2014 brooks rand instruments interlaboratory comparison ......(v/v) bromine monochloride (brcl)...
TRANSCRIPT
-
2014 Brooks Rand Instruments
Interlaboratory Comparison Study for Total
Mercury and Methylmercury
(Intercomp 2014)
Joel Creswell, Ph.D
Julian Metz
Annie Carter
Colin Davies
Brooks Rand Instruments
4415 6th
Ave NW
Seattle, WA 98107 USA
www.brooksrandinc.com
April 23, 2014
http://www.brooksrandinc.com/
-
1
Table of Contents
Introduction ..................................................................................................................................... 2
Methods........................................................................................................................................... 3
Sample Collection and Distribution ............................................................................................ 3
Data Analysis and Calculations .................................................................................................. 5
Results ............................................................................................................................................. 8
Holding time data ........................................................................................................................ 9
Total mercury data, Site WH ..................................................................................................... 12
Total mercury data, Site UB ...................................................................................................... 17
Total mercury data, Site SP....................................................................................................... 22
Summary of total mercury scores .............................................................................................. 27
Methylmercury data, Site WH ................................................................................................... 29
Methylmercury data, Site UB .................................................................................................... 34
Methylmercury data, Site SP ..................................................................................................... 39
Summary of methylmercury scores ........................................................................................... 44
Discussion ..................................................................................................................................... 46
Method Data .............................................................................................................................. 48
Future Studies ............................................................................................................................... 52
Acknowledgements ....................................................................................................................... 52
References ..................................................................................................................................... 53
-
2
Introduction
The Brooks Rand Instruments Interlaboratory Comparison Study (Intercomp) for Total
Mercury (THg) and Methylmercury (MeHg), now in its fourth year, was initiated to take the
place of the Mercury Round Robin, which was conducted by the Florida Department of
Environmental Protection (DEP) on an annual basis between 2001 and 2009. This exercise
provides a reliable means for laboratories to test their competency in the analysis of total
mercury and methylmercury in natural waters, as well as a metric for assessing the
intercomparability of data generated by different laboratories. Brooks Rand Instruments
undertook the organization of the Intercomp in 2011, upon learning that the Florida DEP no
longer planned to continue the exercise.
Intercomp 2011 was organized to replicate, to the extent feasible, the methods followed
in the Mercury Round Robins organized by the Florida DEP. The four week analysis and
reporting period for methylmercury data, introduced and validated in Intercomp 2012 (Creswell
et al. 2012), was maintained in Intercomp 2014 (this study). The organization and methods of
Intercomp 2014 were kept as similar to Intercomps 2011-2013 as possible, with the goal of
producing a multiyear record of laboratory performance and data intercomparability.
Some of the key features of Intercomp 2014 were a broad invitation to participate;
anonymous data submission, analysis, and reporting; and the inclusion of analytical method
reporting. The initial invitation to participate was sent to Brooks Rand Instruments’ email
database of more than 3,000 unique addresses. 64 laboratories in 17 countries signed up to
participate and were sent samples (some of these laboratories registered twice and were sent two
sets). Participants submitted 63 sets of total mercury results and 41 sets of methylmercury
results. Intercomp 2014 had one less participant than Intercomp 2013, which was, to our
knowledge, the largest interlaboratory comparison study for total mercury and methylmercury in
un-spiked natural waters ever conducted.
By requesting participating laboratories to report detailed information about their
analytical methods, we were able to assess the performance of specific protocols, reagents, and
equipment. These data can help identify methods that should be widely used and those that
should be avoided.
-
3
Methods
Sample Collection and Distribution
Samples were collected on January 15, 2014 from a site in Woods Hole, Massachusetts,
and on February 1, 2014 from two sites in the western Washington State, USA. Site WH (Woods
Hole Oceanographic Institution; Woods Hole, Massachusetts) was sampled by Carl Lamborg
from the Woods Hole Oceanographic Institution Environmental Research Laboratory’s seawater
distribution system. Site UB (Union Bay Natural Area; Seattle, Washington) is a freshwater
wetland located within the Union Bay Natural Area on the University of Washington campus
(Figure 1). Although we did not conduct a detailed study of site hydrology for this exercise, the
main water source for this wetland appears to be nearby Lake Washington. We sampled this site
near the parking lot. Site SP (Sunset Pond; Bellingham, Washington) is an impoundment of
Squalicum Creek, a small stream draining a primarily suburban watershed (Figure 2). This site
was sampled from the park on James St.
Figure 1. Sampling site UB
Figure 2. Sampling site SP
At site WH, 20 L fluorinated polyethylene (FLPE) carboys were filled from the seawater
distribution system, after passing the sample through 0.45 µm inline filter capsules (Voss
Technologies, San Antonio, Texas). At sites UB and SP, samples were collected with two
peristaltic pumps, using the same filters and carboys as at site WH. We used FEP sampling lines
connected to silicone tubing through the pump head. The carboys were each tested for
contamination prior to use by analyzing blanks of deionized water that had been stored overnight
-
4
in each vessel. Only containers with blank concentrations lower than 0.4 ng/L were used for
sample collection. Ten percent of the filters in each manufacturer’s lot were checked for
contamination by analyzing DI water passed through them. The average total mercury level was
less than 10 pg per capsule. Filters were rinsed with sample prior to filling carboys. We followed
strict trace metal clean techniques (U.S. Environmental Protection Agency 1996) at the field sites
(UB and SP) to avoid sample inhomogeneity to the extent possible. Approximately 120 L of
water was collected at each of the three sites.
Samples from site WH were preserved prior to shipping to Brooks Rand by adding
concentrated (18M) Trace Metal Grade sulfuric acid to each carboy for a final acid concentration
of 0.2% (v/v) or approximately 36 mM. Samples from sites UB and SP were preserved upon
returning from the field, by adding concentrated (12 M) Trace Metal Grade hydrochloric acid to
each carboy, for a final acid concentration of 0.4% (v/v) or approximately 50 mM. We
distributed the samples into 500 mL FLPE bottles for shipment to laboratories the following day,
February 2, 2014. In an effort to maximize sample uniformity, we shook and transferred samples
from the field carboys into fluorinated HDPE 30 gallon (113.6 L) drums that had been cleaned
and tested in the same manner as the carboys (described above). We filled the 500 mL FLPE
bottles through a fluoropolymer spigot installed in the bottom of the barrel. These bottles were
all from the same manufacturer’s lot, and prior to use we tested a random selection of ten percent
of the cases received. Ten percent of the bottles from each case selected were filled with a 1%
(v/v) bromine monochloride (BrCl) solution, stored at room temperature overnight, and analyzed
the following day. All of these bottle blanks contained less than 0.4 ng/L total mercury.
The 500 mL FLPE sample bottles were placed in single plastic zip-top bags and stored in
shipping boxes overnight. 66 sets of samples were sent to 64 laboratories on Monday, February
3, 2014, by express courier. Most deliveries were completed within three days, but in a few
cases, international customs delays prevented timely delivery and samples were not delivered for
several days or even weeks. Participating laboratories were asked to analyze samples for total
mercury and/or methylmercury following their standard operating procedures, and were given no
further guidance on analytical methodology. All results were requested to be reported by March
3, 2014, four weeks after the samples were shipped; however, results were generally accepted for
samples analyzed by the March 3 deadline. This deadline is based on the results of the 2012
-
5
bottle storage study, which demonstrated that analyte concentrations did not change substantially
over a four week holding time (Creswell et al. 2012).
Three laboratories participated in a holding time study as part of Intercomp 2014. Brooks
Rand Labs, the “Jožef Stefan” Institute Department of Environmental Sciences, and the U.S.
Geological Survey Wisconsin Mercury Research Lab each received two replicate sets of samples
to analyze at the beginning and end of the reporting period. The results of these analyses were
used to determine whether the analyte concentrations changed over the course of the study.
All results were reported to an independent third party, EcoChem, Inc. (Seattle,
Washington, USA), a data validation company who had no role in the study other than data
management. At EcoChem, the dataset was compiled, and a unique identifier was assigned to
each laboratory, before it was transmitted to Brooks Rand. Following delivery of this report,
each participating laboratory received an e-mail containing their own unique identifier, but
identifiers were not disclosed to any other parties, including Brooks Rand Instruments. This
research design ensured that there would be no bias by the preparers of this report against any
participating laboratory and that participants could submit data without fear of embarrassment
from a possible low score.
Data Analysis and Calculations
Each laboratory was asked to report an analytical result, detection limit, and date
analyzed for each sample and analyte. These data are the basis of the scores in this report. In
addition, each laboratory reported information on sample preparation and analytical
methodology and equipment. These data were used to compile assessments of the performance
of various analytical methods, but were not used in laboratory scoring.
Statistical data analysis was carried out largely following the recommendations of Lin &
Niu (1998). Because a true value of the Intercomp samples was not known, it was necessary to
remove obviously outlying data points before calculating a grand mean. The method of modified
Z scores (Filliben 2012) was selected for outlier removal because it is robust for both small and
large sample sizes, and is not affected by the magnitude of extreme outliers. The modified Z
score is calculated as follows:
-
6
( ̃)
Zi Modified Z score for laboratory i
xi an individual measurement of a sample by laboratory i
̃ the median measurement of sample x across all laboratories and
replicates
MAD median absolute deviation = (| ̃|) across all
measurements of a given sample
Any measurement with a modified Z score greater than 3.5 was labeled as an outlier (Filliben
2012) and excluded from the calculation of the consensus mean and median and laboratory
performance scores, described below. For all calculations, individual samples were grouped by
site as shown in Table 1. If any of a lab’s results for a given site were flagged as outliers, that lab
received a score of 0 for that site, and its results were excluded from calculation of that site’s
statistics. A series of exploratory statistical plots was then generated to verify that the removal of
outliers resulted in a dataset with a normal (Gaussian) distribution. The method of outlier
removal used in this study did not always result in a normal distribution of data (e.g., Figure
13a), however outlier removal improved normality for every site.
Table 1. Sample identification table.
Site
Sample
Numbers
WH (Woods Hole Oceanographic Institution) 1, 4, 6
UB (Union Bay Natural Area) 2, 7, 9
SP (Sunset Pond) 3, 5, 8
Participating laboratories were requested to report undetected values as being less than
their detection limit. A value of half of the detection limit was substituted for any undetected
result, for the purpose of calculating a modified Z score. If that modified Z score was above the
outlier threshold (3.5), the undetected value was flagged as an outlier and the laboratory’s results
excluded from calculations of the consensus mean and median and performance scores for that
site. If the modified Z score for a substituted result was below the outlier threshold, that result
was included in subsequent calculations of the consensus mean and median and performance
-
7
scores. Scores generated by this substitution method are marked with asterisks in the data tables
below to indicate that they are at least partially based on speculative substitution of a result.
In order to assign a score to each laboratory’s performance, we calculated a t-statistic and
Cook-Weisberg distance (Cook and Weisberg 1980; Lin and Niu 1998; Niu and Tintle 2008; Niu
and Miller 2009) for the measurement of each analyte at each site by each participating
laboratory. For example, if a lab analyzed total mercury and methylmercury in all nine samples,
with no outliers, that lab would receive six scores: one for each analyte for each of the three sites.
If a lab that did not report results for all three samples from a given site, the lab did not receive a
score for that site. The t-statistic was calculated as follows:
̅ ̃
̃√ ⁄ ⁄
ti t statistic for lab i
̅ the mean measurement for site Y by lab i
̃ mean of all measurements by labs with no flagged outliers
̃ standard deviation of the residuals of all measured values for site Y
r number of replicates from site Y analyzed by lab i
n0 total number of measurements taken by laboratories with no
flagged outliers
The t-statistic reflects the distance of an individual laboratory’s mean value from the consensus
mean for the site.
The Cook-Weisberg distance (C-W distance) was calculated as follows:
( ̅ ̅)
Di C-W distance for lab i
r number of replicates of site Y analyzed by lab i
̅ mean of all measurements of site Y by lab i
̅ the consensus mean of site Y
p number of labs with no flagged outliers
s2 variance of the residuals of all measured values for site Y
The C-W distance reflects the influence that each laboratory’s average value has on the
consensus mean.
-
8
We assigned scores following the rating system of Niu and Tintle (2003; Table 2).
Table 2. Rating system for laboratory performance (Niu and Tintle 2003).
Rating t-value C-W distance
5 (Very good) 0.00 ≤ t-value ≤ 2.00 C-W distance < 3.00
4 (Good) 2.01 ≤ t-value ≤ 4.00 C-W distance < 3.00
3 (Satisfactory) t-value > 4.00 C-W distance < 3.00
2 (Questionable) t-value > 4.00 3.00 ≤ C-W distance ≤ 10.00
1 (Poor) t-value > 4.00 10.01 ≤ C-W distance
0 (Unacceptable) With one or more outliers With one or more outliers
Results
We received 65 unique sets of results. 63 of these contained total mercury data and 41
contained methylmercury data. Because some labs registered for more than one set of samples,
the results may include more than one submission from some participants. We accepted results
with analysis dates up to the reporting deadline of March 3, 2014.
Figure 3. Particulate matter visible in some samples. Photo courtesy of Joël Knoery.
Some participants reported small amounts of leakage or particulate matter in some of the
sample bottles (e.g. Figure 3). Although great care is taken to minimize sample inhomogeneity,
-
9
the occurrence of these two problems (leakage and precipitation) seems to be unavoidable. No
participants reported a correlation between observed leakage or precipitate and sample results.
Holding time data
Due to differences in sample receipt and preparation times, holding time samples were
not analyzed on the same dates by all three laboratories participating in the holding time study.
However, the dates of analysis shown in Figures 4 and 5 below span the majority of the reporting
period for both analytes. There are several instances in which one or more labs observed a
change in analyte concentration in a sample greater than ± 1 standard deviation of three replicate
measurements (indicated by error bars). However there is no trend of a consistent, directional (up
or down) concentration change over time in all results from any one lab. If a lab exhibited
increasing concentrations over time in all samples, this might indicate a contamination problem,
while the reverse might indicate poor sample storage conditions. The lack of any such trend
indicates that the changes over time in these holding time samples exemplify the variability that
can be expected in all samples over the course of the reporting period.
-
10
Figure 4. Holding time data for total mercury. Error bars are ± 1 standard deviation of replicate
measurements. Where error bars are not visible, they are smaller than the chart symbol to which
they pertain. Note that the vertical scales are not identical. BRL = Brooks Rand Labs; JSI =
“Jožef Stefan” Institute; USGS = U.S. Geological Survey.
50
55
60
65
70
75
80
85
90
95
100
2/10/14 2/15/14 2/20/14 2/25/14 3/2/14 3/7/14
Tota
l Me
rcu
ry (
ng/
L)
Date Analyzed
Site WH
BRLJSIUSGS
1.55
1.65
1.75
1.85
1.95
2.05
2.15
2.25
2.35
2.45
2/10/14 2/15/14 2/20/14 2/25/14 3/2/14 3/7/14
Tota
l Me
rcu
ry (
ng/
L)
Date Analyzed
Site UB
BRLJSIUSGS
5
5.5
6
6.5
7
7.5
8
8.5
9
9.5
10
2/10/14 2/15/14 2/20/14 2/25/14 3/2/14 3/7/14
Tota
l Me
rcu
ry (
ng/
L)
Date Analyzed
Site SP
BRLJSIUSGS
-
11
Figure 5. Holding time data for methylmercury. Error bars are ± 1 standard deviation of
replicate measurements. Where error bars are not visible, they are smaller than the chart symbol
to which they pertain. Note that the vertical scales are not identical. BRL = Brooks Rand Labs;
JSI = “Jožef Stefan” Institute; USGS = U.S. Geological Survey. *In the Site WH graph, the
USGS concentrations were below the detection limit (0.04 ng/L). The plotted value is half of the
detection limit (0.02 ng/L) and the error bars span the range from 0 to the detection limit.
0
0.01
0.02
0.03
0.04
0.05
2/10/2014 2/15/2014 2/20/2014 2/25/2014 3/2/2014 3/7/2014 3/12/2014
Me
thyl
me
rcu
ry (
ng/
L)
Date Analyzed
Site WH BRLJSIUSGS*
0.1
0.11
0.12
0.13
0.14
0.15
0.16
0.17
0.18
0.19
2/10/2014 2/15/2014 2/20/2014 2/25/2014 3/2/2014 3/7/2014 3/12/2014
Me
thyl
me
rcu
ry (
ng/
L)
Date Analyzed
Site UB BRLJSIUSGS
0.08
0.09
0.1
0.11
0.12
0.13
0.14
0.15
0.16
2/10/2014 2/15/2014 2/20/2014 2/25/2014 3/2/2014 3/7/2014 3/12/2014
Me
thyl
me
rcu
ry (
ng/
L)
Date Analyzed
Site SP BRLJSIUSGS
-
12
Total mercury data, Site WH
Table 3. Laboratory performance for total mercury, site WH (samples 1, 4, 6). An asterisk
indicates that the lab mean includes at least one undetected value that was substituted with half
of that lab’s detection limit.
Consensus Mean = 76.49 ng/L, Consensus Median = 77.54 ng/L
Lab Mean t value C-W Distance Score
1 91.3 9.24 1.45 3
2 84.7 4.96 0.42 3
3 84.7 4.96 0.42 3
4 61.7 9.77 1.62 3
5 91.1 9.11 1.41 3
6 59.8 10.96 2.04 3
7 37.8 With outliers 0
8 137 With outliers 0
9 63.2 8.81 1.31 3
10 76.5 0.26 0.00 5
11 81.0 2.60 0.11 4
12 63.2 8.79 1.31 3
13 78.6 1.09 0.02 5
14 77.6 0.40 0.00 5
15 77.5 0.35 0.00 5
16 17.1* With outliers 0
17 80.2 2.10 0.07 4
18 99.9 14.71 3.67 2
19 67.8 5.83 0.58 3
20 72.0 3.14 0.17 4
21 83.6 4.28 0.31 3
22 83.5 4.22 0.30 3
23 59.7 11.06 2.07 3
24 81.8 3.15 0.17 4
25 87.6 6.87 0.80 3
26 73.4 2.28 0.09 4
27 78.2 0.79 0.01 5
28 80.4 2.25 0.09 4
29 80.2 2.12 0.08 4
30 86.5 6.12 0.63 3
31 67.8 5.85 0.58 3
32 69.7 4.66 0.37 3
Table continued on following page
-
13
Lab Mean t value C-W Distance Score
33 71.4 3.52 0.21 4
34 71.9 3.22 0.18 4
35 88.1 7.17 0.87 3
36 5.97 With outliers 0
37 81.1 2.68 0.12 4
38 68.8 5.21 0.46 3
39 73.2 2.37 0.10 4
40 77.8 0.58 0.01 5
41 95.1 11.66 2.30 3
43 79.7 1.76 0.05 5
44 58.3 11.92 2.41 3
46 80.6 2.38 0.10 4
47 74.7 1.42 0.03 5
48 79.3 1.50 0.04 5
49 71.1 3.76 0.24 4
50 68.7 5.26 0.47 3
51 98.0 13.49 3.09 2
52 89.8 8.28 1.16 3
53 72.9 2.58 0.11 4
54 68.6 5.32 0.48 3
55 84.9 5.09 0.44 3
56 66.9 6.45 0.71 3
57 74.8 1.34 0.03 5
58 80.4 2.22 0.08 4
59 91.2 9.13 1.41 3
60 72.7 2.73 0.13 4
61 67.3 6.15 0.64 3
62 Unreported values
63 84.8 5.07 0.44 3
64 66.5 6.68 0.76 3
65 59.8 10.96 2.04 3
-
14
Figure 6. Total mercury results from all participating laboratories for site WH.
Figure 7a. Histograms of reported total mercury concentrations for site WH. All concentrations
(left); non-outlier values (right).
0
20
40
60
80
100
120
140
160
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63
Tota
l Me
rcu
ry C
on
cen
trat
ion
(n
g/L)
Lab
Site WH Total Mercury Results
Lab Results
Consensus Mean
-
15
Figure 7b. Box plots of reported total mercury concentrations for site WH. All concentrations
(left); non-outlier values (right).
Figure 7c. Normal quantile-quantile plots of reported total mercury data for site WH. All
concentrations (left); non-outlier values (right).
-
16
Figure 7d. Density functions of residuals (residual = individual measurement - laboratory mean)
of total mercury data for site WH. All values (left); non-outlier values (right).
Figure 7e. Residual (residual = individual measurement - laboratory mean) vs. fitted value plots
of reported total mercury concentrations for site WH. All values (left); non-outlier values (right).
-
17
Total mercury data, Site UB
Table 4. Laboratory performance for total mercury, site UB (samples 2, 7, 9). An asterisk
indicates that the lab mean includes at least one undetected value that was substituted with half
of that lab’s detection limit.
Consensus Mean = 1.90 ng/L, Consensus Median = 1.90 ng/L
Lab Mean t value C-W Distance Score
1 2.56 6.82 0.86 3
2 1.83 0.57 0.01 5
3 1.90 0.13 0.00 5
4 1.93 0.40 0.00 5
5 1.88 0.10 0.00 5
6 2.53 With outliers 0
7 2.39 5.10 0.48 3
8 6.73 With outliers 0
9 1.50 3.90 0.28 4
10 1.98 0.89 0.01 5
11 1.93 0.43 0.00 5
12 1.21 6.82 0.86 3
13 1.93 0.47 0.00 5
14 1.67 2.23 0.09 4
15 1.84 0.44 0.00 5
16 8.44* With outliers 0
17 1.67 2.19 0.09 4
18 5.70 With outliers 0
19 2.77 With outliers 0
20 2.28 3.96 0.29 4
21 1.94 0.50 0.00 5
22 1.76 1.28 0.03 5
23 1.11 7.86 1.14 3
24 4.50 With outliers 0
25 1.98 0.95 0.02 5
26 1.95 0.67 0.01 5
27 2.06 1.78 0.06 5
28 2.00 1.14 0.02 5
29 1.88 0.04 0.00 5
30 1.67* 2.22 0.09 4
31 2.20 3.12 0.18 4
32 1.74 1.51 0.04 5
Table continued on following page
-
18
Lab Mean t value C-W Distance Score
33 1.47 4.23 0.33 3
34 1.70 1.88 0.07 5
35 1.42 4.70 0.41 3
36 0.683 With outliers 0
37 1.64 2.45 0.11 4
38 1.73 1.55 0.04 5
39 2.17 2.82 0.15 4
40 1.80 0.84 0.01 5
41 2.10 2.18 0.09 4
43 1.88 0.07 0.00 5
44 1.54 3.53 0.23 4
46 2.01 1.24 0.03 5
47 1.71 1.82 0.06 5
48 1.93 0.40 0.00 5
49 1.94 0.54 0.01 5
50 1.93 0.47 0.00 5
51 3.33 With outliers 0
52 1.99 1.04 0.02 5
53 2.64 7.56 1.06 3
54 2.18 2.99 0.17 4
55 2.03 1.44 0.04 5
56 1.53 3.58 0.24 4
57 1.95 0.67 0.01 5
58 4.11 With outliers 0
59 2.10 2.15 0.09 4
60 1.90 0.17 0.00 5
61 2.16 2.75 0.14 4
62 2.99 With outliers 0
63 2.03 1.44 0.04 5
64 1.96 0.73 0.01 5
65 1.77 1.14 0.02 5
-
19
Figure 8. Total mercury results from all participating laboratories for site UB.
Figure 9a. Histograms of reported total mercury concentrations for site UB. All concentrations
(left); non-outlier values (right).
0
5
10
15
20
25
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63
Tota
l Me
rcu
ry C
on
cen
trat
ion
(n
g/L)
Lab
Site UB Total Mercury Results
Lab Results
Consensus Mean
-
20
Figure 9b. Box plots of reported total mercury concentrations for site UB. All concentrations
(left); non-outlier values (right).
Figure 9c. Normal quantile-quantile plots of reported total mercury data for site UB. All
concentrations (left); non-outlier values (right).
-
21
Figure 9d. Density functions of residuals (residual = individual measurement - laboratory mean)
of total mercury data for site UB. All values (left); non-outlier values (right).
Figure 9e. Residual (residual = individual measurement - laboratory mean) vs. fitted value plots
of reported total mercury concentrations for site UB. All values (left); non-outlier values (right).
-
22
Total mercury data, Site SP
Table 5. Laboratory performance for total mercury, site SP (samples 3, 5, 8). An asterisk
indicates that the lab mean includes at least one undetected value that was substituted with half
of that lab’s detection limit. Consensus Mean = 6.16 ng/L, Consensus Median = 6.12 ng/L
Lab Mean t value C-W Distance Score
1 8.00 8.89 1.41 3
2 6.00 0.67 0.01 5
3 6.00 0.66 0.01 5
4 6.51 1.78 0.06 5
5 8.90 With outliers 0
6 5.77 1.77 0.06 5
7 6.41 1.31 0.03 5
8 10.3 With outliers 0
9 5.30 4.00 0.29 3
10 5.82 1.52 0.04 5
11 6.17 0.16 0.00 5
12 5.51 3.02 0.16 4
13 6.03 0.50 0.00 5
14 6.21 0.33 0.00 5
15 5.69 2.12 0.08 4
16 17.3* With outliers 0
17 5.77 1.77 0.06 5
18 11.8 With outliers 0
19 6.97 3.97 0.28 4
20 6.94 3.84 0.26 4
21 6.00 0.66 0.01 5
22 6.88 3.54 0.22 4
23 4.47 7.96 1.13 3
24 9.27 With outliers 0
25 6.88 3.57 0.23 4
26 6.24 0.49 0.00 5
27 6.30 0.76 0.01 5
28 5.95 0.90 0.01 5
29 5.94 0.94 0.02 5
30 7.22 5.18 0.48 3
31 5.53 2.92 0.15 4
32 5.65 2.31 0.10 4
Table continued on following page
-
23
Lab Mean t value C-W Distance Score
33 5.83 1.45 0.04 5
34 5.36 3.70 0.24 4
35 5.71 2.04 0.07 4
36 1.18 With outliers 0
37 6.66 2.50 0.11 4
38 5.20 4.48 0.36 3
39 6.57 2.05 0.08 4
40 5.89 1.17 0.02 5
41 6.77 3.03 0.16 4
43 6.42 1.35 0.03 5
44 5.00 5.44 0.53 3
46 6.23 0.44 0.00 5
47 5.76 1.80 0.06 5
48 6.42 1.34 0.03 5
49 6.08 0.27 0.00 5
50 7.07 4.45 0.35 3
51 7.62 7.08 0.89 3
52 6.21 0.33 0.00 5
53 7.33 With outliers 0
54 8.00 8.91 1.42 3
55 6.78 3.07 0.17 4
56 5.49 3.11 0.17 4
57 6.18 0.22 0.00 5
58 7.67 With outliers 0
59 7.03 4.29 0.33 3
60 5.81 1.56 0.04 5
61 4.49 7.88 1.11 3
62 4.61 7.29 0.95 3
63 6.52 1.85 0.06 5
64 5.90 1.15 0.02 5
65 5.79 1.66 0.05 5
-
24
Figure 10. Total mercury results from all participating laboratories for site SP.
Figure 11a. Histograms of reported total mercury concentrations for site SP. All concentrations
(left); non-outlier values (right).
0
5
10
15
20
25
30
35
40
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63
Tota
l Me
rcu
ry C
on
cen
trat
ion
(n
g/L)
Lab
Site SP Total Mercury Results
Lab Results
Consensus Mean
-
25
Figure 11b. Box plots of reported total mercury concentrations for site SP. All concentrations
(left); non-outlier values (right).
Figure 11c. Normal quantile-quantile plots of reported total mercury data for site SP. All
concentrations (left); non-outlier values (right).
-
26
Figure 11d. Density functions of residuals (residual = individual measurement - laboratory
mean) of total mercury data for site SP. All values (left); non-outlier values (right).
Figure 11e. Residual (residual = individual measurement - laboratory mean) vs. fitted value
plots of reported total mercury concentrations for site SP. All values (left); non-outlier values
(right).
-
27
Summary of total mercury scores
Table 6. Summary of total mercury scores. Scores range from 0 (unacceptable) to 5 (very good).
An asterisk indicates that the score is based on at least one result that was below the reported
detection limit and was replaced by half of the detection limit.
Lab
Site
WH
Site
UB
Site
SP Mean
1 3 3 3 3.00
2 3 5 5 4.33
3 3 5 5 4.33
4 3 5 5 4.33
5 3 5 0 2.67
6 3 0 5 2.67
7 0 3 5 2.67
8 0 0 0 0.00
9 3 4 3 3.33
10 5 5 5 5.00
11 4 5 5 4.67
12 3 3 4 3.33
13 5 5 5 5.00
14 5 4 5 4.67
15 5 5 4 4.67
16 0* 0* 0* 0.00
17 4 4 5 4.33
18 2 0 0 0.67
19 3 0 4 2.33
20 4 4 4 4.00
21 3 5 5 4.33
22 3 5 4 4.00
23 3 3 3 3.00
24 4 0 0 1.33
25 3 5 4 4.00
26 4 5 5 4.67
27 5 5 5 5.00
28 4 5 5 4.67
29 4 5 5 4.67
30 3 4* 3 3.33
31 3 4 4 3.67
32 3 5 4 4.00
Table continued on following page
-
28
Lab
Site
WH
Site
UB
Site
SP Mean
33 4 3 5 4.00
34 4 5 4 4.33
35 3 3 4 3.33
36 0 0 0 0.00
37 4 4 4 4.00
38 3 5 3 3.67
39 4 4 4 4.00
40 5 5 5 5.00
41 3 4 4 3.67
43 5 5 5 5.00
44 3 4 3 3.33
46 4 5 5 4.67
47 5 5 5 5.00
48 5 5 5 5.00
49 4 5 5 4.67
50 3 5 3 3.67
51 2 0 3 1.67
52 3 5 5 4.33
53 4 3 0 2.33
54 3 4 3 3.33
55 3 5 4 4.00
56 3 4 4 3.67
57 5 5 5 5.00
58 4 0 0 1.33
59 3 4 3 3.33
60 4 5 5 4.67
61 3 4 3 3.33
62 0 3 1.50
63 3 5 5 4.33
64 3 5 5 4.33
65 3 5 5 4.33
-
29
Methylmercury data, Site WH
Table 7. Laboratory performance for methylmercury, site WH (samples 1, 4, 6). An asterisk
indicates that the lab mean includes at least one undetected value that was substituted with half
of that lab’s detection limit.
Consensus Mean = 0.0281 ng/L, Consensus Median = 0.0249 ng/L
Lab Mean t value C-W Distance Score
2 0.0721 13.96 5.72 2
4 0.00250* 7.86 1.82 3
5 0.0100* 5.51 0.89 3
7 0.0243* 1.02 0.03 5
8 1.21 With outliers 0
9 0.0200 2.38 0.17 4
10 0.0319 1.35 0.05 5
12 0.0870 With outliers 0
13 0.0533 8.07 1.92 3
15 0.0200* 2.38 0.17 4
17 0.0242* 1.07 0.03 5
18 0.0357 2.54 0.19 4
20 0.0233 1.33 0.05 5
21 0.0100* 5.51 0.89 3
22 0.0347 2.23 0.15 4
25 0.0100* 5.51 0.89 3
27 0.0200* 2.38 0.17 4
32 0.0250* 0.81 0.02 5
33 0.0100* 5.51 0.89 3
34 0.100* With outliers 0
35 0.533 With outliers 0
36 0.0100* 5.51 0.89 3
37 0.0125* 4.73 0.66 3
41 0.0153 3.84 0.43 4
42 0.0170 3.32 0.32 4
45 0.0250* 0.81 0.02 5
47 0.0250 0.81 0.02 5
48 0.0273 0.08 0.00 5
49 0.0701 13.33 5.22 2
54 Unreported values
55 0.0500* 7.03 1.45 3
Table continued on following page
-
30
Lab Mean t value C-W Distance Score
56 Unreported values
57 0.0244 1.00 0.03 5
58 0.0591 9.89 2.88 3
59 0.0349* 2.31 0.16 4
60 0.780 With outliers 0
61 0.0333 1.80 0.10 5
62 Unreported values
63 0.0300* 0.76 0.02 5
64 0.0151 3.93 0.45 4
65 0.0340 2.01 0.12 4
Figure 12. Methylmercury results from all participating laboratories for site WH.
0
0.5
1
1.5
2
2.5
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62
Me
thyl
me
rcu
ry C
on
cen
trat
ion
(n
g/L)
Lab
Site WH Methylmercury Results
Lab Results
Consensus Mean
-
31
Figure 13a. Histograms of reported methylmercury concentrations for site WH. All
concentrations (left); non-outlier values (right).
Figure 13b. Box plots of reported methylmercury concentrations for site WH. All concentrations
(left); non-outlier values (right).
-
32
Figure 13c. Normal quantile-quantile plots of reported methylmercury data for site WH. All
concentrations (left); non-outlier values (right).
Figure 13d. Density functions of residuals (residual = individual measurement - laboratory
mean) of methylmercury data for site WH. All values (left); non-outlier values (right).
-
33
Figure 13e. Residual (residual = individual measurement - laboratory mean) vs. fitted value
plots of reported methylmercury concentrations for site WH. All values (left); non-outlier values
(right).
-
34
Methylmercury data, Site UB
Table 8. Laboratory performance for methylmercury, site UB (samples 2, 7, 9). An asterisk
indicates that the lab mean includes at least one undetected value that was substituted with half
of that lab’s detection limit.
Consensus Mean = 0.126 ng/L, Consensus Median = 0.126 ng/L
Lab Mean t value C-W Distance Score
2 0.119 0.81 0.02 5
4 0.148 2.03 0.11 4
5 0.109 1.86 0.09 5
7 0.118 0.91 0.02 5
8 0.117 1.00 0.03 5
9 0.117 1.07 0.03 5
10 0.136 0.80 0.02 5
12 0.249 With outliers 0
13 0.143 1.56 0.07 5
15 0.150 2.21 0.13 4
17 0.0944 3.26 0.29 4
18 0.113 1.43 0.06 5
20 0.154 2.61 0.18 4
21 0.0930 3.40 0.31 4
22 0.202 7.37 1.47 3
25 0.111 1.63 0.07 5
27 0.167 3.85 0.40 4
32 0.123 0.45 0.01 5
33 0.0433 8.29 1.86 3
34 0.120 0.74 0.01 5
35 0.497 With outliers 0
36 0.096 3.11 0.26 4
37 0.146 1.85 0.09 5
41 0.148 2.02 0.11 4
42 0.135 0.70 0.01 5
45 0.221 9.24 2.30 3
47 0.126 0.15 0.00 5
48 0.0973 2.97 0.24 4
49 0.0537 7.27 1.43 3
54 0.100 2.74 0.20 4
55 0.110 1.73 0.08 5
56 0.0827 Unreported values
Table continued on following page
-
35
Lab Mean t value C-W Distance Score
57 0.174 4.54 0.56 3
58 0.111 1.61 0.07 5
59 0.138 1.00 0.03 5
60 0.172* With outliers 0
61 0.130 0.24 0.00 5
62 0.109 1.79 0.09 5
63 0.155 2.67 0.19 4
64 0.304 With outliers 0
65 0.163 3.53 0.34 4
Figure 14. Methylmercury results from all participating laboratories for site UB.
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62
Me
thyl
me
rcu
ry C
on
cen
trat
ion
(n
g/L)
Lab
Site UB Methylmercury Results
Lab Results
Consensus Mean
-
36
Figure 15a. Histograms of reported methylmercury concentrations for site UB. All
concentrations (left); non-outlier values (right).
Figure 15b. Box plots of reported methylmercury concentrations for site UB. All concentrations
(left); non-outlier values (right).
-
37
Figure 15c. Normal quantile-quantile plots of reported methylmercury data for site UB. All
concentrations (left); non-outlier values (right).
Figure 15d. Density functions of residuals (residual = individual measurement - laboratory
mean) of methylmercury data for site UB. All values (left); non-outlier values (right).
-
38
Figure 15e. Residual (residual = individual measurement - laboratory mean) vs. fitted value
plots of reported methylmercury concentrations for site UB. All values (left); non-outlier values
(right).
-
39
Methylmercury data, Site SP
Table 9. Laboratory performance for methylmercury, site SP (samples 3, 5, 8). An asterisk
indicates that the lab mean includes at least one undetected value that was substituted with half
of that lab’s detection limit.
Consensus Mean = 0.103 ng/L, Consensus Median = 0.100 ng/L
Lab Mean t value C-W Distance Score
2 0.0961 0.77 0.02 5
4 0.117 2.01 0.10 4
5 0.0947 0.96 0.02 5
7 0.108 0.84 0.02 5
8 0.126 3.13 0.25 4
9 0.0933 1.14 0.03 5
10 0.120 2.40 0.15 4
12 0.179 With outliers 0
13 0.113 1.50 0.06 5
15 0.113 1.50 0.06 5
17 0.0766 3.34 0.29 4
18 0.101 0.15 0.00 5
20 0.124 2.95 0.22 4
21 0.0784 3.11 0.25 4
22 0.139 4.84 0.60 3
25 0.0867 2.02 0.10 4
27 0.143 5.46 0.76 3
32 0.100 0.26 0.00 5
33 0.0600 5.53 0.78 3
34 0.0382 8.41 1.81 3
35 0.370 With outliers 0
36 0.0773 3.25 0.27 4
37 0.127 3.26 0.27 4
41 0.101 0.17 0.00 5
42 0.104 0.31 0.00 5
45 0.172 9.20 2.17 3
47 0.100 0.26 0.00 5
48 0.0903 1.53 0.06 5
49 0.102 0.05 0.00 5
54 0.0970 0.65 0.01 5
55 0.0667* 4.65 0.56 3
56 0.0605 Unreported values
Table continued on following page
-
40
Lab Mean t value C-W Distance Score
57 0.137 4.58 0.54 3
58 0.0849 2.25 0.13 4
59 0.125 3.09 0.24 4
60 0.0324* 9.18 2.16 3
61 0.0767 3.34 0.29 4
62 0.118 2.07 0.11 4
63 0.119 2.25 0.13 4
64 0.0926 1.23 0.04 5
65 0.123 2.73 0.19 4
Figure 16. Methylmercury results from all participating laboratories for site SP.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54 56 58 60 62
Me
thyl
me
rcu
ry C
on
cen
trat
ion
(n
g/L)
Lab
Site SP Methylmercury Results
Lab Results
Consensus Mean
-
41
Figure 17a. Histograms of reported methylmercury concentrations for site SP. All
concentrations (left); non-outlier values (right).
Figure 17b. Box plots of reported methylmercury concentrations for site SP. All concentrations
(left); non-outlier values (right).
-
42
Figure 17c. Normal quantile-quantile plots of reported methylmercury data for site SP. All
concentrations (left); non-outlier values (right).
Figure 17d. Density functions of residuals (residual = individual measurement - laboratory
mean) of methylmercury data for site SP. All values (left); non-outlier values (right).
-
43
Figure 17e. Residual (residual = individual measurement - laboratory mean) vs. fitted value
plots of reported methylmercury concentrations for site SP. All values (left); non-outlier values
(right).
-
44
Summary of methylmercury scores
Table 10. Summary of methylmercury scores. Scores range from 0 (unacceptable) to 5 (very
good). An asterisk indicates that the score is based on at least one result that was below the
reported detection limit and was replaced by half of the detection limit.
Lab
Site
WH
Site
UB
Site
SP Mean
2 2 5 5 4.00
4 3* 4 4 3.67
5 3* 5 5 4.33
7 5* 5 5 5.00
8 0 5 4 3.00
9 4 5 5 4.67
10 5 5 4 4.67
12 0 0 0 0.00
13 3 5 5 4.33
15 4* 4 5 4.33
17 5* 4 4 4.33
18 4 5 5 4.67
20 5 4 4 4.33
21 3* 4 4 3.67
22 4 3 3 3.33
25 3* 5 4 4.00
27 4* 4 3 3.67
32 5* 5 5 5.00
33 3* 3 3 3.00
34 0* 5 3 2.67
35 0 0 0 0.00
36 3* 4 4 3.67
37 3* 5 4 4.00
41 4 4 5 4.33
42 4 5 5 4.67
45 5* 3 3 3.67
47 5 5 5 5.00
48 5 4 5 4.67
49 2 3 5 3.33
54 4 5 4.50
55 3* 5 3* 3.67
56
Table continued on following page
-
45
Lab
Site
WH
Site
UB
Site
SP Mean
57 5 3 3 3.67
58 3 5 4 4.00
59 4* 5 4 4.33
60 0 0* 3* 1.00
61 5 5 4 4.67
62 5 4 4.50
63 5* 4 4 4.33
64 4 0 5 3.00
65 4 4 4 4.00
Table 11. List of the 64 participating laboratories to which samples were sent.
Laboratory Name Country
ACZ Laboratories Inc. USA
AEP Dolan Chemical Laboratory USA
Alpha Analytical Laboratories, Inc. USA
ALS Environmental (Vancouver) Canada
ALS Environmental (Winnipeg) Canada
ALS Technichem Sdn Bhd Malaysia
American Assay Laboratories USA
Basic Laboratory, Inc. USA
Battelle Marine Sciences Lab USA
Brigham Young University Mercury Lab USA
Brooks Rand Labs USA
Caltest Analytical Laboratory USA
City of San Jose ESD Laboratory USA
Clean Water Services Water Quality Laboratory (Oregon) USA
CNR Institute of Atmospheric Pollution Research Italy
CSIRO Land and Water Australia
Dartmouth College Trace Element Analysis Lab USA
Desert Research Institute USA
Dr. Carol Ptacek Lab, University of Waterloo Canada
Duke University/Hsu-Kim USA
Environment Canada - Burlington Canada
Environment Canada PYLET Canada
Flett Research Ltd. Canada
Geosciences Environment Toulouse France
Gustavus Adolphus College USA
IAEA-Environment Laboratories Monaco
IFREMER/LBCM France
Table continued on following page
-
46
Laboratory Name Country
Institut FA Forel, University of Geneva Switzerland
Institute of Geology, ASCR, v.v.i. Czech Republic
Instituto de Quimica/UNICAMP Brazil
IVL Swedish Environmental Research Institute Ltd Sweden
Jozef Stefan Institute Slovenia
Kentucky Environmental Services USA
Laboratory of Environmental Chemistry, Univ. of Athens Greece
Lakehead University Environmental Laboratory Canada
LEEQ - Environment Canada Canada
Lumex Analytics GMBH Germany
Marine Environmental Monitoring Centre of Guangxi Province China
Metro Wastewater Reclamation District USA
Microbac Laboratories, Inc. - Chicagoland Division USA
NIVA (Norwegian Institute for Water Research) Norway
North Carolina DENR Laboratory USA
Oak Ridge National Laboratory USA
Pace Analytical USA
Southeast Laboratory San Francisco PUC USA
SRC Environmental Analytical Laboratories Canada
SRCSD-EL (Sacramento) USA
Texas Tech University/Dr. Danny Reible Lab USA
Trace Element Laboratory, University of Oulu Finland
TU Braunschweig, Inst. für Geoökologie, Umweltgeochemie Germany
UNIR/Laboratorio de Biogeoquimica Ambiental Brazil
University of Alberta, Biogeochemical Service Laboratory Canada
University of Illinois USA
University of Ottawa Environ. Geochem. & Microbiol. Lab Canada
University of South Carolina USA
University of Utah Geology & Geophysics USA
University of Western Ontario, Analytical Services Canada
USGS Menlo Park USA
USGS Middleton USA
USGS Southeast Ecological Science Center USA
WE-Energies USA
Westminster College USA
Wisconsin State Lab of Hygiene USA
Woods Hole Oceanographic Institution USA
Discussion
The performance of most laboratories was satisfactory (mean score of 3) or better. Of the
63 labs that submitted total mercury results, 50 (79%) received average scores of 3 or higher, 35
-
47
labs (56%) received average scores of 4 or higher, and 8 labs (13%) received the maximum
average score of 5 (Table 6). Of the 41 labs that submitted methylmercury results, 36 labs (90%)
received average scores of 3 or higher, 24 labs (60%) received average scores of 4 or higher, and
3 labs (8%) received the maximum average score of 5 (Table 10). The percentage of labs
receiving THg scores equal to or greater than 3 (satisfactory) has decreased from 2013 and
previous years, while the percentage of labs receiving MeHg scores equal to or greater than 3 has
increased since 2013 (Table 12). Because individual labs are not identifiable in these results, it is
impossible to know if a change in the percentage of labs achieving a given score is due to
improvements or declines in performance, differences in the participant pool from one year to
the next, or differences in the samples being analyzed. It is important to note that sample
variability can have a significant effect on the number of labs receiving satisfactory scores. If a
sample has high natural variability in one analyte, such as total mercury at site MP in Intercomp
2013, it will create a wide range of satisfactory concentrations, driving up scores for that site.
Table 12. Number and percentage of labs achieving scores greater than or equal to 3, 4, and 5 in
each of the four years of the Brooks Rand Interlaboratory Comparison Study.
2011 2012 2013 2014
n % n % n % n %
Mean THg Score ≥ 3 33 79 40 82 50 85 50 79
Mean THg Score ≥ 4 22 52 36 73 45 76 35 56
Mean THg Score = 5 10 24 13 27 18 31 8 13
Mean MeHg Score ≥ 3 25 83 33 92 36 84 36 90
Mean MeHg Score ≥ 4 22 73 28 78 20 47 24 60
Mean MeHg Score = 5 8 27 13 36 6 14 3 8
The fact that the majority of laboratories participating in this study achieved a
satisfactory score for total mercury and methylmercury suggests that the intercomparability of
data among the majority of the participants is good. In every case, laboratories with an average
score of less than three (i.e. large differences from the consensus mean) had outliers for one or
more sites, indicating a significant analytical problem. One of the aims of this study is to help
these laboratories identify their analytical problems so that they can isolate the causes and
eliminate them. Labs with scores greater than three but less than five (i.e. smaller differences
from the consensus mean) may be able to improve their scores by assessing different sample
-
48
digestion/preparation methods, renewing their analytical standards, or cleaning/servicing their
analytical system, but likely do not have significant problems.
Method Data
By collecting information on the methodology and equipment each laboratory used to
analyze the samples, we attempted to assess the effectiveness of different analytical approaches.
We compared methods using t-tests on mean performance scores from each lab (we only
compared groups with n ≥ 4, and used α = 0.05). In Intercomp 2014, as in previous years, there
were few instances in which any method underperformed relative to any other (Tables 13-17).
The reason for the lack of significant differences in performance is likely due primarily to the
size of this study. Although the study involved a large number of participants, the wide variety of
methods in use and inconsistencies in the reporting of method characteristics often resulted in
very small groupings of results for any one method attribute. In many cases, robust comparisons
of performance between methods will require a more targeted study. However, another
conclusion that can be drawn from the lack of significant differences between methods is that a
wide variety of methods and equipment are capable of producing acceptable data. Participants
must assess for themselves whether their score indicates a method problem or other analytical
issue.
We found a significant difference between total mercury analysis using an automated
system (“automated” was defined as having an autosampler) vs. a manual system (p = 0.04;
Table 13), with automated instruments performing better; however, there was no significant
difference between automated and manual methylmercury systems (p = 0.12).
Table 13. Comparison of automated and manual systems. Automated systems are defined as
those with an autosampler. Manual systems are defined as all others. n = number of labs.
THg Median Score THg n MeHg Median Score MeHg n
Automated 4.00 44 4.17 28
Manual 3.83 18 4.00 11
-
49
Unlike in Intercomps 2012 and 2013, in Intercomp 2014, there was no significant
difference between total mercury amalgamation methods (dual gold, single gold, or no
amalgamation; Table 14), although the median score for analysis with no amalgamation is lower
than the others. There were also no significant differences between detector manufacturers
(Table 15), however the vast majority of the participants used Brooks Rand and Tekran
instruments, making it difficult to compare the performance of other instrument manufacturers.
Table 14. Comparison of total mercury amalgamation methods. n = number of labs.
Median Score n
Dual Gold 4.00 36
Single Gold 4.00 19
No Amalgamation 3.33 7
Table 15. Comparison of detector manufacturers. n = number of labs.
THg Median Score THg n MeHg Median Score MeHg n
Agilent (ICP-MS) 1.50 1 3.75 2
Analytik Jena 4.17 2
Brooks Rand 4.00 28 4.00 23
CETAC 4.33 4
Home-made 4.33 1
Leeman 4.67 3
Lumex 1.33 1
Nippon 3.33 1
PerkinElmer (ICP-MS) 4.33 2
PS Analytical 3.00 3
Tekran 4.17 18 4.33 9
Thermo (ICP-MS) 3.33 1 2.00 2
There were no significant differences between sample preparation methods for
methylmercury analysis (Table 16), although distillation-based methods have higher median
scores. As in previous years, the vast majority of labs continue to use distillation, following EPA
Method 1630, however an increasing number of labs are using non-distillation methods. In
Intercomp 2014, 11 labs (of 41 that submitted methylmercury results, 27%) did not distill
methylmercury samples; in Intercomp 2013, that number was 16%; in Intercomp 2012, 25%; in
Intercomp 2011, 20%.
-
50
Table 16. Comparison of methylmercury sample preparation procedures and reagents. n =
number of labs.
Median Score n
DCM Extraction 4.67 1
All Distillation 4.33 28
Distillation / APDC 4.33 10
Distillation / CuSO4 3.67 5
Distillation / KCl/H2SO4 4.50 2
Distillation / L-cysteine 4.33 5
Distillation / No reagents 4.17 6
Direct Ethylation 3.67 6
Other Non-Distillation 3.50 4
The majority of labs performing methylmercury analysis use packed GC columns (71%;
Table 17). Every lab that reported their packing material is using 15% OV-3 on Chromosorb
W/AW, the material specified in EPA Method 1630. Four labs reported using packed columns
but did not report their packing material. Of the labs using capillary columns, the majority are
using a DB-1 x 15 m length x 0.53 mm inner diameter x 5 µm coating thickness, the standard
column in a Tekran automated methylmercury system. The other capillary column type reported
was HP-5. One lab reported using a capillary GC column but did not describe it. One lab is using
ion chromatography (liquid chromatography) to separate mercury species. There were no
significant differences between packed and capillary GC columns, or between any specific
column types.
Table 17. Comparison of methylmercury separation methods. n = number of labs.
Median Score n
All Capillary Column 4.33 9
DB-1 x 15 m x 0.53 mm x 5 µm 4.50 6
HP-5 x 15 m x 0.32 mm x 0.25 µm 0.00 1
HP-5 x 15 m x 0.53 mm x 1.5 μm 4.67 1
All Packed Column 4.00 29
15% OV-3 on Chromosorb W/AW 4.33 25
Ion Chromatography (Dionex CG-5A) 4.50 1
-
51
To investigate possible links between storage conditions and measured total mercury
results, we collected data on the length of time and storage temperature between when BrCl was
added to total mercury samples and analysis. There were effectively no correlations between
storage time and measured concentrations for all three sites (Figure 15). The correlation
coefficients were small and insignificant at the 95% level. There were also no significant
correlations between storage temperature and measured concentrations (Figure 16). Some sample
types have been shown to require more rigorous oxidation than others in order to achieve
complete recovery of total mercury (Olson et al. 1997; Pyhtilä et al. 2012). However either these
samples are not sufficiently complex to require extra oxidation or they require a stronger
digestion than that provided by added time or temperature.
Figure 15. Plot of measured total mercury concentrations (lab means) vs. storage time after BrCl
addition to the samples. Holding times reported as > x hours were not included. Where a range of
values was reported (e.g. 8-12 hours), the mid-point was used.
R² = 0.0308
R² = 0.0061
R² = 0.0178
1
10
100
0 100 200 300 400 500
Me
an T
Hg
Co
nce
ntr
atio
n (
ng/
L)
Storage time (hours)
Site WH
Site UB
Site SP
-
52
Figure 16. Plot of measured total mercury concentrations (lab means) vs. sample storage
temperature after BrCl addition. Room temperature was assumed to be 20 °C. Where a range of
temperatures was given, the midpoint was used. If samples were held at a higher temperature for
several hours, followed by a lower temperature, they are plotted at the higher temperature.
Additional method data beyond what we include in the report did not reveal any
significant differences and were omitted for brevity.
Future Studies
Brooks Rand Instruments conducts this study on an annual basis. Any feedback on this
year’s study or interest in participating in future studies should be directed to Joel Creswell
Acknowledgements
This study was made possible by the dedicated effort and input of many people. Carl
Lamborg at the Woods Hole Oceanographic Institution collected sample WH. Michela
Hernandez at EcoChem, Inc. received, organized, anonymized, and archived all of the data.
Kristina Obu Vazner, Vesna Fajon, Ermira Begu, and Milena Horvat at the “Jožef Stefan”
Institute and John DeWild at the U.S. Geological Survey Wisconsin Mercury Research Lab
generously performed extra analysis for the holding time study. The Hg Group at Brooks Rand
R² = 0.0034
R² = 0.0094
R² = 0.0044
1
10
100
-5 0 5 10 15 20 25 30 35 40 45 50 55 60 65
Me
an T
Hg
Co
nce
ntr
atio
n (
ng/
L)
Storage Temperature (°C)
Site WH
Site UB
Site SP
-
53
Labs (Becky Thorsness, Ashley Pittenger, Ian Joslin) performed holding time analysis and
analyzed site screening samples. James FrederickLindenfelser, Jennie Kim, and Candace
Kittleson packed and shipped sample bottles to participants. Brittany Nelson, Nathan Andersen,
and Frank McFarland provided insight and advice into the statistical analysis. In addition to the
participation fees, additional funding for this study was provided by Brooks Rand Instruments.
We thank everyone involved for helping make this study possible.
References
Cook, R. D., and S. Weisberg. 1980. Characterizations of an empirical influence function for detecting
influential cases in regression. Technometrics 495–508.
Creswell, J., V. Engel, A. Carter, and C. Davies. 2012. 2012 Brooks Rand Labs Interlaboratory
Comparison Study for Total Mercury and Methylmercury (Intercomp 2012). Brooks Rand Labs.
Filliben, J. J. 2012. Chapter 1. Exploratory Data Analysis, In C. Croarkin and P. Tobias [eds.],
NIST/SEMATECH e-Handbook of Statistical Methods.
http://www.itl.nist.gov/div898/handbook/eda/section3/eda35h.htm. Accessed 4/22/14.
Lin, P. E., and X. F. Niu. 1998. Statistical Analysis of Environmental Laboratory Data. Technical Report
LAB012. LAB012 Florida Department of Environmental Protection.
Niu, X.-F., and D. Miller. 2009. Statistical Analysis and Summary of the HgRR9 Mercury Round Robin
Data. Technical Report LAB027. LAB027 Florida Department of Environmental Protection.
Niu, X.-F., and A. Tintle. 2003. Statistical Analysis and Summary of the HgRR3 Mercury Round Robin
Data. Technical Report LAB020. LAB020 Florida Department of Environmental Protection.
Niu, X.-F., and A. Tintle. 2008. Statistical Analysis and Summary of the HgRR8 Mercury Round Robin
Data. Technical Report LAB027. LAB027 Florida Department of Environmental Protection.
Olson, M. L., L. B. Cleckner, J. P. Hurley, D. P. Krabbenhoft, and T. W. Heelan. 1997. Resolution of
matrix effects on analysis of total and methyl mercury in aqueous samples from the Florida
Everglades. Fresenius J. Anal. Chem. 358: 392–398.
-
54
Pyhtilä, H., P. Perämäki, J. Piispanen, M. Niemelä, T. Suoranta, M. Starr, T. Nieminen, M. Kantola, and
L. Ukonmaanaho. 2012. Development and optimization of a method for detecting low mercury
concentrations in humic-rich natural water samples using a CV-ICP-MS technique.
Microchemical Journal 103: 165–169.
U.S. Environmental Protection Agency. 1996. Method 1669: Sampling Ambient Water for Trace Metals
at EPA Water Quality Criteria Levels.