United States
Environmental Protection
Agency
Environmental Monitoring and
Support Laboratory
Cincinnati OH 45268
Research and Development
EPA/600/S4-86/025 Sept. 1986
ŁEPA Project Summary
USEPA Method Study 7,
Analyses for Trace Elements
in Water by Atomic
Absorption Spectroscopy
(Direct Aspiration) and
Colorimetry
John A. Winter and Paul W. Britton
This report describes a study of
USEPA Method 200.0, Atomic Absorp-
tion (AA) Spectroscopy (direct aspira-
tion), and Colorimetry, for ten ele-
ments: aluminum, arsenic, cadmium,
chromium, copper, iron, lead, man-
ganese, selenium and zinc, at trace lev-
els in water.
The study design was based on You-
den's original non-replicate plan (1) for
collaborative evaluation of analytical
methods.
Six sample concentrates, each con-
taining the ten trace elements, were
sealed in glass ampuls, verified as ho-
mogeneous and stable over time, and
provided to the analysts with instruc-
tions for sample preparation and analy-
ses. The analysts were instructed to
add an aliquot of each concentrate to a
volume of distilled water and to a vol-
ume of natural water (river, lake, estu-
ary, tap, wastewater, etc.), and to per-
form AA or colorimetric analyses for the
ten elements. Results from distilled
water evaluated the proficiency of the
method free of interferences while by-
difference results from natural waters,
analyzed with and without the incre-
ment, reveal interferences with the
method from a particular sample ma-
trix, if present. Samples were prepared
in pairs, in which the concentrations of
analytes were slightly different for each
sample of a pair. Three pairs of samples
were used: one contained the trace ele-
ments near minimum detectable levels,
a second contained intermediate levels
and the last contained higher levels, but
all were to be within the linear re-
sponse ranges of the calibration curves.
The analysts performed single analyses
on each sample for each analyte as in
routine analyses. Analysts were cau-
tioned that the lowest levels of analyte
in the study might require concentra-
tion or extraction before measurement.
The data from a study which was
conducted in 1972, were originally eval-
uated with the computerized collabora-
tive study program (COLST) and used
to produce precision and bias state-
ments for the 1979 edition of USEPA's
manual of water methods (2). These
same data were re-evaluated for this re-
port utilizing the current (FY86) EMSL-
Cincinnati statistical program, Interlab-
oratory Method Validation Study
(IMVS), to make the results comparable
to similar studies involving the same
metals by AA-Furnace and Inductively
Coupled Plasma (ICP), also evaluated
by IMVS. Mean recovery, overall stand-
ard deviation and single-analyst stand-
ard deviations were calculated for each
element at each concentration and for
the two water types. Regression equa-
tions for recovery and precision were
generated, and practical effects of
water type noted.
-------
This Project Summary was devel-
oped by EPA's Environmental Monitor-
ing and Support Laboratory, Cincinnati,
OH, to announce key findings of the re-
search project that is fully documented
in a separate report of the same title
(see Project Report ordering informa-
tion at back).
Introduction
The study was designed to evaluate
the AA flame spectroscopy and col-
orimetry methods of analyses recom-
mended for those metals most common
in industrial discharges or most often
cited in NPDES permits as contami-
nants: Aluminum (Al), Arsenic (As),
Cadmium (Cd), Chromium (Cr), Copper
(Cu), Iron (Fe), Lead (Pb), Manganese
(Mn), Selenium (Se), and Zinc (Zn).
Since wastewater discharges to
streams and lakes ultimately become
public water supplies to other commu-
nities, most of these elements are of in-
terest for analysis of drinking water sup-
plies as well.
It is important that USEPA laborato-
ries and water-oriented laboratories of
other federal and state agencies and the
private sector use methods of analyses
which have been evaluated statistically
for precision, bias, and general interfer-
ences under the controlled conditions
of an interlaboratory study.
The participant laboratories should
be representative of laboratories moni-
toring discharges, that is, permittees,
consultant laboratories and the state
and federal regulatory agencies. Conse-
quently, regulatory needs established
the elements, their concentrations, the
methods of analyses and the laborato-
ries who participated in this study.
Statistical parameters from the study
were originally estimated using a com-
puter program (COLST) developed in
the mid-1970's. These statistics were
published as the precision and bias
statements in the current manual of
USEPA methods (2). Several years ago,
however, the statistical estimation pro-
cedure for method studies was modi-
fied to be compatible with the ASTM
Standard D2777-77 (3). To make results
of this study comparable to the recent
studies conducted for the AA-furnace
and ICP methods, it was necessary to
re-evaluate the study data using the cur-
rent computerized evaluation program,
IMVS, to prepare this report. When the
USEPA methods manual is updated, re-
sults from this re-evaluation will be
used as the basis for relevant precision
and bias statements.
Summary
The Quality Assurance Branch (QAB)
of the Environmental Monitoring and
Support Laboratory - Cincinnati (EMSL-
Cincinnati) conducted an interlabora-
tory study on the USEPA methods of
analysis for aluminum, arsenic, cad-
mium, chromium, copper, iron, lead,
manganese, selenium and zinc in water.
The methods evaluated in this study
which are described in the 1971 edition
of USEPA's methods (4) and included in
Appendix A of the full report, are direct
aspiration atomic absorption (AA) tech-
niques, except for arsenic and selenium
analyses which are silver diethyl-
dithiocarbamate and diaminobenzidine
colorimetric procedures, respectively.
Three sample concentrates were pre-
pared as pairs with slightly different
concentrations for each of the elements,
with the intention of representing the
full range of linear response. An aliquot
of each concentrate was added to dis-
tilled water and natural water samples
by the analysts. One measurement was
made on the natural water as back-
ground and one measurement was
made on each of the distilled water and
natural water samples with the added
increments. Recoveries from the natural
water samples were calculated by dif-
ference. Statistical estimates such as
mean recovery, standard deviation, and
bias, were calculated from the results
with each water sample and increment
level. A review of revised detection lim-
its in the current USEPA manual for the
eight AA analytes by direct aspiration,
shows that the lowest concentrations
studied were too low for dependable re-
sults by AA, direct aspiration. There-
fore, the regressions for AA methods,
shown in Table 1, were developed from
the estimates at the medium and high
concentration levels. The regressions
for colorimetric analyses of arsenic and
selenium used all three concentration
levels. The regression equations esti-
mate the mean recovery (X), overall
standard deviation (S) and single-
analyst standard deviation (Sr) which
may be expected in routine work within
the concentration (C) range indicated.
Results
The full report presents tables of all
data received, data actually evaluated
(see limitation below), and final basic
statistics. Plots are also provided show-
ing all statistics and the regressions
given in Table 1.
Need to Limit Data Sets to 50
or Fewer Responses
The original AA study design utilized
a very large number of laboratories (17
USEPA and 99 other federal, state, local
agency, private and industrial laborato-
ries). However, the present IMVS com-
puter program limits to 50 the number
of data sets it can treat statistically.
Since the IMVS program is extremely
complex and too difficult to modify in a
limited time, it was necessary in these
instances to randomly reduce data sets
for a metal to 50 responses to accom-
modate use of the IMVS program.
Discussion and Conclusions
The statistical results of the study are
fully detailed in Tables 52-69 and Fig-
ures 1-20 of the full report. However, a
brief summary for each analyte follows:
Aluminum
At the 500-1200 ng/L levels, recover-
ies by direct aspiration were an excel-
lent 96-103%, with overall relative
standard deviation (RSDs) of 17-43%
and single-analyst RSD of 12-19%. At
the lowest level tested, 15-35 M-9/L,
mean recoveries averaged 236-1270%
and overall RSD ranged from 69-119%,
demonstrating the inadequacy of direct
analysis at this concentration level. By
comparison, concentration by evapora-
tion for the low level samples showed
improved but still high recoveries of
176-245% in distilled water and low
level recoveries of 45-80% in natural
waters. Overall RSDs from the concen-
tration data were also improved but
with somewhat higher RSD levels of 61-
82% for distilled water than the 22-52%
overall RSDs for natural waters.
Arsenic
Recoveries of arsenic by the silver di-
ethyldithiocarbamate colorimetric
method were similar (75-89%) over the
entire range tested, 20-292 n-g/L, for
both distilled and natural waters. In con-
trast, the overall RSDs reduced with in-
creased concentrations of arsenic for
both distilled and natural waters; 42-
51% at 20-29 ^g/L levels, 28-33% at 67-
80 jjLg/L levels and 20-27% at 266-290
levels of arsenic.
Cadmium
Recoveries of cadmium by direct as-
piration were proportionately lower
than with the concentration technique,
-------
Table 1.
Analyte
Summary of Statistical Regressions USEPA Method Study 7, Trace Metals by
AA/Colorimetry
Concentration
Range of
Studied Data
Used,
Recovery from
Distilled Water
By Direct Aspiration
Recovery from
Natural Water of Choice
By Direct Aspiration
Aluminum
Arsenic*
Cadmium
Chromium
Copper
Iron
Lead
Manganese
Selenium*
Zinc
500 to 1205
20 to 292
14 to 78
74 to 407
60 to 332
350 to 840
84 to 367
84 to 469
4.7 to 49
56 to 310
X = 0.979(0) + 6.16
S = 0.066(30 + 125
Sr = 0.08600 + 40.5
X = OSSOfCJ - 0.25
5=0.798(30 + 5.93
Sr = 0.122
-------
Manganese
At manganese levels of 84-469 (ig/L,
recoveries by direct aspiration were ex-
cellent in both distilled and natural
waters, at 97-104%, with relatively small
overall RSDs of 14-17% at 84-106 n-g/L
and 5.4-8.4% at 426-469 ng/L levels of
manganese.
Recovery of manganese at levels of
11-17 (jtg/L by direct aspiration, how-
ever, were higher at 110-196% with
overall RSDs of 41-96%, indicating how
questionable direct analysis is at this
level. By comparison, the extraction
technique also showed rather poor re-
coveries of 91-227% with overall RSDs
of 45-117%, while recoveries by concen-
tration improved to 76-100% with over-
all RSDs of 10-32%.
Selenium
Contrary to arsenic, percent recover-
ies of selenium by the diaminoben-
zidine colorimetric method, were re-
duced with increased concentrations of
selenium both in distilled and natural
waters. At the 4.7-7.8 |ig/L level of sele-
nium recoveries were 75-93%, with
overall RSDs of 18-38%. At the 12-19 \i.g/
L levels, recoveries were 60-77% with
35-64% RSD. Similarly at the 43-49 p,g/L
levels, percent recoveries were 59-66%
with overall RSDs of 40-56%.
Zinc
Direct aspiration analyses for zinc
showed good recovery (99-102%) at the
56-310 jig/L levels, with 10-29% overall
RSD. However, at the lowest levels (7-
11 n.g/L), recoveries ballooned to 182-
252%, with overall RSDs of 81-102%,
demonstrating the questionable value
of direct analyses at such a low concen-
tration level. The low-level statistics
were due largely to the extreme values
which were not rejected in outlier tests.
Comparison of analytical results for
the low-level zinc samples (7-11 n-g/L)
by direct aspiration, extraction, and
concentration showed similar recover-
ies of 182-252% for direct aspiration, 98-
283% for extraction, and 120-205% for
concentrations with overall RSDs of 44-
109%. One must conclude that the low-
est levels tested were marginal for a
number of laboratories.
Percentage of Data Considered
Outliers
A question that commonly arises re-
garding statistical analysis of method
study data is what percentage of the re-
sponses were considered outliers and
omitted in estimating the statistical
characteristics of the method when
properly performed. Overall, 6291 re-
sponses were evaluated in this study for
the analytical conditions being consid-
ered, i.e., direct measurement at all
three concentration levels, and, where
applicable, extraction and concentra-
tion at the low-concentration level, with
up to 50 laboratories reporting. Of this
total, 16.8% of the values were rejected
as less than values, zeros or statistical
outliers. The rejection percentages for
recovery from distilled and natural
water were similar, 17.0% and 16.7%,
respectively. Overall rejection percent-
ages differed somewhat among the an-
alytes studied, from a high of 23.9% for
selenium to a low of 12.4% for man-
ganese. Where the rejection percentage
was highest it was usually caused by
problems quantifying results for the
low-concentration samples, i.e., fre-
quent zeros and less thans, particularly
from direct aspiration. The exception is
the selenium method, which produced
the highest percentage of rejected data
by a slight margin over cadmium
(23.0%) and aluminum (21.7%), primar-
ily because of a consistent high bias in
the selenium data from several labora-
tories. These results support the previ-
ous conclusion that the lowest concen-
trations studied were too low for proper
quantitation by direct aspiration, but do
not otherwise suggest general prob-
lems.
Natural Water Used in the
Study
Another interesting sidelight is the di-
versity of the "natural water" each par-
ticipant chose to use. Of the 119 labora-
tories reporting recoveries from what
has been called throughout this report a
"natural water," 97 described the
source as a river, stream, creek, pond,
lake, reservoir, spring or well, 15 de-
scribed it as a tap water, 4 a salt water,
and 3 described some type of effluent.
Interpreting Results from the
Comparison Between Waters
Finally, to understand the results
from the comparison between waters
(Table 71 of the full report), it was nec-
essary to develop two more summary
tables. Overall precision was summa-
rized in Table 2 for the study analytes,
while percent recovery was summa-
rized in Table 3, as shown on the follow-
ing pages. Entries in the first, second,
fourth, fifth, seventh, and eighth
columns of both Tables were averaged
(or pooled) from the appropriate pair of
entries in Tables 52 through 69 of the
full report.
The next to last column of Table 2
shows that manganese has the smallest
pooled RSD of all the study analytes and
iron is next to the smallest. Since all
comparisons between distilled and nat-
ural water results, RSD and percent re-
covery, are measured relative to pooled
RSD, tests for manganese and iron are
going to be the most sensitive to differ-
ences.
The last column of Table 2 shows
that, relative to their pooled RSD, iron
and selenium have the largest average
differences between overall RSD esti-
mates for the two waters at the concen-
trations used to fit the regressions in
Table 1, with values of 0.40 and 0.28,
respectively. Similarly, Table 3 shows
that, relative to the pooled RSD, percent
recoveries for the two waters differed
most for iron and manganese, with val-
ues of 0.30 and 0.25, respectively. Thus
it would seem, the statistically signifi-
cant differences between waters pri-
marily stem from differences in preci-
sion of the selenium data, differences in
recovery for manganese, and a combi-
nation of the two for iron.
Review of the relevant tables and fig-
ures in the full report suggest that there
is no practical difference in the recovery
of iron and manganese between the two
waters and hence no matrix effect. How-
ever, the differences in precision for
iron and selenium between the distilled
and natural waters do appear of practi-
cal importance, suggesting that recov-
eries of iron and selenium from some
natural waters are adversely effected
and sample effects are significant.
References
1. Youden, W. J. and E. H. Steiner,
1975. Statistical Manual of the Asso-
ciation of Official Analytical Chem-
ists, AOAC, 1111 North 19th Street,
Suite 210, Arlington, VA 22209.
2. Methods for Chemical Analysis of
Water and Wastes, 1979. U.S. Envi-
ronmental Protection Agency, EPA-
600/4-79-020, Environmental Moni-
toring and Support Laboratory,
Cincinnati, OH 45268.
3. Annual Book of ASTM Standards,
Section 11, Water and Environmen-
tal Technology, 77.07, Water (1) Cur-
rent year. American Society for Test-
ing and Materials, 1916 Race Street,
Philadelphia, PA 19103.
-------
Table 2. Summary of Pooled Overall Relative Standard Deviation (RSD)
Analyte
Al
As
Cd
Cr
Cu
Fe
Pb
Mn
Se
Zn
Low Pair
D* N*
NA**
49.82 44.67
NA
NA
NA
NA
NA
NA
25.06 35.34
NA
Medium fair
Diff. D
29.35
-5.15 30.08
40.47
18.11
23.18
12.80
36.12
14.27
10.28 43. 14
25.09
N
35.00
30.88
35.23
23.90
21.35
21.18
31.91
15.88
55.40
26.65
Diff.
5.65
0.80
-5.24
5.78
-1.82
8.38
-4.21
1.61
12.26
1.56
D
18.35
21.59
17.45
14.37
9.18
7.77
12.40
6.01
40.10
11.80
High Pair
N
18.77
23.73
13.41
14.51
11.52
10.52
20.05
7.47
54.22
10.54
Diff.
0.41
2.14
-4.04
0.13
2.34
2.75
7.65
1.46
14.12
-1.26
Avo
s-irg^.
Diff.
3.03
-0.74
-4.64
2.96
0.26
5.56
1.72
1.54
12.22
0.15
Pooled
RSD
26.34
35.04
29.00
18.14
17.39
14.00
26.83
11.70
43.50
19.94
Avg.
Diff.
Pooled
RSD
0.12
-0.02
-0.16
0.16
0.01
0.40
0.06
0.13
0.28
0.01
'Distilled and Natural water samples.
**"NA" means statistics from the low concentration sample pair were not used to fit the regressions in Table 1.
Table 3. Summary of Percent Recovery
Analyte
Al
As
Cd
Cr
Cu
Fe
Pb
Mn
Se
Zn
Low Pair
D* /V* Diff.
NA**
83.73 77.27 -6.46
NA
NA
NA
NA
NA
NA
88.06 77.68 - 10.38
NA
Medium Pair
D
98.85
83.49
110.16
102.35
101.14
99.34
110.89
97.27
71.45
100.59
N
97.17
81.95
111.80
98.95
96.13
106.68
111.85
102.43
67.08
101.55
Diff.
-1.69
-1.55
1.64
-3.40
-5.01
7.34
0.96
5.16
-4.38
0.96
D
98.67
86.77
96.52
98.74
96.56
99.66
100.43
98.50
65.43
99.95
High Pair
N
103. 19
84.97
97.10
98.13
94.33
100.79
103.53
99.27
60.43
99.66
Diff.
4.52
-1.80
0.58
-0.61
-2.23
1.14
3.11
0.77
-5.01
-0.29
Avo
/ivy.
Diff.
1.42
-3.27
1.11
-2.00
-3.62
4.24
2.03
2.96
-6.59
0.33
Avg.
Diff.
Pooled
RSD
0.05
-0.09
0.04
-0.11
-0.21
0.30
0.08
0.25
-0.15
0.02
'Distilled and Natural water samples.
**"NA" means statistics from the low concentration sample pair were not used to fit the regressions in Table 1.
4. Methods for Chemical Analysis of
Water and Wastes, 1971. U.S. Envi-
ronmental Protection Agency,
16020...07/71, Environmental Moni-
toring and Support Laboratory,
Cincinnati, OH 45268.
-------
The EPA authors John A. Winter and Paul W. Britton (also the EPA Project
Officers, see below) are with the Environmental Monitoring and Support
Laboratory, Cincinnati, OH 45268.
The complete report, entitled "USEPA Method Study 7, Analyses for Trace
Elements in Water by A tomic Absorption Spectroscopy I Direct A spiration) and
Colorimetry," (Order No. PB 86-208 709'/AS; Cost: $22.95, subject to change)
will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA 22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Environmental Monitoring and Support Laboratory
U.S. Environmental Protection Agency
Cincinnati. OH 45268
United States
Environmental Protection
Agency
Center for Environmental Research
Information
Cincinnati OH 45268
Official Business
Penalty for Private Use $300
EPA/600/S4-86/025
... vi«.
0000329
CHicfeoEA*80** STREET
------- |