
hw
JZ
O
0
.c
0
CO
03
JZ
a
03
c
o
•4— «
03
CO
0
£1
CO
•a
c
o
a
E
o
o
*o
c
o
"03
03
a
0
CO
L!
"o
Q.
l
C
O
c
CO
•o
o
•4—"
0
E
CO
•4-*
c
'0
+-•
c
CO
0
o
c
0
5
•+—
^^»
^^»
"™*
c
0
CO
CO
0
;hieved
u
03
CO
/1\
0
X
E
CO
03
0
CO
0
03
kanes
03
l
C
•*
•o
o
JZ
0
E
CO
c
•*-<
c
'o
Q.
0)
C
^—
'o
0
E
c
o
c
0
>t %
0
*o
c
o
'•*-'
03
0
O
0
0
JZ
0
c
E
Q)
0
•o
0
-t->
CO
0
o
c
0
0
^
^— V
1C
CM
O
CO
O
CO
0
c
03
"03
i
C
0
JZ
+-
o
CO
^^
c
'o
Q.
O)
C
'o
ja
c
o
c
0
~
•4-1
i
0
^
..
o
03
o
CO
c
o
•1-^
^
CO
'•5
c
'o
Q.
D)
'5
0
"03
E
X
o
Q.
OL
"o
CO
c
0
1—
o
Q.
>4—
o
CO
03
0
03
XL
03
0
Q.
**—
O
CO
E
CO
«*-
0
c
o
"03
N
E
o
c
£
H—
O
ff*
CO
c
'o
a
O)
'5
-Q
C
0
0
0)
O)
D
0
grams
o
15
E
o
o
0
^u
JZ
CO
0
0
C
-------
65
RETENTION TIMES AND BOILING POINTS OF n-ALKANES
FOR DETERMINATION OF BOILING POINT DISTRIBUTION
OF GASOLINE TO DIESEL RANGE TPH IN SOIL USING
DESIGNATED INSTRUMENTAL PARAMETERS
1
BP
°C
,^T 3Q
| 69
98
126
GASOLINE /
RANGE 151
174
196
216
~236
253
270
287
302
316
329
343
402
Retention
Time (min)
2.15
4.09
6.85
9.55
1 1.93
14.03
15.92
17.65
19.26
20.76
22.18
23.51
24.77
25.98
27.1 1
28.20
35.99
Alkane
Marker
n-C5
n-C6
n-C7
n-C8
n-C9
n-C10
n-C1 1
n-C12
n-C13
n-C14
n-C15
n-C16
n-C17
n-C18
n-C19
n-C20
n-C25
-------
66
H
Z)
E9 LU
H a
co -z.
— LU
O CO
Q_ LU
So
_J H
O LU
CD Z
LU _l
t?°
< CO
s <
x^
O
cc
Q.
CL
UJ
CO
I
UJ
UJ
UJ
CO UJ
UJ Z
Q
O O
+
r! a!
O =>
CO CO
Q
LU
O
— oc o
9 "- o
O O
co co
LU
9
DC
o
-; i
*. o o
2 2 LU
LU
_l
>-
O UJ I-
co oc x
LU
O
CO
-------
67
ALTERNATE GASOLINE RANGE METHOD
1
-------
68
GASOLINE RANGE TPH
i:
• Extraction of soil samples with methylene chloride
or methanol results in inability to determine most
gasoline components below C6. As much as 25% of
gasoline is in this range.
• To assess properly gasoline range TPH as well as to
estimate boiling point distribution of contaminants
in soil, an alternate method was developed.
THE METHOD USES n-TETRADECANE (BEYOND
GASOLINE RANGE) FOR EXTRACTION FOLLOWED
BY ANALYSIS OF EXTRACT USING GC-FID
-------
69
Q
LU
IT
D
0.
2
z
UJ
3
o
v>
z
UJ
Z
UJ
X
I
a
ui
Z
UJ
X
o
o
00
o
•
UI
o
•
rg
«
UJ
Z
UJ
D
z
UJ
o
I
h-
UJ
10
o
to
o
o
o-'
o
•o
Q.
CO
O
CO
u.
o
H
o
<
DC
H
X
III
LU
o
HI
Q
5 CC
Q.
LU Q.
*- o
li. O
o ^
o -J
o o
h- co
DC
I-
o
in
O
fM
O
O
-------
70
TETRADECANE EXTRACTION METHOD
RECOVERY STUDIES. SAND AND
SANDY LOAM (*) SPIKED WITH GASOLINE
SPIKED
TPH
ug/g
50
50
50
50
100
100
100
100
500
500
500
500
1000
1000
1000
1000
1020
1020
*1000
*1000
*1000
*1000
12000
12000
ADDED
WATER
%
0
0
10
10
0
0
10
10
0
0
10
10
0
0
10
10
0
10
0
0
10
10
0 1
10 1
FOUND
TPH
ug/g
51
48
52
51
101
98
95
92
470
460
370
340
880
880
950
950
1000
990
940
950
900
870
1600
0500
TPH
RECOVERY
%
102
96
103
102
101
98
95
92
94
92
74
68
88
88
95
95
98
97
94
95
90
87
97
95
AVERAGE RECOVERY: 93±8%
-------
71
0
~
°
Q
O
r\
o
n
-
O
w
o
CO
><
O
CO
W
2
O i Q
t
O O
C/3 CO
O
3
P O
CO CO
X
u
05
w
E-
o
I
G
i—i
£
Q
U
E-
X
O
CO
ooooooooooo
-------
72
Q
LU
Q
Q
LU
Z
O
o
LU
DC
•D
co
O
LL
O
I-
o
111
LL
LL
LU
O
o
O
LU
O
LL
LL
LU
O
<
DC
H
X
LU
CO
Q
LU
Q
Q
DC
LU
I
O) C\J
O) CO c\J
o ^ S co
O 00 CM ^
O>
CM °> g ° o) o
O 00 CM T- G* ?
lO co O
O) °° CM
g cq
g oo
CD oo
I
Q.
H
Q
LU
DC
CO
LU
:>
LU LU
Z Z
LU LU X
N Z) Q.
Z -I h-
LU O .
03 H >-
o Q£
LU LU i£
W 2 LU
LU LU ffi
LU LU
Z Z
LU LU
N ID
Z _J
LU O
m H
DC DC
LU LU
> >
O O
o o
LU LU
DC DC
T— *
-------
73
SUMMARY
DEVELOPED SOIL EXTRACTION PROCEDURES FOLLOW
BY GC-FID ANALYSIS THAT PROVIDES INFORMATION ON
• PRODUCT TYPE FINGERPRINT
• INDIVIDUAL TARGET COMPONENTS
• GASOLINE TO DIESEL RANGE TPH
• APPROXIMATE BOILING POINT DISTRIBUTION
• THE TPH RANGE TESTED WAS 50 TO 10,000 PPM IN
-------
74
APPLICATIONS TO DATE
GASOLINE TO DIESEL RANGE TPH
SPIKED SOIL SAMPLES
SOILS FROM SERVICE STATIONS AND DISTRIBUTION
TERMINALS
DRILLING MUDS
u
J
-------
75
Q _
~ 2
u, 5
i—«
Q <
O 2
2 w
Q
2 < 5
~ a <
cu < &
CO
E-
U
PQ
O
c/3
o
w
O
CN
-4
U
05
K
U
ffi
a,
E-
0
CO
E-
U
E-
X
W
V
1
6
$
(C
^
«
«
1
4
V
•»
U
2
O
U
Q
e- E
oJ5 i
PL, U
0
Q
H-l
U,
1
U
CJ5
2
O
i— i
t-
u
•->
2
h-H
E-H
U
w
(i!
1— 1
Q
s~<.
^4
2
P*
PL.
O
Q
i— 1
ffi
O.
1 • 1
E-<
2
PL,
O,
lO
E-
CO
^
W
D
PL,
g
O H
S £
< 5
U H
SS
>• <
E- p
G
-------
76
MR. TELLIARD: I would like to
read an announcement that I received from our laboratory in
Cincinnati, EMSL. They're looking for a few good volunteers.
They're looking for some labs to participate in three round robin
studies, one on Method 83ISA for the determination of carboxyl
compounds by HPLC and that's going to be during June of '91, a
method for marina nitrates for the determination of nitrate,
nitrites in estuarine and coastal waters by automated color
metric analysis and that's for June, '91, and method 524.2,
determination of purge-able organic compounds in water by
capillary column GC/MS and that's for October/November of '91.
If any of your folks from any of the labs are
interested in participating, there's a sign-up sheet with a
little more information on the back table in the corner for those
of you who would like to.
Our next speaker is Rick Beach. Rick is the Director
of Research and Development for Hydrosystems in Sterling,
Virginia, and he's going to talk to us on a screening method for
total polynuclear aromatics.
Do you want to use the walk-around microphone or do you
want to use the podium microphone?
-------
77
MR. BEACH: Good morning. Can I
have the first slide?
My talk today will focus on a screening method for
total polynuclear aromatics (TPNAs) that I developed at
Hydrosystems. Actually the method goes a little farther in that
it can resolve groups of polynuclear aromatics, the specifics of
which I will discuss a little later in my talk.
I'd like to thank a number of people in my laboratory
who contributed to this work: Kelly West, Lyle Silka, Mike
Albertson and Arkady Gilchenok.
The technique I will be discussing today was developed
for use in evaluating creosote residues at the L.A. Clarke
Superfund site in Fredericksburg, Virginia.
The method works well for almost any PNA contamination,
whether from creosote, petroleum hydrocarbons, or from coal tar.
In addition, we're investigating some other applications right
now.
This is a picture of actual operations at L.A. Clarke
in, I believe, the 1970's. This photo illustrates the creosote
wood treating operations used here during that time. Railroad
ties, contained on railroad cars, were treated in large
autoclaves filled with creosote and exposed to elevated pressures
and temperatures. Under these conditions, the creosote permeates
the wood. After treatment, the creosote was drained off, the
autoclave opened, and the railroad car containing the treated
ties was removed to what is called "drip tracks". At this point
creosote was allowed to "drip" onto the ground below the drip
tracks, resulting in substantial soil contamination. In the
newer plants operating today, the drip process occurs in a
concrete liner or some other type of liner, to prevent soil
contamination.
Screening techniques are highly effective in evaluating
large contamination sites such as the L.A. Clarke site. During
on-going operations, it's very easy to tell where the
contamination is; it sticks to the bottom of your shoes.
-------
78
However, a couple of years later when the drip tracks are ripped
up, the soil is moved around, and you have a little bit of
surface biodegradation, you can't easily tell, other than by
historical photos, where the main problem area was located.
Furthermore, you can't always determine from drainage ditches cind
other spill activities where the highest concentrations of the
contaminants may be on the site. Therefore, a screening
technique is a highly efficient and cost effective tool for use
in evaluating contamination of large sites.
For the work at L.A. Clarke, Hydrosystems inherited a
screening method as part of a consent order between the
Responsible Party and the EPA prior to our taking over the
Superfund work. The initial RIFS work on the first operable unit
was done by R.F. Weston. They developed a technique in 1985 or
'86 that was based on a fluorescent scanning technique of
acetonitrile extracts of soils. The technique was set up to
quantitate two different groups of PNAs by evaluating 2 & 3
ringed compounds and the second group was to evaluate 3 to 6
ringed compounds.
R.F. Weston had used this method to report areas of
major contamination at the L.A. Clarke site. We had planned on
using their method in our subsequent investigations there. In
fact, part of the clean-up standards for this site are based on
the Weston screening technique.
However, before tackling the L.A. Clarke site, we
decided to use the Weston screening method for a separate
investigation of oil and fuel contamination at a railroad
switching yard. We surveyed approximately 50 acres of the 500
acre site. This slide is a photo of a portion of the whole
switching yard. Pictured here is an area containing a braking
system and an automatic lubricator, as well as the beginning of
the distribution yard. There is extensive contamination in these
areas resulting from general railroad activity, as well as from
accidental spills, leaks from cars, etcetera.
Unfortunately, the Weston method did not work well
-------
79
here. We found that we could not get consistent matrix spike
recoveries or sample precision on the analyses. We are really
disturbed about this because we had just obligated to doing
several hundreds of thousands of dollars worth of work on the
L.A. Clarke Superfund site with a technique that had significant
problems. So we went back to the data that Weston had generated
from the RIFS for this site to hopefully resolve the
discrepancies with the method.
The first thing we noticed when we went through the
data was that no data for matrix spike recoveries or sample
duplicates existed in the RIFS report. I asked some of the
people that were involved in the work about the ommision and they
explained that they had not performed such QC measures. They
considered their method to be a very gross screening technique
and therefore felt that the matrix spikes and duplicates weren't
warranted.
Their raw data did contain a range of dilution results
from a number of individual samples. We used this data to
estimate whether or not they actually encountered the same matrix
effects that we did. To make this estimate, I looked at the
ratio of the diluted concentration of a given sample corrected
for its dilution factor to the non-diluted concentration. In the
absence of matrix interferences or non-linearity of response,
this value should equal 100 percent. I did this for all of the
samples within a factor of two of the calibration limits. For
the two groups of PNSs identified by the Weston method, their
analyses resulted in values of over 200 percent. In my view, a
two and a half fold increase in response is stretching it a
little bit too far for only being at most a factor of two out of
the linear working range.
To constrain these results further, we narrowed down
the samples to just those in the linear working range. These are
samples in which the original analyses and the diluted analyses
are both in the linear range. Unfortunately, only nine samples
from group 1 and none from group 2 fit this criteria. In fact,
-------
80
in this case, the ratio got even higher than the previous group,
suggesting that there may be a strong negative interference - the
greater the dilution, the greater the deviation above 100
percent. This result was very disconcerting, and overwhelmed any
attempt to use matrix spike recoveries in conjunction with the
Weston method.
We were therefore forced to scrap the Weston method and
develop a new screening method that could use matrix spike
recoveries and sample replicates to evaluate accuracy and
precision. The results of these QC measures could then be used
to insure that the project data quality objectives are met.
Weston used their method to map large areas of the site
and to identify certain problems, but we were at a disadvantage.
We took over the method with the job and had to incorporate data
quality objectives into our quality assurance plan prior to our
discovering the inherent problems with the technique. I had made
up what I thought were a fairly wide range of data quality
objectives but found that with the Weston technique, we would not
be able to achieve the goals. So we tried using the same
acetonitrile extraction that Weston used, but substituted an HPLC
setup with a low resolution, reverse phase column to give us more
group information than what the Weston method had provided. We
calibrated on equal concentrations of the 16 priority pollutant
PNAs. It's also a very easy standard to come by commercially.
This slide illustrates what you get on your
chromatogram with the setup that we utilized. The first thing
you may notice is there are four groups identified and the first
group is titled, one to three rings. Well, if you have one ring,
it's not really a PNA. But what we found out is that it was
going to be too much work to try to resolve the difference
between the earliest eluting PNA, which is naphthalene with two
rings, and some natural or non-natural fluorescing one-ringed
compounds that are co-eluting. So what we did is to lump them
all together in this case. Therefore the title is somewhat
misleading. However, we are doing more studies to try to narrow
-------
81
this group down without making the method too unwieldy.
For soils, however, we found that there wasn't very
much in the way of early eluting compounds that would need to be
separated from naphthalene. The signal was very small in this
case, so we just let it ride.
The other groups are four ring compounds, five and six
ring compounds, and what I have indicated with a question mark is
greater than six ring compounds. Primarily, they're substituted
six ring compounds. Later, I'll explain why we wanted these last
two groups resolved this way.
Let me back up for a second.
The distribution of the priority pollutant PNAs is
this: there are six priority pollutant compounds in the one to
three ring group, four compounds in the four ring group, and six
compounds in the five to six ring group. So there's a nice
distribution of the compounds across the majority of the groups
that we can resolve.
The calibration curve takes a little interpretation if
you're not used to group concentrations. The x-axis on this
slide is the concentration of each of the PNA compounds in the 16
compound standard mixture. The y-axis represents the cumulative
area for each of the groups of PNA compounds.
To establish calibration factors, you first multiply
the compound concentration by the number of compounds in each
group to determine the group concentration. The group
calibration factor is the group concentration divided by the
cumulative area of the compounds within that group.
The curve shown is one of our first calibration curves
from last summer. Since then we've extended the linear range to
10 parts per million liquid concentration. The soil
concentration gives a range fairly nice dynamic range from
approximately 0.5 parts per million of each TPNA group to between
400 and 500 parts per million of total PNA. This is a very nice,
large range for doing screening analyses.
This slide is courtesy of Hewlett-Packard. Our
-------
82
analyses were conducted on a Hewlett-Packard 1050 quaternary HPLC
system. We're using a binary format right now with a program
going from 40 percent acetonitrile to 100 percent acetonitrile at
eight minutes with water as the balance of the solvent. The
fluorescence detector we're using is the model 1046, set up with
excitation at 235 nanometers, emission at 420 nanometers, and a
cutoff filter at 380 nanometers.
The sample preparation involves extracting two grams of
non-dried soil or sediment with 10 mLs of acetonitrile in a
simple vortexer, allowing it to settle, and filtering the
suspension with a 0.2 micron a crodisc. It's a very simple
procedure with the matrix spike incorporated prior to the
extraction.
The data that you see in this slide is actual QC data
from the project. This is not a traditional method of validation
where you take one fairly unitern soil type, homogenize it
completely, spike it, run duplicates and create nice statistics;.
These project statistics were generated from actual field
samples.
For this project our matrix spike data quality
objective was 40 to 160 percent recovery. The recovery we got
for the three groups that we have standard compounds for ranged
between 126 and 142 percent recovery. This includes all rejected
data. By rejected data, we mean data from soil samples with
other evidence of being heterogeneic. If you excluded the
rejected data, the average percent recovery would drop
approximately 20 percentage points here. So the data for the
actual project on the average worked out very well. The actual
number of samples that we used to calculate statistics was 300
screening analyses. We ran matrix spikes and duplicates on one
out of every 10 samples. We also ran GC/MS confirmations on the
basis of one in 10 samples, which you'll see the results of a
little bit later.
This slide shows the sample duplicate results. We
evaluated the precision as the relative percent difference
-------
83
between the soil duplicates. The data quality objective was set
at less than 67 percent. This DQO is easy to evaluate,if the
duplicate values are within a factor of two of each other, the
data quality objective of less than 67 percent is attained. The
average relative percent difference ranged between 42 and 45
percent. Again, the data would be better if we had excluded the
rejected data, but we did not. These are real sample points.
The number of samples evaluated as part of the
statistics varied depending on whether the matrix spikes were
diluted out or below detection. The same criteria applies for
the sample duplicates.
This work was performed over a one and a half month
period back last summer. All analyses were done under one
initial calibration. The initial calibration agreed with check
standards performed daily on every 10 to 20 samples for the
entire period. My acceptable criteria was a relative standard
deviation of not more than 25 percent. In actuality, trend
analyses showed that the relative standard deviation increased
with time. So if I had put a smaller RSD on my check standards,
I would have perofrmed a new intital calibration and gotten a
better set of data for the overall project. But again, it wasn't
needed for these project goals, which was to evaluate large areas
of the site. The idea was to come up with a very economical
screening technique that fit within reasonable data quality
objectives, which is what the new method did.
This slide demonstrates the increased accuracy that we
got out of our approach. The ROD for the site specified cleanup
standards based on the Weston screening technique and its'
correlation to the standardized EPA methods. We were confused as
to what to apply as cleanup standards if we weren't using the
same exact screening techniques. Therefore one of the things we
had to do was to come up with a new correlation between the GC/MS
data and the new screening technique.
What we have plotted on this slide with a log-log scale
is a sum of all the PNAs that we reported from the GC/MS
-------
84
confirmation analyses. The GC/MS data included three groups of
PNA compounds. The first group was called carcinogenic PNAs
(CPNAs). This group of compounds is the result of a lot of risk
assessment work and has created particular problems for the
analytical methods. The CPNAs consist of 24 compounds, seven of
which are priority pollutants, 17 are not. Of those 17 that are
not priority pollutant compounds, you can only get commercial
standards 'for seven of the PNAs. That left 10 PNAs without
standards. You have to estimate concentrations for these
compounds using relative response factors of the closest elutirig
standard.
Therefore the first group of GC/MS data was
carcinogenic PNAs. The second group was the remainder of the
priority pollutant PNAs, which consisted of nine other compounds,
and the third group which was the tentatively identified
compounds based on library searches and estimated concentrations.
After you lump GC/MS data from all 3 groups together,
you get a number which we included as the TPNAs for the GC/MS.
We plotted those values versus the TPNAs of all the four groups
that we had from the TPNA screening technique and we got this
great correlation of 0.942. For a screening analysis, this was
totally amazing and I was really surprised by this result. I
confess that I did throw out one data point out stat 25 values.,
The data point was located two orders of magnitude off the line.
When I first saw it plotted on the graph, I went back and
searched for the archived sample, planning to re-analyze the
soil. Unfortunately I couldn't find the sample. I went through
the calculations and chromatogram for the sample and found some
inconsistencies in the TPNA groups. Therefore given this
uncertainty in the result, I felt somewhat justified in excluding
that one sample point. I can undoubtedly justify the exclusion
statistically if I use a Q test on it.The other nicety of the
correlation information was that the slope came out close to one.
This was really surprising, considering the fact that I had all
of these non-compound specific calibrations. But the slope of
-------
85
0.885 resulted in an easy comparison of GC/MS to screening data
for project personnel. The same correlation coefficient, to
three decimal places, was calculated from the correlation of
CPNAs to the two comparable groups of the TPNA screening
technique, the five to six ring and the substituted six ring
group. It came out quite well. The same correlation
coefficient, but, the slope, 0.787, fell off a little bit. But
we can still use this comparison to plot a limited sub-set of the
data that's very useful in terms of evaluating remedial
alternatives for the CPNAs.
We've used this information to the data in terms of CLP
equivalents because that's the procedures upon which we're basing
our standardized analyses. This approach also makes the results
more palatable to project people looking at the data. They can
get confused when interpretting results of screening analyses.
Therefore we put the values in familiar terms and units, even
though in this case there wasn't very much of a change in the
actual numbers. The screening results discussed here are from
the analyses of soils and sediments from a floodplain area of the
site. The extensive data set allowed us to develop isopleths
over the whole flood plain area with very good delineation of
areas of contamination. It was a particularly troublesome area
to investigate. The area was 1,700 feet long and 700 feet wide.
The floodplain is on an old stream bed which gets flooded several
times a year. Over the years, all the streams' paths have
meandered and resulted in changes in the depositioned areas.
During creosoting operations over the last 50 years, some of the
drainage ditches from the site emptied to the floodplain and you
may hav had pure creosote contaminating the area. But due to the
changes in the stream paths you can't just go to the stream beds
where they exist now to locate contamination. Unfortunately, I
can't present that data right now, but we found contamination in
areas you wouldn't have located if you had to use a more limited
number of samples.
The final flood plain survey consisted of 360 screening
-------
86
analyses and 36 GC/MS analyses on soils from 180 stations (two
depths). Given the approximate 10:1 cost ratio between the two
types of analyses, the costs for each type of analyses were the
same. The total analytical cost for this investigation turned
out to be approximately $36,000.00. If you had used the budget
only for GC/MS analyses, it would have only given you 72 analyses
on soils from 36 stations. When you begin evaluating data from a
very complicated site, with the potential of isolated soil source
areas, you really need that fivefold increase in site
information. The increased information is particularly useful if
you can evaluate contamination within a factor of two and be
assured of quality control measures that you can use to verify
that the system is under control. This approach is advantageous
if the TPNA group results are adequate for your specific data
usage.
Our current work in progress using the TPNA method
involves analysis of soil types other than flood plain soils and
sediments. We're analyzing different types of geologic matrices,
including clays, loams, and rocks. We're also analyzing TPNAs in
surface and groundwaters. There are some problems that we're
working on for water samples, primarily the lack of resolution
between the one and two ringed compounds. The resolution of the
early fluorescence peaks becomes very critical in the water
samples. This problem is aggravated in water samples due to the
high percentage of one and two ringed compounds relative to that
found in soils. There's very little in the way of the higher
ringed compounds in the water samples since these compounds are
very insoluble and are so "particle active."
We've also used the TPNA screening method for
monitoring bioremedial activities on the L. A. Clarke site.
We're currently using batch reactors, thin film reactors, and
artificial wetlands. We plan on using it for monitoring
activities associated with land farming and soil flushing. We
have a fairly wide latitude from EPA in terms of evaluating
bioremedial techniques and we're at the pilot scale on a number
-------
87
of the different remedial studies for this particular site.
We're also using the TPNA method to create an internal
database to further expand the application of the method. Any
sample that comes into our laboratory as part of a UST
investigation, which could include contamination from gasoline,
lubricating oils, and motor oils, is also analyzed for TPNAs. We
don't normally report the results. We add the results to the
internal database that we can correlate with other petroleum
hydrocarbon technigues, primarily the TPH technigue by IR which
is the most commonly reguested method in our area.
In summary, we've come up with a screening technigue
that's applicable to PNA contamination from a variety of sources
and believe that the method is a very economical technigue which
can utilize QC measures to confirm the validity of the results
and provide valuable site information. It's reliable and rock-
solid in terms of the analytical technigue and may have a lot of
different types of applications.
Are there any guestions?
-------
88
QUESTION AND ANSWER SESSION
MR. BEACH: Any questions? Are
you letting me off easy?
Thank you.
MR. TELLIARD: Thank you, Rick.
-------
89
CO
o
s ^
? o
^ CC
i- 3
in **•
cc
rr<
QC UJ
zori
= LJ. O
UJ
LLJ
DC
O
CO
o
Q.
CO
^ j*
S °
CO c
^ 0)
*J°
8-^
^3
^ c
"S £
0)
^ ^5
CO <
O .
CO Q
CD "o
"E ^
c5 .o
o
£
.r o
|.^co
.Io =
0
^^^^_ ^HH
s Q)
CO
^ a
UJ O
h- u
CO >
> £
CO O
O-o
cc c
Q CO
52 o
c ^
0) «-
acsi
co^
" co
o'S
2'5)
o
o
CM
O)
S
-------
90
111
CO
CD
LU
o:
o
O
O
LJLJ
h-
o
CO
O
UJ
or
o
LL
O
UJ
o
CO
-------
91
LLI
Q
cc
>
CD
Z
X
O
O
O
DC
LL
O
UJ
a
CO
-------
LU
CC
_j
O
<; LU
J 3
^" ^1
^1 ^^
0) O
CO -J
m n
cc9
< z
z
CL Z
Z^^
Z
0<
f/5 o
Ml "•!
!> H
$ ID
LL —
00
zx
o t
^ ^
^
>
LU
LU
D
^
>
^^^
Q
1
O
z
Q
LU
h-
O
vO
0s
=
a.
cc
CD
Q.
cc
CD
Z
0
QL
CC
O
CO
LU
O
^
1-
^
O
92
CD
^- ^"
0 II
CM "
C
CD
00 *"
% "
C
LU
CD
a:
LU
<
— J
-j!
^<
cc
LU
>
o
<
z
„«
^c
UJ
CD
CC
•
OQ
^
O
Z
z
>
^^
.
CD
>
^
0)
D)
C
CO
-Q
To
o
c
*F
*•*
'i
CO
T3
O
C
s
0)
.Q
CO
^O
"a
a
CO
i
^
Z
-------
93
Q
O
X
h-
LU
5
£=.
<
Q.
L_
^^
Q.
<
O
H
LU
Z
CO
o
CO
LL
^^••i
o
o
1-
0
<
cc
1—
X
LU
LU
E
H
z
o
h-
LU
0
<
1
Q.
DC
00
i
o
•
o
CO
LU
DC
O
i
O
Q.
I
i
O
f-
o
LU
H
LU
Q
LU
O
LU
O
CO
LU
DC
O
LL
i
w
<
a.
Q.
Q.
CD
LU
X
H
LL
O
CO
Q.
o
DC
CD
Z
o
CO
<
O
1
-------
94
CO
Q.
LL
O
Q.
UJ
CO
O
CO
UJ
DC
O
CL
0-
i
CO
O
z
£
CO
CO
CD
Z
£
CO
CD
Z
DC
CD
CO
CD
£
r*.
«*s
in
A
-------
95
C
o
to
n
"5
O
<
0.
CO CO
D) D)
C CO C
if gif
CO jf CD
sjunoQ
-------
96
LJJ
o
Q.
I
a
LD
o
o
oc
o
<
Q_
h-
LJJ
LJJ
I
LL
O
LJLJ
G
-------
97
CO
_
o
CO
DC
o
LL
E PREPARATION
a.
|jj
CO
•
r—
LLI
S
Q
LIJ
CO
*fw
O
CO
Q
LLI
CC
Q
Z
o
Z
CD
q
CM
1
X
§
.
o
1
O
CO
Q
O
DC
O
^
D
04
6
I
H
DC
LLI
h-
LL
Q
Z
LLI
H
^^
LJJ
CO
i
-------
98
CO
HI
^
•MM
E
C/)
^^ ^
X
cc
1—
<
^
5
^_
o _
CO vo
M ^
U- o
O co
>- *~
CC '
m^^
C3
> 5-
^^ VI
8 •
111 0
cc o
a
£ Q
i
^
O
<
cc
o
o
z
a.
CO
CD
z
cc
CO
1
LO
RINGS
xt
CO
CD
Z
CC
CO
1
o
CO
T™
CM
xt
CO
CM
>-
cc
LLJ
o
o
LJJ
CC
CD
>
<
r^ xt
CO CN
ID xt
r^ CNI
CO ID
xt CM
•
>
LJJ
Q
ez
CO
r"
o
O)
CO
o
00
1
o
O)
CM
CM
CO
Q
LJJ
•••
^
cc
o
H.
cc
LU
Q.
O
a
-------
99
CO
LU
t_
<
O
Q.
D
Q
— ^*.
O >g
« IN
LL CO
0 v
Q ii
0. "
ft^
c 0
• M
-1 O
Z Q
0 "
CO
O
UJ
DC
QL
<
Q.
H
CO
O
z
DC
CO
i
in
CO
O
Z
cc
•
^f
CO
CD
DC
CO
i
*""
^t 00 CO
^t CO r-
[2 c) o
* CO T-
CM CO CO
^ ^ CM
1$
§e2
§»
0
0)
CO
o
00
1
o
OJ
CNI
CM
CO
Q
UJ
[^"
i>
DC
O
LJL
DC
LU
Q.
O
a
-------
100
O
(/>
z
a.
o
O
g
LL
liJ
o
<
CO
en
75)
E
u?
O
CD
CD
O
c\j
6>|/Buj '(OldH)
9CT1
-------
101
o
(0
0.
O
O
o
LU
*
DC
o
<
\
ID
CO
10
IT)
o
IT)
CO
CO
in
oj
i/6iu '(OldH) s6u|y 9< - 9 '
OO1
-------
102
Q
O
HI
CO
WCD
2P
1
§£
ob
zco
8§
LULL
o:
O
LL
O O
o o
o q
od cd
0
0
CO
CO
II II
o
o co
CO CO
CO
UJ
z
O cc
o
LL
co z.
LJJ
co m
w i^"
^ CO
< s
< H
CO CO
^ UJ
§ ^
CM *>
^ II
S «
—1 ggj*
5 o
o
§ s
o ^
^ ^
-------
103
CO
CO
UJ
cc
CD
O
cc
Q.
z
^
cc
o
z
a.
h-
co
o
CO
h-
z
UJ
cc
UJ
LL
LL
G
1
CO
cc
UJ
H
<
G
Z
O
cc
(D
Q
Z
<
UJ
0
<
LL
CC
D
CO
i
CO
UJ
H
>
|—
O
G
LU
LLJ
CC
O
DQ
CD
Z
CC
0
i-
z
o
^
1
(/)
h-
co
D
^^m
^
O
cc
LL
CO
LU
D
G
CO
UJ
CC
D
UJ
O
CC
UJ
a.
i
-------
104
MR. TELLIARD: Continuing on with
the petroleum part of the program, our next speaker is Greg
Douglas. Greg is presently Senior Research Scientist with
Battelle Ocean Sciences in Duxbury, Massachusetts, and his work
is again talking about fingerprinting petroleum hydrocarbons in
water.
-------
105
MR. DOUGLAS: The currently approved
analytical protocols that are traditionally recommended for oil-
spill response and remediation programs are often without the
necessary sensitivity or selectivity to address program
objectives and regulatory limits. As a result, we've developed
our own techniques for the quantitation and identification of
hydrocarbons in marine and terrestrial systems to better
understand the fate and transport of these materials in the
environment.
These methods either are in the marine chemistry
literature or are modifications of standard Environmental
Protection Agency (EPA) methods. In many cases, we are able to
enhance the quality, improve the sensitivity by almost three
orders of magnitude for many of the analytes, and provide
detailed information to help us to understand the fate and
transport of petroleum products in the environment.
The methods have been developed to support natural
resource damage assessment programs, so they have a very high
level of quality assurance and quality control.
The analytical objectives after an oil and gasoline
spill are to provide chemical information to evaluate the
presence or absence of the petroleum in the sample. We would
like the analysis to yield a reasonably quantitative measure of
total petroleum hydrocarbons, assist in product identification,
and provide the analytical information to evaluate the degree of
weathering. Such information helps us to understand the
potential toxicity and transport of the spilled product in the
environment.
The method that one selects may vary according to the
requirements and the particular situation. How quantitative does
it have to be? Can one rely on a semiquantitative procedure?
What kind of sensitivity is necessary? What matrix will one be
working with? Does one need to know the identity of the product,
or is that not important in this particular situation? Are there
any special-interest compounds that one wishes to track? Does
-------
106
one require any relative weathering information? Cost is always
a factor when considering methodologies, and so the required
turnaround time. The more sophisticated the analytical
procedures, the longer the time needed for completion. Also, is
the work to be used in litigation? That consideration could
affect the methods used and the type of deliverables that one
must produce.
A few of the petroleum hydrocarbon methods currently
in use include Method 418.1, modified for soils and for waters;
ultraviolet fluorescence analysis in marine systems and ground
waters; EPA Method 8270, a gas chromatography with mass
spectrometry (GC/MS) analysis used for a variety of compounds,
including the polynuclear aromatic hydrocarbons; a modified EPA
GC/MS methodology that we use in our own program; and a gas
chromatography with flame ionization detection (GC/FID) analysis,
which is a modification of EPA Method 8100.
I shall begin by discussing the applications and
limitations of one of the most common methods in use today: EPA
Method 418.1. The sensitivity of this method in water is
approximately 1 mg/L (1 ppm). In soils one can achieve a
detection limit of 1 to 10 mg/kg dry weight. The sensitivity of
this method is rather limited because the solubility of many of
the compounds in petroleum is significantly less than 1 ppm. The
procedure is not very quantitative. At best, it's
semiquantitative, and should be viewed as a screening tool. EPA
Method 418.1 doesn't really provide product identification unless
one modifies the method and uses a full-scan Infrared (IR)
Spectrometer and reviews the data. This method does offer a
rapid turnaround time, and it should be used primarily when data
are needed to make rapid field decisions. Method 418.1 involves
a freon extraction of the water or sediment and analysis on an IR
system at a very specific wavenumber, 2930 cm-1. The
calibration standard is composed of an n-hexadecane,
chlorobenzene, iso-octane mixture to replicate the molar
absorptivity of a standard petroleum product (fuel oil). Figure
-------
107
1 shows that the 2930-cm-l wavenumber region encompasses chiefly
the CH2stretch.
Unfortunately, this method is not very quantitative
when comparing different hydrocarbon products. For example, we
find that the analysis of a gasoline standard, yields only 49%
recovery because of the dominance of the CH2. Some of the wide-
band IR systems provide a little better quantitation in that
respect inasmuch as they overlap both wavelengths in many cases.
For fuel oil #6, there is a relative dominance of the
CH2stretch that results in about 115% recovery. For the creosote
oil, which is composed primarily of aromatic components whose
peaks are seen in Figure 1, there is a shift away from the CH2 to
the aromatic C-H stretch; only 16% is recovered in a creosote
standard.
The problem is that the molar absorptivity of the
defined calibration standard differs from that of many petroleum
products in the environment. This results in a wide variation in
response, depending on the product type.
The apparent percent recovery using IR, for a variety
of products is charted in Figure 2. The results range from 145%
to about 16% recovery, depending on the product.
In addition to these calibration problems, products
undergo physical and chemical weathering, which also changes the
relative amount of CH2 stretch. Figure 3 shows a GC/FID
chromatogram of a groundwater sample containing fuel oil #2. We
see the unresolved complex mixture and the resolved aliphatic
components. We also have the isoprenoids pristane and phytane,
which often are used as weathering indicators. When this
material is released into the soil, the more soluble aromatic
components of petroleum will be washed out and dissolve into
percolating rainfall. The peaks in the lower chromatogram of
Figure 4 are mostly aromatic compounds. As a result of the
different solubilities of the individual components of diesel, a
groundwater sample may be contaminated primarily with aromatic
-------
108
hydrocarbons, including naphthalenes and fluorenes and a variety
of other soluble aromatic components. If one were to analyze
water samples of the water-soluble fraction of petroleum, using
the IR method, one may get not-detected (ND) responses because
the method is insensitive to those aromatic components. We have
seen situations where one could smell the contamination, but the
IR method yielded only an ND. There is a slight absorbance of
aromatics because of the alkyl side chain associated with those
compounds, but the relative absorbance is low when compared to
the total mass.
Biodegradation and weathering of products create
further difficulties with the interpretation of results when
using Method 418.1. Figure 5 shows chromatograms of a crude oil
and of a sample of the same oil after degradation. A series of
aliphatic hydrocarbons is evident on the upper chromatogram of
the crude oil. They comprise a much higher percentage of the
sample than they do in the lower chromatogram of the
environmental sample. If one were to compare the results by IR
for the same mass of material, there would be much less mass
detected for the lower sample because the specific IR absorbing
species had been degraded to a large degree.
In addition, natural hydrocarbon interferences will
not be removed by the silica cleanup step because plants
produce n-C2S,n-C27, n-C29, and n-C31 plant waxes. There have been
many situations near peat bog areas, swamps, and sewage systems
where one obtains high numbers for IR measurements because of
background interferences.
In summary, there are a number of problems associated
with quantifying petroleum hydrobarbons with the EPA Method
418.1. Analytical results obtained by using it should be
interpreted with an understanding of the applications and the
limitations of the technique.
Figure 6 illustrates a protocol that we follow. Most
of the methods either are modifications of EPA protocols to
-------
109
improve selectivity and sensitivity or they are reported in the
marine environmental literature.
In this example, we have a sediment sample (but it
could be extracted water, as well, following the EPA Method
3510). The sample is spiked with appropriate surrogate compounds
(Table 1) and extracted following EPA protocols for sediments
(and for waters, using a separatory funnel, liquid-liquid
extraction). We then filter the extract, and concentrate it
using Kuderna-Danish (KD) apparatus or rotary evaporation
techniques. An aliquot of the extract is then weighed to obtain
a total extractable weight, prior to column cleanup. We then
process the extract through an alumina column cleanup, slightly
modifying standard EPA Method 3610 by combining the Fl saturated
hydrocarbon and F2 aromatic hydrocarbon fractions to obtain a
total recoverable hydrocarbon fraction. Next, we again
concentrate the extract to a small volume. We then weigh an
aliquot of the extract to get pre- and post-column total oil
values. We calculate the percent polars, {(precolumn oil-
postcolumn oil)/precolumn oil}, which can be very useful in
evaluating the degree of biodegradation. Then, we add our
internal standards and analyze the sample extract by GC/FID
(modified EPA Method 8100). The method is calibrated very
specifically, using a suite of n-C8 to n-C36 alkanes with alkane
response factors relative to the internal standard androstane.
Total recoverable petroleum hydrocarbon is calculated
from the total GC/FID baseline corrected area and the average
response factor (RF) of the n-C8 to n-C36. With care, one can
obtain RFs across that whole range that vary by no more than 10%,
even at the heavy end (i.e., greater than about n-C27) , where one
can run into some mass discrimination problems. All in all, the
detector responses are uniform, and one can, in fact, quantify
the GC/FID-detectable components with reasonable accuracy.
If the sample was extremely weathered, we then examine
the aromatic compounds by using GC/MS. We also look at another
-------
110
class of compounds that many may not be aware of: the
triterpanes, steranes, and diterpanes. These compounds are very
specific to various sources of petroleum. They are resistant to
degradation, and they have specific fingerprints. As we examine
the more weathered oils, those fingerprints can help us to
identify the kinds of products that are present. Because
steranes, triterpanes, and diterpanes are resistant to
degradation, one can use them as internal normalizers to
calculate the amount of oil lost after a spill.
The analytes determined via GC/FID and GC/MS
analyses are shown in Table 1. For analyzing petroleum products,
this list is more selective than is Method 8270. Rather than
just look at, for example, the 16 priority pollutant PAHs, of
which only a few are present in petroleum, we have expanded the
analyte list for PAHs to include the alkylated PAHs and
dibenzothiophenes because the ratios of the sulphur compounds to
the nonsulphur aromatics are also characteristic of specific
sources. We still analyze for priority pollutant PAHs, including
the pyrogenic PAHs, benzapyrene, fluoranthene, and pyrene. These
PAHs help to separate pyrogenic sources from petroleum sources of
PAHs. In marine systems, we often find evidence of a number of
pyrogenic sources, including atmospheric deposition of combustion
products into the bays and harbors.
We also study the full suite of alkanes from n-C7 up to
n-C36. We also examine the isoprenoid hydrocarbons, pristane arid
phytane. The ratio of n-C17 to pristane or n-Ci8 to phytane may be
used as an indicator of weathering because the organisms in
marine systems soils and ground waters tend to preferentially
degrade the straight-chained alkanes relative to the isoprenoicls.
The detection limits for total petroleum hydrocarbons
insediment is around 1 to 10 mg/kg (parts per million). Alkane
detection limits are typically 0.1 mg/kg, and the PAH detection
limits are 1 to 10 mg/kg.
In water samples, we can detect PAHs down to about 1 to
-------
Ill
10 parts per trillion. Such low detection limits are useful for
plume tracking because one obtains much information in advance of
the arrival of the heavily contaminated portion of a plume.
Also, one gets data on a lot more analytes, which allows one to
start fingerprinting more effectively by having a full suite of
compounds.
In addition to these components, we use the triterpanes
and the steranes as conservative oil indicators, mainly in the
heavier oils, fuel #4, fuel #6, lubricating oils, and crude oils.
There is also a suite of diterpanes in diesel fuels. I'm also
looking into jet fuels to see if we can find those components
there as well. Again, these are components that originally were
biologically derived, are very resistant to degradation, and
have very specific fingerprints.
Table 2 lists products that we have analyzed and
included in our computer library. We have analyzed the products
by using, GC/FID or GC/MS and measured their biomarker
components. We are using principal-component analysis and ratio
analysis to try to cluster various classes of components so that
even under weathered conditions we can still identify the
product.
Figure 7 shows two GC/FID chromatograms of gasoline and
its water-soluble fraction. Again, the quantification is based
on the total area response relative to an internal standard. We
also quantify each individual alkane.
Figures 8 and 9 also show chromatograms of products and
their water-soluble fractions. Our standards include not only
the pure products but also degraded products and water-soluble
fractions, especially for groundwater studies. To get good
correlation between sample results and possible contamination
sources, one looks at the components in the water-soluble
fraction of the potential sources.
We take the data that we generate and enter them into
an electronic spreadsheet and plot the distributions of the
individual n-alkanes, pristane, and phytanes. Figures 10 through
-------
112
15 illustrate the results obtained for selected petroleum
products.
One problem to be aware of when working with GC/FID
analysis of petroleum hydrocarbons is one of mass discrimination.
One can actually lose a large fraction of the heavier components
of a sample if the capillary column is not appropriately
positioned and the injection liner is not of the proper diameter.
Mass discrimination should be monitored routinely.
Figure 16 shows two superimposed chromatograms of the
same oil run twice: one where there is a problem with mass
discrimination and one where there isn't. As much as 50% of the
product could be lost in some cases of mass discrimination. To
monitor mass discrimination, we examine the n-C36 response factor
relative to the n-C21 response factor. They should be within a
few percent of each other. In many cases, we have seen the ratio
of n-C36 to n-C20 response factors near 0.1 or 0.2, indicating that
there is a severe problem that could result in false negatives.
Figure 17 depicts an example of an alkane standard that
we run. It shows the response of each individual alkane standard
of the same concentration from n-C8 to n-C36. The two largest
peaks in the chromatogram are quantitation internal standards
(QIS) and surrogate internal standard (SIS) compounds. The
response should be linear across the full alkane range as it is
here. The peaks should not drop off after n-C2S, which
unfortunately is common in much of the GC/FID work that we have
reviewed.
One would use this procedure to look at an actual
sample. One would examine the distribution of the unresolved
complex mixture, where it begins and where it ends. And one also
would examine the distribution of these resolved compounds
because pristane and phytane (Figure 3) are still there even
after fairly significant degradation.
As shown in the lower chromatogram of Figure 5, the
straight-chained alkanes have been almost completely degraded in
-------
113
this sample.
Comparing the lower chromatogram to the upper one is
not fruitful because of the extent of degradation that has taken
place. As weathering proceeds, the qualitative accuracy of the
measurement decreases. In this case, one would go to the next
level, using aromatic distributions and biomarkers.
The kind of problems that can occur when using the
standard EPA protocols are illustrated in Figure 18, which shows
the GC/FID output for diesel and JP5 and their water-soluble
fractions. Assume, for example, that there has been an oil
spill, and that one is interested in the potential toxicity of
the contaminated water to marine life. You would collect the
water sample, send it to the analytical laboratory and request a
Method 8270 analysis. What you might get back is depicted in
Figure 19. You might get a plot of the PAHs to include in some
cases, the naphthalene, perhaps C, naphthalene and perhaps
phenanthrene. The detection limit for this method is about 10
parts per billion, and does not measure the alkylated PAHs. If
one uses the modified GC/MS approach for the same sample (Figure
20), one obtains a full distribution of parent and alkylated PAHs
that can be used to interpret the data and determine the source
of the oil spill and the transport, fate, and potential effects
of it.
Using the modified GC/MS procedure (Figure 6), one can
analyze a whole spectrum of products (Figures 21 through 26),
including creosote, which has a very distinct distribution where
the major peaks are the parent PAHs. The PAH concentration
within a homologous series in creosote decreases rapidly as
alkylation increases.
The used of petroleum-specific GC/FID and GC/MS methods
can improve our understanding of the proceses that influence oil
degradation in the environment. Biodegradation rates of
hydrocarbons are dependent on the type of bacteria, presence of
limiting nutrients, temperature, and types of hydrocarbons.
-------
114
Bacteria generally degrade hydrocarbons according to the
following sequence: n-alkanes>branched alkanes>aromatic
hydrocarbons>cyclic alkanes. If a conservative compound in the
spilled oil can be identified, then, as degradation of the oil
proceeds, the concentration of that compound will increase. This
increase can be compared to the source oil and the percent
depletion of the oil can be estimated according to the following
equation: % Total Oil Depletion = (1-Co/Cj) xlOO, where C, is the
conservative compound concentration in the degraded oil and C0 is
the conservative compound concentration in the source oil.
Individual analyte depletion can be estimated by the
equation: % Analyte Depletion = {1 - (A^A,,) x (Co/Cj) } x 100,
where A! is the analyte concentration in the degraded oil and Afl
is the analyte concentration in the source oil.
If the source oil cannot be identified at the site, C0
can be substituted with the conservative compound concentration
in the oil prior to application of the remediation agents (water,
bacteria, nutrients, oxygen).
There are technical advantages to using the above
approach rather than traditional methods to estimate site
remediation. Because of spatial variability, the traditional
attempts to evaluate the effectiveness of oil spill remediation
often require large numbers of samples to determine the mass loss
with reasonable precision. The use of an internal chemical
indicator reduces spatial variability, thereby reducing the
number of samples required to monitor effectiveness of the
remediation option. By adjusting the operational remediation
parameters, it may be possible to tune the selected remediation
approach to maximize the degradation of the more toxic
components.
Butler et al. (1991) presented a paper at the In Situ
and on-site Bioreclamation symposium in San Diego using the above
relationship. In their paper, they assumed 17 (H) , 21B (H) -
C30Hopane(hopane) was the most conservative analyte measured in
-------
115
their degraded crude oil samples. Figure 27 from Butler et al.
(1991) shows the GC/FID chromatograms from that study of oiled
beach sediment exhibiting increasing degrees of degradation.
Butler et al. (1991) then examined traditional measures of
degradation for four weathered samples (Figure 28a). Their
results demonstrated that, in the four degraded samples, the
traditional weathering indicators were not very useful. They
then examined the percent total oil depletion, n-Ci8, and phytane
depletion relative to hopane. The results presented in Figure 28
indicate that as weathering proceeded percent total oil depletion
increased to 45%-50% relative to the 30% depleted NSC reference
oil. The n-C18 and phytane demonstrated substantial percent
depletion as the oil degradation increased. These data suggest
that the use of the n-C18/phytane ratio ceases to be a good
weathering indicator in moderately to heavily degraded oil
samples.
In conclusion, the GC/FID and GC/MS methods presented
today provide the analyte selectivity and sensitivity required to
evaluate the fate and transport of spilled petroleum products in
environmental samples. The results from these analyses can be
compared to other large environmental databases such as from NOAA
Mussel Watch program or to established and proposed sediment and
water quality criteria to evaluate potential environmental
effects and relative levels of contamination. Finally, the use
of conservative chemical indicators, such as hopane to estimate
oil degradation may improve our ability to evaluate the
effectiveness of remediation options after an oil spill.
I would like to thank Dr. Rojer Prince of Exxon
Research and Engineering (Clinton, NJ) for his review of this
work. This work was supported by the Battelle Research and
Development Program.
-------
116
QUESTION AND ANSWER SESSION
MR. CASTLE: Do you utilize the
pristane/phytane ratios, and, if you do, have you found that, as
the weathering increases, the pristane/phytane ratios remain
stable or do they alter?
MR. DOUGLAS:They alter somewhat in
heavily degraded samples. I can't say that they degrade at equal
rates, but they do vary enough where it's hard to use them after
they degrade beyond Sample D (in Figure 30), say, which would be
difficult to use.
MR. CASTLE: That's what we found
also. Thank you.
In your simultaneous ion monitoring work, typically how
many ions would you follow?
MR. DOUGLAS:We're looking at
between 6 and 15 ions.
MR. WESTON: Would you attempt to
optimize your SIMS sensitivity, for example, by lowering the mass
spectral resolution? Would you attempt to optimize sensitivity
specific to the SIMS method?
MR. DOUGLAS: We tried to reduce
the number of ions that we had to look at, which would improve
sensitivity. We calibrated it in favor of, say, the midrange
ions. Our sensitivity was good enough to detect about 100 ng of
analyte per milliliter of extract.
MR. WESTON: Right. Thank you.
MR. STANKO: In one of your slides;,
you showed that your detection limit with the modified Method
8270 was somewhere in the range of 1 to 10 parts per trillion.
MR. DOUGLAS:Right.
MR. STANKO: I'd like to ask you,
what did you do to Method 8270, which is essentially a part per
billion, at best, to get it down to the parts per trillion level?
MR. DOUGLAS: Basically, we were
able to increase sensitivity by using single ion monitoring
-------
117
(SIM), concentrating the final extract to 250 ml and increasing
sample size to 2 L.
MR. STANKO: Thank you.
MR. TELLIARD: Thanks a lot, Greg.
I'd like to thank this morning's speakers for their
efforts.
It's break time. Please get your goodies and get back
in here in about 15 minutes. Thank you for your attention.
-------
118
TABLE 1. PETROLEUM FINGERPRINTING ANALYTE LIST
Polycyclic Aromatic
Hydrocarbons
naphthalene
Cl-naphthalenes
C2-naphthalenes
CS-naphthalenes
C4-naphthalenes
acenaphthylene
acenaphthene
fluorene
Cl-fluorenes
C2-fluorenes
C3-fluorenes
phenanthrene
anthracene
C1 -phenanthrenes
/anthracene
C2-phenanthrenes
/anthracenes
C3-phenanthrenes
/anthracenes
C4-phenanthrenes
/anthracenes
dibenzothiophene
C1 -d ibenzoth iophenes
C2-d ibenzoth iophenes
C3-dibenzothiophenes
fluoranthene
pyrene
Cl-fluoranthenes
/pyrenes
benzo[a]an thracene
chrysene
Cl-chrysene
C2-chrysene
C3-chrysene
C4-chrysene
benzo[6]fluoranthene
benzo[Jt] fluoranthene
benzo[a]pyrene
dibenzo[a,/t]anthracene
benzofo h, i\ pery lene
indeno[l,2,3-aflpyrene
Aliphatic Hydrocarbons
Normal Alkanes
n-C7
n-Cg
n-C9
n-C10
n-Cu
n-C12
n-C13
n-C14
n-C15
n-C16
n-C17
n-C18
n-C19
TARGET REPORTING LIMITS
n-C21
n-C27
n-C31
n-C32
n-C33
n-C35
n-CM
Waters
Total PHC = 50.0 /zg/L
Alkanes = 0.2 /xg/L
PAHs = 0.01 /xg/L
Oils
Sediments
Total PHC = 10 rag/kg
Alkanes = 0.1 mg/kg
PAH's = 0.0005 mg/kg
PHC = 80,000 mg/kg
Alkanes = 320 mg/kg
PAHs =16 mg/kg
Surrogate Compounds
PHC =
PAH =
Ortho terphenyl
dg naphthalene
d,n fluorene
d12 chrysene
Internal Standard Compounds
PHC = 5a Androstane
PAH = d,0 acenaphthene
d,0 phenanthrene
d,2 benzo[a]pyrene
Isoprenoid Hydrocarbons
1380
1470
1650
pristane
phytane
The highlighted PAH compounds represent the 16 priority pollutants list.
-------
119
TABLE 2. PETROLEUM FINGERPRINT PRODUCT LIST
PRODUCT
1. Gasoline
2. White Gasoline
3. Paint Thinner
4. Turpentine
5. Jet A
6. JP4
7. Kerosene
8. Diesel #1
9. Diesel 12
10. Fuel Oil #2
11. Fuel Oil #4
12. Fuel Oil #6
13. Lubricating Oils
14. Syltherm
15. Transformer Oils
16. Creosote Oil
17. Coal Tar
18. Prudhoe Bay Crude Oil
19. NBS 1580 Shale Oil
20. NBS 1582 Crude Oil
21. EPA Bunker Oil
22. Gunk Super Oil
23. NBS Coal
24. NBS Fly Ash
25. Combustion Soot
26. Aviation Gasoline
27. Mineral Oil
28. NBS Residual Fuel Oil
-------
120
Total Recoverable Hydrocarbons By Infrared
Analysis - EPA Method 418.1
Gasoline
•49%
Recovery
• Low CH2
• High CH3
Calibration Standard
3200 2800 2600
Wavenumbers
Asymetric
CH3
Symetric
CH3
CH2
(In Phase H)
3200 2800 2600
Wavenumbers
Fuel Oil f 6
Creosote Oil
• 115%
Recovery
• High CH2
• 16%
Recovery
• High
Aromatic
CH
3200 2800 2600
Wavenumbers
3200 2800 2600
Wavenumbers
Figure 1. Infrared absorbance patterns of a calibration oil, gasoline, fuel oil #6, and
creosote vs wavenumber (2600 to 3200 cm"1). These samples were prepared in
freon and analyzed according to EPA Method 418.1 by using a quantification
absorbance of 2930 cm"1.
-------
121
CO
CO
if
cc>
0
£
LU
\\\\\\\\\
1 1 1 1 1 1 1 1 1
ooooooooc
.« J
8.J
%
5
BO
-------
122
pjfputlS |tu-i»lu| ^4 "64
It
0-N «E'JJ
'0-N
o
0)
tr
>
C
S
IB
CN
'S
.3
T3
U
co
C
8
i>
"a,
CO
~a
c
2
BO
03
(4—1
O
6
05
E
2
a
h— 4
s
o
CO
I
-------
123
Gas Chromatographic Analysis of JP5 and
Diesel Oil, Water Soluble Components
Diesel Fuel
JP5
J&-
s
s
*,
CO CO
t— 1 H
CO O
Diesel Fuel Water Soluble Fraction
JP5 Water Soluble Fraction
s •
CO CO
Elwt ion t VTW
Figure 4. GC/FID chromatograms of diesel and JP5 fuel oils and their respective water-
soluble fractions. The water-soluble fractions were prepared by floating the
products on 1 L of distilled water in a separately funnel for 5 days, after which
time the water-soluble fraction (WSF) was drained from the bottom. This WSF
is composed primarily of 2- to 3-ring aromatic hydrocarbons. Using WSF
standards for fingerprinting improves water sample product-identification
accuracy.
-------
124
Standard
V)
»-t
o
CM
TH
u
c
V
l
I
MM*
Sample
Figure 5. GC/FTO chromatograms of fresh and degraded crude oil. The quantitation
internal standard (QIS) is ortho-terphenyl.
-------
125
rrtSi
Add Sumgm Compounds
Extract Sample with 1:1 Methytene
Chloride/Acetone EPA Method 3550
> filter Extract
Concentrate Extract
Kudema-Dansh Technique
Alumina Column Cleanup
EPA Method 3610
Concentrate Extract
Kudema-Danish Technique to 0 5 ml
. w«gh Extract
. Add mwmal Sandarc*
OC/FID ANALYSIS
Prudhoe Bey Crude. Detefence OH - 30% Depleted
Ortho-terphenyl
Surrogate * 5-a Androstane
Phy,ane i ~ Internal Standard
RETENTION TIME
QC/MS SIM ANALYSIS
POLYCYCUC AROMATIC HYDROCAHBQN8 (PAH)
QC/MS SIM HOPANE ANALYSIS
Prudhoe B«y Crude. Reference Oil - 30% Depleted
Ill 1
iljL l
pLMMUr
u
V*
-C3
1
k'J,
0170(H)21P^
- C3017P(H
Internal S
JwuLUJv
RETENTION TIME
Figure 6. Battelle Fingerprinting Program method summary. For water-sample analysis,
EPA Method 3510 would be used to extract the samples.
-------
o
Is-
o
"o
126
Gasoline
co
M
CO
CO
I—I
o
T*
1O
EI ut i oo T im»
2O
AO
1 I '
SO
Minut«s
Gasoline Soluble Fraction
i
r
i
(Jill
CO
I—I
CO
CO
I—I
CJ
10
20
Efution Tim*
3O
4O
50
Mi nut«i
Figure 7. GC/FID chromatograms of gasoline and its water-soluble fraction. The surrogate
internal standard (SIS) is ortho-terphenyl and the quantitation internal standard
(QIS) is 5-a androstane.
-------
127
NBS Crude Oil
co
n
CO
co
i— i
a
co-
co-
o.
a
o
«
C I
o
10
20
3O
50
Elut ion T ime
Mi nutes
NBS Crude Oil Soluble Fraction
•
•>
o'
ex
o
J
JJL
ii^iiiuij-
CO
3
CO
J*
10
Elut ion T ime
20
4O
SO
Mi nute«
Figure 8. GC/FTD chromatograms of National Bureau of Standards (NBS) crude oil
standard and its water-soluble fraction.
-------
128
Fuel Oil #2
E1u t i on T i
50
Mi nutes
Fuel Oil #2 Soluble Fraction
Ilk
iu
20
3O
*-!—*•
4O
cn
• i •
5O
E I ut i on T ime
Mi nutes
Figure 9. GC/FID chromatograms of fuel oil #2 and its water-soluble fraction.
-------
129
w
(cpunnoqjj
HIO-0>I/OW NOI1VU1N30NOO 3NV>nV
I
x:
•O
+~t
co
c
.g
.0
'C
-*^
en
•3
T3
G
co
G
•B
2
1
§
to
-------
130
o
LLJ
O
D
cr
0)0
ffl
>;
=!<
LU
PQ
I
I
u.
ca
I
I
2
cd
W3
<*-*
o
c
.o
•a
c
ea
c
.0
I
§
3
oo
1IO-EM/9IAI NOIlVyiN3ONOO 3NV>«1V
-------
131
CM
CO-J
coo
0!2
tew
WIU
8
-OR
-8°
vyxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
'VX^^^^XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
XX/XxxXXX/XxxXXXxXX/XXxX///xxXXxXxxXXXxxxX/X//xxxxxX////X/x
'XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX.
/•xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
xxxXXXXXXXXXXXXXXXxxxXXXXXXXXX
TSSSSSSSSSSSSSfSSJ
OJ
u
10
u
n
O
^
^
U
u
(N
a
_G
en
I
I
CO
3
•t-1
03
on
.0
•
T3
_o
4— »
05
a
§
bO
E
NOI1V«JLN3ONOO 3NV>nV
-------
132
0)
a: en
I/9W NOIlVyiNBONOO 3NWIV
-------
133
CO
tfc
coo
Will
'S////////Z6Q 7j o
BSJ^E^Bflfti g
8
0)
.3
I
TD
U
C
.o
•4—*
3
.0
°
T3
C
M
C
§
(3
1)
>-c
(SO
E
1IO-0>
-------
134
UJ
COO
coco
UJ
T5
c
03
.2
03
-t~i
«
?
a
1IO-0>I/OW NOIIVU1N30NOO 3NVX1V
-------
135
tr
o
en
t-H
a
en
m
u
4J
3
— O ^
03
_ O
in
_ o
ttl
3s
3i'
asuodsau
1
01
E
£
'5
a
3
oL
ca
S
£
'?
CS
a
ca
"O
•3
U
CS
fl
2
es
c
9i
^ s
Q 2
O E
-------
136
9fiO
980
(98'et)
280
(91-66)
Z20
920-
(80-88) 920 •
(ez'ie) t-20 •
(ee-oe) 820
(98Y2) 1-20 -
020
dlO
810
(LZ'02)9ueisud
(69'02) ZLO
(69-81) 9LO
910
U9>I.) t-LO
(8€'2l) CLO
(LO'Ot) 210
(89'Z) 110
i^i'i^i'igi'-'^'-'^'-'ai'^'^'^-'^''-
«
0)
CM
_O
r
C
o
» c
^ >
CO
« U
•g B
™
H.'S
3g
en
u u
S
»
T3
G>0 W5
o «
£8 c
B 2
O -3
t—* Cy
S 1
•o
T3 tn
Ui CQ
03 C
0) 'TT-
C 2
ca c
-
c
a «
10 CO
T— 1
CO O
s °
ca c
S a
O D
1-1 en
'
Q-B
HH "2
fc ^
U S
r^ 4>
&0
ssuodssy
-------
137
Gas Chromatographic Analysis of JP5 and
Diesel Oil, Water Soluble Components
Diesel Fuel
JP5
Diesel Fuel Water Soluble Fraction
JP5 Water Soluble Fraction
1JI
Figure 18. GC/FID chromatograms of diesel and JP5 fuel oils and their respective water-
soluble fractions. The water-soluble fractions were prepared by floating the
products on 1 L of distilled water in a separatory funnel for 5 days, after which
time the water-soluble fraction (WSF) was drained from the bottom. This WSF
is composed primarily of 2- to 3-ring aromatic hydrocarbons. The use of WSF
standards for fingerprinting improves water sample product-identification
accuracy.
-------
138
o
Q 00
•a
o
LO
LU
o
o
o
en
i i
o o
03 r^
o
m
UJ
UJ
Q
D
I
s/j -a ao
n O .£
- £ 5
DUO
a e &
KJ XI " .
s2
a 5
1/3 Q . .
2 tl -8
'— e: •—
,= 03 w
S .P a ^
0 CO
\
u "n -^ §
-1'3i8
,_ J VH fc.
C < >»«„
.0 AH S O
~ 3 CJ
* C X3 -o —•
=3 1 1 1
a 1
2
a
E
I/OH NOIlVdiNBONOO HVd
-------
139
•0 T3
_Q CO
-J
O
LL.
Q
ii
,3 CM
LL CO
"O "O
CO O
if
T,2
10 UJ
o
Q
o
CM
0
t^
LU
5
D
I
™ US
o I
X! *"*
I -
<->-
o «
S g
•a «
52 s«
- 'fs
3 C3
8.3
8
5T-S
o 2
S-SJ
III
jj -g 2
liS
OT -^s *!?
I/OH NOIlVdlN30NOO HVd
*s • rt
O C *rf
^i M
CO D
-ii Z-i
•Bs I
3 C B
-S o>
PL, .5 ^2
I
00
E
-------
140
(/> _l
Is
^ 1—
Zuj
Q
^^T w^\
z<
^^1 ^^™
^•^* ^^1
d<
QQ-
:;:::!
! i i i i i i
^
\'SSSSS/S/SS//SSS/SS/SSSSSSSSSSSSSSS/SSSSSS/SSSSSSSSSSSSSSS/S*
if^Hfk'SSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSSfSSSSfSSSSS*
± 1 1 1 1 1 L
1IO O>I/OW NOIlVyiNBONOO HVd
0 ^
:s"£
-0° CO
M U ^\
vj ^ ^rf
aJJ
"S" o
jj§
:a8|
:a^
^^ .^^^^^^rr^^^H
k^
o
a
•«^
4>
e
•^^
en
.o
CO
k-
§
•o
CO
o
so
-------
141
CO
lH
o
c
t-l
4-»
e
8
8
•o
C3
I
3
.£5
(N
1IO-O>I/OW NOIlVaiNBONOO HVd
-------
142
IARD ANALYSIS
SIS-FUEL OIL #4
HZ
dx
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
o
Q.
a
c
.O
CO
1
8
T3
CO
C
O
CN
eso
1IO-0>I/OW NOI1VU1N3ONOO HVd
-------
143
CO
O
_l
lU
OCC0
xxxxxxxxxxxxx/xxxxxxxxxxxxxxxx,
^XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXx
VXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX/
xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
'XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX
en
0)
C
o
I
•a
a
nj
C
CN
t>
1IO-0>I/OW NOI1VU1N30NOO HVd
-------
144
O
in
i ANALYSIS
OE BAY CRUDE
• ^1 M^H
or
JJfJ)J777777777JfJJJJi
»»»»»»»»»»»»»»»»»»»»»»»»»»
^^^^^^^•^^^^^^•'xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxy
§ I § i §
IIO-O^/OW NOIlVaiN30NOO HVd
9k «**•
0 T-
» PY BA C1 C3 BB BP DAh
FL FP C C2 C4 BK IP B
rDROCARBONS (PA»
: a =
-2«g
"""SI o
t^ ^* *
u. «J
r*2 f ^
Ks >
•^tVZV
(U
•o
n
•a
3
.o
2
u
ca
o
(A
G
3
£
-------
145
0)0
£
5fc
<(0
ZS
(O
//////X/X//////X//XX////X///X//X/^X/Xyr//X////X//XX//XXXX
•—
s
0
kH
u
o
ncen
-a
c
CO
C
.g
•*—»
X3
VC5
CN
3
00
£
1IO-OX/OVM NOI1VH1N30NOO HVd
-------
146
RETENTION TIME
RETENTION TIME
RETENTION TIME
SAMPLE A
SAMPLE B
SAMPLE C
SAMPLE D
RETENTION TME
GC/FTD Chromatograms from Selected Sediment Samples.
Figure 27. GC/FID chromatograms of a crude oil samples that exhibit increasing amounts of oil
degradation.
-------
147
Traditional Measures of Degradation
<
tr
SAMPLE A SAMPLE B SAMPLE C SAMPLE 0
INCREASING DEGRADATION > »
C17/PRISTANE
C18/PHYTANE
IAROWEATHRATIO
% Depletion Based on Hopane
o
K
LU
a
LU
o
SAMPLE A SAMPLE B SAMPLE C SAMPLE D
INCREASING DEGRADATION » >
*18
i PHYTANE
TOTAL HC
Figure 28. (a) Traditional measures of degradation vs increased oil degradation.
(b) Percent depletion of n-C18, phytane, and total hydrocarbons (HC) relative to
17a(H),21j8(H)-C» Hopane.
-------
148
MR. TELLIARD: Can we get the
folks in from the outside with their coffee and their toast or
whatever they're bringing?
It kind of reminds me of the story about the Aggie that
was on the boat and this guy from the University of Texas and
another fellow from the University of Houston and the ship ran
aground and sank and the three men were left in a lifeboat for
months on end. Finally, a small bottle appeared in the water eind
they opened it and a genie came forth and granted them for
freeing him one wish each. So the fellow from the University of
Texas said, I can't wait. You just send me right back there to
Austin in them hills and I'm going to be happy. And (snap) he's
gone.
The other fellow said, I'd like to go back to Houston
and have a little crab and sit around and eat some. (Snap) He's
gone.
And there sat the Aggie, bobbing in the lifeboat all by
himself, looking around. And the genie said, what would you
wish?
He said, well, actually, I'd like to have my friends
back.
Our first speaker in the pesticide section is Gary
Jackson. Gary is presently associated with Colorado State
University in a company called Support Systems.
Gary is going to talk about a novel approach to
extraction and analysis of chlorophenoxy-acid herbicides. I
think novel is kind of neat. I've never heard of a novel
approach.
-------
149
ANALYSIS OP CHLOROPHENOXY-ACID HERBICIDES
IN SOIL AND WATER
Gary B. Jackson* and Steven M. Workman
Department of Agronomy, Colorado State University, Fort Collins, Colorado
80523
ABSTRACT
Due to the chemical nature and behavior of chlorophenoxy-
acid herbicides in soil and water, and the traditional use of
non-specific or semi-specific detection systems, currently
practiced analytical methods often fail to provide accurate and
precise measurements for making environmental and toxicological
assessments. As a result of the phenoxy-acid group -OC^H^COOH,
phenoxy-acid herbicides of 4 C or less are easily subjected to
alkaline and acid hydrolysis. In order to partition the free
acid form of the herbicides from soil and water to organic
solvent, the entire soil matrix is required to be at a pH of less
than 2. This often becomes difficult to achieve, especially in
calcareous soils where CaCO3 buffers the addition of acid. Once
the herbicides have been favorably partitioned to organic
solvent, difficulty in detection and identification may arise
from non-analyte interferences that are electrophilic and respond
to electron capture detectors. Often, hydrolysis cleanup
procedures are required to remove or minimize interferences.
However, these cleanup procedures have been shown to have an
unfavorable effect on the accuracy and precision of the target-
analyte measurement.
The objective of this study was to develop a new extraction
and analysis procedure for chlorophenoxy-acid herbicides in soil
that eliminates acidifying the soil, minimizes hydrolysis cleanup
steps, eliminates the use of ethyl ether as an extraction solvent
-------
150
in exchange for a less water soluble, non-flammable solvent,
maximizes the accuracy and precision of the measurement, and uses
a specific detector that does not respond to traditional electron
capture detector interferences.
In order to meet the above objective, chlorophenoxy-acid
herbicides spiked in soil were subjected to alkaline hydrolysis
and partitioned to the aqueous phase of a soil suspension. Upon
flocculation of the soil suspension, an aliquot of the supernate
Vas acidified and extracted with dichloromethane. The extracted
herbicides were methylated with diazomethane and analyzed by gas
chromatography using a halide-specific electrolytic conductivity
detector. Reported accuracy and precision results are equal or
superior to those reported in U.S. EPA analytical methods.
INTRODUCTION
Due to the chemical nature and behavior of chlorophenoxy-
acid herbicides in soil and water, and the use of non-specific or
semi-specific detection systems, currently practiced analytical
methods (1,2,3,4,5) often fail to provide accurate and precise
results for making environmental and toxicological assessments.
First, chlorophenoxy-acid herbicides fall under the general
structure of ROC^H^COOH, where R represents a chlorinated
benzene ring. For example, 2,4-D has the following structure:
OCH COOH
Cl
As a result of the phenoxy-acid group -OCnHn+2COOH, phenoxy-
i
acid herbicide compounds are easily subjected to alkaline and
acid hydrolysis.
-------
151
Alkaline hydrolysis
R C~~ OH
I
W
R - C
S c r ongly
nucleophilic
1
OH
OH
RCOO
Acidic hydrolysis
R— c
w
R— C
OH
I
R— C— OH.
I
W
Weakly
nucleophilie
R— C
OH
H : W H
In alkaline hydrolysis, the hydroxide ion acts as a strong
nucleophilic reagent and replaces the leaving group,-W (-OH, -Cl, -
OOCR, -NH2, -OR1. Whereas in acid hydrolysis, the hydrogen ion
attaches itself to carbonyl oxygen and renders itself vulnerable to
attack by the weak nucleophilic reagent, water. As expected, salts
formed during alkaline hydrolysis are soluble in water but insoluble
in non-polar solvents. The free acids of four C or less, formed
-------
152
during acid hydrolysis, are both soluble in water and in organic
solvent. However, only with very polar solvents or when the pH is
less than 2, will the free acid herbicides favor the organic solvent.
OH
RCOOH 1
Because chlorophenoxy-acid herbicides act as strong organic
acids, standard methods for the extraction of herbicides in soils are.
often inadequate, tedious, and time-consuming. For example, EPA
Method SW-846; 8150, requires sufficient HCL be added to 50 g of soil
to obtain-a pH of less than 2. Continuous monitoring of the pH for 15
min is required to insure the pH remains below 2. Achieving a pH of
less than 2 is often difficult, especially in calcareous soils where
CaCOj can play a significant role in buffering the addition of HCL.
After the pH is stabilized, three 20-rain sequential extractions with
acetone and diethyl ether, respectively, are required. The solvent
extracts are combined with a fourth extraction using only diethyl
ether. The solvent extract is then subjected to alkaline hydrolysis
by the addition of reagent water and KOH. Evaporation of the diethyl
ether on a water bath is required for 90 min. The alkaline solution
then undergoes further diethyl ether extraction to remove non-ionic
neutral and basic organic substrates.
At this point in the extraction process, the herbicides have
undergone acid hydrolysis, solvent extraction, and alkaline
hydrolysis. The next step is to convert the salts of the acids to the
free-acid form, partition the herbicides into diethyl ether, and
remove any residual water with the addition of acidified anhydrous
-------
153
Na2S04. The solvent extract is then concentrated and esterified with
diazomethane (6,7). Finally, after removal of residual diazomethane,
the extract is ready for GC analysis employing an electron capture
detector (BCD).
The use of an ECD in the analysis of chlorophenoxy-acid
herbicides may also lead to mis-identification and/or poor integration
of the chromatographic peaks. Because the ECD responds to molecules
that are deficient of electrons in their outer most energy shell, a
wide variety of compounds can be detected. For example, compounds
that contain atoms of Cl, Br, I, F, N, S, P, and 0 respond to the ECD.
Compounds that contain pi bonds such as aromatic rings, also respond
to the ECD. The most common interferences in the analysis of
herbicides are organic acids that include chlorinated phenols and
phthalate acid esters. However, with highly contaminated waste soils,
many different interferences can be encountered. Additionally, ECDs
are sensitive to GC oven temperature programming; Thus, causing
chromatographic base-line rise that could interfere with peak
integration.
The purpose of this study was to develop a new extraction and
analysis method for chlorophenoxy-acid herbicides in soils and water
that would meet the following criteria: 1) simple to use in that the
method minimizes hydrolysis and clean-up steps; 2) replaces diethyl
ether in favor of a less water-soluble, non-flammable, and non-
explosive solvent; 3) maximizes the accuracy and precision of the
analysis; 4) uses a specific detector that minimizes or eliminates
interference from non-target substrates and avoids chroroatographic
base-line rise due to oven temperature programming.
-------
154
MATERIALS AND METHODS
Extraction
Soil from 0- to 10-cm depth of a Larimer County Paoli (8) sandy
clay loam was air-dried and reduced to a particle size of less than
2.0 mm. The soil had 24 g kg"1 organic matter, a pH of 7.7 (1:1
soil/water), an EC of 28 ds m"1, and a CEC of 17.4 cmole kg"1. An
herbicide spiking solution was prepared in methanol and contained 25
Hg mL"1 of Silvex (2,4,5-trichloropropionic acid), 2,4-D (2,4-
dichlorophenoxy-acetic acid), Dalapon (2,2-dichloro-propanoic acid ),
2,4-DB (4-(2,4-dichlorophenoxy)butyric acid), Dicamba (3,6-dichloro-2-
methoxy-benzoic acid), Dichlorprop (2-(2,4-dichlorophenoxy)-propanoic
acid ), 2.,4,5-T (2, 4 ,5-trichlorophenoxy-acetic acid), MCPA (4-chloro-
2-methylphenoxy-acetic acid), MCPP (2-(4-chloro-2-methylphenoxy) -
propanoic acid isooctyl ester), and a surrogate, (2,4-dichlorophenyl-
acetic acid).
A 0.5 mL aliquot of the herbicide solution was added to 100 g of
soil and mechanically tumbled for 0.5 h at 32 rpm. The soil was then
saturated with 1500 mL of reagent-grade water, pH adjusted to 12 with
6M NaOH, and mechanically tumbled for 0.5 h as previously described.
The study was run in quadruplicate at an herbicide concentration of
125 fig kg"1. Assuming a 100% partitioning of the herbicides from soil
to water, the water concentration would be 8.3 /ng L"1. The soil
extract was then treated with 18 nmol of CaCl2 and allowed to stand
for 10 min. A 500 mL aliquot of the soil extract was brought to 1000
mL with reagent water, adjusted to a pH of less than 2 with
concentrated sulfuric acid, and extracted with dichloromethane in a
continuous liquid extractor for 18 h. The solvent extract was then
passed through acidified Na2S04 and concentrated to less than 2 mL
-------
155
using a Kuderna-Danish concentrator. Immediately prior to analysis,
the extract was esterified with diazomethane and exchanged to hexane.
The final extract volume was adjusted to 1.0 mL under a stream of
purified nitrogen.
Analysis
All analytical standards and sample extracts were analyzed using
an HP-5890 GC coupled with an 01-4420 electrolytic conductivity
detector. The detector was set up in the halogen mode using a reactor
temperature of 950° C. The reactant gas, H2, was set at a flow rate
of 100 mL min"1. Chromatographic separations were performed using
either a 30 M J&W DB-608 or DB-1701 fused silica megabore capillary
column. The liquid phase film thickness used was 0.83 urn and 1.0 urn
respectively. The carrier gas, He, was set at a flow rate of 8 mL
min"1 and the makeup gas at 30 mL rain"1. The GC oven conditions were
65° C, for 4 min; 25° C min'1 to 150° c for O.Sroin; 3° C min*1 to 225° C;
20° C min"1 to 270° C. Standard and sample extract injection volumes
were 3 /iL. Finally, Chromatographic peaks were collected and
integrated using a Maxima 820 data system developed by Dynamic
Solutions.
RESULTS AND DISCUSSION
A summary of the chlorophenoxy-acid herbicides from soil and
water is shown in Table 1. Accuracy used here, is defined as the mean
percent recovery and error associated within a particular
determination, and is calculated as:
Mean Percent Recovery = 100 (S Replicate Values/ No. Values) (1)
True Value
and
Mean Percent Error = 100 (Measured Value - True Value) (2)
True Value
-------
156
Precision, defined as the variability amongst the replicate
values and measured as the percent relative standard deviation (RSD)
or coefficient of variation is calculated as:
Percent Relative (3)
Standard Deviation = 100 Standard Deviation of Measured Values
Mean of Measured Values
Relatively high recovery and low RSD (coefficient of variation)
was achieved as illustrated in Table 1. Only dicamba failed to be
recovered above the 90% level and showed a mean percent error above
10%. However, the reported mean recovery and RSD for dicamba is
consistent with results reported in other EPA analytical methods.
Figures 1 and 2 are chromatograms of the herbicide standards and
a single soil extract analysis. It has been Our experience with the
ELCD is that instrument detection limits are approximately 50 times
lower than the standard concentration of 8.3 rag L"1 represented in
Figure 1.
The ability to achieve such accuracy and precision is believed to
be directly related to the following: 1) partitioning the salts of the
herbicides directly from soil to water; 2) minimizing adsorption of
herbicides to colloidal surfaces by flocculating the soil suspension
with the addition of CaCl2; 3) minimizing or eliminating further
hydrolysis procedures for matrix clean-up purposes; 4) the use of a
non-or low water-soluble extraction solvent such as dichloromethane;
and, 5) using an electrolytic conductivity detector that is selective
only for halogenated species and is unaffected from temperature
programming of the GC oven as shown in Figure 2.
CONCLUSIONS
By performing a simple soil-water extraction under basic
conditions and reducing the zeta potential of the soil suspension with
-------
157
CaCl2 to encourage flocculation, significant partitioning of
chlorophenoxy-acid herbicides from soil to water is achieved. Further
extraction of the flocculated soil suspension with dichloromethane,
derivitization with diazomethane before solvent exchange, and analysis
of the solvent extract using an electrolytic conductivity detector,
yields accuracy and precision results that are equal or superior to
those reported in EPA analytical methods.
ACKNOWLEDGMENTS
We are grateful to Bob Jump of ATI for performing the GC/ELCD
analyses. We also would like to thank Drs. Ken Barbarick and John
Tessari and Ms. Kristy Lynch for their review of this manuscript.
LITERATURE CITED
(1) U.S. EPA, National Pollutant Discharge Elimination System,
Appendix A, Fee?. Reg,, 38, No. 75, Pt. II, Method for
Chlorinated Phenoxy Acid Herbicides in Industrial Effluents,
Cincinnati, Ohio, 1971.
(2) U.S. EPA, "Extraction and Cleanup Procedure for the Determination
of Phenoxy Acid Herbicides in Sediment," EPA Toxicant and
Analysis Center, Bay St. Louis, Mississippi, 1972.
(3) U.S. EPA, "Method 615. The Determination of Chlorinated
Herbicides in Industrial and Municipal Wasterwater,"
Environmental Monitoring and Support Laboratory, Cincinnati,
Ohio, 45268, June 1982.
(4) U.S. EPA, "Method 8150. Test Methods For Evaluating Solid Waste,"
1986.
(5) U.S. EPA, "Method 515.1. Determination of Chlorinated Acids in
Water by Gas Chromatography With An Electron Capture Detector,"
Environmental Monitoring and Support Laboratory, Cincinnati,
Ohio, 45268, 1990.
(6) Aldrich, Technical Information Bulletin No. AL-180.
(7) Black, T.H., The Preparation and Reactions of Diazomethane,
Aldrichim. Acta, 1983, 16 (1), 3.
(8) Soil Survey of Larimer County Area, Colorado. U.S. Department of
Agriculture Soil Conservation Service, December, 1980.
-------
158
O
u
Q)
-*
Stta
P4
Q)
n
^tts
a,
0)
CM
=tfc
P^
OJ
H
=*fc
QJ
(U
OS
~
'tn
K^
Cn
a.
o
c
0
u
(1)
•o
•H
u
£1
0)
K
II
II
II
II
II
II
II
II
II
II
jj
II
II
II
II
II
II
II
II
II
II
II
i
H
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
II
||
||
in
in
^
o
1-1
H
ft
o
n
*£>
H
H
en
CM
rH
in
CM
rH
0)
a
o
3
r-
VO
(Tl
CM
in
CM
VO
o
H
ft
(N
rH
in
CM
rH
Ol
u
2
VD
H
CO
CO
VO
H
in
rH
o
o
H
iH
H
rH
in
CM
rH
jflj
CO
y
5
H
Q
cn
in
co
CM
H
rH
in
CM
rH
<
O
£
rH
Q
0
rH
0
CO
CO
(M
V£
H
H
CO
CM
rH
in
CM
rH
0,
O
OH
OH
OH
O
X
M
Q
H
(N
O
H
*3«
ft
in
n
o
CM
H
CM
CM
rH
in
CM
H
Q
1
v
(N
C*
£^
VD
CTi
fT)
CM
CM
CM
n
H
H
CM
CM
rH
in
CM
rH
X
M
>
H
CO
c* n
lA r*»
0 0
H rH
^* V0
n n
rH rH
t"* ^0
f) fO
-------
159
OO'ZS
•i !
I—
CO
I
OJD
£
O
2
o
CO
1
rz «•«
NOJV1VO 01'S
i-SVZ IS'IZ _
X3A1IS 89'U _
ddDU Z6'£T
j-ms <]JSH 5?'ft _
o
ui
vjou srsi -'
JO't -
o -
Hi
,J N
9S'8_ j
_
"''-
C
Q
c
c
C C
in
-------
160
eg
•o
a
a
CO
3,
a
60'S
i c ^
if! —'
16-9 -
! .
!_.
on
-------
161
QUESTION AND ANSWER SESSION
MR. YOUNG: Yes, my name is
Michael Young from NET, Atlantic.
I have a question regarding the hydrolysis step on your
method. As I understand it, the hydrolysis step of Method 8150
is to convert naturally esterified forms, that is, the phenoxy
acids that are not in the free acid state that are esterified
either to natural alcohols or phenolic groups present in, say,
humic matrix or something of that sort in the soil.
Are you sure from your analysis method that a half an
hour shake at room temperature at pH2 will accomplish this
hydrolysis and have you considered doing a method recovery study
using esterified forms of your phenoxy acids in order to
accomplish this?
MR. JACKSON: I haven't gone into
any rigorous tests. Basically, what I've done is taken the free
acids in a methanolic solution and spiked them into soil. The
soil was calcareous so they readily formed a salt and then added
the water and partitioned them. So, I don't really have any data
to look at that.
MR. VANOPAL: My name is Howard
Vanopal and I work at the Army Environmental Hygiene Agency at
Aberdeen.
I've been doing some extensive work with this type of
these herbicides like you have for the last couple of years and
I've discovered a couple of things. In reference to the
gentleman's other question, I also have grave doubts that this
method will hydrolyze, for example, some of the long chain
herbicide esters and these are the type that are used in
agriculture. You have some of these esters that are quite long
like isopropyl esters, propylene glycol, butyl ether esters.
These require a fairly rugged hydrolysis with heat in order to
completely work.
In reference to the dicamba, I think the problem with
low recoverage is due to its volatility. It is far more volatile
-------
162
than any of those others and you must take care when you do
solvent transfers and things like this. Otherwise, you can get
good recovery of it. But overall, your approach is somewhat
similar to what we are working on.
MR. JACKSON: This method was
basically tested only on the 8150 compounds and from the
chemistry, it works very well on four carbon phenoxy acids and
less. The higher carbons do have a problem. Since these were
not the targeted compounds, they were not addressed and this
method may not be applicable at all to those other ones that are
being used.
MR. ELLEMAN: Dave Elleman of
Columbia Analytical Services.
Have you actually done this in a real world sample, a
side-by-side comparison between the two methods?
MR. JACKSON: In the laboratory,
8150 was practiced...before it was practiced commercially, the
four replicate precision and accuracy test was done and the
results were acceptable as to the method. That was done
approximately two years ago and then this method here was
developed, oh, about six to seven months ago and we took these
results and compared them and these were equal or superior. So,
we did do a side-by-side, but not at the same time.
MR. ELLEMAN: What I really was
comparing is on a real world matrix, when you're actually taking
a soil sample and you've done one soil sample by one method and
the same soil sample by the other. We did some supercritical
fluid extractions on hydrocarbons and it worked great in sand,
but it didn't work a hoot on a real clay kind of sample. I
wondered if that might play a role also in this tumbling kind of
episode...get into the matrix with a water by comparison to using
a real solvent to get into it.
MR. JACKSON: Well, the half hour
tumbling time was arbitrary. We had to start somewhere and I was
looking to reduce method time. So I started at a half hour and
-------
163
when I got the results back for that typical soil, and that's a
sandy clay loam that has morilanite type clays in it. It also
has some very simple minerals, kaolanites and then gypsite and
things like that. So there's lots of active sites on that soil
that you can get hydrogen bonding or you can get bridging. You
get absorption and the half hour seemed to work real well. I
haven't found a soil yet that the surrogate recovery, and that's
all basically that we can monitor in real samples...that the
surrogate recovery was low. We've been getting very good
surrogate recoveries.
MR. ELLEMAN: But there's a
difference between a spike and a surrogate. It's already
dissolved in some methanol by comparison to the soil itself
because everything is already contaminated inside of it for a
side-by-side comparison.
MR. JACKSON: Yes, I don't really
have any exact side-by-side comparisons. Based on the results
here, what I tried to do was show that this method was equivalent
or better to the regulatory method.
MR. TELLIARD: Thank you, Gary.
-------
164
MR. TELLIARD: Our next speaker is
Dr. Hodgeson. He's from our laboratory in Cincinnati, EMSL,
Environmental Monitoring Methods Systems...
MR. HODGESON: Laboratory.
MR. TELLIARD: Laboratory. I
always forget the systems. He's going to talk to us on the
measurement of acid herbicides and disinfection products in
aqueous media. Jimmy?
-------
165
ADVANCED TECHNIQUES FOR THE MEASUREMENT OF ACIDIC HERBICIDES
AND DISINFECTION BYPRODUCTS IN AQUEOUS SAMPLES
Jiramie Hodgeson, Research Chemist
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
26 W. Martin Luther King Drive
Cincinnati, Ohio 45268
Jeffrey Collins and David Becker
Technology Applications, Inc.
26 W. Martin Luther King Drive
Cincinnati, Ohio 45268
ABSTRACT
Acidic herbicides and haloacetic acids are two important
classes of organic acids, which may occur in aqueous matrices and
are subject to current and pending regulations, respectively. The
haloacetic acids are formed along with the trihalomethanes as
ubiquitous components of chlorinated drinking water supplies. The
techniques currently used for monitoring both classes of compounds
are complex liquid-liquid extraction procedures, which employ large
volumes of organic solvents. The hydrophilic nature of many of
these analytes places a significant limit on method performance.
Another problem with current gas chromatographic methods is the use
of diazomethane as the methylation reagent. Many laboratories are
reluctant to use this reagent, even when applied in a relatively
safe manner.
This presentation will discuss our recent work on liquid-solid
extraction as a promising means for both simplifying the
methodology and improving method performance. The acidic
herbicides may be extracted from aqueous samples by either reverse
phase or ion exchange media. The hydrophilic haloacetic acids are
efficiently recovered by ion exchange. Recovery data will be
presented for packed columns as well as commercially available
filter disks. Data will also be presented on the use of acidic
methanol as an alternative reagent to diazomethane. This reagent
provides recoveries comparable to diazomethane for phenoxyacid
herbicides and haloacetic acids.
-------
166
INTRODUCTION
Among the most complicated of EPA drinking water methods is
515.1 for acid herbicides. This paper will discuss a much
simplified approach for this class of analytes, as well as an
improved means for the measurement of haloacetic acids in
drinking water. The latter are ubiquitous components of
chlorinated supplies and are subject to regulation under EPA
disinfection byproduct rules in the 1994-1995 timeframe. Several
of the acid herbicides are currently regulated and more are
subject to future drinking water regulations.
A brief summary will be given of the current methodology for
acid herbicides (EPA 515.1) and haloacetic acids (EPA 552).
These methods are published in EPA drinking water manuals
available from the National Technical Information Service. Two
major means for the simplification and improvement of these
methods will be discussed. These are the use of liquid-solid
extraction (LSE) for sample pre-concentration and the use of
acidic methanol for methylation as an alternative to
diazomethane. Many laboratories in this country have been
reluctant to use diazomethane, even when provided with the quite
safe micromolar procedure of Method 515.1. As a point of
emphasis, the two techniques discussed here are still in the
developmental stage. When completed, these procedures will be
published in the open literature and will be available to the
public as EPA Methods 515.2 and 552.1 for the acid herbicides and
haloacetic acids respectively.
DISCUSSION
ACID HERBICIDES; The analytes of 515.1 are listed in the slide
presentation. These include the commonly employed phenoxyacid
herbicides, some phenolic compounds and benzoic acid derivatives,
bentazon and dalapon (2,2-dichloropropanoic acid). Several of
these compounds were not determined quantitatively by this method
during the National Pesticide Survey because of poor precision,
namely acifluorfen (blazer), chloramben, dalapon and
4-nitrophenol. The latter compound does not methylate and does
not belong in the analyte list of 515.1 or the new 515.2
discussed below. Dalapon is a hydrophilic analyte, which does
not partition favorably by liquid-liquid extraction or by the
reverse phase LSE procedure of 515.2. It is efficiently
extracted from water by anion exchange LSE and has logically been
included in the analyte list for the new LSE procedure for the
haloacetic acids (552.1). Both acifluorfen and chloramben are
efficiently recovered and quantitatively determined by the
reverse phase LSE method outlined below.
Method 515.1 begins with an aqueous phase hydrolysis step at
pH 12 to convert commercial ester forms of the herbicides to the
free acids. At the basic pH the acids are present as anions and
-------
167
the aqueous sample is washed with methylene chloride to remove
potentially interfering neutral and basic organics. After
acidification, the 1 liter sample is serially extracted with
ethyl ether. Drying of the ether extract with sodium sulfate is
a tedious step with the potential for analyte losses. The ether
volume is reduced to 1-2 mL by Kuderna-Danish evaporation with a
solvent exchange to methyl-tert-butyl ether (MTBE). A florisil
clean-up of the extract is included to remove potential
interferences from, e.g., high humectant ground water. The acids
are methylated by a safe procedure employing micromolar amounts
of diazomethane. The analytes are determined quantitatively by
gas chromatography with an electron capture detector (GC-ECD).
Overall, this is a tedious, time-consuming method requiring a
high level of operator skill. Even a skilled operator is limited
to approximately six samples a day for the sample preparation
step alone.
Two mechanistic options are available for the extraction of
acid herbicides from water, anion exchange and reverse phase
extraction on a hydrophobic substrate. We have done sufficient
work on anion exchange to demonstrate feasibility. However,
reverse phase extraction is a simpler approach. Of the reverse
phase options, extraction by 3M Empore 47mm disks is faster and
more efficient than by extraction cartridges. We have
successfully employed both C-18 and the newer resin disks. The
resin disks provide somewhat better efficiency and certainly
faster extraction times than C-18.
An outline of the procedure is given in the slide tables.
Anhydrous sodium sulfate (20% w/w) is added to a 100 mL sample
and the sample pH is adjusted to 1.0 with reagent grade HC1. The
disks are conditioned by sequentially passing through the disk 20
mL 10% methanol in MTBE, 5 minutes air (under vacuum), 20 mL
methanol and 20 mL reagent water. The sample is then extracted
without allowing the disk to dry and under a vacuum of 5 inches
Hg. The disk is air dried under vacuum for 20 minutes and eluted
with two 2 mL aliquots of 10% methanol in MTBE. Approximately 8
mL of MTBE is used to rinse the sample bottle, frit and funnel.
The combined MTBE extract is dried through a wide bore Pasteur
pipet packed with anhydrous sodium sulfate (described in Method
552). The sample volume is reduced to approximately 4 mL under a
nitrogen purge and the sample methylated by the diazomethane
technique of Method 515.1 or 552.
All of the data in the slide tables was obtained with a
sample volume of 100 mL and a concentration factor of 20. Larger
sample volumes and concentration factors are feasible. However,
a significant drop in recovery was observed for some of the
analytes at a sample volume of 1 liter. The determination of
optimum sample volume and concentration factor for overall method
performance are parameters yet to be determined.
Representative recovery data are presented in three slide
tables. These tables compare C-18 versus resin for unsalted
-------
168
reagent water, C-18 versus C-18 salted, and the effect of washing
the sample bottle for a salted high huraectant ground water.
Several conclusions can be drawn from these data. Salting the
water sample is necessary for optimum recovery and this salting
largely washes out the significant differences between C-18 and
resin recoveries for unsalted reagent water. The recovery data
on fortified ground water show that high humectant levels do not
adversely affect recovery. These data also indicate that the
florisil clean-up procedure of Method 515.1 will not be reguired.
The disk appears to effect sample clean-up in that a substantial
portion of the humics appear to pass through without retention.
Any humics retained are apparently not eluted with MTBE. This
same table illustrates the importance of rinsing the sample
bottle. The reagent water data on 5-hydroxy-dicamba is suspect,
possibly because of a problem with the fortification standard
used. We believe this analyte is recovered greater than 60% and
the ground water data is supportive.
The data indicate that all of the analytes may be recovered
sufficiently, with the exception of Dalapon. Method 515.2 will
be written to require the extraction of agueous standards to
correct for extraction efficiencies. Dalapon is too hydrophilic
to be collected by reverse phase and it is thus included in the
anion exchange procedure for haloacetic acids discussed below.
HALOACETIC ACIDS AND DALAPON; While not as complicated as Method
515.1, Method 552 also requires serial extraction with MTBE and
evaporation of the considerable excess solvent. The latter must
be accomplished by a gentle nitrogen purge because of the
volatility of the monohaloacetic acids. The acids are methylated
by the same diazomethane procedure used in Method 515.1 and
analyzed by GC-ECD. A skilled analyst is still limited to 6-8
samples per day.
The anion exchange extraction procedure was developed with a
number of objectives in mind - method simplification and
pollution prevention (i.e., eliminate evaporation of large
volumes of solvent), development of a much needed method for the
regulated analyte, Dalapon, and inclusion of a simple alternative
methylation procedure to diazomethane. While method development
has not been completed at this time, accomplishment of these
objectives appear imminent. Accomplishment of the latter is
particularly satisfying in that methylation is efficiently
accomplished directly in the elution solvent, acidic methanol.
A brief summary of the method is given. The anion exchanger
is analytical grade AG 1-X8 resin, 100-200 mesh, from Bio-Rad.
The commercial resin is washed liberally with deionized water and
stored under deionized water. One mL solid phase extraction
tubes (Supelco) are loaded with the resin dropwise as a slurry
until a height of 10 mL wet resin is attained. The resin is
maintained under water until used. The column is conditioned by
sequential addition under vacuum of 10 mL aliquots of methanol,
reagent water, 1 M HC1 in methanol, reagent water, 1 M NaOH and
-------
169
reagent water, without allowing the bed to dry. A 100 mL sample
is extracted immediately after addition of the last reagent water
aliquot. The analytes are eluted with a small aliguot (2-4 mL)
of H2S04/methanol and quantitatively converted to their methyl
esters by simply heating this aliquot with a small aliquot of
MTBE added as cosolvent for 30 minutes at 50°C. For a 4 mL
aliquot of 10% H2S04/methanol (the probable final eluant of
choice), the methyl esters are partitioned into the organic phase
by the addition of 10 mL of 10% by weight of sodium sulfate in
reagent water. The organic phase is removed and the aqueous
phase is washed with several more small aliquots of MTBE to a
total volume of 8-10 mL. The small excess of solvent is removed
by nitrogen purge to a final volume of 5 mL for a concentration
factor of 20. The analytes are analyzed by GC-ECD.
The bar graphs shown in the slide tables show the recoveries
obtained for all the analytes with 4 successive 2 mL aliquots of
25% H2S04/methanol. The recovery is essentially maximized after
the first two aliquots and the final method version will probably
call for a single 4 mL aliquot of H2S04/methanol. The next slide
table shows some preliminary haloacetic acid and Dalapon
recoveries in reagent water and simulated high ionic strength
water. The presence of high concentrations of competitive anions
is expected to present the greatest challenge to the method.
These data in fact show reduced recovery for the monohaloacetic
acids. We have since shown that the anion responsible for the
displacement is sulfate. One solution for high sulfate waters is
sample dilution, but this would raise MDL's. Another possibility
is sulfate clean-up with cation exchange resins in the Ball form,
but this approach has not been tested.
ESTERIFICATION EFFICIENCIES: The remaining slide tables provide
esterification efficiencies using both diazomethane and the
H2S04/methanol technique. Absolute esterification efficiencies
have been determined for those analytes, for which methyl esters
are available. Esterification is quantitative for all the
haloacetic acids, Dalapon, all the phenoxyacids and a few others.
A critical factor for the acidic methanol procedure is the
addition of an organic cosolvent. This serves to shift the
equilibrium esterification reaction to the right.
Toluene was added as a cosolvent for the esterification data
shown for the acid herbicides. Only the true, relatively strong
acids are methylated by this procedure. No yield is obtained for
phenolic compounds. In our current procedure for Dalapon and the
haloacetic acids, the elution solvent is 4 mL of 10% by volume of
H2S04 in methanol and 2.5 mL MTBE is added prior to heating for
30 min. at 50°C. Under these conditions, the esterification is
quantitative for all of the analytes, contrary to the data shown
in the final data table.
-------
170
QUESTION AND ANSWER SESSION
MR. McCARTY: To remind
questioners, please come to the microphone and identify yourself
and your affiliation.
MR. VANOPAL: Yes, I've got one
question on this. Is this designed to do just acid forms of
herbicides or can you do all forms that would be commonly found
in the environment?
MR. HODGESON: The techniques that
I've been talking about, this liquid-solid, you first would have
to do the same thing on the front end of the method. You'd have
to hydrolyze. You know, I didn't address that, but you'd still
have to hydrolyze these acids. By the reverse phase, you could
collect these compounds, but then you're going to have to still
do something before you do your final analysis in terms of a
hydrolysis step probably.
MR. VANOPAL: Okay.
MR. HODGESON: The halo acids, of
course, are not a problem. They're present only as a free acid.
-------
171
CO
o
cooc
HI " CO
DaS
OzO
LU Q
O ID
otr
O IN
Z T
<
O
<
u.
>»
k.
o
00
«k
o -^
.t^ (0
s5
E
|
>
LJJ
CO
h-
I
O)
CO
10
I
CO
IS
-------
172
o
o
co >,
h- -Q
Z CO
PI
g*
6
CO
Q
O
I
^
a 3
UJ CO
§'l
^ '5 °
^* .ii Q
o^§
z^^
* -1-1
z ' ,
|— LO LO
Z -o T5
0)
cjS
fe I
2g
I- UJ
X c
HI o
9 i
O w
CO TJ
Q 2
5^
o
a>
D)
C
CO
«n
s^
£g
*-»
05
7S <1>
gcr
5 o
£ o
o
-------
173
CO
^
URRENTL
O
•
• ^^
lO
U)
Q
O
I
LU
2
X
O
co
z
•*-»
$
CM
I
a
a
E
03
CO
0
•*-»
_l
T-
co
0
>
1
0
•o
^
N
j*
O
•o
>*
O
4-»
D
0
r
s
03
W
CM
O
CM
I
O
_«.._ _»
•t™
'i
J3
CO
1
1
1^^
"o
^ Jib
C/3
H2SO4 and
•4-rf
$
CM
I
Q.
4-»
CO
^D
*n
\«/
<
0
-C
0
">
_C
"0
4-»
'>
>
CO
c
"•f-*
o
2
X
0
^
'u.
0
CO
CO
I
0
0
2
+
111
GO
h-
2
^BH
13
•o
03
Q
k ^
^
azomethane
^ ^r
^
JZ
•4-»
•^
5
0
^.^
OJ
>
_T"
^B.
4-'
0
S
O
LU
i
O
y column G
k.
_co
'5.
03
O
• •
•^*
i
^a t
0
C
E
•
5J
0
Q
D
O
O
1
00
c
CO
0
a
E
03
CO
N*
1
to
• •
0
E
'•+-•
a
0
V-
a
0
a
E
03
W
-------
174
CO
ULJ
LL
o
H
CO
-------
175
z
O
H
Q.
FHOD 552 DESCR
Haloacetic Acids
i —
LU
*>
4C—
RVATION
ith NH4C
UJ £
LU "O
COLLECTION AND PRI
nL Sample Dechlorinate
LU 1-
-1 0
0- 0
^ *-
< •
CO
CO
JBrf
CLEANUP
lination and Organic We
LJLJ CO
-1 ^
0- <
^ ,
<
CO
ier(MTBE
J_
^-»
0
I
JQUID EXTRACTION
ification
act with Methyl-t-butyl
i ~° £
Q'o x
^ r ^
=: < LU
_J
01 i
1 '
_J
T CONCENTRATION
o
^^
<
DC
H
X
UJ
UJ
z
<
UJ
ICATION WITH DIAZOM
omolar Procedure
LL ^
&£
uiS
h- ,
CO
UJ
,RY GC-ECD ANALYSIS
<£
-j
Q.
<
O
-------
176
o
cc
CO
Q
O
"co
O^i
[_ 0
LU ^
O g
^3
Q
LU
CO
Z
• —
LJL
NALYTE
< < CM 00
Z Z ^ O) CO
CM
oq < co iq oq
d z r^' ^t co
2 !°
'o 'o
« 2 2
2 .2 "9 £ -43
CD
O
CO
O
00
o o o
O O^ g
O g
§
< <
Z Z
<
0 Z Z
o —
0
000-^
b h-
omoace
D
a
6-Trich
oro
c
;g
'o
o < £ ^ ^
en b of CM
-
o
0
E
03
CO
T3
-------
177
CO
z
o
\-
Q_
O
q
o
<
CL
h-
X
LU
Q
1
_J
O
CO
1
I
Q
HIT
3
rx
CO
o
c
0
CL
*s^
CO
"O
'o
<
.g
05
0)
6
Phase
e verse
GC
CO
D
CO
5
0
O)
C
CO
o
X
HI
C
g
'c
<
CO
_*
CO
vx
Q
•D
0
CO
CO
D
1
c
»f-
ft
CO
D
CO
0
CO
c
E
"o
o
c
g
"•+-»
o
CO
L_
J_J
^^^"
X
LJJ
0
0
N
Divinylbe
^
0
c
0
>
4-^
CO
CO
D
CO
k_
5
/Tl
TO
ao
(O
X
DC
H
<
^••B
:E
m
D
•
CO
a.
i_
0
E
>»
o
a.
o
O
monium)
E
<
>
^
03
r~
s(Quarter
0
•»-•
CO
O)
c
0
*-•
CO
• •
LU
O
Z
<
X
o
X
LJJ
Z
z
<
'E
€D
'c
O
E
<
>
i_
um(Tertia
T3
0
2
CO
D
u.
0
0
D
XX
Q.
•«
^O
0
CO
CO
JD
CO
0
'co
^
1
0
•o
c
00
1
O
• •
LU
CO
<
I
CL
LJJ
CO
V J
cc
LJJ
>
LJJ
QC
CO
0
,0
•*-•
i-
03
CL
C
tn
\ii
0
CC
CQ
>
D
i
CO
a.
-------
CD
D)
178
O
0)
^CIDS
Procedur
^^
O D)
Z03
JC
i t
0
O
73
'o
JC
•;r
^
0
D
LLI
•^^
O
O)
i
O
10
03
C
E
o
CO
1
IO
>
03
c
<(
-------
179
-
-
~
-
-*
w
y*i
=
=N
-
~S P
.=. : ^
^ ~ :^=
^
^
i=f
===== i.-
5 fi
==
-
-T- ' ~ ..-
-
^ ^
__— -*~ _a^-'~ i
: £ :
O O O
j~! i™4 *~1
In =rr rri
= : •- =:-r
-S3- 1 Ofi
0 0
0 0
N T*
, ni x
-------
180
RECOVERY OF DICAMBA AND 24D
COMPARISON AT PH 8 AND PH11
DICAMB
PH8
DICAMB
PH1 1
24D
PH 8
24D
PH 11
100
DC
111
O
(J
UJ
DC
t-
UJ
O
DC
HI
a
4 1
ALIQUOT
-------
181
RECOVERY OF DALAPON AND 24DB
COMPARISON AT PH 8 AND PH1 1
DALAPO
PH8
DALAPO
PH11
24DB
PH 8
24DB
PH 11
1OO
QC
HI
O
O
HI
d
Hl
U
CC
LU
CL
4 1
ALIQUOT #
-------
182
C7) o
Q t3
< x
UJ
O
CO
OC co
O a>
CD
cc
Mt±
CO
CO
Q
c
o
c
o
O
CM
V
I
Q.
(\3
0
CL
E
03
CO
E
o
o
"o
03
^»
"x
LU
LLI
m
H
^
-e>
D)
^&*
>S
>
^^
"o
\^
CO
Cu
o
Q_
H—
O
£
1
CM
"i
-*-•
D
LLI
O
0
C
03
£
0
E
o
N
03
Q
£
.-0
§1
x o
UJ 2
£o
0) CO
UJ I
Q
O
UJ
i
O
o
c
£
^
0
O
^
'a
OJ
0
.a
CO
'co
"c5
c
-------
L
0
V
IT
in
N
8
u
(f
183
CS EXTRACTION DISK ELUTION TEST
•JO
80
70 -
60 -
50-
40 -
30-
20 -
1\ \
\
/\
•V
\
\
X
10 J
••/J\\
//K\
5 * 2MI EUmONS RED DYE
/\\y//!
_f__ {_ * > T r^L-ri
1
ELUTiON
#2
SAMPLE
q-so
-------
184
CO
CO
Q
i
z
o
QCO
QC 0)
H-CC
B:
UJ 0)
CO'C
< 3
<
z
CM
+ i
CM
IO IO
+ l +l
CM
CO
CO
CO
i +i +i +l
O O) (T)
CO O) OT
O)
CO CM
°2
+ 1
CM
??=$2
+ i +i +l
CO CD CO CD i- CO o
co N r^- co 10
o w
0)
a
o
E
03
O
O O
Q. 03
4 =
<« 0) ,, .
Q QL O iO
ydrox
oram
03
c
co
O
o
N
03
19
•_ ^
H CD J=
X ' S **
0 m LJ
>
entazon
Q.OCQDCMCOCMCMQCQ
O)
03
CO
(D
03
-------
185
CO
^
CO
D .
• 5
z Q
o •
^ 2
H™ •+-'
o <*>
<.
_J
DC 0)
h- DC
X +,
mco
m/i\
u/
CO 'C
< 5
I^p
o
Q- o
in °
LU of-
co"-
<£ S
LU -2
>5
m^^f
^
B- +-
IBICIDES
Reagen
LL
LU
z
TJ
(D
+«•
03
CO
CO
\nj
T~
1
O
CO
*^^
i
O
NALYTE
<
CM CO £ co
+ 1 4- 1 + 1 . .
+ 1
00 lO CO rn
CM ^ CD OT
^ ^
+ i
T ' + 1
ogooo
— 03
o S
1 i
Q. .2
o c -Q
X ° >»
5 J3 X
C £ £ 0
£82-0
JS c ^ i
03
-------
186
CO
CO
D
i Q
z <0
0*
h- +l
O co
< »
QC co
111 5
LLI -73,
DC o
CO 0
LLI ^)
O CD
QCO
CQ
CC
LU
Si
"cO
CO
c
'co
CD
CC
"U
^)
"5
CO
CO
1
o
LLI
H
>3
<
-z.
o
CO
°^
"So
CM^CD
+ 1+1
00 00
+ , +1 +1 +1
°0 CO rn K ^ CO
CVI CO v N- N CO
CO
E
co
o
c S
2 >•
-Q X
o
a.
co o
£25?
2P 25
>* o -
ntazon
— — 3171 O
CO -C | .2 CO CD
< Q O 10 CL Q CQ
O)
10
"S
CO
0)
I-
CO
(D
CO
o
-------
187
CO
CO
D
i
z
o
o
o:
X
UJ
UJ
CO
UJ
f^\
^JJ
DC
LJJ
>
UJ
CL
CO
UJ
Q
O
CD
cc
UJ
0)
D
•
35
"55
^
tt\
CO
<
MB
Q
Cl)
^
52
03
^
^
c
D
2
o
CO
03
5
CO
00
wV
T—
i
O
CO
CO
*^\f
1
O
LU
>j
<
Z
^
CO
o>
CD
O
C
o
a
03
03
D
00 CD
T- CO
00 N
^ r^
CM ^t
K CD
>MBB
O
C
a
o c
2 -^
0 03
03 Si:
^^ C^
CD -C
Q.O
O
CM
N-
00
CD
03
E
03
O
P^
x
0
1
1C
o
CM
CM
O)
CM
O
E
J5
o
a.
O) O O
CO CO O)
00 CM
O CD °°
O) T- CO
.Q S
CD i_ "2
co a> c
O N O3
C 03 O
b CD b
N 0)
O N
O 00
?= CD
CO i-
O) CO
Q X
T §
^ —
CM" co
O)
o
o
CO
CD
O)
K
10
(^
CM
CO
CM
CO
00
CD
GQ
D
^F
CM"
?co
^t CO
CO O)
CO T-
CO CO
CD 00
_
03 O
.C N
•^ &
JZ -i-«
03 (D
O CO
O
03
C
03
O
E
D
X
g>
±
O)
IO
13
CO
(1)
CO
03
-------
188
ORGANIC ACIDS
DOWEX STRONG ANION EXCHANGE RESIN
PROCEDURE
0 100 ML WATER SAMPLE, PH = 7
0 AG1-X8 RESIN COLLECTION, HYDROXIDE FORM,
0.2 ML VOLUME
0 ELUTION WITH 2 ML 10% HCL/METHANOL
0 SOLVENT EXCHANGE TO MTBE
0 ESTERIFICATION
0 CAPILLARY COLUMN GC-ECD
-------
189
EXP 8 ELUTION 4X2 ML 25% H2SO4 MEOH
ELUTION SPEED / 1 ML/MIN SAMPLE PH 5
MCAA
MBAA
DALAPO
DCAA
DC
0
u
UJ
d
u
d
100
10
O.1
12341
2341234
ALIQUOT #
1234
-------
190
EXP 8 ELUTION 4X2 ML 25% H2SO4 MEOH
ELUTION SPEED / J\ ML/MIN SAMPLE PH 5
TCAA Jill BCAA DBAA
DC
UJ
0
O
UJ
OC
t-
LU
O
OC
UJ
a
10O
10
O.1
I
1234
ALIQUOT
-------
191
CO
LLJ
O
-z.
LU
O
111 .£?
>
-Q * X
E c ^ .Q u_ °
frt ^5 ^^ Q) UJ ^!J
v-Q-EcoQa^^
o co co o i i *'--£
!E "co .2 .£ ^L ^ ^ i
O Q Q Q c\T cvT cvT 10
DO
<
^rophenol*
u
Z E
O (0
^< . w s^/
Co *— ^s
-£ O 0
c "o ^
Q. E CO
c
CO
"co
LLJ
">
0
^
C
O
0
CO
CO
-------
192
LJJ
DC
Q
LU
O
O
DC
Q.
O
LLJ
g
DC
ID
U_
CO
o
CM
I
+
CO
I
o
CM
O
0
cc
II
3
o
<
o
CO
I
o
+
I
CM
0
CC
(MTBEJoluene
_i
E
CM
I
0
0
E
D
o
^[^
«^
•*-•
O
«T<
03
v_
•+-•
X
^ ^
111
0
O
D
•o
0
CC
0
c.
03
r-
••••
0
^
^
\
^t
O
CO
CM
X
c£
10
CO
1
E
T3
T3
<
O
O
G)
I
0
IO
_l_^
03
•
C
E
O
(O
1
LO
o
03
0
CC
•a
0
Trt
vU
CO
u.
0
03
4^
c
0
D)
03
0
CC
_J
E
10
•a
•o
<
0
CO
«*y
03
a.
o
f—
03
O)
6
•
_J
E
10
i
CM
O
+-»
03
k_
4-»
X
HI
-------
193
CO
LU
O
z
LJJ
O
u.
LL
LII
O
O
"••
LL
DC
LLJ
CO
LU
^
0
rrt
vw
O
;g
'5
^
o^
•
>
Q)
0
0)
DC
10
O)
0
>
CO
^—
•
5
o
•*-»
ft/
a
o
o
CO
+•*
c
CD
Q.
O
c
0)
hloram
0
IO
JQ
E
CO
o
T3
X
0
I
1
10
00
JO
0
0
Q_
o
inoseb
D
r^ T- o) N- o> K
CDOJOOOOOOOOO
—: C
2
C
CD
< Q CM CO C\T CM~ Q CO
h! .b o X ^ = ^T ^ oj
O
03
i_
4-»
X
LU
-------
194
CO
O
H
g
as
•^•^••I^B
O
DC o
LU
CO c
05
0
CO
(D
•*-•
0
CO CO
c
O
D
D
0
_ DC
O ^
•D
0
O
LJJ
DC
•
•••••
^^^i^
DC
LJJ
Q
LJJ
U_
^ ^r
O
2
0
T3
'o
Q
^
•^
0 *
DC
—
6
~n ^
0
O
0
-i-»
>
^^
"co
c
<
a
o
J5
o
f~\ ?$ 03
ii CD "t^
'. ^ (^
^. = 0
C\J CO Q.
03
03 03
C C
5oo||||
O)
C
C
o
CO
D)
0
CL R
5>
03
o
l
O -9
-------
195
MR. McCARTY: Our next speaker
this afternoon is going to be Bob Beimer from S-Cubed
Laboratories. For those of you who have been here before, Bob is
a fixture at Norfolk. As a matter of fact, we're not sure he
ever actually goes home. That's where the tan comes from, I
think.
He's going to be speaking today on temperature-
programmable on-column injector for GC/MS work. He got off easy
because Telliard would have raked him over the coals.
-------
196
MR. BEIMER: I have long been
convinced that the EPA believes all detectors are linear and a
corollary to that premise is that detector response curves, all
pass through the origin.
It's a frustrating fact that virtually all of the
methods EPA publishes require response linearity of the
detectors. I'm associated mainly with mass spectrometers and I
guess I've been doing GC/MS now for about twenty years, as a
result of this, I've determined that mass spectrometers response
is not linear.
In a continuing effort to improve detector response, we
have recently focused our attention on at the front end of the
system to try to determine if the injector on the gas
chromatograph has an influence on linearity.
I have pleaded with various instrument manufacturers
and think maybe they're starting to listen and will at least look
at linearity of response as an important aspect of selling
hardware. Reliability, dependability, precision, accuracy,
resolution, sensitivity, all of this is important, but nobody's
really bothered with response linearity.
As a part of this effort, we embarked on a program in
association with Hewlett-Packard to evaluate a cool on-column
injection system. The main focus of this effort was response
linearity.
We have compared the traditional splitless injection
system to the new on-column system by using a single instrument.
The instrument was a Hewlett-Packard 5970 MSD, equipped with a
splitless injector on one side and an on-column injector on the
other. We literally set the instrument up in the splitless
mode, ran a calibration curve and then turned around and took the
same column and hooked it into the cool on-column injector and
ran another curve to compare the two.
I'm going to give you just a brief tutorial on what a
splitless injector looks like compared to an on-column injector.
Slide 1 shows a Hewlett-Packard splitless injector. In the
-------
197
splitless injector, you shoot your sample onto a liner, which is
located inside the injection area here, which is typically heated
to 300 degrees centigrade. A small flow of gas goes down through
here and onto the column. When the purge is off, the gas is
diverted out in this direction and you don't have a purge gas
blowing through the injector. As a result a slight flow of one
milliliter a minute or so pushes the sample onto the column and
then after some pre-determined period of time, a couple of
minutes, usually this valve opens and allows the purge gas to
flow through the injector. At this point anything that's left
in the injector is blown out through this second line. The
technique results in some discrimination, both thermal
discrimination because of the volatility of your compounds, but
also if you don't get all of your sample out of that injector
while the purge valves are closed, it's going to be blown away.
Some of the data that I'll show you later will illustrate this
graphically.
Slide 2 is a schematic of an on-column injector. It's
a very low mass injector so that it can be rapidly heated or
cooled. In use, it is cooled to the same temperature as the
starting point of the column program. It's then temperature
programmed with the column so that it heats as the column heats
and is cycled back for the next injection.
There are a lot of guides in here. This area right in
here serves as a needle guide. There's another needle guide
located about here so that the needle from the syringe actually
gets onto the column. You're dealing with some fairly fine
needles and they tend to be a bit prone to bending, so guiding
them is extremely important.
At this point down here, the needle deposits the sample
directly onto the column. About two meters of 0.5 mm ID uncoated
fused silica column is butt connected to a traditional .2
millimeter ID DB5 GC column for determination of semi-volatile
analytes.
As a result all of the sample is placed on the column
-------
198
directly as a liquid. It disburses over this two meter uncoated
column and then as the temperature is programmed, it's
accumulated at the front of the analytical column and
chromatographed.
The cool on-column injector has some fundamental
advantages and I'll just touch on a couple of these.First of all,
if your analyzing thermally-labile materials, the low temperature
minimizes degradtion. The injector is never any hotter than the
GC column and you're not having to vaporize the sample in the
injection port.
The amount of sample that can be injected using cool-on
column is not as high as that which you can get for the splitless
injector, but when doing trace analysis, it's certainly adequate.
Slide 3 is a chromatogram of a series of base neutral
compounds, the typical CLP/BNA standard, run using splitless
injection. Please note the relative intensities of the later
eluting compounds. These two right here are higher molecular
weight polynuclear aromatics eluting late from the column.
Slide 4 is the same sample using cool on-column
injection you can see the relative height of these peaks is
significantly greater. In fact, there's a general trend to
higher transmission of material as the retention time increases.
More data will point this out graphically. It should be noted
that this is not a distortion in the cool-on column; it's because
the splitless injector discriminates against higher boiling
materials. That is they are not getting onto the GC column
during the injection process and they're being blown away when
the split valves open.
Now, to get back to my favorite topic, linearity.
Slide 5 is the response of phenol as a function of concentration.
This axis is the area generated for the quantitation mass for
phenol, versus the concentration at five points...20, 50, 80, 120
and 160. The sad part about this curve is, as I was;
hoping that I would be showing this type of curve for the
splitless injector, and a straight line going through the origin
-------
199
for the on-column injector. But, as you can see, they both have
the same shape. I started this study with a wonderful plan about
how this was going to solve the world's problems. But the fact
is, the non-linearity in this case is not associated with the
injector.
The RSDs that you see printed on the curves are the
traditional EPA calculation for response. The percent RSD of 18
for the on-column injector and 4 for the splitless injector is a
measure of the linearity of the lines. I should point out here
is that phenol is a fairly polar compound. It elutes early in
the chromatographic program and the splitless injector and the
on-column injector yield primarily the same results.
A little later in the chromatogram, fluoranthene
exibits much the same behavior. RSDs are a little different,
but, not significantly. One is five and one is eight. They
certainly meet all EPA criteria.
But if you notice the absolute intensity of the peaks
for the on-column injector, they are almost twice that of the
splitless injector. What that's saying is that even for a
compound like fluoranthene, the splitless injector is
discriminating by about a factor of two.
It should be pointed out that these runs were set up so
that all voidables were controlled as closely as possible. We
literally took the column off of one injector and put it onto the
other. We changed no mass spectrometer conditions, used the same
auto sampler for the two runs and ran them essentially back-to-
back.
I should also point out that the splitless injector has
a pressure programming feature and as a part of that pressure
programming feature, we were pressure bursting the front end of
the chromatographic run so that we would help drive as much of
the material out of the injector and onto the column as possible.
We were able to put this pressure ramp in the front end of the
chromatogram to help transfer as much of the material onto the
column as possible. The standard injector on the HP-5890 does
-------
200
not have this feature and therefor the factor of two is probably
closer to eight or ten in terms of the transmission efficiencies,
especially for the later eluting compounds.
Pentachlorophenol behaves similarily to fluorthane.
Finally, I thought I'd look at a base. This is 4-
nitroanaline and the results again are about the same. You see
this difference of about a factor of two. If I were to show the
late eluting polynuclear aromatics, such as the benzo/pyrenes and
perylene, there would be about a factor of three, using the
pressure burst injection and without it, a factor of eight to
ten.
Not everything is rosy, however. This is total ion
trace for three of the analytes at three different
concentrations. This is 20 nanograms, 60 stet nanograms and 160
nanograms. The three analytes in this trace are benzyl alcohol,
2-methylphenol. Between the 20 and the 80 peak broadening. At
160 there's radical peak broadening occurring. We puzzled over
this and puzzled over this for sometime, played with the length
of the pre-column and finally when all else failed consulted the
literature. Sure enough, this was expected. It is suggested
that if you use a thicker film column and this problem will go
away. Sure enough, we put in a thicker film column and the
problem went away.
This is basically column overload because of the
injection efficiency of the on-column. We were using thin film
columns initially and had to change to thick film.
This slide shows the same set of data with splitless
injection some peak broadening is observed as you go to higher
concentrations but the peak broadening is much less than with the
on-column.
As a result of what we have seen, we are limiting on
column to low concentration work, drinking water type analyses as
opposed to hazardous waste and higher concentration materials.
The title of this talk mentioned on-column injection
and the HP 5971 GC/MS. When I put the title together, I really
-------
201
thought we were putting the on-column system on a 5971. It turns
out we put it on a 5970 and used the 5971 for looking at low
concentration, high sensitivity analyses such as Method 525.
I've got four slides here I'll just run through quickly
to show that the 5971 works quite well for this application.
I'm not sure the 5971 is the answer for Method 525
since the detection limits are challenging.
This is a linear regression response curve for
phenanthrene. This axis is the amount ratio that represents the
amount of material injected, divided by the amount of the
internal standard. The internal standard concentration is five
and the concentrations injected were from . 1 up to 10 or two
orders of magnitude of concentration range.
This axis is the area ratio on the area of the analyte,
divided by the area of the internal standard. This results in
some normalization correcting for injection vandelites.
Method 525 allows for a little flexibility in terms of
response since you can use linear regression or even a polynomial
without having to force the origin. As a result it is probably
one of the more progressive EPA methods. Phenanthrene is an easy
material; it chromatographs very well and has a very strong
molecular ion.
This is a spectrum of the low concentration point of
the curve, 100 picograms of the phenanthrene. Note the molecular
ion is clearly visible.
Hexachlorobenzene, a little tougher compound
to deal with. It's got a lot of chlorine on it and doesn't like
to form positive ions. Again, the response is reasonably linear.
Again, it fit with the linear regression.
Finally, the spectrum. A lot more grass on the
baseline. You've really got the gain cranked up on the
instrument in order to achieve these detection limits.
We've been pleased with the system. It's worked out
very well and I think there are some real applications for it.
-------
202
QUESTION AMD ANSWER SESSION
MR. PRONGER: Greg Pronger,
National Environmental Testing.
Have you tried HP's nanoliter adapter in the injector?
MR. BEIMER: I don't even know
what it is.
MR. PRONGER: One of our
laboratories has been playing with the same technique for the
standard 8270 type of analysis and what they've found is they get
almost scary straight lines when they go down to a .2 microliter
injection, taking advantage of the better transfer of analyte
into the column. They're making .2 microliter to .5 microliter
injections. Some of the problems with the percent RSDs just
completely went away. They saw no increase in loss of the more
sensitive analytes like 4-nitrophenol, 4-nitroanaline, penta-
chlorophenol. They remained with very strong response at the low
end standards and very good results that way. It might be
something worth looking at in that aspect.
MR. BEIMER: We'll try.
MR. BEACH: Richard Beach from
Hydro-Systems.
On your cool on-column work, you're showing the
difference in the response curves there. Was that with the
pressure programming or without? Did you try just cool on-column
without the pressure?
MR. BEIMER: The cool on-column
was without pressure programming. MR. BEACH:That's what you were
showing there?
MR. BEIMER: Yes.
MR. BEACH: And so if you used
pressure...
MR. BEIMER: The only pressure
programming that was used was on the splitless injector. We did
not use any of the other pressure programming features.
MR. BEACH: So you didn't use the
-------
203
pressure programming on the on-column work either?
MR. BEIMER: Right.
MR. BEACH: Thank you.
-------
UJ
204
ffi
<
O
(5
1
h-
O
W
Q.
UJ
x
^^^
u.
O
QC
O
QC
Q.
Ul
QC
D
£
<
QC
UJ
Q.
in
QC
O
s
UJ
"Z
"Z
3
-J
O
DC
E
OC
2
O
n
O
O
UJ
Q
(0
•«
Q
UJ
m
O
(0
<
UJ
O
UJ
0.
Q.
O
UJ
-------
205
hQ
§!
8i_
X£
It
COO
s
2
-------
206
4>
0)
I
o>
o
5
ffl
N
-------
207
m
i.
Gtf
3
o
-------
208
c
3
O
.2
CO
CL
o
o
55
8
w
Q.
V)
0>
|
i
s
"EL
CO
O
O
CD
3
•
* 75
1 £
> CO
% £
M
ii
§
ill
E E
ill
S
m
1 * ?
2 S .3
l
c
0
: I
e j» £
*HI
liii
i a
c
18 o
a.
0
II
f£
^ 3
« O
•
*5?
•
£
•^
c
3
(5
E
3
E
c
-J
]1
£ •.
SI
i:
o
a.
E =
r
c
at
V*
"Z
o
-------
209
240000^
220000J
200000^
leoooo:
160000;
140000^
120000^
100000^
6000O
400001
£000<>
0^
&
' 12" ' •
I
111
1
t
II
1
16 ' 20 ' 24 ' 26 ' 32 ' 36 ' 40
•100
•90
-80
-70
-60
•50
-40
•30
•20
-10
-A
44
SPLITLESS CHROMATOGRAM (80 ng)
-------
210
280000-
260000-
240000-
220000-
200000-
180000-
160000-
140000-
120000-
100000-
80000-
60000-
40000-
20000-
8
! .
1
1
1
fl
1
1
J
1
I J
11
1
M
11 1
Jiuni
1
1
F
1
,
•100
-90
-80
•70
-60
-50
-40
-30
-20
•10
.A
12 16 20 24 28 32 36 40 44
ON - COLUMN CHROMATOGRAM (80 ng)
-------
211
i
5
l
O Q
I
o
<0
CM
1*
i
2
00
o
10
o
CM
I I
I I I I
s
CM
g 5 8 8
Y- T- T- T-
B8JV
8 § Q
CM
-------
212
CM
00
D)
C
o
I
U)
CM
88888888
o> co r»
-------
213
I
O Q
I I
CM
O
-I
t
c
o
7
CB
4^
§
o
in
CM
CM
? 8
(O
g
CM T-
B8JV
-------
214
I
- S
1
8
I
CM
J I
s
CM
s a §
S § 8
-------
215
8000-
6000-
4000^
2000;
0-
2-Methylphenol
41
4-Mathylphanol
1 1
-..
11
12.2 12.4 12.6 12.8 13.0 13.2 13.4 13.6 13.8 14
-100
-80
-60
-40
•ao
-0
.0
12000-
8000-
4000-
80 ng
-100
-80
-60
-40
12.2 12.4 12.6 12.S 13.0 13.2 13.4 13.6 13.8 14.0
12000-
8000-
4000-
•100
-80
-60
-40
-20
iL0
12.0 12.2 12.4 ' 12.6 12.S ' 13.0 ' 13.2 13.4 ' 13.6 ' 13.8 ' 14.0
ON - COLUMN RESOLUTION WITH INCREASING AMOUNT INJECTED
-------
216
6000-
4000-
2000-
2-Methylpheno)
Banzyl Alcohol
4-Methylphenol
13. 13.2 13.4 13.6 13.3 14. 14.2 14.4 14.6
•100
•30
-60
-40
-20
-0
20000-
16000-
12000-
3000-
4000-
0-
A A
i
A
A
\
/ L / V /
80 ng
^
13.0 13.2 13.4 13.6 13.8 14.0 14.2 14.4 14.6 14
•100
-so
-
•60
-40
-20
-0
.8
30000-
20000-
1000'>
Q_
i:
fi
X
2.S 13.0 13.2
/I
fl
V /
13.4 13.6
fl
v /
13.8 14. C 1-
160ng
•-
t.2 14.4 14.6
-SO
-60
-40
-20
-
•0
SPLITLESS RESOLUTION WITH INCREASING AMOUNT INJECTED
-------
2.0—,
1.5—
217
DC
"c
:a
o
1.0 —
0.5 —
0.0
I I T
0.0
0.5
' I ' '
1.0
i i i r
1.5
Area Ratio
PHENANTHRENE RESPONSE CURVE (METHOD 525)
-------
218
-------
219
c
o
2.0—1
1.5 —
1.0—
0.5—
0.0
I | I | I | I | I | I 1 1 | T
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
Area Ratio
HEXACHLOROBENZENE RESPONSE CURVE (METHOD 525)
-------
220
o
o
m
CO
fvl
f»
(M
c5
f
J
-------
221
MR. McCARTY: Our next speaker this
afternoon is Margaret St. Germain from Midwest Research
Institute. Margie has been doing a lot of work with Bill
Telliard's program, certainly with isotope pollution methods over
the years. But she's taken a really unique approach. She's
actually looked at historical data. When you go back and look at
the data, you've got to answer all sorts of questions about what
happened, and I think Margie's going to tell us, maybe, things we
don't want to hear. But she's going to compare some historical
data for complex samples in which the isotope dilution methods
did not work very well, I think, and look at the criteria that
are specified in the latest revisions, Method 1624 and 1625C.
Margie gave me a little bit of a background for an
introduction and one of the things that I find very interesting
is that when she got out of college, she had one job interview
and was immediately hired at MRI and has stayed there ever since.
That kind of longevity, I think, is also a reason she can look at
historical data, because she's got some. If you bounce around,
you can always blame it on..."well, that was the old lab and they
didn't know how to do it". So, hopefully, Margie will be able to
tell us something about what's wrong with the criteria, as well
as what's right.
-------
222
THIS IS MX UNEDITED VERSION OF THE PRESENTATION BY THE SPEAKER
MS. ST. GERMAIN: Good afternoon.
I thought I'd start out by telling you that I'm a quarter Polish
and I thought there was a rash of Polish jokes awhile back and I
thought that had died until I got to the conference and I heard a
whole bunch more. So, just to let you know that during that last
rash of Polish jokes I heard about a gentleman who decided that
with all of that publicity, he thought that he'd become Polish
because he thought there had to be something good about Polish
people. And so he looked around and he finally went to a surgeon
and said, well, I really want to become Polish. Is it possible
and what does that entail?
So, the surgeon said, well, you know, it's an awfully
serious operation. I have to remove half of your brain.
And the doctor said, I'm not going to let you make that
decision. You go home and think about it.
And the guy thought, well, it's possible. And he got
real excited and said, I definitely want to become Polish.
So he went back to the surgeon and said, let's do it.
And so they scheduled the surgery and got through the
surgery. Everything went smoothly. The guy was recovering from
the surgery and was starting to wake up and the surgeon came in
and said, we made a very small mistake.
The guy just kind of looked at him and the surgeon
said, instead of removing half of your brain, we removed 95
percent of your brain.
And the poor old guy just looked at him and said, mama
mia.
So, this was the wake-up call. So, I'd like to get
onto some of this data. When I got into it, I thought that sure
enough I've got this database to look at all of this historical
data. It should go real smoothly, very fast. Needless to say,,
it's taken me three months to get through about 18
compounds...just trying to get all of the data loaded, organized,
-------
223
go back and check for all the problems. So what I'm going to do
is try to give you an overview of some of the things that I did
find and also remind you as it compares to two other specific
types of methods.
The other thing I should let you know is I have been
doing isotope dilution for two and a half years. In that time,
I've analyzed over 550 samples from six different matrices,
including pesticide industries, pulp and paper, creosote paper
plant, municipal sludges, Superfund sites, and oil and gas. In
trying to collate this volume of data, I ended up focusing on the
OPR criteria to start with because I knew that if that did not
meet the method, then we should look back at the method. My next
step would be then to go back to these different matrices and see
what kinds of recoveries I got on my surrogates for each matrix.
And so I was concerned to see how close to the method I really
was and what can happen over that two and a half year period.
Just to give you an overview, what I've done is set up
in the next few slides references to two other methods... two
other series of methods, the solid waste methods, including 8240
and 8270 for volatiles and semi-volatiles, respectively, and
those are for solid waste and for the water samples, the 624, 625
samples and then, of course, 1624 and 1625 for complex samples.
Now, as you notice in this slide, the comparison of
this data, what I focused in is what the QA requirements or
criteria were. So on this slide we've indicated what the
internal standards are, how many there are, how many surrogates,
how many target analytes and this is specific for the analytes
that must pass criteria, additional compounds and tentatively
identified compounds.
If we go through the volatile method, there are three
internal standards for each of the methods, although for 8240
they are different internal standards. The surrogates will range
from three for 8240 up to eight for 624, although they do
recommend a minimum of three surrogates. Up to 31 for the
isotope dilution method. So, already we're generating a lot of
-------
224
volume of data just for volatiles.
The target analytes, on the 8240 there are 10 analytes
that I've listed and these are what they call the SPCCs and the
CCCs for the system performance check compounds and the
calibration check compounds. All of the other compounds listed
in that method are recommended to meet the criteria. They are
not required. And so that's why I've listed them as I have
there.
Okay, the isotope dilution methods, the 1624, 1625,
also list criteria for how to identify additional GC peaks, which
is not found in the other methods. The same kind of information
is available for the semi-volatile method, as listed on the
table.
Going on, let's look at the internal standards for
volatile methods just to point out that the first two methods
have the same internal standards. 8240 maintains the
bromochloromethane...
Hmmm? No. Uh-oh. Okay. There's one slide out of
line, so please try to remember that one. I'll discuss that
later. Okay.
Bromochloromethane is the only common internal standard
to all three methods.
Going onto the next slide is the internal standards for
the semi-volatile methods. And you'll notice that two of the
methods only require one internal standard. 8270 requires six of
them.
The other thing we want to look at are the surrogates
for the volatiles methods and you'll notice that 8240 has three
surrogates, 624 has up to eight, although not all eight are
necessarily used. But keep in mind that isotope dilution, which
you can't read, has 31 of them and this just starts giving you an
indication of what kind of QA is actually built into the method.
Okay, if we go on and do the same thing for semi-
volatiles, we have a similar situation with six surrogates for
8270, up to 17 for 625 and 81, which takes up two slides to even
-------
225
half-way read the slide, with about 81 surrogates.
Okay, that one slide that was out of line, that had to
do with developing valid data and you have to have several pieces
of information about QA. What I'm going to focus on is
instrumental QA and method QA. So, the next couple of slides
will focus in on that.
The instrumental QA basically has to do with the
calibration of the instrumentation and monitoring of that
calibration via an internal standard. And so, all the methods
require some kind of mass spectrometer calibration, including the
mass assignment and the mass ratio or mass intensity. The
internal standards, they monitor area and retention times for two
of those methods, although one of them does not specify any
criteria.
The gas chromatographic criteria has to do with
resolution more than anything else and the 8240 does apply some
of that by using CCC compounds. The isotope dilution does have
that same kind of criteria for all of its surrogates and also
includes a resolution criteria of D8 toluene-toluene. However,
that criteria right now is impossible on a packed column.
We've come up with a recommendation for a different
resolution that truly indicates degradation of the packed column.
We can go and look at the same kinds of information for the semi-
volatile methods and an additional item, if you go through it and
look at it. The additional item for the semi-volatiles is that
they've thrown in penta-chlorophenol to monitor tailing and that
is what monitors the degradation of the capillary column.
Isotope dilution has also taken that same standard, added ODD and
is looking for degradation of the standard, as well as the column
chromatography. When DDD degrades, it goes to DDT and so they're
looking for additional degradation of the standard and
degradation of the column within one's standard mix.
Okay, that was the instrumental QA. Now we're going to
the method QA and as we can see, the isotope dilution
methods...what I call isotope dilution methods or the 1600 series
-------
226
methods...have much stricter...have a lot more criteria, although
they are not quite as strict and I'll be able to show you how
that can be done. So, we're looking at initial calibration which
all the methods require, and a continuing calibration. Again,
most of the methods have very specific criteria that are written
into the methods. A lot of additional criteria are added because
of the statements within the methods that you must maintain a QA
program and as part of that program, some of the additional QA
that are in other methods are being incorporated into running
624.
Okay, along with that method QA you also have to
monitor that QA for samples and some of the things that they
monitor are the internal standard areas, the internal standard
retention times, surrogate recoveries, and target identification
by retention time and by spectra verification.
All the methods do require some kind of matrix spiked
samples...at least recommend, if not require, and duplicate
samples. And that's just to observe some of the statistics
involved with each matrix.
Okay, we can look at the same kind of thing for the
semi-volatile methods, including initial calibration with the
curves, IPRs, number of samples, continuing calibration. What I
found interesting is that two of these three methods did specify
a maximum shift length where at the end of that shift you must
perform more QA in order to continue past that number of hours.
Going on, we can look at some of the sample criteria
and, like I said, this is just a quick review to give you an idea
on the perspective of what we're trying to do here.
As far as that information, what we do is we receive
the sample, extract it, analyze it, transfer the quantitation
data to a PC that has a database and that's how we can monitor
our ongoing precision and recovery and check to see if we're
within specs or not. And so, what we've done is we've taken the
information from the quan reports...I'm ready to strangle this
thing...and for each compound in the quan reports for the isotope
-------
227
dilution methods, we have 10 items that are collated into that
database for each compound and they include the sample file name,
the EGD number, the compound name, retention time in seconds and
scans, relative retention time, area, the amount calculated by
the isotope dilution method, relative response factor, and the
average response factor from the current curve for that data.
In compiling the data that I did just for the ongoing
precision and recovery standards, I had 231 volatile files for 92
compounds. Of those, only 67 of them I have enough data to even
look at. The other 29 are additional compounds that do not have
necessarily very strict criteria.
For semi-volatiles, I had 171 files for 257 compounds,
of which 157 of them I really do have lots of data. So if you
look at that, I've got on the order of 641,000 pieces of
information in the database and it's taking up about 35 meg of
space in my 40 meg hard drive, so I'm running on a very tight
system right now.
Okay. Some of the key features of the data that you're
going to see is that this data was acguired over a two year
period. We generated several calibration curves based on failing
curve or daily responses. We used several columns and on
occasion, we have been able to pass previous curves because we've
been real careful about the capillary columns we've purchased.
We've also been careful about trying to reproducibly pack our own
packed columns. We've even collated the data from several
instruments for the same compounds and for each OPR summary that
you will see, there are over 150 points.
When looking at all that data, it was obvious when
there is truly a problem and when it was a statistical outlier.
In the data that you will see, the X axis will always be the file
number, the Y axis will change and I'll try to discuss that as we
go through some of the data.
We're going to look at the internal standards for
retention times in areas and three compounds from volatiles and
semi-volatiles. I purposely picked one CCC and one SPCC so you
-------
228
can see what some of the relationships are between the other
methods.
Okay, so this next slide is the retention times of the
three internal standards and the method specifies very specific
ranges for retention times for those internal standards. The Y
axis in this case is the scan number which converts to...each
scan is three seconds. And so, the bromochloromethane, the lower
line of dots, has a requirement of 653 seconds to 782 and for the
most part, we did meet the criteria. There is one outlier that
I'll discuss in a second. The same is true for the
bromochloropropane and then, finally, for the dichlorobutane.
They all have method criteria, specific retention time windows.
The one outlier...! went back and looked at the file
and the notebook read, started the run...computer didn't
start... started the computer acquisition 30 seconds late.
Therefore, we have 30 seconds to add onto that to make it more of
a real time and it does fall in at that point.
And I thought, well, you know, this is nice. It shows
straight lines. But what does that tell me about my internal
standards? So, I decided to look at the areas and they're really
spread out. And I started looking at some of the data and came
up with an additional criteria that my lab is starting to
recommend and that's for bromochloromethane that the area counts
be between 10,000 and 30,000. Below 10,000 we notice that we
start having detection limit problems. Above 30,000 we notice
that we have saturation problems for the high level standards.
And so if we can maintain that internal standard within a tighter
internal standard area, then we should be able to have a lot more
reproducible data.
I looked at the same kind of thing for 2-bromo 1-
chloropropane and, again, they're spread all over the place and
came up with new limits for that also and they end up focusing on
similar ranges, although the absolute intensity is much higher
for this compound.
And then we looked at 1,4 dichlorobutane. The first. 60
-------
229
files or so are significantly different from the scatter that you
saw previously and we are currently trying to identify what the
possible difference is for that.
Okay, so let's go on and look at an SPCC for volatiles,
tetrachloroethane. This is real data with nothing thrown out.
The center solid point that's way above that you can see... The
hollow points you may not be able to see. That point...we went
back and it's a labeled compound. We discovered that it was
calibrated incorrectly. When we recalculated it on the correct
peak with the correct areas, then it fell in line. There are two
additional hollow points and they were standards that were run at
the end of the day after high level samples, so it's obvious that
there was carry-over. But in looking at the data, the next
morning, the point fell in and so therefore the system cleaned up
overnight and we were ready to continue on.
That's a lot of points to look at, so what I've done is
focused in on two areas. One is from file 130 to 160 and the
first point that's low was again an end of the day standard where
we had some very dirty samples. You can see the very next day
the point is back up again. You also notice two more points
later on that were low and we discovered that our column was
degrading. We packed a new column and we were again to pass the
same curve in this case and we were back up to where we were once
we had packed a new column.
Tetrachloroethane is one of those compounds that we
analyzed on two separate instruments during the month of
February, 1990, and, as you can see, all the solid points are
scattered generally over a range. The hollow points are the same
and the hollow points are all the natives on two different
instruments and the acceptable criteria range for
tetrachloroethane in the method is from 7 to 34 micrograms, which
we definitely passed in this case.
Going on to a CCC, again we have wide scatter. We do
have a couple of points that are outside of the criteria.
Acceptable ranges for the labeled compound...the method says,
-------
230
detected up to 65. After the method was written, then they
changed the labeled compound concentration to 100 micrograms and,
of course, then you fail the criteria. So, by multiplying that
range by a factor of five, which is how they raise the standard,
then you have detected to 325 to monitor as a criteria base that
your labeled standard should be, right around 100.
What we've done is looked at files 130 to 160 again.
The hollow points are staying fairly consistent, right at 20
micrograms, where they ought to be since our native is at 20
micrograms and you can notice that there's a general downward
trend of the labeled compound at the conclusion of that
particular run and then we made up a new solution and started
over and we are back up at 100 micrograms again. So, we've been
using that as a criteria for how far it degrades and then change
over rather than just go until it goes to not detected.
Okay, we've got similar data on two instruments. And
again, you can see that the general scatter is within an
acceptable range for the two instruments.
Going on to chloroform, this was one of the few
compounds that I did not have any outliers and I was, frankly,
surprised. The acceptable range was eight to 30 micrograms and
concerning how tight those spots are or those data points, it
looks real good to me, concerning some of the other plots that
you've just seen. However, I do have about eight outliers within
that two year period that need to be addressed that I haven't
gone through yet to identify the problems.
Let's focus in on a couple of areas. From file 130 to
160 again and within that criteria, within that set data, we did
meet the criteria all the time very easily. And I got to
thinking, say, well, how does this compare to the 624 Method arid
the 8240? So, what I did was I took chloroform. I have shown
the response factor for chloroform, calculated by the isotope
dilution method. The response factor should be right around one
and has been. I do have one outlier in that respect.
Taking that same data and recalculating the areas,
-------
231
based on an internal standard method using bromochloromethane, we
can see we've got quite a few more outliers. Most of them are
right around a general response factor of one. But towards the
end, we start seeing a lot of scatter. So in that case, it's not
working as well as the isotope dilution method.
Okay, going into the semi-volatiles, we'll go through
the same kind of data and I'll probably go through this fairly
quickly.
We have the retention times for the internal standard,
1,4-difluorobiphenyl and as you can see, we did meet the criteria
of 1,078 seconds to 1,248 seconds. Again, I looked at the areas
and decided that any areas above 600,000 counts, which we've got
about four points on that graph, I start having difficulties with
saturation of the mass spectrometer for higher levels.
Okay, I'm going to look at phenanthrene, which is a
good reactor. And as you can see, the criteria for the method on
the semi-volatiles the labeled compounds have a wider range. The
native compounds are much tighter. And so, the labeled compounds
for phenanthrene have an acceptance window of 34 to 168
micrograms, based on a 100 microgram standard, and this is
micrograms per liter. The natives, however, are from 87 to 126,
so we're looking generally at a plus or minus 20 percent for the
natives.
Going on and looking at a selected window from files 40
to 60, we can see that was real easy to meet. Things were fairly
consistent, although I do have a section where things didn't work
so well and I do have two outliers. The first one at this point
I cannot explain. The second one was at the end of a day of a
dirty sample and so we were seeing some things...but the very
next day we were back in spec. So, we're seeing that something
does happen if you run shifts back to back...that there are some
difficulties.
The next thing I tried was 4-nitrophenol...first off, a
bad reactor and also a SPCC. And I do have quite a few points
that are above the majority of the points, the solid points being
-------
232
the labeled compounds and the hollow points being the native
compounds. And again, for the labeled compound on this 4-
nitrophenol, the method says there's no specification for what
kinds of numbers you get. What's important is that your native
compounds must meet a 51 to 178 microgram per liter criteria.
Looking at that, we obviously had a small problem
towards the end of that section from about file 135 to 150. In
looking at that data, all the acids failed. At that point, I
refused to run acid extracts. However, all of my base neutrals
passed. So at that point, I ran only base neutral fractions in
order to try to meet holding times and at the same time meet
specific criteria.
Another compound that I looked at was
hexachlorobutadiene. The criteria for the labeled compound was
not specified on the low end. It went up to 413 micrograms on
the high end. The native, however, was from 43 to 287 micrograms
and, as you can see, for over 150 files, I did meet that
criteria.
I decided to do the same kind of thing to compare this
to the other two methods by response factor information since
that's the one thing I could manipulate. And so, here we had the
response factor for hexachlorobutadiene using the isotope
dilution calculation and we do have a spread, but generally most
of the points are between 1 and 2.5. Calculating against D,10-
phenanthrene, we have a range from 0 up to .6 with most of the
points being around .3. Okay, when you run into response factors
this low, you start running into difficulties in quantitation and
reproducibility. I went and looked at that against D,8-
naphthalene which would be the 8270 Method and saw a similar
situation, although in this case, I did find a lot more scattex.
So, in conclusion, I was surprised at the number of
points that did meet criteria and the ones that I could identify
what the problem was either related to previous sample analyses
or a bad column situation. I was able to identify additional
internal standard area criteria that should help me monitor
-------
233
what's going on in real time a lot sooner than the day or two
later and have to rerun a set of samples. With the database that
I've developed, I can compare this data to other methods since
most of the surrogates and internal standards are already in the
isotope dilution mixes. So we have the potential of re-
evaluating OPRs based on other methods, other surrogates and also
looking at different matrices and surrogate recoveries.
So I'd like to conclude with expressing my thanks to
Bill Telliard for sending me all of those scuzzy samples to deal
with and to Jim King and his group for sending us all the samples
and doing all the footwork for the organization and especially to
the MRI staff who performed the analysis. Without them, we could
not have gotten through that number of samples.
I'll take any questions now.
(No response.)
MR. McCARTY: Thanks, Margie.
-------
234
NO SLIDES AVAILABLE
FOR THIS PRESENTATION
-------
235
MR. McCARTY: Our next speaker is
going to be Lance Steere, who is also from S-Cubed Laboratories.
Lance tells me he's a member of an endangered species; he's a
native San Diegan. Apparently, the reintroduced species out
there are getting pretty bad. The water shortage seems to drive
the native species away for some reason.
He's going to be talking today about heated purge and
trap GC/MS analysis as a way of dealing with the water soluble
compounds and alcohols that are often attempted by Method 1624C.
-------
236
MR. STEERE: Good afternoon. I'm
here to talk about heated purge and trap and isotope dilution.
What originally got me into this area was back during the
National Sewage Sludge Survey in the summer of '88 through the
winter of '89, S-Cubed analyzed over 200 sewage sludge samples
and I personally had the opportunity to run, about 150 of these
analyses myself on the volatiles which, of course, means smelling
them and dealing with them very up-front and personal. What I
came across in about 25 percent of these samples was the
simultaneous QC failures of the so-called water soluble
compounds. These are acetone, acrylonitrile, diethyl ether,
methylethylketone, acrolein and p-dioxane and it was interesting
to see QC failures among this same special group of compounds. I
sort of picked up the pattern after a couple of months of doing
sewage sludge samples and, gee, these same ones keep failing in
some samples and I wondered why. This really was of concern
because among the corrective actions for when you have QC
failures in isotope dilution methods, the first corrective action
is to rerun the sample at a 1:10 dilution. Of course, this is
not always necessary as far as analyzing the targets to bring
them in the calibration range. It's just a matter of seeing
whether the method is working on that particular sample matrix.
So this problem generated a lot of reruns, a lot of extra work.
Another thing you have to do for corrective action is
check your daily calibration, your ongoing precision and
recovery, or OPR as they call them in the isotope dilution biz.
Our system always seemed to check out, so it really did seem to
indicate a problem with the samples themselves and not with our
system. And, of course, if your dilution also fails, then you're
supposed to calculate the target amounts based on internal
standard quantitation, which is equivalent to tossing out the
isotope dilution method in favor of the old internal standard
method and the issue arose as to whether which set of data would
be more reliable, isotope dilution quantitated values or internal
standard quantitated values?
-------
237
Actually, the two sets of data weren't all that
dissimilar. They varied by about a factor of two, depending on
the magnitude of the XS recovery that the labeled standards were
showing.
Well, originally when I was running these samples, I
was pretty much convinced that there was some non-purged compound
in there that was somehow modifying the solvent and creating this
phenomenon of enhanced purgability. So, like Bob, I went ahead
and wrote up an abstract real quick saying that this was going to
be the topic of my discussion here today. Which was OK until I
went to do the actual experiment and nothing that I seemed to add
to the aqueous media I was testing seemed to give this same
enhanced purgability effect.
The only other plausible explanation that occurred to
me was the mild heating that we routinely gave these solid
samples. The method recommends that you run high solid samples
around 40 degrees, as opposed to room temperature and it helps
get all of the nasties off the solid matrix that you're running.
So, I ran a standard at 40 degrees and I observed excess
recoveries of these water soluble compounds very similar in
magnitude to what I observed with the sewage sludges. So, my
idea was to test the isotope dilution method under hot purge and
trap conditions.
My goals for this were several. First was to establish
that the hot purge and trap method would be useful in an actual
analytical situation. I wanted to verify that the detection
limits would be adequate. I expected that they would be better
since the hot purge should get these compounds out of solution
easier. I also wanted to verify that there would be a
sufficiently wide calibration range. In case it was too
sensitive, you too narrow a range and it's not really useful for
running routine samples. Secondly, I wanted to establish that
the method performance is satisfactory under the hot purge and
trap conditions and towards this end, I decided I would run an
initial precision and recovery analysis. And, lastly, I wanted
-------
238
to see if the isotope dilution quantitation method would correct
for temperature variances. I thought that it would, but I kind
of wanted to push the system and see just how much temperature
change could be tolerated. And then looking back at the National
Sewage Sludge Survey problem, I wanted to see if this information
might shed some light on what had been happening during the
sewage sludge excess recoveries.
For my tests, I focused on the six water soluble
compounds listed there and I also selected three alcohols.
Ethanol and isopropanal are of special interest in the
Pharmaceuticals industry and I've been asked on a number of
occasions to use hot purge conditions to analyse for these
particular compounds. Also, I had to limit myself to compounds
for which I already had labeled analogs available. After all, I
was on overhead for this whole project and couldn't really
justify to my boss buying a lot of expensive standards for this
project.
So, after setting up my compound list, and narrowing it
down to these compounds, I did the instrument detection limit
studies with the results shown here. Most of the compounds are
very well detectable around the 5-ppb range.
Acetone experienced somewhat of a problem due to the
use of bulk solvents in the extraction labs upstairs at S-Cubecl.
There's always a background of acetone in the air and you have to
go to sort of extreme conditions, such as pre-purging your water
and using heat to get the acetone on out of it. The background
level is around five to ten ppb, but if you took those
extraordinary measures, you could actually detect fairly
consistently in the one to five ppb range.
Ethanol and n-butanol both experienced some sort of
interference that I never really did trace down. Even if you
didn't purge anything on the system, you got peak areas that, at
the base temperatures I established at 75 degrees, would
correspond to ethanol at about 10 ppb and n-butanol at about
five...even if you didn't purge anything. So, there was some
-------
239
sort of interference going on there.
There's substantial improvement in sensitivity for
these compounds using the hot purge and trap conditions of 75
degrees.
Here I show the hot purge and trap curve at 75 degrees
that I ran three weeks ago with the average response factors
listed there and the relative standard deviations in those
response factors, as compared to a room temperature curve that I
ran for the PE sample for the recent isotope dilution RFP a
couple of months ago. Across the board, about an order of
magnitude improvement.
Notice that acrolein and p-dioxane, two of the ones
that show the lower sensitivities, show great improvement as far
as relative standard deviation once you bump up the response
factor to something that's a little more palatable to our system.
I went on and ran a calibration curve at that point. I
chose the same calibration range that the six water soluble
compounds get in the method, which is from 50 to 1,000 ppb.
Because of their general lower sensitivity in aqueous media,
they're given a calibration range that's actually five times
higher than all of the other analytes, such as chloroform and
tetracholoethylene and like that. Those are usually analyzed in
the 10 to 200 ppb range. And I got these results. Actually, the
sensitivity for four of the compounds, acrylonitrile,
isoproponal, diethyl ether and MEK were so much better that I got
a little bit of saturation on my system while running the 1,000
point. You can see some of the relative standard deviations for
the targets. Here's MEK at about 19, diethyl ether around 21.
Those are approaching what I consider to be the warning limit in
the method. The method allows you to use average response
factors, as long as your relative standard deviation in your
curve is below 20 percent. Above that, you have to go to all of
the trouble of deriving linear regression response factors or,
even worse yet, graphical response factors. Generally, linear
regression will work fine for these.
-------
240
If I exclude the 1,000 point, the relative standard
deviations drop by about a factor of two so some of the variance
displayed by the target compounds here is due to the 1,000 point
being biased somewhat low.
With a calibration curve in hand, I went on to an
initial precision and recovery analysis, which is four replicates
all spiked at the 100 ppb level. Ideally, you get a mean
recovery in all cases right around 100 and you can see that these
conform pretty well, except for acrolein which is really one of
the bad boys. We've only learned to detect acrolein on our
system just in time for the last PE sample. Historically, we've
never been able to see that one.
One thing I noticed...you might have noticed on my very
first slide, that when the labeled standards of the acetone,
acrylonitrile and MEK were high in sewage sludge samples, I'd
actually see the labeled acrolein. So I knew that it existed and
knew that it might be able to be detected on my system, but I had
to find the trick and eventually I did.
The relative standard deviations in the IPR's are all
much more than adequate. Anything under 10 percent is just
great. Up to 20 percent is good. Up to 30 percent is fair.
Around 40 percent is, well, marginal and 50 percent okay.
Actually, for the six compounds that are included in the method,
there are established criteria for IPRs and all of these
compounds pass. Even acrolein manages to slip under the wire
here. So, standard deviation criteria under the IPRs are really
rather liberal for these compounds, because historically they're
of lower sensitivity. Which goes hand-in-hand with higher
variability. But, under hot purge and trap conditions, you get
better sensitivity, so correspondingly lower variability in the
data.
Here's a really bad slide that I'm not going to bother
showing you because they said you ought to graph this stuff so
it's easier to see.
Here's the results I got, running different
-------
241
temperatures...seeing the temperature variability effect.
Here's room temperature at 25 degrees and the "boiling"
purge and trap at 90 degrees. The labeled alcohols...see,
there's a lot of condensation you get in the system. You have to
worry about water carrying over onto your trap and screwing up
the next analysis. You see, the labeled alcohols shown here all
show the type of response you'd expect to see.
Here's the reference temperature of 75 degrees where
all the calibrations took place, so you'd expect them to cross
the 100 line here at 75 and they do. At higher temperatures, you
get improved purgability and at lower temperatures, you get
miserable purgability. But the target compounds here all get
corrected with the isotope dilution method to be around 100,
which is the level where they were spiked.
Notice that ethanol and n-butanol really take off when
you get really low recoveries of the labeled alcohols. They are
biased very high. Ethanol is something like 240 ppb. This is
due to the background interference levels that were seen on the
system. If you have a small area that's there no matter what you
purge on, then that's going to have such a tremendous effect. By
the time you correct for less than 10 percent recovery of labeled
alcohol, the values there go sky high. But isopropanol didn't
have such a problem and so it's recovery is about 100 all
throughout.
Acetone, MEK and acrylonitrile all were very well
behaved compounds. They show the classic low recovery at low
temperatures, higher recovery at higher temperatures and about
100 at 75 degrees and all of the target compounds are corrected
adequately to show right around 100.
Here's diethyl ether and p-dioxane. P-Dioxane shows
the classic behavior, temperature dependent low recovery, high
recovery, the target compound biased somewhat low below 100, but
pretty constant throughout the range. Diethyl ether, on the
other hand, shows both the target and the labeled compounds are
about 100 throughout the entire temperature range. That's
-------
242
because the boiling point of ether is down around 34 degrees and
this may have had some influence on its getting out of the water
solution fairly easily. Maybe they should drop diethyl ether
from this water soluble list. You don't tend to think of it as
being water soluble, although it does have the capability of
hydrogen bonding, which is probably why it's included here.
So, the implications for my problems back on the sludge
survey... It appears that the high observed recoveries of those
labeled compounds was due to the mild heating and my results in
this study indicate that the target values reported using isotope
quantitation were probably perfectly adequate and therefore we
didn't have to go to any great extra effort to recalculate using
internal standard areas and so forth and, of course, skip all of
the reruns that weren't needed because of the target being
outside of the CAL range.
So, in general, what I want you to go away with today
after this talk is for you to know that the hot purge and trap
method works just fine with isotope dilution. You get very good
detection limits, very good calibration range, although perhaps
you get too much sensitivity at the high end and you have to trim
that a bit, the IPRs are as good as at room temperature, and it.
corrects for temperature variances just thumbs up...just great.
It's too bad I didn't have these results in at the time
when I did the National Sewage Sludge Survey because I had this
little voice in my head saying, trust your labeled compounds,
Lance. Use the power of the method. But, unfortunately, I was a
little bit more pragmatic in those days and went strictly by the
book.
Are there any questions? I guess I should have seeded
a few questions.
I thought of a question, why didn't I calibrate my
system for the 40 degree...
MR. McCARTY: Do you want to
identify yourself, please?
MR. STEERE: My name is Lance
-------
243
Steers and I work at S-Cubed and I ran some of these myself and
what I want to know, Lance, is why didn't you calibrate your
system at 40 degrees in the first place?
Well, the answer to that, young lad, is that the Sewage
Sludge Survey samples came in literally one or two a week for
several weeks and we were running them along with a lot of
aqueous samples. So, it was very convenient just to set it on up
with your calibration curve already in place. We really only
experienced troubles with a few compounds and a few samples, so
it didn't really seem to be calibration curve related. As I got
more and more into it, I finally began to see the pattern. But,
I didn't have time at that point to think about running extra
calibration curves for sewage sludge samples because we were then
doing 20 of them a week and that was just about impossible. So,
the problem kind of crept up on me and I didn't think of the
solution until I had to come up with a topic for the talk today.
No questions? Thank you very much.
-------
244
HEATED PURGE AND TRAP GC/MS
ANALYSIS OF WATER SOLUBLE COMPOUNDS AND
ALCOHOLS BY METHOD 1624
BY
LANCE R. STEERE
Senior Chemist
S-Cubed, a Division of Maxwell Laboratories, Inc.
-------
245
SIMULTANEOUS QC FAILURES FOR 'WATER SOLUBLE'
COMPOUNDS IN -25% OF SEWAGE SLUDGE SAMPLES
Labeled Compound
d6 Acetone
d3 Acrylonitrile
d10 Diethyl Ether
daMEK
d4 Acrolein
Acceptance Criteria
(ppb)
35-165
NS-204
44-156
42-158
37-163
Range of Values in High Recovery
Samples (ppb)
165-300
204-350
156-250
158-300
Observed*
'Usually not detected in any calibration or daily standard (OPR) analyses.
Instrument: Finnigan 4021 with Tekmar LSC-1 purge and trap
-------
246
CORRECTIVE ACTIONS FOR LABELED
RECOVERY OUTSIDE RECOVERY LIMITS ...
1 Rerun sample at 1:10 dilution.
2 Check system calibration by running another daily
standard (OPR).
3 Calculate target amounts based on internal standard
quantitation method.
-------
247
WHY THE XS RECOVERY OF LABELED
'WATER SOLUBLE' COMPOUNDS?
SOLVENT MODIFIER? - not supported by experiment
MILD HEATING (-40°) - observed XS recoveries
similar to sludges
Test the ability of Method 1624C to correct for
purge temperature variability
-------
248
GOALS FOR ISOTOPE DILUTION
HOT PURGE AND TRAP DEMONSTRATION ...
1 HPT method is useful for analyses:
• Adequate detection limits.
• Sufficiently wide calibration range.
2 Method performance is satisfactory under HPT
conditions.
• Initial precision and accuracy analysis.
3 Isotope quantitation can correct for temperature
variances.
• How much change in temperature can be tolerated?
• What do results suggest for sewage sludge survey
sample problems?
-------
249
Method 1624C Water Soluble Compounds:
• Acetone
• Acrolein
• Acrylonitrlle
• Diethyl Ether
• Methyl Ethyl Ketone
• 1,4-Dioxane
Selected alcohols:
• Ethanol
• Isopropanol
• n-Butanol
-------
250
INSTRUMENT DETECTION LIMITS (IDL) UNDER
HPT CONDITIONS (75°C)
Compound
Acetone
Acrolein
Acrylonitrile
Diethyl Ether
MEK
1 ,4-Dioxane
Ethanol
Isopropanol
n-Butanol
IDL
(ppb)
1-5
1-5
0.1-2
0.1-2
1-5
5-10
10-15
1-5
5-10
Backround level 5-10
High variability, raises effective detection
limit
Interference near 10-ppb level
Interference near 5-ppb level
-------
251
IMPROVEMENT IN SENSITIVITY FOR HPT CONDITIONS
VERSUS ROOM TEMPERATURE PURGE
Labeled
Compound*
Acetone
Acrolein
Acrylonitrile
Diethyl Ether
MEK
p-Oioxane
HPT Curve at 75°C
(04/15/91)
Average Rf
.815
.042
2.50
1.66
1.08
.169
%RSD
6.5
9.8
5.7
2.1
1.8
8.9
Room Temperature Curve at
25°C (01/21/91)
Average Rf
.101
.0064
.412
.957
.082
.011
%RSD
12.8
23.7
2.0
1.5
3.7
26.3
'Labeled compounds are spiked at the 100-ppb level in all CAL runs.
-------
252
HPT CALIBRATION RESULTS
Labeled Analog/
Compound
de Ethanol
Ethanol
de Acetone
Acetone
d4 Acrolein
Acrolein
d3 Acrylonitrile
Acrylonitrile
d3 Isopropanol
Isopropanol
d10 Diethyl Ether
Diethyl Ether
d8 MEK
MEK
d8 p-Dioxane
p-Dioxane
d7 n-Butanol
n-Butanol
CAL Range 50-1000 ppb
Mean R,
.103
1.76
.815
1.36
.042
19.3
2.50
.982
1.73
1.08
1.66
.859
1.08
1.27
.169
1.19
.177
1.28
%RSD
9.7
13.4
6.5
15.0
9.8
15.3
5.7
11.6
3.5
9.3
2.1
20.8
1.8
18.7
8.9
5.1
3.6
14.3
%RSD
(Excluding 1000 ppb
CAL)
6.6
5.4
11.6
11.4
-------
253
INITIAL PRECISION AND RECOVERY (IPR) RESULTS
UNDER HPT CONDITIONS
IPR - 4 replicates spiked at the 100-ppb level.
Labeled Analog/
Compound
d6 Ethanol
Ethanol
d6 Acetone
Acetone
d4 Acrolein
Acrolein
d3 Acrylonitrile
Acrylonitrile
d8 Isopropanol
Isopropanol
d10 Diethyl Ether
Diethyl Ether
de MEK
MEK
de p-Dioxane
p-Dioxane
d7 n-Butanol
n-Butanol
Range of Recovery
87-117
92-102
120-124
86-95
145-169
60-74
105-110
99-104
97-116
94-94
100-105
109-114
110-113
95-100
85-113
90-97
99-118
88-93
Mean Recovery
102
96
122
91
154
69
107
101
106
95
102
111
111
98
99
93
109
90
% RSD
13.6
4.8
1.4
4.0
7.1
8.5
1.7
2.2
8.6
2.4
2.0
2.2
1.2
2.6
13.1
3.5
7.9
2.8
-------
254
COMPARISON OF HPT IPR RESULTS WITH METHOD 1624C
IPR CRITERIA (SECTION 8.2; TABLE 6)
Analog/
Compounds
Acetone
Acrolein
Acrylonitrile
Diethyl Ether
MEK
p-Dioxane
Acceptance Criteria Versus HPT Results
Mean
77-153
32-168
70-132
75-146
66-159
65-135
HPT
122/91
154/69
107/101
102/111
111/98
99/93
Standard
Deviation
51
72
16
44
57
36
HPT
4
8
2
2
3
4
-------
255
HPT RECOVERIES UNDER VARIABLE TEMPERATURE CONDITIONS
All compounds and analogs spiked at the 100-ppb level.
Analog/
Compound
d6 Ethanol
Ethanol
d6 Acetone
Acetone
d4 Acrolein
Acrolein
d3 Acrylonitrile
Acrylonitrile
de Isopropanol
Isopropanol
d10 Diethyl Ether
Diethyl Ether
ds MEK
MEK
d8 p-Dioxane
p-Dioxane
dy n-Butanol
n-Butanol
Purge Temperature (°C)
90
189
95
160
88
199
46
119
97
172
96
100
111
123
101
171
94
176
90
80
171
92
157
84
237
39
119
98
159
93
106
108
130
95
163
91
165
88
75*
102
96
122
91
154
69
107
101
106
95
102
111
111
98
99
93
109
90
60
28
104
67
92
108
64
76
102
36
99
100
112
70
100
36
94
37
99
50
23
120
54
105
80
95
69
105
29
102
100
115
60
101
30
96
29
100
40
13
144
41
93
78
67
54
102
17
102
99
111
44
94
20
91
16
109
25
2
340
21
107
51
62
31
102
6
111
95
114
21
87
8
86
5
145
'Average of 4 IPRs.
-------
256
LU
QC
1
LU
CL
s
DC
B!
o
o
111
QC
O
o
o
o>
o
00
_ o
&
I
10
41)
•5
CL
_ o
o
CO
Zrf
I I I I I I I I I I I
I I I I
o
o
CM
O
CM
-------
257
LU
cc
1
LU
Q.
LU
5
DC
UJ
O
O
UJ
CC
LU
E
QC
O
LU
LU
O
LU
o
I I I I I I I I
§
O
(O
£
i
o
u>
I
-------
258
1
UJ
oc
s
oc
UJ
0.
CO
OC
UJ
O
m
UJ
z
o
5
i.
Q
DC
UJ
UJ
>•
1
i i
i i
O
O)
o
00
CD
O
10
O
o
2
2
Q.
c>
o
CO
1 1
1 1 1 1
1 1
o
o
CM
O
to
o
10
CM
-------
259
IMPLICATIONS FOR NATIONAL
SEWAGE SLUDGE SURVEY RESULTS ...
1 High observed recoveries of water soluble labeled
compounds was caused by mild heating.
2 Target values reported using isotope quantitation are
more reliable.
3 Recalculated values using IS method are not needed.
4 Reruns at dilution are not needed if original recovery is
within CAL range.
-------
260
GOALS FOR ISOTOPE DILUTION
HOT PURGE AND TRAP DEMONSTRATION
1 HPT method is useful for
analyses:
You Bet!
Adequate detection limits.
Sufficiently wide calibration
range.
Order of magnitude
better than RT!
May need to trim high
end!
Method performance is
satisfactory under HPT
conditions.
Of Course!
Initial precision and accuracy As good as RT!
analysis.
Isotope quantitation can correct
for temperature variances.
• How much change in
temperature can be
tolerated?
• What do results suggest for
sewage sludge survey
sample problems?
Like a Champ!
± 15°Cor more!
"Trust your labeled
compound quantitations;
Use the power of the
method!"
-------
261
MR. McCARTY: That's the problem
when you throw up a lot of data. People have to assimilate it.
Lunch is settling in at this point.
Our next speaker this afternoon is Bruce Colby from
Pacific Analytical Laboratories.
I'm going to be real nice to Bruce. I'm not going to
give him a hard time about the money he borrowed from me in
Europe last September or the grappa we drank until all hours or
anything like that.
Bruce has done a lot of work over the years on library
search results. He's one of the few people I know who drives to
work thinking about library search data, and goes to bed thinking
about library search data, and wakes up at 4:00 in the morning
and writes a paper, and actually submits the abstract before
there's been a call for papers.
He's going to talk about some improvements in the TIC
spectral assignments through a technigue known as peak centroid
analysis.
Bruce?
-------
262
MR. COLBY: Thanks, Harry. I'm
going to break this afternoon's tradition here in two ways.
First of all, I'm going to talk about something I did before I
wrote the abstract. I guess I just don't have enough imagination
to do it any other way. Secondly, I'm going to talk about
qualitative aspects of methodologies.
I don't really want a slide yet.
The qualitative aspects of any of these GC/MS methods
have to be dealt with before you worry about the quantitative
aspects. If you can't make a qualitative analysis correctly...in
other words, identify the compound correctly...there's not much
point in worrying about how well you can do your quantitation.
Well, the specific part of qualitative, spectral
dealings I'm going to concern myself with is the TIC area. The
technology I'm going to describe is actually applicable to
identifying overlapped target components, as well. But I'm going
to concentrate on the TICs because that's what I started out to
deal with.
The reason I got involved in the TICs in my own mind
recently was a consequence of watching people in a lab struggle
with trying to identify non-target analytes in relatively messy
samples, like isotope dilution samples or anything that our
illustrious Mr. Telliard might send us.
The problem we run into is that the non-target analytes
are often co-eluted or closely eluted with materials we've spiked
in, labels compounds and base neutrals. Margie said we had about
80 of them; that sounds around right and in the volatiles,
roughly half that many.
The problem we run into in terms of what happens when
we look at data, what I've got here is a nice peak on the left
hand side. It's nicely shaped, it's pointed, there's not much
background. The reaction to that is, yes, that must be some kind
of compound.
All of the examples I'm going to show, by the way, are
actually for targets. But I'm going to treat them as if they
-------
263
were not targets because going into it, I needed to know what the
compounds were.
You can see on top the spectrum we get for a retention
time of 14.03 minutes and the best library search is a spectrum
below it. At first glance there is a lot of agreement there, but
there are a few things that are a little bit off in left field.
You'll see there are some peaks out in this area in the library
that don't correspond with any in the sample. The same in here.
Worse yet, when you get into this collection of peaks, there's a
big 104 not there. There are differences.
The first reaction to it might be, yes, that must be
it. It must be this benzene ethanamine beta methyl. The only
problem is that we're looking at a purge and trap sample and
there's no way that could have come out of a purge and trap
device and made it through the column in that time frame. So
what we have is a mystery compound.
The problem with the TICs in particular is a
combination of things. There are, I believe, two issues in
particular that we have to fight. One is the database that we
search our unknown spectra against and come up with reference
spectra; it is incomplete. It's only got a little more than
50,000 or so entries in it and there may be a few million
possibilities. So there's not a very good coverage there. It
also has a lot of things in it that won't find their way through
a GC column and I thought some about trying to work on that, but
decided that was too big of a problem for me to deal with, so I'd
concentrate on another aspect of the problem and that is a lot of
times the spectra that we feed into our library search program
are the result of more than one component generating peaks in the
mass spectrum. In case you're not already guessing it, there's
probably more than one compound here and we'll look at that a
little bit later.
Well, if we've got a situation with two components,
either very closely eluted to each other or overlapping to some
extent that makes it difficult to use normal data system
-------
264
technology to extract clean spectra for each component, then
there has to be some other way that we can go at it, unless we're
just going to give up and say, that's an unknown. So, I thought,
what is it that we have? What pieces of information do we have
that we might be able to extract from the data file to give us
this additional handle to work on the library search? In other
words, how are we going to get a cleaner spectrum?
Basically, there are only two things we have. One is
that we can have sort of a supposition that if we have two
components next to each other, they have different mass spectra.
If we say that, then we have something we might be able to work
with.
The other thing that we can say fairly emphatically is
that if we have two separate components here, each of those
components will have its own retention time. Now, obviously if
they have a different enough retention time, you'll see two peaks
on the TIC. But I'm going to try to push that technology to
separate things that give you peaks like that.
What I'm going to do basically is to attack my data
file, in particular, a peak that I believe to be multi-component
and extract for each mass that has a peak associated with this
particular collection of masses, that spectrum up there...extract
a mass chromatogram and calculate for that mass chromatogram a
centroid corresponding to the retention time for the elution of
the component that that mass belongs to. We'll then do that for
each mass in the spectrum and see if I come up with more than one
group of centroids.
Now, before we get that far, it would probably be be-.st
to try something very simple and clean. So, take our old friend,
DFTPP, which we have many spectra of in the laboratory. This is
an NBS library spectrum. It's a nice spectrum to work with
because it's got lots of peaks across the mass range...down here
at 51 all the way up to 442. It's got little peaks, 365 and what
not. So, it's a nice kind of test compound.
I went to the lab, grabbed a floppy disk, hauled off a
-------
265
couple of data files and tried the centroid calculation and what
I've done here is plot that TIC trace for DFTPP and next to it
I've plotted the time at which the centroids for all peaks
greater than or equal to one percent of the base peak happen to
come out at. In other words, how closely do we calculate a
centroid for all of those roughly 65 peaks or whatever.
It turns out that the centroids for DFTPP masses were
within... Well, the standard deviation was .04 seconds. We were
scanning the instrument at one scan per second over sort of a
traditional BNA mass range. So, if in fact there was another
component in here, three standard deviations away, that would be
about .15 seconds away, we should get two groups of centroids
showing up there.
Now, if we know that mass X belongs to the first group,
then we say, okay, that intensity belongs to this compound and
then any mass that belongs to the second group of centroids
belongs to another compound and then all of a sudden we're
starting to separate these spectra, one from the other.
Just to show you what the spectra actually look like
for the DFTPP, we've got the...these are the centroids that
eluted within a plus or minus three standard deviation window of
the center of the centroid and this is that NBS reference
spectrum. Basically, they're indistinguishable from each other.
So, certainly we're not hurting the data by messing around with
these centroid things.
I thought, well, let me take something where I know
there's probably going to be a little bit of an
interference...something that will be eluted fairly closely. I
took a peak and in this case for... This is the TIC trace here.
It's a peak for D34-hexadecane. When I break that down into
groups of centroids, I get one major centroid and one little
bitty group right next to it at a slightly longer retention time.
In fact, it's .48 seconds difference. And again, you can see
that the peak up here, you know, would be pushing 10 seconds
wide. Anyway, this first collection of centroids produces a
-------
266
spectrum...a very nice spectrum for the D34-hexadecane.
The second group of centroids produces another nice
spectrum. Masses are all offset by one in the lower mass
direction, the less-deuterated compound having a longer retention
time than the deuterated. Well, obviously that's the impurity in
this deuterated hexadecane. There's a D33-hexadecane that was in
there when it got spiked into this particular sample and we can
see that we can actually separate those. They're less than a
half second apart in retention time. So, I'm starting to get
happy.
Let's take a look at that thing we had up there first,
the spectrum for one sort of peak that gave us a crummy library.
We run the centroid analysis and lo and behold, we get two
groups. Now, those people in the audience who are highly
familiar with volatiles analysis ought to already know what those
are. For those of you who don't fall into that category, I'll
show you the spectra... or try to.
Here is a spectrum of the first group of centroids.
Any volatiles analyst will pick that one out very quickly. The
second group of centroids, most of you should recognize that.
Anyway, if you don't, we stick these into the library search.
Here's the first component in our first group of centroids.
Actually, it's orthoxylene, so it's a 1,2-dimethylbenzene, but in
these spectra these are indistinguishable from each other. So,
now we're getting somewhere. We've got a pretty good fit on that
one.
The second group of centroids, the masses in there
correspond to this spectrum and what we've got is styrene,
another pretty good match.
Well, maybe that was too easy of an example, so let's
move on here and try something a little bit more complicated.
Here we've got a couple of components coming out here. I'm going
to concentrate on the more intense of the two. If we take the
data spectrum for the 19.61 retention time and library search it,
we get something that is clearly useless. We can look at the
-------
267
sample spectrum and we can say certainly there's chlorines in
here and, yes, this library search produced chlorines. But also,
there's no way that's the right compound. So, let's run out
little centroid analysis thing on it and we get a nice single
collection of centroids corresponding to this and here we have
three groups of centroids. So, well, let's have a look and see
what those are.
The first of this collection of three gives us this
spectrum. The second gives us this spectrum and the third one
gives us this spectrum.
Now, if we take those and submit that to library
search, we get a very good match for dibromochloromethane,
followed by a very good match for the 1,1,2-trichloroethane.
This, the middle one of the three, is a very match for
1,3-dichloro 1-propene. We know really what those particular
components are because, again, they are targets in this
particular sample and this particular sample is, in fact, a
volatile standard I've run on a packed column. The previous data
we worked with were volatiles run on a capillary column. This is
run on an instrument... about a five or six year old instrument.
The other one was a much newer instrument. It doesn't seem to
make much difference what kind of data we attempt to deal with.
Anyway, that was only three things. So, let's attack a situation
where maybe it's even worse.
Here we've got a TIC trace. It's got a shoulder on
there, so we know going in that we've got some kind of potential
interference situation.
If we take the spectrum for the peak top, 18.81, we get
a spectrum that has some peaks in it...all kinds of peaks in it.
If we library search it, it says it's a hydrocarbon. That
doesn't really look like a hydrocarbon. It looks like it may
have some chlorines in it out here. It looks like it may have
some aromatic character to it, based on those intense higher mass
peaks. One thing we can do with this, going in and knowing that
there's a shoulder on it, is we can run some spectral cleanup
-------
268
software on it. This is a technology called or basically
referred to as a Biller-Beaman cleanup. If we do that to the
spectrum, now we've...this is like a single scan enhance in a
Finnegan system. You can see that those chlorine clusters went
away and we still have some hydrocarbon character and aromatic
character; we still get a hydrocarbon answer that the spectrum
comes up with for the best fit. Still not too impressive, so
let's run this centroid business on it and see what happens.
Well, this one's even worse yet, isn't it? This time,
we've got one, two, three, four, five, six different components
showing up in this and since I don't have any good way to show
six spectra all at once, I'll just pick these three here in the
middle since they're the ones that are most closely eluted to
each other. Now we have spectra for those three middle
ones at the top. Here's a spectrum. It's about .9 seconds
later. This is capillary column BNA one second scan stuff.
About 1.3 seconds later, we get this spectrum with some
chlorinated things in it. That's kind of encouraging. So, we
send that off and do some library searching on it. The first
component that we looked at, biphenyl. Well, that's dead easy,
isn't it?
The second one...well, it's some kind of hydrocarbon.
They all give these sort of similar spectra, so that's about what
we do as we look at it and say, yes, it's a hydrocarbon.
And the last one, we get a very nice comparison with 2-
chloronaphthalene.
So, what we've managed to do is come up with a way that
allows us to take spectra for components that are eluted fairly
close together and, in theory, based on the standard deviation
for the centroids of the DFTPP spectrum, you would predict that
you have about a 99 percent change of getting a good separation
of spectra that are eluted as components...eluted as closely as
about .15 seconds for a capillary column run. I haven't found an
example that close. The closest we've managed to find one was
that deuterated hexadecane. That was about .48 seconds. But
-------
269
clearly, we're separating things that are coming very, very close
together in retention time.
So, now we have a better show at sending a good
spectrum off to library search and I think this has the potential
of improving our ability to do reasonable TIC identifications.
Well, what's the down side? There's got to be one; it
can't be all good. Two things...if two compounds have the same
masses in their spectrum, the calculated centroid will come out
between the two. So that's a little bit of an issue. The
styrene and ortho-xylene that were in the first example, if you
recall, weren't fully separated one from the other in the
centroid groupings. That is a consequence of some common low
mass ions.
If two compounds had exactly the same retention time,
they'll have exactly the same centroids and we won't be able to
tell those apart.
So, those are the two limiting factors on it that we're
aware of at this point. We continuing to pursue some other
variations of this and I expect you'll be seeing it in commercial
data systems, I would say within the year. It's been quite fun
working on it. If anyone has any kind of question to ask, I'd be
more than happy to try to answer.
-------
270
QUESTION AND ANSWER SESSION
MR. McCARTY: Bruce, one question
I have is in terms of practicality, what would you estimate is
the time increase this would put on TIC searches for, say, your
10 to 20 most intense TIC type of situation?
MR. COLBY:It appears that the
centroid analysis time, processing that, well, certainly it's
going to be system-dependent and data-dependent. It adds in the
vicinity of about four or five seconds per collection of things.
So, basically it will double the amount of time people spend on a
TIC.
MR. McCARTY: From four seconds to
eight seconds?
MR. COLBY: Four to eight.
MR. McCARTY: Any other questions?
MR. COLBY: Everybody knows how 3
did those calculations? I didn't think so.
MR. McCARTY: There will be a test
later.
-------
271
**
U
p-
ID
CD
^
00
H
H
- I
l-i
H
vfi
9
1ft
9-
-W
04
H
W
W
N
'
-p-
co
ID
P-
CO
P-
, • •
I*
ID
Ift
>;
j 9
J 9
^
'Z
H
A
I
X
•ii
CBN
9X
I
O
9e
X
-------
272
04
fc
•^
O
•**
•ft
I
O
U
in
Win
W
a,
&-
1*
53 5"
f->
o
§
-------
273
PH
IX,
§-4
t*.
A
«6
O
>f^
•ft
O
o
VOOH
in»-i*
V0H
«*>
oo
CO
00
8
SO
in
CM
•H
c
«*
s
cv
-------
274
Ok
&
-------
275
LO
CM
O
CD
•ft
o
•r*
CM
m
o
s
^
CTi
PQ
O
O
en
CM
CM
CM
H
"91:
CO-:
00.
00
ve
o
3
SO
CO
in
m
2
sfi
o
X
8
in
OH
o
p
-------
276
-------
277
O
A
+»
Q>
O
O
00
9
na
ou
: 9
:, 9
H
(A
•9
CO
:co
H
9C
IDH
-------
278
sfl
T3
O
0)
sOH
CO
9
CS1
9
9
fa
X
9C
00
00
51
••*
*
O
8'
M
00
cr»
•JD
CD
CD
S
I
O
s-
CS1
in
CO
in
CM
9
1-9
pa
O
b
CPl
r*.
..
r- -
SO -
SO
tA
ID
64
X
in
-------
279
U
It
U
ess
CM
CO
«
OQ
i
W
sD
^
*-»
cn
i
&
./
«*».
in
in
in
- o>
t-A
i
M
Z
M
.-P-
• v
fa
8
H
CO --
in.
in
00
CO
H
'"t«
• I
tt
X
rS
9
\-n
.9
X
tf!
4£
-------
280
2
**•>
tt
ts
o
8
-------
281
-------
282
^
9
tt
Q
•i>4
*P
-------
283
* ^
sfl
•g
U
U
• 04
CM
•p*
U
«
CL,
CD
LD
I
(S3
Os
9
9
C9
U
3Q
"»
00
SN
(5
CA
X
CSC
9
hN
N
00
OJ
9
boo
9
sot-so
9_
in
9
h9
ON
r-
00
so
SS
C
9
9
9
H
_OS
9C
X
00
9
-N
9
hoo
so-so
CO
r-
OJ
Ul
cr^
H
9
/— =
CTi W
00
9
•9
9
00
9
SO
^•S
9
00
9
SO
ON
-------
284
sJO
It
O
8
S
CS3
CM
J2
(0
CO
CO
oa
-p
en
w
Lf)
CO
9_
«
in
9
04
0>
O
X
o
«
0?:
t-g
-OS
S_j S
9
-Pfl
9
-9
04
9
-08
9
-so
w
9
OB
9
SO
tt
Sl. f
X
S
S
9S.
•HC
-------
285
TP
3
1
tf
I
U
U
it
a*
i
OJ
(A
<*•
o
"*•»
W
o
Ln
en
w
1 i"
K>X^
GO
tn
in
I
W
A
K
O
A
I
t£<
U
m_
H
8
N
Ift
CO
K
X
y.w
A
W
r
5
X
8
-------
286
O
U
•v*
CM
•P4
U
£
s
1
1
CD
LD
1
5
J=f?
s"
W
CO
CM
«
X
GO
v\J
8
sO
00
H\
2 °*~
LO 2~
O
•+•*
ro
in
t"T o
5^to ^
« »\
O 09v ""
oo
00
*^
LO
2
^^_ C^
2 SOx^N
T"^ wH —
^^ ** ®/
SSO
TH
•s
G^
CO
3
.ft .j. *
^1 H X
in
in
-
-
-
-
sO
00
-K.
w-
n-
- 1 ^
«
o
S
.«
P^ CM
.W (hv ^"5
.H B 5 s
i ^ON
w ^
-Z
X
H
" W
CO ^0\.
00 ^4 —
"S ** CB''
sO_
^*
"in pi-;
CO rti~
OB
p-
"ta w-^
H 8 ^ d
* H X
-
-
-
-
.
-
.
-
•
Z
H '
-ft
1
W
Z
-5
S s ' s ' a
W| 0 C
*H l^
-
OB
•
09
•0
04
S
-oc
C9
-sO
H
CB
H
8
-OJ
-------
287
in
03
in
04
H
W
00
w
<£
O
W
M
<*>
oa
00
<*>
W
in
w
a
X
s« S
K ®
(S
•HC
-------
288
LO
•JD
•rH
O
tt
O
•p*
^
?l
t—*
O
•|M
e«4
•^4
O
CX4
rS
.in
CD
-in
-8
9
S
.09
in
CON
-Hi
I
O
se
O
(A
-------
289
-------
290
LH
s£
O
Hi
so
0>
04
O
•r+
*P
jl»
00
*_
O
-8
8
Csl
8
SO
8
5
OH
V.
X C
00
00
SO
N
00
S
00
04H
CO
LH
*H
CO
sO
co
in
•8
8
04
CBN
X £
tt
04
CM
-8
:8
04
8
Hoo
V in
so *£
so-
Csl
8
f-so
8
8
h«
en
CO
09
8
3-2
r-.
oo
00
.8
SO
CO
8_.
X
C9H
£
-------
291
•w
O
•i-<
•ft
O
•p*
CM
•fN
O
OH
0)
CO
M
• •
M
r-
fsj
CO
s
*H
O
00
O
O
00
^^^
a! r
10
in
so
in •
in
in
04
in-
w
CQ
i
5
in
in
in CB
H in
in
8
H
SO
w
A
I
W
i—i
05
O
a
O
a
CA
fa
X
-------
292
in
CM
0)
o
••*
"f»
71
*
-------
293
in
0)
ITS
O
-------
294
MR. McCARTY: Our last speaker in
this session immediately before the break on GC/MS issues is
William Eckel from Viar and Company. Bill has been with Viar for
something on the order of 10 or 11 years. He started out as a
member of the group that works on one of our projects for
Superfund scheduling samples, and he has worked his way up to the
point where he can now deal with exciting topics like reverse
search compounds. Bill is also the two time apple bobbing champ
at the Viar Company picnic. He works very hard at this. He
spends a hell of a lot of time with his head in a bucket,
sometimes under water.
Bill, if you still want to go on, it's up to you.
-------
295
REVERSE SEARCH COMPOUND STUDY: METHODS 1624C AND 1625C
William P. Eckel, Thomas A. Jacob And Peter J. Isaacson
Viar And Company
Sample Control Center
300 North Lee Street
Alexandria, Virginia 22314
ABSTRACT
Four laboratories participated in an inter-laboratory study to establish Relative Response
Factor (RRF) and Relative Retention Time (RRT) data for 118 Reverse Search (RS)
compounds to be published in ITD methods 1624C and 1625C. The laboratories performed
five-point calibrations using standards of the Reverse Search compounds supplied by one of
the participating laboratories. The resulting data were reported to, and statistically analyzed by
the Sample Control Center. The results of the statistical analysis indicate that the best
qualitative and quantitative results for Reverse Search compounds would be achieved by
requiring each laboratory which uses these methods to establish its own RRF and RRT data.
INTRODUCTION
A inter-laboratory study was conducted in the first half of 1990, under SCC episode
1878, to revise the RRF and RRT specifications for the Reverse Search (RS) compounds in
methods 1624C and 1625C, which are presently based on one laboratory's data. One of the
participating laboratories obtained standards for the 118 RS compounds, prepared combined
standards for the two methods, and distributed the combined standards to the three other
participating laboratories for analysis. The study results were reported to the Sample Control
Center and were then subjected to statistical analysis to determine the suitability of the data
for establishing new RRT and RRF specifications for the RS compounds in methods 1624C
and 1625C.
METHODOLOGY
Laboratory Analysis
The participating laboratories performed five-point calibrations as specified in methods
1624C and 1625C, using the combined RS compound standards supplied to them. Data were
reported in both computer-readable and hardcopy formats.
Statistical Analysis of Laboratory Data
1. Relative Retention Times (RRT)
The RRT data submitted by the four participating laboratories were analyzed by the
same methods used for the inter-laboratory study which established the RRT windows for the
Target compounds in method 1625 (SRI International, "Interlaboratory Validation of U.S.
Environmental Protection Agency Method 1625A", Draft Final Report, June 1984). Briefly,
the GC temperature programs used by each of the laboratories were reviewed for adherance to
-------
296
those specified in the methods. Data from laboratories which did not use the temperature
ramps specified in the methods were rejected. (This resulted in the rejection of one
laboratory's data in both fractions). Data from the remaining laboratories was subjected to an
outlier analysis, which rejected 2% of all the data points as outliers. Next, 95% Prediction
Intervals (95% PI) were calculated from the remaining data to establish RRT windows. These
windows were compared to the RRT specifications for RS compounds presently given in
methods 1624C and 1625C to see whether or not the 'old1 RRTs fell within the 'new' RRT
windows.
2. Relative Response Factors
The RRF data were reviewed using Analysis of Variance (ANOVA), by calculating
Percent Relative Standard Deviations (%RSDs) for the five-point calibrations, and by plotting
RRF versus concentration for each compound and laboratory. The ANOVA was used to
detect differences between laboratories, between laboratories and the method specifications,
and between concentration levels within laboratories. The %RSD calculations were used to
follow up the results of the ANOVA, and to evaluate calibration linearity and stability of
Response Factors.
RESULTS AND DISCUSSION
1. Relative Retention Time data
The 'new' RRT windows for RS compounds calculated in this study are presented in
Tables 1 (method 1624C) and 2 (1625C). Compounds whose 'old' method RRTs do not fall
within these windows have an asterisk in the 'flag' column. The discrepancies between the
'old' RRTs and the 'new' RRT windows are detailed in Tables 3 (1624C) and 4 (1625C). In
Tables 3 and 4, the column 'deltaRRT' shows how far outside the 'new' RRT window the 'old'
RRT is. Figures 1 (method 1624C) and 2 (1625C) plot the midpoint of the RRT windows
versus the widths of the windows.
Most of the RRT values currently in method 1625C fell within the RRT windows
calculated by the above-mentioned methods; only 14 of 94 RS compounds' 'old' RRT values
fall outside the new windows (see Table 4). For eight of these compounds, the difference
between the 'old' RRT and the RRT windows from this study is marginal (0.014 RRT units or
less). The other four compounds have more substantial differences.
For the volatiles, over half (14) of the 24 compounds studied have 'old' method RRT
values outside the RRT windows from this study (see Table 3). Again, however, the
differences are marginal (0.019 RRT units or less) for seven of the twelve. The two with the
largest differences (the xylenes) are the last compounds to elute. This may be due to
differences in the final isothermal period of the GC temperature program used (between this
study and the one used to establish the original single RRT values). The last three compounds
in Table 3 are all "hot purge" compounds and none of the three was detected at all five levels
of the initial calibration by the three laboratories (this may be because the "hot" purge was not
used).
Figures 1 and 2 demonstrate, as expected, that the width of the RRT windows becomes
relatively larger as the RRT departs from 1.0. (Note the difference in the vertical scale in
figures 1 and 2). This appears to serious for early-eluting compounds in method 1625C, for
-------
297
some of which the width of the window is greater than the RRT. For example, the RRT
window for l,2:3,4-diepoxybutane is 0.143 to 0.459 ('old' RRT, 0.352). Such wide RRT
windows would seem to be of little use in compound identification. Referencing compounds in
this part of the chromatogram to an early-eluting internal standard (or labelled compound)
would alleviate this problem. The obvious outlier in Figure 2 at an RRT just below 1.2, is
1,4-naphthoquinone. The calibrations for this compound were badly saturated, which may
have had some effect on the RRT data.
In the course of the analysis, it was noted that there was no overlap in RRTs between
some of the laboratories using different GC temperature programs. RRT data from two
laboratories did not fall within a 95% Confidence Interval (which is narrower than the 95%
Prediction Interval mentioned above) calculated from all four laboratories' data. This points
out the necessity of matching the analysis conditions which a laboratory is using to those used
in generating the reference RRT data, in order for the reference RRT data to be valid for
compound identification under the conditions of use.
2. Relative Response Factor Data
Analysis of Variance (ANOVA) was performed on the RRF data to answer three
questions:
1) Are there significant differences between RRFs at different calibration levels
within laboratories?
2) Are there significant differences between mean RRFs between laboratories?
3) Are there significant differences between laboratories' mean RRFs and the present
RRF specifications in the methods?
The ANOVA results allowed us to conclude that:
1) Except for three compounds, there were no significant differences (95% level)
between RRFs at different calibration levels within laboratories. Percent Relative
Standard Deviations (% RSDs) were, in general, acceptably low (less than 25%).
2) There were significant differences (95% level) between laboratory mean RRFs (for
all compounds which all of the laboratories were able to detect). At most two of
the four laboratories had mean RRFs that were not significantly different for a
particular compound.
3) There were significant differences (95% level) between laboratory mean RRFs and
the method-specified RRFs. All four labs had significant differences for 43
compounds, three labs had significant differences for 49 compounds, two labs for
19 compounds, and one or no labs for only six compounds. Clearly the majority of
the results were different than the method-specified RRFs.
Because each laboratory's data appeared well-behaved, we concluded that the differences
were due to normal inter-laboratory variability. However, because the differences are
significant, no one set of RRF specifications can be recommended for inclusion in the
methods that would give equally accurate results from the four study laboratories.
-------
298
Figures 3 through 10 show plots of RRFs versus calibration levels for selected
compounds. These plots illustrate the differences between laboratories, and between the
laboratories and the method-specified RRF, which is shown as a horizontal line.
CONCLUSIONS AND RECOMMENDATIONS
We recommend that laboratories using methods 1624C and 1625C be required to make
their own RRT and RRF measurements, on a less-frequent basis than is required for the
Target compounds. This could be done at the same time that Relative Retention Time
measurements are being made, as mentioned above in the RRT section. A reasonable schedule
might be to require a five-point calibration for the Reverse Search compounds at the time of
the Initial Precision and Recovery (IPR) study, and then to require a one-point calibration at a
lesser frequency than is required for the Ongoing Precision and Recovery (OPR) analysis (say,
once a week or once a month). An early-eluting internal standard for method 1625C would
help to narrow the RRT windows for early-eluting Reverse Search compounds. Measurements
of RRT and RRF data for 'hot purge' compounds should be made under both hot and room-
temperature conditions by individual laboratories.
We believe that the development of laboratory-specific RRT and RRF data is the best
method to ensure the reliable identification and quantitation of Reverse Search compounds.
The problems encountered by the laboratories in this study illustrate the qualitative and
quantitative errors which could potentially be made by using RRT and RRF data which do
not apply to the specific analytical system being used.
If laboratories are not to be required to generate their own RRF and RRT data for the
reverse search compounds, then we recommend that this study be repeated, with strict
adherance to the analysis conditions in the methods, and with proper Quality Control
guidance. The methods would then need to be modified to require strict adherance to the
analysis conditions, to ensure that use of the reference RRT and RRF data would result in
reliable identification and quantitation of reverse search compounds.
-------
299
QUESTION AND ANSWER SESSION
MR. McCARTY: Nobody seems to
wants to ask questions about search compounds.
MR. ECKEL: It's a messy, messy
subject.
MR. McCARTY: Nobody wants to do
them either!
Surprisingly, we've finished up a little bit early.
What I would like to do, in order to give you people more time to
relax before we go out for our evening cruise this evening, is to
take our break now and come back earlier than 4:15, since I see
our 4:15 speaking standing at the back of the room, chafing at
the bit. What say we reconvene about 4:00 here? Go out and have
a few refreshments outside, and then we'll get out about 5:00,
hopefully.
-------
300
<
****** *****
IE
114
O. 1-
I O O «l O
I- 3
III O
tti:
£8
- 55
I (X
tu <
ac lu
in
u ui
z
~ ui
2°=
> u
M O
u.
u
•a-
CM
•O
o
Ifl
O-
a
o
3
o
OOOOOOOOOOOOOOOOOOOOOOOO
u
u
i
O -I Z
J.S2
I IH IU
z a u
IU <
a. z o
o o a
a -
o x
a: ui
a. a
a -J
CE <
Si
55
i o:
10 U
z
Is
IU <
is
O IU
o: iz
m o
w i:
a o
i tx.
CM CO
o z
a: <
ss
g£
n a
?g
«s- -J
-5
I W
-
m u
-• Q f- -*
i v- ;
m x :
w (- i
U iu i
o
s
u
iu -J
!2<
I X -I
: i- >-
: iu t-
: E o
i o ca
: a o
o m
CM ui:
• Sz
IU -I <
-!>-»-
M CE Z
K O IU
t- < a.
M x i
Z I- CM
IU IU I
z i: -i
IU >-
CL -I X
O >- I-
K X IU
a.t- n
i iu i
CM z: -a-
iu
x iu ui
,£55
: o x a.
St- o
iu a
-i z: a.
go o
CE a:
< o o iu
a: 3 -J »-
t--ix<
IU U. O t- UI
f- O M UI Z
i a tx u ui iu
CM O I- < Z -I
--II UI >-
-t 3: ro _i -i x
- o ->- >-
*< W M Z X CL
- CE -Ml +
-• t- ~* > z: o
a
Ul
a m
u> z:
-------
301
3
« * *
K
IU
B. *-
i CM ** o e c*i ^« »* •*<
O •« O O M M <
I O O CM X «4 -**o-4CM~oo<\i»«~4 —i<
z o>
HI HI HI HI HI
Z Z ZZ I Z
HI HI HI M O HI
8
I IU
IE
.8;
!J5i
ggggg g
igggE.nz'e
I -I -1 -> M II M
: xx xz --iz
I O U O I HI M I
II I I N Z Z HI
z1 §
.*. i-i M aeoxo^cucumin
iu m o i z H- N < i i i o o
x o HI w HI t- n HI z o o o a M
Czzmi-zuziu-i£i:£O3
O>-ll-<<-IM-OOO*J-'
f-r-JZZNOM I NKKKXO
HiZHiz0ffia)u»-
(JIZIOfUIUIUIoHII I I I I I I
<*-<~-i
-------
302
a. t-
a. H
X -I
in
o-
ioooooo>«oaca<
o
§
ss
X -J
in
>OOOOOO>4OOCM<
OO-*OOO-HOOO-«
O r* O O •-!
i- in
z a
ss
Ul O
a: o-
CM »- x
< O
u -i <
-J ui a:
g £S
Ul Ul
z£
.. Ul
>
— Ul
M ce
ui o:
in o
u
in
CM
•o
Q
O
I
: CE
CK
_ ui o
25SSS
• ~i ui -J ui
a. a: ui >- z KI
o o OL x r '
. W -I h- Ul H- HI
OOX «UIZUIX
aauujguioH
MM(-<(->-n^
ZZZXZCEUiui
I I Ul H- IU Ul X X
ZZO.U1O.O.GLIL
Q CO
ts z:
ui 3
-------
303
O f- IA -4
o- O-OO>r-O-KlKls3-O
•*~4OOO*-4OOO~4«4~4*4'
a:
iu
x
in
iu i
z:
>ooooooooooooo
>-
IU IU
z 2:
: Z i
> O I
< o
I X K
: t- o
x -i
O X:
O O I
a M :
o o:
M >-
-I O. IU -I M
>- O -J O tt
gCE M IU X H-
IU O tt Z O -I w
< ( II--uj<
-I < I D- E K- CL O
x uj >- - Q: o: fli IE o
IU I M • I M «) I X
a
o
o>o>inininininininminininin
-------
304
— BJ
Sz
M a:
z oa
i i
< a. • i
—i -a t-t 3 M it
-------
305
CJ
-------
306
*
*
«0
H •
HI
*
•a
^
* *
*
CM
iu
CM
•O
IU
in
>
ui
m
*
*
*
*
«
*
* * *
* *
DC
* *
CM IU
oc
tK
° 5
•
o
-O
O
in
o
o
o
c
I
.a
o
ui
o
-------
307
FIGURE 3
VOA : 539 refrrf=1.350 DIBROMOMETHANE
20
S 15
o
cd
10
0)
>
• rH
+J
CO
0
0 50
100 150
concentration
+ REFRRF
^ LAB4
* LABS
0 LAB2
• LABI
200 250
-------
308
FIGURE 4
VGA
550 refrrf-0.150 4-METHYL-2-PENTANONE
O
-t-J
«
cu
CO
a
O
P<
Cfl
O)
^(
-------
309
FIGURE 5
SEMI : 560 refrrf=0.150 BENZANTHRONE
7
O
6 -
5 -
CD
§ 4
O
P<
CO
-------
310
FIGURE 6
SEMI : 579 refrrf=0.110 2,3-DICHLORONITROBENZENE
0.8
0.7
o 0.6
a
C/3
0.5
o
OH 0.4
cn
a;
£ 0.3
•f-4
•+->
cd
0.1
0.0
0
50 100 150 200
concentration
+ REFRRF
^ LAB4
•& LAB3
0 LAB2
• LABI
250
-------
311
FIGURE 7
SEMI : 587 refrrf=0.240 1,4-DINITROBENZENE
0.25
0.20
o
-*->
o
8 0.15
o
P,
in
0.10
0.05
0.00
0
+ REFRRF
^ LAB4
-------
312
FIGURE 8
SEMI : 913 refrrf=0.110 P-NITROANILINE
o
£ 2
cu
en
o
O,
en
cu
cu
0
0
+ REFRRF
^ LAB4
iSr LAB3
0 LAB2
• LABI
50 100 150
concentration
200
250
-------
313
FIGURE 9
SEMI : 915 refrrf=0.470 N-NITROSODI-N-BUTYLAMINE
0.8
0.7
o 0.6
*j
o
<3
~ 0.5
-------
314
FIGURE 10
SEMI
926 refrrf=0.150 PHENOTHIAZINE
6
5 -
o
03 4
0)
en
G
a 3
to
QJ
-- 2
-»-> ^
CO
<«4
CD
0
+ REFRRF
^ LAB4
* LAB3
0 LAB2
• LABI
0
50 100 150 200
concentration
250
-------
315
MR. TELLIARD: Our last session
today is going to deal with the joys and relationships with
chlorinated phenolics, one of the more exciting things.
Two speakers are going to be addressing two programs
that the agency has underway to help out the pulp and paper
industry. As you know, a number of them have felt neglected so
we haven't written any regulations for them and in an attempt to
dispel that feeling of neglect, we're busily trying to get them
more regs so they'll have something to talk about at their
meetings. In lieu of that, our first speaker today is Larry
LaFleur who is from the National Council of the Pulp and Paper
Industry for Air and Stream Improvement. Over the last couple of
years, Larry and I have been seeing each other in very strange
places and talking about very strange things. Larry has taken on
the dubious pleasure...honor...whatever... of being responsible
for tying off many of the analytical issues that are facing both
the agency and the industry and trying to measure some of these
compounds in this rather complex effluent. So, he's going to
speak a little bit today on chlorinated phenolics in the pulp and
paper industry.
-------
316
MR. LaFLEUR: Thanks, Bill.
Much of the pulp that's produced in the United States
and throughout the world is bleached in order to improve the
brightness, cleanliness, strength properties, color stability and
absorptive properties of the final pulp. Current practices for
bleaching utilize oxidation with chlorine and chlorine dioxide to
break up the lignin molecule which is then subsequently dissolved
in a strong caustic solution.
The reactions responsible for the de-lignification that
occurs have been extensively studied and reported in the
literature. A number of de-methylation reactions and de-
alkylation reactions occur to break ether bonds, giving an
increased phenolic content to the residual lignin and to begin to
break up the lignin polymer backbone. There are a series of
reactions on the ring, itself, some forming quinones and others
leading to ring opening that increases the carboxyl content.
The next reaction that's important in the de-
lignif ication process is electrophilic displacement of the three
carbon side chain. All of those reactions lead to increased
solubility of the residual lignin, breaking up of the lignin
backbone and reducing the molecular weight, thereby, facilitating
the subsequent extraction with sodium hydroxide.
Concurrent with those reactions are some reactions that
do not improve the bleaching properties; specifically, these
reactions include electrophilic substitution reactions on the
aromatic ring and some substitutions on some definic groups that
are on the propyl side chain.
Amongst the wide variety of chlorinated organics that
are produced as a result of the bleaching process, is a whole
series of chlorinated phenolic materials. The bleaching of
softwood pulps, yields chlorinated phenols, chlorinated
guaiacols, (ortho-methoxy phenols) chlorinated catechols (ortho
dihydroxy bezene) and chlorinated vanillins which have the basic
structure of the guaiacol with an aldehyde group para to the
hydroxy group.
-------
317
Due to the difference in the chemical makeup and
structure of lignin that is in hardwood species, all of these
types of materials are formed, along with chlorosyringols, (2.6
dimethoxy phenol) and syringaldehydes, (2.6 dimethoxy-4-formyl
phenol).
A common misconception is that the bleaching process
produces essentially all possible isomers of chlorinated
phenolics. In actuality there is a number of very characteristic
compounds that are formed.
This slide summarizes the most frequently detected
materials organized by different process streams.
Generally, what is observed is a predominance of two to
four chloro substitutions. The 4-chloro substituent occurs
largely as a result of that de-alkylation reaction and breaks up
the lignin molecule and subsequently releases the lower molecular
weight chlorinated phenol.
Higher concentrations of chlorocatechols are observed
in the C stage (first chlorination stage) than in some of the
other stages because they are more stable under the high acid
conditions. In the E stage because the purpose of that is to
extract these materials as their phenolate salts. Thus, the E
stage typically has the highest concentrations in the bleaching
process.
The next table summarizes the least frequently detected
compounds. The reactions that are responsible for the formation
of these tend to be electrophilic aromatic substitution. The
thermodynamics dictates ortho para substitution so what you see
is anything that has substituents in the three and five position
are virtually never found.
There are some other compounds, such as the catechols
which are only occasionally detected in the E stage, but are
frequently detected in the C stage.
You'll also note that the frequency of detection in the
final treated effluent is significantly lower than any other
process streams. That's in part due to the dilution of the
-------
318
bleach plant waste waters with the pulping process waste waters.
But it's also due to the removal of these compounds in the
biological treatment system.
Chlorinated phenols came to the attention of the pulp
and paper industry in the early 1970s when it was determined that
they were significant contributors to fish toxicity in some
untreated process waste waters. The next groups chloriodrenolics
by chlorination level and plots them versus the toxicities to
salmonids and/or fathead minnows. The toxicities range from the
300 to 500 ppb all the way up to several ppm. If you look within
each chlorination level, you see quite a range of toxicities.
But in general, if you look at the average for each chlorination
level, there is a trend towards increasing toxicity with an
increasing degree of chlorine substitution.
A number of studies have been undertaken to determine
the environmental fate in the receiving environment of
chlorinated phenolics, particularly of the guaiacols and
catechols. What's begun to emerge is what is being described as
the hypothetical guaiacol cycle. In biota, it has been shown
that both the guaiacols and the catechols are readily conjugated,
forming constituents like glucuronides. We've measured and
determined that the uptake rates by the fish and the depuration
rates are quite rapid. As a result, overall the bio-
concentration potential, particularly for the lower chlorinated
compounds, is quite low.
In the aqueous environment, which is essentially an
aerobic environment, a number of studies have looked at the kinds
of microbial transformations that occur. What seems to be the
predominate metabolism is either demethylation of the guaiacol to
form catechol structures which then can behave essentially as a
catechol would normally behave or alternatively...methylation of
the free phenol group to form a veratrole compound. Veratroles
are fairly important when you consider that they may, in fact, be
co-contaminants of the chlorinated phenolics and guaiacols in the
receiving environment.
-------
319
There's a number of other metabolic pathways that have
been looked at, many of which lead to ortho-hydroxylation or
para-hydroxylation, both with or without subsequent
dehalogenation. Those products then ultimately undergo the same
sort of demethylation and methylation chemistry mentioned before
so the complexity of the compounds that one might encounter in
the environment are substantially greater than what is observed
in the bleach plant.
Materials that find their way into the sediments, have
been shown to bind very quickly and very tightly with the solids.
Generally, simple solvent extraction procedures are not very
effective in recovering these materials and it required
methanolic KOH to release what the authors of the work are
terming bound materials. Sometimes the ratios of bound versus
free are on the order of 100 to 1,000 times.
Some of the early analytical schemes that were used to
monitor for chlorinated phenolics involved simple solvent
extraction and then derivitization; often with diazomethane.
These methods were found to be not very useful. A series of
different substitution patterns become indistinguishable when
methylated so that these different trichloroguaiacols are
indistinguishable as the corresponding veratrole derivative.
Catechol recoveries in the solvent extraction systems
were found to be very low and highly variable. The catechols
that were extracted also methylated to produce an
indistinguishable veratrole-type compound. Using this approach,
you couldn't tell the difference between guaiacols, isomeric
guaiacols, catechols and any veratroles that might have been
present in the sample.
In early 1981, Ron Voss of the Pulp and Paper Research
Institute in Canada reported the procedure shown here and it
resolves many of those problems. In the procedure, a 50 mL
sample...is neutralized, and 2,6-dibromophenol is added prior to
derivatization as an internal standard. The sample is then
buffered to a pH of 11.6, using potassium carbonate. Acetic
-------
320
anhydride is added to the separatory funnel and mixed into the
sample. A carbonate buffer forms phenolate salts of the phenols
which then react with the acetic anhydride and directly form
chlorophenol acetates in-situ, that is, right in the aqueous
phase. Once that reaction has occurred, the resulting neutral
chlorinated phenolics are much more amenable to extraction with a
non-polar solvent like hexane. Following extraction, the
analyses in the original method were performed on an electron
capture detector with an unconcentrated sample.
One important aspect of this technique is that the
analytical standards were prepared in an exactly identical
manner. That is, for calibration purposes, you spike internal
standards and un-derivatized target analytes into reagent water,
buffer, acetylate, extract and analyze. Thus, the resulting
response factors or calibration curve reflects not only the
electron capture response, but the overall derivitization and
extraction efficiency of the analytes.
The figure illustrates that the resulting acetates are,
in fact, chemically distinguishable and the chlorophenol
accetates can be separated and analyzed uniquely.
To summarize the advantages of the Voss procedure, the
method overall was quite sensitive, using a 50 mL sample for
compounds with two or more chlorines. One could obtain low part
per billion detection limits which made the method ideally suited
for the laboratory bleaching studies, for which it was originally
developed. It had improved selectivity, both in terms of being
able to differentiate the different isomers of chloroguaiacol
that are formed and it could also differentiate guaiacols from
catechols and guaiacols and catechols from veratroles.
The accuracy and precision I'll talk a little bit more
about, but it was certainly better than anything we had seen to
date using more traditional approaches.
Finally, the method was quite rapid and that really
lent itself to the research environment where you're trying to
generate a great deal of information.
-------
321
In support of the current effluent guidelines program
review that Bill Telliard just mentioned, EPA has been utilizing
an adaptation of this in-situ acetylation procedure. The method
was originally developed as a research tool with an idea of
generating large databases for essentially comparative purposes.
Considering it in the context of a regulatory compliance method,
you need to go back and review the suitability of the method in
the context of single analyses compliance/non-compliance
determinations. Ideally, a method that's to be used for a
regulatory purpose ought to have well-defined and known accuracy,
precision, and ideally it should be reliable and rugged.
What I'd like to do is talk about how the in-situ
acetylation procedure stacks up against these kinds of criteria.
Starting with accuracy, in the context of the in-situ
acetylation, you're dealing with a balance between specificity
and sensitivity. The in-situ acetylation procedure itself
provides a means of cleaning up the sample. It optimizes the
recovery of the target analytes and minimizes the recovery of
undesirable co-contaminates. Obviously, detector selection is
going to be important to both specificity and selectivity.
Column selection is probably most important to
specificity and certainly there's an obvious relationship between
the sample size and the overall method sensitivity.
Looking first at the detectors, a number of detectors
have been used in connection with this methodology. The original
work done by Ron Voss used an electron capture detector. Shortly
after Dr. Voss published his work, MCASI reviewed the method and
published a GC/MS full scan version and recently in the Effluent
Guidelines Program, EPA has been looking at the use of an
electrolytic conductivity detector. If you compare those in the
context of selectivity, we feel that the GC/MS offers, at least
in full scan mode, clearly the best selectivity and we think
there's reason to believe the electrolytic conductivity is more
selective than the original electron capture. If you view those
detectors from the point of view of sensitivity, the order is
-------
322
essentially reversed. We're very interested in which of these
detectors is going to provide the optimum compromise between
selectivity and sensitivity. In order to do that, we went back
to re-evaluate the original electron capture method. We looked
at 23 different paired samples, covering the entire gamut of
different matrices experienced in the pulp and paper industry
collected from a variety of different pulp mills. We split the
samples, analyzed them by both electron capture detection method
and GC/MS. We used the GC/MS data as our reference point and
only made comparisons where we had comparable detection limits.
We flagged samples that had a positive by the electron capture
detector, but showed non-detect by the GC/MS as a false positive
and samples that showed a factor of two higher result in the
electron capture detector, compared to the GC/MS as a positive
bias.
For a wide number of different analytes that we were
monitoring at this point, we flagged at least some percentage of
the samples for false positives, some for a positive bias, some
for both.
Overall, when we look at the full set of 23 paired
samples, there were only actually three samples or 13 percent
that had complete agreement between the GC/MS and the ECD data.,
Based upon this, it was determined that this is probably not a
real good compliance monitoring tool.
EPA had adapted the Voss method to use a two column
confirmation approach to enhance the selectivity of the Hall's
detector that was used and thus complimenting its improved
sensitivity over the full scan GC/MS. This sounded like a very
good approach, so we set about evaluating it. The first step in
that process, we thought, would be to do the best job we could of
getting the best separations between all potential analytes and
potential interferences.
We spent a couple of weeks working on different
temperature programs and what you see here is the results of that
effort. It's a pretty complicated program. We optimized the
-------
323
separations on two columns. The original DB1 is the column that
MCASI used since the beginning when Ron Voss originally published
his work and the Rtx-35 column was the one that EPA was using as
their confirmatory column.
Even after spending a couple of weeks trying every
program we could think of, there were still a number of
unresolved pairs. There's basically several sets of compounds
that can't be resolved. On the Rtx-35 column, there was even a
larger number of compounds that couldn't be resolved.
Improving resolution of any given pair of close elating
compounds simply results in degrading the resolution in another
area of the chromatogram. The temperature programs given in the
figure offer the optimum compromise in terms of getting the
maximum number of analytes separated. This is only with known
standards or known constituents of pulp mill effluents and a few
of the veratroles that might be co-contaminants in environmental
samples.
Even if we could get all of the standards separated,
that doesn't necessarily solve all potential problems. This is a
chromatogram of 10 ppb standard on the Hall's detector, using a
DB1 column and a typical C-stage filtrate, both run under
identical instrumental conditions. The C stage filtrate
chromatogram clearly illustrates that finding the target analyte
in the presence of a myriad of potential co-eluting peaks is the
challenge.
The next series of evaluations we performed involved
six different typical pulp mill samples. These included some E-
stage filtrates, C-stage filtrates from the process and some
final treated effluents. We analyzed them both by the
electrolytic conductivity detector and by the GC/MS. Since the
GC/MS detection limits weren't as low, we enhanced the
sensitivity by analyzing the samples first by a full scan
procedure which we had been using for years and then repeat the
analyses using a selected ion monitoring approach. Again, we
used the GC/MS data as a reference point.
-------
324
For the dual column data shown in the table we used the
criteria that EPA had established for positive identifications.
The analyte had to show up at the right retention time on both of
the columns and the concentration had to agree within a factor of
two. There are 40 different analytes in the six different
samples giving a total of about 240 determinations.
Using the two column confirmation procedure, you see a
fairly small number of false positives, but a fairly significant
number of false negatives. One other thing we considered was the
selectivity achieved using the Hall's detector without two column
configeration. As you might expect, we about doubled the number
of false positives, but we significantly reduced the number of
false negatives. What's probably happening here is on one or the
other columns we have a major chemical contaminant which is
moving under the target analyte and then throwing the
concentrations off.
The other point that I'd like to make is that this
extends down into a concentration range below one part per
billion. Some of the values were as low as half a part per
billion. I'm sure that this whole treatment of the data would
change significantly if a different lower detection threshold was
used.
Based upon these preliminary evaluations, we feel that
the electrolytic conductivity may offer some advantages over the
electron capture detector, but it clearly has significant
problems with process streams. We still feel that in terms of
the optimum compromise between sensitivity and selectivity, the
GC/MS procedure is the preferred method.
Due to the unique nature of the in-situ acetylation
procedure, there's a number of factors that can contribute to the
overall precision of the method. The first is the acetylation
efficiency or yield and I'm going to talk more about that in a
minute. The second factor is the extraction efficiency. We
found that the extraction efficiency using the hexane is greater
than 95 percent. We've done some studies looking at more polar
-------
325
solvents to see if that might further improve that, but what we
found was that you run into two problems: One, you start
extracting some of the acetic acid, which was formed as a
byproduct of acetic anhydride in water. This causes a lot of
chromatographic problems. Secondly, you start to pull out more
interferences. Given these two observations, hexane remains the
optimum solvent.
The third factor affecting precision is concentration
losses. Unless you boil the extract down to dryness, we haven't
seen a real problem with concentration losses of these
chlorophenol acetates. We think the method is fairly rugged with
respect to concentration.
The last item here is the GC analysis. Clearly, that
is a problem, particularly for the chlorinated catechols. We
frequently have to replace the injection port liners and are
often cutting off the front part of the columns. So, the method
does reguire fairly careful chromatography.
In order to give you an idea of what the
reproducibility of some of the spike recoveries are, I've tried
to tabulate some of our older electron capture data. This
represents approximately 100 determinations of matrix spikes made
in various different pulp and paper industry waste waters and
process streams. The recoveries are uniformly high with the one
exception of the tetrachlorocatechol, which is in the 73% range.
From this data, it is readily apparent that high recoveries can
be routinely maintained.
The chlorinated phenols and guaiacols, the relative
standard deviation of the recoveries are fairly consistently in
the 16 to 23 percent range.
The vanillins recovery relative standard deviations are
30 to 38 percent. Thus, they are not as reproducible as the
guaiacols and phenols.
The relative standard deviations for the
chlorocatechols are 40 to 57 percent making them the least
reproducible. It is very hard to maintain consistent higher
-------
326
recoveries with the chlorocatechols.
Looking more specifically at method precision, we
summarized some of our GC/MS data. These are seven replicate
analyses of C-stage filtrate, an E-stage filtrate and a final
treated effluent. A select group of analytes are included to
illustrate the precision of the different analyte groups.
The overall precision of the chloroguaiacols and
phenols is the best of all the analytes and the relative standard
deviation is fairly low. The 4,5-dichloroguaiacol was at about a
2 ppb level so we probably are running into problems with
concentration there. The vanillin relative standard deviations
are fairly similar but slightly higher than the chloroguaiacols
and phenols, the chlorocatechols consistently show poorer
precision.
Clearly, the catechols are the weak point in the
analytical methodology. If there is anything wrong with this
method, it's the poor ruggedness or reliability of catechol data.
Looking into this a little more closely, I think a
brief review of basic catechol chemistry begins to help
illustrate some of the reasons why this might be the case.
Catechols are known to undergo oxidation reduction reactions
forming the corresponding orthobenzoquinone, particularly in
alkaline solutions. This is an eguilibrium-type reaction and the
distribution of the products is very much dependent upon the
presence of other oxidants in the sample, including oxygen.
A second reaction that may give rise to the degradation
of the catechols is just simple base degradation. In fact, this
reaction occurs very extensively in a caustic extraction stage.
Compounds like tetrachlorocatechol are known to form chloranilic
acid in the presence of the strong base and that's essentially
why we see very little catechol in the E-stage filtrates.
These observations give us a little bit of an idea of
how we might be able to go about solving the problem. The first
thing that you can look at is the addition of a preservative, a
reducing agent such as thiosulfate and bisulfite that are added
-------
327
to quench residual chlorine and is generally practiced as a field
preservation technique. Addition of thiosulfate has been
pratical for years and the catechol problem still exists. Thus,
that isn't the solution to the problem.
We have seen that the addition of ascorbic acid can, in
fact, reduce the orthobenzoquinones back to the catechols and
thus offers at least one approach to solving the problem.
Another consideration we might explore is to lowering
the pH of the buffer. If these compounds are degraded in a
strong base, use a weaker base. Along those same lines, we're
looking at the possibility of shortening the exposure time to the
high pH.
Of these three general approaches, clearly the most
commonly practiced is the addition of ascorbic acid. What we
wanted to do is evaluate just what that does to the results.
Generally we can control our catechol recoveries fairly well. A
series of paired analyses of samples with and without the
addition of the ascorbic acid were performed. What was observed
is that in the E-stage where most of these base labile compounds
don't exist, there's no difference between the two analyses.
Also, virtually no difference was observed in the treated
effluents. The most significant differences were observed in the
C-stage where there is a great deal of labile material and in one
of the primary influents. The differences are fairly consistent
with the one exception of the 3,4-dichlorocatechol where it's
significantly greater. We think that these differences are
directly attributable to the presence of chloroquinones.
We looked into this 3,4-dichlorocatechol a little more
closely. Going back to the literature, back in the early '80s a
compound 2,6-dichloroparaquinone was reported as a constituent in
C-stage filtrates. This compound hadn't received a great deal of
attention because nobody had good analytical methods for
quinones. We've taken the compound and have shown that it is, in
fact, reduced to the corresponding hydroquinone, (i.e. 2,6-
dichloro-p-dihydroxy benzene) in the presence of ascorbic acid.
-------
328
The diacetate that is formed upon acetylation has exactly the
same retention time as 3,4-dichlorocatechol. By inspection of
the spectra of the two compounds, it is readily apparent that, it
would be hard to tell, if those two compounds were co-eluting,
whether it was from a catechol or a paraquinone.
In general, we feel that although the ascorbic acid may
offer some advantages in solving this catechol problem, it does
so at the expense of the selectivity of the method. In
particular, that means that we're no longer talking about
catechols; we're talking about catechols plus quinones and, in
fact, we may even be grouping in paraquinones with orthoquinones.
We are currently investigating solutions for solving
this catechol problem which do not compromise the selectivity of
the method. We went back and looked at the pH of the buffer.
The figure summarizes a series of experiments performed
investigating buffers from 8.2 up to pH 12. The original Voss
procedure uses a buffer at 11.6, which is included in our study
for reference. For a number of compounds, we actually saw a
slight peak but otherwise, these are pretty flat curves. So
first of all, there's very little sensitivity to pH. Secondly,
if there is any sensitivity, it would actually indicate that on
some of the compounds, we actually saw an increase at pH 9.9.
There was very little difference between the original pH 11.6
buffer that we had originally used and buffers as low as pH 9.1.
We concluded that we may have a handle on minimizing the catechol
degradation and the data certainly suggests it's worth pursuing.
What remains to be seen is whether or not the dropping the pH
from 11.6 to 9.9, or even as far as 9.1, will be sufficient to
adequately stabilize the catechols and at the same time we have
to make sure that these buffers will work well on some of the
samples. The experiment was just spiked reagent waters.
Another approach we're looking at involves reviewing
the way we conduct the buffering in order to try and minimize the
time it takes to get the sample acetylated after the pH
adjustment. In order to look at that, we plotted here titrations
-------
329
of several different samples against the carbonated buffer we use
for the calibration standards.
The samples themselves have a native buffering capacity
in the range of pH 8 to 10. In the early stages of our method,
we actually found that when we added the buffer, we were only
getting the pHs in the 11 range and we thought that was important
at the time and so we added a step to increase the pH up to a
target equal to that which we were using for calibration. All of
that takes a lot of time. You've got to remove the pH electrode,
rinse it off, take out the stir bar, and transfer the sample from
the beaker to the separatory funnel, all of which extends the
time that the catechols are sitting there degrading.
We are presently investigating a modification rather
than making the initial pH adjustment to seven, the adjustment is
made to pH 9 to 9.5. This takes the pH beyond the inherent
buffering capacity of the sample. Then we go through all the
business of taking the pH electrode out and transferring the
sample to a separatory funnel. At that point, you are at a
milder pH and the time may not be as critical. The carbonate is
added and there is no check on the pH; the analyst immediately
adds the anhydride. Hopefully, that will get us beyond any pH
degradation. Preliminary experiments show that samples will not
always hit pH 11.6, but from the data presented in the previous
slide, we're convinced now that that's not anywhere near as
important as we originally thought.
Both of these later modifications to the method are
being considered specifically to enhance the ruggedness of the
method, without compromising the accuracy or the precision.
I'd like to conclude that basically, we feel that the
GC/MS provides the best selectivity, particularly in light of the
complex nature of the effluents we were trying to analyze and we
feel it has adequate sensitivity for most applications. The
precision for all of the analytes is acceptable, with the
exception of the chlorinated catechols. The reliability of the
catechol analyses certainly needs improvement. Most people
-------
330
struggle, trying to get 20 or 30 percent recoveries. Clearly,
the method needs to be more rugged with respect to those
analytes.
MR. TELLIARD: Thank you.
-------
331
QUESTION AND ANSWER SESSION
MR.TELLIARD: Questions?
MR. CASTLE: Bill Castle,
California Fish and Game Department.
I was wondering if you had tried what seems to be the
obvious use assumed under GC/MS for quantitation of these
compounds?
MR. LaFLEUR: The use of what now?
MR. CASTLE: Selected ions?
MR. LaFLEUR: Well, we've done it
in the research context, but generally we didn't think that we
needed detection limits much lower than we had for most of our
research purposes. When you're working in the process streams,
detection usually isn't a problem. You're not really worried
about a sensitivity issue.
MR. CASTLE: Well, it just seemed
in your two comparisons that you put it last under detection and
you could have increased that by going to that...under your
sensitivity that you listed on the one slide.
MR. LaFLEUR: Oh, okay, yes. If
we were primarily after sensitivity, that would be one option to
consider. The target analyte list we work with in our program
typically is about 45 compounds and when you get that many
compounds what you have to resort to, to use selected ion
monitoring is only a couple of ions for each peak and at that
point you have to really ask yourself, is that sufficient
information to an unambiguously identify a contaminant.
MR. TELLIARD: Thank you, Larry.
-------
332
ln-Situ Acetylation Analysis of Chlorinated Phenols
in Pulp and Paper Industry Wastewaters:
Further Investigations and Refinements
L. LaFleur, J. Louch, G. Wilson, and G. Dodo
National Council of the Paper Industry
For Air and Stream Improvement
ncasi
-------
333
-c-
O
-c/
ELECTROPHILIC r
DISPLACEMENT (C1,)V ~C~
QUINONE
FORMATION
(C102)
ELECTROPHILIC
SUBSTITUTION (C1J\
DEHETHYLATION
(CIO,, C1Z)
RING OPENING
(CIO,)
DEALKYLAT10N
OCH,,
SOFTWOOD EFFLUENTS-
-H
OH
CHLOROPHENOLS
OCH,
OH
CHLOROGUAIACOLS
CH.,0
OCH,
OH
CHLOROSYRINGOLS
CHO
CHO
CH3O
OCH,
OH OH
CHLOROCATECHOLS CHLOROSYRINGEALDEHYDES
HARDWOOD EFFLUENTS
-H
-------
334
MOST FREQUENTLY DETECTED CHLORINATED PHENOLICS
FREQUENCY OF DETECTION
2,4-Dichlorophenol
2,4,6 - Trichlorophenol
4,5 - Dichloroguaiacol
3,4,5 - Trichloroguaiacol
4,5,6 - Trichloroguaiacol
3,4,5 - Trichlorocatechol
Tetrachlorocatechol
6-Chlorovanillin
5,6-Dichlorovanillin
C E
Stage Stage Process Treatment Treated
Filtrate Filtrate Sewers Effluent Effluent
n - 23 n • 23 n • 39 n - 3O n - 63
48%
57%
78%
100%
43% 87%
35% 74%
13% 96%
100%
78%
57%
13%
70%
13%
83%
83%
59%
64%
59%
67%
49%
85%
49%
64%
56%
77%
80%
27%
38%
90% 57%
83% 30%
60% 22%
93%
63%
83%
83%
75%
22%
57%
56%
LEAST FREQUENTLY DETECTED CHLORINATED PHENOLICS
FREQUENCY OF DETECTION
2-Chlorophenol
4 -Chlorophenol
2.6-Dlchlorophenol
3,5-Dlchlorophenol
2,3-Dlchlorophenol
3,4-Dlchlorophenol
2,3,6-Trichlorophenol
2,4,5-Trichlorophenol
2346-Tetrachlorophenol
Pentachlorophenol
4-Chlorogualacol
4,8-Dichloroguaiacol
Tetrachlorogualacol
4-Chlorocatechol
3,4-Dlchlorocatechol
4,5-Dlchlorocatechol
34 6-Trichlorocatechol
5-Chlorovanlllln
Chlorosyrlngaldehyde
Trlchlorosyrlngol
3,5-Dlchloro-4 -hydroxy-
benzaldehyde
C
Stage
Filtrate
n - 23
0%
43%
0%
0%
O%
0%
0%
0%
0%
0%
22%
0%
9%
48%
65%
65%
70%
4%
13%
30%
E
Stage
Filtrate
n - 23
9%
43%
4%
O%
0%
0%
0%
0%
9%
4%
35%
35%
65%
4%
9%
4%
17%
57%
57%
87%
Process
Sewers
n • 39
0%
49%
0%
0%
0%
0%
0%
0%
15%
10%
25%
31%
56%
13%
54%
54%
89%
33%
15%
33%
Treatment
Influent
n - 30
3%
73%
0%
O%
0%
0%
0%
0%
0%
0%
83%
80%
20%
43%
47%
83%
37%
83%
20%
30%
Treated
Effluent
n - 83
0%
5%
0%
O%
0%
0%
0%
0%
6%
3%
0%
17%
14%
14%
5%
51%
8%
24%
3%
3%
0%
65%
38%
57%
3%
-------
335
12
10
Chlorinated Phenolics Fish Toxicity
en
E
64
-i-
i
n
O O.5
1.5 2 2.5 3 3.5 4 4.5
Chlorine Atoms per Molecule
5.5
Chlorocatechols
Chloroguaiacols
Chlorophenols
HYPOTHETICAL "GUAIACOL CYCLE1
OConJ
OConl
CcO
OCon)
A
(Cl
//
\
OH
OH
,OH
OM«
OM« j^ OM.
S«d
(cO
OM«
.Bed
O' !,
OM«
BIOTA
WATER
SEDIMENT
-------
on
336
OCM-
OH
ci
OCH.,
OCH-
CI
on
on
cK\ / "Ci
ci
METHYLATION
PRODUCT
EFFLUENT: IN SITU METHOD
Sample is made
up to 50 mis
and adjusted to
pH7
ISTD
added
I ~*nujf I
KjCOj
added
acetylation
collected
Extraction
with hexane
(2 min)
-------
337
OCH,
Cl
OCH,
Oil
Cl
OH
Cl
C!
ACETYLATION
PRODUCTS
Advantages of Voss In-Situ Acetylation Technique
• Sensitive
Small sample size
• Improved Selectivity
Differentiates Isomeric Chloroguaiacols
Differentiates Chloroguaiacols and
Chlorocatechols
• Improved Accuracy and Precision
Rapid
-------
338
Requirements for Compliance Methods
Accuracy
• Precision
• Reliability
(Ruggedness)
Column
Selection
1
Accuracy
Derivitization
Detector
Selection
I
Sample
Size
-------
339
Detector Comparison
Chlorophenol Acetates
Selectivity
GC/MS > ELCD > ECD
(full scan)
Sensitivity
ECD > ELCD > GC/MS
(full scan)
Comparibility of GC/ECD vs GC/MS Data
False Positive
Positives Bias
2,6-Dichlorophenol 35%
2,4-Dichlorophenol 17% 13%
3,5-Dichlorophenol 9%
2,3-Dichlorophenol 17%
3,4-Dichlorophenol 22%
2,4,6-Trichlorophenol 9%
2,3,6-Trichlorophenol
2,4,5-Trlchlorophenol 4<*>
2,3,4,6-Tetrachlorophenol 17%
Pentachlorophenol 13%
4,6-Dichloroguaiacol 17% 4%
4,5-Dichloroguaiacol 22% 17%
3,4,5-Trlchloroguaiacol 4%
4,5,6-Trichloroguaiacol 4<*
Tetrachloroguaiacol 4% 4%
3,4-Dichlorocatechol 4%
4,5-D!chlorocatechol ^
3,4,6-Trichlorocatechol
3,4,5-Trichlorocatechol 4%
Tetrachlorocatechol 13%
6-Chlorovanillin 4% 17%
5,6-D!chlorovanillin 4% 4%
Chlorosyringaldehyde 9% 9%
Trichlorosyringaldehyde 9%
-------
340
DB-1
2°
16
15° /min
(1 min)——" •""-"« - •' • > 100
/min 0 l°/min
*• 128 *•
Qo 40° /min , 25Qo (1Q mjn)
63 minutes
Rtx-35
0 . 20° /min
2°/min ,ioo"10°/min,
0 o 20° /min „
45 minutes
Compounds that Co-Elute
DB-1
6-Chloroguaiacol
3,4-Dichlorophenol
2.4,5-Trichlorophenol
4,5-Dichloroveratrole
4,6-Dichloroguaiacol
3,5-Dichloro-4-
Hydroxybenzaldehyde
5-Chlorovanillan
3,5-Dichlorocalechol
3,4,5-Trichloroguaiacol
Tetrachloroveratrole
Rtx-35
3.5-Dichlorophenol
2,6-Dichlorophenol
2,4-Dichlorophenol
4-Chloroguaiacol
5-Chloroguaiacol
3-Chlorocatechol
3,4,5-Trichlorophenol
4-Chlorocalechol
4,5-Dichlorogiiaiacol
Tetrachlorophenol
3-Chloro9yringol
3,4-Dichlorocatechol
4,5-Dichlorocatechol
5-Chlorovanillin
Telrachloroveratrole
-------
341
Comparibility of GC/ELCD vs GC/MS Data
All Matrices
2 Column DB-1
Confirmation Only
False Positives 4% 9%
Positive Bias 17% 24%
False Negatives 23% 2%
Negative Bias 3% 7%
Total Number of Samples » 6
Total Number of Analytes • 40
-------
342
Percentage of Dual Column Confirmations
in Replicate Analyses (>/• 5 ppb)
Final
Effluent
n • 7
4-Chlorophenol
2,4 -Dlchlorophenol
2,4,6 -Trlchlorophenol
Tetrachlorophenol
Penlachlorophenol
8 -Chlorogualacol
4-Chlorogualacol
4,8-Dlchlorogualacol
3,4-Dlchlorogualacol
4.5-Dlchlorogualacol
3,4,8-Trlchlorogualacol
3,4,5-Trlchlorogualacol
4,5,8-Trlchlorogualacol
3,6-Dlchloroealechol
3,5-Dlchlorocatechol
3,4 -Dlchlorocateehol
4,5-Dichloroeatechol
3.4,8-Trlchlorocalechol
3,4,5-Trlchlorocatechol
5,6-Dlchlorovanlllln
100%
88%
C Stage
Filtrate
n • 6
100%
100%
100%
33%
17%
33%
87%
100%
100%
100%
33%
E Stage
Filtrate
n • 4
100%
100%
100%
100%
100%
5O%
100%
100%
100%
100%
100%
100%
25%
75%
50%
25%
100%
1OO%
Percentage of Dual Column Confirmations
in Replicate Analyses (
-------
343
In-Situ Acetylation Procedure
Factors Contributing to Precision
Acetylation Yield / Efficiency
Extraction Efficiency
Concentration Losses
GC Analysis
-------
344
GC/ECD Matrix Spike Recovery Summary
3-84 to 1-86
_ Relative
X Standard Deviation
2,4-Dichlorophenol 110 23%
2,4.6-Trichlorophenol 102 16%
4,5-Dichloroguaiacol 103 19%
3,4,5-Trichloroguaiacol 99 19%
Tetrachloroguaiacol 93 18%
4,5-Dichlorocatechol 83 57%
3,4,5-Trichlorocatechol 94 58%
Tetrachlorocatechol 73 49%
6-Chlorovanillin 108 38%
5,6-Dichlorovanillin 83 32%
GC/MS PRECISION
Relative Standard Deviation
C Stage E Stage Treated
Filtrate Filtrate Effluent
n-7 n-7 n-7
2,4-Dichlorophenol 7% 4% 2%
2,4,6-Trichlorophenol 8% 5% 4%
4,5-Dichloroguaiacol 4% 3% 11%
3,4,5-Trichloroguaiacol 6% 3% 1%
Tetrachloroguaiacol 4% 5% 2%
4,5-Dichlorocatechol 11% 16% 3%
3,4,5-Trichlorocatechol 5% 16% 4%
Tetrachlorocatechol 9% 17% 49%
6-Chlorovanillin 8% 5% 8%
5,6-Dichlorovanillin 13% 5% 5%
-------
345
CHLOROCATECHOL pH STABILITY
OH [OxidJ
(Red]
C'x Clx
CHLORO-O-BENZOQUINONE
CHLOROCATECHOL
OH
OH
Clx
'x-2
CHLORANILIC ACID
Possible Solution to Chlorocatechol Problem
• Addition of Preservative
Reducing agent - ascorbic acid
thiosulfate
bisulfite
• Lower pH Buffer
• Shorten Exposure Time
-------
346
Percent Increase Due to Use of Ascorbic Acid
C-Stage E-Stage Primary Treated
Filtrate Filtrate Influent Effluent
Analyte
n-3 n-1 n-2 n-6
4-Chlorocatechol 0-11 0 0 0
3,4-Dichlorocatechol 0-311 0 0-137 0
4.5-Dlchlorocatechol 0-73 00 0
3,4.6-Trichlorocatechol 0-56 0 0 0
3,4,5-Trichlorocatechol 0-42 0 0 0
Tetrachlorocatechol 0-38 0 0 0
>• 100
t:
w 50
«
DC
>. 100
«5
o>
c
d) 50
w
cc
iv
^ 3
6
^3 2
6
. f
i* 1 tid •
,4-Dichlorocatechol Diacetate
178
S\
i
, 07 11 3 i.|g 149 i 193 21
" T ' <' \ / K
' ' ' \ / u.l
OAc
1
^-^
Cl
" - ' \
.-.-i-,-!i1+-»— .-- r-r--~^— i -V
'eV T 120 ' 160 200 2-10
m/e
,6-Dichlorohydroquinone Diacetate
178
\
5 97114 ,
85 / ,' 142151 / 8(
/ / / / / |,| /
OAc
C|VNX>
V
OAc
220
}3 / 2fi?
• OAc
•Cl
160
200
240
m/e
-------
347
Relative Recovery vs. Buffer pH
o
o
cc
Chlorophenols
Chloroguaiacols
Chlorocatechols
Chlorovanilllns
12.3
TITRATION CURVES
12
11
10
I 9
Q.
8
7
— K2CO3 Buffer
-J- Effluent
-W- C Sfngw
-B- E Stage
g J_LU_L|_JLJ_U_) J_LJ_1 (_I_U_1.{JJ-JJLJ-J_1_(_L | I I I I | I I I I [ t I I I | I I II ^ I I I I ^ LLLI
0 2 4 6 9 11 14 19 23 25 27 29
MLs 0.1 M HCL
-------
348
Conclusions
• GC/MS Provides Best Selectivity
with Acceptable Sensitivity
• Precision Acceptable for all Analytes
Except Chlorocatechols
• Reliability for Chlorocatechol Analyses
Needs Improvement
-------
349
MR. TELLIARD: I saved the best
until last. Our next speaker has presented here before. Peggy
presented some work that she had done on isotope dilution when
she was at Weyerhaeuser. Since then, she had a total conversion
and now works for our Region X laboratory in Seattle and is going
to continue the same thing, the in-situ acetylation and look at
the method for chlorinated phenolics again in pulp and paper
effluent.
-------
350
MS. KNIGHT: I thought I'd give
you a look at where I currently come from. This is the EPA
Region X BSD Lab. It's located in Manchester, Washington, very
close to Port Orchard, Washington, a little bit south of
Bremerton. One of the advantages of the Manchester lab
is it's not located in the exact place as the administrative
offices - that can be quite an advantage. The administrative
offices are located across the water. This is one of the views;
that we see from the laboratory, a ferry boat going across the
Puget Sound. If you look in a slightly different direction, you
would see Seattle.
This slide is included for Bill's benefit. I know that
he likes ships and this is the Nimitz. We can frequently see the
Nimitz coming into Bremerton. We get aircraft carriers...not
battleships.
Larry has given you an outline of how the NCASF's
method CP-86.01 works. I'll continue a discussion of that and
how we at the Region X lab have been using the GC/MS portion of
the method.
We've made a couple of changes that we think are
appropriate for us. One of them is scaling up from combining
three separate 100 milliliter extractions in a separatory funnel
to extracting up to three liters by stir bar.
There isn't much you can really see with this slide
except the vessels we do the stir bar extraction in. They are
the sample containers that come into the laboratory. A stir bar
is put into the bottom of the containers and a very strong vortex
is formed.
Actually, I'm a little bit ahead of myself. I wanted
to mention what the other changes were first. We have had
trouble losing catechols, so we add a reducing agent, ascorbic
acid. Another change we've made to the method is the
incorporation of an injection internal standard or an instrument
internal standard, whichever the current terminology is,
something that's added before the analysis of the sample on the
-------
351
GC/MS so that we can determine the recovery of the method
internal standard, either 3,4,5-trichlorophenol or 2,6-
dibromophenol, both of which have been used with success. We
like to monitor the recovery of the method internal standard to
find out how well we are doing. Along the same lines, we also
add a series of surrogate compounds to monitor the method.
About the stir bar method. One of the characteristics
of the in-situ derivitization procedure is that it proceeds very
rapidly so mixing of the sample is exceptionally important.
After the addition of acetic anhydride, the pH rapidly drops from
11.6, if you've reached that valve, down to about 7 within about
30 stet and the reaction is complete.
In order to make sure that everything is well mixed, we
pull a very strong vortex as we're mixing the sample, adding the
buffers, adjusting pH and adding the acetic anhydride is added.
It seems to ensure good mixing during a very rapid reaction.
Carbon dioxide is evolved rapidly during the reaction. The green
vessel does not build up pressure as a separatory funnel would.
The same vessel is used for extraction with hexane of
the reaction products. This is a picture of the apparatus that
we use to extract the reaction products with hexane. We use a
vacuum system. The parts that contact the sample are teflon or
glass. The top hexane layer is pulled off into the sep funnel,
then we use the sep funnel to separate aqueous phase from the
hexane, the water is returned to the sample container for further
extraction, and the hexane is saved as the extract.
You might think that the stir bar method would prevent
emulsion formation. Pulp and paper mill effluents form terrific
emulsions. However, the stir bar extraction method does help.
We still have emulsions with these samples and usually use a
centrifuge to break them.
As I mentioned before... Larry, could I have the first
overhead. We've had a fairly substantial problem with catechol
recovery even in our standards, so we now add ascorbic acid. The
chlorinated catechols, as you can see from this slide, which is
-------
352
derived from the same source as Larry's hypothetical guaiacol
cycle, readily oxidize to orthobenzoquinones under alkaline
conditions. In the presence of ascorbic acid, the benzoquinones
are reduced to catechols which can then be derivatized and
extracted. There is industry concern, especially in process
streams that quinones and catechols will both be lumped together,
but we need to find some way to at least have a look at the
catechols.
Some Fathead Minnow larvae testing that we have done on
tetracholo-orthobenzoquinone indicates that it is on the same
order of magnitude toxicity as tetrachlorocatechol. We plan on
testing other species, perhaps oyster larvae.
The injection internal standard that we use to monitor
the recovery of our internal standard is 2,6-dibromophenol.
NCASI in the past has used 3,4,5-trichlorophenol, which is not
apparently found in pulp mill effluents. Both of these stet hcive
problems. 3,4,5-Trichloroguaiacol is formed from
tetrachlorophenol and also from pentachlorophenol under anaerobic
conditions on some matrices. 2,6-Dibromophenol has been found to
be produced naturally in infaunal worms and we've also found it
in English sole. So, if one were to use this method, you'd have
to be careful in internal standard selection to use it with
marine tissue.
We selected a group of surrogates monitor how the
method was proceeding and the ones that we selected were ones
that are relatively easily available, D5-phenol, 2-fluorophenol,
2-ethoxyphenol, which I was convinced was going to mimic
guaiacol, 2,4,6-tribromophenol and D6-resorcinol. After we had
done about 20 matrix spikes, we looked at the correlation of
every compound to every other compound to see what reflected what
with SCOUT, a statistical package developed by Lockheed for EPA,
turned out that the surrogates that we had selected were really
only monitoring the recoveries of non-chlorinated and low
chlorinated phenolics. We didn't really have appropiate
-------
353
surrogates for any of the tri and tetrachloroguaiacols or
catechols. So, what we'd really like is to have a 13C labeled
representative of each of those groups.
Well, how do the method mosifications on these work?
We've got a couple of ways that we've tested this. This slide
represents our method detection limit determination. It was done
in blank water, so it represents the best case and detection
limits. Of course, you could not expect this in real mill
effluents and I'll show what happens with those in a minute.
There are a couple of things I should say about this
slide. All of the detection limits that we determined were sub-
parts per billion. The spiking level was 0.83 micrograms per
liter and we extracted seven replicates of three liters of sample
using the stir bar procedure. There are a couple of compounds,
4-propenylguaiacol and tetrachlorocatechol, which should be
spiked at a higher level in order to more accurately determine
the method detection limit. There are another four or five
compounds that we should actually spike at a lower level in
another set of seven replicates. I'm a little hesitant to try
and spike less than .8 parts per billion.
Next slide.
This slide shows what the mass spectrum one of the
trichloroguaiacols looks like at this level. There are enough
masses there that I think we can identify the compound.
Next slide.
These are three masses present in this
trichloroguaiacol and they maximize within plus or minus one
scan.
Next slide.
Clean water is one thing and real samples are another.
This is a set of 18 matrix spikes that were run on everything
from in-process streams to pulp mills effluent samples. There
were a couple of receiving water samples in here. The reason
that they don't all have 18 valves is some had to be tossed
-------
354
because the spiking level, which varied from 15 to about 50 parts
per billion, sometimes was greatly exceeded by the natural amount
in the sample so we couldn't include that data. There were also
a couple of compounds that were added as we went along in this
study.
These data were found using stir bar extraction and
ascorbic acid. These data compare well with the NCASI precision
and accuracy data, except for the catechols which are better, as
you might expect from the addition of ascorbic acid.
Next slide, Larry.
This is representative of a real sample I thought you
might like to see what it looks like. Somewhere around here is
the region that we will be looking at a little later.
Next slide.
This is that same sample. The top trace represents a
25 microgram per liter spike of that sample. You can't really
see many differences. Overall it's difficult to tell that the
spiked sample is any different then the unspiked.
Okay, Larry.
The unspiked sample contained five parts per billion
3,4,5-trichloroguaiacol. This isn't a lot of area so 0.8 parts
per billion for this sample would be very difficult. What you
can detect mostly depends on what else is in the matrix.
Okay, Larry.
Again, this shows three masses and the addition of a
fourth that is in this particular compound, but you wouldn't be
able to see the maximum of that particular mass due to matrix
effects.
Okay, Larry.
I just included this to show you what the spike would
look like with 25 parts per billion added.
Okay.
This is a different sample, a treated effluent whereas
the previous one was untreated.
Okay, Larry.
-------
355
Although it looks messier, if you looked at the
intensity on the right hand side, you would see that it's about
20 percent of the intensity of the untreated sample. This is
actually quite a bit cleaner. We'll be again looking between
2,700 and 2,800 scans. Okay, Larry.
It is a lot easier to detect the compound in this
sample at 2.5 parts per billion than in the untreated effluent.
These have been our experiences with the GC/MS method
of in-situ acetylation on pulp and paper mill effluents. The
detection limits are ultimately determined by what kind of sample
you have, whether it's influent into a primary treatment or a
secondary effluent. It also depends on the mill process. If we
were to need lower levels, we'd probably need a cleanup
technique, foe example silica gel or alumina.
If there are any questions, I'll entertain them.
-------
356
QUESTION AND ANSWER SESSION
MR. TELLIARD: Any questions?
Thank you, Peggy.
Thank you very much. I'd like to thank this
afternoon's speakers.
Thank you for your attention. See you tomorrow morning
and see the rest of you at the boat tonight.
Thank you very much.
-------
357
fel
o c
«•*,
CJ
1 O
~i n
O
-~ ' n
'"V
a,
GO
t-<
v ° ai o
--• !*•• , - .—«
f
0 W
" 2
^3 XJ
0 0
«S'
^Si
o
s
o
t-.
s
5
;
n '
V j
JO'
o:
u1
U '
a:
cr
o
^
\^
t>%
M
MH
S
L —IL.
CO
^"o
- Ki S
-JDL,
o
cu
CO
CV2
o
X fi
j^
00
.1-1
6
0)
o
0)
f—4
a
l/J
f-Hf
0
,-H
M
*O
SH
O
CO
CD
PC:
p{
(0
a>
t .
rj
a,
'o
a
0)
"p.
o
0)
03
o
o
S-.
o
o
T3
a
cct
O
CO
0)
0,
K
i—<
o oo
(0 CO
'32
3
00 -
O T~l
O T-l
I—I I
5s
n o rt
'3 a
8 » »
0> fl .
OJ
0) O
g
I-
«o
fl
o
^
• r-l r-1
>m
c! o
N
o> M.
o
£10
cd
. o
0)
O
CO
-------
358
o
o
c
T- Q
Q Gt
h- IO
h. CO
Gl Gt
CM ^
Gl Gl
^
00 CO
Gl Q
—
O
C
o>
a.
^
X
— .c
o *-•
c
•
CM i-
Gt Gt
00 CD
Gt Gl
Gt Gt
t>- CO
O) h-
Gt Gt
— —
O O
c c
- Gl
o
c
*
(D .C
£ *-•
o O
^^ ^z
>s 1
£ CO
«- 1
-
Q —
1 J=
TT O
i
• 1
CM -
CM CO
i- CM
GIG,
•t h-
Gt Gt
QGi
CM CD
co r-
Gt Gt
— —
o o
c c
r-
Gt T-
Q Gt
in in
CO CM
Q 1-
—
o
c
0>
£: —
a. o
o c
o .c
— a
c. 0
O i_
— o
l_ —
1— t~
\ 0
in to
« •»-•
•* c
CM CL.
0)
^^
^^
CM
Q
0)
S
CD
CD
Q
•^
0
O
ro
—
ro
0
-* co
O) i-
CM
\ >i
T- CD
IS Gl
in o>
Gt T-
SGI
00 h-
Is- f-
IS ^
...
o —
o o
ro o
— ro
ro —
3 ro
Oi 3
O D)
i_ —
O >>
— —
.c —
0 <
Ii
I
Gt O)
Gt **
00 O)
r- CD
Gl Gl
CD CM
Gt CM
so
CO i-
oo - -t-l
1 1
CD m
• .
m •<»•
•
0
0)
CM
0
•"•
oo
r- m
Q Q
c
— —
o —
at —
c —
— c
a. ro
!_ >
•—
O i-
l_ >s
O co
— o
^ L-
0 0
Q i:
1 0
CD —
in (-
-------
359
-------
360
ru
-------
361
IS CO
-------
362
-------
363
CM
GO
CO O
z:®
o ocj _
a:
-------
364
-------
365
CO CO CO CO
O <5>-H ,
S£S!~
O X UJ . ..
•-•CM_imui
Of —^ ft <-N f 1
Q x zrzrz
•-* ^^o
-------
366
CM
i
10U> •-"
CO CO XT
00 00 U
f>- N- IJJ
CD ^^ E^ *
LUUJ
ins-
CM CN
.
o to
.0
k ••
I CM
occ
-------
367
coin
CO
CO
LU -.
mo
C3
CO
CO
COCO
.. .. LU
-------
368
CM
ID
00 O
SB
• •
u> CD
CM
s
••
u?
CM
-------
369
O
O
cn
H
D
^Q
O
- I (D -
cr» O rt 4*.
D M 0» Ul
H-O D I
O if 3* rt
3* o H n
h-1 < O H-
O 0» ^ O
^303*
O H- O M
< M 0» O
0» M ft l-<
3 H-(D O
H-3 O O
M O rt
H- M (0
3 O
- (D
ui rt
D P»
H- o
O 3*
Ul *>
^ ^
a\ u
I I
- I £ (D
Ul O 0> 3
i rr H-rt
D M (U 0>
H- O O O
n h o 3*
tO NJ tO
Ul
I
" I
a\ D
I H-
O
O
(D
3
O
O \Q
O C
rt H-
(D (U
o n
3" O
O M
H- H-
O O
tr 3"
o o
o o
PJ Q)
H- H-
pj PJ
O O
o o
f"< 0»
O H-
lQ O)
d o
pi O
H- M
O
O
O
^
o
(D
3
O
H- H-
O O
cr 3"
o o
O O (D
T3 H3 3
3-3-0
(D (D M
3 3
O O
3- o
3*
(D
3
O
ONJVDOOOOOO
Ul 00 O £» CT\ M 03 03 00
t-« M
i
0>
<
(D
*1
P
v£l 03
(D
0>
^ W
(D
O 0>
o n
<5
(D
^
en
O
«
o>
to
•o
Effluents
o
0-
Laboratory
H-
0>
(D
H-
rt
(D
-------
370
MR. TELLIARD: Good morning. We'd
like to get started this morning, please. If the folks in the
lobby could bring in your strawberry and sit down.
Our session this morning will start off with a subject
that we've heard a lot from you over the last few years on
methods consolidation and efforts along that line. Every time
we've had a discussion along that line, there's always some young
person in the audience who says, geez, why don't you guys get
your act together and bring the message together...showing
complete lack of understanding of the government.
But in an attempt to do that or the appearance of an
attempt to do that, which is even more important in government,
the agency has formulated, as usual, a committee...when in
doubt...to look at methods consolidation, laboratory
certification and along that lines-and we're going to discuss
some of those issues this morning.
Our first speaker is the Secretary of the Environmental
Monitoring Management Council. Ramona Trovato has been with the
agency 13 years...14 years?
MS. TROVATO: Eighteen years.
MR. TELLIARD: Eighteen years. So,
Ramona is going to talk about the EMMC this morning and then
we'll follow it by two presentations on methods consolidation.
-------
371
MS. TROVATO: Thanks, Bill. I
worked in an EPA Regional Laboratory for 13 years and I really
loved it. About five and a half or six years ago, I decided to
move to headquarters. About a year ago the new Deputy
Administrator, Hank Habicht, was contacted by EPA's Region 111
Administrator, Ted Erickson, who raised the question of methods
consolidation and recommended the formation of a committee to
address these issues.
Regional Administrator Erickson recommended that we
establish a group to coordinate agency-wide policies concerning
all environmental monitoring issues. This recommendation was an
important step. The Deputy Administrator valued data quality and
was willing to set uup an agency-wide committee to address
monitoring issues.
The Section 518 Report, which dealt with the adequacy,
availability and comparability of testing methods provided a road
map for establishing the environmental monitoring management
committee. The EMMC is co-chaired by the Assistant Administrator
for Research and Development and a Regional Administrator. The
establishment of the EMMC is exciting since it addresses cross-
program method issues. I was especially interested because after
having worked 13 years in a lab; having to do certain things for
RCRA and certain things for Superfund and other things for
drinking water and yet other things for NPDES and not really
thinking the quality of the data was much different from method
to method or program to program, I was looking forward to some
cross-program consolidation. At this time, the Deputy
Administrator asked me to serve as the Executive Secretary.
The EMMC charter includes seven charges.
The first is to coordinate agency-wide environmental
methods research and development needs; they have started doing
that. Second is to foster consistency and simplicity in methods
across media and programs. Third is to coordinate long and short
term strategic planning and implementation of method development
needs. Fourth is to promote adoption of new technology and
-------
372
instrumentation. Fifth is to coordinate development of quality
assurance and quality control guidelines. Sixth is to evaluate
the feasibility and advisability of the National Environmental
Laboratory Accreditation Program. The seventh one is a catch-
all: Any other activities that influence environmental
monitoring programs. So that sort of gives us a free hand to get
involved anywhere the committee identifies issues for resolution.
The EMMC is structured to allow the Deputy
Administrator to decide any issues that we can't resolve
ourselves. The Deputy Administreator gets at least an annual
update on what we're doing and what progress we've made.
The Policy Council identifies and addresses monitoring
issues. The Steering Committee goes ahead and starts looking
into the details of what we need to do and the ad hoc panels are
the actual people who are doing the work. Right now, we have
five ad hoc panels.
The Policy Council is co-chaired by the Assistant
Administrator for Research and Development since EPA's Office of
Research and Development plays such a key role in all of the
methods development issues, by Ted Erickson the Regional
Administrator from Region III. The members consist of our Deputy
Assistant Administrators, Deputy Regional Administrators, and
Office Directors.
This is the first time in the 20 years of the Agency
that we've had this kind of attention focused on the methods
issues, so we're very excited and believe that we have an
opportunity to make a difference in the way things are working.
The Steering Committee is run by the Director of our
Office of Modeling, Monitoring Systems and Quality Assurance.
The Executive Secretary of the Steering Committee is
David Friedman. Many of you may know him from his work on SW846.
The members of the Steering Committee are division directors arid
branch chiefs from our regional and headquarters offices.
The five panels that are set up right now include: The
quality assurance services panel which will focus on funding of
-------
373
quality assurance activities includings performance evaluation
samples.
The Methods Integration Panel is looking at developing
uniform test procedures across programs.
The Automated Methods Compendium Panel is developing an
Agency-wide data base of our analytical methods-called EMMI.
The Analytical Methods and Regulatory Development Panel
will ensure that our methods are factored into our regulatory
development process, earlier, so that we can have the method
ready to go when it's time to implement the regulation.
Last of all, the National Lab Accreditation Panel will
look into the advisability and feasibility of a national
environmental lab accreditation program.
The quality assurance services group is struggling with
the budget issues. They are addressing: How are we going to
fund QA and once we do get some funds, how are we going to
prioritize where the money goes? We were hoping to have a budget
initiative for FY '93, but it looks like it's not going to be
ready until FY '94.
The issue of National Environmental Lab Accreditation
was raised because the private lab community said that they are
being audited by states and they said that this is costing them a
fortune and it's taking a lot of time. The laboratories feel
that the audits aren't that different and that there could be a
better way. It was also recommended in the Section 518
Report that EPA investigate the feasibility and advisability of
national environmental laboratory accreditation. We believe that
it could benefit the agency's programs to have a National Lab
Accreditation program. The next question to address is its
advisability? And the advisability question revolves around a
number of issues. One is, are the states willing to adopt
reciprocity. I see that as one of the very big, important steps
that we have to come to agreement on.
One of the others is how are we going to administer the
program and how will it be funded? That's one of the things that
-------
374
the panel is now looking into and we're in the process of setting
up a Federal Advisory Committee so that we can have industry
associations and industry members join us in making
recommendations on the design of a program and how it should be
administered. Jeanne Hankins is going to be the Executive
Director of that group, and we plan to have our first meeting
sometime this summer.
All of you know Bill Telliard. Fred Haeberer is here
also. They both chaired the Automated Methods Compendium Panel.
This grew out of the list of lists that Bill has been running on
a shoe string for a long time. Bill and Fred made a presentation
to the panel about what they've been doing and what the system
included and the panel decided to adopt it as the agency's
Automated Compendium of Methods. Each of the programs agreed to
contribute to its support and update. Bill's the lead on taking
care of the Compendium. Again, this is the first time that the
agency has had in any one place all of our methods pulled
together. Bill was telling me that he's planning on having it
available in September of this year and I think that it's going
to make a big difference to us in all of our methods
consolidation efforts because it will show us where we have
methods overlap now. It's going to show us some of the
precision, bias, and detection limit information on each of our
methods. So I think this is a major step forward and a major
accomplishment for this group. It couldn't have happened if Bill
hadn't already been working on it for all of these years.
When I was in the lab, oftentimes I would be told,
we're going to start up this priority pollutant program or RCRA's
going to roll or whatever, and I would turn around and go try cind
find the method that I was supposed to use, pull it off the
shelf, start trying to set it up, and one or another of a number
of things could happen to me. First, maybe I couldn't find a
method at all. If I found a method, it couldn't get the
detection limit I needed. Sometimes, I couldn't get it to work
at all. So, one of the things we thought was very important was
-------
375
to get the methods development activities included early in a
regulatory development process. We will begin by asking: Are
methods available? Are methods already in existence that we can
use? If they're already in existence, then we are one step along
the way of consolidation. This will allow us to use what we
already have so that each program doesn't develop their own
methods unnecessarily.
Inclusion of methods development in the regulatory
development process has just started. We're going to try it for
a year and then at the end of the year, we're going to evaluate
how well it worked and see if we need to make any changes to make
it work better.
The last panel I want to talk about is methods
integration. Joan Fisk, who is up next, is going to talk in
detail about this, but I'll just say that generally this is the
group that's taking a look at what methods we have now, where the
possibility for consolidation exists and then recommend that all
of the programs adopt, perhaps, an existing method or make minor
modifications to an existing method so that it will be acceptable
to all.
They're looking at five methods right now including:
VOCs, metal digestions, ICP, semi-volatlies and microwave
digestion. They're hoping to have some results this fall and
Joan will be able to tell you more in detail about that.
So, with that, if there are no questions, I'll just
turn it over to Joan and she can get into the specific methods
integration activities that we're addressing right now.
Thank you.
-------
376
QUESTION AND ANSWER SESSION
MR. TELLIARD: Any questions? We:
have a question.
MR. MYERS: My name is Harry Myers.
I'm from Keystone Environmental Resources. My question is, is
there going to be any opportunity for industries analysts to
participate in this?
MS. TROVATO: In the Federal
Advisory Committee, once we get that set up for laboratory
accreditation, that will include folks from the laboratory
industry, laboratory associations, users of lab data like
industries, states, federal government and other federal
agencies. In some of the other panels... Right now, the Methods
Consolidation Panel only includes representatives from EPA and
other federal agencies. We start to get into a real problem
anytime we try to include industry folks because when we're
writing regulations, industry folks can't help us write them.
And so, the only way we can get there specific input is to ask
them to give us a factual presentation or, if we set up a Federal
Advisory Committee, then they can give us advice. But that is
the only way that we can do that legally. So right now, the only
panel that's going to have industry input, in the form of advice,
is going to be the lab accreditation panel.
MR. MYERS: I appreciate what
you've said, but also I believe there are good folks out there in
industry that know best the kinds of things they have to look for
with respect to their own waste waters and I suspect that they
have methods that might be useful to you that aren't presently
recognized.
MS. TROVATO: Yes, there's no doubt
about it, I'm sure. Ron Kites is here and he's going to talk
about one of the opportunities where we did have industry
involved.
MR. KITES: Sure.
MS. TROVATO: Yes, I thought so. I
-------
377
wasn't real involved with this. My husband, who also works in
the chemistry lab, was involved in this and they had a real
wonderful approach to coming up with which methods could be
consolidated, which ones we should skip, which ones we should
keep, and...
You're going to talk in detail about that right?
MR. KITES: At length.
MS. TROVATO: At length, yes. He's
going talk at length, so I don't need to say anymore. But, yes,
we did have some industry...
MR. TELLIARD: Fred brought up an
interesting point. If there are some methods out there that
people are using in an industrial category, we'd certainly like
to reference them at least in the list of lists and have them
available as far an information source. So, if you want to make
that available to us, we'll be glad to look at that since ours
isn't really a regulation; it's an information system. We could
do that.
NO SLIDES AVAILABLE
FOR THIS PRESENTATION
-------
378
MR. TELLIARD: Our next speaker
is Joan Fisk. Joan used to work in the Office of Water and then
she lost religion and went some place else. Joan used to be much
taller when she worked in the Office of Water, but now she's in
Superfund and you know how that stuff will hurt you.
Joan's going to talk about the Methods Consolidation
Panel and where the agency is presently standing.
-------
379
MS. FISK: Thank you. Bill. I
have the illustrious Dr. Fred Haeberer helping me with my slides.
I decided I had to go very high in the agency, in order to assure
the quality of the turning of my slides, so I've gone to the
Quality Assurance Management Staff.
The first thing I want to mention is that I did hear a
little riddle the other day about something that happened in
Cleveland. For those of you who don't know, this is where Bill
Telliard hails from. It's about a busy corner in Cleveland and
there were four people standing on that corner. One was Santa
Claus, one was the Easter Bunny, one was the perfect man and the
other, the perfect woman. The riddle is, on this corner in
Cleveland, which of these four people crossed the street first?
I thought the answer was real easy, myself...the
perfect woman. The others are just a figment of your
imagination!
The first slide shows the first three tenets of the
charter for EMMC as previously described by Ramona.
The next slide has the rest of them. I have six,
though there are several versions of them. Sometimes it shows
six and sometimes it shows seven. So, I pulled a couple of the
redundant ones together.
See, there's the first three tenets. What I'm pointing
out here are the ones that have an X next to them. Those are the
ones that I consider the methods integration panel as being
involved in. I think it's pretty clear that if you're
coordinating methods research and development, the method
integration panel must be involved. This is it's primary
function - fostering consistency and simplicity.
In order to coordinate methods research and
development, we must focus on consistency and simplicity if we're
going to be doing a better job and improving our methods.
The next one is "promoting and facilitating adoption of
improved and new instrumentation and technologies." We are
doing this with other government agencies.
-------
380
The gentleman who asked Ramona the question before
about whether the industry would have any participation in this,
Ramona was correct in that the panel and the work group do not
have any contractors or industry people involved. However, there
will be at some point in time mechanisms for contractor
participation.
The next one is "coordinating development of QA/QC
guidelines." I wouldn't call that QA/QC. I would say QC. The
quality control that we put in our methods in order to make sure
that they meet our needs is an intrinsic part of the methods that
we write, and while it may vary from program to program, it is
within the methods. So this also part of the charge of the
Methods Integration Group.
The next one is the only one without an X in front of
it and that's Ramona and Jeanie's charge of evaluating the
feasibility of a National Lab Accreditation Program. The methods
integration really would not be concerned about that at all.
And the last one, as Ramona said, is that kind of
bureaucratic one that allows us to do pretty much anything we
need or want. Definitely the Methods Integration Panel is doing
that.
Presently there are two ongoing efforts in the Methods
Integration Panel. Obviously, we have this plethora of methods
that Ron will be probably be showing you much more graphically
than I'm going to do, and figuring out how to integrate them is
one important part of the panel's job.
One of the things the panel did do to help with this is
establish a work group. I believe it's the only one of the
panels that actually has gone to another level.
As you can see from Ramona's description EMMC is kind
of hierarchical. I remember at the original panel meetings we
talked about a core work group that would be a panel who had
enough authority that they could make decisions and didn't have
to go several levels above to make final decisions. We also said
we needed rotating subgroups tailored to the different methods
-------
381
that we deal with. Now, I have not seen that show up in most of
the summaries of meetings, but this is the design we did have in
mind.
The other thing, of course, as was already mentioned is
there is interaction.
The other effort of the Methods Integration Panel is
probably the one I consider even more important because we're
coming together on our own to make the plethora of methods go
away. That's how we're going to do things in the future so we do
not have everybody doing their own thing, driving you people in
the industry crazy.
One of the things the panel decided to do, and the work
group agreed upon, was in order to sell our ideas that we really
wanted this to work, we thought that we would deal with some of
the easier situations first. We want to have early successes to
show Hank Habicht and Ted Erickson and all of the other people
who initiated EMMC that it was a good thing to do.
The three methods that were originally dealt with were
the digestion procedure for solids, which our air people at
AREAL/RTP did, integrating the various digestions that were
around. EMSL, Cincinnati, did the capillary GC/MS method for
volatile organics and EMSL, Las Vegas did the ICP spectroscopic
method for elements. Ultimately the validation of the ICP and
the digestion will occur together in that they are related. In
order to check out your digestions, you do have to do some
analysis and so we all felt that it made good sense to do a
coordinated validation of the two methods.
As far as the capillary column GC/MS, right now EMSL
Las Vegas is looking into the volatiles in soils and AREAL/RTP is
looking into the volatiles in air analysis. Cincinnati did take
the lead to do this with water and now the other procedures have
to be added.
The work group has chosen six other methods. We said
that we could do about six integrations a year. Ron Kites will
probably give you an idea of the number of methods that have to
-------
382
be integrated or can be integrated. We say this has a finite
end, and I'm not sure when that finite end is when we will have
resolved all of the issues with multiple methods.
One of the things that was decided also was that as we
start getting these methods integrated, we will no longer support
the cold methods after a period of time. We will be starting
Federal Register notices and give everybody time to be aware of
what's coming down the pike before we actually make these the
methods that should be approved cross programs whenever
possible. After methods are integrated and so forth,
they would be approved by the panel, the Steering Committee and
ultimately the Policy Council and would become the integrated
method. One thing I do want to mention here is if it can be
justified to have more than one version of a method, the panel
feels very strongly that this will be allowed. No one is going
to tell Bill that he can't use isotope dilution anymore, because
there is a need for him to do something different than what we in
Superfund are doing. Therefore, there would certainly be
allowances for different variations of methods to do the same
thing when there is, say, a regulatory impact or a regulatory
need, a need for, say, better precision, recovery, whatever. So,
there could actually end up being more than one version of an
analysis when there is a need.
Okay, the next group that was proposed to the Steering
Committee to be looked into were also ones that I think are all
very well-known to all of you, the semi-volatile organics by
capillary GC/MS. We're talking about Method 625, Method 8270,
the CLP Method. 8270 and the CLP Method are pretty much
identical now. While they don't look the same, they are
literally. If you follow procedures in the lab, they are exactly
the same.
Microwave digestion, this is one, again, where there
are a few different versions out there and continuous liquid-
liquid extraction. These three are already started; they are in
process. Here, I would think logically when we went through the
-------
383
validation, we would combine the continuous liquid-liquid
extraction and the semi-volatile organics by capillary column.
That certainly makes good sense.
Later on in the year, at the end of FY '91, we would be
looking into the organochlorine pesticides and PCBs. The
dioxin/furans by high res GC/high res mass spec. We're not
talking about the low res mass spec method. And last, graphite
furnace AA with a selection of elements.
The Steering Committee has been thinking about how the
Methods Integration Panel could be improved. One of the things
they are proposing is to have multiple work groups. It was my
understanding in the beginning to have tailored groups for the
different methods, such as separate workgroups for water, soil
and air aith tailoring to deal with spinfus (such as organic or
inorganic, simple preps, etc.). Also, we reiterated the need to
do the easy integrations first, the ones we that we can do
quickly and get the successes out there. The last thing, which I
think is very, very important is we should be emphasizing
pollution prevention. We should be looking into things like
solid phase extraction, micro extractions, things where we're
getting rid of our solvents or as much as possible in the process
and prevention of lab waste. This is a very, very important
piece that the Methods Panel has to be thinking about all the
time.
Future methods development. Here's where we're going
to have possibly a little bit of, shall I say, politics involved?
There are many programs in EPA. Every program is very
possessive. We all work very hard and we come up with what we,
ourselves, think are the best ways to do things and this is why
we have this plethora of methods and this is why we now have EMMC
to make this plethora of methods go away. There's been a couple
of options proposed on how to deal with methods development in
the future. I, myself, believe personally and I believe that the
other people on the panel agree that we can go with a couple of
different options and use them all and use them all effectively.
-------
384
The first option I've mentioned up there...and it will
go onto the next overhead in just a second, is one that really
has ORD, the Office of Research and Development, pretty much in
control of the process from beginning to end, with the work group
participation, of course, coming from the programs and the
regions and so forth. This would be where a priority setting
process would be developed and then these priorities for
chemical, physical, immuno-assay and sampling methods would be
addressed by these various work groups. We would need to add
people with a bio assay background to help in prioritizing
because this is something...we do not haverepresented now.
It is important to identify regional needs for methods.
There is a lead region who does a needs survey and I don't know
if they do this every year, but they do it quite routinely and
this is done from the Office of Regional Operations and State and
Local programs. We within Superfund have always...and also
OSW...tried to get this information. In fact, Ramona was our
source of this information for several years...to try to keep
apprised of where we needed to be going for Superfund. We had
lots of other ways of doing this, too. We look into our Special
Analytical Services requests to see where our contractor, Viar,
who does subcontracting to do special things for Superfund
through the CLP, to see where we are getting large volumes of
needs demonstrated. That's how we came up with the dioxin/furans
and the air toxic, radionuclides and things of that nature.
We also go to our clients on a routine basis. We
examine the ARARs. That's the "Applicable, or Relevant and
Appropriate Regulations to see where other legislations are
forcing us to perhaps make some changes to our methods or add
methods. Anyway, we look into all of those kinds of things and
these are certainly the kinds of things I think this work group
will look into, just as we do. This process would ultimately
implement these priorities as we've come up with them.
The other thing that has been proposed, and the panel
liked it, would still provide a little bit of ownership and pride
-------
385
of ownership to the various programs. E.G. The program has a
burning need. It does not want to sit around and wait for the
Office of Research and Development to get it prioritized and have
ORD address it. What we've proposed is that when we have a
need...something we really need badly...OSW does...water
does...pesticides...whoever... that they could go forward and take
the lead, but get EMMC...to coordinate participation in the
development of the method and make sure that we have all of the
right people there, the right Federal agencies, etc., and help us
coordinate that process and still be able to carry it through in
the same kind of work group manner that Superfund has been using
traditionally.
This is where I want to answer the question that the
gentleman asked before. We in Superfund have always included the
contractor community in our methods development work groups
because we always have believed that you are the guys who knows
what works in a lab if you're doing research versus what happens
in a large volume laboratory. It's a lot different if you're
trying to push 200 or 300 samples out a month than it is in a
more researchy-type atmosphere. We would like to think that we
can continue that process because we do have the attention of
upper management with EMMC, we would be able to get all of the
right people in the agency to participate to make sure all of
their needs were recognized.
Ultimately, after we have our work group product, we
would put it through the panel and Steering Committee for
approval, going through the Federal Register process and so
forth. We do believe that these two things are complementary and
that ORD certainly should be taking the lead in any of the ones
that are really, really researchy. But the problems themselves,
when it comes to just revisions or things to update or adapting
things, we really believe that this can be done with the program
in the lead.
The next thing I wanted to mention is that RCRA and
Superfund have worked very well together for about four or five
-------
386
years now. We are in the same office, the Office of Solid Waste
and Emergency Response. And we kind of made a pact. We found
that some of these little differences were driving our community
nutty. We decided to avoid redundancy in the methods development
effort and by avoid the redundancy, we were able to get more
methods developed. Instead of ORD working on just say, for
example, a volatile capillary method for Superfund and another
one for RCA, they could do the volatile capillary one for both of
us and then they could be working on a pesticide method for both
of us, too. This also, we felt, would be a tremendous relief to
our laboratory community so that they didn't have to keep their
RCRA samples and their Superfund samples separate all of the time
because they were changing temperatures or changing surrogates or
whatever.
That's why we decided to do it. We thought it made
sense. Now, how do we do that? We do our Research Committee
process together. We sit down together on and off over many,
many, many weeks and identify our common needs. Both programs
have research committee monies. You cannot mix our monies; So
we'll say, okay, here's all the stuff we need. Why don't we
recommend that we do all of X, Y, Z methods, using Superfund
money and we'll use RCRA money for all of these other methods
over here on this side and then we both kind of monitor the
progress of ORD together.
We also do methods development that does not involve
Research Committee monies. We pay for it directly out of our
program funds and this can be either to ORD contractors who do
methods development or our own contractors. Here again, we sit
down and we decide who's going to take the lead on each one. It
was kind of like, really, which one do you want to do? Which one
do you think is more fun? And then we just again kept each other
completely in the process, knowing what was happening.
We do have RCRA participation in all of our work groups
and we in the Analytical Operations branch have been
participating for many years on the RCRA SW846 work group.
-------
387
These are some of the things that we have done
together. There's a much longer list, but these are the ones
that came to my mind.
The first is the low res dioxin/furans the lead being
Superfund's Operations Branch. We're going to have an
Invitation For Bid out on the street within days right now and
OSW is in the process of rewriting it in their own format. The
method will not be different, but it will be in the SW846 format.
We also have what we call our Quick Turn-around GC
methods. This is for a 48 hour turn-around, once the samples hit
the laboratory. These are all kinds of streamline/methods so
that our clients can get the answers a lot faster.
The organochlorine pesticides, we did a lot of updating
and rewriting the methods for wide bore capillary columns.
The radionuclides and mixed wastes is a new effort we
have ongoing right now because we've got 116 odd sites on the
national priority list that are federal facilities, many DOE, and
some are EPA leads. This is something we're having to address.
Air toxics, which is a real big thing, particularly
with the Clean Air Act being reauthorized. We added new analytic
to the target compound list.
The last one, I didn't mention on the slide is field
development efforts methods. The majority of the field
analytical methods are being funded by Superfund and we are the
ones developing a compendium of methods. This is something that
RCRA is, of course, very interested in also.
RCRA has taken the lead on some of the, shall I say,
more sophisticated things. The LCMS, both thermospray and
particle beam. In fact, RCRA did a really fine job coordinating
the efforts. They've actually got the vendor
community...different members of the vendor community doing
different pieces, the different EMSL laboratories, everybody
doing their own part and I think it's worked very well. They're
also taking the lead in supercritical fluid extractions and have
been able to give many papers about their progress and I think
-------
388
we're going to probably have something working in the near
future.
The microwave digestion is kind of divided. There is a
soil and water digestion. We've got the lead on the water and
RCRA on the soil. RCRA is taking the lead on cations and also on
the high res mass spec for dioxin/furans. This is one where we
did try to coordinate and keep as much similar in the low res
mass spec methods as possible.
There are many others. As I said, RCRA has lots and
lots of methods that we don't need and we really only would want
to be kept informed of some of the things they're doing. We
certainly would want to have the methods available to us and
would not think about changing them if these were things that we
suddenly came up with a need for. We certainly don't plan on
going out and doing our own TCLP.
That's it. Any questions?
-------
389
QUESTION AND ANSWER SESSION
MR. TELLIARD: Questions?
MS. FISK: I'd like to thank
Dr. Haeberer for the quality assurance of my slides.
SPEAKER: I have a question.
I'd like to know if you are...
MR. TELLIARD: Could you tell us
who you are and who you are with?
MS. ASHCRAFT: Oh, I'm Merrill
Ashcraft, with the Navy Public Works Center. I would like to
know if you have privileged information as to when the new
addition to SW846 is being published?
MS. FISK: No, I don't. I think
possibly, if Jeanie Hankins is in the audience, she may know. Is
Jeanie here?
Jeanie, do you have an answer to the question?
MS. HANKINS: Well, we had all
hoped that it would be ready before the symposium in the second
week in July, but it looks like it may be coming out later this
summer.
MS. ASHCRAFT: Thank you.
MS. FISK: Anymore?
MR. TELLIARD: Thanks, Joan.
Thanks, Fred.
-------
390
8
5 2
P ^
G g
» &
I
B
*• V5
I §
§ £
I I
I
C) B
^^ ^^
§ H
I I
*i*
tl
*!
k^ ^
I*
«1
It
^^
^)
g ^J
^ ^
«5|
•«&;
!3
?!
^ ^
2?|
11
%*8
.
!
-------
391
1
o
I
&
63
&
<&
I
8
1
I
I
•«
&i §
S 3
§
Coordina
includin
.5
f
I
&
§
I-
r
^•^
•c *
^ G<9
18
b^ Q
-a
S
^
•«
0^0 «3 Q ^
. s ^ •« :$
I 8 I E $
llHI
ON Q 8 a S
-------
392
I
I
8
S
•-S
Coordinating
they apply to
S
S
*l
^ fc
^ ^<
If I
Q ^j
.S *Bs
it
k] o
I
5
I
§
I
.« s
3 ^
^ $A
•^^ ^"^*
«.s
Si
ii
« s
I
1
•9
^
1
^
^
ti
-------
393
a
I
1
O
s
£
•8
^
1
:«
e
1
*
^
*
^
*
§
§
•>
i
§
^
I
s
••§
•I
I
I
~S
-------
394
1
.eg
I
1
-------
395
I
p
I
:?
I
**
•s
1
111
s.
I
•^ •&
S KV»
»
I
SI
3
*b
•s
1
-------
396
I
*B
«w
£
03
2
3
S a
3
-
s
**
' 2-
a
•si
-------
I
^
s
397
I
•a
i
^^^^w
I
I
I
•*
i
•s
i
i
ll
M Q
fc ^
-------
398
s
Market Survey good idea for
methods development needs.
i
i
1
§
§
Si
tl
s^r^
S «s S
-------
399
**
I
*»»
I
I
i
«
1
1
I
i
i •
I|
^ •«
•sra
S^
§ s
I-I
Cd S
^ 8
«S
•^ 45
^ S5
§
1
I
6
a.
|
I
5N
"8
'42
^
^
s
fi
1
f
i
I
8
•8
EMMC Me
coordinatin
&> ^
1 ^
§J>s
•S
« s
« Q
-------
400
5
•8
? Methods Development (cont'd)
«
^
'«
Program lead workgroup would June
TQM manner as presently is done in
Superfund methods development.
s ^s
S* *25 7*
S MS *J
*SA ^ 1*^
o Workgroups include all EPA Pro±
other government Agencies, EPA
ORD, Headquarters, & the contrt
laboratory community.
9
EMMC approval process would folio]
§
-------
401
a 3
s §
ft ^
11
redunda
&
ethods
Si
-------
402
ts -
x **
5^
*> »« _•> »•>
Research Committe
>th our needs could
Hotted dollars/FTEs
^^ ?
s g
^ ^
^ *s*
^
^fe
•5^
^ &
2 ^
^
^ ^
^ 2
^ ?
S :2
s^:
a
itt
communicate
-------
403
g
K
fc
3
i
l
^^4
a
a;
>w
H
CO
5
^
§
•ft
^
•i
8.
1
^
^
Q
X
8
i
?
1
I
I
3
i
•a
>w
s
j
.8
I
5
i
-------
404
MR. TELLIARD: I think that you
can see there are a lot of efforts and a lot of things haven't
gelled yet, mainly because we've only been at it a few months.
Along the same line, we've kind of initiated an effort
to try to consolidate, at least in a small way, part of the
program and that is in the water methods. Ron Kites is with us
from the University of Indiana and he did a study, which we
affectionately refer to as the Kites Report. That could become
infamous. It's going to describe a little bit of what they went
through and how they did it.
I think that if you look at the methodologies that we
have been dealing with here, we've all kind of used the same
instruments. If you're using SW846 or 500 or 600 series methods,
when you get to the mass spectrometer or the ICAP or the AA, it's
pretty much the same. Where we find that there's a large
discrepancy is in the QA and QC and I think that's one of the
biggest areas where we're going to try to come to closure.
Since we know that QA/QC is pure science, which means
it's negotiable, that's going to be probably the biggest area of
work over the next year or two, to come to closure on some of
that. And that all ties into Ron's paper and I'd like to say
right now that it was the first shot at this effort and I think
if you haven't seen the report, if you want a copy or something,
we'll try to get them to you. Ron?
-------
lrLLi
405
THE U.S. EPA'S ANALYTICAL METHODS FOR ORGANIC COMPOUNDS
IN WATER AND WASTEWATER: THE NEXT GENERATION
Ronald A. Hites
School of Public and Environmental Affairs
and Department of Chemistry
Indiana University
Bloomington, Indiana 47405
and
William L. Budde
Environmental Monitoring Systems Laboratory
U.S. Environmental Protection Agency
Cincinnati, OH 45268
HISTORY
By the late 1970's, it had become clear to the U.S. Environmental Protection
Agency (EPA) that organic compounds were polluting many of the nation's waters. By
1977, as a result of a lawsuit by several environmentally concerned plaintiffs, the EPA had
focused on a list of 114 "priority" organic pollutants (7). These included such compounds
as the trihalomethanes (by-products of the water chlorination process), polycylic aromatic
hydrocarbons (well-known human carcinogens), and numerous compounds of industrial
significance (nitroaromatic compounds, for example).
The EPA's first task was to assess the prevalence of these compounds in both
wastewater and drinking water. Their long term goal was the regulation of specific
compounds that were found to pose significant environmental problems, a daunting task.
-------
Ir"-1-, 4Q6
Tens of thousands of samples needed to be -measured by hundreds of different laborato-
ries. Clearly, there were concerns about the comparability of data among laboratories.
The result was a series of laboratory-based analytical "methods".
These EPA methods are detailed, step-by-step directions (recipes) that describe
everything the analyst needs to know to complete a satisfactory analysis. Analysts must
have basic chemistry laboratory skills, which are not covered in the methods. The methods
include sample collection, preservation, shipment, and storage; instrumentation, appara-
tus, glassware and reagents; analytical calibration and quality control; and the calculation
and reporting of results. In most cases, the methods are directed toward specific environ-
mental matrices, for example, industrial wastewater. This approach is similar to that used
by standard-setting organizations (for example, "Standard Methods for the Examination of
Water and Wastewater" published by the American Public Health Association). The EPA
felt that their methods would eventually be used for regulatory purposes; thus, a tightly
focused set of directions would best serve the regulator and regulatee and minimize dis-
putes about analytical results.
During the 1970's, the first set of methods were developed; these were the "600
series" methods for the analysis of organic compounds in wastewater. Some of these
methods are applicable to a relatively small set of compounds (for example, Method 606 is
aimed at 6 phthalate esters), while others are aimed at a much larger set of analytes (for
example, Method 625 focuses on a list of 72 compounds which can be extracted from water
with methylene chloride). The 600 series methods are now required for monitoring efflu-
ents under National Pollution Discharge Elimination System permits and were promulgat-
ed in the regulations under Title 40 of the Code of Federal Regulations (40 CFR) Part
136, "Guidelines Establishing Test Procedures for the Analysis of Pollutants" under the
Clean Water Act (2).
-------
lrLh
407
In 1979 and in the 1980's, a set of "500 series" methods, focusing on drinking water,
was developed. For example, Methods 501.1, 501.2, and 501.3 are for the 4 trihalometh-
anes, and Method 525 is for 43 compounds that can be isolated from water by liquid-solid
extraction. The 500 series of methods was developed in response to the requirements of
the Safe Drinking Water Act (SDWA). Many of these methods are mandatory test proce-
dures under National Primary Drinking Water Regulations (40 CFR, Parts 141 and 142)
(3), and a number of other 500 series methods have been proposed for future regulations
under the SDWA (4). The complete 600 series methods are printed in the Federal Regis-
ter (2), and the 500 series methods are available from the National Technical Information
Service (5).
Another series of methods, the "1600 series", has evolved over the last several years.
Methods 1624 and 1625 are both described in the Federal Register rule of 1984 (2). These
two methods are the same as their lower numbered cousins (624 and 625), except for the
isotope dilution calibration procedures, which require an isotopically labeled standard for
each analyte. These standards are available, but they are expensive (a complete set costs
$2500-3000). These methods are particularly useful for extremely complex samples (such
as sewage sludge). For this reason, they should be retained in their present form, and we
will not consider them further.
By now, many of the 500 and 600 series methods are in wide-spread use, and it is
clear that there are considerable overlaps among the methods in terms of both procedures
and analytes. For example, Methods 524.1, 524.2, and 624 have 30 out of 62 analytes in
common, and they all use similar purge and trap gas extraction procedures.
There are also considerable differences between these methods. They reflect
various levels of analytical technology because of the different times at which the methods
were developed. There are also differences in the detection limits of the methods; the
drinking water methods generally have lower detection limits than the wastewater meth-
-------
408
ods. There are also some trivial variations among the methods. For example, gas extrac-
tion times and flow rates vary for the different purge and trap methods; there is a large
range of initial and final sample volumes; there is a bewildering array of maximum storage
times for the samples; mass spectrometric scan rates and mass ranges vary widely. These
differences have proven to be troublesome for analytical labs that must faithfully execute
these procedures. In some cases, this means having different sets of equipment to analyze
virtually the same set of compounds. This is both expensive and wasteful.
In response to these concerns, Section 518 of the Water Quality Act of 1987 direct-
ed the EPA to study the availability and adequacy of field and laboratory test procedures
and methods to support the provisions of the Act. This study resulted in a report to
Congress titled "Availability, Adequacy, and Comparability of Testing Procedures for the
Analysis of Pollutants Established Under Section 304(h) of the Federal Water Pollution
Control Act" (6). The report included the finding that "Improved coordination is needed
in the Agency's methods development program to avoid duplication in the development
and standardization of test procedures and inconsistencies in quality assurance and quality
control guidelines." As a result of this finding, Indiana University was asked by the EPA to
consider the question "Is it possible to revise or eliminate some of the 500 and 600 series
methods and effect a savings of time and money?" This and related questions were studied
and recommendations were developed.
Our study considers only the fifteen 600 series methods and twenty-four 500 series
methods given in Table I. These methods were selected because these are the methods
that are either mandated in the regulations or were recently proposed for inclusion in the
regulations under the SDWA. The EPA's Office of Emergency and Remedial Response
(Superfund) contract laboratory program (CLP) uses several methods that are derived
from 500 and 600 methods. While these methods are well-known because of the require-
ments of numerous EPA contracts, the unnumbered CLP methods are not part of any
-------
regulatory requirement; therefore, they were not considered in this study. Similarly, the
methods published by the EPA's Office of Solid Waste in the manual "Test Methods for
Evaluating Solid Waste" (7) contain many numbered methods that are derived from the
600 series methods, but the SW-846 methods are generally not mandatory in the Resource
Conservation and Recovery Act regulatory program; therefore, they were not included in
this study either. Although inclusion of the CLP and SW-846 methods is beyond the scope
of the current study, they should be included in a broader-based, consolidation study that
might be done in the future.
ISSUES
As we began our task, we had several concerns in mind. First, many of the methods
are somewhat out of date in terms of analytical technology. For example, capillary gas
chromatographic columns are rarely specified in the 600 series methods. Hydrophobic
traps are not used in the purge and trap methods. In some cases, hazardous solvents and
preservatives are specified; for example, diethyl ether and arsenic and mercury salts are
required in some methods. It seemed clear that a revision of these methods should bring
them in line with modern analytical technology.
Our second concern centered on the different sizes and types of laboratories which
use these methods. These include "Mom and Pop" and large contract laboratories and
small and large drinking and wastewater treatment plant laboratories. Some laboratories
are concerned with only one or two methods, while other laboratories are concerned with
virtually all of the methods. As we considered this issue, it seemed clear that revisions of
these methods should not make it prohibitively expensive for the small laboratory with
limited resources to meet their needs, while at the same time, the revisions should be
extensive enough to streamline the operations of large laboratories.
-------
Third, we were concerned with the great level of detail now specified in these
analytical methods. We wondered if this was necessary, either to produce good data or to
facilitate the operations of the laboratories.
Because our laboratories at IU or the EPA do not routinely use these methods, we
convened an expert panel of 24 "method users." This panel met for one and a half days,
and the discussions ranged over all aspects of the methods. This panel helped us form the
recommendations which follow. Despite their valuable assistance, we should make it clear
that the recommendations that follow are our own and are not necessarily the consensus of
the panel.
RECOMMENDATIONS
Ground-rules. There are several advantages to revising and eliminating various 500
and 600 series methods: The total number of methods would be less; thus, there would be
fewer methods with which the various laboratories would need to be familiar. This would
lead to less paperwork and result in cost savings for everyone. With fewer methods, there
would be more uniformity, and presumably, data quality would improve.
Consolidation should not be taken to its extreme. Rather, methods should be
consolidated only as needed to reduce redundancy. Let us explain. The complexity of a
method increases dramatically as the number of analytes the method is designed to cover
increases. This is illustrated in Figure 1; complexity is plotted in arbitrary units against
the number of analytes (note that this is a log-log plot). Two lines are shown in this plot.
One is for methods based only on gas chromatography (GC) and the other is for methods
based on gas chromatographic mass spectrometry (GC/MS). This graph shows that a
GC/MS method reaches its maximum complexity when one attempts to measure about
120 analytes at the same time. A GC method which does not use mass spectrometry
-------
lrLLi 411
reaches its maximum complexity at about 40 analytes. The numbers on this graph are
somewhat arbitrary, but they were derived from discussions with the panel mentioned
above. The lesson from this graph is that existing methods should not be consolidated in a
way that pushes the number of analytes beyond about 40 for a GC-only method or about
120 for a GC/MS method.
Another issue affected our thinking. The 500 and 600 methods can be broadly
divided into three groups: (a) those which use GC with a selective detector (such as an
electron capture detector), (b) those which use GC/MS, and (c) those which use high
performance liquid chromatography (HPLC). The last group represents only a very few
methods, but it is a group that will grow in the 1990's as more attention is focused on non-
volatile compounds. As we considered the pros and cons of the GC-only methods, it
became apparent to us that these were valuable methods, and they should be retained.
There were two reasons for this decision. First, these methods are useful as screening
methods. It makes no sense to use a relatively expensive mass spectrometer to analyze a
sample that has virtually no analytes in it. GC-only methods can screen out those samples
which do not require a follow-up. Second, GC-only methods can be useful in small labora-
tories, which measure only a few analytes in well characterized samples. For example,
until recently, small drinking water utilities only needed to measure the four trihalometh-
anes. There was no need for these laboratories to invest in more sophisticated equipment
when a GC-only method would meet their requirements.
Despite our acceptance of GC-only methods, laboratories which use them are
cautioned that the analytes must be properly identified. The only qualitative data that
GC-only methods produce is the analytes' retention times and responses on selective GC
detectors. This is frequently not enough information for an exact compound identification.
Additional qualitative data can be obtained by rerunning the sample on a GC column with
a truly different liquid phase. Another possibility is the simultaneous use of two or more
-------
r 412
selective GC detectors. For example, Method 502.2 specifies a photoionization detector
followed by an electrolytic conductivity detector. The ratio of detector responses helps to
identify a compound. Of course, if the sample were more complex than anticipated, GC-
only methods would not produce satisfactory results, and GC/MS would be essential.
As mentioned above, most 600 series methods do not specify capillary GC columns;
rather, they provide detailed instructions for the use of packed columns. Packed columns
do, in fact, have several advantages: They are less expensive than capillary columns (both
the GC instrumentation and the column itself). Packed columns are also simpler to oper-
ate, have a higher on-column analyte capacity, and sometimes allow for faster analysis
times. On the other hand, capillary columns have considerably higher resolution per unit
time. Thus, retention times can be measured more accurately and precisely, and many
more compounds can be distinguished from one another based only on their retention
times. Using wide-bore capillary columns, on-column analyte capacities can be reasonably
high, and analysis times can be reasonably low. Once a laboratory changes over to capil-
lary columns, analysts usually find that these columns are no more complicated to operate
than packed columns. Thus, we recommend that all methods should use capillary col-
umns. The only exception might be in cases where a screening method is needed. If it is
expected that a large fraction of the samples will be blank, then screening these samples
with a packed column could be acceptable.
Specific Recommendations. Having laid out the ground rules of our strategy (a
minimal consolidation to keep the number of analytes relatively low and retaining GC-
only methods), we can now present our specific recommendations. These are summarized
in Table II.
Our first recommendation is to eliminate Methods 501.1, 501.2, and 501.3. These
vintage methods use packed GC columns to measure the four trihalomethanes. Because
current drinking water regulations require monitoring for a broader group of volatile
-------
organic compounds than just the trihalomethanes and because these compounds are
covered adequately by other, more general methods (see below), these old methods are no
longer needed.
Methods 502.1, 502.2, 503.1, 601 and 602 are GC-only, purge and trap methods,
which use photoionization or electrolytic conductivity detectors. A total of 63 analytes are
covered by these five methods; 58 analytes are in two or more of these methods. Clearly,
considerable consolidation of the 5 methods is possible. In practice, we recommend that
this consolidation be done by simply keeping Method 502.2, which is already a consolida-
tion of Methods 502.1, 503.1, 601, and 602. These latter four methods are based on
packed GC columns while Method 502.2 uses capillary GC columns. While the number of
analytes (60) covered by Method 502.2 is above our implied limit of 40 (see Figure 1),
these are simple compounds which should be easily distinguished from one another on
capillary GC columns serviced by two different detectors.
Methods 524.1, 524.2 and 624 are also purge and trap methods, but they use
GC/MS as the detection system. These can be consolidated to form one method. This
new method would have a total of 62 analytes, 42 of which are in two or more of the exist-
ing methods. Again, this represents a major reduction of overlapping methods. This
consolidation is a fait accompli; Method 524.2 is the capillary column upgrade of Methods
524.1 and 624. Thus, we recommend dropping these latter two methods and retaining
524.2; this is a paper exercise that requires no further research.
Methods 515.1, 552 and 604 are for the measurement of halogenated acids such as
2,4,5-trichlorophenoxyacetic acid or 2,4,5-trichlorophenol. They all start with liquid-liquid
extraction followed by derivatization to form methyl esters or ethers. They represent a
total of 31 different analytes, only 4 of which overlap among the methods. Nevertheless,
because of the similarity of the extraction and derivatization procedures, we recommend
t
that these 3 methods be combined into one. This is a case where laboratory work is
-------
lrlh
414
needed to develop a revised method. The major experimental issues are: (a) Diazo-
methane is used in Methods 515.1 and 552, but it may not work for the less acidic phenols
of Method 604, some of which chromatograph well without derivatization. (b) Some of
the phenols now covered by Method 604, which specifies a flame ionization GC detector,
may not respond well in the electron capture detector specified in Methods 515.1 and 552.
(c) Method 515.1 uses diethyl ether, a hazardous solvent; this must be omitted without
sacrificing the quality of results, (d) It is not clear whether the volatile haloacetic acid
methyl esters (covered in Method 552) and the phenoxyacetic acid methyl esters (covered
in Method 515.1) can be separated in a reasonable amount of time on the same column.
Given these problems, it is entirely possible that these 3 methods could only be consoli-
dated into 2 methods.
Methods 525 and 625 address a total of 93 compounds; these methods, start with an
extraction step followed by GC/MS. There is some overlap between the methods: 22
compounds appear in both. Unfortunately, simply combining these two methods is fraught
with difficulties: Method 525 is an integrated liquid-solid extraction, capillary column
GC/MS method that works well with particulate-free water. There are data in the recent
literature that suggest that this method will fail even with particulate-free water containing
large amounts of dissolved humic acids (8). This method could be modified by adding the
much more robust methylene chloride extraction for wastewater and particulate-laden
waters and the cleanup sections of Method 625. This would probably result in an exces-
sively complex method which would be more difficult to implement than the current ones.
As a result of this thinking, we recommend keeping Method 525 as it is and revisiing
Method 625 so that it includes all the modern features of Method 525 but retains the
methylene chloride extraction and cleanup features of the old Method 625.
Phthalate esters are measured by Methods 506 and 606, and virtually all of the
analytes overlap between the two methods. These compounds are also covered by Meth-
-------
ods 525 and 625. Furthermore, phthalates are frequently laboratory artifacts. It requires
heroic efforts to make sure that these compounds are really in the sample at the levels
measured. For these two reasons we recommend that both of these methods be dropped.
Polycyclic aromatic hydrocarbons are measured by Methods 550, 550.1 and 610
using HPLC with UV fluorescence detection. The analyte lists for these 3 methods are
identical. We recommend that Method 610 be dropped; it is badly out of date. Unfortu-
nately, Methods 550 and 550.1 probably cannot be combined. Method 550.1 is limited to
particulate-free water with low humic acid content, while Method 550 is for particulate-
laden water. For this reason, both of these methods should be retained.
A total of 39 chlorinated pesticides are measured by Methods 508 and 608. There
is considerable overlap between the methods: 24 compounds appear on both lists. These
are both GC-only methods using electron capture detection. Method 608 should probably
be dropped in favor of Method 508; of course, the latter would need to be recertified for
wastewater use. This will probably require including the cleanup procedures from Method
608.
Methods 551 and 612 have no compounds in common. Nevertheless, we recom-
mend combining these two methods. Methods 551 and 612 both deal with chlorinated
organic compounds; they both use liquid-liquid extraction and electron capture GC.
Method 551 is a microextraction method, but 612 is not. Microextraction methods use
very small volumes of sample and solvent; typically 35 mL of water and 2 mL of the extrac-
tion solvent. We suggest that the combined method should be a microextraction method,
having a total of 27 analytes, a number that can easily be handled by a GC-only method.
Clearly laboratory work is needed to be sure that all of the 27 analytes can be separated
from one another and be precisely measured.
A variety of other methods have problems. They should probably be dropped and
the analytes which are left uncovered should be added to some other, more modern
-------
416
method. Method 603 for acrolein and acrykmitrile is a purge and trap method at an ele-
vated water temperature. This results in large amounts of water vapor entering the analyt-
ical system, which in turn, results in poor method performance. Clearly, more research is
needed to develop a satisfactory analytical method for these two compounds. Methods
605, 607, 609, and 611 deal with benzidines, nitrosamines, nitroaromatics, and haloethers,
respectively. The analyte lists cover a total of 14 compounds. These methods have serious
problems: For example, Method 605 uses chloroform, a solvent which many laboratories
are not permitted to use, and it uses an electrochemical detector which requires special-
ized skills to operate. Research is needed to incorporate these 14 "orphaned" compounds
in other, existing methods.
Methods 504 and 505 are GC-only microextraction methods for halogenated pesti-
cides and polychlorinated biphenyls, which use electron capture detection. There is no
overlap with any 600 method. These microextraction methods should be retained, and as
discussed below, given more emphasis.
Methods 513 and 613 both focus on only one compound: 2,3,7,8-tetrachlorodiben-
zo-p-dioxin (2378-TCDD). Method 613 is about 15 years old. This method is out of date,
and it should be dropped—there is little to salvage. Method 513 is a high resolution mass
spectrometric method, which was adapted from Method 8290 (7). This latter method
covers all of the dioxins and dibenzofurans containing between 4 and 8 chlorines. Method
513 was adapted lifted from the more comprehensive and complex Method 8290. This is
an approach which should be repeated in the future.
There were several 500 series methods that have no parallels in the 600 series.
These include Methods 507 and 531.1. Both of these methods are for nitrogen- and
phosphorus-containing pesticides, a compound class which is not addressed by any 600
series method. Therefore, both of these methods should be retained. Methods 547, 548,
and 549 are for glyphosate, endothall, diquat, and paraquat, compounds which are not
-------
417
addressed by any 600 series method. Thes& compounds present unusual analytical difficul-
ties, and these 3 methods should be retained.
Other Recommendations. Beyond the specific revisions outlined above, we have
several other recommendations. First, we recommend that the EPA should standardize
the format of official methods throughout the entire Agency and that the process of writ-
ing new methods be modularized. Each method should incorporate all necessary modules
to be complete.
Sampling needs to be addressed in a module. Specifications on the sample contain-
er should be included; toxic preservatives such as mercury and arsenic salts should be
avoided; and the minimum set of sample preservation specifications (including holding
times) should be given.
An extraction module should specify the physical procedures for extracting the
sample. These could include a continuous liquid-liquid extraction system or shaking or
stirring the sample with the extraction solvent. Liquid-solid extraction should also be con-
sidered in this module. Micro-extraction methods could be included as options for screen-
ing in each method.
A clean-up module should explain how the extract is fractionated so that the ana-
lytes of interest can be measured without interference. Beyond the existing liquid-solid
chromatography, this module could include gel permeation chromatography as an option
for samples that are especially "dirty". HPLC is also a useful clean up procedure and could
be included where appropriate.
The analysis module should specify how the isolated extract is to be measured for
the various analytes. It should give exact GC and GC/MS procedures. It should also be
clear about identification criteria for the analytes. For example, it should specify how a
compound is to be identified from its mass spectral or gas chromatographic data. Of
-------
lrlh
418
course, the identification criteria should be no less rigorous for an analyte measured by gas
chromatography than by mass spectrometry.
The calibration module must specify how the raw data from the analytical instru-
ment (either the GC or GC/MS system) are to be converted into concentrations. In
general, an internal standard for each analyte is the optimum calibration, but when this is
not feasible, a 3 to 5 point external standard working curve may be suitable. This working
curve should be defined by a regression analysis, and it need not be linear. It is, however,
important to bracket the analyte's concentration by the standard's concentrations. It is not
justifiable to extrapolate a working curve too far above or below the highest or lowest
measured concentration. Other issues that the calibration module should address are
calibration frequency and the acceptance criteria for the calibration. For example, the
maximum scatter of the data about the calibration working curve could be specified,
perhaps as a standard error of the estimate.
A reporting module is needed. The specifications on quantitation limits should be
given here. In a method for both drinking water and wastewater, the quantitation limits
will be different depending on the matrix. In a drinking water sample, which is presuma-
bly much cleaner than a wastewater sample, it should be possible to measure analytes at
much lower concentrations. We suggest that each method include a "practical quantitation
level" which is defined as the sensitivity for a given analyte in a given type of matrix that a
competent lab could routinely achieve (3). There would be different "practical quanlita-
tion levels" for each compound in each matrix encountered.
The reporting module should address another issue: All of the 500 and 600 meth-
ods focus on a set of target analytes. On occasion, mass spectral or chromatographic data
may be obtained indicating that a compound in the sample is abundant, but it is not one of
the target analytes. If the abundance of these compounds are high, their presence should
be reported even though their identity is not known. It would be a shame for a laboratory
-------
419
to report that all of the target analytes' concentrations are low but not to report that some
other, unknown compound's concentration was very high. This information should be
reported back to the organization that requested the analysis, and it should be their re-
sponsibility to decide what to do next.
The last module should be the quality assurance and quality control (QA/QC)
module.. This should include a section on the demonstration of competence with the
method and a section on reproducibility and accuracy. It is important, however, that the
quality assurance and quality control procedures be as simple as possible and thus maxi-
mize the ratio of data quality to cost. It does no good to have QA/QC procedures that
routinely take up 30-50% of the cost of running these methods. A way must be found to
minimize the cost of QA/QC, while at the same time assuring the regulator and regulatee
that the data are of sufficient quality to meet their needs. The QA/QC module should
include specifications and acceptance limits for all of the QA/QC data that are required.
For example, field duplicates are now required by QA/QC, but the methods do not tell the
laboratory how to interpret the data. The same problem exists for laboratory fortified
blanks and spikes.
Data should be continually gathered on the usage of the various methods. This
could be done by an annual survey or questionnaire sent to laboratories using the meth-
ods. Data could also be gathered in connection with annual performance evaluation
studies which are required for certification under the SDWA regulations. By gathering
data in these ways, the EPA can keep track of which methods are most important and
focus development efforts on those methods. Methods which are seldom used over a
period of 2-3 years should be proposed for decommissioning.
We recommend that all revised and new methods should have expiration dates.
When a widely-used method is about to expire, the EPA would need to recertify it, which
would allow for public comment on the method. An expiration date would ensure that the
-------
lrlh
420
most recent analytical technology would be considered for the method. Expiration dates
might cause an additional expense if many methods need to be extensively revised, but we
are convinced that the benefits outweigh the cost.
Before we address the issue of recipe-based versus performance-based methods, let
us define what we mean by a performance-based method. Rather than having an exact
specification of how all of the details should be carried out, the method simply specifies
the goals of the various parts of the method. For example, a large number of extraction
possibilities could be given, and the analyst allowed to choose among the various proce-
dures as long as a certain percent recovery is achieved. The advantages of performance-
based methods include flexibility and the ability to add new analytical technology as it
develops. The disadvantages include a need for a significantly higher level of chemical
skill and knowledge on the part of the method user. We have no firm recommendation on
which is better, but we do suggest that new methods should include ample options so that
an experienced method user can take advantage of technological improvements. "Ad hoc"
improvements, which eventually prove to be effective can be incorporated in the next
version of the method, presumably when it is recertified after its expiration.
Revisions to the EPA's analytical methods should aim to significantly reduce waste
from laboratory operations, especially hazardous wastes. One of the EPA's goals for the
1990's is "pollution prevention". Unfortunately, millions of liters of methylene chloride are
either evaporated up the hood or buried in the ground as a direct result o£ its use in offi-
cial EPA analytical methods. Methods based on liquid-solid and supercritical fluid extrac-
tion, which can reduce solvent consumption by about 90%, can contribute to this goal of
pollution prevention. For particulate-laden water samples, these extraction methods may
not work; thus, we need to retain liquid-liquid extraction methods. In this case, microex-
traction methods have many advantages. These methods are excellent for screening
samples; in some cases, they may be sufficient for the final measurement of relatively
-------
lrLh
421
clean samples. Microextraction methods use less solvent; therefore, they are less expen-
sive and generate less waste than the "macro" methods. Furthermore, the micro methods
are frequently much more rapid. For all of these reasons, we recommend that the use of
microextraction methods be expanded.
In considering these issues and in developing the recommendations presented here,
it became evident that the EPA faces a potential long-term problem: Where are the envi-
ronmental chemists of the future going to come from? The development of new methods
and the implementation of existing methods will require increasingly well-trained technical
personnel. Unfortunately, the number of new scientists entering the work force continues
to decline, in some cases (such as chemistry) dramatically. Clearly, a problem lies ahead,
and it behooves the EPA to do something about it. We recommend that the EPA fund a
system of training grants to academic institutions for the express purpose of increasing the
flow of trained scientific personnel into the chemical work force.
SUMMARY OF RECOMMENDATIONS
1. Some existing 500 and 600 series methods should be revised and some others
should be dropped. The set of changes that we recommend is given in Table II; the result-
ing methods provide for the determination of about 340 compounds and mixtures (such as
PCBs).
2. Gas chromatography in any revised or new method should use capillary columns.
This is necessary to provide sufficient resolution to distinguish analytes from one another,
especially when they are measured only with gas chromatography.
3. The format of official EPA methods should be standardized throughout the
Agency, and the writing of revised methods should be modularized. The different methods
should then incorporate the appropriate modules. This would promote uniformity among
-------
422
the methods and eliminate trivial but irksome differences.
4. A usage survey of the EPA's analytical methods should be conducted. This
would help in the development of new methods, in dropping old methods, and in the revi-
sion of current methods.
5. Revised or new methods should have expiration dates. This would force the
EPA to either drop the method or recertify it. Clearly, this would lead to the incorpora-
tion of the most up-to-date analytical technology.
6. The development of microextraction methods, using small volumes of sample
and solvents should be encouraged. These methods can be excellent but inexpensive
screening tests. They also prevent pollution.
7. The EPA should establish a system of training grants to help increase the pool of
environmental chemists.
ACKNOWLEDGEMENTS
We thank the expert panel for their help. This panel consisted of Roy Araki (EPA
- Region 10), Frank Baumann (California Department of Health Services), Fernando
Calera (Michigan Department of Natural Resources), Bruce Colby (Pacific Analytical
Inc.), Joseph Comeau (Aquatec Inc.), Alan Curtis (EPA - Region 8), Mike Daggett (EPA -
Region 6), Andrew Eaton (Montgomery Laboratories), James W. Eichelberger (EPA -
Office of Research and Development), Jerry Fair (Mead Technical Center), Floyd Geni-
cola (New Jersey Department of Environmental Protection), Edward Click (EPA - Office
of Drinking Water), Bob Greenall (EPA - Region 7), Nabih Kelada (Metropolitan Water
Reclamation District of Greater Chicago), Mary Khalil (Metropolitan Water Reclamation
District of Greater Chicago), Bart Koch (Metropolitan Water District of Southern Cali-
fornia), Pamela Kostle (Iowa Hygienic Laboratory), Arthur Lepley (Maryland Department
-------
of Health), Raghu Sharma (City of Indianapolis Department of Public Works), Dick
Siscanaw (EPA - Region 1), Joseph Slayton (EPA - Region 3), Kent Sorrell (EPA - Office
of Drinking Water), Myron Stephenson (EPA - Region 4), and Bob Teece (Pima County
Wastewater Management Department).
Several students at Indiana University did considerable groundwork, and we thank
them. They are Tony Borgerding, Louis Brzuzy, Donna Carter, Kelly Dodson, Carolyn
Koester, Mark Krieger, Voon Ong, Sandra Panshin, and Anne Starrett.
REFERENCES
1. Keith, L. H.; Telliard, W. A. Environ. Sci. Technol. 1979,13, 416-423.
2. Federal Register October 26,1984,49 (209), 43234-43439.
3. Federal Register July 8,1987,52 (130), 25690-25717.
4. Federal Register May 22, 1989, 54 (97), 22062-22160; and July 25, 1990, 55 (143),
30370-30448.
5. Address: 5285 Port Royal Road, Springfield, VA 22161; order numbers PB-89-
220461 and PB-91-108266.
6. EPA report number 600/9-87/030 September, 1988; available from NTIS at the
above address.
7. Office of Solid Waste, U.S. EPA, SW-846, Third Edition, November 1986; available
by this reference from the U.S. Government Printing Office.
8. Gauthier, T. D.; Seitz, W. R.; Grant, C. L. Environ. Sci. Technol 1987,21, 243-248.
-------
lrLh
424
TABLE I
LIST OF 500 AND 600 SERIES METHODS CONSIDERED
501.1 Trihalomethanes in finished water by the purge and trap method using gas
chromatography with electrolytic conductivity detection. 4
501.2 Trihalomethanes in drinking water by liquid-liquid extraction using gas
chromatography with electron capture detection. 4
501.3 Trihalomethanes in drinking water with gas chromatographic mass spec-
trometry and selected ion monitoring. 4
502.1 Volatile halogenated organic compounds in water by purge and trap gas
chromatography with electrolytic conductivity detection. 40*
502.2 Volatile organic compounds in water by purge and trap capillary column gas
chromatography with photoionization and electrolytic conductivity detection
in series. 60
503.1 Volatile aromatic and unsaturated organic compounds in water by purge and
trap gas chromatography with photoionization detection. 28
504. 1,2-Dibromoethane and l,2-dibromo-3-chloropropane in water by microex-
traction and gas chromatography with electron capture detection. 2
505. Organohalide pesticides and commercial polychlorinated biphenyl products
in water by microextraction and gas chromatography with electron capture
detection. 24
506. Phthalate and adipate esters in drinking water by gas chromatography with
photoionization detection. 7
507. Nitrogen-and phosphorus-containing pesticides in water by gas chromatog-
raphy with nitrogen-phosphorus detection. 46
-------
lrLL!
425
508. Chlorinated pesticides and polychlorinated biphenyls in water by gas
chromatography with electron capture detection. 38
513. 2,3,7,8-Tetrachlorodibenzo-/>-dioxin in drinking water by gas chromatogra-
phy and high resolution mass spectrometry. 1
515.1 Chlorinated acids in water by methylation and gas chromatography with
electron capture detection. 16
524.1 Purgeable organic compounds in water by packed column gas chromato-
graphic mass spectrometry. 48
524.2 Purgeable organic compounds in water by capillary column gas chromato-
graphic mass spectrometry. 60
525. Organic compounds in drinking water by liquid-solid extraction and capillary
column gas chromatographic mass spectrometry. 43
531.1 N-methylcarbamoyloximes and N-methylcarbamates in water by direct
aqueous injection HPLC with post column derivatization. 10
547. Glyphosate in drinking water by direct aqueous injection HPLC with post-
column derivatization. 1
548. Endothall in drinking water 1
549. Diquat and paraquat in drinking water by high performance liquid chroma-
tography with ultraviolet detection. 2
550. Polycyclic aromatic hydrocarbons in drinking water by liquid-liquid extrac-
tion and HPLC. 16
550.1 Polycyclic aromatic hydrocarbons in drinking water by liquid-solid extraction
and HPLC. 16
-------
lrLh
426
551. Chlorination disinfection byproducts and chlorinated solvents in drinking
water by microextraction and gas chromatography with electron capture
detection. 18
552. Haloacetic acids in drinking water by liquid-liquid extraction and gas
chromatography with electron capture detection. 8
601. Purgeable halocarbons in municipal and industrial wastewater by gas
chromatography with electrolytic conductivity detection. 29
602. Purgeable aromatics in municipal and industrial wastewater by gas chroma-
tography with photoionization detection. 7
603. Acrolein and acrylonitrile in municipal and industrial wastewater by heated
purge and trap gas chromatography with flame ionization detection. 2
604. Phenols in municipal and industrial wastewater by gas chromatography with
flame ionization detection. 11
605. Benzidines in municipal and industrial wastewater by gas chromatography
with special electrochemical detection. 2
606. Phthalate esters in municipal and industrial wastewater by gas chromatogra-
phy with electron capture detection. 6
607. Nitrosamines in municipal and industrial wastewater by gas chromatography
with nitrogen selective flame ionization detection. 3
608. Organochlorine pesticides and PCBs in municipal and industrial wastev/ater
by gas chromatography with electron capture detection. 25
609. Nitroaromatics and isophorone in municipal and industrial wastewater by
gas chromatography with flame ionization and electron capture detection. 4
610. Polycyclic aromatic hydrocarbons in municipal and industrial wastewater by
high performance liquid chromatography. 16
-------
r 427
611. Haloethers in municipal and industrial wastewater by gas chromatography
with electrolytic conductivity detection. 5
612. Chlorinated hydrocarbons in municipal and industrial wastewater by gas
chromatography with electron capture detection. 9
613. 2,3,7,8-Tetrachlorodibenzo-/>-dioxin in municipal and industrial wastewater
by gas chromatographic mass spectrometry. 1
624. Purgeable organic compounds in municipal and industrial wastewater by gas
chromatographic mass spectrometry. 31
625. Basic, neutral, and acidic organic compounds in municipal and industrial
wastewater by gas chromatographic mass spectrometry. 72
* The bold faced number is the total number of analytes on the target list of a
given method.
-------
lrLh
428
TABLE II
PROPOSED METHOD REVISIONS
Number Number
of of
analytes dupli-
cates
501.1. 501.2, 501.3
502.1. 502.2. 503.1.
524.1, 524.2, 624
515.1, 55^. 604
525, 625
506, 606
550. 550.1. 610
508, 608
551, 612
603
605, 607, 609, 611
504,505
513, 613
507, 531.1
547, 548, 549
Trihalomethanes
601, 602 Purge and trap by GC-only
Purge and trap by GC/MS
Haloacids by GC
Extractables by GC/MS
Phthatates by GC
PAH by HPLC
Chlorinated pesticides by GC 39
Misc. chloro compounds
Acrolein and acrylonitrile
Nitrogen compels. & haloethers 14
Halogenated compds.
2378-TCDO
Nitrogen & phos. pesticides 56
Unusual pesticides
4
63
62
31
93
7
16
39
27
2
14
26
1
56
4
4
58
42
4
22
6
16
24
0
-
0
0
1
0
0
Reconrnendat ions
Drop alI three
Keep 502.2, drop all others
Keep 524.2, drop all others
Combine into one method
Keep 525, revise 625
Drop both
Keep 550 and 550.1, drop 610
Revise 508, drop 608
Combine into one method
Drop
Drop alI four
Keep both
Keep 513, drop 613
Keep both
Keep all three
-------
429
1U
7
5
r^i
V^MM'
mmmi
X
r-H
O< 3
O
CJ
1
» t
/ /
/ .
/ /
> /
f r
t /
f f
t ^
t /
t .
• ' /
t f
/ f
t J
1 f
t /
/ /
/
/
/ /
/ /
//
GC only
GC/MS
__«___.
12 5 10 20 50 100
Number of analytes
-------
430
QUESTION AND ANSWER SESSION
MR. TELLIARD: Any questions for
Ron? Good morning, George.
MR. STANKO: Good morning, Bill.
George Stanko, Shell Development Company.
In keeping with the past history of this conference,
I've been asked to inject some controversy. I think the time is
opportune.
Ron, it seems to me...I don't know if this is a case of
the committee or yourself or your students, but there's a general
bias towards the 500 series methods. If the bias towards the 500
series is for the more comprehensive quality control, I think we
can concur that that was a good decision. If the bias towards
the 500 series methods was because it has, I must say,
unrealistically low detection limits, then we think that is wrong
and we have some concerns there.
You may not have been given all the facts, but I would
like to tell you that all of the criteria that was developed and
reported in the 500 series was actually done on the 600 series
methods and it's a mystery to me how you can have a lower
detection limit using the same data set. But that can happen.
One other thing that's obvious to me is that none of the methods
in ASTM, which are consensus standards, were ever shown on any of
your slides or did not appear to be even considered. ASTM has
the consensus standard methods for low molecular weight
chlorinated compounds which covers trihalomethanes, which is an
excellent method. It is an environmentally accepted method
because it's a mini-shake procedure and it's used by a lot of
drinking water plants. It's a very cost effective method,
environmentally acceptable method. I would like to have seen
that one considered.
Another thing, there's a PCB method in ASTM that is
also a consensus standard method and EPA was party of the round
robin and EPA, Cincinnati, is on Committee D1906 that had a major
impact on that method itself. One advantage of the ASTM methods
-------
431
is they are re-certifled on a five year cycle. This is one of
the things you recommended.
Currently, Method 524.2 is being considered for a round robin in
ASTM D1906. In fact, there's a Shell person who is a task group
chairman and industry and trade organizations have supported the
ASTM activity towards elevating Method 524.2 to a consensus
standard method. But nothing...either you weren't given all the
information about the ASTM activity in this area and I think it
really needs to be considered.
MR. KITES: Let me respond to
three points. The ASTM area is, in fact, one that we
deliberately avoided. We were told to focus only on the 500 and
600 series methods and, of course, we weren't in a position to
say, oh, we should do much more than that. I think that's
probably a more important policy issue for the agency as a whole
and, in fact, commended. I think if ASTM is going to do round
robin studies on official methods, that's absolutely great.
Let me respond to the issue of the 500 versus 600. We
did not discriminate against the 600 based on detection levels.
We recognize that the detection levels are going to be vastly
different between the two methods because the matrix is vastly
different and so we did not discriminate on that basis. We did
not discriminate on the basis of QA/QC. I hate QA/QC; it's just
too boring. So, we didn't even pay much attention to that. We
discriminated on the basis of age. The 600 level methods are old
technology. The 500 methods are...and this is an
oversimplification...the 500 methods are newer technology and
that's it. We recommend newer analytical technology.
MR. STANKO: Ron, I have one
further comment. Not only should you drop the methods for
phthalates, you really ought to drop the phthalates.
MR. KITES: I agree.
MR. VINOPAL: My name is Howard
Vinopal from the Army Environmental Hygiene Agency and I'd like
to say that in general I agree with your recommendations. One
-------
432
area, though, that I'd like to get a little comment from EPA
on... At a recent meeting, I talked to some people and they felt
that they weren't going to pursue the microextraction procedures
much further and I hear that you're recommending them, such as
505, which is a microextraction method. I think we need a little
better direction from EPA in this area because we're considering
some method development efforts in microextraction, in
improvement of the 505, which is a somewhat weak method in the
way that it extracts compounds. You must spike the samples with
standards and carry them through the procedure in order to
overcome deficiencies in recoveries. I would like to see a
little more direction on whether microextraction is what we
should be spending more effort on.
MR. TELLIARD: I think from the
Office of Water standpoint, it is something we're looking at and,
again, it's in relationship to the pollution prevention effort.
If we don't need a 55 gallon drum sample and then have to dispose
of it... And, of course, it's very dominant in the dioxin issue
where we have laboratories sitting around with whole storerooms
full of trash that they can't get rid of. So, we're going to
pursue it. In fact, this coming year we're going to fund some
efforts, both in the contract lab program and also with some of
our regional labs to look at not only microextraction, but
additional pollution prevention of solvent substitution and along
that line. So, it is something we're concerned about and we are
intending to move that way.
MS. FISK: I can second that for
Superfund and some of the research committee money. I was
talking before about the mystery process. A lot of that is
diverted now toward pollution prevention. I don't know where you
got your information because I don't think it's true.
MR. TELLIARD: The gentleman over
here.
MR. MOYE: Yes, good morning.
Anson Moye, University of Florida.
-------
433
Ron, in your considerations on how to reduce costs and
time in combining and doing away with some of the 500, 600
methods, one of the criteria has to be sample throughput yet you
did not mention that in your discussion. Did you have sample
throughput criteria, and if so, what were they?
MR. KITES: We didn't really have
that as an explicit criteria. It seemed in reading over the
various methods, the 500 and 600, there were not major
differences in throughput between them. I agree with you. That
certainly is an important consideration and I've got to believe
that as we go to more and more modern technology, that it has to
increase the sample throughput.
MR. MOYE: Well, I hope you're
right, but there are certain steps that can be taken in the
development of a method with old technologies that will greatly
reduce sample throughput, particularly in the extraction end of
the method, as you well know.
MS. ASHCRAFT: Merrill Anderson
Ashcraft from the Navy Public Works Center.
This is really directed at the panel. I don't know if you can
answer this question. I'm not someone who works a whole lot with
organic matrices so I don't know what the detection limits are in
those processes, but I do have to interpret a lot of data and use
that data in environmental regulations and apply it to our
industry. One of the problems that I've seen is, for example,
like in metal analysis, hazardous waste limits for chrome is
five. Well, I don't see a need to measure...
MR. KITES: Five what?
MS. ASHCRAFT: Five parts per
million. I don't see a need to measure one part per billion if
my limit is five. I would like to see a lot of the methods be
directed at regulatory limits. Also, some of the actual organic
species like tricholoethane, which is an F category waste, it
says 10 percent solvent. Now, 10 percent is a large number and I
would like to see some of the methods detection limits directed
-------
434
at actually what the regulatory limits are so that you could do
rough screenings and not worry about detection limits to find out
whether or not you have a hazardous waste. That's my comment.
MS. FISK: One of the things
that's happening at the agency, which you've probably all been
hearing about for years, is data quality objectives and the user
of the data is supposed to determine up front what quality of
data he needs and one of the things he would be concerned with
would be the detection or quantitation limits for whatever the
regulation was or whatever his use and he should then be
selecting the appropriate method or at least reviewing the data
from the point of view of that method. It may be that the method
may have a much lower detection limit than what you really need,
but when you're looking at it from your data review perspective,
all you're concerned about is the five parts per million. This
is the kind of thing where we in Superfund are constantly trying
to provide a gamut...a whole gamut of methods. Like I mentioned
before, the quick turn-around GC. These are basically screening
type things. Is there any of that junk in there or not? And
kind of direct where you do your sampling. You may find hot
spots and you say, okay, these are hot spots. There's going to
be real high concentrations and then you may want to send samples
on to a fixed laboratory and you would certainly not be asking
the laboratory to look for two parts per billion if you knew
these things were at two parts per million or 200 parts per
million. It's an ongoing process in the EPA today.
MR. HODGESON: Jimmie Hodgeson,
USEPA.
I was glad to hear most of your conclusions. Just one
issue though. On the organic acids, Methods 515.1 and 552, I
gave a talk on these yesterday on the research we're doing...not
on combining the methods, but on methods for simplification by
the use of solid phase extraction. They are not both for halo
acids. 552 is for haloacetic acids which are components of
chlorinated drinking water. 515.1 is for organic herbicides.
-------
435
These are, in fact, quite different in their chemical and
physical properties and the forms in which they are found. It
may or may not be possible to combine them.
MR. KITES: I agree. That's a
good point.
MR. TELLIARD: It's break time. I
want to thank the panel for this morning. If you folks will get
your strawberries and get back in here, we'll continue on.
Thank you.
-------
436
CO
Lu
U.
u tjjj
- ^r
<
z
<
U H
> O
U
MO
H
a >-
O Ct
O K- O
O H H
3 Z <
O tt
O
CD
<
< Z
-------
437
O
111
(0
O
O
Ill
z
U
H
*
111
u
III
(0
III
a
•
111
Z <
-------
438
>- I- 111
00 I- CO X
Ot H 111 t-
iH J h-
H
U. GO >. t-
O < OL *
J O
H C
O
GO
Ii
S°.
o
2
CO I Ul tt I
<0 Z X ul 3
ul O I- O -J
£ O Z J
log"
SI'-SK
?..aSh
< s g j
og^
u Q .r: u
o
in
ul
x
LU
ui
i-
111
0 H < CO CO
S2 Q ° S O
3 P g i 55
O < 111 H
5 ui o >
2*aoo
£ M z or C*
£ o < Q. o.
u
CO
Q M u H i *
u S i- d o H
I- < O O
D ^ •"• "l
si:*is
? H ro I-
H co z
-i H z o
i H CO O O
L 111 Cfl >- H
£ -I < -J I- Z
w H tt < U O
.. H < Z Ul M
2 H a < co H
-------
439
i Id
z z
^" f™^
O O D O
H Z Z Z
RA
t 5
< M o
f 08
^ O.
id
O Q
Z u H
S S! Z
g Z id
H §
^ en O
Id
Id
Z |S S
Kg3
°- S of
O g 3
-i S to
>§3
Q °- .
w H H a
£ -I Z
H Id < H
i Id
of n:
u i-
H fif
Z id
=> o
H
< to
z z
< o
H O
o
z o
^^ t™
o
H
Ob ^0 &^
O Id Z
H <
Of
Id Id m
to to sr
H 5
Id
* iO u.
O
o o
I- z to
< o
-|S
o H
id t to
§ ^ §
^ u ^^
_J M T
o S H
' o g
I
CO
fc ^
Of s
O O >-
Q. Id O
bJ > Z
Of O Id
Of O
Id °-
u.
z °
o
H z
H4fm
«r O
s£
-3
•jBk n
O id
O
Id
M U.
Of H
" O
to -I
Id O
H of
O H
z z
Id O
I- O
to
Q Q
H Z
O <
< to
Z H
O -J
U <
H O
ss
* S
S^
v)
Id
Of
Id
gu>
tfl u
2 H
u. o
I- O Id
H U.
Id ^
^> 1 W
r S Q
Z
o
Id
H
to
cn
H H O >-
V) Z Z Id
Id H H Z
3 -I Id O
O Id Z Z
-------
440
o ui
o z
VO O
H 111
I- tt
H m
2 «
LI 3
i§
3P
^ 8
z S
OH
H LJ
O U-
Z O
o
H
H
H •
HI Z tO
H- HI £
O X
' H
g III 2
2- E °
1!<
tt Q^
O o «*>
O t/l t/l Q
5 UI Qi O H
S M UI I 3
§ C* X I- O
5 UI J- III CO
g to o z <
111
-------
441
111
III
02
111
X
o
III
X
o
CO
I-H
LU
Q.
g
O.
O
C\J
LU
CO
O
Q.
O
in
in
- z
CM <
• X
«H h-
O 111
LO z
O
- _J
LO
CM
10
CM
- >• O
i-H J LT>
O Z
O O.
111
111
rH CD
CO >-
O 10
to
Q.
* <
CM a
• H
CM
o o
LO Z
i-H 111
• o
CM Di
O 3
LO Q-
o.
o
oi
CM
CM
LO
111
O Ul
CO
>-
CM CO
VO
Q.
* <
CM a
• H
CM O
LO Z
r- 111
• O
CM 3
LO Q-
-------
442
o
111
111
z
o
o
z
Id
Z
M
CD
O
O
in
CNJ
vo
u
u
in
CM
in
o.
u
u
o
0.
o
0*
O CD
(O
- CO
CM
in -
u
K-
o <
vo -i
m
-------
443
VO
Q.
O
00
VO
a
s
O
o
111
z
III
in
ID
CO
LO
LJ
H
UJ
o
z
H
UJ
LO
LO
O.
Ill
UJ
vo
- o
LO
LO
LO
LO
O
CD
CO
CO
UJ
O
H
O
H
UJ
00 CL
O
VO O
£
* O
00 -1
O X
LO O
(0
O
o
o
o
o.
o
o
o
<±
o
CM X
i-l O
to
•
- o
rH (0
LO H
LO
-------
444
Q.
O
O
U.
O
CQ
Q.
Ill
111
Q.
g
Ill
H
I-
H
O
o
o
UJ
X
111
O
08
rH
-------
445
co
vo
a
8
o
o
Q.
U
111
111
U
tt
CO
U)
Q.
U
UJ
id
o
H
o
o.
U
bl
U
a
O
oc
o
o.
O>
CO O
I
- 00
CO 1^
rH CO
in CM
CO U
LH O
O
I-
O HI
U
o
H
h-
u
- a
oo
<
3
*
-------
446
H S
< -
is
° 3
<
I
0 o
o
"I
CM
o
H
0
111
HI X
O D
O 0
tt Z
Q. H
o i-
(0
111
O
H
Z I
in o
X O
-J CC
< x
H O
u
U o
ui
u
3 < O
O
(/>
111
111 O
Z U
O tt
>
Ul
U
•~ o <
C* 111
I- U- Z
• 111 (/) 111
- >
=> U. -I >- X
-I U. < id O.
O 3 Z X <
O
-------
447
H
O
O
OI
111 Ik
o o
Q S S
O <
X 111
H I- >-
111 < h-
Z H H
tt Z
I- a a
zoo
u a u.
cd a H
in a z
u. < 3
-------
Id i
I- O
O
L.
O HI
UJ O
448
l O H
Q. -I Z
O O HI
111
O 3
Z O
U
O
< 2
(/) O
^ 111
o
H
to
-I 111
111 * £
X <0
O U
O O -r
O
3
H tO
h- H
z >\ iE
2 <*-
o
z
<
111
o
U
U
5 u- - •
^ • o to to
O 00
to in o o
^ I- I- X X
< O Z H H
Q. 3 UJ HI HI
a E E E
-------
449
O QJ
H Z
> o
111 H
£
H
0.
111
O.
O
§
•
Z -J
I- tt
H <
U U
O
(0
O
111
Z
II. >
o o
o
85
P2
111
o
o
in
U >
Z Ik
I- H
in a
u u
a o
o u
u- a
o
H
i
o
a
o
u
u
o
B
Ul <
Z Z
I- <
-------
450
I
H
Ill
z
o
<
3
IU
>
U
O
U U U
ui -j iE
CQ III ^
O
o S
o
* I-
z S
2 P
bl
Q.
fl^j ^^^
z u
o
O
• ^^« B^B
Z H
^ HI Z
< • O. Ill
10 O X >
III III III
IL O Z 0£
o < H a
U
-------
RECOMMENDATION 7. THE EPA SHOULD ESTABLISH A
SYSTEM OF TRAINING GRANTS TO HELP INCREASE
THE POOL OF ENVIRONMENTAL CHEMISTS.
in
-------
452
MR. KING: The next session is on
other methodologies. We're talking about everything from robotic
BOD to LCMS and there could be a change of schedule with the
LCMS. Bill Budde has a late arrival, so we're anticipating
switching Henry Kahn's 1: 45 paper on analytical variability
into the 11: 15 slot for playing purposes. So, with that,
we'll get ready to start.
Wayne Michalik of Shell Oil is going to talk a little
bit about automated BOD. Wayne has been working on laboratory
automation and robotics for about six years now and is just
trying to meet the limits and all of that good stuff that you
guys are really familiar with.
-------
453
AUTOMATION OF THE BIOCHEMICAL OXYGEN DEMAND (BOD) TEST
FOR WASTE WATER
W.A. MICHALIK, R.M. ZERKEL and J.B. MAYNARD
SHELL OIL COMPANY
P.O. BOX 262
WOOD RIVER, ILLINOIS 62O95
SUMMARY
The procedure for determination of the Biochemical Oxygen
Demand (BOD) of waste Mater given in Standard Methods 16th
Edition (1985) has been automated with a Zymark Laboratory
Automation System using a custom BOD workstation. The BOD
workstation does all glassware and sample manipulations
except movement of the filled BOD bottles to and from the
incubator. The system handles all sample transfers, as well
as the additions of dilution water, biological seed and
phosphate buffer. Dissolved oxygen readings are made before
and after incubation with a YSI Dissolved Oxygen Meter. The
sample transfers and additions of dilution water are made
with calibrated peristaltic pumps rather than using
conventional pipe.ts. In addition, the robotic system has a
workstation to wash and rinse the BOD bottles after use.
The precision of the BOD results obtained with the robotics
system has been equal to that obtained using the manual BOD
procedure and has met the precision requirements prescribed
in the standard method for the glucose/glutamic acid mixed
primary standard.
Analyst involvement with BOD testing has been reduced by
about 75X since the procedure has been automated.
-------
454
INTRODUCTION
Most analyses of environmental samples such as waste water
require a considerable amount of analyst time to do
repetitive tasks such as sample transfers, sample dilutions,
reagent additions and pH adjustments. The determination of
the Biochemical Oxygen Demand (BOD) of waste water, as
described in Standard Methods, 16th Edition (1), indeed
requires a high level of analyst involvement and testing is
often required 24 hours per day, seven days each week. The
automation of BOD testing should allow an increase in
labo'ratory productivity by freeing the analyst from routine
analytical procedures.
The BOD of waste water is proportional to the amount (ppm) of
soluble, biodegradable organic compounds present in the
water, and as such, is an indication of the performance of an
effluent.biotreating system. Most operating refineries or
chemical plants have maximum BOD levels for the waste water
that they can discharge from the manufacturing facility which
are stipulated by governing and controlling agencies, such as
the Environmental Protection Agency. Thus, accurate
measurement of the BOD is necessary to assure the quality of
plant waste water.
The BOD method consists of completely filling a bottle with
a waste water sample to exclude contact with air and
incubating the bottle at 20 Degrees C for five days. The
dissolved oxygen (DO) in the waste water is measured before
and after incubation. Most waste waters contain more oxygen-
demanding materials than the amount of DO in air—saturated
water. Therefore, the waste water samples are diluted to
several different concentrations before incubation in order
to bring the DO demand and supply into appropriate balance.
Active bacterial growth requires nutrients and a fairly
narrow pH range. To supply these needs, a biochemical seed
solution and phosphate buffer solution is added to each
sample dilution before the incubation is started. The BOD
(mg DO/L) is computed from the difference between the initial
and final DO readings divided by the volume fraction of
sample in the bottle.
To meet environmental testing requirements at the Shell
Manufacturing Complex in Wood River, Illinois, BOD analyses
are done four days each week on a number of different aqueous
effluent streams. As many as seven dilutions are required
for each waste water sample, and this amounts to an average
of 3O sample preparations on each of the four days. The
manual manipulations required for a BOD analysis consist of
-------
455
transferring water samples to BOD test bottles, adding
biological seed solution, buffer solution and dilution-water
to the test bottles, taking dissolved oxygen (DO) readings on
the resulting sample solutions and calculating the BOD for
the sample. The time spent by an analyst doing these manual .
manipulations amounted to 16-20 hours of a 40-hour week. In
addition, the analyst who prepared a series of BOD samples
would often be working a different shift five days later when
the samples were removed from the incubator. Thus, another
analyst had to make the final DO determinations on the
samples and calculate the BOD. Robotic automation of the BOD
test (except for moving the BOD test bottles to and from the
incubator) has significantly reduced the amount of analyst
involvement with the test, and by repetition of each step in
exactly the same manner, has removed inconsistencies that
could have occurred when more than one analyst was involved
in running the test.
This article describes the laboratory robotic system that
performs these manual manipulations required for the
determination of BOD in refinery waste water. The automated
method adheres to the standard protocol specified for the
manual procedure (I).
EXPERIMENTAL
A Zymark "Zymate II" Robotic Laboratory Automation system has
been used to automate the BOD analysis of waste water. Since
BOD analyses require multiple sample dilutions (with the
addition of biological seed solution and phosphate buffer
solution to each dilution), it was necessary to determine the
most efficient way to do these multiple dilutions and reagent
additions robotically. The Zymark Robotic Pipet Hand is
accurate, but extremely slow when transferring volumes from
1 to 250 milliliters (ml). Since the sample load in our
Environmental Laboratory averages 30 dilutions per day,
parastaltic pumps and pinch valves were used to deliver
sample and dilution water accurately and in a reasonable time
period. Sample and dilution water transfer lines are 1/4-
inch OD Tygon tubing.
The Zymate II Robotics System is installed on a mobile bench
equipped with 5-ft x 7-ft upper and lower bench areas. A
schematic of the equipment located on the upper bench is
shown in Figure 1. This consists of a BOD bottle wash
station, a BOD work station, a 15-bottle incoming-sample
rack, two 15-bottle sample-dilution racks and a robotic-hand
-------
456
parking station. The glass BOD bottles used with the system
are conventional 3OO-ml volume BOD bottles with glass caps
purchased from the Wheaton 51 ass Company.
The BOD bottle-wash station consists of the following:
- A BOD bottle holder with an optical sensor to confirm the
bottle is in place.
- An aspirator for removal of sample from the bottle.
- A single-head variable-speed peristaltic pump to add soap
solution to the BOD bottle.
- A dual-head variable-speed peristaltic pump to aspirate
sample from the BOD bottle, add rinse water to the BOD
bottle and to pump the drain of the wash station when
adding soap solution or rinse water.
- A pinch valve to allow aspiration of sample from the
bottle or draining the bottle holder.
- A pinch valve to control the addition of soap solution or
rinse water to the bottle.
The pumps are located on the lower bench and the pinch valves
on the upper bench near the bottle-wash station. Aspirating
the sample from the BOD bottle, washing and then rinsing the
bottle are being done while the final DO readings are being
made after incubation. After the final DO reading is made on
a sample, the BOD bottle is removed from the BOD work station
holder, positioned under the wash station aspirator and the
sample removed. The bottle is then inverted and placed in
the bottle washing section, where the position of the bottle
is confirmed optically. A strong, low-foaming soap solution
is pumped into the bottle coating all surfaces on the inside.
Rinse water is pumped into the bottle to remove the soap.
While this process is being done, the DO reading is being
taken on the next sample dilution.
The BOD work station consists of the following:
- A holder for BOD sample bottles, which includes a sample
transfer cannula, cannula wash station and stirrer.
- A holder for empty BOD dilution bottles, which includes
the YSI DO probe, nozzles for addition of the biological
seed, buffer solution and dilution water, a level-sensing
thermistor and an air line to remove water droplets from
the thermistor.
-------
457
- An apparatus to remove the glass stoppers from the BOD
bottles and an optical sensor to confirm removal of the
stopper.
- A 100 rpm fixed-speed peristaltic pump and pinch valve
used to transfer sample from the original sample bottle
into the BOD bottle.
- An AT&T personal computer, the System V controller for
program operation and System control and an Okidata
Microline 182 printer.
The sample-transfer pumping system is calibrated each day
before analyses are started. The BOD work station dispenses
the biological seed and buffer solutions using a Zymark
MasterLab station. The MasterLab station is equipped with a
lO-ml syringe for the seed solution and a 5-ml syringe for
the phosphate buffer solution. Both solutions are added
directly to each sample dilution.
The sample dilution water and DO probe rinse water are
dispensed using a dual-head, 600 rpm, variable-speed
peristaltic pump. One pump head dispenses dilution water and
the other dispenses the rinse water. The rinse water flushes
the DO probe between readings and washes the sample-transfer
line and sample-transfer cannula. The dilution-water pump
is calibrated before sample dilutions are started.
A low-temperature thermistor on the BOD work station is used
to insure the proper amount of dilution water has been
dispensed before initial DO readings are taken. After the
initial aliquots of dilution water, biological seed solution
and buffer solution have been dispensed, an initial amount of
dilution water is added that is needed to bring the total
volume of water in the BOD bottle to 306-ml (the average
neck-full volume of the 3OO-ml BOD bottles). If the
thermistor is not in contact with the water in the bottle
after the initial amount of water has been added, dilution
water is added in 1-ml increments until the thermistor
contacts the liquid. This insures that there is an adequate
amount of water in the bottle for the DO probe to make a
proper reading, with no free air trapped in the bottle. The
maximum volume of any of the BOD bottles used was found to be
315-ml and provisions were made in the control program so
that no more than a total of 315-ml can be added to any
bottle, regardless of the thermistor reading. The thermistor
is also used after the DO reading is completed to insure that
there is enough water in the bottle to make a water seal when
the glass stopper is replaced.
-------
458
The DO readings are obtained using the YSI Model 58 DO Meter
equipped with a self-stirring BOD-bottle probe. This meter
has been easy to calibrate and maintain. It is interfaced
directly to the robotic system, which controls the stirrer
and captures the DO readings in memory. The DO probe is
placed in the BOD bottle and allowed to stabilize for 90
seconds. Then, DO readings are taken every 2 seconds until a
stable reading of ± 0.05 mg/1 is achieved. This final DO
reading is captured in memory.
A schematic of the equipment on the lower bench is shown in
Figure 2. This equipment consists of:
- Five 20-liter plastic reservoirs (three used for rinse
water, one for soap solution and one for dilution water).
-All pumps and controllers except for the sample transfer
pump.
- The MasterLab Station for adding the biological seed and
buffer solutions to the diluted samples.
- The containers of the biological seed and buffer solutions.
- The Power and Event Controller.
- The Liquid Handling and Event Controller.
Dilution water is conserved by a loop arrangement. When the
dual-head pump is moving rinse water to the BOD work station,
the dilution water is circulated. While dilution water is
being dispensed, rinse water is delivered to the two wash
stations in the BOD work station. The three rinse water
reservoirs have been plumbed together and a constant fill
apparatus installed. The dilution water and soap solution
reservoirs are filled independently. The cleaning solution
is purchased from Baxter Scientific Products (Micro
Ail-Purpose Liquid Cleaner).
Before sample preparation is begun, the analyst adjusts the
pH of each waste-water sample (pH = £>. 5 - 7.5) and places the
samples in the designated locations in the incoming sample
rack. Table 1 lists the basic steps of the BOD procedure.
-------
459
TABLE 1
STEPS FOR ROBOTIC BOD ANALYSIS OF WASTE WATER SAMPLES
- CALIBRATE SAMPLE AND DILUTION WATER DELIVERY PUMPS.
- MOVE SAMPLE AND BOD BOTTLES FROM RACKS TO THE BOD
WORK STATION.
- ADD 5O ML DILUTION WATER, 3 ML BIOLOGICAL SEED AND
1 ML PHOSPHATE BUFFER TO THE BOD BOTTLE.
- ADD NECESSARY AMOUNT OF SAMPLE TO THE BOD BOTTLE.
- FILL BOD BOTTLE WITH DILUTION WATER.
- TAKE INITIAL DO READING FROM THE BOD BOTTLE.
- STORE INITIAL DO READING IN MEMORY.
- REPLACE SAMPLE BOTTLE AND BOD BOTTLE IN THE
RESPECTIVE RACKS.
- WHEN ALL SAMPLES HAVE BEEN DILUTED, THE ANALYST
MOVES THE BOD BOTTLE RACK TO THE INCUBATOR.
- INCUBATE THE SAMPLES FOR 5 DAYS.
- AFTER 5 DAYS, THE ANALYST MOVES BOD BOTTLE RACK
FROM THE INCUBATOR BACK TO ROBOTICS SYSTEM BENCH.
- TAKE FINAL DO READINGS FOR EACH BOD BOTTLE.
- CALCULATE AND PRINTOUT THE BOD READINGS FOR EACH
DILUTION OF EACH SAMPLE.
- WASH AND RINSE EACH BOD BOTTLE AND REPLACE THEM
IN THE BOD RACK.
-------
460
Following is a more detailed step-wise procedure for
calibration of the sample and dilution water dispensing
systems, for sample preparation and DO determinations.
Step 1 : The program is started by the analyst. The
BOD hand is retrieved and the DO Probe storage
bottle (a) removed and placed in a reserved
position in the incoming sample rack.
(a) The DO probe is stored in this bottle
partially filled with water when not in use.
Step 2 : The analyst is asked by the system whether or
not to do pump calibrations. The pumps are
calibrated before each set of sample
dilutions.
Step 3 : If calibration is desired, the analyst inputs
the amount of time (120 seconds) to warm up
the pumps before calibration.
Step 4 : After pump warm up, the analyst decides which
pump to calibrate and enters (1) for the
sample-transfer pump, or (2) for the dilution-
water pump.
Step 5 : The transfer line is purged with dilution
water and the analyst is prompted by the
system to check for air bubbles trapped in the
line. If no air bubbles are found, the
calibration continues. If air bubbles are
observed, further purging is done.
Step 6 : If pump calibration is required, the analyst
places a lOO-ml graduated cylinder under the
proper water dispenser and informs the system
that the cylinder is in place. Then, 5O-ml of
water are dispensed if calibrating the sample-
transfer pump, or 100-ml of water are
dispensed if calibrating the dilution-water
pump. The analyst observes the amount of water
dispensed into the cylinder.
Step 7 : The analyst inputs the volume of water that
was dispensed and that value is saved in the
system memory. Then, a calibration factor for
that pump is calculated and is automatically
entered into memory.
8
-------
461
Step 8 : Steps 5-8 are repeated for calibration of
the other pump.
Step 9 : After calibration of the pumps has been
completed, initialization of the system is
done automatically.
Step 10: Testing samples is started by removing the
first sample bottle from the incoming-sample
rack and placing the bottle on the appropriate
sample holder at the BOD work station. The
sample-transfer line is then purged with the
sample.
Step 11: While purging the sample-transfer line with
sample, the first dilution bottle is obtained
from the dilution bottle rack, the glass
stopper removed and the empty bottle placed
into the dilution-bottle holder on the BOD
work station.
Step 12: Next, 5O-ml of dilution water, 3-ml of the
seed solution and 1-ml of the phosphate buffer
solution are dispensed into the BOD bottle.
The designated volume of sample for the first
dilution is transferred into the dilution
bottle. The amount of dilution water that is
needed to bring the total liquid volume in the
bottle to 3O6-ml is added to the bottle. The
level of water in the bottle is checked by the
low-temperature thermistor to assure that an
accurate DO reading can be obtained. If the
water level is low, 1-ml increments of water
are added until the correct level is obtained.
Step 13: The sample transfer line and the dilution
water line are removed and the DO Probe is
inserted into the BOD bottle for the initial
DO reading.
Step 14: The DO Probe is given 9O seconds to stabilize
before taking the first DO reading. Readings
are then taken every 2 seconds until a stable
measurement is achieved. This value is stored
in memory.
Step 15: A check is also made to assure that enough
water is in the bottle to obtain a proper
water seal using the low temperature
thermistor, as described previously.
-------
462
Step 16: The dilution bottle is removed from the
holder, the glass stopper inserted and the
bottle placed into the same location in the
dilution-bottle rack.
Step 17: Steps 10 - 16 are repeated for all dilutions
required for each sample to be analyzed.
Step 18: After all samples have been diluted, the
analyst is alerted by a flashing blue strobe
light which means that the sample dilutions
and initial DO readings have been completed.
Step 19: The analyst transfers these initial DO
readings from memory to a disc file, places
plastic caps on each BOD bottle to preserve
the water seal and places the racks containing
the diluted samples into the incubator, where
they remain for five days.
Step 20: After the five day incubation period, the
analyst replaces the racks c.ontaining the
diluted samples on the robotics bench and the
system is activated to make the final DO
readings for each dilution of each sample.
Each BOD bottle is washed after the final DO
reading is completed.
Step 21: After all the final DO readings are completed,
the BOD values are calculated for each
dilution of each sample and the final report
is printed. Washing of the BOD bottles is
continued until all have been washed and
replaced in the BOD bottle rack.
DISCUSSION
Since the BOD analyses has been automated with the Zymark II
Robotic System, the time spent by an analyst in BOD sample
preparation has been reduced from 16 - 20 hours per week to
less than 5 hours per week. Lowering the amount of analyst
involvement in BOD testing has resulted in at least 15 hours
of additional analyst time each week that can be channeled
into more involved, non-routine analytical procedures.
10
-------
463
The accuracy and precision of the BOD test obtained with the
robotic system has been equal to that obtained when running
the BOD tests by the manual method. The prescribed standard
for measuring the accuracy and precision of the BOD test
given in the Standard Method (1) is a 50/50 mixture of
glucose/glutamic acid. Acceptable performance of the BOD
test for this standard mixture is an average BOD of 200 ± 37
mg/1. This standard mixture is routinely analyzed by the
robotic system and was analyzed by the manual BOD method for
comparison. For 4O analyses, the automated BOD analyses of
this standard yielded an average value of 205 ± 31 mg/1 BOD
compared with values of 218 ± 34 mg/1 BOD for the manual
method. Thus, the automated BOD results obtained for the
standard mixture have been more accurate at essentially the
same level of precision.
Since the robotics system was installed in September 1989,
good reliability has been obtained. However, a problem was
encountered in delivering the correct amount of dilution
water to the BOD bottles. The water volume dispensed into
the bottles was 5-1O ml short of the 306-ml final volume
required. It appeared that the thermistor for sensing the
liquid level in the BOD bottles was malfunctioning. However,
it was determined that short volumes of dilution water were
being dispensed. The pinch valves were being opened at the
same time the pumps were started, but the pumps require at
least five seconds to attain full speed. As a result,
smaller amounts of water were being dispensed than the pumps
had been calibrated for. In this instance, the thermistor
system for detecting the water level in the bottle did not
continue to add water because the calibrated delivery system
had already determined that the amount of water in the bottle
exceeded the maximum volume (315-ml) that had been previously
determined for the BOD bottles being used. Thus, with this
low level of water in the bottle, the initial DO reading was
in error because the DO probe was suspended in air, not in
the water sample. Inserting a 5 second wait after pump start
before starting to dispense the dilution water solved the
problem.
After 5 months of operation, a dilution water line ruptured
from fatigue of the Tygon tubing material. To assure that
this does not occur again (and result in losing a substantial
number of BOD measurements), the dilution water lines are
replaced every two weeks. Further preventative maintenance
procedures includes replacement of the YSI DO Probe membrane
every two weeks and replacement of the sample transfer lines
at least every 3 months.
11
-------
464
CONCLUSIONS
The automation of the BOD procedure from Standard Methods (1)
with the Zymark II Robotic Laboratory Automation System has
been a success, resulting in accuracy and precision equal to
that obtained when performing the test manually. Further, a
75% reduction in analyst involvement with the BOD test has
been realized. Acceptance of the robotic system by the
analysts has been good because it is easy to understand and
operate.
REFERENCE
Standard Methods for the Examination of Mater and Wastewater.
16th Edition, American Public Health Association, 1O15
Fifteenth Street NW, Washington, DC, (1985), pp 526-531.
12
-------
465
ROBOTIC SYSTEM UPPER BENCH
FIGURE 1
BOD
WORK STATION
INCOMING SAMPLE
BOTTLE RACK
DILUTION BOTTLE
RACK NO 1
DILUTION BOTTLE
RACK NO 2
BOD BOTTLE
WASH STATION
HAND
PARKING
STATION
PERSONAL
COMPUTER
SYSTEM
PRINTER
-------
466
ROBOTIC SYSTEM LOWER BENCH
FIGURE 2
DILUTION WATER LOOP
LIQUID
HANDLING
A
EVENT CONTROLLER
AIR
LINE
POWER & EVENT
CONTROLLER
MASTER LAB STATION
MAKE UP HATER TEED -»-
RINSE WATER TO WORKSTATION
RINSE WATER FEED
DILUTION WATER
PUMP CONTROLLER
RINSE WATER PUMP
CONTOLLER
RINSE WATER TO WASH
SAMPLE A8PERATOR
A
RINSE WATER PUMP
SOAP WATER TO WASH
SOAP WATER PUMP
DILUTION WATER
RESERVIOR
RINSE WATER RESERVOIR
RINSE WATER RESERVOIR
RINSE WATER RESERVOIR
SOAP WATER RESERVOIR
wrntoop
-------
467
MR. KING: Our next speaker is
Merlin Bicking. He's with Twin City Testing. He's currently the
Manager of Research and Development there and he's going to talk
a little bit about supercritical fluid extraction.
-------
468
THIS IS AN UNEDITED VERSION OF THE PRESENTATION BY THE SPEAKER
MR. BICKING: We've got a slight
audiovisual problem.
Okay, can you put the first one up...preferably the
other way? Thank you.
When Bill and Harry called us earlier in the spring a.nd
asked if we could talk, we said, sure, I'd love to talk about
SFE. And Harry's comment was, fine, as long as you don't say the
same things you did two years ago. So, Harry, if you're out
there, I've changed the title slide and a few other things since
then.
Could I have the first slide, please, and the
transparency can go off. Turn the transparency off for now.
First slide, please.
Really, what I'd like to talk about is an experimental
design approach that we've been using for optimizing
supercritical fluid or SFE conditions for a number of
environmental applications.
Next slide, please.
We have two primary objectives here. One is we're
going to be using an experimental design approach for
optimization. In keeping with our method development activities,
we feel this is a better and more efficient approach to
optimizing methodology and we'd like to use that information to
look at some environmental applications.
Next slide, please.
SFE, if you're not familiar with it, is an extraction
process that uses a supercritical fluid rather than a typical
liquid solvent.
Next slide.
In terms of potential advantages, in general, you get
faster extractions, with 15 to 16 minutes being typical
extraction times versus an overnight soxhlet for many of our
standard methods.
-------
469
Your selectivity is different because we're using a
different extraction medium and also because we have other
modifiers we can add that will also affect our selectivity. In
addition to that, we can vary the selectivity through the
operating conditions, which are temperature and pressure. In
other words, we really have almost... With these two factors
here, we almost have the equivalent to an infinite number of
solvents at our disposal.
Finally, the SFE approach is compatible with a number
of other analytical techniques.
Next slide, please.
There are some concerns which you're probably already
familiar with. Certainly, what are the optimum conditions for
extraction? At this point in time, we have a limited database of
applications. There are some excellent examples published where
SFE works very well, compared to standard methodologies and,
obviously, there are a number of conditions where SFE does not
work well and, of course, those are generally never published.
Certainly, if you can optimize the conditions, will your
extraction efficiency be what you expect it to be? If you can
get everything out, will you get out other things you don't want,
the usual problems with selectivity and matrix interference. And
finally, if you can get it out of your sample, can you collect it
at some point?
These are really, I think, the four critical parameters
that have to be solved or at least optimized if we're going to
use SFE as a really viable extraction method.
Next slide.
We were using a statistical approach rather than trial
and error for optimization and I'll discuss the central composite
design in a minute.
Our approach is to simultaneously optimize two
experimental values and in the process generate data over a wide
range of operating conditions, with a relatively few number of
experiments. This statement is actually incorrect. We are
-------
470
fitting a mathematical model to the data, not the other way
around.
Next slide.
As much as I'd like it to be the other way around,
we're actually doing it the right way.
The central composite experimental design approach,
also called a star square design, allows us to look at
temperature and pressure and in this case, we will choose three
levels in each factor, generating nine separate extraction
conditions and then a regression using a full second order
polynomial model involving six coefficients. We have first order
terms in temperature and pressure and second order terms in
temperature and pressure plus an interaction term.
Next slide.
This is a combination phase diagram and experimental
design summary. Every talk dealing with supercritical fluids is
required to have a phase diagram of carbon dioxide so here it is.
This is one of the compulsories for SFE talks. If you're
familiar with the critical point for carbon dioxide, 73
atmospheres 31 degrees centigrade... Above that point, we're in
the supercritical fluid region or the supercritical region.
The nine dots there represent the central composite
experimental design. It really is a combination of two
techniques.
The star which you see in the middle is really the
conventional one variable at a time optimization scheme. You set
the, in this case, the temperature. By the way, the numbers here
are...the first number is the temperature and the second number
is the pressure. You set the temperature and you look at several
pressures and then you set the pressures and you look at several
different temperatures and that's really the way most of us have
done our optimizations over the years.
Superimposed over this is a square or a box and that is
a conventional factorial design where we're, through these four
experiments simultaneously looking at high and low values for
-------
471
each of the variables. If you throw all of those together, you
get the central composite design and it does allow you to work
with a full second order polynomial as a result.
Next slide, please.
Again, a summary, nine extraction conditions. We're
using a dynamic extraction for 60 minutes. That is a continuous
flow through the cell and we're collecting the sample in a vial
which contains glass beads and a small volume of solvent and then
we are using conventional analytical techniques for analysis. We
are not using SFC.
Next slide.
These experiments were performed in a Super X SFE50.
Next slide.
The sample is held in this sample cell. We're using a
5 mL extraction vessel. There are some switching valves which
control the pressurization of the system and the effluent exits
from the top of the oven out through the restrictor area, which
is shown on the next slide.
We've mounted our restrictors a 40 centimeter length of
50 micron fusilica capillary. It's simply a simple pressure
restriction device. It's heated externally at about 50 degrees
centigrade and then our collection is in the vial with glass
beads here. The tip of the restrictor extends down into the
glass beads.
Next slide.
There are a number of extraction cells available. This
is an old Super X model. There's a new finger-type version which
is available commercially and then this is marketed by Keystone
Scientific, the TCT modification of the Keystone Scientific one.
This is what happens if you tell the technician to
tighten it until it doesn't leak anymore.
The TCT modification will probably not be used in most
laboratories.
Next slide, please.
We have two main applications. One of them is oil and
-------
472
grease and the other is a subject that's dear to Bill Telliard's
heart. That's the dirty, deadly and dangerous dioxins.
Actually, you're looking at a suite of EPA methods here with oil
and grease. There's a 413.1, in addition to the 413.2, and also
a 418.1. Our reference mixture is hexadecane and chlorobenzene
in this case and in both instances, we're using clay or celite or
diatomaceous earth as our reference majors for our optimization.
Next slide, please.
This is a summary of the oil and grease methods, at
least of 413.2, a 20 to 40 gram sample, a soxhlet for four to
eight hours in freon. The sample or the extract is concentrated
to 100 mLs and analyzed by FT-IR. This is an infrared method at
29-30 wave members so we're looking at aliphatic CH stretches.
So it's a general screening tool for general... for aliphatic
hydrocarbon contamination.
The 413.1 is simply a gravimetric analysis from this
extract.
418.1 involves passing the sample with silica gel which
removes the esters and leaves the petroleum hydrocarbons behind,
the simple aliphatics and the 418.1, called the TPH or total
petroleum hydrocarbons.
Our comparable SFE method uses a smaller sample size,
about six mLs of freon and a 60 minute extraction versus four to
eight hours. In this case, we dilute to 10 mLs, rather than
concentrate and then again can do FT-IR.
Next slide.
Just as an overall summary of how these methods are
working, our soxhlet procedure generated 97 percent recovery for
four...actually in this case eight replicates. This was spiked
hexadecane onto celite. We have a room temperature shaker method
which we also use, which gives slightly lower recoveries and then
the SFE method generated about 91 percent recovery. This is a
summary of the 13 extractions performed for the experimental
design. Not all of them optimized. So even including the non-
optimum conditions, we still get a very high recovery with SFE.
-------
473
Next slide.
If you pull out the results from the regression from
the experimental design approach, we get a profile that looks
like this in terms of temperature on this axis and pressure on
this axis. It's a fairly flat response surface. But in this
triangle in the lower left half, it's really the high recovery
region. All of the recoveries in this area are in excess of 90
percent and the recoveries fall off a little at high temperature
and high pressure. But basically over a wide range of
conditions, we can get high recovery of hexadecane from celite.
We were very pleased with that result. It meant we
probably had a fairly rugged method. We next evaluated SFE with
several real samples and one low level soil by all three methods,
the soxhlet, shaker and SFE approach. We found a very good
agreement between the soxhlet and the SFE approach...comparable
standard deviations. These are for a minimum of four replicates.
High level soils...higher level, I guess... around 1,000
milligrams per kilogram. We find a little bit lower recovery in
the SFE numbers, about 60 to 70 percent of the corresponding
soxhlet number and somewhat better agreement with some of the
shaker...one of the shaker numbers.
We're encouraged by this because we're certainly within
a factor of two in each case for the SFE versus soxhlet.
Next slide. Could we have the transparency on now,
please?
These are some more recent data we have which are a
more rigorous comparison of the methods. Again looking at the
gravimetric numbers, we actually have two different SFE
experiments now, a dry collection which is simply allowing the
C02to evaporate into a volumetric in this case and then weighing
the viable before and after and then collecting in freon and
using the conventional EPA methods for a workup.
We see for the dry collection excellent agreement for
the oil and grease number between those two methods. A little
-------
474
bit low, but not too bad for the TPH number. The freon
collection is just over half the oil and grease number and
probably 70 to 80 percent of the TPH number. In this case, the
SFE vial was actually heated to constant mass to make sure there
was no water in that gravimetric determination and this sample
was extracted as received in all experiments and these are
summaries of at least four replicates in each case.
Next transparency, please.
This is a second sample which has somewhat higher
levels. In this case, we get good agreement, again, between the
SFE numbers and the soxhlet values by gravimetric means. The oil
and grease value for the freon extract from SFE is very
comparable...very good agreement with the soxhlet number. And
again, the TPH numbers are also in good agreement. In this case,
the dry collection seems to be a little bit low. I'm not exactly
sure why. In think in general we're finding the SFE numbers vary
between 50 and 100 percent of the corresponding soxhlet values.
This sample was moist and it was mixed with sodium sulfate before
extraction, which is the conventional procedure for soxhlet. We
also used that same approach for the SFE experiments. But you
can see in general we're generating comparable data sets with the
SFE approach.
That's all for the transparency. If I could have the
next slide, please?
The next thing we wanted to look at was a comparison of
SFE with the conventional procedures and with our existing setup,
we felt a single technician could prepare about eight soxhlet
samples a day and roughly six SFE samples. We get high recovery
in both cases. The cost of materials is weighted heavily in
SFE's court there. Most of this cost...in fact, all in this case
is the cost for freon. We're estimating with rinsing and
glassware cleaning, it requires about 500 mLs of freon. Freon is
presently costing us about $100.00 for four liters. The cost
here for SFE are roughly 10 mLs of freon and about $1.25 for the
-------
475
actual supercritical fluid, the supercritical C02' So, there's
considerable savings in this case, especially when we're only
charging about $30.00 for the analysis to begin with. To have
half of it tied up in solvent costs means it's not a terribly
profitable venture.
The SFE approach offers a lot of advantages there and
this six per day is a very conservative estimate on the low side.
This does not incorporate any automation which is going to be
coming in the next generation of SFE instruments.
Next slide, please.
In summary, we've got excellent recovery from a model
matrix over a broad range of operating conditions and the
results, we feel, are comparable, probably realistically in the
50 to 110 percent range, compared to the soxhlet.
One of the big issues in environmental methods, we are
minimizing the use of freon. We've eliminated 98 percent of the
freon required for this method. We've reduced the use from 500
mils, roughly, to about 10. That's going to have a tremendous
advantage in our waste stream. It has a tremendous advantage in
the air quality since much of that 500 mLs is often evaporated
during workup. And certainly, automation is going to improve
things considerably.
Next slide.
I can't go away without talking about dioxins very
briefly. EPA Method 8290 or 1613 involve soxhlet extractions and
three column steps. We have worked on an SFE method which, in
this slide, eliminates all of the column steps. That's probably
unrealistic across the board. We expect that we'll probably have
to incorporate the carbon column in a routine method, but we feel
we can probably eliminate the silica gel and alumina steps.
Next slide.
If you look at the regression information from the
experimental design, you get a much different profile for TCDD.
Basically, we find optimum conditions for extracting TCDD are at
-------
476
much higher temperatures than this conventional wisdom.
Conventional wisdom has most people working up in this upper
corner here where things are a more slower recovery. It partly
explains why a lot of the work with TCDD and carbon dioxide has
been very disappointing. In fact, some of our work we reported
two years ago was disappointing because we were working in this
region. We now have more information that temperatures are much
more important... .variable. Unfortunately for C02' we still don't
get good recoveries with dioxins.
The next slide explains partly why that is. Looking at
the recoveries as a function of time, you see up to two hours
from some samples. We are still extracting samples out. In
fact, they're just beginning out. So we have in this case a
kinetic problem with extraction.
The next slide summarizes the same results for the
furans and, again, we see at two hours we're just beginning to
extract the majority of the materials. So, we think we have a
kinetic problem with C02that will probably prevent it from being
used. Fortunately, we have some other alternatives. One of
those is summarized in the next slide and that's another picture
with a COmethanol mixture. This is, again, an experimental
design summary. We get a different profile...more interesting,
but still it's telling us that high temperature and pressure are
probably the way to go in this system and we don't have
corresponding dioxin extraction for that same matrix. I believe
that kinetics will be faster here.
The next slide shows one other application we've
developed and that's using the empore disks, which are marketed
by 3M and Analytic M. They involve a Q8 matrix or C^S particles
which are embedded in a teflon membering matrix and you simply
pass your aqueous sample through the membrane and the organics
are trapped inside. We have spiked dioxins... the suite of
dioxins into water samples, extracted with the empore disk and
now we're looking at a supercritical fluid extraction of the
-------
477
empore disk, rather than a. soxhlet or convention solvent
extraction. The results are summarized in the next slide.
We can see spiking at 160 picograms per liter for the
tetra, hexa and octas. Spiking directly on the disk, we can get
pretty much quantitative recoveries at the higher levels of
chlorination. A little bit lower for the tetras, but probably
within the range that we see for some of our other methods. If
we actually do the spikes in water, we see recoveries drop a
little bit, which could be due to a number of other factors. But
basically, even at this spike level, we're getting most of it
back at the higher levels of chlorination and with a little bit
of fine tuning, we can probably improve the tetra numbers, too.
But again, we're doing the C02methanol extraction of
the empore disks. So in this case, we've eliminated the use of
organic solvents except for the dissolution of the final extract
for GC/MS analysis.
Next slide.
In summary, we've demonstrated slow extraction rates
with C02for the dioxins and furans. Our preliminary data
indicate that C02methanol is going to offer some improvement.
There is information in the literature that nitrous oxide also
may be even better. We feel for a number of samples we're able
to reduce the analyte enrichment requirements, eliminating
perhaps two or three of those open column steps. But certainly
at this point, we need to do some more optimization.
Last slide, please.
I'd like to acknowledge my colleagues at Twin City
Testing and Super X Corporation which has also provided partial
support for this work.
Thank you.
MR. KING: Questions?
-------
478
NO SLIDES AVAILABLE
FOR THIS PRESENTATION
-------
479
MR. KING: Our next speaker is
Bill Budde. He's found his way here from EMSL, Cincinnati. He's
going to talk about recent work that EMSL has been doing on
LC/Mass Spec.
-------
480
MR. BUDDE: Thank you very much.
I want to thank especially Bill Telliard. I don't know if Bill
is in the room today. Is Bill here?
MR. KING: He's upstairs putting
out a fire.
MR. BUDDE: Bill's upstairs
putting out a fire, I was told.
Anyway, I am sure all of you know Bill Telliard and his
enthusiasm for his work, his marvelous sense of humor, his
dedication to environmental protection, and his friendly style at
all times and I want to thank him for inviting me here and asking
me to say a few words about liquid chromatography-mass
spectrometry, which I refer to as an emerging technology for
environmental analysis.
Before I go on and this is off the record, so please
don't take this down. I wanted to show you a side of Bill
Telliard that's not often seen. You all see him running around
here organizing this symposium, doing his other work. You don't
often see him back at his job where he works in Washington as 61
person who covers his in box well... covers his in box well and
his out box. So, sometimes you might want to ask Bill about
that.
Before I begin, I want to say that we've had a program
of research to develop liquid chromatography-mass spectrometry
and methods for environmental analyses for a number of years.
Among other co-workers, I'd like to introduce Thomas Behymer who
has been with us almost four years now. He received his PhD from
Indiana University from Ron Kites, who was up here earlier this
morning. Also, Thomas Bellar has been working on this project.
Tom is the inventor of the purge and trap method for volatile
organics in water. Jim Eichelberger has been working with me for
about 19 years and we're still working together, believe it or
not, and Jim Ho, a chemical engineer on our staff who is also
working on this project. It gives you a little hint of what you
might need to make liquid chromatography-mass spectrometry work;
-------
481
that is, a PhD in chemical engineering.
I don't have pictures of all of the staff and all of my
co-workers here, but here's certainly Tom Behymer, and for those
of you who do not know him, Tom Bellar. Jim Ho and Jim
Eichelberger are not on this picture and the other person there
no longer with us ... regrettably.
The objective of our work is to develop a broad
spectrum liquid chromatography-mass spectrometry method for the
simultaneous identification and measurement of what I would call
non-gas chromatographable materials. In other words, we'll
define non-volatile as being not amenable to gas chromatography
toxic organic environmental pollutants. This would be the LC/MS
equivalent of capillary column GC/MS. That is an ambitious goal,
very ambitious, and we're not there yet, but perhaps we will be
in a few years.
Why do we want to do this? We want to do this because
there are a number of compounds which in some cases are marginal
GC analytes and in other cases simply are not amenable to gas
chromatography at all. I give you an example of the carbamate
pesticides. Carbamates contain a carbonyl group...nitrogen on
one side, oxygen on the other and often are a methyl carbamates
because these compounds are derived from methyl isocyanate. If
an R group was alpha naphthol the carbamate would be carbaryl.
Carbamates are sensitive to temperature and to gas
chromatographic injection system or column. They tend to lose
the elements of methyl isocyanate.
Another group of compounds which are pesticides are the
thio-ureas which contain a carbon sulfur double bond with a
nitrogen on either side for example, ethylene thio-urea, which is
not only a pesticide but a metabolite of other pesticides. Other
compounds are the ureas with a carbonyl group and nitrogen on
either side. A number of these are used commercially: diuron,
linuron and siduron would be three examples.
I'm just showing you examples of the types of
compounds. Another one which is rotonone, a natural product that
-------
482
happens to be a pesticide also. It's been used to kill fish, if
you want to get rid of all the fish in a pond. It's a very
sensitive compound and has one, two, three, four, five ether
linkages and unsaturation in a carbonyl group and in a gas
chromatographic injection port of a GC column, this has no change
whatsoever. It decomposes quite easily.
Another group of compounds are the benzidines.
Benzidines are important for a reason which I'll give you in the
next slide. Benzidine is basically a diphenyl system, two amino
groups, and often substituted in the three and three prime
positions with methyl, methoxy, chlorine and so on. Benzidines
are important because they are used in a class of azo dyes. This
happens to be one example. Color index Direct Red 2 is an azo
dye, a sulfonated azo dye. But you see there's an azo group
right here and an azo group right here and these azo groups are
susceptible to reduction under anaerobic conditions as in
anaerobic sediments, sewage, landfills and biological systems and
reduction of this azo group here to form an amino, reduction of
this azo group to form an amino here would release three, three
prime dimethylbenzidine to the environment or to the biological
system and that's a known human carcinogen. In fact, all of the
benzidines I showed on the previous slide are known human
carcinogens. So, the benzidines are of great interest because
these azo dyes are widely used in large volume commercial
chemicals and a number of them still incorporate the benzidine
groups.
I think, before we proceed, I will just give you a
little review of the status of liquid chromatography among EPA's
approved analytical methods. I'm particularly referring to the
drinking water methods, which would be the 500 series, waste
water methods, the 600 series, and the 1600 series, perhaps.
If you look at what has happened over time, beginning
in '79 up through about '89, for a 10 year period, these methods
are the number of gas chromatography methods... just the number of
methods which employ gas chromatography among those EPA approved
-------
483
methods... and these are the number of methods which employ liquid
chromatography for non-volatile compounds. As you see, there's
been a tremendous growth over the years from something like 10 in
1979 to over 33 in 1989 and that's still grown a little bit until
1991 in gas chromatography, whereas liquid chromatography methods
were rather dormant. We were stuck on two for a long
time...recently jumped up to seven. I'm predicting that for the
1990s we'll see very little growth in gas chromatography methods.
I think the compounds and the environmental interest that can be
exploited and analyzed by gas chromatography has been pretty well
accomplished. There will be continuing improvements in these
methods, to be sure, but not many new methods. I think there
will be a continuous and an accelerating growth in liquid
chromatography methods as we look at more and more of these non-
gas chromatographable compounds in environmental samples.
Let's see if I can focus here. I don't have a remote
focus, do I? Can someone touch up that focus a little bit?
This is in the handouts, so you may not want to write
it down, but I thought I would throw this up. This gives you an
idea of some of the activity we've been engaged in, particularly
journal articles published by EMSL, Cincinnati. This will be in
the proceedings because I left it with the people who are
publishing that. We had an article on thermospray in Analytical
Chemistry in 1988, a very definitive article, I think, a wide-
ranging evaluation of thermospray. Later, we switched our
attention to the particle beam and the carrier effect, particle
beam in another recently published paper, and a paper in
Environmental Science & Technology, on thermospray. So these are
the sources of more information about what I'm going to say this
morning. I simply cannot cover all of the material and all of
the information that's been developed in this short presentation.
So, if you're interested in more information, I would suggest any
of these articles.
The kind of liquid chromatography/mass spectrometry I
want to discuss today has to do with sprays and aerosols. There
-------
484
have been many different types of liquid chromatography/mass
spectrometry interfaces proposed and studied over the last 10 or
15 years. Some of these have been more successful than others.
The one that we have been concentrating on lately in our work has
to do with sprays and aerosols and this picture is courtesy of
Marvin Vestal of Vestec Corporation, who is one of the inventors
of this type of interface. In effect, what you have with a spiray
on aerosol generator is a liquid chromatographic effluent coming
out here and by some means or another, we convert this into a
spray or an aerosol which is billions and billions of tiny
droplets of the liquid phase. And, of course, the non-volatile
analytes that are swimming around in these little droplets do not
realize what has happened to them. They think they are still in
a perfectly happy liquid state. They're dissolved; they're in
solution. They don't know that they're flying through the air
and about to become a gas phase analyte. That's why this type of
interface actually is so attractive and why it works.
The ways to generate sprays and aerosols and there
are...Four I've listed here for common ones. One is the
pneumatic approach in which you use a gas...helium, for example,
and mix that gas with your liquid, emerging from the liquid
chromatograph and form a spray.
Another variation on that is the heated pneumatic in
which you not only heat the liquid, but you also apply the gas.
A different version and really the first one...this one
should have been at the top...is thermospray. Thermospray does
not use a gas. It does not use a pneumatic approach. What it
does is it heats the liquid chromatographic effluent. It heats
the mobile phase and converts some of that mobile phase into a
gas which then acts as the nebulizing agent to form the spray or
the aerosol. Recently, we've seen some work published and some
advances in using ultrasonic systems to form sprays or aerosols
for liquid chromatography/mass spectrometry.
Most of the work I'm going to talk about is going to be
concerned with the pneumatic and the heated pneumatic. I'm not
-------
485
going to say a whole lot about these other, but in general, we
can use any of these techniques to form an aerosol.
I want to take a little bit... If you could touch that
up and see if that's a little better in focus there, I would
appreciate it.
This is kind of a generalized schematic diagram of
particle beam LC/MS interface. This is the mobile phase from the
liquid chromatograph coming in under high pressure and an aerosol
nozzle. If we're using a pneumatic system, that is, if we're
using helium or a heated nebulizer to form a spray, then we need
to supply that gas. If this was a thermospray or an ultrasonic
nebulizer, we wouldn't need this supply here, but we'd need it
over here anyway. This sprays into a chamber which is usually
heated. We have our billions and billions of little droplets of
mobile phase now in a spray or aerosol form and regardless of the
type of nebulization we use here, we now have to...if we haven't
added a gas here, we have to add it here, as in the case of
thermospray or ultrasonic in order to provide a carrier effect to
push these particles in the direction we want them to go, which
is over here, which is the ion source of the mass spectrometer.
This nozzle or beam columnator, you might call it...a
nozzle, separates this desolvation chamber from this chamber
which is evacuated with a pump and a skimmer, which has a small
hole in it, a second chamber, another vacuum pump, and finally, a
chamber over here which leads directly into the ion source of a
mass spectrometer. What happens, of course, is as we move these
particles containing our non-volatile materials through this
desolvation chamber which may be heated...usually it is and not
very high...we get some evaporation of solvent. The particles as
they move through the nozzle pick up speed because they're being
pushed in this direction and we're pumping on this side. They
may reach supersonic velocities over here. Because of the
momentum of these particles, their mass and their velocity, they
tend to move straight ahead whereas solvent vapor, helium and
other gases entering this chamber tend to be pumped out. So, in
-------
486
fact, from maybe slightly less than one Torr, you get down to 10
to the minus five over here and you have a beam of particles and
these are little droplets of solvent containing our non-volatile
materials. As they pass through these systems, more and more
solvent evaporates.
Now there are many variations on this basic design
commercially available. Usually the desolvation chamber walls
are solid walls. They're metal and they're heated, as I pointed
out. In one variation, however, by one company, in particular
due to Marvin Vestal, these walls, in fact, are made of teflon or
some other membrane and there's another gas on the outside
sweeping along and sweeping away solvent and what happens is that
the vapor and the solvent molecules pass through the membrane and
are swept away and so we get some reduction in solvent in this
chamber, as well as this chamber and this chamber. So there are
several variations.
There's another device, I've heard, which uses a three
stage momentum separator. This is the momentum separator region
where you're separating the particles and the gases by their
momentum. That's the basic design. Some of you would probably
recognize that if you chopped off the momentum separator and
threw it away, drilled a hole in the solvation chamber and stuck
a mass spectrometer right here, you'd have the classic
thermospray device. That's what a thermospray originally was as
invented by Vestal. It didn't have the momentum separator on it.
Putting the momentum separator on is merely a refinement to bring
thermospray and these other methods of generating aerosols up to
a higher state of performance.
The method which I'd like to discuss briefly is what
we're calling Method 553. It's the determination of benzidines
and nitrogen-containing pesticides, the ureas, thio-ureas and
carbamates I mentioned, in water. We use both liquid-liquid
extraction and liquid-sold extraction and I might also add that
we're experimenting with the techniques similar to what was
described by the previous speaker; that is, using the liquid-
-------
487
solid extraction and then using supercritical fluid C02to extract
the cartridges or the disks in this particular methodology. We
also use reverse phase high performance liquid chromatography
particle beam mass spectrometry and we've been working with three
different systems, three different designs of the basic particle
beam interface. This happens to be rev 1 of this method which
was made available in July of 1990. We expect to have rev 2, or
maybe it's 1.2 or something like that... 1.1...I'm not
sure...available this October. As I mentioned before, Tom and
Tom and Jim have all worked diligently on this and are
responsible for most of the information that I'm presenting.
This is not a typical total ion current profile from
our LC/MS particle beam system. This is absolutely the best one
we have. There's no reason to show one of the ones that isn't
the best. But so far, this is one of the best. This is done
with a reverse phase CIS column packing by Waters, called
Novapak. It is a very low bleed. It's very important to have a
low bleeding column. We're seeing in liquid chromatography
pretty much the same thing we saw in gas chromatography. In the
beginning of gas chromatography, 20 or so years ago, when they
used certain types of detectors, non-mass spectrometry detectors,
they didn't realize the column packings bled. When someone put a
mass spectrometer on, they found the bleed and then the
manufacturers fixed that problem and got non-bleeding columns.
The same thing in liquid chromatography. There's a. lot of LC
columns out there that have been used with UV detectors and no
one knew they were bleeding until we put a mass spectrometer on
it. We found the bleed and the problem with the bleed. Now
we're seeing much advance in liquid chromatography columns and
the bleeding problem is going away.
This is a group of about 15 compounds that are included
in Method 553. As I mentioned, most of them are cleanly
separated and identified by abbreviations in the slide but, for
example, this is 3,3'-dichlorobenzidine and linuron and rotenone.
-------
488
If you remember, rotenone is the natural product with a complex
structure and five ether functions and a carbonyl and
unsaturation.
In addition to this being a total ion current profile,
this happens to be a gradient elution. In the early part of the
chromatograph where you're running about 25 percent acetonitrile,
75 percent water, and as we go through the gradient elution, we
end up with about 25 percent water and 75 percent acetonitrile.
So, it's a gradient elution and we also have present 100th molar
ammonium acetate in the solution in order to facilitate the
liquid chromatography and do some other things, which I can't
really go into at this time.
The mass spectra that you get though really do look
like electron impact spectra. I know this isn't new information.
It's been known for a few years that by whatever process that
goes on, when these little particles enter the heated electron
impact ion source of a mass spectrometer, they form...the solvent
goes away, the ammonium acetate disappears and what you see are
what appear to be classical electron ionization spectra.
In this particular case, for example, we're using
3,3' -dimethoxybenzidine, molecular ion at 244. There are no
signs of ions here due to ammonium ions of any type, M minus 15
and then a loss of CO, presumably from this, a spectrum which
looks very much like the spectrum that you would obtain from
GC/MS. In this particular case, this compound is what we would
call a marginal GC/MS analyte. It doesn't work very well, but
you can actually get it through and compare the spectra and they
pretty much look alike.
Let's see. Another example...well, here's rotenone,
the one I was talking about before. You have no hope of getting
rotenone through a gas chromatograph, but it's electron
ionization spectrum is known, using a direct insertion probe and
the spectrum from the liquid chromatography particle beam mass
spectrometer is remarkably alike, similar to the direct insertion
probe. They're even showing a little bit of a molecular
-------
489
ion...about 10 percent at 394 and a major fragment at 392. So
the spectra that we're getting from these systems, all of the
systems that I've talked about, are real El spectra.
As you might expect, we've investigated the possibility
of finding a test compound, one that we could use to evaluate the
performance of an LC mass spec system and we tried our favorite
compound, which is decafluorotriphenylphosphine and that was
simply too volatile for this system. But there is an oxide that
forms from DFTPP which is really a very nice compound. It's a
little bit more polar, liquid chromatographs very nicely, and
I'll show you the spectrum of it later. I just put the masses in
there. The molecular weight of the compound is 458 and we're
calling it DFTPPO. I'll show you the composite spectrum of that
in a few minutes.
Estimated detection limits using this methodology
varies. Now in this particular case, we're showing three
systems: System A, System B and System C. They're all
commercial products and if anybody wants to know what they are,
I'll be glad to tell them. Just see me after the talk. But the
remarkable thing is that we've seen very similar performance with
all three of the commercial particle beam systems. You can see
ethylene thio urea, approximately five nanograms. This is amount
injected in nanograms.
These are the estimated detection limits, based on a
3:1 signal to noise ratio for the quantitation ion. For many of
these compounds, if you drew a line across here, you can see many
of these compounds are well below that. Many of them are well
below even 10 nanograms. In particular, the benzidines, although
there are some differences in interfaces. Look at benzidine
right here. System A and System B around five nanograms... System
C up around 30 nanograms. That persists to this day and we don't
quite understand this.
This is methylbenzidine. In that particular case,
System A and System C did better than System B. I think this is
methoxy. I'm sorry. This is 3,3'-methoxy. This is 3,3'-
-------
490
dimethyl.
Here's the 3,3'-dichlorobenzidine. You see there's 30
nanograms here or so...20 nanograms here. But for System C, 100
nanograms. So there are some remarkable differences. Rotenone
is an interesting one. You see, rotenone for System A requires
almost 70 nanograms, System B, 40 nanograms. System C is a
rotenone machine. It will do five nanograms. So, there's some
differences here which we don't understand and this data is
pretty reproducible. This doesn't represent just one experiment,
but any number of experiments that are summarized on this graph.
For the reason that...a number of reasons...we
conducted a small multi-laboratory study. It was actually
conducted by our Quality Assurance Research Division and we
concentrated only on benzidine, 3,3' dimethoxy, dimethyl and
dichloro. We took the four benzidines and included them in the
study and we did not include the extraction step in that
particular multi-lab study. Again, I want to acknowledge our
Quality Assurance Research Division in EMSL, Cincinnati, which
orchestrated this study.
If you could focus that up, I would appreciate it.
The four benzidines were included and it was only the
determinative step of the...that is, the liquid chromatography
and the mass spectrometer. We didn't include extractions in this
preliminary multi-lab study. However, we did include 13
laboratories. I think it should add up to 13. I didn't check
it. When I made out the slide, there were four Hewlett-Packard
systems, which were pure Hewlett-Packard particle beam LC/MS
systems. There was one system using a Vestec interface and HP
otherwise. There were two systems using an HP...I'm sorry, a
Vestec interface and a hybrid Vestec HP spectrometer with a
Techninet data system. There was one pure Extrel. There was
sort of an Extrel interface with an Extrel spectrometer and
somehow an Incos data system. I don't know who put that
together. Some of you may know these labs by the strangeness of
the combinations here; I don't. There's an Extrel interface with
-------
491
a Finnigan TSQ spectrometer and an Incos data system and a VG
with a Finnigan-MAT. That's a magnetic machine with an Incos
and, finally, a General Electric interface. That's an ultrasonic
with a Joel mass, a magnetic machine. So, there are some 13
laboratories that participated in this little study of the
determinative step of Method 553 with only the four benzidines.
We wanted to see how this would work on a multi-lab basis, and
the answer is on the next slide.
This is the easy slide. This is multi-laboratory
accuracy. I will say of the 13 laboratories, two laboratories'
data had to be discarded because they were clear statistical
outliers and it wouldn't be fair to average those numbers in with
the other 11 because they were just screaming statistical
outliers. Even I could accept that. But, if you take the
remaining 11 sets of data and average them out...this is
benzidine...this is methoxybenzidine...this is the
dimethylbenzidine and this is the dichlorobenzidene...there were
two different concentrations. The double Crosshatch is 10
micrograms per liter and the other one is 100 micrograms per
liter equivalent and the mean and multi-lab accuracy in measuring
those four compounds is across here and you can see it's right
around 97 percent or so. Those are mean accuracies again. That
doesn't say anything about precision at this point. So, the
accuracy we're very pleased with in terms of measuring those four
compounds.
The precision is in this slide. Again, there are two
bars because there's the 10 microgram per liter and the 100
microgram per liter equivalent. These are RSD numbers, relative
standard deviations, going from zero to 22.
This is the multi-laboratory precision of the 11
facilities. You see if you draw a line across here, you'd see
that, in fact, all of them at 100 micrograms per liter are under
10 percent. When you go to 10 micrograms per liter, even one of
those...benzidine itself is under 10 percent. One of them gets
up as high as 20 percent and the other ones are in this range.
-------
492
Now, the statisticians that we have on our staff by
some wondrous process that I don't understand are able to convert
multi-laboratory precision data into single analyst precision
data. That is, of course, an estimate of what a single analyst.
would get, based on the multi-lab data and you can see, again,
the RSDs for these, if you draw the line across here, are all
under 10 percent for a single analyst except this one measurement
of...that would be 3,3'-dimethoxybenzidene at 10 micrograms per
liter.
So we were very pleased with these results and, again,
I think you'll be hearing more about this study as time goes on.
I also wanted to tell you a little bit about
decaflurotriphenylphosphineoxide. This is not a single spectrum
measured, but this is the composite spectrum measured by all 13
laboratories in the study. What we did is we simply averaged the
intensities together of the ions that you see and this is the
composite, a molecular ion at 458, about 25 percent...M
minus...let's see that would be a loss of a fluorine, I
think...38...58...no, it's HF...loss of HF. It must be knocking
off one of these two HFs or M minus 20 at 438. Base peak
generally at 271 and a number of ions and you can see why this
compound is so attractive as a performance test compound. It has
a range of ions of significant intensity, ranging all the way
from mass 77, even mass 69, all the way through mass 458. We've
looked at some ranges of reasonable performance and we've also
found two correlations where if certain conditions are present,
which are not desirable on the LC/MS, this ion will drop down
like mad and this one will go up like mad. We also see other
conditions here these ions will disappear and will get weighted
down here. So we look like we have a good performance test
compound and it's diagnostic for problem conditions in the
system.
Conclusions: I would like to make, and I didn't show
you any slides on this and you'll really have to consult some of
the papers and the literature because we just don't have time to
-------
493
go into all of these things, but there is a strong tendency
towards non-linear calibration in this particular type of system.
We used second order regressions for quantitative analysis.
There are some alternatives. For example, isotope dilution
analysis will straighten this out and make it linear.
There is a co-elution effect and that's the explanation
for this. I won't go into this at this time. It's in the
literature. We don't think this is necessarily true anymore.
External calibration is certainly one option and that's what was
used in the multi-lab study. But it looks like if the
isotopically-labeled compounds are available and they were in a
few cases that we've studied, you can use those very nicely and
get a very linear calibration.
The performance of the three systems are similar for
most compounds, but I did mention the fact that rotenone and
there are some other exceptions where one or more other type of
instrument may give you significantly improved performance.
The precision and detection limits have been improved
significantly over the last three years or so by incorporation of
method controls. Now, you've seen those method controls if you
looked at any of the 500 or 600 series methods and in particular,
one that I mentioned happened to be column bleed. Column bleed
can have a significant effect on precision and detection limits.
Detection limits and short term precision are
acceptable for environmental analysis and we think that long term
stability will be improved and is being improved in the current
commercial instruments.
So this is the end of my presentation. I have one more
slide which is off the record again. I don't know if Bill
Telliard has come into the room, but I want to congratulate Bill
for this wonderful conference because he is not doing what is
depicted on this slide, which says: Let's leave the public in
the dark. It works for government; it should work for business.
Bill doesn't do that and he has arranged a marvelous meeting here
which I know is in its 17th year in which we all have the
-------
494
opportunity to share some of the things we're doing with all of
you folks in the public.
Thank you very much. I'll be glad to address any
questions.
-------
495
QUESTION AND ANSWER SESSION
MR. BICKING: Bill, the high res
dioxin methods use a lock mass approach for monitoring
performance during the run. Do you anticipate or have you looked
into possible applications for lock mass with LC/MS?
MR. BUDDE: Lock mass is made
capable by the fact that you do have a high resolution mass
spectrometer.
MR. BICKING: Right.
MR. BUDDE: That's what makes it
possible.
MR. BICKING: Right.
MR. BUDDE: We've been doing all
of this work with a nominal mass resolution mass spectrometer and
so we don't have that capability. But if we ever find another
dioxin and it's not volatile, we can put an LC on a high
resolution mass spectrometer and certainly that would be a viable
approach.
DR. HAEBERER: Fred Haeberer, QAMS,
EPA.
Bill, you very cavalierly addressed the ammonium
acetate liquid phase modifier and said, it went away...it goes
away. Yet in your spectrum of DFTPPO, I noticed a strong 77 MO
ratio. Was that run with the acetate modifier in there? Is that
what we're looking at? What's going on?
MR. BUDDE: Well, in that
particular case, no. When I said ammonium acetate goes away, we
don't see any trace...any ions in the spectra that are
attributable to adduct ions from ammonium acetate. The 77 in
that case is due to the unfluorinated phenyl ion. On DFTPP,
there is one ring which does not contain fluorine and probably
forms an ion and that's probably the reason for that ion.
Ammonium acetate is very important in these analyses
and I didn't take the time to go into it.
-------
496
DR. HAEBERER: I understand that
completely.
MR. BUDDE: Yes, ammonium acetate
facilitates chromatography, but more important, it does give us
some enhanced signals for some of these compounds and we just
don't have time to go into that. But, it's...
DR. HAEBERER: But you're not
seeing it at all?
MR. BUDDE: No, there's no
evidence that ammonium acetate participates in the mass spectrci.
However, it would be there, of course, in lower masses and we
don't scan those masses. We don't scan much below 77.
DR. HAEBERER: Very good. Thank
you.
MR. BUDDE: Thank you again for
your attention. Let's go to lunch, I guess.
MR. KING: Thank you very much.
Let's break for lunch and get back here promptly at 1:15. The
statisticians are going to do their magic and I'm sure you don't
want to miss any of that.
(WHEREUPON, a brief recess was taken for lunch.)
-------
497
OBJECTIVE
TO DEVELOP A BROAD SPECTRUM LIQUID CHROMATOGRAPHY -
MASS SPECTROMETRY (LC/MS) METHOD FOR THE SIMULTANEOUS
IDENTIFICATION AND MEASUREMENT OF NON-GAS
CHROMATOGRAPHABLE (NON-VOLATILE) TOXIC ORGANIC
ENVIRONMENTAL POLLUTANTS.
(THE LC/MS EQUIVALENT OF THE CAPILLARY COLUMN GC/MS
METHODS)
-------
498
CO-WORKERS AT EPA
0 THOMAS D. BEHYMER, PH.D
0 THOMAS A. BELLAR
0 JAMES W. EICHELBERGER
0 JAMES S. HO, PH.D.
-------
499
m
- -
V
CD
oc
<
o
LU
z
o
u_
CO
m
DC
<
o
o
O
I
l-
ai
LU
I-
<
m
oc
X
111
CO
tr
<
o
o
I
H-
UJ
-------
-------
501
UJ
o
o
QC
-------
502
X = H, CH3, OCH3, Cl
Benzidines
-------
503
(VJ
O
UJ
an
o
LU
-------
504
CD
TD
O
0)
TO S
o i
CfU
D>
< J
0)
>
O
Q.
Q.
U)
CN
CD
(O
1 I I I I I
CJ
o>
CO
2 o
v\
O
N
O)
N-
CD
I I I I I
O CO (O T CM O
paAOjddv
-------
505
METHOD 553
DETERMINATION OF BENZIDINES AND NITROGEN-CONTAINING
PESTICIDES IN WATER BY LIQUID-LIQUID EXTRACTION OR
LIQUID-SOLID EXTRACTION AND REVERSE PHASE HIGH
PERFORMANCE LIQUID CHROMATOGRAPHY/PARTICLE
BEAM/MASS SPECTROMETRY
REVISION 1.0
JULY, 1990
THOMAS A. BELLAR
THOMAS D. BEHYMER
JAMES S. HO
WILLIAM L. BUDDE
ENVIRONMENTAL MONITORING SYSTEMS LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
CINCINNATI, OHIO 45268
-------
506
EMSL-CINCINNATI LC/MS JOURNAL ARTICLES
0 THERMOSPRAY: ANAL. CHEM.. 1988, fiQ, 2076-2088.
0 PARTICLE BEAM CARRIER EFFECT: JL AM. SOC. MASS
SPECTROM. 1990, 1, 92-98.
0 PARTICLE BEAM: ANAL. CHEM.. 1990, £2,
1686-1690.
0 PARTICLE BEAM: i. AM. WATER WORKS ASSOC.
1990, 82, 60-65.
0 THERMOSPRAY: ENVIRON. SCI. TECHNOL. 1990, 24,
1748-1751.
-------
507
SPRAY/AEROSOL GENERATORS USED WITH
PARTICLE BEAM LC/MS
PNEUMATIC
HEATED PNEUMATIC
THERMOSPRAY
ULTRASONIC
-------
508
8
C 3
O O O
>i
i-H
Q'
a
N
42 ro
0) O
O
Vj
0
Q)
CJ
rc
QJ
4->
C
I-H
I
ro
ro
a.
rc
>M
O
S-l
tn
ro
•^i
D
U
•H
4->
I
G)
!j
P
I,"'
I
^
ft,
-------
509
o
o
o
°r
o
01
o
_o
T
m
o
nj
©
_o
_o
o
ui
ec
t-
z
o
t-
UJ
o
en
Hs>
CD
O
HS>
3
-J
O
o
m
O
a.
CD
ho
in
Hs>
HS>
CD
Ui
H
<
Z
o
ai
o
o
z
-------
510
cxj
cu
u
z
D
N
z
UJ
m
>-
X
O
I
i-
LU
2
Q
J
CO
co"
CD
CXJ-
nj
CD
CD-
GD
in-
en
~
rtu
ECU
rttJ
U
= -OD
cn - •
01
OJ-
CO
00
•CD
.r®
ItXJ
= r€)
::©
-C£>
-------
511
16
F F O F F
31
F F
DFTPPO
MW 458
H
-------
512
o
O o
LJ 5
O o
Q 2
LJ _j
I— z
< *i
(/)
LJ
t>OCvNXsN>X^OvXVO;^O^N>^^
Y//////S
k\\\\\\\
Y/////7/.
CL
QQ
§
QQ
m
M
m
o o
OO
O
OO
O
o
o
^
LJ
(—
CO
!=! CD
O 00
I ^
O pn
0 x^
l/ZI
00
00
(6u) NlAiniOO NO a3103PNI
-------
513
MULTI-LABORATORY LC/PARTICLE BEAM/MS STUDY
DIVERSITY OF EQUIPMENT
No. INTERFACE SPECTROMETER DATA SYSTEM
4 HP HP HP
1 VESTEC HP HP
2 VESTEC VESTEC/HP TECHNIVENT
1 EXTREL EXTREL EXTREL
1 EXTREL EXTREL INCOS
1 EXTREL FINNIGAN-TSO INCOS
1 VG FINNIGAN-MAT INCOS
1 GE JOEL JOEL
-------
514
u
CD
Q
D
O
U
CO
O
<
cr
o
oo
CD
D
Nl
m
o
o
o
O)
o
CO
o
ID
o
m
-------
515
22
20
18
16
14
12
10
8
6
4
2
0
SINGLE ANALYST PRECISION
BZ
MB
LB
COMPOUND ID
10 ug/L RSSSS1 100 ug/L
DB
MULTI- LABORATORY PRECISION
22
20
18
16
14
12
10
8
6
4
2
0
BZ
MB
LB
DB
COMPOUND ID
10 ug/L RSSS3 100 ug/L
-------
516
•0
n—
u>
to-.
M
10
irt
N
N
<0
O—
O
O
en
C
CO
CJ
00
o
-iO
—o
CD
—o
O
-«0
-------
517
CO
o
CO
O
O
cc
DQ
O
CC
LU
Z
CO Z
QO
CO
m
cc
>- UJ
ooc
IE
ZQ
Ho
o
V-J Q
£UJ
H CO
CO
m
LU
Q r-
H
CO
o
CL
o
LU
LL
LL
LJJ
Z
O
H
Z>
LU
O
O
D
LU
CC
O
LU
CC
DQ
DQ
O
CC
Q.
O
CC
DQ
_J
<
O
-l
<
Z
cc
LU
K
X
LU
CO
O
CC
O
LL
CC
CO
LU
CC
CO
2
111
CO
CO
CO
LL
O
LU
$2
21
oc ^
LU O
0.0
LU
LU
DQ
LU
X
CO CO
cc
-J H-
LU O
gf
Q LU
13 b=f
SION
OVED
111
CC
Z
O
CO CO
O CO
LU >-
Z
xz
CO O
-J Hi
|5
HH
00-
O £ LJJ LU
UJ^ HO
DC ^ LU O
CL ^ Q <
I-
Z
LU
2
LU
O
CC
CL
CO
Q
LU
LU
Z
DQ
£
CO
CC
LU
I-
i
O
o
-------
518
MR. TELLIARD: Our first speaker
today is Dr. Hau. He is Assistant Professor at the Center for
Quality and Productivity Improvement at the School of Business at
the University of Wisconsin in Madison. He's going to talk about
an interesting paper that I saw about a year and a half or so ago
on the use of statistics as it relates to...particularly has a
lot of implications in the NPSD program down the road and I think
you will find it very interesting.
-------
519
DR. HAU: Well, I'm going to talk
about a paper called, Judging Compliance and the Limit of
Detection. This is a joint paper with Professor Mac Berthouex.
He is a professor of Environmental Engineering at U of Wisconsin-
Madison.
For the last two days I feel I am in a foreign country.
It seemed like I'm in Japan. I'm Chinese anyway, so Japan is as
foreign to you as to me. I heard names such as GC and MS and CFC
and all of those terms and I don't know what they were. Well,
just to feel closer to you, we changed our name actually. We
changed our name to M-A-C and I-A-N. From now on, you can call
me I-A-N and you can call Mac, M-A-C. It sounds more chemical.
Well, the problem we're going to talk about is about a
detection limit. Recently, we have some regulation and the water
quality limit is required even below the detection limit. As a
consequence in that situation we will see a lot of cases where
most of the measurements are reported as below detection limits.
It looks like that if our effluent quality is below the detection
limit, we can expect a lot of measurements below detection
limits, i.e. only a very small portion of the numbers are
reported. If we take a lot of samples, we see this kind of
situation: we get a distribution on the effluent quality and
only a very small part above detection limits. In that case, we
really cannot talk about the average or the standard deviation.
The only meaningful measure or statistic is the proportion of
measurements above the detection limit.
Recently, we saw a regulation similar to this. I call
this Regulation I that I want you to remember because we are
going to refer to it a lot. "An effluent will be judged out of
compliance if there is a measurement above detection limits." It
looks like a very reasonable regulation because if the limit is
below the detection limit, we should expect that not many of our
measurements above detection limits. So this is a reasonable
regulation...it seems like. However, the fact is that all
effluents will be judged out of compliance under this regulation
-------
520
no matter what!
We are to understand why this is so. I think most of
you know better than me what is detection limit, but let's go
through it quickly. Detection limit mostly is set to three
standard deviation of the blank measurements. The idea is that
we do not want to report any small number which we may get from
measuring blank. How do we do that? We set the detection limit
so that there is a very small chance for a blank measurement to
exceed that limit. How small? One percent chance to see a
measurement from blank falling above the detection limit. So if
you actually see a number that is above the detection limit, then
we are pretty sure that it is not coming from a blank.
So this is a graphic representation of the statement
and if we measure the blanks, we get a distribution and we set
the detection limit 3 standard deviations away from zero. The
idea is make the tail area to be one percent which is very small
chance. If we see a number out here, we are pretty confident
that it is not coming from a blank.
To see why Regulation I has a problem, we use this
example. Suppose we make 100 measurements on blank and by
definition, each measurement has only one percent chance greater
than detection limit. However, the chance of getting no detect
out of 100 measurements is .99 to power 100 and that number
actually decreases very fast. That number is in fact equal to
.37. Although there is very little chance to get one detect out
of one measurement, there's a very small chance to get no detect
out of 100, or, in other words, the chance of getting at least
one detect out of 100 measurements is very high, namely, 63
percent. Okay. And what we mean by this 63 percent was in
computation of it. Well, remember this is the blank so this
number means that under Regulation I, if you take 100
measurements of blank, there is a 63 percent chance that we would
be declare blank out of compliance. And as you can see, if we:
increase the number of observations, there is a chance you'll get
even higher chance and finally you'll get up to 100 percent.
-------
521
An analogy: suppose you want to judge if a coin is fair
or if a coin is balanced. What do we do? We toss the coin seven
times. If we see all heads then we declare it fair. What's the
rationale behind it? Well, we think that if the coin is fair,
there is a very small chance to see seven heads in seven tosses
and that's very reasonable. The problem is that we do this every
week for the same coin, and this is the regulation similar to
Regulation I. We perform this experiment to the same coin every
week and whenever we see seven coins out of seven tosses, we
declare them as fair and, as you expect, sooner or later we're
going to see seven heads out of seven tosses just by chance and
this is exactly the problem of the Regulation I. Although there
is very little chance to see a detect in one measurement, but in
the long run the chance is getting very high and eventually gets
up to 100 percent.
So in the following, we are going to propose a
regulation, a form of regulation that takes into account the
variability of the measurements.
To review the situation, suppose we have effluent
quality limit below the detection limit. Then, as we said
before, the number that measures the quality of the effluent is
the proportion of measurements expected to fall above the
detection limit. This is actually a measure of the effluent
quality. Graphically, it looks like that. This is the effluent
quality and since this is below detection limits, most of the
measurements we cannot or we won't report. So we use this shaded
area, which is the probability that we just talked about. The
area here is the proportion of the measurement above detection
limits. For the blank, the P is one percent and if the effluent
quality is not as good as blank, but below detection limits like
this, the P is five percent as in the second figure. If we move
the distribution to the right and the P will get higher. So the
P is indication of the quality of the effluent. So we can write
the permit in terms of P. For example, suppose we say that, when
P is greater than 10 percent, then we'll declare noncompliance.
-------
522
That is, if we see 10 percent of measurement of detection limits,
then we'll declare noncompliance. Otherwise, when P is less or
equal to . 1, we'll declare compliance. This is the number to be
determined by chemists as a lot of people here.
But however, P here is unknown. We cannot observe the
P because we don't know the distribution and so this needs to be
estimated by the sample proportion in a particular sample, or
even more convenient, by the number of detects in a sample of
observations.
So although the permit was written in terms of P, we
need an operational rule that we can actually carry out which is
based on the sample value. What is the sample value? We said
that the number of detects is the statistic we want to use. What
we need to do is to determine the critical value, CVu. If the
number of defects is greater than CVu, then declare compliance.
We want to determine critical value so that if the
number of detects is greater than this number, then we will
declare noncompliance. For example, let's take the example where
P is equal to . 1. That means in the permit we just mentioned, it
is in compliance. And how should we determine the upper critical
values, CVu? Let's suppose we take a sample of 20 observations.
Which number we should take as the upper critical value? Well,
in that case we have to calculate the chance of seeing a
different outcome in 20 observations. If P is really .1, we
should expect to see two detects, (20 times .1.) However, it's
quite likely that we can see three defects in a particular set of
samples or it's also likely to see one detect. However, it's
quite unlikely to see five or above. And actually, the chance of
seeing five or above is less than five percent and that is the
critical value we should set. That way the chance of seeing the
number of detects above that critical value is very small and
this chance we call the discharger's risk. This is the wrong
decision we are making. And we want to minimize that risk and in
this example we want to control that risk to be less than five
percent. Then, according to this probability, we should set the
-------
523
critical value to be five. That means the regulation is that if
we see five or more detects, we will declare noncompliance. And
the idea is to minimize the risk of making the wrong decision
that we declare noncompliance when it is, in fact, in compliance.
For example, this is the probability of different
outcome and P equal to . 1 when we have 20 observations. As you
can see, we have a 12 percent chance to see zero detects and so
forth and, as you can see, the highest we have is 28
percent...which is the highest to see two detects and although
it's not that weird to see one detect or three detects. But, as
you can see here, we have only a three percent chance to see five
defects and a very small chance to see six and so the chance of
seeing five or above is less than five percent and this is the
upper critical value we are choosing. The rationale is that we
chose the critical value such that it discharges risk which is
the probability of declaring noncompliance, when, in fact, in
compliance.
In this example is the probability of the number of
detects greater than the critical value when P is equal to .1,
which is in compliance. We want to control the critical value
such that the discharge's risk is small enough. How small is
enough is a difficult question. It depends on our risk analysis.
When can we declare in compliance? Well, similarly, we
define another critical value, the lower critical value, such
that the regulator's risk is small. What is the regulator's
risk? Well, this the probability of declare non compliance when
in fact it is in compliance. What is the risk in this example?
That is the probability of seeing the number of detects below the
lower critical value when P=0.1. We want to control the critical
value so that this risk is small enough.
If you see the graph again, this side, up about
five...the area is small...and this side as low as we can get, at
zero. When you calculate, the number is 12 percent. We are
getting as low as we can get. You can make a regulation that we
have to see no detects. The regulators still have ... of 12
-------
524
percent making a wrong decision. But that's the lowest we can
get so we have to set the lower critical value in that value,
zero.
Well, to summarize, we can set the critical values
according to what the value of P we require. So for each P, we
can calculate the chance of seeing different outcomes. Take P
equals .15 as an example. When P is equal to .15, that means we
want the permit limit to require P to be less than .15 in order
to be in compliance and above .15 to be noncompliance. When P is
equal to .15 out of 20 observations, it's very little chance that
we see zero detects and it's very little chance to see above six
detects. So the upper critical value we should set is when we
declare noncompliance, we receive six or more detects and we
declare in compliance when we receive zero detects.
What should we decide if we see one, two, three, four
or five detects i.e., between the two critical values. We can
not declare noncompliance and we cannot in compliance. Well,
when the number of detects is less than the upper critical value
and above the lower critical value, then there's not enough
chance (enough evidence) to declare either compliance or
noncompliance because making either decision we will have high
either regulator's risk or the discharger's risk. In that case;,
we should issue a warning. What does that mean? Well, that
needs to be discussed in detail. The idea is to get more
evidence by getting additional N samples. Okay. And if one
detect is seen in that additional N sample, then we will declare
noncompliance. Otherwise, we would declare in compliance. But
what is N? That's the question.
Well, this is a table of what N should be. Of course,
the N depends on the P that requires a proportion of value above
detection limit. This is the P and, in most cases, we want the
rule in the form: if we see one value above detection limits, we
will declare out of compliance. In how many additional samples?
Well, it depends on the P. When P is equal to .01, then we
require six more samples. Let's take this as an example. When P
-------
525
equals .05, then if we issue a warning, what it means is that we
require to take five more samples. If we see receive one detect
out of the five, then we'll declare noncompliance. How did we
get these numbers? Well, these numbers come from the fact that
we want to control the discharger's risk. This number is set so
that the discharger's risk is small. So, you can calculate the
chance of getting a detect out of. N additional samples for
different P. If that is small, then that's the number we want.
To summarize what the proposed regulation is, well,
first, we have to determine the P value which is the expected
proportion of detects you want to control and the limits should
be in this form: when P is greater than (then we will declare in
noncompliance and when P is less than equal to) then we declare
compliance.
But in order to carry out this rule, we need an
operation rule that depends on observation. If the number of
detects is greater than the upper critical value, we would
declare noncompliance and if the number of detects is less than
the lower critical value, we declare in compliance and if it's
in-between, then we issue a warning and additional N observations
are required, and if we see at least one detect in the additional
observations, then declare noncompliance. Those values, the
upper critical value is determined so that the discharger's risk
is small so that we don't wrongly declare it in noncompliance.
The lower critical value is determined so that the regulator's
risk is small so that we don't wrongly declare in compliance.
When it's fuzzy, i.e. when we don't have enough evidence, we
require more evidence and the additional number of sample N is
determined so that the discharger's risk is small because we are
going to make a decision based on N to declare noncompliance.
Of course, there are some issues that we haven't
discussed such as how to determine P. Some other issues that we
haven't discussed that is in the paper that Mac and I wrote. If
you are interested, we can talk about it, afterwards.
So to summarize what I've done in this talk is, first,
-------
526
to introduce the risk associated by regulations. Okay, that is
the discharger's risk and regulator's risk and if we're not
careful, by regulation I that looks like a very reasonable
regulation. But is has actually not much meaning because we know
that sooner or later we will be declare every discharger
noncompliance according to that regulation. Also, we propose a
form of regulation so that we can minimize or control the
discharger's risk and also the regulator's risk.
-------
527
QUESTION AND ANSWER SESSION
MR. HAU: Should I take questions
up there?
MR. TELLIARD: Yes, come on up
here. Questions?
MR. PRONGER: This is Greg
Pronger, National Environmental Testing.
I'm not really certain who this question is for, but
how do you differentiate between a natural variance in a waste
stream discharge where you're going to be measuring which
basically is steady state system and looking at the statistics
around that or true variances in the concentration where you
would have a positive hit because there's been a change in the
outfall? As a regulating community, how do you differentiate
between those two occurrences looking at an argument like this
for the statistics? I can see where the Regulation I was
possibly trying to approach the problem in the variances in the
outfall itself where this is addressing the issue if it's a
steady state outfall.
MR. HAU: This is an excellent
question. I am not a regulator, so I'm not able to answer that.
DR. KAHN: That's a tough
question. Basically, you'd have to make... I mean, the data
that you would use to establish the compliance standard should be
consistent with the rules that you use to measure compliance. I
mean, that's the way that we try to do it. It doesn't always
work out that way.
MR. WHITE: There's also a couple
of different kinds of waste water regulations. There are waste
water regulations that are based on water quality and there are
waste water regulations that are based on technology. With the
technology regulations, you would be assuming that there's a
waste water treatment technology that can reduce pollutants to a
certain level and you would not be allowing peaks. You would say
that is the level at which you're going to regulate and even if
-------
528
you're working on a batch process and every once in awhile you've
got a heavy load of this chemical coming in, we say you've got a
technology that can keep the pollutant concentrations down to
this level. With the water quality standard, I don't work with
those as much and I can't...
MR. TELLIARD: Did you get an
answer?
MR. PRONGER: Thank you, I think
so.
MR. HAU: I think this issue is
about determining all sources of variations. It's really
important and I think that we need to look more.
MR. HAEBERER: Fred Haeberer.
While I understood about 20 percent of what you said and am in
complete agreement with you, my observation is that perhaps a
more efficient way of bringing reason to what we're doing would
be to introduce more and more rationale risk assessment process
and to develop methods that are more sensitive. We're going to
have a very, very, very difficult time introducing those concepts
into the regulatory process.
MR. HAU: Well, maybe I should
clarify the difference here. The risk analysis, I think, should
be done here. When we determine what is the expected proportion
of detects we want, that is the risk analysis here. Once this is
determined, what we are talking about is how to carry out this
rule. So, I think it's two different issues. You can be
stringent here, but because of the random errors, including the
sampling error or the measurement error, there is some
uncertainty involved that when we carry out this rule, if you're
not careful, we are making a lot of mistakes on this
regulation...on this part. But it's two different issues here.
I think most of us are concerned with P, which is very important.
If this was set wrong, it doesn't matter what you do afterwards,
you won't correct the situation. But if this is set right here
and still there is the random error that we should be careful how
-------
529
to carry out this part.
MR. TELLIARD: Thank you. Any
other questions? I can't see over there. Anyone over there?
One more.
MR. PRONGER: How do you correlate
this information with the different working ranges of a piece of
measuring equipment where you have traditionally a detection
limit...a range up to your limit of quantitation...the range. I
think part of what I'm seeing as confusion is I'm trying to
relate this to being an analyst. There's this grey area whenever
you're running an instrument that's from your MDL to your LOD.
It seems to me that this has some application in that range.
Once you get above a limit of quantitation, you have a pretty
sound measurement and it would seem unlikely that the statistics
are on that bell curve. If you extrapolated that thing out
infinitely, you could have an accidental discharge of a pH of
two. That would be statistically possible, but hugely unlikely.
There seems to be a factor in here that's being missed that takes
into account that at some point, the probability just goes to
zero, rather than looking at just possible hits. There also has
to be a relation to the distance from the mean.
Am I making myself at all clear to anybody?
DR. KAHN: Let me try to respond
to that, sort of partly in defense of Dr. Hau, who I think has
provided us with a very interesting framework to look at what is
really a general class of problems. What he's talking about, I
think, can be extended to other situations. That is, where
you're not just looking at detection limit compliance. In other
words, you can see a P value that would correspond in compliance
at any level. So, I think that would...some more work, some more
generalization of what he's done, I think, could fit the
situation that you're describing, although we could argue about
that grey area between the so-called detection limit and the so-
called LOQ.
MR. PRONGER: Whatever you wanted
-------
530
to. . .
DR. KAHN: Whatever you want to
call it.
MR. PRONGER: Just so you...
DR. KAHN: The PQL, the XYZ,
whatever it is.
MR. HAU: I think that's right. I
think the method is general. But, of course, when we talk about
LOQ, then the probability has to be adjusted accordingly. I
think the framework is general.
MR. TELLIARD: Thank you.
-------
531
Regulation Design Related to
Detection Limit
by
Dr. Ian Hau
-------
Problem:
532
Water-quality limit is below the
"Method Limit of Detection" (MDL)
Consequence:
Most measurements are not
reported.
pt reported
Effluent Detection
Quality Limit
Limit
-------
Regulation I
533
The effluent will be judged out of
compliance if there is a
measurement (the concentration of
the restricted substance) above
MDL.
Fact
Under Regulation I, all effluents
(even "blank") will be judged out of
compliance.
-------
What is MDL:
534
Let s be the standard deviation
of the repeated measurements on
"blanks", where
MDL = 3s
Idea:
Do not want to report small numbers
which we may get from measuring
"blanks11.
Set MDL so that there is only 1%
chance to see a measurement from
"blanks" falling above MDL.
Distribution of
blank measurements
\- 3s
-------
535
Example:
~ 100 measurements on "blanks"
~ Each measurement has 1% chance
greater than MDL.
~ Chance of getting no detects out
of 100:
100
0.99 = 0.37
~ Chance of getting at least 1
detect out of 100:
1 - 0.37 = 0.63
i.e. The chance of wrongly declared
an effluent (blanks) out of
compliance when 100 measure-
ments are taken is 63%.
-------
Analogy:
536
Judge if a coin is "fair" by:
Tossing a coin 7 times, if all
'heads', then declare "unfair".
Perform this test for the
same coin once every week; when-
ever we get all 'heads', declare
"unfair".
When the EQL < MDL, the
parameter
P = proportion of measurements
expected to fall above MDL
is an indicator of an effluent's
quality.The smaller the P, the better
quality the effluent.
-------
537
Measurement distribution.
for blanks
-3 -2-1
-3 -2 -1
-3 -2 -1
-------
Permit limit:
538
If P > 0.1,
then declare noncompliance,
If P < 0.1,
then declare compliance.
But P is unknown, it needs to be
estimated by sample proportion OR
the number of detects in a sample of
n observations.
Operational Rule
Determine the critical values CVu :
If the number of detects > CVu,
then declare noncompliance.
Suppose P = 0.1, i.e. in compliance,
calculate the probability of seeing
different outcomes ( # of detects )
out of 20 observations.
-------
n=20
539
s
0 0.12
1 0.27
2 0.28
3 0.19
4 0.09
5 0.03
6 0.01
7 <0.01
Choose CVu such that the
discharger's risk :
Probability (declare noncompliance when in fact in
compliance) = Probability ( S>CVu | P=0.1 )
is small enough (say 0.05).
Similarly:
Choose CVL such that the
Regulator's risk :
Probability (declare noncompliance when in fact in
noncompliance) = Probability ( S
-------
540
expected # of detects
compliance
\
noncompliance
01234567
-------
541
What happen if ...
CVL < # of detects < CVu
... then there is not enough evidence
to declare either compliance or non-
compliance.
Warning is issued. Additional N
samples are required.
If a detect is seen, then declare non-
compliance. Otherwise, declare
compliance.
-------
542
Summary of Proposed Regulation:
Determine the desired po s.t.:
If P > po, then in noncompliance;
If P < po, then compliance.
Operational Rule:
If # of detects > CVu,
declare noncompliance;
If # of detects < CVL,
declare in compliance
If CVL < # of detects < CVu,
issue warning. Additional
N observations are required.
If see at least 1 detect,
declare noncompliance.
Determine CVu so that Discharger's
Risk is small.
Determine CVL so that Regulator's
Risk is small.
Determine N so that Discharger's
Risk is small.
-------
543
C
E
-------
544
Issues need to considered further:
- defination of MDL
- definition of "blanks"
- determine p0
- serial correlation
-------
545
MR. TELLIARD: Our next speaker is
Henry Kahn. Dr. Kahn is a chief of the statistical section in
the Engineering and Analysis Division in the Office of Water.
DR. KAHN: It just seems that way,
Bill.
MR. TELLIARD: Henry is going to
talk on some work that he did looking at analytical variability
in a study that was carried out regarding the discharge of
dioxins and furans from the pulp and paper industry.
-------
546
Assessment of Analytical Variability of Dioxin
and Furan Measurements in Environmental
Samples From Bleached Pulp Mills
Henry D. Kahn and Maria D. Smith
Office of Water, EPA
Kirk Cameron
Science Applications International Corporation
Presented at the 14th Annual Environmental Protection Agency Conference
on Analysis of Pollutants in the Environment, May 8-9, 1991, Norfolk,
Virginia
-------
ABSTRACT 547
Assessment of analytical variability of dioxins and furans in environmental
samples from bleached pulp mills
Henry D. Kahn and Maria D. Smith, US EPA, and Kirk Cameron, Science Applications
International Corporation
This presentation considers data collected in 1988 by EPA and the paper industry as part of a
cooperative study to evaluate the discharge of dioxins and furans from mills in the United States that
bleach wood pulps with chlorine or chlorine derivatives. Bleaching wood pulp with chlorine has been
shown in other studies to be a source of dioxins and furans. The study included the measurement of
2,3,7,8-TCDD and 2,3,7,8-TCDF in samples of effluent, pulp and sludge taken at all mills in the
United States that bleach pulp (referred to as the "104 Mill Study"). The analyses of the samples were
conducted using an analytical method developed by the paper industry known as Method 551. The data
are used in a number of statistical analyses which are discussed in this presentation. These analyses
include an assessment of analytical measurement variability and the combined effect of analytical
measurement variability and field sampling variability using components of variance analysis. Both
these components of variability are shown to be small relative to other components of the total
variability in the data. Other analyses of the data discussed include distributional properties of detected
measurements, distributions of reported detection levels and the effect that different amounts of
chlorine usage and chlorine dioxide substitution have on levels of 2,3,7,8-TCDD and 2,3,7,8-TCDF.
-------
548
In 1988, the Environmental Protection Agency and the paper industry
conducted a cooperative study to evaluate the discharge of dioxin and
furan from all 104 mills in the United States that bleached wood pulps
with chlorine or chlorine derivatives; the study is referred to as the
"104-Mill Study". These mills were the subject of investigation because
other studies demonstrated that the bleaching of wood pulp with chlorine
is a source of dioxins and furans. An important objective of the 104
Mill Study was to measure dioxin and furan levels in samples of
effluent, sludge, and pulp from the mills. The measurements were
performed using a high resolution GC/MS and isotope dilution analytical
procedure (known as Method 551) developed by the paper industry for the
most toxic of the dioxin and furan congeners: 2,3,7,8-
tetrachlorodibenzo-p-dioxin (TCDD) and 2,3,7,8-tetrachlorodibenzofuran
(TCDF). The effluent samples were obtained from treated wastewater at
the mills. Very few untreated wastewater samples were obtained in the
study. The sludge samples were obtained from semi-solid residue from
the treatment system. The pulp samples were cellulose fibers after
conversion from wood chips. Most of the effluent, sludge, and pulp
samples were collected in mid to late 1988. Paper industry
representatives managed the collection of the samples, coordinated the
laboratory work and forwarded the analytical results to EPA. The
Engineering and Analysis Division in EPA's Office of Water coordinated
quality control reviews with industry representatives, constructed
computer files containing the data and performed analyses that include
the results described here. The data and a thorough description of the
statistical analyses are contained in references [1] and [2].
-------
549
Information useful in preparing this paper was provided in paper
industry reports (references [4] and [5]) and EPA documents (references
[3] and [6]).
This paper will concentrate on a number of results and conclusions
based on our analysis of the 104 Mill data. Of primary interest is the
analysis of a number of laboratory and field replicates collected in the
study which demonstrates that for effluent, sludge and pulp: (1)
variability in the data due to analytical measurement is small; and (2)
the combined variability in the data due to analytical measurement and
field sampling is small. For (1) and (2), only the results based on
analysis of effluent data will be presented here; the results for sludge
and pulp are similar. Also of interest are: (3) greater chlorine use is
associated with higher TCDD and TCDF discharges from the mills when
combined mill output from effluent, sludge and pulp is analyzed; and (4)
increased chlorine dioxide substitution for chlorine is associated with
slight reductions in TCDD and TCDF. For (3) and (4) only the results
for TCDD are presented. Results for TCDF are similar. Also, as part of
our analysis we determined that, for effluent, sludge, and pulp: (5)
detected values are approximately lognormally distributed; and (6)
detection levels of 10 ppq for effluent and 1 ppt for sludge and pulp
are achievable based on reported detection levels reported (these
detection levels were established as goals at the beginning of the
study). For (5) and (6), only the results for effluent will be
presented here.
As part of the cooperative agreement, industry agreed to collect
-------
550
the data and send the results to EPA. The National Council of the Paper
Industry for Air and Stream Improvement (referred to as NCASI) managed
the program for industry. NCASI provided guidance on taking the
samples, developed the laboratory method for 2,3,7,8-TCDD and 2,3,7,8-
TCDF, submitted the samples to labs and reviewed the results before
forwarding them to EPA. The Engineering and Analysis Division (formerly
the Industrial Technology Division) in EPA's Office of Water was
responsible for coordinating the submission of data from NCASI to EPA
and data quality control reviews, entering the data into computer files
and performing a variety of analysis tasks including the work described
here.
Data
Each mill was required as part of the agreement to provide one
sample from each of effluent, sludge, and fully bleached pulp from each
bleach line at the mill. These samples were composite samples taken
over a 5-day period. This generated about 400 samples with about 80
additional samples for QA/QC. The mills also submitted process
information corresponding to the dates of sampling. EPA also received
from NCASI a limited amount of QA/QC information (recoveries and ion
ratios). The paper industry used only two labs for all of the
laboratory analyses. Both labs did some samples from each of effluent,
sludge, and pulp; however, the bulk of the analyses for any particular
matrix was limited to one lab. The two labs were Wright State
University in Dayton, Ohio which did about 80% of the pulp analyses.
Enseco-California Analytical Laboratories in West Sacramento, California
was the other laboratory and did 89% of the effluent samples and 81% of
-------
551
the sludge samples. None of the effluent or sludge samples were
analyzed by both laboratories. Since there were only a few pulp samples
done by both labs, inter-laboratory variability could not be estimated.
The emphasis in this paper is on the results for TCDD measurements
in effluent at kraft mills. There are several reasons we chose to focus
on effluent. The primary reason is that the conclusions based on
analyses of effluent data are similar to those based on sludge and pulp
data. In addition, there were confounding factors in the measurements
of sludge and pulp which are not present in effluent. In some cases,
the sludge samples were difficult to obtain physically, and the results
may not reflect completely the effectiveness of the treatment system.
Pulp is the final product rather than a by-product as are effluent and
sludge. However, the pulp was sampled before going through the drying
process and the water from this drying process became part of the
effluent which was then sampled. This may have resulted in some double-
counting of TCDD and TCDF levels between pulp and effluent.
We also decided to focus on the results for TCDD for this paper
because TCDD and TCDF are highly correlated and, accordingly, the
results for TCDD and TCDF are similar. Figure 1 (Figures and Tables are
located at the end of this paper) shows the strong relationship between
TCDF and TCDD for effluent from kraft mills. The linear regression of
the data shown in Figure 1 yields an R2 of 79% for detected measurements
of TCDD and TCDF.
In addition, we are focusing on kraft mills for a number of
-------
552
reasons. The processes at kraft and sulfite mills are very different
and, in our analysis, we found a significant difference in the data from
the two types of mills. Sulfite mills would be expected to produce less
TCDD and TCDF because they use less bleaching than kraft mills and In
fact most of the measurements from sulfite mills were reported to be
close to detection level. In addition, sulfite and kraft mills tend to
use different types of wastewater treatment so that mill type and
treatment type would be confounded in the data. There were also some
difficulties with the TCDD and TCDF analyses of samples from sulfite
mills. Part of this was apparently due to the low levels of TCDD and
TCDF in the sulfite samples, and part was due to analytical
interferences.
Lognormal Data; In general, we found that the detected measurements
are approximately lognormally distributed across mills. This finding is
not surprising but important in that it provides the basis for the use
of logarithms of the detected measurements in our analyses. The
probability plot in Figure 2 demonstrates the approximate normality of
the Kraft mill effluent TCDD data.
Non-Detect Measurements: Measurements reported as non-detect
constitute an important segment of the data; for example, 28% of the
effluent samples in kraft mills were non-detects for TCDD. Because all
of the mills had detected concentrations in either effluent, pulp, or
sludge for TCDD or TCDF, we concluded that a non-detect was more likely
to be an amount too small to measure rather than an indication that the
TCDD and TCDF were not in the sample.
-------
553
On the basis of the reported data, we concluded that a detection
level of 10 ppq for TCDD and TCDF in effluent was reasonable. The use
of 10 ppq for effluent was established as a goal at the beginning of the
cooperative study. There were 30 effluent measurements for TCDD from
both kraft and sulfite mills that were reported as detection level
values. The minimum detection level is 3 ppq, the maximum 17 ppq, the
mean 7.7 and the median 7.5 ppq. The cumulative distribution (Figure 3)
shows that about 80% of the non-detect measurements were reported at or
below the target detection level of 10 ppq. In addition, after
completion of the 104 Mill Study, EPA received new data from industry
which report almost all detection levels as less than 10 ppq for
effluent. Reports developed by NCASI (reference [5]) now support the
use of detection levels of 10 ppq.
For the analyses presented in this paper, we considered two
possibilities for treatment of the non-detect measurements: substituting
of a value equal to one-half the detection level or substituting a value
equal to the detection level for measurements reported as non-detect.
The choice of substituting a value of one-half the detection level for
non-detects offered a reasonable "middle ground" approach and the
results did not differ much from substituting the detection level.
Analytical and Field Sampling Variability: Evaluation of
analytical and field sampling variability was possible using a number of
laboratory and field replicate sample measurements collected as part of
the study. A statistical procedure referred to as components of
variance analysis (see, e.g., [7]), we were able to estimate: (1) the
component of variation in the data due to analytical variability
-------
554
utilizing the laboratory replicate measurements; and (2) the component
of variation due to the combined effect of field sampling and analytical
variability utilizing field replicate sample measurements. The results
of the components of variance analysis demonstrated that the
contribution to the total variability in the data due to analytical
variability and field sampling variability was low as measured by
laboratory and field replicates.
Laboratory replicates are individual samples that are split in the
laboratory to form separate aliquots of the same sample; the separate
aliquots are then analyzed to generate separate measurements of the
concentration of the sample. Variation in the replicate measurements is
attributable to variation in the analytical measurement process.
Repeated laboratory measurements on the same samples (i.e., laboratory
replicates) may be used to estimate variability in concentration
measurements due to analytical measurement variability.
Field replicate samples are physically distinct samples collected
at the same time using identical field sampling and handling protocols.
Measurements on distinct field replicate samples may be used to estimate
variation due to the combined effect of field sampling techniques and
analytical variability.
In this study, about 30% of all samples were either field or
laboratory replicate samples. For TCDD concentrations in effluent,
there were 107 samples from 84 kraft mills of which 34 were replicate
samples from 15 mills. The number of combined laboratory and field
replicates from each of these 15 mills varied from 2 to 3 samples. (Not
all mills provided replicate samples.) There were 15 laboratory
-------
555
replicates from 6 mills and 19 field replicates from 9 mills.
We examined analytical and field sampling components of variability
because we intended to average the replicates to determine mill specific
values for use in other analyses and we wanted some indication of the
effect that this averaging would have. In addition, a paper industry
study (reference [8]) had asserted that the analytical variability in
measuring TCDD and TCDF is high. This assertion is not supported by a
components of variance analysis which provides a specific estimate of
the component of variance in the data attributable to variation in
analytical measurement.
Components of variance analysis may be used to examine any
components or sources of variation given the availability of appropriate
data. In this case, for instance, examination of variability of inter-
lab effects and variability due to field sampling alone would be
desirable. However, there were not enough data to evaluate the
variability due to inter-laboratory effects because only a few samples
were analyzed by both laboratories. Also, evaluation of variability due
to field sampling alone was not possible because no lab replicates of
field replicates were analyzed.
A plot of the laboratory replicate effluent TCDD measurements for
kraft mills is shown in Figure 4. A pair of replicate measurements that
agree perfectly would be plotted on the diagonal dashed line. To show
the approximate variability, the figure also shows a 95 % confidence
ellipsoid for the data. The correlation coefficient between replicates
-------
556
is 0.98. In performing the analysis, we assumed the data are
approximately lognormally distributed and set non-detect measurements
equal to one half the detection level. The data for field replicate
effluent TCDD for kraft mills are shown in Figure 5 and demonstrate
similarly good results. The correlation coefficient is 0.99. The
results for sulfite mills were not as good (correlation coefficient of
0.73) although there were only five laboratory replicates for TCDD. The
laboratory effluent TCDD replicates for sulfite are shown in Figure 6.
There was only one field duplicate pair from sulfite mills.
The results of the components of variance analysis are summarized
in Table 1. The amount of variability due to analytical measurement was
small, 1.4% of the total variability. The amount of variability due to
the combination of field sampling and analytical error is also small,
0.8% of the total. Some assurance that these small percentages are
valid throughout the range of the data obtained in the 104 Mill Study is
provided by the plot shown in Figure 7. The plot shows the cumulative
distribution of TCDD effluent measurements by replicate type for the
kraft mill data. There are more non-replicate measurements but, as
shown in the plot, the range covered by non-replicate and replicate
measurements is virtually the same.
The following conclusions are supported by the components of
variance analysis: (i) the component of variability in the data due to
laboratory analytical measurement variability, or analytical
variability, is small; (ii) the component of variability in the data due
to the combined effect of field sampling-variability and analytical
-------
557
variability is small; (iii) given (i) and (ii) , it was reasonable to
average replicates to determine mill specific values for use in other
analyses; and (iv) analytical variability or field sampling variability
alone are not sufficient to explain the variability observed in the TCDD
and TCDF data. These results suggest that the observed variability in
the data may be due to differences among the mills in production,
manufacturing processes and other factors that could be controlled at a
specific mill.
We were able to examine some other sources of variability in the
bleaching operations. This analysis examined the combined output of
TCDD from effluent, sludge, and pulp with output adjusted for the amount
of pulp production in each mill. The results are not strong and tend to
support working hypotheses generally accepted in the industry concerning
relationships among plant operations and generation of TCDD and TCDF.
The data collected in this study were not intended to support an
analysis of what factors would account for increases in TCDD and TCDF.
However, we were able to examine two factors that were presumed to
influence TCDD and TCDF levels using the available data. These factors
were chlorine usage and chlorine dioxide substitution for chlorine
usage.
Chlorine Usage; Chlorine is important because it is used in
bleaching to whiten the pulp and other studies have shown that most TCDD
and TCDF are produced in the chlorination stage. Different amounts of
chlorine are required in bleaching to produce different products. For
example, high grade writing paper requires more bleaching than diapers.
-------
558
Using the 104 Mill data it is possible to demonstrate a weak positive
relationship between the chlorine use and TCDD formed. A plot of the
data is shown in Figure 8 overlaid by the estimated regression line a.nd
a 90% confidence band about the estimated regression line. This
relationship accounts for only about 30% of the variation in the data.
One problem may be that over-chlorination in the chlorination stage even
for a very short period of time may lead to excess TCDD and TCDF
although the overall chlorination may remain about the same as usual
(see reference [3]). This problem may provide a partial explanation of
the relative weakness of the estimated relationship. In Figure 8, the
vertical axis is the total TCDD in Ibs/ton of air-dried brownstock pulp
adjusted for amount of production. The horizontal axis is the amount of
chlorine in Ibs/ton of air-dried brownstock pulp. The estimated
regression equation is
log,0(total TCDD)= -0.449 + 0.010*C12
with an R2=32%. The plot shows an upward trend in the data although the
relationship is not well defined.
Chlorine Dioxide Substitution: Chlorine dioxide is substituted for
chlorine in bleaching to improve effluent quality and to reduce TCDD and
TCDF (reference [3]). Very few mills substituted chlorine dioxide for
more than 30% of their chlorine usage (most substituted between 0 and 20
%) and not all mills substituted. In the 86 kraft mills, 59 bleach
lines out of the 165 bleach lines did not use any substitution.
Regression analysis of data from mills that practice substitution
demonstrated a relationship which accounted for at most 16% of the
-------
559
variation in the data. The increased use of substitution produced
slight reductions in TCDD formation. The data are shown in Figure 9.
The estimated regression equation is
Iog10(total TCDD) = 1.145 - 0.693*%ClO2substitution
with R2=16%. The plot in Figure 9 shows the weak relationship between
the total TCDD and the percent ClO2 substitution, which accounts for the
low R2.
Analysis of chlorine and chlorine dioxide substitution separately
is problematic. The order that these chemicals are added in the
bleaching process may affect the amount of TCDD and TCDF formed. Adding
the chemicals in stages instead of in one dose may reduce TCDD and TCDF.
In laboratory and field studies, it has been found that there is
competition between the chlorine and chlorine dioxide and this
competition may increase the amount of chlorine related to the formation
of TCDD and TCDF (reference [5]).
EPA is continuing to collect data and study the pulp and paper
industry to support the development of water pollution control
regulations. We are aware that the industry is dynamic and responding
to the challenge of reducing and controlling TCDD and TCDF discharges.
The situation represented by the data collected in 1988 is changing.
Preliminary analysis of some post-1988 data indicate some changes in the
amounts of TCDD and TCDF in effluent, pulp and sludge from paper mills
have may occurred. Changes in the levels of TCDD and TCDF discharged
should not, however, affect the conclusions based on the 1988 data
presented here. EPA is sampling a number of mills (total of 16 to 19
mills) with 4 or more of these planned for long term sampling. In these
-------
560
sampling episodes, EPA is collecting TCDD and TCDF concentrations at
more places in the process than were collected in the 104-Mill Study.
This should provide a more complete database and support the evaluation
more factors influencing variability in TCDD and TCDF measurements. In
addition, a detailed questionnaire has been mailed to these facilities.
Additional self-monitoring data and process information should be
provided with the completed questionnaires which will add substantially
to the base of information available to support evaluations of the
industry and the Record for EPA's rulemaking activities.
-------
561
REFERENCES
[1] USEPA/Paper Industry Cooperative Study. The "104-Mill Study".
Statistical Findings and Analyses. U.S. EPA, Washington, D.C., July 13,
1990.
[2] Preliminary Draft Report. Follow-up Analysis of Analytical Data
from the 104 Mill Study. U.S. EPA, Washington, D.C., July 31, 1990.
[3] Summary of Technologies for the Control and Reduction of
Chlorinated Organics from the Bleached Chemical Pulping Subcateaories of
the Pulp and Paper Industry. U.S. EPA, Washington D.C., April 27, 1990.
[4] USEPA/Paper Industry Cooperative Dioxin Study; The 104 Mill
Study. Technical Bulletin No. 590. National Council of the Paper
Industry for Air and Stream Improvement, Inc., New York, New York, May
1990.
[5] An Intensive Study of the Formation and Distribution of
2.3.7.8-TCDD and 2.3.7.8-TCPF during the Bleaching of Kraft Pulps.
Technical Bulletin No. 591. National Council of the Paper Industry for
Air and Stream Improvement, Inc., New York, New York, May 1990.
[6] Development Document for Effluent Limitations Guidelines and
Standards for the Pulpr Paper, and Paoerboard and the Builders7 Paper
and Board Mills Point Source Categories. U.S. EPA, Washington D.C.,
October 1982, EPA 440/1-82/025.
[7] Box, G.E.P., Hunter, W.G. and Hunter, J.S., Statistics for
Experimenters. John Wiley & Sons, 1978.
[8] A Study of the Variability of 2.3.7.8 - TCDD and 2.3.7.8-TCDF
in Bleached Kraft Mill Pulps. Sludges and Treated Wastewaters. Technical
Bulletin No. 568. National Council of the Paper Industry for Air and
Stream Improvement, Inc., New York, New York, July, 1989.
-------
562
QUESTION AND ANSWER SESSION
MR. TELLIARD: Any questions?
MR. KAHN: Larry, no questions?
MR. BERTHOUEX: Not to ask a
question, but to add a comment.
My name is Mac Berthouex from the University of
Wisconsin.
I like this kind of a study very much and I wish there
were more analysis of variance studies made. I'll just briefly
tell of a similar experiment that was done in Denmark on dioxin
and furan emissions from incinerators, a designed experiment with
loading rate to the incinerator, operating temperature of the
incinerator, type of waste going in and so on. When it was all
designed, they found out they didn't have enough samplers of the
same kind to study all of the incinerators in the study. So,
just as a matter of chance, they were forced to add one variable
to the experiment which was the kind of a sampler. Both samplers
were widely used and approved and when all of the data was in, it
turned out that the most significant variable of all the things
under study was this sampler. The difference between samplers
explained more variation than analytical procedures which were
done in different laboratories and a lot of other things.
I suspect if we saw more studies of this kind, we would
find out that very often we may blame the chemist for variations
in the data which is really not their fault and it is coming from
other sources and it's very important for us to know what the
sources are.
MR. TELLIARD: He's a damn
engineer.
MR. BERTHOUEX: I am an engineer.
You can blame the engineers if you like.
MR. TELLIARD: Thank you.
-------
563
MR. KAHN: Well, I think that's
part of the message to this group. It's when you see these high
levels of variability, it's not the chemist's fault, it's
s omebody else.
MR. TELLIARD: Thank you, Henry.
-------
564
CO
O
CO
H
Z
cd
^*
CO
Q
Q
3
3
Q
Q
(Odd) JQD1
-------
565
r<
O
E
Q
Q
U
H
H
z
•J
u-
• *
H
O
H
3
H-I
PQ
PQ
O
o
CO
N*
5
H
o
O
o
o
o
eu
.S
a
o
o
c
o
U
U
H
-------
566
n
a
E
o
X
a.
<
3 *
u 0
si
Q >
H
H P
§*
D
(^ ti
fc H)
a
OH
C
•*4
"^
Q
Q
00
6
G
0
U
00
6
NO
6
cs
•
o
O
6
jo noi)jodoij
-------
567
C/D
w
H
<
U
PQ
fa.
Q
Q
U
H
O
1/3
i—i
2
H
tiI I i i i
• mill i i
o
p
o
o
o
o
p
o
O PLH
~ PL,
CX
O
O
o
0
O
O
C
0>
2
§
U
P
Q
o
p
o
o
o
O O
p p
o o
o —
bdd u!
o
o
o
o
QQ3X
p
^ o
p
o
-------
568
u
«
{^
P
O H
HH >^r
fe «
o
Q
U
H
o
t/3
o
o
o
o
o
cx
p-l
c
• v-4
G
o
c
o
U
Q
Q
U
H
i...
o
o
o
o
o
in
-------
569
2
PH
H
<
U
PQ
•
Q
Q
U
H
s
H
i.
I • III I I 1 I 1 1
o
p
o
o
o
o
p
o O^
O (X,
-• A*
o -
§ I
o
o
cj
Q
Q
-! H
o
o
O
p
o
o
o
o
p
o
o
o
p
o
o
o
o
o
^ o
p
o
QQDX
-------
03
H
Z
570
o
4-i
cd
o
Z
,
o
•0
.2
.0
Q
Q
U
H
fc Nj
o 5
w
HH 04
£ H
> UH
L^ C^
HH
Q
P.
H
U
PQ
o
o
~^ o
o
o
OH
-3
a
o
'+-»
cd
o
a
o
O
x
o
»^^
Q
oo
O
"t
o
u
-------
571
00
W
5
O
hH
fe
Q
Q
U
H
J
<
H
O
H
Q
H
H
c/3
!D
>->
Q
<
^
gr
O
CO
i—*
2
H
UH
•<
^
*
o o
11 i i i i
O
o
fN
CO
Q
o c
o o
.-H *•»
(A
fS
0
o
p
o
o
o
p
o
o
o
o
o
p
o
80+H
°o;/sqt)
-------
Q
Q
U
H
572
ON
H
O
H
Q
W
H
3 2
5 O
PQ
O
U
z
o
%
H
DO
I I I I t I 1
O
O
CU
CO
CQ
Q
e
o
•<_>
(A
o c
— o
CO
X)
9
O
p
O
O
o
p
o
o
o
o
<—I
o
p
o
80+3 * (JSaaV noj/sqi) QQDX
-------
573
TABLE 1
COMPONENTS OF VARIANCE RESULTS
EFFLUENT REPLICATE SAMPLES FROM KRAFT MILLS
N SSI SS1% SS2 SS2%
Lab Replicates
Log10 (TCDD) 15 5.57 98.6 0.08 1.4
Field Replicates
Log10 (TCDD) 19 4.75 99.2 0.04 0.8
SSI = Between replicate set sum of squares, i.e., the sum of
the squared deviations of the replicate set means from the
overall mean
SS2 = Within replicate set sum of squares, i.e., the sum of
the squared deviations of individual sample measurements from
their respective replicate set means
N = Number of replicate sets
SS = Total Sum of Squares = SSI + SS2
SS1% = (SS1/SS)*100
SS2% = (SS2/SS)*100
-------
574
MR. TELLIARD: Our next speaker is
Chuck White. Chuck is also with the Office of Water.
Chuck is going to talk about a study that we did a year
ago...almost a year and half ago...on a national evaluation of
domestic sewage sludge from some 180-something publicly-owned
treatment plants and he's going to address a multi-approach to
the various analytes that were analyzed in that study.
Chuck?
-------
575
STATISTICAL ANALYSIS OF
ANALYTE CONCENTRATIONS IN
MUNICIPAL SEWAGE SLUDGE
WITH MULTIPLE DETECTION
LIMITS
Chuck White, USEPA
Henry Kahn, USEPA
Kathleen Stralka, SAIC
Presented at
The Fourteenth Annual EPA Conference on
Analysis of Pollutants in the Environment
May 9, 1991
-------
576
Statistical Analysis of Analyte Concentrations in
Municipal Sewage Sludge with Multiple Detection
Limits
Chuck White, USEPA
Henry Kahn, USEPA
Kathleen Stralka, SAIC
Washington, D.C.
This paper will present the current analysis approach to multiple "detection limits" contained
within physical analytical data from the National Sewage Sludge Survey (NSSS). The term "de-
tection limit" generically describes an analytical response where it has been determined that the
concentration of an analytical material, e.g., zinc concentration, is below a certain level although a
single numeric estimate for that concentration is not produced. We will provide basic background
for the survey, our motivation to seek out innovative methods of statistical analysis, describe some
of the methods we considered, present selected results of our analyses, and discuss the usefulness
of the selected statistical approach to the analysis of these NSSS data.
Background on the Survey
The NSSS was conducted in order to obtain technical, economic, and chemical analytical data that
will support analyses required for the development of sewage sludge use and disposal regulations
under the Clean Water Act. These analyses include the regulatory impact analysis and the aggregate
risk analysis. The regulatory impact analysis will estimate the cost and potential for closures of
Publicly Owned Treatment Works (POTWs). The aggregate risk analysis will estimate the health
impact of current use and disposal practices.
The survey included questionnaire and physical analytical portions. The questionnaire portion
requested both physical process and economic condition information. The physical analytical
portion physically sampled final process sewage sludge and analyzed these samples for 412 analytes.
In order for the Agency and the Public to better understand the potential risk associated with sludge
use and disposal practices, the current statistical analysis has focused on estimating national dis-
tributions for 28 pollutants of concern identified by analyses of previous data collections.
The NSSS was restricted by design to POTWs that practice secondary or better wastewater
treatment. The population of POTWs practicing secondary or better treatment was determined
from responses to the 1986 NEEDS Survey conducted by EPA's Office of Municipal Pollution
Control. For the physical analytical portion of the survey, this population of 11,407 POTWs was
divided into strata by four flow groups. These four flow groups were:
Statistical Analysis or Analyte Concentrations in Municipal Sewage Sludge with Multiple Detection Limits 1
-------
577
WET Weight Concentrations of ZINC
Versus Percent Soflds
IjDQOyOQQUOO ~*
IQuOQQjOO ~>
1JPQPQQ
tOOuOO -g
•nun
.Alr-
to
10UO
•NKLO
Conc8fifrB00fis of ZINC
VBTBUS Percent Solids
1,OOQuOO
10QUOO
10JOQ
++. + * *-
0.1
1JO
100
1000
Statistical Analysis of Analyte Concentrations in Municipal Sewage Sludge with Multiple Detection Limits 2
-------
578
Flow < = 1 Million Gallons per Day
1 MOD < Flow < 10 MOD
10 MGD < Flow < = 100 MGD
100 MGD < = Flow
A stratified probability sample of 208 POTWs was then selected from within these strata. This
probability design provides the basis for calculating unbiased national estimates with calculable
sampling variability. However, the stratification used in the design requires that the sampling
fraction for each stratum be taken into account when calculating such unbiased national estimates.
The sampling fraction was the probability that a particular POTW within a particular stratum
would be selected for participation in the survey.
Characteristics of the Physical Analytical Data
Innovative statistical techniques were adopted because of the chemical analytical results obtained
in the survey. Relevant characteristics of these results were caused by a combination of the percent
solids correction required to compare analyte concentrations in sludges and by the behavior of the
"detection limits" reported from the survey samples.
In general, there is an increasing relationship between whole sample concentrations of pollutants
and the percent solids in a sewage sludge sample. This relationship is illustrated by the plot titled
Percent Solids ofNSSS Samples Versus Wet Weight Concentrations of Zinc. However, converting
all sample results to a dry weight basis removes this relationship. The plot titled Percent Solids
ofNSSS Samples Versus Dry Weight Concentrations of Zinc illustrates that this trend is removed
by the dry weight conversion. This log-log plot shows dry weight Zinc concentrations varying in
a random fashion about some median concentration value independently of percent solids.
The form of "detection limit" used in the NSSS is called a Minimum Level. The Office of Science
and Technology in EPA's Office of Water defines the Minimum Level to be the lowest acceptable
concentration level used to establish the calibration relationship. That relationship describes the
response of the measurement system to analyte concentrations. The plot titled Percent Solids of
NSSS Samples Versus Wet Weight Concentrations of Mercury illustrates the whole matter con-
centration results and Minimum Levels for a pollutant where a noticeable portion of the sample
results were reported to be below the Minimum Level. The plot titled Percent Solids of NSSS
Samples Versus Dry Weight Concentrations of Mercury illustrates the mixture of quantified values
and Minimum Levels^ that required the search for innovative statistical methods. In particular,
notice that Minimum bevels for this analyte occur throughout the same range of concentrations
as the quantified values.
Estimating the Frequency Distribution for Pollutant
Concentrations within a Flow Group
The occurrence of "detection limit" observations in environmental data is not unusual and there
are a number of statistical methods for analyzing such data. Several computationally simple
methodologies and the Agency's current methodology for most analytes in municipal sewage sludge
are discussed briefly.
Statistical Analysis of Analyte Concentrations in Municipal Sewage Sludge with Multiple Detection Limits 3
-------
579
WET Weight Concentrations of MERCURY
Versus Percent Solids
HTtrartlon
Cmg/l)
moo
too
"M
10
10JD
•KILO
DRf Weight Concentrations of MERCURY
UBTBUS Percent Solkte
(nrifj/Ka
10000
KLOO
1JDO
0,10
+ +
oo
i
ai
104)
KKLO
mart Sri*
Statistical Analysis of Analyte Concentrations in Municipal Sewage Sludge with Multiple Detection Limits 4
-------
580
Computationally Simple Methodologies
We will discuss four computationally simple methodologies for estimating the frequency distrib-
ution within flow groups for pollutant concentrations when data from a particular pollutant con-
tains several "detection limit" values. The first two of these methods will be used in illustrating the
estimates calculated by EPA's current method.
One method is to assign the Minimum Level ("detection limit") value to sample results reported
below the Minimum Level. However, the true concentration value could be anywhere between
zero and the Minimum Level. Hence, the method of assigning the maximum possible concen-
tration value for the sample will result in overestimates for the frequency of occurrence for con-
centrations of the pollutant at the reported Minimum Levels.
Another method is to assign the sample results reported below the Minimum Level to the value
zero. This method is likely to underestimate the frequency of occurrence for concentrations of the
pollutant that are actually between zero and the Minimum Level.
A third method is to discard sample results reported below the Minimum Level. This method
will overestimate the frequency of occurrence within the flow group for higher concentrations of the
pollutant because pollutant concentrations for pollutants not detected at low levels will be treated
as if they do not exist. This method discards valuable information contained in the survey results
and is not recommended.
The fourth method is to assign sample results reported as below the Minimum Level to some
fraction of the Minimum Level. This method requires a policy decision as to what fraction to use
and the precision of any estimates produced from this procedure will depend on how well the
chosen fraction models the true concentration values censored by the detection limit. This method
is commonly used as a computationally simple manner for analyzing censored data.
EPA's Current Methodology
EPA's current methodology for sewage sludge concentration data was adopted in order to maxi-
mize the use of information contained within the available data. It uses quantified observations,
assumptions about the shape of the curve that will describe the frequency distribution for the
pollutant concentrations, and Minimum levels in order to pick the optimum estimate for the scale
of the curve with the assumed shape. The assumed shape of the curve is that of the lognormal
distribution and the optimum estimate for the scale of the curve is expressed as a combination of
the log mean and the log variance for the distribution represented by the curve. This statistical
method, called Maximum Likelihood Estimation for Left Censoring Points, was developed for use
with the normal distribution and single "detection limits" by Cohen in 1959. Cohen and others
have extended this method to multiple "detection limits" and other distributions through a series
of papers published in the 1970's and J980's. An explicit description of how EPA has used this
method with the NSSS is provided in the Technical Support Documentation for Part I of the Na-
tional Sewage Sludge Survey Notice of Availability.
Estimating the National Distribution of Pollutant
Concentrations
Flow group means and variances were combined in the usual fashion for calculating overall mean
and variance estimates from a stratified probability design for sampling. This usual method requires
knowledge of the probability for selecting a POTW during the design of the survey in order to
weight that POTWs response
Statistical Analysis of Analyte Concentrations in Municipal Sewage Sludge with Multiple Detection Limits 5
-------
581
Cummutofre DttrtwOon Function*
ZINC
10uOO
I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' '^ T ' T ' ' I
000.1020304050*07080910
RMfon of POIW VWuw Lnt Itai OanovMlon
O -
Mnfenum
MMbutfon
MERCURY
Concentration
a/R
VOOQjQO -
1000 -
1JDO -
Rvctton of POTWVUuw LiwThvi Conofrtradon
o«SuMUiZlHO • - Subeftuft Mrtmum
Statistical Analysis of Analyte Concentrations in Municipal Sewage Sludge with Multiple Detection Limits 6
-------
582
when calculating national estimates. These combined mean and variance estimates were then as-
sumed to be the mean and variance for a lognormal distribution of pollutant concentrations across
the nation. In particular, the estimated distribution is used to represent the distribution among
publicly owned treatment works of pollutant concentrations in dry weight sewage sludge that is
ready for disposal and that is generated by secondary or better treatment of wastewater.
Fit of the Estimated Distribution
Since the true underlying distribution of pollutant concentrations is unknown, it is difficult to
evaluate the fit of the maximum likelihood estimates. However, using two substitution methods,
it is possible to approximate upper and lower bounds for each pollutant's cumulative frequency
distribution. The upper bound is obtained by assigning the value of the Minimum Level to samples
reported below the Minimum Level and then calculating percentiles of the observed distribution in
the usual fashion for data from a stratified probability design. The lower bound is obtained by as-
signing the value of zero to samples reported below the Minimum Level followed again by the usual
calculation for percentiles of the observed distribution.
The plot titled Cumulative Distribution Functions: Zinc illustrates the curve estimated by these
three procedures when all sample results are reported above the Minimum Level. A point on a
curve indicates what percentage of the POTWs across the nation have pollutant concentrations
below the concentration quantified on the left axis. The curve accented with solid circles is the
frequency of occurrence for Zinc concentrations reported in the survey and as adjusted for POTW
membership in each appropriate design flow group. Since all sample results are reported above the
Minimum Level, this curve illustrates the results from both substitution methods. The curve ac-
cented with cross marks is the curve estimated by the maximum likelihood methodology. This
estimate appears to model the observed distribution quite closely.
The plot titled Cumulative Distribution Functions: Mercury illustrates the curve estimated by this
procedure when a noticeable portion of the sample results are reported as below the Minimum
Level. The top curve is the frequency of occurrence for Mercury concentrations reported in the
survey, as adjusted for POTW membership in each appropriate design flow group, and where the
concentration value for a sample reported as below the Minimum Level is set equal to the Mini-
mum Level. The bottom curve is the frequency of occurrence for Mercury concentrations reported
in the survey, as adjusted for POTW membership in each appropriate design flow group, and where
the concentration value for a sample reported as below the Minimum Level is set equal to zero.
This maximum likelihood estimate, generated in the presence of multiple 'detection limits," appears
to be reasonable.
However, it is important to note that there are at least two cases where this maximum likelihood
method does not work with these sludge data. In the first case, if there are no sample concentration
values quantified above the Minimum Level for any one of the survey design flow strata then no
national estimates will be produced. In the second case, if the data dp not approximately follow
the lognormal curve than national pollutant concentration, estimates will tend to be unreasonable.
The plot titled Percent Solids ofNSSS Samples Versus Dry Weight Concentrations ofPCB 1254
illustrates a pollutant that does not appear to follow a lognormal distribution. In particular, the
log-log plot does not show the uncensored dry- weight PCB-1254 concentrations varying in a ran-
dom fashion about some median concentration value. The plot titled Cumulative Distribution
Functions: PCB-1254 shows that the curve estimated by the current methodology falls between
those extreme curves estimated by the substitution methodologies. However, the curve produced
by the maximum likelihood estimates does not appear to be reasonable. In particular, the inflection
point at the 50th percentile, i.e., the log mean, of the estimated lognormal distribution does not
appear on the plot. It does not appear on a plot that shows the range of all observed values and
"detection limits" recorded from the survey for this pollutant. Additionally, the gap between the
Statistical Analysis of Analyte Concentrations in Municipal Sewage Sludge with Multiple Detection Limits 7
-------
583
Commufatfre Dfctribuflbfi Functions:
PCS-1264
Goncenrtrortli
Cug/R0>
00
Rwflufi of POTW^MuM LMI Ttan Connnlraflon
• - SUbaflUte MHrun
M^gfit Concantrettons of PC8-1254
Versus Percent Solids
Concentration
0/lcg)
10,00000
WOOD
1000
o.
o
ai
to
100
KJOO
o o o
Statistical Analysis of Analvte Concentrations in Municipal Sewage Sludge with Multiple Detection Limits 8
-------
584
quantified values and the recored minimum levels provides evicence that this PCB does not vary
log symetrically about some median value as would be expected if these data followed a lognormal
distribution.
Mean Pollutant Concentration Estimates
The table titled National Pollutant Concentrations for the National Sewage Sludge Survey Using
Lognormal Maximum Likelihood and Arithmetic Substitution Estimation Procedures presents mean
pollutant concentration and standard deviation estimates for zinc, mercury, and PCB-1254. The
mean estimates for zinc are consistent with the assumptions of lognormality since the lognormaJ
mean estimate is "close" to the arithmetic estimate. Mean estimates using a methodology specific
to the lognormal distribution'are expected to be lower than those generated by an arithmetic
methodology since the arithmetic methodology allows for more influence from unusually high
concentration values. The mean estimates for Mercury are as desired since the lognormal mean
estimate is between the arithmetic estimates. However, mean estimates for PCB-1254 provide evi-
dence that the lognormal model does not fit the data for this pollutant. In particular, the estimated
mean of the log normal distribution is three orders of magnitude greater than either arithmetic mean
and in fact exceeds the highest PCB-1254 concentration value reported in the survey.
National Pollutant Concentrations for the National Sewage
Sludge Survey using Maximum Likelihood and A rithmetic
Substitution Methods
ANALYTE
Zinc
Mercury
PCB- 1254
UMTS
mg/kg
mg/kg
ug/kg
PERCENT
DETECT
100
63
8
ESTIMATION
PROCEDURE
Substitue Zero
Maximum
Likelihood
Substitute
Minimum Level
Substitute Zero
Maximum
Likelihood
Substitute
Minimum Level
Substitute Zero
Maximum
Likelihood
Substitute
Minimum Level
MEAN
1,320.34
1201.88
1,320.34
3.90
5.22
6.58
281.33
118,118,762.55
515.86
STANDARD
DEVIATION
3,820.82
1554.42
3,820.82
8.78
15.54
8.94
1,347.08
81,514,762,398.43
1,324.97
Conclusion
The maximum likelihood procedure for left censoring points allows, in many cases, the calculation
of national estimates for pollutant concentration distributions that are more accurate than simpler
methods. When the data approximately follow the lognormaJ distribution and the lognormal dis-
Statistical Analysis of Analyte Concentrations in Municipal Sewage Sludge with Multiple Detection Limits 9
-------
585
tribution is the model then the maxirnimum likelihood method method provides excellent results.
However, this method does not work in cases where there is an extreme departure from the model
distribution. In these cases it is useful to investigate other models since the general maximum
likelihood method can be used with other probability models and it is quite powerful. The Office
of Water continues investigagc appropriate methods to use with data from the NSSS.
Statistical Analysis of Analyte Concentrations in Municipal Sewage Sludge with Multiple Detection
Limits
-------
References
586
Water Quality Data, Water Resources Re-
search, Vol. 22, No. 2, February 1986, 147-155.
Aitchison, J. and Brown, J.A.C., The
Lognormal Distribution, Cambridge University
Press, 1963.
Barakat, Richard, Sums of Independent
Lognormally Distributed Random Variables,
Journal of the Optical Society of America, Vol
66, No. 3, March 1976.
Cochran, W.G., Sampling Techniques, Third
Edition, John Wiley & Sons, Inc., 1977.
Cohen, A. Clifford, Jr., Simplified Estimators
for the Normal Distribution When Samples are
Singly Censored or Truncated, Technometrics,
Vol. 1, No. 3, August 1959, 217-37.
Hansen, M.H.; Hurwitz, W.N.; and Madow,
W.G., Sample Survey Method and Theory
Volume II: Theory, John Wiley & Sons, Inc.,
1953.
Helsel, D.R. and Gilliom, R.J., Estimation
of Descriptive Statistics from Multi-Censored
Helsel, D.R. and Cohn, T.A., Estimation of
Descriptive Statistics form Multi-Censored Wa-
ter Quality Data, Water Resources Research,
Vol. 24, No.12, December 1988, 1997-2004.
Travis, Curtis C.; Land, Miriam, Estimating
the Mean of Data Sets with Nondetectable
Values, Environmental Science and Technol-
ogy, 1990, 24, 961-962.
U.S. Environmental Protection Agency,
Standards for the Disposal of Sewage Sludge;
Proposed Rule (40 CFR Parts 257 and 503),
February 1989.
U.S. Environmental Protection Agency, An-
alytical Methods for the National Sewage
Sludge Survey, Industrial Technology Divi-
sion, Office of Water
U.S. Environmental Protection Agency,
Technical Support Documentation for Part I of
the National Sewage Sludge Survey Notice of
Availability, October, 1990.
References
11
-------
587
QUESTION AND ANSWER SESSION
MR. TELLIARD: Are there any
questions?
(No response.)
MR. TELLIARD: It's break time.
If you could get your coffee and your cookie and come on back in
here on time, we will try to keep it moving.
-------
588
National Pollutant Concentrations from the
National Sewage Sludge Survey
Using Lognormal Maximum Likelihood and
Arithmetic Substitution Estimation Procedures
Analyte Units Percent Estimation
Detect Procedure
Zinc mg/kg 100
Mercury mg/kg 63
PCB- ug/kg 8
1254
Substitute
Zero
Maximum
Likelihood
Substitue
Minimum
Level
Substitute
Zero
Maximum
Likelihood
Substitute
Minimum
Level
Substitute
Zero
Maximum
Likelihood
Substitute
Minimum
Level
Mean
1,320.34
1,201.88
1,320.34
3.90
5.22
6.58
281.33
118,762.55
515.86
Standard
Deviation
3,820.82
1,554.42
3,820.82
8.78
15.54
8.94
1,347.08
81,514,762,398.43
1,324.97
-------
589
WET Mtyt Concentrations of ZINC
Versus Percent Solids
Concentration
(mg/i)
10,000,000.00
1,000,000.00
10,000.00
1,000.00
100.00
0.1
4-
1,0
10,0
100,0
Percent Solids
Detection: +
Yes
-------
590
DRYlVe/gWConcenffafionsofZINC
Versus Percent Solids
Concentration
(mg/kg)
100,000.00
10,000.00
1,000.00
100.00
10.00
++'
-t
0.1
1.0
10.0
Percent Solids
100.0
Detection: L + Yes
-------
591
WET Weight Concentrations of MERCURY
Versus Percent Solids
Concentration
(mg/l)
10,000.00
1.00
o
o
0.1
1.0 10.0
Percent Solids
100.0
Detection: + + Yfes o o o
-------
592
DAY Weigh! Concentrations of MERCURY
Versus Percent Solids
Concentration
(mg/kg)
1.00
0.10
o
0
0.1
+ ;*wnv
++ + +
+ + +
O +
oo
I I i I I I I I I I I I I I I I
1.0 10.0
Percent Soids
100.0
Detection: + + YB o o o NO
-------
593
Cummu/ative Distribution Functions:
ZINC
Concentration
10,000.00
1,000.00
100.00
10.00
0.1 02 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0
Fraction of POTW Values Less Than Concentration
+ = Maximum Ukehood o - Subside Zero • = Substitute Minimum Level
-------
594
Cummu/aifl/e Distribution Functions;
MERCURY
Concentration
(mg/kg)
1,000.00
1.00
0.0 0.1 02 0.3 0.4 05 0.6 0.7 0.8 0.9 1.0
Fraction of POTW Values Less Than Concentration
= Maximum Likelihood c = Subsitute Zero •= Substitute Minimum Level
-------
595
CummuJative Distribution Functions:
PCB-1254
Concentration
1,000,000.00
100,000.00
1,000.00
1.00
0.10
11
0.1 0.2 0.3 0.4 0.5
0.7 0.8 0.9 1.0
Fraction of P01W Values Less flian Concentration
+ = Maximum likelihood o = Subsitute Zero • = Substitute Minimum Level
-------
596
DRY Mfe/g/tf Concenfrafons of PCB-J254
Versus Percent Solids
Concentration
(ug/kg)
10,000,00
1,000,00
100,00
10,00
o
-f
o
o
o
i 1
0.1
1.0 10.0
Percent Solids
100.0
Detection* + + Kes o o o
-------
597
MR. TELLIARD: We'd like to get
started, please. Could we get the folks in from the back a
little bit? Have them bring their cookies.
Our afternoon session is on quality assurance. Our
first speaker is Rick Johnson. Rick is with the Environmental
Protection Agency's... this is good...Science Systems Staff in the
Office of Information Research Management... wow. Basically, he's
an OIRM down in RTF.
-------
598
MR. JOHNSON: Can everybody hear
me okay?
I want to thank Bill for inviting me to talk today.
This is going to be kind of a different talk for you all. One
thing I promise is you won't see a chromatogram. You won't see; a
piece of instrumentation or equipment and you probably won't hear
a chemical name said unless I slip one out by accident. This is
an entirely different arena from most of the stuff that was
talked about today. However, I think it's a very serious and
important part of a laboratory operation. As I move through the
talk, I'll try to give you a feel for why.
I've given this talk to USDA, FDA and EPA laboratory
auditors, North Carolina Quality Assurance Group, Contract
Laboratory Data Management, CAUCUS and ORD's retreat recently.
How many people have heard this before so I don't repeat?
(No response.)
Oh, good, I can tell the joke then. Fred, pardon for
having to hear it over again.
There were two computer salesmen who had landed a
rather lucrative contract in a city which they had never been to
before. They went out celebrating that night and went sort of
bar hopping and in the course of it, had a little bit too much to
drink. Trying to make their way back to their hotel room,
unbeknownst to them, they wandered into the city zoo and they're
sitting on one of the benches in the zoo. Suddenly, there's a
roar of a lion and the one computer salesman gets up and he
starts to run away and the other one is still sitting there on
the bench waiting and kind of curious. Then one turns around and
says, well, aren't you leaving? The other one says, no, man, I'm
staying to see the movie. So, with that in mind, I'd like you to
kind of stay and bear with me for a few minutes.
My purpose here is to describe to you our program for
assuring integrity of data in laboratories as they automate.
I'll try and give you sort of a feel for what happened and why we
went about it and what we found out and where we are today.
-------
599
As Bill has indicated, the document is in review.
There are about 500 copies that have been distributed to date.
My name and address are in the speaker book, so if you want to
get a hold of me or would like a copy later, I'll certainly be
glad to make one available to you. We're currently having more
copies made. I expect them to be out of the print shop at the
end of next week forward.
I put this up here because I spent a lot of time with
two folks on contract that said it couldn't be done. This was
done on computer with computer graphics and computer coloring and
all that. They said they couldn't put the circle around it and
couldn't do the coloring and there was one reason or another and
we just worked with it for a day and half. So, just to show it
can be done, I wanted to put it up to show you that you can do
things like this.
In a nutshell, where my organization is, you can see
here there is the administrator. Under the administrator are the
major offices of Research and Development, Pesticides and Toxic
Substances, Program Planning and Evaluation, Office of
Administration and Resource Management and Office of Air and
Radiation. Unfortunately, I forgot another key one in here,
OSWER, which Joan reminded me at the last meeting. The point is
that there are numerous different folks reporting to the
administrator and we basically are under the Office of
Administration and Resources Management in what's called the
Scientific Systems Staff.
Basically, the mission of our office is really
threefold to essentially deal with hardware and software
technology to improve the information flow and management in the
agency, develop and manage the agency's information assets, and
promote data sharing and integration. I think this project
nicely fits under what I would consider to be data
sharing...quality data sharing.
Why did we get into this program in the first place?
There were a couple of things about two and half years ago that
-------
600
led us down the road of thinking we might need to look at
laboratory practices as they now exist and with automation moving
into the laboratory.
First, as many of you know, automation is moving into
the laboratory. We're steadily replacing one person after the
other one, body after the other, with one kind of piece of
software or hardware or combination thereof.
Secondly, problems were beginning to surface. We were
seeing some problems that ultimately wound up in fines,
disbarments, and a number of other penalties on folks who were
doing business with the agency because they were beginning to
tamper with data in an electronically-oriented environment.
Third, our laboratory auditors for years have been
operating with laboratory notebooks and principles thereof and
folks at the same time were moving to automate and it became
clear that the auditors... there now becomes a generation gap
between the auditors and what they know and also with what's
going on in the laboratory and the movement to automation. Many
of our auditors when they go into labs and you mention the
computer, they simply just turn around and walk away from it.
They are virtually computer illiterate. So, another reason, an
indication of why we wanted to do this was to help bring our
auditors into the 21st century with some updating for them.
Third, in spite of EPA having all kinds of edicts and
we have all kinds of measurement requirements and monitoring
methods and all of that, we did not have one single set of
uniform principles in one place to guide laboratories as they
move to automate. As a result, in the course of this talk, I'll
show you numerous laboratories that made a lot of mistakes in
trying to automate their labs because they didn't have
appropriate guidance from the agency.
Finally, to add to all of this, there were a number of
requirements underway or being developed in a variety of
different areas in the agency that seemed to be kind of scattered
about and they needed to be brought together. So, with these
-------
601
things in mind, we started this project on a shoestring about two
and a half years ago.
What we did was, first of all, we worked in a couple of
different areas. One, we wanted to go out and find out what the
current situation was in laboratories. Was it, in fact...
Are we getting an echo here? Maybe it's me. The trip
last night, I guess, on the boat.
First, we wanted to find out what the condition in the
laboratories was. We've had some indications that there were
problems and we wanted to go out firsthand and see for ourselves
so we conducted many site visits across the country at
laboratories that provide data to the agency, either through
contract programs or through requirements to generate data for
the agency to decide whether or not a chemical should be
registered for pesticide use or for introduction into the
environment as a new chemical.
Secondly, we wanted to look at what kinds of procedures
already existed in a lot of places that were already there to
guide people into automating. Automated financial systems have
been around for many years. Many of you folks use the teller
machine and the principles and understandings of data exchange
through that kind of electronic media have all been established
and several court cases settled a lot of different disputes about
different issues related to electronic data transferring stuff.
We felt like it was imperative on us not to reinvent the wheel,
but find out what lessons were learned in that arena.
Third, clinical laboratories have been moving to
automation very rapidly over the last 15 years, particularly your
forensic toxicology labs, your drug testing programs and all of
that. There were lessons there to be learned, so we also looked
into the clinical arena.
Then we decided to look at hardware and software.
There are a number of different changes going on in the hardware
and software arena. Maybe there are lessons. Maybe there are
some fixes there that probably could straighten this thing
-------
602
out... that you really have a very simple fix available.
Fourth, there are lessons from EPA's Good Laboratory
Practices, guiding principles for management when they generate
data for the agency.
Finally, there were requirements being developed that
we also felt like we needed to track.
So this is basically the seven methods or ways we went
about looking and putting this thing together.
What we found out in a nutshell through about 150
different surveys sent out and 10 different laboratory visits and
some other voluntary submissions was in a nutshell that physical
security in laboratories was lacking. People could walk in and
out of automated laboratories without much chance of running into
any kind of a problem. System access was not protected. Joe Doe
could get on the computer right behind Joe Schmoe and make data
changes and Joe Doe would never have known it and the data
becomes corrupted.
Third, and probably one of the key things was data
verification. People were manually entering data, keying it in,
and where a lot of the problems in laboratories were coming in
were not necessarily the corruption...intentional corruption of
data, but rather the fact that errors were occurring in data
input that were not being caught because there were not proper
data verification procedures going on or some type of way to
determine, in fact, that the data that were being entered
manually were, in fact, correct. I'll speak to that a little bit
in a minute.
Documentation was very sketchy. In a number of cases,
people couldn't tell what version of the software was used to
create what data set. They didn't even have the version of
software around. A sad story on this to illustrate how poignant
this situation can be, one of our sister agencies, whose name I
shall not mention but who deals a lot in outer space, has lost
some one and a half billion dollars worth of data because, while
the data are there in resident on computer tapes, the software
-------
603
that was used to create it is gone and they can't read it. They
have no way of knowing what the data are. They are lost for
future use. So, what I'm leading up to is that private industry
is not alone in some of these problems.
There were several other problems, too, that I will not
go into now.
When we looked at financial systems, we found a number
of things in place that we thought were lessons there to be had
to be put into the automated laboratory environment.
First of all, they always do a security risk analysis.
Whenever a company puts in...typically, a large company puts in
an automated financial system, they evaluate where their security
risks are, where their problems are, where there may be some
reasons to suspect that there could be some corruption or
misrepresentation of dollars since they're coming back and forth
through transmission lines and they evaluate that and put
together a portfolio of where all the different problems are.
Secondly, they implement... Based on that report, they
implement what's called a risk management program and some of the
things that typically go into a risk management program include
system access management programs where tellers come and go from
the organization. Their passwords are automatically taken out.
People move on to other jobs and their passwords are changed and
so on and so forth. Depending on the level and the type of
person, they have access to further levels of security within the
system.
There are verification procedures, typically double
entry. Somebody will enter the data from a strip chart or
something like that or it comes off of a roller from a register
and they'll enter it and somebody will come in behind and re-
enter it to double check it or they may have double verification
going on at the same time. The point is, anytime data are
entered manually, there always is some type of verification
procedure in place.
Third, audit trails. Whenever data are changed in a
-------
604
computer system in the financial arena, there is a trail to
indicate, number one, that data have been changed; number two,
who changed it; number three, when they changed it and the reason
why they changed it, and the original data are still there to be
able to see. This is an important point because one of the
things we found out in our laboratory visits and also through the
survey were a number of people had been sold what they thought
was an audit trail that really amounts to a transaction log.
What I mean by that, without getting into a lot of detail, is
data are dumped onto another tape and the new data replaces it so
that they don't know per se that the data have been changed and
there is no direct link between the new data and the original
data to satisfy the conditions that I just mentioned. Which
reminds me, what's the difference between a... Pardon me,
anybody, if I hurt anybody's feelings here, but what's the
difference between a computer software salesman and a used car
salesman? The used car salesman knows when he's prevaricating.
Finally, hard copy retention. This is an important
point that I found out kind of oddly enough through the course of
this. You know the little receipt you get when you do a mini-
bank transaction? That is your official legal copy of your
transaction and if anything happens or if any error occurs with
that, that is your actual legal tender to prove anything about
that transaction. I've been throwing them away. I started
keeping them after I found this out.
When we looked at hardware and software technology, a.
couple of things: One, there were no breakthroughs in hardware
or software to guarantee data integrity. Nobody had come along
with a great software system or nobody had come up with an
optimum hardware fix to be able to ensure integrity in automated
laboratories and this kind of surprised me. There were no
established software standards to ensure data integrity. The
business I just mentioned audit trail, there was never any
official proclamation from any organization or the other that,
here, these are the conditions that constitute an acceptable
-------
605
audit trail and in order to be able to be selling this as an
audit trail, you should follow these conditions. There isn't
such a thing in place. But, however, software can be customized
much like the software salesman I mentioned. They will sell you
about anything you want. They can do it about any way you want.
The only thing is, let the buyer beware.
Audit trails can be provided. They are now being
provided in some cases, I think, as a large result and password
protection is pretty well standardized.
There are some interesting hardware advances that some
of you in your laboratories may be already moving to: Optical
scanners, such as you have in the grocery store. They're now
being used in some labs. Magnetic ink readers, such as that kind
that are on your checks. You know, your checks are now all
standardized across the country. That's why when you write a
check in California, by the time you get back home, you may find
to your surprise it's already cleared your account. Electronics
is a marvelous thing when it comes to taking money away from you.
Finally, smart cards. This is something that is just
now coming along. We all have those magnetic strips on the back
of our credit cards and all. Well, technology is moving in such
a way now that you virtually can put almost an entire person's
life history on the back of it. Some hospitals are doing this,
for example, for patient history in certain cases and the patient
is carrying it around. This also has incredible application in
the laboratory. It's not really a forefront technology as it
sounds like. It's on its way in and we found some innovative
laboratories moving to use it, too.
With any of these things I might mention, what you're
doing is reducing the need for verification because you're
reducing the need for manual entry. You're enhancing the
reliability of the electronic environment. But again, as I'll
mention, there does need some caution with this.
Finally, when we looked at our own good laboratory
practices requirements that are in place in a number of offices,
-------
606
we found several things out. One, you can use whatever data you
want...whatever median you want in a laboratory and call it raw
data with the one proviso that wherever the data are first
recorded, that ergo is the median for the raw data. Without
getting into a lot of legalese and stuff about this, some of
which I don't even know, if, for example, you're typically
recording data and you're reading it off of an instrument and you
write it down to a piece of paper, that instrument isn't actually
recording it, itself, but that piece of paper is the raw data.
When you go to enter it into the computer system, that data are
not considered the raw data, but the original data written on the
piece of paper are and they typically have to be retained under
the conditions of our GLP.
Secondly, wherever a computer is used during the course
of a study, it also is part of and subject to the conditions of
good laboratory practices. Anytime a computer is used in the
course of data generation in a laboratory, it must also follow
the things that are required of other instruments through the
good laboratory practices, such as maintenance requirements and a
log if the system goes down or problems occur and what corrective
actions were taken and so on and so forth. I'll speak about
those a little bit in a minute.
Finally, documentation requirements of personnel
qualifications for the computer system are also, therefore, under
the aegis of the GLP.
Oversight of a quality assurance unit... Many of you
have had the quality assurance people, I'm sure, up and down your
backs. I'm sure you are all good buddies by now.
And SOPs must be in place.
When we started looking at our own requirements that
were underway, there were several things we found. One was
there's an electronic transmission standard whereby people can
electronically transfer data to the agency directly. That needed
to be factored into this and our own information resource
management policy that we lean heavily on that I'll show you in a
-------
607
minute.
Finally, there were a number of things under
development. The Federal Electronic Reporting Standards, Federal
Computer Security Act and our own System Design Developed
Guidance, all of those had to be factored into this, too.
So, what we ended up with through all of this was
clearly... It looked like there was a need for the agency to
come out with a standard set of good automated laboratory
practices and as a companion to that, implementation guidance for
how people can successfully implement any one of the practices
that we would put in and, finally, guidance for laboratory
auditors to determine if a lab is in compliance.
So, we published December 28th and actually distributed
February 13th The "Good Automated Laboratories Practises"- GALP.
There were some very important issues that arose between the
December 28th date and the February 13th distribution date. One
of them was, is it appropriate to put acknowledgments in the
document? We spent about a week discussing this issue. There is
a quote in the document by Francis Bacon. Another week and a
half passed discussing whether or not it was proper to put a
quote in a government publication. And then there was about a
three or four day discussion about what it meant. People
couldn't seem to grasp it. At any rate, nonetheless, it went out
on the 13th of February.
All of these things I've mentioned...all of these
things come into it. In a nutshell, its got two basic elements
that it relies heavily on: Good laboratory practices and our IRM
policy that relates to the development of hardware and software
and implementation thereof, stress testing and so on and so
forth, some of which I'll go through in a minute, and then
several other statutes come into play and are brought into it, as
well.
National Archives and Records Administration is
mentioned up here. They have guidelines out for dealing with
such things as...if you're data are on magnetic tape, how often
-------
608
should that magnetic tape be refreshed? If it's on optical
disks, how often and so on and so forth... Those requirements
are all effected in here, too. There are statutory requirements
from each one of the other federal EPA statutes that we deal with
that are referenced through in here and must be dealt with...the
Computer Security Act and then a couple of other things, as well.
What I'm going to do real quickly here is just run
through the different areas that the GALP speaks to and I'll show
you...and then I'm going to focus in on one of them in
particular... the data change requirement, to show you basically
how and what's going on in the document.
There are several areas and within each one of these
areas, there are several different subgroups of specifications
that are stated.
Personnel requirements. Those things relate to such
things as the documentation of the training of the personnel.
Are the personnel adequate? Are there sufficient personnel to
handle the computer and stuff like that.
Laboratory management responsibilities that they have
are under the aegis of the GALP.
Responsible person. He's a character who can be
anybody in the organization chain, including management, but he
is somebody who has certain very specific charters within the
GALP and would be able to sort things out and identify them. So
he's the guy who you kind of point to, to be able to say he's the
one who is responsible, typically like a system manager or
something like that...has those kinds of roles.
Quality assurance unit. Their responsibilities are not
only relative to the laboratory and its manual operation, but
also certain things when automating the laboratory. There are
about eight or 10 different specifications for the quality
assurance unit.
Facilities. Certain requirements on the facilities to
assure that the facilities are adequate to accomplish the need of
the automated laboratory.
-------
609
Equipment Requirements.
Security. There are four or five different areas of
security standard operating procedures. It specifies about six
or eight different standard operating procedures that must be
maintained as you automate a laboratory. In doing this, for
example, and also consistent with the software requirements, if
they were in place, we probably wouldn't have this problem with
about three and a half billion dollars worth of data sitting
around unable to know what to do with.
Date entry requirements a set of procedures, mostly
surrounding data verification and the like and transmission of
data, documentation, and who entered the data and so on and so
forth.
Raw data requirements. We could spend an entire career
dealing with the whole idea of raw data. We try and deal with it
in two pages in the GALP.
The bottom line, interestingly enough without getting
into a whole lot of detail on it, probably is a legal issue in
the long run. There was a recent court decision where Dalkon
Shield was being sued by somebody because apparently something
didn't work right and the company claimed that they kept the raw
data consistent with the federal FDA requirements and that they
threw it away after the requirement mandate had expired and they
were using that as justification in their case for not having to
be prosecuted any further and the judge threw that out and said
the FDA requirements did not supersede common sense in certain
situations. So what this kind of means, unfortunately for a lot
of the industry, is they're basically keeping everything as long
as they can. I don't know what to say about that except to say
that if it wasn't on paper and it was on an electronic media,
there might be a little more room in the house for other things.
Records and archives. That deals with record
retentions and the like and archiving of software and hardware,
as well...software and hardware maintenance records and so on and
so forth.
-------
610
Reporting requirements and finally the last thing,
comprehensive ongoing testing that periodically...and again this
is up to the lab, but probably no more than every 24 months, but
there should be a periodic stress test done of the system or, iC
there are any changes in the hardware or software configuration
of the system in the interim, there should be a stress test done
following those changes. But at the minimum, certainly no more
than 24 months, regardless.
What I'd like to do now is kind of focus in on one area
in particular, the data entry requirement which I mentioned
before. This is the one in particular.
These are the two different groupings under the data
entry requirements: Integrity, the tracking person, the tracking
equipment, data change and data verification. So this is one of
the 13 different elements. There's a total of 82 sub-elements
within the GALP so this one particular element has four or five.
What I'd like to do now is just focus in on the data
change requirement. I think that does it for that...yes.
This is a specific element from the GALP, the data
change requirement. I'll read it for those of you that can't see
it in the back of the room.
When the laboratory uses an automated data
collection system in the conduct of a study, the
laboratory shall ensure integrity of the computer-
resident data collected, analyzed, or maintained on the
system. The laboratory shall ensure that in automated
data collection system:
1. The individual responsible for the
direct data input shall be identified at the time of
data input.
2. That the instrument transmitting
data to the automated data collection system shall be
identified and the time shall be documented.
-------
611
And three, the one that seems to have given a
lot of trouble over the years:
3. Any change in automated data entry
shall be made so as not to obscure the original entry,
shall indicate the reason for the change, shall be
dated, and shall identify the person responsible for
the change.
In addition to identifying these requirements in the
GALP, we also provide guidance in the GALP on how to implement
each requirement.
The implementation guidance is set up much like this
for each one of the 82 standards and in a nutshell up in the
upper left corner, there is an icon directing your attention to
the particular category that it sits in, the fan for the
facilities, the responsible person with a little RP on his or her
chest, a checklist for standard operating procedures and the
like. I got into that because I have a Mac and Macs are heavily
icon-oriented and I thought it would be a good idea.
Finally, then the specification...the exact statement
from the GALP about what the requirement is or what the specific
standard is.
Then there is an explanation of what it means...what
it's all about, an example of how one might be in compliance with
the specification. And then there are codes that relate to two
things. One, the person...the operational role. There are six
operational roles. Without getting into a lot of detail, there
are six operational roles and it relates to who probably should
be in charge of that specific...this specific GALP entry and the
principle...the underlying principle behind why this standard was
put out in the first place. There are six of those basically.
-------
612
And any special considerations... The thing I just mentioned
before about the Dalkon Shield thing in the raw data definition,
that thing was elucidated and brought out and spelled out so as
to give the reader, when they get through with this thing, a full
feel for the spectrum of issues and ideas and concepts that are
going into each one of these things.
And finally, there is a notes section where you can put
in any other additional information or reference other things
that you see fit.
Let me show you how this works for one in particular.
Here is our little beast, the data change requirement. Do you
see the icon in the top left? We chose something that attemps to
look like a PC.
Next a restatement of the GALP requirement followed by
an explanation of the need for this requirement.
The data change requirement is nothing new. This has
been in place for many years in accounting. It's been in the FDA
Good Laboratory Practices since the '70s and it was considered
and drafted into the EPA draft GLP back in the '80s. It's
interesting that a lot of software houses somehow or another
decided that their data change specifications probably could be
different from this. I have no other explanation for that.
Next we have an example of how to comply with the
principle. The codes down here show the responsible person. In
this case, the responsible person is suggested to oversee data
changes and probably be the one to do it. Next we have the
underlying principle behind this requirement.
Finally, here's a special consideration in this case.
Laboratories may consider adopting a policy by which one
-------
613
individual may be authorized to changed data rather than
implementing a system that records the name of any and all
individuals making data changes.
For example, in this case, the responsible person would
be the only one authorized to make a data change and you might
have the software set up in such a way as to do that.
And here... This is from the manual, as well. It's
kind of a pictorial of what...one picture is worth 1,000 words,
so the picture had better be good. We worked on this one for
awhile. I think it kind of tells the story of what you mean by a
successful data change.
And then this notes section at the bottom here is in
each one of the GALP guidance areas to allow people to be able to
go back to and find other information relative to it to be able
to determine if they want more information or need to have a
further understanding of what's going on.
In most cases, this audit trail requirement is
mentioned in all of the background documents that were prepared
that I mentioned earlier because it's something that comes out of
the automated financial systems. It's something that comes out
of the clinical laboratory area. It's something that comes out
of a lot of different places. It does not come out of chemical
laboratories right now for some reason.
Finally, in a nutshell, none of this is new. Most
everything that I mentioned here is already requirements that
have been in place for a number of years in different places.
They just never were put together.
In the case of facilities and all the various statutes
that have been in place, they just never were all brought
together...and so on and so forth for those areas and procedure
software documentation and the like...and things related to
-------
614
software backup and recovery and all of that. All of these
things have been around for some time. There's really nothing
new in it.
It looks like we have time for one or two questions.
-------
615
QUESTION AND ANSWER SESSION
MR. TELLIARD: Any questions?
MS. ASHCRAFT: Merrill Ashcraft
from Navy Public Works Center.
Is there a documentation number? Is this document
available?
MR. JOHNSON: Yes, if you'll write
to me or call me. My phone number and address are all on the
speaker list in the book. I'll be glad to get one to you.
Unfortunately, there aren't any now. There will be some by the
middle of next week. We ran out. We had 500 made. We only had
250 people on the original mailing list and now they're all gone.
But we'll have more and I'll be glad to make you a copy...have a
copy for you.
MR. TELLIARD: Rick, is this
intended to become a rule?
MR. JOHNSON: No, let me explain,
too. Thanks, Bill. These are recommendations from our office to
EPA programs and to all people developing data to provide to the
agency. They are not requirements that are being laid on
anybody. It is going to be up to each one of the program offices
within the agency to decide how and what they want to do with
them and how they want to implement them. Part of the reason for
that is... One of the biggest reasons for that is I'd like to
get onto working on something else within the next 10 years and
the second reason is that each one of the different programs has
different constraints that... They may, in fact, want to go
beyond this or they may already have this and some of these
elements in place in their individual programs. So rather than
pick and sort through all of that and try and come down to a core
minimum set of requirements, we put these out as recommendations
to the programs and to folks doing business with the agency in
the hopes that each one of them would come to grips with one or
the other of them and deal with them accordingly. For example,
the Contract Lab Program, Joan, you might just mention very
-------
616
quickly where you all stand with it.
MR. TELLIARD: Joanie, we're
running late.
MS. FISK: You know I'm a man of
few words. As far as the Contract Lab Program is concerned, the
CLP is very anxiously awaiting this document. They are aware of
it because Rick has spoken at our meetings in the past and our
perspective right now is some of the things we consider
absolutely critical, we are going to put into the contracts and
have in some cases. We have included a lot of data management
requirements in personnel and, as we learn more, we certainly
intend to do more. Don't think we would ever make it a
contractual requirement that this document an actual piece, but.
we would certainly intend our laboratories to use it. But we
would expect them to use it as guidance. They may find other
ways that work better for them to implement it and we would want
them to monitor that, track it, write SOPs, the whole bit.
The other thing that we did want to do is we've talked
to Rick about helping us in setting up better audit procedures so
that we can learn more about the people's automated systems up
front.
MR. JOHNSON: One of the things I
might mention along that line of personnel procedures, we have
put into the 1992 budget... We have put together a planning
program for personnel, a training program or some sort of a
program ...
MR. TELLIARD: Thank you.
-------
617
14th Annual
EPA Conference on
Analysis of Pollutants
in the Environment
May 8-9,1991
GOOD AUTOMATED LABORATORY PRACTICES
EPA'S RECOMMENDATIONS FOR ENSURING DATA INTEGRITY
IN AUTOMATED LABORATORY OPERATIONS
WITH IMPLEMENTATION GUIDANCE
-------
618
O 3 C
all
s so
ft o
0)
-------
619
-------
620
t/3
c
po
13
I*
OX)
0>
(0
(0
« 2P
«1
-^ O
cc c
O)
D£
a) ;
DC
O
••=? fi
fl
eo
oj c^
.-is
O)
s
o
L.
-------
621
OPTS
ADMINISTRATOR
OPPE
Scientific
Systems Staff
OIRM*
OTHERS
* Office of Information Resources Management
-------
622
S
Q,
O
C/}
^ w
A
s
o
* v
S
o 5
o
DJD -t5
c "S
O)
CJ
c
C/)
PH
C
p
o
-------
623
0)
D
O
3
O
O
o
13
o
CO
in
O
^
a
DJD
^^"^^
.2
c
c
"O
amne
03
§CLJ
O
13
o
(3
'3
-------
624
(0
o
o
Pu,
>,
L*
O
ts
&*
o
O
O
N
c
c
O)
9
o<
0)
t<
OD
tfi
en
"H
0)
03
VN
O
a
JM
O
a
C
OD
a
S
c
a>
s
C3
L*
H
-------
625
(0
£
O
c
O
O
O
"5
O
&
(0
0)
^^^^^
O
O
C3
£* o
3
o
c»
13
a>
o>
s£
c»
o
a
c
o
•PN
c
-------
626
0)
E
0)
"S
>»
en
75
'o
c
(0
c
iZ
£
OB
E
o
3
g
LN
.O
a
VI
03
fl
I*
9
u
OJ
ented
^Oj
"a
S
DJD
O
IN
OH
"S
a>
s
<3J
OD
CQ
C
S
a
JM
OD
o
b
PL,
'S
(L>
g
CJ
OD
03
C
03
CJ
CJ
w
O
U
PH
C
O
fica
c
o
"3
g 8 ^
K IS **
-------
627
O)
o
o
0)
0)
c5
o
c/>
c
£
OX)
"3
03
"O
c
03
03
DJD
O
L-
03
-O
C
03
-C
OJD
O
JC
are s
soft
a;
03
N
•P^
O
^^
O5
0
O
^ •=
'3
IN
C
03
o>
LN
e
e
o
O5
Protectio
*- O
-------
628
D)
O
O
c
O
O
CD
(5
03
c
C3
c
e
a
O
.a u
G
OD
03
03
E
-------
629
•o
o
o w
o*
0)
= b
82
TS
-------
630
O CD
0.0
(0
o o
OS
08
£
o
"3
CQ
C
2?
•-
o>
5
PM 5
O ^
5/3
c
o
• pp«
IS
^-^
a
0->
s
3
u
O
Q
s
a
CM
O
Sf
S
a>
•V
tSl
ON
O
-------
631
0)
E
o>
DC
O)
(0
LU
•S3
•a
e
5>
IP
S
I
tu
-------
632
c
0)
Q.
_O
0)
•»•*
K
i
o
o>
O
2
•8
•8
•o
C
3
c
o
e
o
si
r «
O)
s
C
D
§
OX)
s evo
nce
EPA
G
**
2
0)
cc
-------
633
1
!•
wm
D
5
3.
5
b
E
2
D)
O
Q->
O
C
A
r/5 Prt
** ^^
3J «M
•t 5
c c
^ ^^^ ^^
C 2
£ *3
^ «
«l-4 ^^
^
0)
^^
'oS
"2
'3
O
0)
o
Q.
O
"(0
o
O>
a
s
c
a
C = a
^™ /i \ ^^
a
S
o
U
-------
634
United States
Environmental Protection
Agency
Office of Administration
and Resources Management
Draft
December 28,1990
EPA Good Automated
Laboratory Practices
Recommendations For Ensuring
Data Integrity In Automated
Laboratory Operations
with Implementation Guidance
-------
635
Computer Security
Act of 1987
Statutory Requirements
for Environmental
Programs:
• Superfund
Resource Conservation
and Recovery Act
Clean Water Act
Safe Drinking Water Act
Others
GALP
EPA IRM Policy:
EPA System Design and
Development Guidance
EPA'S Operations and
Maintenance Manual
EPA Information Security Manual
EPA'S Data Standards for Electronic
Transmission of Laboratory
Measurement Results
Findings of EPA's Electronic
Reporting Standards Work Group
National Archives and Records
Administration's Electronic
Records Management
Regulations
EPA's Good Laboratory
Practices
Federal Insecticide,
Fungicide, Rodenticide
Act GLP (40 CFR Part 160,
August 1989)
Toxic Substances Control
Act GLP (40 CFR Part 792,
August 17, 1989)
-------
636
APPENDIX A: INVENTORY OF COMPLIANCE DOCUMENTATION
RECORD PURPOSE SUBSECTION REFERENCE
Organization and Personnel i
Personnel Records
Quality Assurance
Inspection Reports
Ensure competency of
personnel
Ensure QA oversight
7.1
7.4
FIFRA GLPs 160.29
TSCA GLPs 729.29
FIFRA GLPs 160.35
TSCA GLPs 792.35
Facility , :
Environmental
Specifications
Ensure against data loss
from environmental threat
7.5
FIFRA GLPs 160.43
TSCA GLPs 792.43
Equipment
Hardware Description
Acceptance Testing
Maintenance Records
Identify hardware in use
Ensure operational
integrity of hardware
Insure on-going operational
integrity of hardware
7.6
7.12
7.6
7.12
7.6
7.12
FIFRA GLPs 160.61
TSCA GLPs 792.61
EPA Information Security
Manual for Personal
Computers
System Design and
Development Guidance
FIFRA GLPs 160.63
TSCA GLPs 792.63
Laboratory Operations
Security Risk
Assessment
Standard Operating
Procedures
• Security Procedures
• Raw Data
Definition
Identify security risks
Ensure consistent use of system
Ensure data integrity secured
Define "computer-resident'
records subject to GLPs
7.7
7.8
7.8
7.8
Computer Security Act
FIFRA GLPs 160.81
TSCA GLPS 792.81
Computer Security Act
FIFRA GLPs 160.3
TSCA GLPs 792.3
-------
637
APPENDIX A: INVENTORY OF COMPLIANCE DOCUMENTATION
RECORD
PURPOSE
SUBSECTION
REFERENCE
Procedures for data
analysis, processing
Procedures for data
storage and retrieval
Procedures for
backup/recovery
Procedures for main-
tenance of computer
system hardware
Standard Operating
Procedures
• Procedures for
Electronic Reporting
• SOPs at bench/
workstation
• Historical Files
Ensure consistent use of system
Ensure consistent use of system
Ensure consistent use of system
Ensure consistent use of system
Ensure consistent use of system
Ensure consistent use of system
provide historical record of
previous procedures in use
7.8
7.8
7.8
7.8
7.8
7.8
7.8
FIFRAGLPs 160.87,160.107
TSCAGLPs 792.81. 792.107
FIFRAGLPs 160.81
TSCAGLPs 792.81
EPA Information Security
Manual for Personal
Computers
FIFRAGLPs 160.63
TSCAGLPs 792.63
Transmissions Standards
Electronic Reporting
Standards Workgroup
FIFRAGLPs 160.81 (c)
TSCAGLPs 792.81 (c)
FIFRAGLPs 160.81 (d)
TSCAGLPs 792.81 (d)
Software Documentation
Description
Life Cycle Documentation
Design Document/
Functional
Specifications
Identify software in use
Ensure operational integrity
of software
Ensure operational integrity
of software
7.9
7.9
7.9
FIFRAGLPs 160.81
TSCAGLPs 792.81
Computer Security Act
System Design and
Development Guidance
see above
-------
638
APPENDIX A: INVENTORY OF COMPLIANCE DOCUMENTATION
RECORD
PURPOSE
SUBSECTION
REFERENCE
Life Cycle
Documentation
AcceptanceTesting
Testing
Change Control
Procedures
Procedures for
Reporting/Resolving
Software Problems
Historical File
(version numbers)
Ensure operational integrity
of software
Ensure operational integrity
of software
Ensure operational integrity
of software
Ensure reconstruction of
reported data
7.9
7.9
7,9
7.9
EPA Information Security
Manual for Personal
Computers
see above
see above
see above
FIFRAGLPs 160.81
TSCAGLPs 792.81
Operations Records/Logs
Back-up/Recovery Logs
Software Acceptance
Test Record
Software Maintenance
(Change Control) Records
Protection from data loss
Ensure operational integrity
of software
Ensure on-going integrity
of software
7.12
7.12
7.12
EPA Information Security
Manual for Personal
Computers
System Design and
Development Guidance
see above
-------
639
Q.
<
0)
c
c
o
(0
<5
Q.
C
(1)
0)
D)
CO
c
CO
o
s
o
^
(0
CM
O
(0
0)
Q.
.O
"55
c
o
Q.
(0
O
DC
CO
•
h-
•E
Ol
CD
-------
640
Q.
<
CD
o
c
CO
CO
CO
o
CO
-------
641
0)
3
Q.
o
•o
c5
•o
c
CO
CO
00
co
O)
CD
-------
642
CL
0)
o
<
<
o
LII
03
Q
GS
"53
O
CO
DC
c
(0
0)
o
o
O)
DC
CM
-------
643
CL
<
0)
c
O)
c
"5
o
c
O
0)
"55
c
0
-------
644
Q
CL
<
C
W
Q
CM
o
c
o
ox
fl
OJD
Q JJ
OX)
0>
<-*
c
^
Q
c
0o
15
u
£
*n
4J
<^
•
t>
-------
645
GALP IMPLEMENTATION GUIDANCE
This section is intended as a key to using die Guidance. The model below, with commentary
footnotes, illustrates the implementation guidelines provided for each of the standards.
GALP Category Name
GALP subsection
Icon depicting the
GALP category
Specific and officially approved wording of the particular GALP standards.
ID cases where a GALP has general specifications with distinct subsections or subspeci-
fications, the general specification will always appear with each subspecification with two or
three pages of discussion of that subspecification; the next subspecification win repeat the
general specification, and follow with its dis
r-Jjg^-T»-rffcr'.»^.-*w*w-^j-,
•EXAMPLE
jj ,^m., ,, .,. f*
CONSIDERATIONS
A paragraph exposition defining the key terms of the standards and
explaining the intent of the standards.
Discusses the kind of compliance evidence that might be gathered or
acceptable ways in which the standards has been or may be met
Twocodesareprovided: theRESPONSIBILITYcodeidentifyingtherole
(or persons(s) assigned the role) expected to implement the standards; and
the PRINCIPLES code; providing general guidance into the theoretical
intent of the standard.
Provides potentially relevant facts or noteworthy factors that may be
relevant for certain laboratory settings, computer equipment, EPA
statutes, or court decisions that may take precedence.
NOTES: The GALP Guidance is a working document. An area on the right-hand page is provided
to allow annotation as needed. The size of this area is determined by the space available to
complete a page. This variation is not meant to imply any difference in the extent of comment
anticipated. Sources for additional guidance are also listed here.
-------
646
7.10 Data Entry
1) Integrity of Data
3) Data Change
When a laboratory uses an automated data collection system in the conduct of a
study, the laboratory shall ensure integrity of the computer-resident data col-
lected, analyzed, processed, or maintained on the system. The laboratory shall
ensure that in automated data collection systems:
3) Any change in automated data entries shall not obscure the original entry,
shall indicate the reason for change, shall be dated, and shall identify the individual
making the change.
EXPLANATION
EXAMPLE
When data in the system is changed after initial entry, an audit trail
must exist which indicates the new value entered, the old value, a
reason for change, date of change, and person who entered the
change.
This normally requires storing all the values needed in the record
changed or an audit trail file and keeping them permanently so that
the history of any data record can always be reconstructed. Audit
Trail reports may be required and, if any electronic data is purged,
the reports may have to be kept permanently on microfiche or
microfilm.
CODE
Responsibility:
Principle:
Responsible Person
3. Audit
SPECIAL
CONSIDERATIONS
Laboratories may consider adopting the policy by which only one
individual may be authorized to change data, rather than implement-
ing a system that records the name of any and all individuals making
data changes.
184
-------
647
7.10 Data Entry
1) Integrity of Data
3) Data Change
AUDIT TRAIL
• NAME OF PERSON
ENTERING DATA
• DATE OF ENTRY
i
- - - - - 134,y
CHANGE PROCESS
ORIGINAL
DATA
• NAME OF PERSON
MAKING CHANGE
• DATE OF CHANGE
• REASON FOR CHANGE
CHANGED
DATA
Notes...
For additional guidance, see: FIFRA GLPs 40CFR 792.130(e); TSCA GLPs 40CFR
160.130(e); Automated Laboratory Standards: Evaluation of Good Laboratory
Practices for EPA Programs, Draft (June 1990); Automated Laboratory Standards:
Evaluation of the Standards and Procedures Used in Automated Clinical
Laboratories, Draft (May 1990); and Automated Laboratory Standards: Evaluation
of the Use of Automated Financial System Procedures (June 1990).
-------
648
MR. TELLIARD: Our next speaker
probably doesn't need any introduction. It's George Stanko from
Shell Development. George has probably the dubious honor of
having been to as many of these conferences as I have. In fact,
I think you outrank me by one.
George is going to talk about one of his
favorite subjects: Contract laboratory performance. George hcis
taken this on as a personal trust and is going to save the world.
-------
649
PERFORMANCE EVALUATION STUDY
OF
ENVIRONMENTAL ANALYTICAL CONTRACT LABORATORIES
Author: G. H. Stanko
Shell Development Company
Houston, Texas
Presented at: 14TH Annual EPA Conference on Analysis
of Pollutants in the Environment
Norfolk, Virginia
May 8,9, 1991
-------
650
ABSTRACT
A performance evaluation (PE) study of environmental analytical contract
laboratories used by Shell was conducted. Spiked groundwater samples
were blindly submitted to 24 laboratories for analysis by 6C/MS Method
8240 for volatile organics, ICP metals, and selected general parameters.
Samples for the study were prepared by a contractor who also made
arrangements to have the samples analyzed. Sampling kits were obtained
from laboratories and returned for analysis. The results of the
performance evaluation study were used to assess the performance of
commercial laboratories from initial contacts; receiving and returning
sampling kits; analysis of samples; and through final reporting of data.
A second contractor was employed to statistically analyze the data and to
evaluate the results for the study. The study was also designed to gain
a better understanding for recovery correction of environmental data and
to determine the qualitative and quantitative performance of commercial
laboratories for non-target analytes commonly called "tentative
identified compounds" (TIC's). The results for the PE study are presented
in the paper.
-------
651
PERFORMANCE EVALUATION STUDY
OF
ENVIRONMENTAL ANALYTICAL CONTRACT LABORATORIES
INTRODUCTION
A performance evaluation study of 24 environmental analytical contract
laboratories was conducted in late 1990. To accomplish this, a contract
was negotiated with a third party to contact the laboratories; make
arrangements for receipt and shipment of sampling kits; prepare spiked
matrix samples for the study; and to compile the data and significant
information resulting from the study. The study was conducted with real
matrix samples and the samples were submitted to the commercial
laboratories without identifying them as performance evaluation samples.
PE studies conducted in such a manner reflect actual real-time
performance at commercial laboratories.
The current PE study was designed with one major goal and two secondary
goals. The major goal was to assess the performance of a select group of
laboratories for the analysis of groundwater samples for volatile
organics, metals, and a limited number of general parameters. The
secondary goals were to determine how well commercial laboratories can
identify and quantify non-target analytes by GC/NS and to gain a better
understanding for recovery correction of analytical data as proposed in
the draft version of the new Chapter 1 of SW-846. These non-target
analytes are commonly referred to as "tentative identified compounds" or
"TIC's". To minimize bias for the study's major and secondary goals, a
second firm was hired to statistically analyze and evaluate the resulting
data for the PE study and to summarize the findings in a report to Shell.
This paper includes the information and results reported by independent
contractors to Shell.
PERFORMANCE EVALUATION STUDY
Twenty-four laboratories plus Shell's Westhollow Research Center were
included in the PE study. Each of the laboratories were contacted by an
engineering firm where arrangements were made to have some groundwater
samples from a cleanup site analyzed for volatile organics and TIC's, oil
and grease, BOD, pH, COD, TOC and ICP metals. EPA Method 8240 was
specified for the organics and TIC's. ICP was specified for the metals
and EPA approved methods were specified for the general parameters. To a
limited extent, laboratories were given the impression that the
engineering firm was doing an investigation associated with a leaking
service station.
Laboratories were evaluated on their response and the help provided to
the customer and how knowledgeable they were with respect to the
customer's needs. Since Shell relies on contract engineering firms to do
a lot of this type work, the resulting information was important.
Sampling kits were requested and shipping instructions were obtained.
-------
652
The program was arranged so that all of the laboratories eventually
received their samples within one day of the same time.
The groundwater used to prepare the PE samples was a real groundwater
matrix collected from a site with well characterized groundwater. The PE
samples were prepared to simulate the kind of background matrix one might
expect from a service station with leaking storage tanks. Some target
analytes of interest were also present in groundwater and this was
accounted for in the theoretical concentrations or made-to values. Three
samples for organics were prepared with varying concentrations of nine
target analytes and nine TIC's. The target analyte concentrations ranged
from 5 to 68 ppb and the TIC's concentrations ranged from 15 to 270 ppb.
The target analytes were prepared so two of the samples represented
duplicates and the third was 2X the duplicates concentrations. These
were randomly done with the three samples. The TIC's were prepared at
IX, 3X, and 6X concentration levels and these were also randomly done.
The volatile organics and TIC's samples were designated MW-1, MW-2, and
MW-3.
MW-4 and MW-5 were designated for ICP metals and a limited number of
general parameters. Six of the target metals were present in the
groundwater matrix. The background levels were accounted for in the
made-to values for these analytes. A total of 11 metals were designated
as target analytes for the study. These two samples were prepared to
represent Youden pairs to assess laboratory performance
MW-4 and MW-5 were also designated for the general parameters. Separate
containers were used for metals, BOD and pH, TOC and COD, and oil and
grease since different preservation steps were involved. The oil and
grease samples were prepared by delivering a known quantity of oil and
grease into the sample bottles provided by each laboratory and diluting
to a known volume. Slight variations in the amount of oil and grease put
into each container made the sample concentrations slightly different for
each laboratory.
Table 1 is a tabulation for the made-to values for all the performance
evaluation study samples. Essentially all of the samples were prepared
to minimize any possible bias on the part of the laboratory and to
disguise the samples from being associated with a PE study. It would
have been difficult for anyone to establish any kind of pattern or that
these were PE samples. All coolers were shipped to the laboratories by
Federal Express on the same day.
LABORATORY INTERACTION
Records were maintained for all laboratory interaction between the
engineering firm and each laboratory in the study. Such items as how
helpful and knowledgeable the contacts at the laboratories were with
respect to sampling requirements and analytical methodology were
documented. The adequacy of the sampling kits, cost of analyses,
turn-around time, and the degree of difficulty that was experienced in
-------
653
trying to obtain missing Information and verify some results were all
documented. Most of this information was included in the individual
laboratory evaluation reports provided by the contractor. These
individual laboratory evaluation reports have since been sent to the
laboratories along with the raw data reports for their samples. All this
was done as part of the quality improvement process and has allowed
laboratories to initiate corrective action where appropriate prior to
complete statistical evaluation of the entire database.
Sampling Kits
There was a wide range of sampling kits that were provided by the
laboratories. These ranged from a cardboard box containing bottles to
well designed ice chests with labeled containers and blue ice included.
Many of the bottles contained acid preservatives and some of these had
leaked during shipment. It was obvious that there was a lot of room for
improvement at many of the laboratories. One cannot expect precise and
accurate data for samples that are collected and shipped with sampling
kits that are inadequate.
Cost of Analysis
Laboratory pricing practices varied considerably depending on whether you
wanted hard copies of the data, supporting QA/QC, and the raw data. Some
laboratories were quite willing to provide CLP data packages for the
analyses and based their bid on that basis. Other charges were quite
confusing as well as the final billing. In one instance we were not
billed for all analyses. There was approximately a 2X range of costs
from the laboratories in the study. Because of the complex nature of the
pricing practices it was not possible to estimate some average cost.
Turn-around Time
It was discovered that there are at least two major components for
turn-around time problems. The first of these is that of holding times
specified for most of the analyses. Only two laboratories did not meet
the 10-day CLP holding time for volatile organics but did analyze samples
within 13 days. The mean and median analysis times were 5 days for the
study. For metals, the mean and median analysis time were 17 and 14
days, respectively. Because of the short holding time for the general
parameters, no delays in reporting data resulted from these analyses.
Reviewing the raw data for when the data were available at the laboratory
and when the data were actually sent to the customer, it was discovered
that on the average it takes 35 days for a laboratory to report the data
after the analyses are completed. This component of turn-around time
leaves a lot of room for improvement. Essentially, this indicated that
turn-around time delays are not caused by instrument limitations or
delays but by clerical and/or quality assurance review.
The fastest and slowest laboratories sent out their data in 15 and 58
days, respectively. Unfortunately, this did not complete the laboratory
-------
654
portion of the study. Five laboratories amended their reports to Include
forgotten data -- metals, raw data packages, and volatile organic TIC's
data. These five reports were not completed from 51 to 103 days after
the laboratories had received the samples. It should be noted that these
five laboratories did not know their reports were not complete until they
were contacted -- In some cases several times. Some information was
never obtained after many attempts so a closing date was set for the
study.
LABORATORY PERFORMANCE
There are a number of acceptable ways to assess laboratory performance.
Two of the common parameters are precision and accuracy (recovery). EPA
has also included precision and accuracy criteria in many of their
analytical methods published in SW-846 or 40 CFR Part 136. Youden plots
are another way to assess performance. Both common practices were
employed to assess laboratory performance.
Outliers
It was not necessary to employ statistical tests to identify specific
outliers. Simple visual comparisons and initial Youden calculations were
sufficient in finding the most extreme outliers. The outliers also
became quite obvious from the bar charts prepared for the data. Data
that were seen as outliers were omitted from some of the calculations
because of the obvious bias they would have given to the data group as a
whole. Where any data were omitted from calculations, outliers were
identified in the various tables and figures and usually enclosed in
boxes.
Accuracy and Precision
The discussion of accuracy and precision is presented by analyte group --
volatiles, metals, and general parameters. Within each group, the
discussion begins with Youden plots because they give a quick visual
review of the data. A scorecard of laboratory performance was then made
from the initial conclusions from these plots. Subsequently, tables and
figures were prepared to give more details for laboratory performance.
It was noted that these tables and figures were entirely consistent with
the initial conclusions from the Youden plots.
Volatiles
The Youden plots for the nine target volatile analytes are shown in
Figures 1 and 2. Noted on each plot are the laboratory numbers that fall
outside the region of values where 95% of a single laboratory's results
are expected to fall (the circle). A scoring system was devised to
summarize the results from the Youden plots in Table 2. As can be seen
in Table 2, there are four possible areas where a laboratory may fall on
a Youden plot.
-------
655
Before the 95% circle was determined for each Youden plot, preliminary
areas were drawn using the data from all laboratories. Those that fell
well outside the 95% area were then designated as outliers and deleted
from the calculation of sample averages and intralaboratory precision
that are used to draw the 95% circles shown in Figures 1 and 2. Those
laboratory numbers that were considered outliers are enclosed within
boxes in the Youden plots.
Some laboratories reported observations for analytes as less-than values
and in some instances no values were reported. These reported
observations could not be used in the Youden plots. Where laboratories
reported less-than values in a sample pair, the symbol "X" was used in
Table 2 for the observation. Where a laboratory did not report any data
for an analyte, the letters "nr" were used for "not reported" in Table 2.
To get an overall picture of each laboratory's performance for the
volatiles as an analyte group, the number of times a laboratory fell into
a particular region were counted as well as the number of times a
laboratory reported less-than values. This information is shown in Table
2 as the "performance summary". The best possible score would be a value
of 9 for falling within the 95% circle for all analytes. Lab-15 and
Lab-23 did have scores of 9. Other laboratories that most often fell
within the 95% region were Lab-3, Lab-4, Lab-5, Lab-10, Lab-12, Lab-19,
and Lab-25.
Laboratories that had a tendency towards higher than average recoveries
were Lab-2, Lab-8, Lab-9, Lab-11, Lab-14, and Lab-17. Higher than
average recoveries do not necessarily mean poorer performance for the
volatiles. In most instances, the recoveries were actually closer to the
made-to values. The made-to values for each sample pair are shown in the
Youden plots in Figures 1 and 2. Judging from the position of the
made-to values shown in the Youden plots, only Lab-2 has a clear problem
with high recoveries that places it well outside the performance of the
other laboratories. As noted in the Youden plots, Lab-2 was identified
as an outlier in the group of laboratories used to calculate the 95%
performance region for five of the nine volatiles.
Laboratories that had a tendency towards lower than average recoveries
were Lab-1, Lab-16, Lab-20, Lab-21, and Lab-24. Lower than average
recoveries are indicators of poorer performance as shown by the positions
of these labs and the made-to values shown in the Youden plots. Some of
these laboratories were also identified as outliers in the group of
laboratories used to calculate the 95% performance regions. Boxes on the
Youden plot identify these outliers.
Table 3 summaries the recovery of volatiles for all the laboratories and
for each of the analytes. This is just another way to summarize the
accuracy data. The outliers identified in the Youden plots were not
deleted from the calculations of mean recoveries shown in Table 3 because
they had little effect on the mean values. Table 3 also summarizes the
mean recoveries for all nine analytes for each laboratory. In fact,
-------
656
Table 3 was rank ordered by mean recovery from lowest to the highest.
The mean recoveries ranged from 55% for Lab-20 to 117% for Lab-2.
Among the individual volatiles, the mean for all laboratories ranged from
59% for carbon tetrachloride to 105% for methylene chloride. The mean
recovery of 88% for trichloroethylene is somewhat misleading. Nearly
half of the laboratories could not detect trichloroethylene in the 5 ug/L
samples (MW-2 and MW-3) although all 25 laboratories reported a
detectable concentration for the 10 ug/L sample. The fact that most
laboratories did report less-than values but had no problem with the 10
ug/L sample suggested that the detection limit for trichloroethylene is
indeed somewhere above 5 ug/L. Given the number of less-than values
reported for trichloroethylene, the mean recovery for all laboratories
would be less than the 88% shown in Table 3.
Bar charts of mean recovery were prepared with the same data and these
are presented in Figures 3, 4, and 5. The bar charts were prepared with
all the data; however, outliers were removed for mean calculations. Both
presentations represent additional ways of assessing the same recovery
data. Another way of looking at the data in Table 3 is through range
charts that show the high, mean, and low mean recoveries by laboratory
and analyte. The range chart shown in Figure 6 shows the variability of
each laboratory in analyzing volatiles with the laboratories ranked by
overall mean recovery. The high value shown is the highest mean recovery
among the nine volatiles for each laboratory and the low is the lowest
mean recovery. The mean is the overall mean for all nine volatiles. In
judging laboratory performance, if the mean recovery for volatiles as a
group is acceptable, the laboratory with the narrowest performance range
indicates overall consistency for the analyte group. For example, Lab-3
has a mean recovery of 86% compared with 84% for the group average and
the range was 71 to 107% for all analytes.
Figure 7 is a range chart showing the variability in mean recoveries by
volatile analyte. The analytes were ranked in order of increasing mean
recovery starting with carbon tetrachloride at 59% and ending with
methylene chloride at 105%. Toluene showed the most variability between
high and low recovery which ranged from 61% for Lab-21 to 286% for
Lab-19. The next highest value for toluene recovery was 124% if one
considers the 286% value an outlier. The new range would be 61% to 124%
which is more in line with the other analyte recovery ranges. This
suggests the 286% recovery value for toluene is an outlier.
Intralaboratory precision was calculated for each laboratory from the
duplicate samples for each of the nine target volatiles. These precision
values were then compared with EPA expected single analyst precision
(intralaboratory precision), calculated from the method performance
equations found in Method 8240 from EPA's SW-846 Methods Manual . Table
4 is a summary of these comparisons. As noted in the table, the
comparison was not made for xylenes because there are no precision
equations for this analyte in the method. In the last column of Table 4,
the number of times that a laboratory met the EPA expected method
performance is summarized. A value of 8 indicates meeting EPA criteria
-------
657
for all analytes where comparison was possible. Laboratories that met
the method criteria for all eight volatiles were Lab-3, Lab-4, Lab-14,
Lab-15, and Lab-17. Other laboratories that did well were Lab-11,
Lab-12, Lab-23, and Lab-25 which met the criteria for seven of the eight
analytes. Laboratories that were on the poor performance end with
meeting EPA criteria for three or less out of the eight volatiles were
Lab-2, Lab-6, Lab-13, and Lab-21.
Interlaboratory precision was used to calculate the EPA expected recovery
range that a laboratory should fall within. Again, EPA performance
equations from SW-846 Method 8240 were used as the criteria. These
results for interlaboratory precision are summarized in Table 5 using the
same nomenclature as with Table 4. Again, because equations for xylenes
are not given in the method, no comparison was made for this analyte.
Only Lab-9 and Lab-14 met the recovery range criteria for all eight
volatiles. Other laboratories that did nearly as well were Lab-3, Lab-4,
Lab-5, Lab-8, and Lab-17 which met criteria for seven of eight analytes.
Only Lab-21 did poorly meeting the criteria for three or less volatiles.
It appears from the number of times that the laboratories failed to meet
the criteria for 1,1,1-trichloroethane, carbon tetrachloride, and
trichloroethylene, that the EPA's criteria are not representative of
general laboratory performance at the concentration levels used for this
study and may be too restrictive (although the study concentration were
within the applicable range of EPA's equations, 5-600 ug/L). Carbon
tetrachloride, 1,1,1-trichloroethane, and trichloroethylene had mean
recoveries of 59%, 69%, and 88%, respectively over all laboratories for
this study. The recovery equations for Method 8240 indicate recoveries
greater than 100% for all three of these analytes which the 24
laboratories could not meet.
Metals
Youden plots for the eleven metals that were spiked into MW-4 and MW-5
are shown in Figures 8 and 9. As with the volatiles, preliminary Youden
plots were made first to visually identify outliers. These outliers
again were then omitted from the calculations of the circular area in the
plots that represent where 95% of a single laboratory's results are
expected to fall. The outliers were identified in the Youden plots by
boxed-in laboratory number. All values for Lab-9 were omitted from the
calculations because it appeared that the laboratory had switched sample
numbers in reporting the results (MW-4 values were reported for MW-5 and
vice versa). It could not be established prior to the closing date that
this mistake had indeed been made. One exceptional outlier was the
sample pair of Lab-18 for sodium. The values of this pair were an order
of magnitude over the made-to values (1,800 and 1,700 mg/L reported for a
made-to value of 176 mg/L for both samples). It is likely that the
laboratory misplaced a decimal in calculating the sample concentration;
however, no information was obtained from the laboratory to refute the
reported values. The sodium data from Lab-18 were omitted from
calculations for the 95% region.
-------
658
A summary of the Youden plots for metals Is shown In Table 6 using the
same symbols and criteria described previously for volatile analytes.
Again, the performance summary column shows the overall picture of each
laboratory's performance for the metals. Lab-7, Lab-11, and Lab-14
results all fell within the 95% region for all metals. Other
laboratories that most often fell within the 95% region included Lab-2,
Lab-3, Lab-5, Lab-6, Lab-12, Lab-13, Lab-18, Lab-22 and Lab-24.
Laboratories that had a tendency towards higher than average recoveries
were Lab-10, Lab-20, Lab-21, and Lab-25. Laboratories with a tendency
towards lower than average recoveries included Lab-8, Lab-16, Lab-17, and
Lab-19. Lab-8 appeared to have a particular problem with recurring low
recoveries for 9 of the 11 metals. Laboratories that had a tendency
towards poorer than expected precision were Lab-1 and Lab-23.
Table 7 summarizes the mean recoveries for all analytes and laboratories.
Again, a summary of the mean recoveries for all metals is shown in the
last column of Table 7. The laboratories in Table 7 are ranked in
increasing mean recoveries for all metals. The overall mean recovery for
all eleven metals was 97%. Among the individual metals, the means for
all laboratories ranged from 89% for cadmium to 107% for zinc.
Recovery bar charts were prepared for each of the metals and all data
from all laboratories were included; however, the mean recoveries
calculated excluded the outliers. These bar charts are shown in Figures
10 and 11. This is just another way to illustrate outlier data that were
omitted from calculating the 95% circular regions of the Youden plots.
Range charts of the high, mean, and low mean recoveries by laboratory and
analyte are shown in Figures 12 and 13. These charts do not show the
range with the outliers included because they tend to distort the overall
view of the data. The range chart in Figure 12 shows the variability of
each laboratory in analyzing metals, with the laboratories ranked by
overall mean recovery. The high value shown is the highest mean recovery
among the eleven metals for each laboratory, the low is the lowest mean
recovery, and the mean is the overall mean for all eleven metals.
Many laboratories showed a fairly narrow range between high and low
recoveries. These were Lab-6, Lab-10, Lab-11, Lab-12, Lab-13, Lab-14,
and Lab-16. The range of mean recoveries for these laboratories was from
90 to 103%.
Figure 13 is a range chart showing the variability in mean recoveries by
metal. The metals were ranked in order of increasing mean recovery.
Again all of the data were included and essentially this chart confirms
those metals where some laboratories had generated exceptionally high
recoveries (outliers).
There were no duplicate samples for metals, except for sodium, that could
be used to compare performance with EPA method performance criteria for
intralaboratory precision. Therefore, only inter!aboratory precision was
evaluated using the performance equations in EPA Method 200.7 for ICP
metals analysis. It should be noted that some laboratories really did
-------
659
not use ICP, but because this was the method requested to analyze the
samples, It was used as the basis for comparison. In some Instances, It
was difficult to establish that ICP was not actually used.
Interlaboratory precision was used to calculate the expected recovery
range that a laboratory should fall within. Table 8 summarizes the
overall (interlaboratory) precision for the 11 metals with respect to
meeting EPA criteria.
For calcium, all laboratories met the EPA criteria. No laboratory was
able to meet the expected recovery range for all 11 metals, primarily
because of their performance with barium. Oddly enough, the recovery
equation for total digestion of barium in Method 200.7 predicts an
unusually low recovery (67%) for this metal that was not experienced by
the laboratories included in this study. The mean recovery for all
laboratories actually was 100%. This predicted low recovery when coupled
with the interlaboratory precision estimate actually gives an upper limit
of recovery range that is below the made-to concentration. This problem
with the barium equations did not affect the relative performance among
the laboratories, so it was merely noted here. It does suggest that
laboratories can do much better in the recovery of barium than the method
equations would predict.
Many of the laboratories did well by being within the expected recovery
range for at least 8 out of the 11 metals. Lab-8, Lab-17, and Lab-25 did
poorly in only meeting the recovery range for less than half the metals.
Review of the tabulated data shown in Table 7 for individual mean
recoveries for each metal as well as each individual laboratory suggested
excellent performance by all laboratories with only a few outliers being
observed; however, when one counts the number of times EPA criteria were
not met in Table 8, one has a different impression. It suggests that the
EPA criteria may be too restrictive in the case of metals. This
possibility has not been thoroughly explored.
General Parameters
Youden plots for the five general parameter analytes are shown in Figure
14. Preliminary Youden plots were made first to visually identify
outliers. These outliers were omitted from the calculation of the
circular area in the plots that represent where 95 percent of a single
laboratory's results are expected to fall. These outliers are identified
in the plots by a boxed-in laboratory number. In total, only 6 sample
pairs were considered outliers. There was no one laboratory that had
persistent outliers; all 6 outliers came from different laboratories. It
is interesting to note that all 6 outliers were in the BOD, COD, and TOC,
analytes that address in different ways the organic content of a sample.
A summary of the Youden plots for general parameter analytes is shown in
Table 9 using the same symbols and criteria described for volatile
analytes. An overall picture of each laboratory's performance for these
analytes as a group is shown in the performance summary column of the
Table 9. The performance summary is the number of times a laboratory fell
-------
660
in a particular region of the Youden plots as well as the number of times
the laboratory reported less-than values.
Laboratories that always fell within the region expected for 95 percent
of a single laboratory's results were Lab-6 and Lab-16. Other
laboratories that most often fell within this region were Lab-5, Lab-11,
Lab-13, Lab-17, Lab-18, and Lab-21.
Laboratories that had a tendency towards higher than average recoveries
were Lab-1, Lab-9, Lab-12, Lab-20, and Lab-22. There is no explanation
for the over 900 percent recovery shown by Lab-9 for TOC as the quality
control sample analyzed by the laboratory had acceptable recovery near
100 percent.
Laboratories that had a tendency towards lower than average recoveries
were Lab-2, Lab-4, and Lab-19. Lower than average recoveries are
indicators of poorer performance as shown by the positions of these
laboratories and the made-to values shown in the Youden plots. Lab-2 and
Lab-19 were the poorest performers for oil and grease as shown in the
Youden plot of Figure 14. They reported only about 10 percent of the
made-to concentrations in the sample pairs. Lab-13 also had a greater
deviation in pH from the true measurement (about 1.0 SU).
No laboratory showed a particular tendency towards poorer precision.
Lab-8, Lab-20, Lab-23, and Lab-24 fell in the Youden plot regions of
poorer precision only once each. No one analyte predominated as all four
laboratories fell in these regions for a different analyte.
Laboratories that often reported less-than-values were Lab-14 and Lab-15.
Among the four organic-related analytes, less-than-values were limited to
BOD, COD, and oil and grease; TOC was always reported above detectable
levels. (Analyses for pH would not be reported as less-than-value given
the nature of the analysis.) Lab-9 had an abnormally high detection
limit reported for BOD (60 mg/L). The laboratory's explanation was that
they did not expect the BOD to be at such low concentrations and they set
their dilutions too high for the analysis. Given that the true
concentrations for the sample pair were typical of treated wastewater
effluents (32 and 19 mg/L), the laboratory should have had no problem in
setting up the dilutions correctly. Setting the dilutions too high is
evidence of poor laboratory practice.
The laboratory performance shown in the Youden plot summary of Table 9 is
reflected again in the mean recoveries shown in Table 10 and the recovery
bar charts of Figures 15 and 16. The outliers identified in the Youden
plots of Figure 14, were not included in any of the calculations of
overall laboratory or analyte mean recoveries shown in Table 10 and the
recovery bar charts because they distorted the overall means too much.
The mean recovery for each laboratory is shown for individual analytes in
Table 10, along with the laboratory's overall mean for all general
parameter analytes. The laboratories are ranked in the Table 10
according to these overall means. Lab-3, Lab-9 and Lab-15 reported
-------
661
less-than-values for a few samples and so the mean recoveries for these
labs could not be Included In the overall mean calculation.
The range in overall mean recoveries for the five general parameters was
55% (Lab-19) to 115 percent (Lab-11). Among the individual analytes, the
mean for all labs ranged from 59% (oil and grease) to 104% (TOC).
The bar charts of mean recovery shown in Figures 15 and 16 tell the same
story as the numbers in Table 10, but highlight the differences among
labs more dramatically. Range charts of the high, mean, and low mean
recoveries by laboratory and analyte are shown in Figures 17 and 18. The
range chart in Figure 17 shows the variability of each lab in analyzing
general parameter analytes, with the laboratories ranked by overall mean
recovery. The high value shown is the highest mean recovery among the
five analytes for each laboratory, the low is the lowest mean recovery,
and the mean is the overall mean for all five analytes.
Laboratories showing the most narrow range between high and low
recoveries in Figure 17 were Lab-6, Lab-12, and Lab-21. The overall mean
recovery for these laboratories was 94% to 109%. The other laboratories
fall somewhere in between the wide and narrow range groups. In judging
laboratory performance, the laboratory with a narrow performance range
indicates overall consistency for this analyte group.
Figure 18 is a range chart showing the variability in mean recoveries by
general parameter. The analytes are ranked in order of increasing mean
recovery starting with oil and grease at 59 percent and ending with TOC
at 104 percent. Again, the outliers were not used in these mean
recoveries although they are noted on the range charts. COO showed the
most variability between high and low recovery (185% for Lab-4 and 23%
for Lab-19). The mean recovery for pH showed the most narrow range
(95-113%) not surprisingly given the nature of the test and ease of
measuring this parameter.
The general parameters do not have method performance equations that can
be used for precision comparisons like the volatiles and metals. Instead,
a recovery range was assumed of ±40% of the made-to concentration in
order to make a performance comparison among laboratories. The ±40% level
was selected because EPA has previously used it in data evaluations of
laboratory performance (specifically the PQL concept for volatile
organics in drinking water).
The comparison of laboratory performance based on this recovery range is
shown in Table 11. Four laboratories were able to met this criterion for
all five conventional analytes -- Lab-6, Lab-12, Lab-20, and Lab-21.
Laboratories that did rather poorly, meeting the criterion for only 2 of
the 5 analytes, were Lab-2, Lab-4, Lab-10, Lab-19, and Lab-23.
Performance Summary
A lot of detail has been presented in various ways on laboratory
performance in accuracy and precision, so much that it becomes difficult
-------
662
to get a clear picture of one laboratory's performance relative to
another. To present the data in a more concise form, a scoring system was
created to rate each laboratory's performance in each analyte group --
volatiles, metals, and general parameters. The performance for TIC's was
also included in summary table. The system was divided by analyte group
because some laboratories had definite strengths in one or two groups,
but may have been weak in others, so it was not appropriate to combine
performance across these groups.
Each laboratory's score is shown in Table 12. The method of scoring is
as follows:
Volatiles Accuracy
The performance within EPA expected recovery ranges for SW-846 Method
8240 shown in Table 5 was the basis of accuracy scoring. The number of
times that a laboratory met the EPA criteria was given these scores:
Excellent --8
Good -- 6 to 7
Fair --_4 to 5
Poor -- 3 or less
Volatiles Precision
The performance against EPA expected single analyst (intralaboratory)
precision for SW-846 Method 8240 shown in Table 4 was the basis of
precision scoring. The number of times that a laboratory met the EPA
criteria was given these scores:
Excellent --8
Good -- 6 to 7
Fair -- 4 to 5
Poor -- 3 or less
Metals Accuracy
The performance within EPA expected recovery ranges for 40 CFR 136 Method
200.7 shown in Table 8 was the basis of accuracy scoring. The number of
times that a lab met the EPA criteria was given these scores:
Excellent -- 10
Good -- 8 to 9
Fair -- 5 to 7
Poor -- 4 or less
Metals Precision
Unlike the volatiles, there were no duplicate samples for laboratories
that could be used to calculate intralaboratory precision and compare
each laboratory against the EPA criteria of Method 200.7. Instead, the
performance of the laboratory in the Youden plots that were summarized in
Table 6 were used to score precision. The number of times that a
-------
663
laboratory fell into the poor precision region of the Youden plot was
given these scores:
Excellent -- 0
Good -- 1
Fair -- 2 to 3
Poor -- 4 or more
General Parameters Accuracy
General parameters do not have EPA performance equations like volatiles
or metals that can be used to judge accuracy. Instead a criterion of
being within ±40% of the true concentration was used to judge accuracy
performance. The number of times that a lab fell within the ±40% range
(shown in Table 5-11) was given these scores:
Excellent — 5
Good --4
Fair -- 3
Poor -- 2 or less
In assessing laboratory performance based on Table 12, the most important
criterion is accuracy because if the laboratory cannot come close to the
made-to concentration, the analysis is of little value no matter how
precise it may be. If a laboratory's accuracy is satisfactory, one
should then look for good precision. When a laboratory has good
accuracy, but poor precision, it means that on average, its analyses are
right on target, but any single analysis may be way too high or way too
low. So one should look for both good accuracy and good precision in a
laboratory.
Some of the best overall laboratories were Lab-12 and Lab-14. These two
laboratories did good to excellent work in all three analyte groups. No
laboratory did poorly in all three analyte groups. Each laboratory rated
at least a "good" in one or more analyte groups.
Within individual analyte groups, the best and worst performers were:
Volatiles
Metals
Best
Lab-3
Lab-4
Lab-14
Lab-15
Lab-17
Lab-11
Lab-12
Lab-14
Worst
Lab-2
Lab-6
Lab-13
Lab-21
Lab-1
Lab-3
Lab-8
-------
664
General Parameters
Lab-6 Lab-2
Lab-12 Lab-4
Lab-21 Lab-10
Lab-19
Lab-23
Two other points thai stand out clearly in the table can be made. One,
more laboratories achieve excellent precision for metals and general
parameters, than for volatiles. Two, where a laboratory does quite
poorly in one group, it may be very strong in another. For example,
Lab-21 and Lab-6 did poorly in analyzing volatiles, but did excellent
work in general parameters. Knowing a laboratory's strength and
weaknesses can be useful if samples can be split according to analysis
and sent to the laboratory that does the better job; although, the prime
quality goal is to improve performance for all environmental analyses.
Results from this study can direct a laboratory to improve in those
weaker areas.
TENTATIVE IDENTIFIED COMPOUNDS
When one requests that a laboratory analyze a sample for volatile
organics by Method 8240, it is implied that only the analytes listed in
the method are to be reported. In some instances, a laboratory may be
asked to report additional analyte peaks identified by the mass
spectrometer on the basis of computer matching of the spectra. These
additional analytes are commonly called "tentative identified compounds"
or TIC's. As their name suggests, the reliability of the analyses is
less than that for specific target analytes listed in the method.
There are many misconceptions concerning the ability of 6C/MS methodology
and instrumentation to identify and quantify non-target analytes or TIC's
in environmental samples. This has been one of the major selling points
for GC/MS methodology. The current investigation for TIC's could be
considered biased to demonstrate the best that one might expect. Method
8240 utilizes the purge and trap concentration technique which not only
concentrates volatile organics by a factor of 1000, but also separates
these volatile organics from the other less volatile (extractables)
organics present in most samples. The purge and trap technique is also a
cleanup step. By limiting the study to volatile and purgeable-type
organics the total number of possibilities was also reduced. From an
analytical point of view, the method employed and procedures used were
the most ideal one could select for TIC's.
The selection of the TIC analytes for the study was also conservative.
All of the compounds were present in some samples that a number of these
laboratories had previously analyzed for Shell and had been routinely
reported at a number of cleanup sites. The concentrations spiked into
samples were at levels to yield chromatographic peaks easily
-------
665
distinguished above the matrix background. It was our intent to make the
TIC's easy to pick out which should have improved the qualitative and
quantitative probability.
The results for TIC's indicated most laboratories do not do very well
with TIC's. Not a single laboratory qualitatively identified the
presence of all nine TIC's. Table 13 shows the qualitative performance
of the laboratories based on the percentage of samples that had
reportable values (any values over the laboratories reporting limit).
Methyl ethyl ketone (MEK) had the highest number of reportable values
with 84% of the samples correctly identified for this analyte. In
contrast, tertiary butyl alcohol (TBA) was not found by any of the
laboratories. The only other TIC that was found in at least 50% of the
samples was diisopropyl ether (DIPE) (59%). All other TIC's had less
than 25% or less reportable values.
Reportable values would be expected to increase as the concentration of
the analyte is the sample increases. The TIC's were spiked into samples
at three levels. The percentages of reportable values at individual
spike concentration are shown in Table 14. Methyl ethyl ketone, the
analyte with the highest percentage of reportable values, shows a clear
trend of increasing percentage with increasing concentration as the
percentage reportable values increased from 72% to 96% with concentration
increases from 30 ug/L to 180 ug/L. If 95% reportable values was chosen
as the level at which the presence of an analyte could be reliably
identified, then the quantitative level for methyl ethyl ketone for this
study is around 180 ug/L.
The other eight TIC's showed a general trend of increasing percentage
with increasing concentration but the concentration range did not extend
far enough to show this clearly, nor to estimate at what concentration
95% of the values would be reportable. Based on the highest spiked
concentration for these TIC's in Table 14, the concentrations at which
95% of the values would be reportable with Method 8240 would be greater
than 90 ug/L for cyclohexane, MTBE, DIPE, napthalene, iso-octane, and
THF; greater than 180 ug/L for TCB; and, greater than 270 ug/L for TBA.
The overall laboratory performance in identifying TIC's was evaluated,
and the performance for each laboratory is shown in Table 15. This table
shows the number of reportable values for each laboratory for each TIC.
If a laboratory found an analyte in all three samples where it was
spiked, then the number of reportable values for that TIC would be three.
If a laboratory found every TIC in every sample, the maximum number of
reportable values would be 27. The highest number of reportable values
was 13 for Lab-25 which is a little less than a 50% identification rate.
Other laboratories that did nearly as well with 10 to 12 reportables were
Lab-2, Lab-10, Lab-13, Lab-15, Lab-17, Lab-19, and Lab-22. Laboratories
that did poorly, reporting 3 or less values, were Lab-1, Lab-5, Lab-9,
Lab-18, Lab-20, Lab-21, and Lab-23.
Table 15 also shows the number of analytes found at least once by each
laboratory. The highest number of analytes found was only five out of
-------
666
nine TIC's In the samples. Laboratories that found at least five TIC's
were Lab-13, Lab-16, Lab-17, Lab-19, and Lab-25. The five TIC's found by
these laboratories were not the same for each laboratory.
The number of laboratories that reported a TIC at least once is shown at
the bottom of Table 15. The number of laboratories ranged from zero
(TBA) to 24 (MEK). Lab-1 was the only laboratory that did not find
methyl ethyl ketone in any of the three samples.
With such poor qualitative performance, one could not expect the
quantitative performance to be good. Table 16 gives the mean recoveries
for each TIC for each laboratory. The laboratories were ranked by
overall mean recovery for all nine TIC's. The range in overall mean
recoveries was much wider than those for the target volatiles, 41% to
344% as compared to 55% to 117% for the target volatiles. The range for
each laboratory was also wider, for example, compare the TIC range of 87%
to 933% for Lab-7 with its 54% to 99% range for target volatiles.
The overall mean recoveries in Table 16 cannot be used very well to judge
each laboratory's performance with respect to the other laboratories
because of the variability in the TIC mean recoveries and the number of
TIC's that the laboratory actually identified. For example, the
performance of Lab-21 is deceiving since its overall mean recovery was
101%, but this was based on only one sample in which it found methyl
ethyl ketone. Since it only reported a single value the laboratory
really had poor performance for TIC's.
To better judge laboratory performance for TIC's, the criteria shown in
Table 17 were used. There are four levels of recovery performance that
were arbitrarily chosen and are shown in Table 17. The table shows the
majority of observations were less-than values or not reported at all.
Of the reportable values, 14% were found within plus or minus 40% of the
made-to values; 5% were between plus or minus 40% to 70% of the made-to
values; and, 7% were more than plus or minus 70% of the made-to value.
RECOVERY CORRECTION
The most recent drafts of SW-846 Chapter 1 requires recovery correction
of all environmental data. Recovery correction is currently mandated by
regulations for some tests such as TCLP. Some in the analytical
community have opposed recovery corrections and have provided EPA with
very explicit comments on why recovery correction should not be mandated.
However, it has been EPA's position that recovery correction always
biases the data more closely to the true value. If this were actually
true, it would be difficult to argue against. The current status of
recovery correction is that EPA is rewriting Chapter 1 to remove the
requirement.
While Method 8240 data are already recovery corrected on the basis of how
the method is run, the data resulting from this current PE study can be
and was used to gain a better understanding for recovery correction. The
-------
667
target analytes were prepared as duplicates and the third sample was 2X
the concentration of the duplicates. One could consider the 2X level
sample as the matrix spike upon which each of the duplicates could be
recovery corrected. Recovery correction calculations were performed
following EPA prescribed procedures on the organic target analytes. We
did not include the TIC's data for obvious reasons. Since the true values
were known for our samples, it was possible to test the EPA hypothesis
that the procedure always biases the data more closely to the true value.
The effects of recovery correction on the nine target volatile analytes
are summarized in Table 18. Table 18 lists the analyte, the percentage
of analyses that were improved by recovery correction (bought closer to
the true value), percentage of those that were not, and the total number
of analyses available for comparison among all the laboratories. These
recovery-correction effects are also shown graphically in the bar chart
of Figure 19.
The key for interpreting these recovery-correction effects in the bar
chart is included in Figure 19 (the interpretation of the numbers in
Table 18 is similar). As shown in the key, if the bar is all the way to
the left, all of the values (100 percent) are improved by recovery
correction. As the bar shifts to the right, fewer values are improved by
recovery correction. When the 50/50 split between improved versus not
improved is reached, one would say that recovery correction has no
advantage whatsoever. In other words, recovery correction would not be
useful because half the time it improved the values while the other half
of the time, it made the values worse (or did not improve them at all in
a very few cases). Continuing to the right past 50 percent, recovery
correction becomes a disadvantage as more and more values are found not
to be improved by recovery correction. Finally, arriving at the far
right of the chart, none of the values are improved by recovery
correction. Therefore, for recovery correction to be considered
advantageous at all, more than 50 percent of the values must be improved.
If there were an improvement for 60% of the values this would not
represent much of an improvement overall since the remaining 40% would
not be improved.
Seven of the nine volatile organics had greater than 50% of the values
improved by recovery correction. Of the seven that were improved, the
greatest improvements were for ethyl benzene (96% improved); carbon
tetrachloride (86% improved); and 1,1,1-trichloroethane (86% improved).
Lesser improvements were found for methylene chloride (65% improved);
benzene (62% improved); xylenes (61% improved); and chloroform (61%
improved). Two analytes that were less than 50% improved were toluene
(48% improved) and trichloroethylene (19% improved). As a group the
average improvement for these nine volatile organics was 65%, with 35% of
the values not improved by recovery correction. For all practical
purposes, recovery correction did improve observations for ethyl benzene,
carbon tetrachloride, and 1,1,1-trichloroethane, but for the remaining
six analytes there was marginal or no improvement at all.
-------
668
The amount of improvement due to recovery correction, however, is negated
in part by the percentage of values that are not improved. Subtracting
the percentage not improved by recovery corrections yields the net
improvement which is summarized for these nine analytes in Table 19. The
net improvement among the nine analytes ranged from negative 62 percent
for trichloroethylene to positive 92 percent for ethylbenzene with an
average of plus 30 percent for the entire group. The negative percents
for trichloroethylene and toluene mean that there was no net improvement,
but rather that a net of 4% of the values for toluene and 62% for
trichloroethylene were made worse.
Looking at recovery correction only as a percentage that was improved or
not improved does not show how much the actual analytical values were
changed. In other words, is the change a significant improvement?
Statistical tests on the interlaboratory means (improved vs. not
improved) were made for each analyte, using the Student's t-test. The
results of these tests are summarized in Table 20 which lists the
calculated "tn along with the two means and pooled deviations for each
analyte. As shown in Table 20, recovery correction did not make a
significant difference in the mean analytical value among the 25
laboratories for chloroform and methylene chloride. This indicated that
the percentage improvement for these two analytes shown in Table 19 is
not significant. For the other seven analytes in Table 20, the change in
mean made by recovery correction either for better or for worse, would be
considered significant.
In summary, recovery correction for volatile analytes improved the
analytical values by bringing them closer to the true value for seven of
the nine analytes tested in this study; however, two of the seven
improvements were not considered statistically significant. Of the
remaining five considered statistically significant, the net improvement
by recovery correction (after subtracting the percentage of not improved
values) ranged from 22% for xylenes to 92% for ethylbenzene. Recovery
correction did not improve values for two of the nine analytes.
Combining all effects for the nine analytes, recovery correction improved
a net of 30% of the analytical values. The results from this study also
show that the effects for recovery correction are different for each
analyte and this fact needs to be considered should EPA find it necessary
to mandate recovery correction for all environmental programs.
For metals data by ICP it was possible to review the results in Figure 12
which summarizes the range of recovery for all laboratories for all
metals. Figure 12 shows that the mean recovery for all laboratories was
97 percent. The tabulated data for Figure 12 is found in Table 7.
Table 7 shows that the mean recovery for all laboratories was 97%. The
means for each metal that resulted in the 97% mean value are listed on
the bottom of Table 7. The individual means ranged from 89% to 107%
which again illustrates the good performance of the laboratories
involved. When one considers the mean recovery for all laboratories and
the narrow range of recoveries for each of the laboratories and the fact
that the study was done blindly, one wonders what would be accomplished
-------
669
by attempting to recovery correct such observations.
CONCLUSIONS
The laboratories Included In the performance evaluation study meet EPA
method criteria for volatile organics and TCP metals with few exceptions.
Performances for general parameters such as oil and grease, pH, BOD, etc.
were disappointing. Considering that the PE study was conducted to
reflect day to day operations and that the samples were blind to
laboratories, one would expect similar performance for most compliance
monitoring samples. We need to more closely monitor the nature of
sampling kits provided by commercial laboratories to ensure the integrity
of samples submitted for analysis. One cannot expect quality data on
poorly collected or preserved samples.
The laboratories included in the blind PE study all were laboratories
with considerable GC/MS experience. The method selected (8240) and the
compounds investigated would be expected to bias the performance for
TIC's toward good performance. Unfortunately, the results clearly show
how poorly we perform on TIC's both qualitatively and quantitatively for
groundwater samples. The performance would be expected to be even poorer
for extraction-type methods (8270) and for solid or sludge samples. It
is time we stop fooling ourselves on the ability of commercial
environmental laboratories to routinely identify and quantify TIC's in
environmental samples.
Recovery correction as proposed by EPA simply does not always bias
results more closely to the true value. It was also discovered that in
some cases just the opposite is true and that the effects of recovery
correction may depend on the specific analyte. The results from the
study for ICP metals clearly show recovery correction would do little to
nothing to observations because of the extremely good precision and
accuracy of the method. If there really is a need for recovery
correction some discretion must be exercised. The results from this
study clearly indicated recovery correction for organics would be
complicated and for analytes such as the metals recovery correction
really is not needed. This may be true for other analytes as well.
-------
670
REFERENCES
1. W. J. Youden and E. H. Steiner, "Standard Manual of the AOAC", 1975,
Published by the Association of Official Analytical Chemists, P.O.
Box 540, Benjamin Franklin Station, Washington, DC 30044.
2. United States Environmental Protection Agency, "Test Methods for
Evaluating Solid Wastes Physical/Chemical Methods" Office of Solid
Waste and Emergency Response, Washington, DC 20460, November 1986,
SW-846 Third Edition.
3. Preamble to Promulgated Rule Revisions for Ground Water Monitoring
for Hazardous Waste Facilities, 40 CFR Part 264 (53 FR 3872).
90PESTDY.PAP
-------
671
QUESTION AND ANSWER SESSION
MR. TELLIARD: Any questions for
George?
MR. SCHUMACHER: Del Schumacher,
Lancaster Laboratories.
Your oil and grease, did you specify which
method, whether a gravimetric or infrared?
MR. STANKO: We specified an EPA
approved method.
MR. SCHUMACHER: Which some of the
spread of results could have been due to the fact that you did
have volatiles in your sample?
MR. STANKO: It could be due to
the fact that the contract laboratory did not use an EPA approved
method, yes.
MR. SCHUMACHER: The gravimetric
method is an approved method.
MR. STANKO: Gravimetric is not an
approved method?
MR. SCHUMACHER: No, it is an
approved method.
MR. STANKO: It is an approved
method. That's what was specified...an EPA approved method.
MR. SCHUMACHER: But the IR method
is also an approved method. One of them is going to give you
volatiles results and the other one is going to slight your
results.
MR. STANKO: That may be possible.
In this particular case, I think it would more than implied that
it was a gravimetric procedure as the EPA approved procedure.
There has been some follow-up work since this time and we have
discovered that, yes, they were pipetting 10 mLs of the sample.
They were not extracting the bottle and the oil was always on the
glass. The IR method will not give you the right answer if you
can't get the oil off the glass. There was some zero percent
-------
672
recoveries which is difficult to explain by any method.
MR. LANG: Ken Lang from the Army
Toxic and Hazardous Materials Agency.
George, did you determine the error in the
preparation of the samples that were sent out for analysis,
especially the volatiles?
MR. STANKO: We are quite
experienced at doing this and there were only two of the target
analytes that were actually in the background matrix and we did
account for that. Very carefully...it was done very carefully to
make sure that the volatiles were prepared properly, as well as
all the samples. There were double checks and balances and to
our knowledge, we believe that the samples actually contained
what we call the made-to values. If you looked at some of the
average recovery, like for all the volatiles, it was about 84
percent. Why the data were biased low was because there were a
couple of labs who did not perform as well. They were not
outliers, but they did bias the data towards 84 percent.
MR. LANG: You also mentioned that
many of the bottles that were shipped out contained acid. Were
all the volatiles acid-preserved or were some not and some were?
What was the...
MR. STANKO: Where they provided
bottles that had acid in, we left the acid in. We did not add
acid to ones that did not have it because the samples would have
been analyzed in sufficient time not to exceed the holding times.
To give you an illustration of the extremes, one of
these 43 mL vials had 20 mLs of hydrochloric acid. It's a little
overkill.
MR. LANG: Yes, I'd just point out
that the holding time study that was conducted by Oak Ridge
National Lab would cast some doubt on those holding times,
especially for the non-preserved samples which account for some
of the variability in the volatiles data. That's why I was
wondering whether or not they were all preserved or just some a.nd
-------
673
not others.
MR. STANKO: I know at least a
half of these were.
MR. LANG: Okay.
MR. STANKO: But we were trying to
simulate what an engineering firm would do once they got the kit.
If it was there, they normally don't carry acid in the field.
But if the acid was in the bottle, they just don't have much in
the way of an option. They have to put what's in there. So
that's what we did. There were some samples, oil and grease
samples, where the sulfuric acid had eaten through the lid and
also into the box.
MR. LANG: Okay. I have one last
question. Did any of the labs report analytes that weren't in
the samples?
MR. STANKO: We had no false
positive. That was unusual. But I think we doctored this
program up that the TICs should have stood out taller than the
target analytes, so that may be the reason. Actually, the
concentration of the TICs were higher than our target analytes.
But that didn't seem to help them any.
MR. LANG: Okay.
MR. STANKO: But there were no
false positives. Nobody reported something that wasn't there.
MR. LANG: Thanks.
MR. TELLIARD: One more.
MR. VINOPAL: I just have one
question. I wonder on the TICs, did you verify if the
laboratories had authentic standards for all of the TICs that you
put in there or did they attempt to identify these without the
aid of authentic standards?
MR. STANKO: It's not required to
use authentic standards, but all of the laboratories in this
study had analyzed samples from Shell at some point in time that
contained these TICs and they were identified by at least one or
-------
674
two or maybe three labs. So, in other words, our list of TICs
were geared for samples that actually had them. So some of these
labs should have seen them before.
MR. VINOPAL: Okay.
MR. TELLIARD: Thank you, George.
MR. STANKO: Thank you.
-------
675
Figure 1
Youdan Plots
Benzene, Toluene, Ethylbenzene, Xylenes
10 20 30 40 50 60
70 80 90
MW-1(g/L)
10 20 30 40 50 60
10
20 30 40
MW-1(0/L)
SO 60
* Tni« concOTtmtor* for Mmpte pair
OuUtof dilafcd from oUcutetton of dreulaf
-------
676
Figure 2
Youden Plots
TCA, Chloroform, Carbon T«trachlorlda, Mathytena ChlorWa, TCE
20 40 60 80
UW-1(g/L)
100 121}
1020304050607080
10 15
MW-KoA)
20
10 15 20 25 30
MW-1(fl/L>
True concentrations tor sample pair
Outlier deleted from calculation of circular
-------
677
Rgure 3
Recovery Bar Charts
Benzene, Toluene, Xylenes
23"
u:
•g is:
3 13"
.0 11*
-> 9
8~
4"
2~
•MMHMHMHM
1 • 1 • 1 • 1
* Bonzan*
mecn recovery
torallatw
«MS90%.
i • i
20 40 80 80 100
Benzene Mean Recovery (%)
120
140
SO
100 ISO 200 2SO
Toluene Mean Recovery (%)
300
1
21 ~
»~
1 *
' 1 ' 1 ' 1 ' 1
XytanflA
WM8ZH
Mow dMKton Imta for tf
20
60
80
100
120
Xylenes Mean Recovery (%)
-------
678
Figure 4
Recovery Bar Charts
Ethylbenzene, Chloroform, Carbon Tatrachlorlda
23"
21"
»:
15"
13"
10"
5"
4"
3"
—
• Elhylbanzerm
for all late
was 74%.
20 40 60 80 100
Ethyfcenzene Mean Recovery (%)
120
M
20"
ie~
8"
3"
^ Chloroform
mean recovery
for all late
wa«98%.
20
40
ao
80
100
120
140
180
Chloroform Mean Recovery (%)
0)
£t
3
A
-------
679
Rgure 5
Recovery Bar Charts
Methylene Chloride, 1,1,1-Trlchloroethane, Trlchloroethylene
al
21"
20n
19"
18"
«B 10"
— ' 9"
7"
19"
(0 10'
7"
6"
3"
*
I 20 40 60 80 100
Methylene Chloride Mei
Methytene chloride
mean recovery
forilllcte
VMS 105%.
120 140 160 1*
in Recovery (%)
^~~ 1.1.1-ICA
tor ill W»
«w«69%.
20 40 60 80 100
1,1,1-Trlchloroethane Mean Recovery (%)
21"
20-
19"
— a—
1 1 ' 1 ' I * 1
* TCE
fTIMfl PMOWy
brill MM
MM 88%.
NotM
20
40
60
100
120
140
Trlchloroethylene Mean Recovery (%)
-------
680
Figure 6
Range in Recovery by Laboratory—Volatiles
300-
250-
3* 200-
£
150-
ec
CO
1
100-..
50-1'
j High
< ' Mean
-L Low
CMCMCM»- »-»-
The mean recovery
for volatiles
for all labs was 84%.
IP
Laboratory Number
(in order of increasing mean recovery)
-------
681
Figure 7
Range in Recovery by Analyte—Volatiles
250-
£.200-
0
8 150-
1
§
Jj 100-
50-!
4
«•
r High
1 Mean
••maun Thft P
foraU
was J
nean recovery •
volatile analytes
14%.
- Low
<
I
• «
• <
i
<
, —
i
i
(
i i i
S
I
3
Volatile
-------
682
Figure 8
Youdan Plots
Barium, Cadmium, Copper, Lead, Nickel, Zinc
0.2
o.4 ae aa
MW-4(mg/L)
OJ 0.5
MW-4(mg/L)
0.05 0.1 0.15
MW-4(mg/L)
0.2
0.8-
0.6-
0.4-
0.2-
I
0.12 mgi
Nickcri
0.2 0.4 0.6
MW-4(mgA.)
0.8
MW-4(mg/L)
Nolw
0.8-
True conc*nttm
-------
683
Figure 9
Youdan Plots
Calcium, Magnesium, Sodium, Iron, Manganese
175
125 150
MW-4(mg/U)
175
10 15 20 25 30
MW-4(mgyL)
40 50 60
MW-4(mg/L)
70
1.5
2.S 3.5
MW-4(mg/L)
4.5
200
180-
160-
140
(1600.1700)
140
160
* Tru« concOTlraiom tor Mmpi* p«Jr
toiKl from exaltation of orcutar
MW-«(mg/L)
-------
684
Figure 10
Recovery Bar Charts
Barium, Cadmium, Copper, Lead, Nickel, Zinc
0-
•'
••
1 • 1 • 1 • 1
• MM 100%.
1 '
40 « M 100
Barium Mean Recovery (%)
120
140
Load rtiMn recover
for •!!•>» except tab-17
93%.
100 190 200 280 300
Lead Mean Recovery (%)
2150
400
rneen recovery
tor el tabs except leb-3
100 ISO 200 250
Cadmium Mean Recovery (%)
300
-* 9
••
I • I • I • I •
*-^™"* Nickel rnflen
recovery Ibr afl
U»ww92%.
i • i ' i
40
80
100
120
140
Nickel Mean Recovery (%)
311-
¥
24"
16"
W
9~
2"
••
1 ' 1 • 1 ' 1
1 lebe except
20
4O
80
M
100
120
140
Copper Mean Recovery (%)
p
s
4
3
2
1
Zinc meen recovery
tor el tabs except
leb-2 wee 107%.
SO
100
ISO
200
250
300
Zinc Mean Recovery (%)
Note*
* Lab reported nondencti or below detection llmit» tor el aamptoe
" NotptottedUbbelevedlohevewrilchedsernplee.
-------
685
Figure 11
Recovery Bar Charts
Calcium, Magnesium, Sodium, Iron, Manganese
25
24"
23"
22"
20"
19"
18"
16"
13"
10"
9"
7"
6"
nr
••
i • i i ' i
mean
recovery tar
I except
lab-A twa
i ' i
w ID
•s »:
E u
3 13"
2 12
JO 11"
* 10"
•J o
8 16
•O 15"
E 14"
20 40 80 80 100
Calcium Mean Recovery (%)
120
140
23"
22"
19"
13"
8"
• *
1 ' 1 ' 1 ' 1
for all late was 92%.
i • i • i • i
20
40
100 120 140 180
180
Iron Mean Recovery (%)
23"
21"
20"
18"
3 13"
2 12
2"
*•
1 ' 1 • 1 ' 1 ' 1
• Magneeium
mean
recovery
tor all hb»
except
I 105%.
**
1 • 1 ' 1 • 1
•• ""• MftnQWM
1 mean
racovofy for
•llbbswas
1 101%.
20 40 80 80 100 120 140
Magnesium Mean Recovery (%)
180
20 40 80 80 100 120
Manganese Mean Recovery (%)
140
23"
18"
9 18J
JO 15-
2 12
J» 11"
•J 9"
8"
2"
•*
' 1 ' 1 ' 1 ' 1 '
* Sodium mean
— recovery for all labs 1
except tab-18 recovery
«m*98%.
i ' i ' i * k l
Nolw
l nondMvdB or tMlowdBtBCtion HmllB forcN Munplcs
Not plotted. Lab believed to ham (witched awnptoa.
20 40 80 80 100 120 140 180 180
Sodium Mean Recovery (%)
-------
686
Figure 12
Range in Recovery by Laboratory—Metals
300-
250-
g.200-
-*•
in
o
I
High
Mean
Low
The mean recovery for all labs was 97%.
Outliers that were not ptotted:
Lab-1:50%Cu
Lab-2:284% Zn
Lab-3:271% Cd
Lab-8:79% Ca
Lab-17:386%Pb
Lab-18:994%Na
Lab-22:137% Mg
8
I
No values were plotted
for lab-9, samples were
believed switched.
I I T
1I
I I I I
T T
Laboratory Number
(in order of increasing mean recovery)
-------
687
Figure 13
Range in Recovery by Analyte—Metals
r
350-
300-
250-
1/\/\
200-
150-
100-
<
50-
0-
-r High i * ,,,
" Mean
-*- Low
' I I x "
'. f I 5 '
The mean recovery for all labs was
Outliers that were not plotted:
Cd: 271% lab-3
Ca: 79% lab-8
Cu: 50% lab-1
Pb:386%lab-17
Mg: 137% lab-22
Na: 994% lab-1 8
Zn: 284% lab-2
¥ ¥ ¥ ¥
. JL J. 1 ]_
97%.
{<
Cd Ni Fe Cu
Pb Na Ca Ba Mn Mg Zn
Metal
-------
688
Figure 14
Youdan Plots
BOD, COD, TOG, O&G, pH
200
150-
100-
50-
SO 100
MW-4(mg/L)
600
150 200
OftG
True
ooncflfttnitfofi
diffsfBd in Mch
12 lab'a sample*
100 200 300 400 500 600
MW-4(mg/LJ
10 20 30 40 iiO
6 6.5 7 7.5 8 8.5
0 20 40 60 80 100 120 140 160
MW-4(mg/L)
* True concantratons lor sample pair
Gutter deleted from calculation of circular i
-------
689
Figure 15
Recovery Bar Charts
BOD, COD, TOC
S18
e 15"
E 14-
25"
11"
9"
8"
7"
8"
2"
1
MIIM11M8
*• 234% '
*
1 ' 1 ' 1
) 20 40 60
« BOO mean
PBcovwy
brill tabs WM 70%
without lab-2. tab- 7
I ' I • i
80 100 120 14
Lab Number
iVittwn
P
s=
1 ' 1 • 1 • 1 ' 1
recovery
! tab-9.
80
100 120 140 180 180 200
BOD Mean Recovery (%)
Mean Recovery (%)
22"
21"
19"
•i 15I
E 14-
3 13-
J1S-
7"
*• 576%
—
—
»• 1183%
* 1 ' 1 ' 1 * 1 ' 1
^m far all MM
wihout
tob-10«nd
* i ' i ' i • i •
i of Mow dMctfon Into lor 41 Mmplw
20 40 80 80 100 120 140 180
COD Mean Recovery (%)
180 200
-------
690
Figure 16
Recovery Bar Charts
O&G, pH
OAG mtut racovwy
for aU late MM 59%.
40 « 80
O&G Mean Recovery (%)
too
120
23"
jib Number
,v,M,a,v,v,-,
-------
691
Figure 17
Range in Recovery by Laboratory—General Parameters
450-
400-
£350-
£-300-
o>
8 250-
c 200-
co
£ 150-
100-
50 -i
0-
•
(
Outliers that
1 Mean Lab-7: 627%
Lab-9: 957%
i/ery for all labs was 88%. .
were not plotted:
BOD
TOG
L Low Lab-1 0:1 183% COD
Lab-1 1:234% BOD
Lab-22: 576% COD
(
»
iln
.Hlill^pW
I i i i i i i i i i i i i i
T
T Jlr.}'
[I1 1
i 1 1 1 1 1 I 1
u>^u>h-oc|jl*h.eo«oc<)«D^r»-CM»-CTeM
CMCNt- i- »-T-»-I-T- CMCMCJi-
Laboratory Number
(in order of increasing mean recovery)
-------
692
Figure 18
Range in Recovery by Analyte—General Parameters
250-
?200-
0
>
§ 150-
OC
i
1 100-
50-
•i
(
i
ean recovery for all
labs was 88%.
Outliers that were not plotted:
1 Mean BOO:
TOG:
• Low COO:
<
(
»
i
J-
lab-7(627%), lab-1 1 (234%)
lab-9 (957%)
lab-1 0(1 183%).
I
lab-22 (576%)
(
O&G
BOD
COD
Analyte
PH
TOG
-------
693
Figure 19
Effects of Recovery Correction
Volatiles
% Improved by
Recovery Correction
100
50
% NOT Improved by
Recovery Correction
50
All samples
improved
Fewer samples
improved
Key to Interpreting .Chart
50/50 split
Recovery-correction improvements
negated by same number of values
not improved
ethylbenzene
carbon tetrachloride
1,1,1 -trichloroethane
methylene chloride
benzene
xylenes
chloroform
toluene
trichloroethylene
100
'More samples
NOT improved
\
No samples
improved
-------
694
Table 1
Sample Concentrations
Target Vo la tiles and TICs
Analyte
Background
Matrix*
Sample Concentration
After Spiking (ug/L)
MW-1 MW-2 MW-3
Target Volatiles
benzene
carbon tetrachloride
chloroform
ethylbenzene
methylene chloride
toluene
1,1,1 -trichloroethane (TCA)
trichloroethylene (TCE)
xylenes
18
43
10
45
16
33
15
30
10
20
68
30
15
16
33
30
10
5
20
43
10
15
26
59
15
10
5
40
Tentatively Identified
Compounds (TICs)
cyclohexane
diisopropyl ether (DIPE)
isooctane
methyl ethyl ketone (MEK)
methyl tert-butyl ether (MTBE)
naphthalene
tert-butyl alcohol (TBA)
tetrahydrofuran
1,1,1-trichlorobenzene (TCB)
15
45
15
30
45
90
45
15
90
45
90
45
90
90
45
270
45
180
90
15
90
180
15
15
135
90
30
Ground water contaminated wtth benzene and ethylbenzene was dHutad wflh reagent water (or the
background sample matrix. The values (or benzene and ethyfbenzene were calculated from the
found in the groundwater source used to make the sample matrix.
-------
695
Table l.cont'd
Sample Concentrations
Metals
' ' '/^J^v
Analyte
% •; '
barium
cadmium
calcium
copper
iron
lead
magnesium
manganese
nickel
sodium
zinc
* Groundwater from
•f^j *•' AI.V- "• y* s % s ^'**f'fy$s&
\ •>' % -X . w fJffya
ae3:J>> i* 4J&. r •• &&?&• fC\ "\ *•• < v*^4> f, * ffqf,
Background
Matrix*
(mg/L)
•.',., ' - ' * - 4.
0.384
-
129
-
-
-
37.9
2.09
-
176
0.045
Sample Concentration
After Spikinq (mart.)
MW-4 MW-5
0.384 0.72
0.114 0.034
155 129
0.939 0.522
19.6 53
0.111 0.158
37.9 42.7
3.44 2.09
0.134 0.382
176 176
0.138 0.258
an uncontaminated well was used as the badqroumd
matrix for the spiked samples.
-------
696
Table l.cont'd
Sample Concentrations
BOD, COD, TOC, pH
Analyte
Sample Concentration
after Dilution (mg/U*
MW-4
MW-5
BOD
COD
TOC
pH
32
53
20
7
19
32
12
7
Lab standard was dluted with
deoinized water to prepare sample*
-------
697
Table l.cont'd
Sample Concentrations
Oil and Grease
Sample
Lab
Number
after Dilution
MW-4
(ma/L)'
MW-5
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
20.0
22.2
20.3
19.8
19.5
19.7
18.5
21.8
20.5
19.0
18.8
20.0
20.0
19.3
20.1
21.1
21.7
20.7
19.7
19.8
19.9
19.8
20.6
17.1
20.3
37.0
41.8
39.8
39.7
39.3
38.7
34.6
41.9
41.7
35.0
36.8
36.6
37.0
33.8
36.4
39.4
42.8
38.8
37.8
38.8
38.1
38.0
38.2
26.5
37.0
Each sample far oi and grease was made
IndMduaty In each lab's containers.
Variations In the amount of ol and
grease added made for sightly different
concentrations among the labs.
-------
698
Table 2
Youden Plot Summary
Target Volatiles
O
A
V
X
nr
D
Volatile Analytes
Lab (see list at bottom of table)
No.
123456789
1 ••VVV0VV*
2 I A I O A A | A | A A A A
3 • • • • • • V •
4 • • • • • • A •
5 • • V • • V • •
6 o o • o o • m •
7 • • • • • X V X
8 X • A A • A A X
9 X • A A A A A •
10 • • • • • X • X
13 V V X • • • X
14 A • • • A A | A | •
16 V00VVXVX
17 AAAA*««t)
18 • • V • • nr X
20 VVVV| O • X V
21 | V | V V X V X V X
22 V • • • • X • X
23 •••• ••••
24 V V V | O | X V X O X
25 ••V«f)t>«A«
iiijjL^^^^^
Performance
Summary
•
»£<&%»*
4
0
8
8
7
4
6
3
3
7
2
8
5
5
9
3
5
6
2
0
6
9
0
7
Ust of Analytes
i— uanznne 4— xyienes /-caraon •vacnionos
A
0
8
0
1
0
1
0
4
5
0
5
0
0
4
0
0
4
0
2
0
0
0
0
0
1
V
5
0
1
0
2
0
1
0
0
0
1
0
2
0
0
4
0
1
0
5
5
1
0
4
1
O
•. V> ft, ft
0
1
0
0
0
4
0
0
0
0
0
0
0
0
0
0
0
0
0
1
0
0
0
2
0
X
0
0
0
0
0
0
2
2
1
2
1
1
2
0
0
2
0
1
0
1
4
2
0
3
0
2— toluene 5—1 ,1,1 -trlchloroelhane 0 nielhylene chloride
3-«thylbenzene ft-chtoroform 9-4ricntoroethylene
Within 95% circle representing values where 95% of single lab* results are expected to fall
Outside 95% circle showing tendency towards poor piedslon
Outside 95% circle showing tendencey toward tower tun average recoveries
Not plotted because at least one value was reported as less-lhan-value
No data reported
Vallum that warn mrmktanvf raiflbint ami nmitfftri frrvn rsifad&tfai nf ItM
circular region representing 95% of a single lab's results
-------
699
Table 3
Mean Recovery—Target Volatiles
Mean Recovery (%j of Volatile Analytes
(see list at bottom of table)
Lab
No.
20
21
24
16
1
13
18
22
5
7
23
10
25
15
12
3
4
6
11
8
17
9
14
19
2
Mean
List of Analytes
1 -benzene
2-toluene
3-ethylbenzene
1
66
46
83
84
80
84
91
81
94
99
84
93
84
91
89
96
87
90
112
101
97
102
99
93
134
90
4-xytones
2
69
61
67
81
87
90
87
89
92
98
89
91
88
86
91
98
87
93
97
94
101
97
99
286
124
98
3
30
37
58
76
68
64
78
73
76
81
79
84
70
79
82
73
71
76
86
80
89
82
78
88
97
74
4
21
*
75
78
69
61
66
88
75
76
89
94
78
78
84
71
76
80
98
102
109
102
78
100
117
82
5
65
60
44
50
53
60
70
57
69
58
71
64
64
68
75
77
77
71
75
83
80
87
71
72
98
69
6
98
66
75
79
96
98
84
89
95
90
87
96
85
87
95
107
97
94
130
111
112
119
124
106
136
98
7
47
*
34
50
38
54
60
57
29
54
58
53
50
59
60
73
68
50
68
75
72
80
81
53
86
59
8
*
73
93
68
65
104
nr
107
96
83
101
90
149
108
107
76
129
144
80
118
102
118
158
107
137
105
9
45
50
71
70
86
75
80
80
97
85
77
77
83
100
84
100
97
113
100
98
100
94
117
100
125
88
Mean
55
56
67
71
71
77
77
80
80
80
82
82
84
84
85
86
88
90
94
96
96
98
101
112
117
84
7-carbon tetrachloride
5-1, 1,1-trtchloroe thane
8— chloroform
0-methylene chloride
9-trlchloroethylene
* Not calculated because lab reported "below detection limits' for all samples
nr No data reported
-------
700
Table 4
Single Analyst Precision (Intralaboratory)
Compared to Method 8240 Performance Criteria
Ariaiytes
Lab
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
(see list at bottom of table)
1
V
no
V
V
V
no
V
*
*
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
2
V
no
V
V
no
no
V
V
V
V
V
V
no
V
V
V
V
V
no
no
V
no
V
no
V
3
V
no
V
V
V
no
V
V
V
V
V
V
*
V
V
V
V
V
V
V
*
V
V
V
V
4 5
no
- V
- V
- V
- V
- V
- V
- V
- V
- no
- V
- V
*
- V
- V
*
- V
- V
- V
- V
*
*
- V
*
no
6
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
*
V
V
V
V
7
*
no
V
V
no
no
*
V
no
*
V
V
*
V
V
*
V
V
*
V
*
*
*
*
V
8
V
V
V
V
V
no
V
V
V
V
V
V
V
V
V
V
V
*
V
*
no
V
V
V
V
9
*
*
V
V
V
V
*
*
*
*
*
*
*
V
V
*
V
*
V
V
*
*
V
*
V
Number
of Times
Criteria Met
V
5
3
8
8
6
3
6
6
5
5
7
7
3
8
8
5
8
6
6
6
2
4
7
4
7
List of Anaiytes
1 -benzene
2-toluene
3-ethylbenzene
4-xylenes
5-1.1
7-carbon tetrachloride
, 1 -trichloroethane 8-methy tone chloride
6-chloroform
9-trichloroethyiene
V Met EPA performance for single analyst precision
no Did not meet EPA performance for single analyst precision
* Could not be determined because lab reported tess-than-values for at least one sample
- No EPA method performance equations
for single analyst precision
-------
701
Table 5
Expected Recovery Range for Volatiles
Compared to Method 8240 Performance Criteria
Analyies
Lab
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
(see list at bottom of table)
1
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
no
V
V
V
V
2
V
no
V
V
V
V
V
V
V
V
V
V
>/
V
V
V
V
V
no
V
no
V
V
no
V
3 4
V -
V -
V -
V -
V -
V -
V -
V -
V -
V -
V -
V -
V -
V -
V -
V -
V -
V -
V -
no
no
V -
V -
V -
V -
5
no
V
V
V
V
no
no
V
V
no
V
V
no
V
no
no
V
V
V
V
*
no
V
no
no
6
V
no
V
V
V
V
V
V
V
V
no
V
V
V
V
V
V
V
V
V
V
V
V
V
V
7
no
no
no
no
no
8
V
V
V
V
V
no no
no
no
V
no
no
no
no
V
no
no
no
no
no
no
»
no
no
no
no
V
V
V
V
V
V
V
V
V
V
V
*
V
*
V
V
V
V
V
9
no
V
V
V
V
no
V
V
V
no
V
no
no
V
V
no
V
no
V
no
no
no
no
no
V
Number
of Times
Criteria Met
V
5
5
7
7
7
4
6
7
8
5
6
6
5
8
6
5
7
5
6
4
2
5
6
4
6
Ust of Analytes
1 -benzene
2-toluene
3-etfiylbenzene
4-xylenas
5-1.1.
1-trtchtoroe thane
6-chloroform
7-carbon tetrachloride
8-methylene chloride
9-trichloroethylene
V Met EPA performance for recovery range
no Did not meet EPA performance for recovery range
* Could not be determined because lab reported tess-than-values for at least one sample
- No EPA method performance equations for recovery range
-------
702
Table 6
Youden Plot Summary—Metals
'•f't '/'.{:.„,,:;', ^ s ' •
Lab Metal
No.
Ba Cd Cu Pb Ni Zn Ca Mg Na Fe Mn
f f -f *
1 • X [j5] O O O • A • • A
3 • [~O~| • ••O«««V«
4 «X«XX«««A*A
5 •••V««««A««
8 V V V • V • ["?"] V V V V
„ "-^A.^T^^v '' U- ;
Performance
Summary
• A
tf f
4 2
90
£
8 0
6 2
9 1
in 1
1U 1
11 n
11 U
2 0
9 Not used. All sample pairs appeared to be switched In lab reports.
10 ••A««*A«A«A
15 ••••V«A««OV
16 ••V«V««V««V
17 V«V|A]«»*VVVV
18 o ••••&••[ A ]•
19 •X^X^AVVV*
20 «AA««««AVV
21 AX0XXAVAA0O
23 ••OA*««OO«
25 AOAA««nrA««A
7 4
11 n
11 U
9n
U
m 1
1U 1
11 n
11 0
7 1
7 0
4 1
8 2
5 1
6 3
2 4
90
£
7 1
m 1
1U 1
4 5
V
0
1
0
1
9
0
2
4
6
0
3
2
1
0
0
O
•f
4
2
0
0
0
0
1
0
0
1
0
0
1
3
1
X
V %
1
0
3
0
0
0
0
0
0
0
2
0
3
0
0
• Within 95% drde representing values where 95% of single lab's results are expected to fall
O
A
V
X
nr
D
Outside 95% circle showing tendency towards poor precision
Outside 95% drde showing tendency toward higher than average recoveries
Outside 95% cirde showing tendencey toward lower than average recoveries
Not plotted because at least one value was reported as less-than-value
No data reported
Values that were considered outliers and omitted from calculation of the
circular region representing 95% of a single lab's results
-------
703
Table 7
Mean Recovery—Metals
Mean Recovery (%) of Metals
Lab
No.
8
17
5
16
15
1
7
12
3
11
24
22
20
14
19
18
13
6
2
10
4
21
23
25
9
Ba
91
85
99
92
95
99
101
99
102 £
104
99
97
106
103
103
94
103
109
97
107
100
108
105
110
| Not used.
100
Cd
68 £
79
74
84
69
100
89
82
27l"i
88
88
84
98
95
96
97
91
100
88
95
123
88
92
86
Ca
~79l
88
95
102
106
98 [
97
91
102
96
99
98
102
99
94
99
104
98
99
104
102
92
98
nr
Cu Fe
80 68
79 76
86 90
84 89
86 99
"~50l 98
95 88
94 89
97 76
96 88
97 92
89 109
99 78
95 93
93 84
97 85
97 92
95 96
92 138
100 101
87 97
90 96
95 91
103 98
Pb
73
"3861
38
77
83
64
55
97
75
105
65
100
103
92
101
105
93
100
103
93
*
127
160
137
Mg
86
97
102
94
103
112
104
103
107
107
109
~J37l
112
105
100
105
111
111
105
111
108
112
106
110
Mn
76
83
94
93
93
111
98
100
102
100
107
96
109
102
97
102
107
105
100
109
113
107
105
111
Ni
73
76
86
82
74
64
94
91
97
93
96
104
85
101
108
88 £
97
96
106
102
105
105
98
101
Na
86
94
102
99
99
97
97
93
97
96
102
96
88
102
90
994
99
98
97
102
105
106
97
101
Zn
103
94
103
92
92
73
103
100
90
103
124
109
106
100
131
[ 129
111
105
284 1
113
105
131
131
122
All
Metals
80
85
88
90
91
91
93
94
94
98
98
98
99
99
100
100
100
101
103
103
104
106
107
108
All sample pairs appeared to be switched in lab reports. I
89
98
92 92
93
105
101
92
98
107
97
* Not calculated because lab reported "below detection limits' for all samples
nr
^^^•M^BHHHB
Lab did not report any analyses for this analyte
I | Boxed-in values are not included in calculations of overafl means.
-------
704
Table 8
Expected Recovery Range for Metals
Compared to Method 200.7 Performance Criteria
f ' ••
Lab
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
-''
, - ,
-
*-
-,-
w*
'."•
**
Metals
Ba
%
no
no
no
no
no
no
no
V
Cd Ca
%
* V
V V
no V
no V
no V
V V
V V
no V
Cu
'
no
V
V
no
V
V
V
no
Fe
'
V
no
no
V
V
V
V
no
Pb
,.
V. V -.
V
V
no
*
no
V
no
no
Mg
y-\
V
V
V
^1
*f
V
V
no
Mn
, ^
%
no
V
V
no
V
no
V
no
Ni
4*
"
V
V
V
V
V
V
V
no
Na
<'
"" , ,
V
•vf
V
V
V
V
v
V
Zn
,/
"•f
V
no
no
V
V
V
V
V
* ' s X
fe^ *i\. % -Wv ' v
Number
of Times
Criteria Met
V
ff f f ••
%V< f f f ff '' VV
7
8
6
6
8
9
9
4
Not used. All sample pairs appeared to have been switched in lab reports.
no
no
no
no
no
V
V
V
v
no
no
no
no
no
no
no
V V
V V
V V
V V
V V
no V
no V
no V
V V
V V
V V
V V
no V
V V
V V
no
V
V
V
V
V
V
no
no
V
V
v
V
V
V
V
no
no
V
V
V
v
no
V
no
V
V
V
V
no
V
V
V
V
V
V
V
V
V
no
no
V
V
V
no
V
V
no
no
V
V
V
V
V
V
V
•vf
V
^
no
no
no
V
V
V
no
V
V
no
V
V
V
no
V
V
no
no
V
no
no
no
^
V
V
V
v
no
no
no
v
V
V
V
v
V
V
V
V
V
V
V
V
V
V
^
no
V
V
V
V
V
V
V
V
v
v
V
V
V
V
V
V
no
V
no
V
no
no
V
8
10
10
9
10
8
7
5
10
9
8
6
7
8
7
5
V Met EPA performance for recovery range
no Did not meet EPA
performance for recovery range
* Could not be determined because lab reported less-than-values for at least one sample
- No data reported for analyte
-------
705
Table 9
Youden Plot Summary—General Parameters
* ' "
, > - -..-
Lab
No.
-
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
f f f .. *• v
f. '-.
'*/'
..;.. -„ v'
General Parameters
B
O
D
•.
•
| A
•
•
•
•
| A
•
X
•
I A
A
•
X
•
•
•
•
V
A
•
•
A
V
V
C
O
D
X
X
A
0
•
•
T
0
C
,v,«.
•
V
•
V
•
•
•
O
A
0
&
G
„'
A
V
X
•
X
•
v
X
I *
V
•
A
•
•
X
•
X
•
V
O
A
A
•
•
•
pH
A
•
V
V
V
V
•
•
•
•
0
A
,
,< %J?-*V
•o •• ••'»'''
<
"v.?, ":.-"
Performance
Summary
•
A V
O
X
, - ,/ \-' *'-'.*? *,\y, *-
3
2
3
2
4
5
3
3
2
3
4
3
4
3
3
5
4
4
1
2
4
3
3
3
3
2
1
0
1
0
0
1
0
2
1
1
2
1
0
0
0
0
0
0
2
1
2
1
0
1
0
2
1
2
0
0
1
0
0
1
0
0
0
0
0
0
0
1
3
0
0
0
0
1
1
0
0
0
0
0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
1
0
0
1
1
0
0
0
1
0
1
0
0
1
1
0
0
0
0
2
2
0
1
0
1
0
0
0
0
0
0
• Within 95% circle representing values where 95% of single lab's results are expected to fall
O Outside 95% circle showing tendency towards poor precision
A Outside 95% circle showing tendency toward higher than average recoveries
V Outside 95% circle showing tendencey toward lower than average recoveries
X Not plotted because at least one value was reported as tess-than-value
D Values that were considered outliers and omitted from calculation of the
circular region representing 95% of a single lab's results
-------
706
Table 10
Mean Recovery—General Parameters
Lab
No.
* * jf*
19
2
8
5
25
15
7
24
10
3
14
17
18
16
13
6
4
1
22
21
23
9
12
20
11
^ ;.• , "
O&G
s ""-^ **
10
9|
18
46
79
*
31 1
75
28
*
81
12
52
47
81
84
57
95
101
92
32
77
112
60
90 1
59
... •• -
General
BOO pH
'"„- ?,'" ' ,A- <;
27
4711
60
57
16
60
627 1
27
80
54
81
83
68
72
50
69
83
51
83
74
122
*
126
128
234 1
70
* Not calculated because lab
",' c "£
^ "i.t $&:&&tf™
23
85
110
88
74
41
81
98
11831
107
55
116
94
112
119
103
185
100
576 1
114
151
144|
105
86
164
102
reported "below detection
| | Boxed-in values are not included in
-w- <
118
64
70
84
103
113
106
97
116
70
107
121
138
123
101
105
45
123
104
107
103
957 1
101
182
104
104
All
General
Parameters
%vn>s;>',»;
55
64
72
76
76
78
80
80
82
82
85
87
90
91
93
93
93
96
97
98
102
108
109
111
115
88
limits" for all samples
calculations of overall means.
-------
707
Table 11
Recovery Comparison
General Parameters
Analyte
Lab
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
V
no
*
B
O
D
no
no
no
V
no
V
no
V
*
no
no
V
no
V
V
V
TJ
V
no
V
v
V
no
no
no
C
O
D
V
V
V
no
V
V
V
V
V
no
V
V
V
no
*
V
V
V
no
V
V
no
no
V
V
T
O
C
' v. Cv. '
V
no
V
no
V
v
V
V
no
V
V
V
V
V
V
V
V
V
V
V
V
v
V
^
V
O
&
G
V
no
*
no
*
V
no
no
V
no
v
V
V
V
*
no
*
no
no
•V
V
V
no
V
V
pH
V
•>/
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
V
Number
of Times
Criteria Met
V
" S-K- * ' "" ^f *•*
4
2
3
2
3
5
3
4
3
2
4
5
4
4
3
4
4
4
2
5
5
4
2
4
4
Within ±40% of true concentration
NOT within ±40% of true concentration
Could not be determined because lab reported
tess-than-values for sample
-------
708
Table 12
Lab Performance Summary
„ xfc. •.-£- "
' vt.^tf'ff "•
Lab
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
js-\_-- ,~'^ s,, f >rv>x
%> M"A%.O.X*v% .w^ }Sft
Volatiles
A P
o o
0 0
® •
® •
<8> ®
0 O
® ®
® ®
• o
o o
0 ®
® ®
0 O
• •
® •
o o
® •
o ®
® ®
o ®
o o
0 0
® ®
0 0
® ®
• Excellent
® Good
0 Fair
O Poor
j>;jK&r ; ,*&*?$&
Metals
A P
0 0
® •
O O
o •
® •
® •
® •
o •
data not used
® •
• •
• •
® •
• •
® ®
O •
O •
• ®
® •
® •
O ®
0 •
® 0
o •
o ®
A — Accuracy
P — Precision
'j^'-->* ~-& ^
x^iSS-x^ ,,
0
0
0
o
•
0
®
o
0
®
•
®
®
o
®
®
®
o
^k
•
®
o ®
® ®
9 •
-------
709
Table 13
Summary of Percent Reportable Values
TIC's
Analyte
Percent
Reportable
Values
Overall*
Cyclohexane
MTBE
DIPE
Napthalene
TBA
Isooctane
MEK
TCB
THF
15
21
59
16
0
4
84
24
12
Over the entire set of samples
(75,3 samples analyzed by 25 labs)
-------
710
Table 14
Percent Reportable Values by Concentration
TIC'S
Spiked
Analyte Concentration
(H9/L)
Cyclohexane 15
45
90
MTBE 15
45
90
DIPE 15
45
90
Napthalene 15
45
90
TBA 45
135
270
Isooctance 15
45
90
MEK 30
90
180
TCB 30
90
180
THF 15
45
90
Percent
Reportable
Values
8
20
16
16
24
24
56
56
64
12
20
16
0
0
0
0
0
12
72
84
96
32
16
24
4
16
16
-------
711
Table 15
Number of Reportable Values
TIC'S
,'lf'<*" -, ', '"-'%'$&* "-. ","' t, ""' ""%" ", *'
Number of Reportable Values
Lab (see key and analyte list at bottom of table)
No. 1234567
1 0010010
2 0033002
3 0002013
4 0020003
5 0000003
6 3030003
7 1030002
8 0000003
9 0000003
10 0033003
11 0030003
12 0010003
13 0230013
14 0330003
15 2030003
16 1322001
17 0130003
18 0000003
19 1230003
20 0000003
21 0000001
22 1230003
23 0000002
24 0022001
25 2330003
Total Reportable Values 11 16 44 12 0 3 63
Labs with Reportable Values 7 7 17 5 0 3 24
8
0
3
3
0
0
0
1
2
0
3
0
0
3
0
0
0
1
0
0
0
0
0
0
0
2
18
8
9
1
0
0
0
0
0
0
0
0
0
0
0
0
0
2
0
2
0
2
0
0
2
0
0
0
9
5
'«f Mf&'tr*. ™
Total
Reportable
Values
3
11
9
5
3
9
7
5
3
12
6
4
12
9
10
9
10
3
11
3
1
11
2
5
13
Number of
Analytes
Found
3
4
4
2
1
3
4
2
1
4
2
2
5
3
4
5
5
1
5
1
1
4
1
3
5
List of Analytes
1-cydohexane 4-Napthatene 7-MEK
2-MTBE 5-TBA 8-TCB
3-DIPE 6-lsooctane 9-THF
3 Found analyte at all three spiked concentrations
2 Found analyte at two of three spiked concentrations
1 Found analyte at one of three spiked concentrations
0 Did not find the analyte at of the three spiked concentrations
-------
712
Table 16
Mean Recovery by Lab—TIC's
•'•y/ ^^f::-- "},,
Lab
No.
* X^ 'i&3il*^ '-^v'-
3
9
5
18
25
24
6
20
23
1
21
22
4
19
13
10
16
14
17
15
12
11
2
8
7
Minimum
Maximum
;*'\i'~
.. ,'
'<,V>
tic
''r.V
Mean
' }Af Ssp f
$#
' '-S& '
$* >
;|rJP>^'£r;
Recovery (%)
(eaa list at bottom of table)
1
,^J'X <
*
37
*
18
*
73
*
227
*
*
41
*
*
83
*
*
*
87
18
227
2
*'-.' '
*
86
*
*
*
102
*
22
69
*
77
120
160
*
*
*
*
*
22
160
3
f(S „
*
105
121
127
260
204
133
179
265
191
227
168
221
411
156
279
830
213
105
830
4
22
*
43
*
*
*
*
*
*
87
161
*
*
*
*
*
89
*
22
161
5
'
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
*
6
11
*
*
*
20
*
*
*
20
*
*
*
*
*
*
*
*
*
11
20
7
^
73
60
63
78
80
82
104
89
90
*
101
113
84
109
116
84
100
80
48
113
193
80
77
57
142
48
193
8
/'*' ''
59
*
97
*
4
*
*
*
*
*
*
109
119
*
*
240
*
*
*
146
528
933
59
933
9
4; ••v,;
*
*
*
*
*
*
12
*
23
*
34
*
*
*
*
19
27
*
*
*
*
*
12
34
Mean
: '' '?, \
41
60
63
78
81
82
83
89
90
97
101
103
109
114
116
120
121
123
137
158
174
179
286
293
344
Ust of Analytes
1-cyclohexane
2-MTBE
3-OIPE
4-Napthalene
5-TBA
6-lsooctane
* Did not find the analyte in
7-MEK
8-TCB
»-THF
any of the three spiked samples
-------
713
Table 17
Overall Recovery by Lab
TIC'S
:::::::::::::::::::::::::::::::::::::::::::
Lab
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Total Samples
Percent Samples
tc
±40%
xt40%to±70%
xt70%
Recovery Performance
Number of Samples
NC ±40% >±40%
to ±70%
24 0 0
16 6 2
18 4 2
22 4 1
24 2 1
18 4 2
20 2 2
22 2 2
24 2 1
15 9 1
21 3 0
23 0^1
15 8 0
18 7 1
17 5 1
18 5 2
17 0 5
24 3 0
16 6 1
24 1 2
26 1 0
16 6 0
25 2 0
22 3 2
14 9 4
499 94 33
74% 14% 5%
X*X*X'X*X'X"X'
>±70%
3
3
3
0
0
3
3
1
0
2
3
3
4
1
4
2
5
0
4
0
0
5
0
0
0
49
7%
Not calculated because lab reported no data or less-than-values
Within ±40% of spiked concentration
Between ±40% and ±70% of spiked concentration
More than ±70% difference from spiked concentration
-------
714
Table 18
Effects of Recovery Correction
Volatiles
Analyte
benzene
carbon tetrachloride
chloroform
ethylbenzene
methylene chloride
toluene
1 , 1 , 1 -trichloroethane
trichtoroethylene
xylenes
Average
% % Not
Improved Improved
62
89
61
96
65
48
86
19
61
65
38
11
39
4
35
52
14
81
39
35
Total Number
of Analyses
48
36
49
48
46
50
42
27
46
-------
715
Table 19
Net Improvement by
Recovery-Correction
% Net
Analyte Improved*
benzene 24
carbon tetrachloride 78
chloroform 22
ethylbenzene 92
methylene chloride 30
toluene -4
1,1,1 -trichloroethane 72
trichloroethylene -62
xylenes 22
Average 30
Net improved is the difference
between percentage improved by
recovery-correction minus the
percentage not improved.
-------
716
Table 20
Recovery-Correction
Test of Means
Analyte
Mean
(Interlaboratory)
No With
Recovery- Recovery-
Correction Correction
Standard
Deviation
for Test
of Means
Calculated
MAM
benzene 39.5
carbon tetrachtoride 6.3
chloroform 14.6
ethylbenzene 11.5
methylene chloride 35.4
toluene 15.4
1,1,1-trichtoroethane 7.1
trichtoroethylene 4.8
xylenes 16.3
46.2
10.7
14.7
14.3
35.6
18.2
10.1
5.3
20.5
1.37
0.41
0.48
0.57
1.56
1.64
0.30
0.31
1.84
-4.88
-10.67
-0.04
-4.77
I -0.13
-1.67
-10.10
-1.66
-2.26
Boxed-in Y shows which means are NOT significantly different. For simplicity, a
z-factor instead of Student's t was used to decide if the means were significantly
different (z-factor of 1.645 at 95% significance level).
-------
717
MR. TELLIARD: Our last speaker
today is Dr. Fred Haeberer from Quality Assurance Management
Staff. Are you going to work from up here or down there?
-------
718
DR. HAEBERER: Up here. Thanks,
Bill.
An observation, maybe it's apocryphal, but QA/QC is
bringing up the rear again, Bill.
MR. TELLIARD: Who said it was
boring? Anyway...
DR. HAEBERER: Well, it's boring
to those people who let their egos get in the way.
MR. TELLIARD: Oh.
DR. HAEBERER: Like Joan Fisk
and Ramona Trovato, I used to have a real job working in
laboratories but as younger people came into the laboratory,
brighter younger people with backgrounds in such esoteric areas
as statistics and Scientology, I knew it was time to get out and
consequently I moved to headquarters, thereby raising the IQ
levels of both organizations.
I'm going to tell you about a project that we have been
kicking around for, oh, a couple of years now. The basic
background of it is that, as we all know, there are error
distributions associated with measurements and each operator
introduces his own level of random errors and even systematic
errors. Specifically, we have developed, through statistical
means, a program for compensating for the difference between
laboratories or analysts' performances; that is, through
replication by the individual laboratories or analysts
participating in a study, they are able to compensate for the
differences in their performance.
Let me say that this program addresses only measurement
errors. It does not address sampling errors and, as I understand
it, the major share of total error is generally due to sampling
design, sampling and sample preservation and sample handling.
Also, this program is ideally suited for homogeneous matrices.
One is looking at soils from a Superfund site where not only the;
individual sample may be non-homogeneous, but gross sample matrix
differences may be encountered when sampling as close as three
-------
719
feet apart, this program would have to address additional
complexities.
Slide 1 illustrates if one has a series of measurements
with an associated error distribution and averages these
measurements, the error distribution of the end result, will be
considerably tighteri.e., the averaged result will be more
precise. That's what I've tried to indicate here with the yellow
distribution "curve".
Slide 2 addresses that this program also addresses the
issue of adjusting for bias. It adjusts sample analytical
results for the method bias that a particular operator or
laboratory incurs when performing matrix spike analyses The
program takes into consideration that the error distribution of
the adjusted result will be broader. This is shown here by the
curve on the right representing the "adjusted" error
distribution, which is somewhat broader than the left curve, the
error distribution associated with the unadjusted results.
In implementing this program the agency would develop
replication plans which would be based on the specific data
quality goals that a particular program or data user required.
Slide 3 is a representation of one such replication plan to
achieve a specific data quality constraint. An operator that
typically is incurring three percent relative standard deviation
could get away with a one, one replication plan. That is to say,
he would have to replicate his analysis once and he would have to
do one matrix spike analysis for the purpose of bias adjustment.
As relative standard deviation increases, replication
requirements increase.
We have heard several people refer to data quality
objectives today. Data quality objectives STET it really refers
to a process. Slide 4, represents that process in diagram form.
We're talking about holding the data user's feet to the fire and,
through an iterative process, make that data user bite the bullet
and communicate the level of total error that he or she is
willing to tolerate in the environmental data that's to be used
-------
720
for decision making. Once that total acceptable error is
communicated, it is aportioned to field and laboratory
operations. The major portion should be allocated to sampling
the big lump has got to go over there and a portion to and a
lesser amount to measurement activities. The laboratory portion
is translated into measurement quality constraints i.e.,
analytical precision and bias.
The intra-laboratory QC approach to achieving these
measurement quality constraints is depicted schematically on
Slide 5. In implementing this program a laboratory or an
operator would initially estimate the variability incurred either
by examining available historical data, from the analysis of very
similar type samples, or by doing a special study using the
sample nmatrix to be analyzed to establish what his/her
variability is. Then, taking that variability, the analyst would
go to the agency provided replication plans, select the
appropriate replication, perform the analyses, adjust the
results, that is, average of the results and then adjust for bias
or matrix spike recovery. If a significant change in precision
has occurred, a different replication plan is selected for
subsequent analyses. The adjusted data are then used for
decision making.
An organization approached Quality Assurance Management
Staff (QAMS) because it had noted significant performance
differences in the results that it was receiving from a group of
commercial laboratories. Slide 6 presents their zinc data.
There is a mistake in this illustration. It should
read "micrograms per kilogram" along the abscissa.
This organization had been providing Performance
Evaluation (PE) samples to its family of contract laboratories,
over a relatively short time and noticed that the laboratories
performance was quite variable and the organization didn't know
whether the performance of these laboratories actually met their
data quality constraints. QAMS took the results from these PE
studies, and derived the error distributions depicted in this
-------
721
slide.
QAMS entered into a dialogue with the organization, and
got them to, in effect, to establish its data quality
constraints. Slide 7 illustrates what we are calling jocularly
the "discomfort curve". The organization defined its action
limit is at 1.2. times the regulatory limit. Below 1.0 it could
tolerate a 10 percent false positive error rate and above 1.4
times the regulatory standard for ambient stream water quality.
It could tolerate a 20 percent false negative error rate. This
input was presented to the statisticians and they developed the
data quality constraints illustrated in slide 8.
The data quality constraints are depicted here as a
"required performance envelope". The envelope is skewed because
the acceptable error rates were skewed, 10 percent acceptable
false positives, 20 percent acceptable false negatives.
Acceptable performance is defined as bounded by the triangle. A
laboratory with zero percent bias can provide the desired data
quality, i.e.,the desired false negative and false positive
rates, with a relative standard deviation ranging as high as 14
percent while a laboratory that has a negative bias of 14 percent
would have to be dead-on in terms of deviation in order to meet
those error bounds. Similarly, a laboratory having positive bias
of 20 percent would also have to have zero percent relative
standard deviation.
We entered into a pilot study with the organization to
establish the actual variability of the participating
laboratories. The essentials of that pilot study are presented
in slide 9. A bulk sample was collected from a stream meeting
the ambient water quality standards. That is to say, the
concentration of copper, iron and zinc were below 5, 300 and 30
micrograms per liter. That bulk sample was divided in half, and
one portion spiked with copper, iron and zinc at the ambient
water quality standard levels. That is to say, if a
concentration of N was contained in the original sample, after
spiking it contained N plus the standard concentration. Both
-------
722
spiked and unspiked samples were submitted to the contract
laboratories as routine samples with no other specific
information provided other than copper, iron and zinc analyses
were desired. Each laboratory received 10 unspiked and 10 spiked
samples.
The results obtained by the participating laboratories
were analized for precision and matrix spike recoveries. The
percents bias and the percents relative standard deviation
achieved bu the participants is presented in slide 10.
Laboratory A had a significant drift in their analytical results
for copper. When graphed chronologically, one can see their
results deteriorating and consequently their copper data couldn't
be used. But looking at these results, one can see that some of
the labs performed acceptably for some of the analytes. None of
the laboratories performed acceptably for all analytes.
Slide 11 compares these performance results with the
required performance envelope. Only STET those three data points
within the triangle are acceptable. The five outside of the
triangle for either reasons of bias or relative standard
deviation would not meet the program's data quality constraints.
Based on those performance results, our statisticians
then calculated the replication plans that would be required to
meet the needed performance constraints. No replication plans
greater than 10, 10 were calculated since clearly such plans
would be economically unworkable. Consequently, you see here 10
plus, 10 plus plans listed for several of the poorer performing
labs.
Slide 13 p[resents how the adjusted results obtained
with various replication plans would compare with the required
performance envelope, if the laboratories continued to perform as
previously indicated. If we one plots the results, adjusted for
matrix spike recovery against the required performance envelope,
the results would all fall on the zero percent bias line. None
of the 10,10 replication plans would result in acceptable
performance from the poorer laboratories, or for outlier
-------
723
analytes.
This is as far as we have gotten with this study. We
have begun the next phase. Those laboratories with performance
requiring 10,10 plans are being drawn into participating. We're
not asking any of the labs to go beyond a 5,5 replication plan;
the program won't pay for it. It's my gut feeling that once
laboratories know that their precision and bias results are
actually being utilized, they will focus on those aspects of
their work and their precision and bias characteristics will
suddenly improve.
The conclusions that we can make from phase 1,
presented on slide 14, are as follows: (1) no single lab is
meeting the data quality constraints for all the analytes, (2)
expect the ILA, the intra-laboratory approach to succeed in
allowing labs with differing performance characteristics to meet
defined data quality goals, in some cases, yes, it's going to be
cost-prohibitive. (3) the program will probably have to relax
its data quality requirements unless it is really willing to bite
the bullet and pay for a significantly greater number of
analyses.
Slide 15 presents some of the planning for Phase 2. We
hope to have all three laboratories participating in it for all
three analytes. The laboratories would monitor and adjust for
performance changes. When we conclude phase 2, we intend to take
a look at the error rates that are incurred with the ILA versus
the error rates that are incurred with the current approach, the
current approach being one matrix spike and duplicate per batch
of 10 or batch of 20, as the case is currently with this program.
Any questions?
-------
724
QUESTION AND ANSWER SESSION
MR. TELLIARD: Any questions for
Fred? Thank you.
(No response.)
-------
725
-------
726
£
O
o
0)
Q,
E
o>
o
0)
DC
O)
c
03
CO
c
o
OS
o
"5.
o
DC
w
o
to
O
CO
-------
727
0)
o
o
o
DC
£
O)
§
o
o
-------
728
O)
CO
O 0)
o °
¥ o
coZT
C .±:
O co
•^ o
03 O
Q.
g
E3.
3
-------
729
Data Quality Objective Process
(Logic Flow)
State problem
Identify a decision that addresses problem I
i
Identify inputs affecting decision
Specify domain of decision
Develop formula for decision-making I
Establish constraints on uncertainty
Optimize design for obtaining data
SLIDE 4
-------
730
Steps of Intralaboratory Approach (ILA)
Step 1: Estimate Variability I
Step 2: Determine Levels of Replication I
Step 3: Perform Analysis I
If precision
has changed
Step 4: Adjust Results
Make Decisions
SLIDE 5
-------
7 31
I 3
si
<0 .£2
QQ
r-
£ •-
gs
O g
—
OS E
.So
o
CM
O
CM
O
CM
O CD
CM £
O ^
oo .2
*" S
t/]
o
to
1 t I [ 1
1 1 1 I
C
s
o g
? o
O
CM
O
O
O
GO
-------
CD
732
o
CD
ST
O
,,^
^*
•—•
CO
2
O
-
-
-
-
-
i i
2 o> o
ii i i i i
-
~
™
^ 2
c
o
in |
T" c
0)
0
c -g
O OJ
•a
T-: o> c
5 2
0^
• O 4->
1~ "^ ^5 f"*
DC 0) w
. . 0) Q
d .2 "
(0
*-•
Is: S
o o
o
o
in
d
(Q ddddddddd'"'
Q
E
CO
D)
O
^^
-8
og
>»=
IE
3"
2§
Q.Z
-------
733
0)
Q.
0)
c
LLJ
0)
O
c
CO
E
0)
Q.
O
2
o
.a
CO
0)
_ IO
in
DC
00
w
CO
CT
(I)
DC
C
O
-------
734
r-
0)
03
f
•M^
Q.
1
"o"
D
E
jrent laboratories
UJ
2
o
CD
O
03
o
H=
CD
Q.
CD
_N
CD
ts
03
C3
6
• •
(0
t>
o
0)
5"
O
<
CD
C
CO
c
03
Q_
.g
"03
"o.
®
Determine i
CD
-»— •
Surface wa
.£
ts
5>
o
c
'N
c:
03
O
CD
Q.
Q.
O
O
c/>
0)
^
"co
^
i
M^
O)
(0
•o
c
CO LO o o
^ 0 CO
.-= °°
D >-
OS
2 8-S.s
13 O-^N
C
0)
5
•<
03
D
CD
03
^
"c
CD
ID
E
03
I
CD
CD
• •
(/>
i Concentratio
w
0)
0
Q.
X
LJJ
0)
•g
03
•a
c
^5
"co
ty standards
03
^
o5
"c
.CD
E
03
"03
CT
LU
GO
O
O)
c
12
'5.
0)
-------
735
CD
O CO
eo.
CD ^
O
C
N
5s
C/>
* I
°2 cd
o" ^
CO ^
LfT irT
8 ' si 3
5 "
formance (^
Copper
in
Z CD"
£
o
2
o
(0
< CD
CO
c\l
3
in
^
CD"
C)
C\J
0
o
l-l
HJ
CO
-------
736
0)
0)
CO
o
o
c
CO
E
0)
a.
o
2
o
CO
•o
o
o
m 2^
as
.1
*-«
IS
a>
oc
- in
^D ^j1 ^^ ^3 CO ^O ^^ ^^ ^3 CO CJ3 ^^ ^^ C3
O4CJCMCMT-1-T-T-T-
0)
-------
737
L-
« 3>
»N ^—
C co
ix cj
gl
^ B
CL o
H
73 »
o t
— en
T^ d>
Q. .*
O '5.
OC w
S 2
o s
0 ii
CE c
00
N
o
o>
Q.
Q.
O
O
ra
o
JD
(0
00
O
OJ
+ +
O O
CD O
-------
738
CM
CO
CL
v\
0)
"5
0)
cc
0)
o
0)
CL
X
LU
0)
Q
O
(0
"~
c
_<0
Q.
o"
10
m
at
CO
m
a)
DC
•J
C/3
.2
A T3
co o)
•»— 4-<
CL O
0) 0)
O CL
o x
< UJ
I I I I I I I I I I I I
in
CMCMCMCM-f-T-i-T-T-
§1
0) CO •*
nr ••^ «*»
-------
739
Conclusions
i
i
CD
(/>
CO
a.
CO
CD
^"*
>*
03
C
03
"o3
£
6
O
Q
"CD
CD
E
c
8
>s
O
s!
o
-Q
JT3
CD
D)
C
"co
o
CD
CD
o
CO
CD
"*_
1-
75
5
_03
0)
C
1
03
_c
-o
CD
%•*
S
r= 03
> Q
> *-«-
c o
c -»-
2 D)
D).£
9 =
n ^
01
J= =?
O
M
i-3
en
-------
740
CM
0
(/)
CO
JBI
Q.
1
>*
TJ
CO
1_
^^
fl
UB
CO
Q
CO
Q.
C
g
"co
.0
"o.
CD
•*-•
^•%
0
CD
Q.
,E
CO
.CD
o
rrt
5
o
JD
JO
-Q
CD
•«
O
CD
CD
CO
precision
c
CO
CD
D)
CO
JZ
o
^
"co
_ZJ
:f5*
\^
co
•D
C
co
i_
0
"c
0
E
CO
CD
^_
O
2?
o
JO
CO
3
__i
v_
^
•Q
CD
CO
to
CD
CD
CO
CO
CD
?
O
v_
LU
•
0
co
o
iZ
O
Q.
CO
"c
CD
i—
•••M
13
o
CD
•£
.0
T3
CD
CO
"•S
CD
CD
i_
CO
CO
CD
^— •
a
E
LU
Q
t-t
hJ
CO
-------
741
MR. TELLIARD: Thank you for your
attention. I'd like to thank the speakers.
Any of you who would like to talk on a subject, please
give me a call or drop me a line for next year. If there's
something you'd like to talk about at one of these meetings and
we haven't talked about it, drop me a line. If there is some
area you think would be interesting either to you or your
colleagues, we'd like to hear from you and see if we can work it
in. We're always looking for new ideas.
I'd like to thank the court reporters for putting up
with all of the strange names.
I'd like to thank Jan. What happened to Jan? Did you
escape again? Jan Sears is the one who puts this thing together;
I'd like to thank Harry McCarty and Dale Rushneck for
working on the speakers' program.
I'd like to thank you for your attention and hopefully
we'll see you next year.
Thank you very much.
-------
14th ANNUAL EPA CONFERENCE
ON ANALYSIS OF POLLUTANTS IN THE ENVIRONMENT
LIST OF SPEAKERS
(Alphabetical Order, Left to Right)
Richard Beach
HYDROSYSTEMS, Inc.
100 Carpenter Drive, Suite 200
Sterling, VA 22170
703-471-9580
Robert Beimer
S-Cubed
3398 Carmel Mountain Road
San Diego, CA 92121
619-587-8448
Merlin K. L. Bicking
Twin City Testing
662 Cromwell Avenue
St. Paul, MN 55114
612-645-3601
William Budde
USEPA, EMSL-Ci
26 W. Martin Luther King Drive
Cincinnati, OH 45268
513-569-7309
Bruce N. Colby
Pacific Analytical
6349 Paseo Del Lago, #102
Carlsbad, CA 92009
619-931-1766
Tudor T. Davies
Director, Office of Science & Tech.
USEPA
401 M Street, S.W. (WH-552)
Washington, DC 20460
202-382-5400
Greg Douglas
Battelle Ocean Sciences
397 Washington Street
Duxbury, MA 02332
617-934-0571
William Eckel
Viar and Company
300 North Lee Street, Suite 200
Alexandria, VA 22314
703-684-5678
Joan Fisk
USEPA - AOB (OS-230), Rm. M-2624
401 M Street, SW
Washington, DC 20460
202-382-3115
Fred Haeberer
USEPA OAMS, RD-680, Room M3828B
401 M Street, SW
Washington, DC 20460
202-382-5785
-------
743
Dr. Ian Hau
Center for Quality and Productivity
University of Wisconsin - Madison
610 Walnut Street
Madison, WI 53705
608-262-9100
Ronald A. Hites
Indiana University - SPEA 410 H
Bloomington, IN 47405
812-855-0193
Jimtnie Hodgeson
USEPA, EMSL-Ci
26 W. Martin Luther King Drive
Cincinnati, OH 45268
513-569-7301
Gary Jackson
Support Systems
3249 Silverthorn Drive
Fort Collins, CO 80526
303-226-3561
Rick Johnson
USEPA, OIRM, MD-34
Research Traingle Park, NC 27711
919-541-1132
Henry Kahn
USEPA, WH-586, Room E731B
401 M Street, SW
Washington, DC 20460
202-382-5406
Peggy Knight
USEPA Region X, Manchester Laboratory
P.O. Box 549
Manchester, WA 98353
206-871-0748
Larry LaFleur
NCASI
720 SW 4th Street
Crovalis, OR 97333
503-752-8801
W. A. Michalik
Shell Oil Company
Route 11 and Madison
Roxanna, IL 62084
Dr. Ileana Rhodes
Shell Development Co.
P.O. Box 1380
Houston, TX 77251-1380
713-493-7702
George H. Stanko
Shell Development Co.
P.O. Box 1380
Houston, TX 77251-1380
713-493-7702
Lance Steere
S-Cubed
3398 Carmel Mountain Road
San Diego, CA 92121
619-587-8448
-------
744
William A. Telliard
Chief, Analytical Methods Staff
USEPA, EAD
401 M Street, S.W. (WH-552)
Washington, DC 20460
202-382-7131
Ramona Trovato
USEPA, H1503, Room W331
401 M Street, SW
Washington, DC 20460
202-382-7077
Charles White
USEPA, WH-586, Room E731C
401 M Street, SW
Washington, DC 20460
202-382-5371, 5411
Margaret E. Wickham St. Germain
Midwest Research Institute
425 Volker Boulevard
Kansas City, MO 64110-2299
816-753-7600 (x8420 FAX)
-------
745
14th ANNUAL EPA CONFERENCE
ON ANALYSIS OF POLLUTANTS IN THE ENVIRONMENT
LIST OF ATTENDEES
(Alphabetical Order, Left to Right)
Pilar Advani
Laboratory Supervisor/Chemist
Rahway Valley Sewerage Authority
1050 East Hazelwood Avenue
Rahway, NC 07065
908-388-0868
Don Anne
Supervising Chemist
McLaren/Hart Environmental Engr. Corp.
28 Madison Avenue Extention
Albany, NY 12203
518-869-6192
Nasim H. Ansari
QA Officer
City of Kalamazoo
1415 N. Harrison St.
Kalamazoo, MI 49007
616-385-8157
Merrill Ashcraft
Chemist
NAVFAC, Navy Public Works
Utilities Dept., Bldg. P-71 - Code 600
Norfolk, VA 23511-6098
804-455-1789
Steve Azar
Supervisor/Environmental Engineer
Atlantic Division, Naval Facilities
Engineering Command - Code 09G1
Department of the Navy
Norfolk, VA 23511-6287
804-444-9718
Shakoora Azimi
Operations Manager
McLaren/Hart
11101 White Rock Road
Rancho Cordova, CA 95670
916-638-3696
Mindy Baldwin
Manager, Organics Analysis
Environmental Labs
9211 Burge Avenue
Richmond, VA 23237
804-271-2440
Louis B. Barber
Chief Chemist
City of Richmond, Public Utilities
1400 Brander Street
Richmond, VA 23224
804-780-5338
Harriet L. Beazley
Supervisor Environmental Lab
Babcock and Mil cox Company
P.O. Box 11165
Lynchburg, VA 24506
804-522-5130
R. A. Bechtold
Engineer
Westinghouse Hanford
MSIN: H4-55, P.O. Box 1970
Rich land, WA 99352
509-376-9017
-------
746
Lisa Benya
Analytical Chemist
American Analytical Laboratories, Inc.
840 South Main St.
Akron, OH 44311-1516
216-535-1300
John Bernatchy
NET Pacific, Inc.
4224 Campus Point Ct., Suite 100
San Diego, CA 92121
619-535-7509
Mark R. Bero
CLP Manager
IEA, Inc.
3000 Weston Parkway
Gary, NC 27513
919-677-0090
P. Mac Berthouex
University of Wisconsin - Madison
61Q Walnut Street
Madison, WI 53705
608-262-7248
Patricia Bickford
Administrator II
NH Dept. of Environmental Services
6 Hazen Drive
Concord, NH 03301
603-271-2841
Daniel Biggerstaff
Technical Director
Trident Labs
125 Wagon Trail Road
Ladson, SC 29456
803-871-4999
Dan Bolt
Environmental Products Manager
Cambridge Isotope Laboratories, Inc.
30 Commerce Way
Woburn, MA 01801
617-938-0067
James S. Bookwalter
Laboratory Technician
Newport News Shipbuilding
4101 Washington Avenue, Dept. 031
Newport News, VA 23607
804-380-7744
Robert L. Booth
USEPA, Region V
26 W Martin Luther King Dr.
Cincinnati, OH 45268
513-565-7362
Brett L. Bordelon
Scientist
ENSECO/CAL
2544 Industrial Blvd.
W. Sacramento, CA 95691
916-372-1393
James Boyle
Chemist II
City of Cincinnati - Sewer Dept
1600 Gest Street
Cincinnati, OH 45204
513-244-5115
Joel C. Bradley
Cambridge Isotope Labs
30 Commerce Way
Woburn, MA 01801
617-938-0067
-------
747
Steve Brady
Staff Chemist
Eastman Kodak Company
Kodak Park, CQS, B-34
Rochester, NY 14652
716-722-3430
Marielle Brinkman
Technician
Battelle
505 King Avenue
Columbus, OH 43201
614-424-5277
Patricia M. Brown-Derocher
AT Kearney
One Wall Street Court, Suite 330
New York, NY 10005
212-425-5470
Nancy A. Broyles
Advanced Chemist
Union Carbide Chemicals and Plastics Co.
3200-3300 Kanawha Turnpike
South Charleston, WV 25303
304-747-4707
Thomas A. Buedel
Nine Miles Services
P.O. Box 12104
Research Triangle Park, NC 27709
919-544-5600
Anne S. Burnett
Quality Assurance Officer
Environmental Testing Services, Inc.
816 Norview Avenue
Norfolk, VA 23509
804-853-1715
Tony Burns
S-CUBED Division, Maxwell Labs,
3398 Carmel Mountain Road
San Diego, CA 92121
619-453-0060
Inc.
Robert G. Butz, Ph.D.
Senior Chemist
JSCF, Inc.
1015 15th St., NW, Suite 500
Washington, DC 20005
202-789-3317
Jennings R. Byrd
Materials Engineer
Maryland Dept. of Transportation
2323 West Joppa & Falls Road
Brooklandville, MD 21022
301-321-3536
William T. Castle
Agricultural Chemist III, Specialist
California Department of Fish & Game
2005 Nimbus Road
Rancho Cordova, CA 95670
916-355-0856
Hank Chambers
Keystone/NEA
12242 SW Garden Place
Tigard, OR 97223
503-624-2773
Lloyd Cheong
Environmental Chemist
Syntex Pharmaceutical Ltd.
P.O. Box F2403
Freeport, Bahamas,
809-352-8171
-------
748
Linda Chicquette
Environmental Chemist
The Boeing Company
P.O. Box 3707, M/S 6301
Seattle, WA 98124
206-237-9228
Elizabeth A. Chisholm
Laboratory Manager
ELI ECO LOGIC INTERNATIONAL, Inc.
143 Dennis St.
Rockwood
Ontario, Canada, NOB2KO
519-856-9591
Ray Christopher
Marketing Manager
Finnigan Corporation
355 River Oaks Parkway
San Jose, CA 95132
408-433-4800
Ida Church
Physical Science Technician
NAVFAC, Navy Public Works
Utilities Dept., Bldg. P-71 - Code 600
Norfolk, VA 23511-6098
804-455-1789
Dr. Eugene Cioffi
Director
Crystal Springs Laboratory
178 Bridge St., Rear
Groton, CT 06340
203-445-1751
Arthur E. Clark
USEPA
60 Westview Street
Lexington, MA 02173
617-860-4374
David F. Clark
Section Chief
VA/DGS Div. of Consolidated Lab Services
1 North 14th Street
Richmond, VA 23219
804-786-8312
John Clayton
Organics Manager
Betz Labs
9669 Grogan Mills Road
The Woodlands, TX 77380
713-367-6201
Martin Collamore
Laboratory Supervisor
City of Tacoma
2201 Portland Ave.
Tacoma, WA 98421
206-591-5588
Jayme Connolly
Associate Environmental Chemist
ABB Environmental Services, Inc.
261 Commercial Street, P.O. Box 7050
Portland, ME 04112
207-7752-5401
William Corl
Chemist
NAVFAC, Navy Public Works
Utilities Dept., Bldg. P-71 - Code 600
Norfolk, VA 23511-6098
804-455-1789
B. Rod Corrigan
Quality Assurance Officer
Environmental Consultants, Inc.
391 Newman Ave
Clarksville, IN 47129
812-282-8481
-------
749
Richard L. Cotter
Research Scientist
Millipore Waters
34 Maple Street
Milford, MA 01757
508-478-2000
Carolyn J. Covey
Assistant to the Executive Director
International Association of
Environmental Testing Laboratories
1911 N. Fort Myer Dr.
Arlington, VA 22209
703-524-2427
David Crane
Agricultural Chemist III (Specialist)
California Department of Fish & Game
2005 Nimbus Road
Rancho Cordova, CA 95670
916-355-0856
Mark Crews
Viar and Company
Sample Control Center
300 N. Lee Street, Suite 200
Alexandria, VA 22314
703-557-5040
John P. Criscio
President
Absolute Standards Inc.
498 Russell St.
New Haven, CT 06513
203-468-7407
Kelly B. Cruey
Volatiles Analyst
Enviro-Tech Mid-Atlantic
1861 Pratt Drive
Blacksburg, VA 24060
703-231-3983
Johanna M. Culver
Chemist
Norfolk Naval Shipyard
Quality Assurance Office, Code 130
Portsmouth, VA 23709
804-396-9307
Michael A Cunningham
Environmental Engineer
Pennzoil Company
P.O. Box 2967
Houston, TX 77252-2967
713-546-6149
Doris F. Curry
Group Leader GL/MS
Eckenfelder, Inc.
227 French Landing Drive
Nashville, TN 37228
615-255-2288
Beth Curtis
Chemist
U. S. Army Environmental Hygiene Agency
Building E2100
Aberdeen Proving Ground, MD 21010
301-671-2208
Thomas L. Dawson
Group Leader
Union Carbide Chemicals and Plastics Co.
3200-3300 Kanawha Turnpike
South Charleston, WV 25303
304-747-5711
Margie Delashmit
Chemist
Indiana Dept. of Environmental Mgmt.
5550 W. Bradbury
Indianapolis, IN 46241
317-243-5166
-------
750
Ivan B. DeLoatch
Chemist
USEPA, GWDW
401 M Street, SW (WH-550D)
Washington, DC 20460
202-382-2273
Michael Deufemia
Env i ronmenta1 Spec i a 1i st
McLaren/Hart Environmental Engr. Corp.
28 Madison Avenue Extention
Albany, NY 12203
518-869-6192
Kathy J. Dien Hillig
Mgr., Ecology Analytical Services
BASF Corporation
1609 Biddle Avenue
Wyandotte, MI 48192
313-246-6334
Twila Dixon
Assistant Lab Manager
RMC - Environmental Services
88 Robinson Street
Pottstown, PA 19464
215-327-4850
James J. Dodson
Laboratory Manager
Environmental Testing Services, Inc.
816 Norview Avenue
Norfolk, VA 23509
703-853-1715
Willard L. Douglas, Ph.D.
Sverdrup Technology, Inc.
John C. Stennis Space Center
Bldg. 2423
SSC, MS 39529
601-688-3158
Art Driedger
Wayne Analytical & Envrn. Services
992 Old Eagle School Road
Wayne, PA 19087
215-688-7485
Joshua Dubnick
Principal Lab Technician
Bergen County Utilities Authority
P.O. Box 122
Little Ferry, NJ 07643
201-807-5853
Rolla M. Dyer
University of Southern Indiana
8600 University Blvd.
Evansville, IN 47712
812-464-1701
Dave Edelman
Lab Manager
Columbia Analytical Services
1317 South 13th Avenue
P.O. Box 479
Kelso, WA 98626
206-577-7222
Kenneth W. Edgell
Section Chief
The Bionetics Corporation
16 Triangle Park Drive
Cincinnati, OH 45246
513-771-0448
Nariman Elfino
All-Organic Chemist
Froehling & Robertson, Inc.
3015 Dumbarton Rd., P.O. Box 27524
Richmond, VA 23261-7524
804-264-2701
-------
751
Edward B. Engel
DeYor Laboratories, Inc.
7655 Market Street, Suite 2500
Youngstown, OH 44512
216-758-5788
Paul S. Epstein
NSF International
3475 Plymouth Rd.
Ann Arbor, MI 48105
313-768-8010
Valerie Evans
Assistant Product Manager
Triangle Laboratories, Inc.
801-10 Capitola Drive
Durham, NC 27713
919-544-5729
Mary Fencl
ENSR Consulting & Engineering
35 Nagog Park
Acton, MA 01720
508-635-9500
Sam Ferro
Laboratory Manger
Kansas City Testing Lab
1669 Jefferson
Kansas City, MO 64108
816-842-7350
Gary Folk
IEA Inc.
3000 Weston Parkway
Cary, NC 27513
919-677-0090
Carleton Fong
Senior Chemist
McLaren/Hart
11101 White Rock Road
Rancho Cordova, CA 95670
916-638-3696
James Forbes
Laboratory Director
Law Environmental Inc
112 TownPark Drive
Kennesaw, GA 30144
404-421-3310
Peter Fowlie
Chief, Laboratory Division
Wastewater Technology Centre
867 Lakeshore Rd., P.O. Box 5050
Burlington
Ontario, Canada, L7R4A6
416-336-4633
Kitty Gallagher
Virginia Institute of Marine Science
College of William & Mary
Toxicology and Chemistry
Gloucester Point, VA 23062
804-642-7228
Linda J. Garner
Hill Air Force Base
00-ALC/TIELC
Hill Air Force Base, UT 84056
801-777-2302
John Garrett
Wright State University
Colonel Glenn Highway
Dayton, OH 45435
513-873-2202
-------
752
Kirby Garrett
Organic Department Supervisor
ENSECO-CRL
2810 Bunsen Avenue, Unit A
Ventura, CA 93003
805-650-0546
Denise S. Geier
Laboratory Director
Analytical Services, Inc.
390 Trabert Avenue, NW
Atlanta, GA 30309
404-892-8144
Michael G. Goergen
President
Fire Environmental Consulting Lab., Inc.
1451 E. Lansing Drive, Suite 222
East Lansing, MI 48823
517-332-0167
Dean Gokel
President
GeoChem, Inc.
2500 Gateway Centre Blvd., Suite 300
Morrisville, NC 27560
919-4608093
Stephen Green
Specialist
Toyota Motor Manufacturing, USA,
1001 Cherry Blossom Way
Georgetown, KY 40324
502-868-2533
Inc.
Zoe A. Grosser
Environmental Marketing
The Perkins-Elmer Corporation
761 Main Avenue, M/S 219
Norwalk, CT 06859
203-834-6874
H. Markus Gudnason
Director SASD
Pace, Inc.
1710 Douglas Drive N.
Minneapolis, MN 55443
612-525-3467
John Gumpper
Section Manager
DataChem Laboratories
960.West LeVoy Drive
Salt Lake City, UT 84123
801-266-7700
Jennie Gunderson
Virginia Institute of Marine Science
College of William & Mary
Toxicology and Chemistry
Gloucester Point, VA 23062
804-642-7228
John Gute
Lab Supervisor
LA County Sanitation District
1965 Workman Mill Road
Whittier, CA 90601
213-699-0405
David Haddaway
Senior Chemist
City of Portsmouth
Lake Kilby Water Treatment Plant
105 Maury Place
Suffolk, VA 23434
804-539-7608
Danny B. Hale
President
Specialized Assays, Inc.
210 12th Avenue South, P.O. Box 25110
Nashville, TN 37202
615-255-5786
-------
753
Robert C. Hale
Virginia Institute of Marine Science
College of William & Mary
Toxicology and Chemistry
Gloucester Point, VA 23062
804-642-7228
Guy J. Hall
President
Environmental Testing Services, Inc
816 Norview Avenue
Norfolk, VA 23509
703-853-1715
Jeanne Hankins
USEPA
401 M Street, SW
Washington, DC 20460
Kim G. Hanzelka
Chemist
Martin Marietta Energy Systems, Inc.
P.O. Box 2009, Bldg. 9115, M/S 8219
Oak Ridge, TN 37831
615-574-1599
Bill Hardesty
Chemist
Viking Instruments Corp.
12007 Sunrise Valley Drive
Reston; VA 22091
703-450-5183
William C. Harris
Group Leader Analytical Services
Union Camp Corporation
P.O. Box 178
Franklin, VA 23851
804-569-4263
Riaz-ul-Hasan
Supervisor
Bergen County Utilities Authority
Foot of Mehrhof Rd., P.O. Box 122
Little Ferry, NJ 07643
201-807-5855
David Haske
Analytical Chemist
Roche Analytical Laboratory
1415 Rhoadmiller Street
Richmond, VA 23220
804-353-8973
Christopher A. Heltzel
Western - Esat Program
2890 Woodbridge Avenue, Bldg. 209
Edison, NJ 08837
908-548-1024
Scott Henderson
Statistician
SAIC
8400 Westpark Drive
McLean, VA 22102
703-821-4672
Mike Heniken
City of Columbus
Div. of Sewage & Drainage, Lab.
900 Dublin Rd.
Columbus, OH 43215
614-645-7016
Michael Herbert
Baxter Healthcare Corporation
Route 120 & Wilson Road
Round Lake, IL 60073-0490
708-546-6311 ext. 6596
-------
754
Joseph F. Hill, Jr.
Chemist
Lancaster Laboratories,
2425 New Holland Pike
Lancaster, PA 17601
717-656-2301
Inc.
Geoff Hinshelwood
Organic Chemist
Jennings Lab
1118 Cypress Avenue
Virginia Beach, VA 23451
804-425-1498
Chris Hoi brook
Specialist
Toyota Motor Manufacturing, USA, Inc.
1001 Cherry Blossom Way
Georgetown, KY 40324
502-868-2532
Dr. Philip Holt
Occidental Chemical
2801 Long Road
Grand Island, NY 14072
716-773-8538
Ben Honaker
USEPA, EAD (WH-552)
401 M Street, SW
Washington, DC 20460
202-382-2272
Dr. Lyman H. Howe, III
Tennessee Valley Authority
Water Quality Dept., 401 Chestnut St.
1101 Market Street
Chattanooga, TN 37402-2801
615-751-3711
IA
Dean Howe11
Naval Supply Center
Fuel Department, Code 700
Norfolk, VA 23512-5100
804-444-2761
Han-Ping Huang
Chief Chemist
James R. Reed & Associates,
813 Forrest Drive
Newport News, VA 23606
804-599-6750
Inc.
Frank H. Hund
Chemist
USEPA, EAD/OWRS
401 M Street, SW
Washington, DC 20460
202-382-7182
John Huntington
President
Phoenix Analytical Laboratories, Inc.
3401 Industrial Lane
Broomfield, CO 80020
303-469-1101
Phyllis Huntington
Vice President
Phoenix Analytical Laboratories,
3401 Industrial Lane
Broomfield, CO 80020
303-469-1101
Inc.
Nang Huynh
Laboratory Manager
National Laboratory, Inc.
3210 Claremont Ave.
Evansville, IN 47712
812-464-9000
-------
755
Patti Isaacs-Hansen
Environmental Chemist
Medlab Environmental Testing, Inc.
212 Cherry Lane
New Castle, DE 19720
302-655-5227
Christopher S. Jones
Laboratory Manager
Serco Laboratories
1931 West County Road C-2
St. Paul, MN 55113
618-636-7173
Francis M. Jungfleisch
Westinghouse Handford Company
P.O. Box 1970
Richland, WA 99353
509-373-5774
Victor F. Kalasinsky
Chief, Div. of Environmental Toxicology
Armed Forces Institute of Pathology
14th Street & Alaska Avenue
Washington, DC 20306
202-576-2434
Laine Kasdras
6C/MS Supervisor
RMC Environmental Services
88 Robinson Street
Pottstown, PA 19464
215-327-4850
Lawrence H. Keith
Senior Program Manager
Radian Corporation
P.O. Box 201088
Austin, TX 78720-1088
512-454-4797
R. Michael Kennedy
Laboratory Supervisor
City of Red Hill, SC/Env Man Laboratory
P.O. Box 11706
Rock Hill, SC 29731-1706
803-329-8704
Jackie B. Key
Environmental Scientist III.
MS Dept. of Environmental Quality - OPC
121 Fairmont Plaza
Pearl, MS 39208
601-961-5183
Mary Khali!
Metropolitain Water Reclamation
District of Chicago,
550 South Meacham Rd.
Schaumburg, IL 60193
708-529-7700 x 281
Jennifer Kiesel
Laboratory Manager
Kiesel Environmental Laboratories
4801 Fyler Avenue
St. Louis, MO 63196
314-351-5500
Jim King
Project Manager
Viar and Company
Sample Control Center
300 N. Lee Street, Suite 200
Alexandria, VA 22314
703-557-5040
David Kirk
Research Chemist
Metropolitan Sewer District
4522 Algonquin Parkway
Louisville, KY 40211
502-540-6735
-------
756
William Kirk
President
Reliance Laboratories, Inc.
P.O. Box 625
Bridgeport, WV 26330-0625
304-842-5285
Dewey R. Klahn
Chief Chemist
Environmental Science Corp.
1910 Mays Chapel Rd.
Mt. Juliet, TN 37122
615-758-5858
Keith Kline
Assistant Mgr. Environmental Lab
ATEC Accociates
5150 East 65th Street
Indianapolis, IN 46220
317-849-4990
Chris Klopp
Chemist III
Wisconsin ONR
P.O. Box 7921
Madison, WI 53707
608-767-0860
David A. Kovacs
ManTech Environmental Technology, Inc.
Kerr Lab Road, RSKERL
Ada, OK 74821
405-332-8800
S. Krishnamurthy
Chemist
USEPA, Releases Control Branch
Region II
2890 Woodbridge Avenue
Edison, NJ 08837
908-321-6796
Dr. William 6. Krochta
Manager Environmental
PPG Industries, Inc., Chemical Group
440 College Park Drive
Monroeville, PA 15146
412-325-5183
Joe Kurek
Chief Chemist
EMS-Heritage Laboratories
7901 W. Morris Street
Indianapolis, IN 46231
317-243-0811
Kenneth T. Lang
US Army Toxic & Hazardous Materials Ag.
Attn: CETHA-TS-C
Aberdeen Proving Ground, MD 21010-5401
301-676-7569
Ivette Larralde
Savannah Laboratories
414 SW 12 Avenue
Deerfield Beach, FL 33442
305-421-7400
Janet M. Lee
Monsanto Company
800 N. Lindbergh Blvd.
St. Louis, MO 63167
314-694-1273
Dr. Marguerite L. Leng
Consultant
Leng Associates
1714 Sylvan Lane
Midland, MI 48640
517-832-2624
-------
757
Solomon W. Leung
Senior Research Engineer
Association of American Railroads
50 F Street, NW
Washington, DC 20001
202-639-2280
H. Nathan Levy, III
President
Analytical & Environmental Testing,
1717 Seabord Drive
Baton Rouge, LA 70810
504-769-1930
Inc.
James W. Lewis
Laboratory Projects Manager
Bionetics Analytical Labs
18 Research Drive
Hampton, VA 23666
804-865-0880
Laurence Libelo
VIMS
Gloucester Point, VA 23032
804-642-9164
William Lock
Vice President/Lab Manager
Gascoyne Labs, Inc.
2101 Van Deman Street
Holabird Industrial Park
Baltimore, MD 21224-6697
301-285-8510
Jon W. Lodge
Supervisor
Research Triangle Institute
P.O. Box 12194
Research Triangle Park, NC 27709
919-541-6905
Lazaro Lopez
Associate Director
Suburban Laboratory, Inc.
4140 Litt Drive
Hillside, IL 60162
708-544-3260
Norman Low
Environmental Product Manager
Hewlett Packard
1601 California Avenue
Palo Alto, CA 94304
415-857-7381
Ted W. Lufriu
Chesapeake Analytical Lab., Inc.
106 A Rockefeller Court
Waldorf, MD 20602
301-932-4775
Chung-Rei Mao
Chemist
Corps of Engineers
12565 West Center Road
Omaha, NE 68144
402-221-7494
Jim Maquire
Roche Analytical Laboratory
1415 Rhoadmiller Street
Richmond, VA 23220
804-353-8973
Michael F. Martin
VA/DGS Div of Consolidated Lab Services
1 North 14th Street
Richmond, VA 23219
804-786-8365
-------
758
R. L. Matsushima
Environmental Manager
Colorado Refining Company
5800 Brighton Blvd.
Commerce City, CO 80022
303-295-4500
Scott Mayo
NUS Corporation
900 Gemini Avenue
Houston Analytical Laboratory
Houston, TX 77058
713-488-1810
Harry McCarty
Senior QA Chemist
Viar and Company
Sample Control Center
300 N. Lee Street, Suite 200
Alexandria, VA 22314
703-557-5040
Robert J. McDaniel, II
Applied Marine Research Lab
Old Dominion University
1034 West 45th St.
Norfolk, VA 23529
804-683-4195
Neal A. A. McNeil!
Chemist
Newport News Shipbuilding
4101 Washington Avenue, Dept. 031
Newport News, VA 23607
804-380-7744
Thomas McVicker
Organic Section Manager
Gascoyne Labs, Inc.
2101 Van Deman Street
Holabird Industrial Park
Baltimore, MD 21224-6697
301-285-8510
Eric Melanson
GCMS Specialist
Varian Instrument Groups
105 Duchess Place
North Wales, PA 19454
215-368-9353
John P. Melvin
President
PEL, Inc.
9405 S. W. Nimbus Avenue
Beaverton, OR 97005
503-671-0885
Carol Meyer
Research Scientist
New York Dept. of Health
Wadsworth Center for Labs and Research
Empire State Plaza, Box 509
Albany, NY 12201-0509
518-486-5670
Ronald Michand
Supervising Chemist
CT Dept. of Health Services Laboratory
10 Clinton Street
Hartford, CT 06106
203-566-3802
Patricia Miles
Chemist
Norfolk Naval Shipyard
Quality Assurance Office, Code 130
Portsmouth, VA 23709
804-396-9307
Harold W. Miller
Department of the Navy
Atlantic Division
Naval Facilities Eng. Command
Norfolk, VA 23511-6287
804-445-7336
-------
759
Kevin H. Miller
Environmental Chemist
NC Division of Environmental Management
1424 Carolina Avenue
Washington, NC 27889
919-946-6481
Toney G. Miller
Analytical Chemist
Amoco Corp., Groundwater Management
P.O. Box 3385
Tulsa, OK 74102
918-660-4447
Ray Mindrup
Supelco, Inc.
Supelco Park
Bellefonte, PA 16823
814-359-3441
Lisa Mitas
Chemist
Texaco, Inc.
329 North Durfee Ave
South El Monte, CA 91733
213-699-0948
Gregory B. Mohrman
Chief, Laboratory Support Division
Program Manager Rocky Mountain Arsenal
AMXRM-LS
Dept. of the Army
Commerce City, CO 80022
303-289-0217
R. Charles Morgan
Battelle Ocean Sciences
2101 Wilson Blvd., Suite 800
Arlington, VA 22201-3008
703-875-2947
Robert Mothershead
Virginia Institute of Marine Science
College of William and Mary
Toxicology and Chemistry
Gloucester Point, VA 23062
804-642-7228
Anson Moye
University of Florida
Pesticide Research Laboratory
Gainsville, FL 32611
904-392-1978
Robert Murphy
Technical Coordinator
Technical Testing Laboratories
1256 Greenbrier Street
Charleston, WV 25311
304-346-0725
Violetta F. Murshak
Vice President
Fire Environmental Consulting Lab., Inc.
1451 E. Lansing Drive, Suite 222
East Lansing, MI 48823
517-332-0167
Harry V. Myers
Senior Project Manager
Keystone Environmental Resources, Inc.
3000 Tech Center Drive
Monroeville, PA 15146
412-825-9818
Deborah Nelson
Environmental Lab Specialist
HRSD
1436 Air Rail Avenue
Virginia Beach, VA 23452
-------
760
Gordon M. Nelson
Chemist
Norfolk Naval Shipyard
Quality Assurance Office, Code 130
Portsmouth, VA 23709
804-396-9307
Alan Newman
Associate Editor
Environmental Science & Technology
1155 16th Street, NW
Washington, DC 20036
202-872-6069
Becky Newman
County Court Reporters,
124 East Cork Street
Winchester, VA 22601
703-667-6562
Inc.
Jeffrey L. Nielsen
City of Tallahassee/Water Quality Lab
3805 Springhill Road
Tallahassee, FL 32310
904-575-5907
Natasha Nimo
Hydrosystems, Inc.
100 Carpenter Drive, Suite 200
Sterling, VA 22170
703-471-9589
Babu Nott
EPRI
3412 Hi 11 view Avenue
Palo Alto, CA 94303
415-855-7946
Greg O'Neil
Analyzer Marketing Manager
Tekmar Company
P.O. Box 429576
Cincinnati, OH 45242
513-247-7000
Alicia P. Ordona
QA Analyst
Commonwealth of VA/Dept. Gen. Services
1 N 14th St
Richmond, VA 23219-3691
304-786-3411
Nancy Osterhoudt
ERCE
3211 Jermantown Road
P.O. Box 10130
Fairfax, VA 22030
703-246-0334
Charles Overpeck
Office of Environmental Management
Operations Bldg.
Municipal Center
Virginia Beach, VA 23456
804-427-4801
James G. Overpeck
1606 Woodmoor Lane
McLean, VA 22101
713-448-7614
Nancy Owen
Chemist
ETS Analytical Services
1401 Muncipal Road
Roanoke, VA 24012
203-365-0004
-------
761
Robert G. Owens
Analytical Services, Inc.
390 Trabert Avenue, NW
Atlanta, GA 30309
404-892-8144
James Oyler
Analyst
Enviro-Tech Mid-Atlantic
1861 Pratt Drive
Blacksburg, VA 24060
703-231-3983
Kurt C. Picel
Assistant Scientist
Argonne National Laboratory
9700 S. Cass Avenue
Argonne, IL 60439
708-972-4018
Marvin D. Piwoni
Laboratory Manager
Hazardous Waste Research & Info. Center
One E. Hazelwood Drive
Champaign, IL 61820
217-333-8724
Rebecca Plemons
Lab Manager
Reliance Laboratories
P.O. Box 625
Bridgeport, WV 26330
304-842-5285
Roy W. Plunkett, Jr.
Analytical Chemist Supervisor
VA/D6S Div. of Consolidated Lab Services
1 North 14th Street
Richmond, VA 23219
804-225-4007
Ruggero Pocci
R & D Chemist
Varian Sample Preparation Products
24201 Frampton Avenue
Harbor City, CA 90710
213-539-6490
Steven R. Pond
Environmental Laboratories, Inc.
9211 Burge Avenue
Richmond, VA 23237
804-271-3440
Richard Potts
President
A. A. Labs, Inc.
P.O. Box 749
Plainsboro, NJ 08536
609-799-8787
Gregory E. Pronger
Technical Director, Organic Labs
NET - Midwest
850 W. Bartlett Road
Bartlett, IL 60103
708-289-7333
Gilberto Quintero
Analytical Chemist
Tennessee Valley Authority
401 Chestnut St., Suite 150
Chattanooga, TN 37402
615-751-3705
Sohail Qureshi
All-Organic Chemist
Froehling & Robertson, Inc.
3015 Dumbarton Rd., P. 0. Box 27524
Richmond, VA 23261-7524
804-264-2701
-------
762
Maryfrances Ran
Staff Chemist
Dames & Moore
2025 First Avenue #500
Seattle, WA 98121
206-728-0744
Marline Raphael
Ohio EPA
1030 King Avenue
Columbus, OH 43212
614-294-5841
William G. Rauch
Technical Director
KAR Laboratories, Inc.
4425 Manchester Road
Kalamazoo, MI 49002
616-381-9666
Katharine M. Raynor
Director, QA Division
Naval Supply Center
Fuel Department, Code 700
Norfolk, VA 23512-5100
804-444-2761
Brian Reddy
Advanced Cleanup Technologies
200 South Service Road
Roslyn Heights, NY 11577
516-625-1860
Stephen E. Reeves
Chemist
Union Camp Corporation
P.O. Box 178
Franklin, VA 23851
804-569-4891
Stephen Remaley
Chemist
USEPA, Region IX
75 Hawthorne Street (P-3-2)
San Francisco, CA 94105
415-744-1527
Lynn Riddick
Viar and Company
Sample Control Center
300 N. Lee Street, Suite 200
Alexandria, VA 22314
703-557-5040
Anita Rigassio
Environmental Chemist
COM Federal Programs Corporation
98 N. Washington Street, Suite 200
Boston, MA 02114
617-742-2659
Major Douglas S. Rinehart
U.S. Army Environmental Hygiene Agency
Bldg. E 2100
Aberdeen Proving Ground, MD 21010-5422
301-671-3739
John Rissel
Laboratory Manager
CIBA-GEIGY Corporation
P.O. Box 71
Toms River, NJ 08754
908-349-5200
Rex Robinson
Senior Chemist
Metro Environmental Laboratory
322 West Ewing
Seattle, WA 98119
206-684-2362
-------
763
Alice J. Rose
Maryland Medical Lab, Inc.
1901 Sulphur Spring Road
Baltimore, MD 21227
301-536-1546
Nancy C. Rothman, Ph.D.
Enseco, Inc.
205 Alewife Brook Parkway
Cambridge, MA 02138
617-661-3111
Anna M. Rule
Chief Laboratory Division
Hampton Roads Sanitation District
P.O. Box 5000
Virginia Beach, VA 23455
804-460-2261
Dale R. Rushneck
ATI - CO
225 Commerce Drive
Ft. Collins, CO 80524
303-490-1522
Mary Rybitski
Virginia Institute of Marine Science
College of William & Mary
Toxicology and Chemistry
Gloucester Point, VA 23062
804-642-7228
Sue Salkin
QC Manager
Analytics Laboratory
1415 Rhodemiller St.
Richmond, VA 23220
804-353-8973
Gwendolyn C. San Agustin
Research Associate
New Jersey Institute of Technology
138 Warren Street, ATC Bldg.
Newark, NJ 07102
201-596-5857
John Scalera
Chemist
USEPA, Tox. Sub.\Exposure Eval. Div.
Field Studies Branch (TS 798)
401 M Street, SW
Washington, DC 20460
202-475-6709
Aisling Scallan
Product Manager
EnSys, Inc.
P.O. Box 14063
Research Triangle Park, NC 27709
919-941-5509
Robert B. Schaffer
Manager of Environmental Services
ERCE
3211 Jermantown Road
P.O. Box 1030
Fairfax, VA 22030
703-246-0274
Bill Schnute
Finnigan Corporation
355 River Oaks Parkway
San Jose, CA 95134
408-433-4800
Cindy Schreyer
Viar and Company
Sample Control Center
300 N. Lee Street, Suite 200
Alexandria, VA 22314
703-557-5040
-------
764
Delwyn K. Schumacher
Group Leader
Lancaster Laboratories, Inc.
2425 New Holland Pike
Lancaster, PA 17601
717-656-2301
Edward J. Schwarz
Supervisor Environmental Laboratory
Campbell Soup Company
1 Campbell Place
P.O. Box 44W
Camden, NJ 08101
609-342-4800 ext.2057
Jan Sears
Project Manager
ERCE
3211 Jermantown Road
P.O. Box 10130
Fairfax, VA 22030
703-246-0306
Shawn M. Shanmugan
Oak Ridge Associated Universities
P.O. Box 117
Oak Ridge, TN 37831-0117
615-576-0048
Raghu Sharma
Department of Public Works, Lab
2700 S. Belmont Avenue
Indianapolis, IN 46221
317-633-5429
Frederick J. Shaw
Analytical Chemist
Martel Laboratory Services
1025 Cromwell Bridge Road
Baltimore, MD 21014
301-825-7790
Peter Shen
Quality Assurance Laboratory
6605 Nancy Ridge Drive
San Diego, CA 92121
619-552-3636
Barry L. Silver
Technical Sales Manager
AnalytiKEM
28 Sprinsdale Road
Cherry Hill, NJ 08003
609-751-1122
Carl H. Simmons
Chemist II
EA Laboratories
19 Loveton Circle
Sparks, MD 21152
301-771-4920
Kate Simmons
Tighe & Bond, Inc.
53 Southampton Road
Westfield, MA 01085
413-562-1600
David W. Singer
Tekmar Company
111 Chesterfield Avenue
Centreville, MD 21617
301-758-3953
Peggy Sleevi
Director of Quality Assurance
Enseco Incorporated
2612 Olde Stone Road
Midlothian, VA 23113
804-378-1851
-------
765
Carol F. Smith
Environmental Quality Manager
SC Department of Health & Envrn. Control
P.O. Box 72
State Park, SC 29147
803-935-7025
Cynthia Smith
Chemist
Mantech Environmental
#2 Triangle Drive
P.O. Box 2313
Research Triangle Park, NC 27709
919-549-0611
Dr. James S. Smith
President/Chemist
Trillium, Inc.
7A Grace's Drive
Coatesville, PA 19320
215-383-7233
Terry Smith
Manager Organic Analysis
USPCI Analytical Services
4322 South 49th West Avenue
Tulsa, OK 74107
918-445-1162
Curt G. Spear
Laboratory Supervisor
Newport News Shipbuilding
4101 Washington Ave., Dept. 031
Newport News, VA 23607
804-380-3244
Dr. J. G. H. Stafford
VG Laboratory Systems
Number 1, St. Georges Ct.
Hanover Business Park
Altrincham, Cheshire, UK WA145TP
(0) 61 941 6159
Sally S. Stafford, Ph.D.
Senior Applications Chemist
Hewlett-Packard Company
P.O. Box 900
Avondale, PA 19311-0900
215-268-5444
Lisa Stamper
Analytical Chemist
American Analytical Laboratories, Inc.
840 South Main Street
Akron, OH 44311-1516
216-535-1300
Annette Stanley
Manager, Environmental Programs
Chemical Manufactures Association
2501 M Street NW
Washington, DC 20037
202-887-1103
Eric Steindl
Chemical Standards Chemist
Restek Corporation
110 Benner Circle
Bellefonte, PA 16823
814-353-1300
Chuck Sueper
Huntingdon Analytical Services
P.O. Box 250
Middleport, NY 14105
716-735-3400
Leroy M. Sutton
Development Chemist
Compuchem Corporation
P.O. Box 12652
Research Triangle Park, NC 27709
919-248-6468
-------
766
E. L. Sweeney
Senior Project Chemist
Auto Research Laboratories
400 E. Sibley
Harvey, IL 60426
708-210-3494
Jawad Syed
Chemist
Indiana Dept. of Envirn. Management
5500 W. Bradbury
Indianapolis, IN 46241
317-243-5133
Joseph Szlachciuk
Environmental Technician
Texas Instruments, Inc.
34 Forest Street M/S 11-4
Attleboro, MA 02703
508-699-1343
Dr. Jay Tice
Mgr. Govrn. Affairs, Science & Tech.
Georgia Pacific
1875 Eye Street NW, Suite 775
Washington, DC 20006
202-659-3600
James C. Todaro
Laboratory Director
Matrix Analytical
106 South Street
Hopkinton, MA 01748
508-435-6824
David Tompkins
President
ETS Analytical Services
1401 Municipal Road
Roanoke, VA 24012
203-265-0004
Felicitas Trinidad
Hoffmann-LaRoche, Inc.
340 Kings land Street
Nutley, NJ 07110
201-235-3131
Robert J. Valis
U.S. Army Environmental Hygiene Agency
Bldge E2100
Aberdeen Proving Ground, MD 21010-5422
301-671-2208
Stan Van Wagenen
Senior Chemist
Reynolds Electrical & Engr. Co., Inc.
P.O. Box 98521 M/S 706
Las Vagas, NV 89193
702-295-7219
Barney J. Venables
TRAC Labs., Inc.
113 Cedar Street
Denton, TX 76201
817-566-3359
Lisa Ventry
QA Administrator
Massachusetts Water Resources Authority
100 First Avenue
Boston, MA 02129
617-241-2329
Joe Viar
Chairman
Viar and Company
Sample Control Center
300 N. Lee Street, Suite 200
Alexandria, VA 22314
703-557-5040
-------
767
Carolyn Vidrine
Physical Science Technician
NAVFAC, Navy Public Works
Utilities Dept. Bldg. P-71 - Code 600
Norfolk, VA 23511-6098
804-455-1789
David L. Vinci
Senior Environmental Chemist
Westchester Public Health Lab
Hammond House Road
Valhalla, NY 10595
914-524-5575
J. Howard Vinopal, PhD
Entomologist
US Army Environmental Hygiene Agency
ATTN: OFCD/Pesticide Analysis
Aberdeen Proving Ground, MD 21010-5422
301-671-2177
Joseph S. Vital is
USEPA Office of Water
Engineering & Analysis Division
401 M Street, SW (WH-586)
Washington, DC 20460
202-382-7172
Kenneth J. Wai its
Laboratory Director
Pacific Treatment Analytical Services
4340-A Viewridge Avenue
San Diego, CA 92123
619-560-7717
Tonie M. Wallace
President
County Court Reporters, Inc.
124 East Cork Street
Winchester, VA 22601
703-667-6562
Robert D. Welker
Chemist
Indianapolis Water Company
1220 Waterway Boulevard
P.O. Box 1220
Indianapolis, IN 46206-1220
317-263-6396
Dr. Charles Weston
Techn i ca1/Di rector
ETC Corporation
284 Raritan Center Parkway
Edison, NJ 08818-7808
908-225-6784
Kirview Wicker
Chemist
Department of the Army
Program Manager for Rocky Mt. Arsenal
AMXRM-LSA
Commerce City, CO 80022-2180
303-289-0194
Jill M. Williams
Associate Microscopist
R. J. Lee Group
10366 Battleview Parkway
Manassas, VA 22110
703-368-7880
Allison S. Wilson
Hampton Roads Sanitation District
P.O. Box 5000
Virginia Beach, VA 23455
804-498-7749
Dr. Hugh Wise
USEPA, EAD (WH-552)
401 M Street, SW
Washington, DC 20460
202-382-7177
-------
763
Katherine Wright
1936 N. Clark Street, Apt. 316
Chicago, IL 60614
312-664-7428
Chwen-Jiuan (Jane) Wu
Chemist
New Jersey Institute of Technology
323 Martin Luther King Blvd.
Newark, NJ 08830
201-536-5857
John E. Young
Senior Scientist A
Westinghouse SRL
Bldg. 773-A, Room B-113
Aiken, SC 29808
803-725-3565
Michael S. Young
NET Atlantic, Inc.
12 Oak Park
Bedford, MA 01730
617-275-3535
Wallace S. Zick, Jr.
Braun Intertec
6800 South County Road 18
Minneapolis, MN 55442
612-941-5600
•&U.S. Government Printing Office : 1992 - 312-014/40077
-------
|