The Science of the Total Environment 304 (2003) 295?303
0048-9697/03/$ - see front matter 
 2002 Elsevier Science B.V. All rights reserved.
PII: S0048-9697 ? 02 . 00576-4
Application of ultrafiltration and stable isotopic amendments to
field studies of mercury partitioning to filterable carbon in lake
water and overland runoff
Christopher L. Babiarz *, James P. Hurley , David P. Krabbenhoft , Cynthia Gilmour ,a, b c d
Brian A. Branfireune
Environmental Chemistry and Technology Program, University of Wisconsin, 660 N. Park St., Madison, WI 53706, USAa
Aquatic Science Center, University of Wisconsin-Madison, 1975 Willow Drive, Madison, WI 53706, USAb
Water Resources Division, United States Geological Survey, 8505 Research Way, Middleton, WI 53562, USAc
Academy of Natural Sciences, Benedict Estuarine Research Center, St. Leonard, MD 20685, USAd
Department of Geography, University of Toronto, Mississauga, Ont., Canada L5L 1C6e
Received 20 February 2002; accepted 15 October 2002
Abstract
Results from pilot studies on colloidal phase transport of newly deposited mercury in lake water and overland
runoff demonstrate that the combination of ultrafiltration, and stable isotope amendment techniques is a viable tool
for the study of mercury partitioning to filterable carbon. Ultrafiltration mass balance calculations were generally
excellent, averaging 97.3, 96.1 and 99.8% for dissolved organic carbon (DOC), total mercury (Hg ), andT
methylmercury (MeHg), respectively. Sub nanogram per liter quantities of isotope were measurable, and the observed
phase distribution from replicate ultrafiltration separations on lake water agreed within 20%. We believe the data
presented here are the first published colloidal phase mercury data on lake water and overland runoff from
uncontaminated sites. Initial results from pilot-scale lake amendment experiments indicate that the choice of matrix
used to dissolve the isotope did not affect the initial phase distribution of the added mercury in the lake. In addition
there was anecdotal evidence that native MeHg was either recently produced in the system, or at a minimum, that
this ?old? MeHg partitions to the same subset of DOC that binds the amended mercury. Initial results from pilot-scale
overland runoff experiments indicate that less than 20% of newly deposited mercury was transported in the filterable
fraction (-0.7 mm). There is some indication of colloidal phase enrichment of mercury in runoff compared to the
phase distribution of organic carbon, but the mechanism of this enrichment is unclear. The phase distribution of
newly deposited mercury can differ from that of organic carbon and native mercury, suggesting that the quality of
the carbon (available ligands), not the quantity of carbon, regulates partitioning. Further characterization of DOC is
needed to clarify the underlying mechanisms.

 2002 Elsevier Science B.V. All rights reserved.
Keywords: Isotopes; Colloids; Runoff; Lakes; Mercury; Methylmercury; Ultrafiltration
 Submitted to the Proceedings of the Sixth International Conference on Mercury as a Global Pollutant Minamata, Japan:
October 15?19, 2001.
*Corresponding author. Tel.: q1-608-265-5085; fax: q1-608-262-0454.
E-mail address: babiarz@cae.wisc.edu (C.L. Babiarz).
296 C.L. Babiarz et al. / The Science of the Total Environment 304 (2003) 295?303
1. Introduction
In response to considerable evidence that atmos-
pheric transport, deposition, and reemission of
mercury are key processes in the movement of this
neurotoxin throughout the globe, scientists are
studying the fate of newly deposited mercury on a
watershed in northwest Ontario, Canada (Hintel-
mann et al., 2002; Harris et al., 2001; Renner,
2001; Rouhi, 2001). In the Mercury Experiment
to Assess Atmospheric Loading In Canada and the
United States (METAALICUS), scientists will be
able to differentiate between the newly deposited
mercury and standing pool of mercury using stable
isotope amendments. These isotopic techniques
will provide the first direct evidence of a watershed
response to changing atmospheric inputs of mer-
cury and will inform pending controls on mercury
emissions that may exceed several billion dollars
to implement (USEPA, 1998).
One of the major goals of the METAALICUS
project is to determine the mobility of this ?new?
mercury through the watershed, and assess the
recovery time for an ecosystem should new indus-
trial emissions of Hg be ceased. One of the key
transport vectors for mercury in overland runoff,
and within lakes, may be the colloidal phase
(roughly 0.7 mm?10 kilo Daltons (kDa)). Recent
studies have shown that the colloidal phase can be
a large component of the -0.4 mm filtered phase
mercury concentration, and that organic carbon
plays an important role in mercury phase partition-
ing in freshwaters (Cai et al., 1999; Babiarz et al.,
2001). Combining stable isotopic amendments
with ultrafiltration presents new analytical chal-
lenges. Before full-scale implementation of the
larger experiment, it is important to verify the
integrity of data produced by novel techniques so
that experimental hypotheses developed from the
data are valid.
In this paper we present the first data on
colloidal phase partitioning of both native mercury
and newly deposited stable isotopes of mercury in
lake water and overland runoff. In particular we
assess the colloidal phase partitioning of native
mercury in lake water and runoff, and we compare
the phase distribution results for isotopic mercury
from different amendment application methods.
These pilot-scale experiments demonstrate the
effectiveness of combining ultrafiltration with sta-
ble isotope techniques to study colloidal phase
transport of mercury. In addition, the data present-
ed informs several hypotheses on factors control-
ling colloidal phase partitioning of newly deposited
mercury that can be tested in future work.
2. Methods
The work presented here is part of the whole-
ecosystem METAALICUS Project, currently
underway within the Experimental Lakes Area of
the Canadian Boreal Forest in northwestern
Ontario.
To evaluate the effectiveness of tangential-flow
ultrafiltration, we applied the technique to subsam-
ples from larger pilot-scale experiments designed
to determine the most appropriate method for
introducing stable isotopic amendments to various
watershed compartments. A full discussion of the
numerous factors that influence a full-scale, non-
invasive, timely, and executable amendment is
beyond the scope of this paper. In general, the
project goals were to select variables (target spike
concentration, final lake concentrations, delivery
mechanism, delivery time, and delivery matrix)
such that defendable scientific method and practi-
cal field operations could both be maintained
(Harris et al., 2001).
Salient details of the two ultrafiltration-specific
experiments are outlined below. The first examines
the effect of the spike solution matrix on the initial
partitioning of the spike in the epilimnion of the
lake. The second experiment examines the relative
colloidal phase mobility of the spike in overland
runoff after wet and dry application of the spike
to the watershed.
2.1. Does the delivery solution affect the partition-
ing of the amendment in surface waters?
The complexation of Hg in natural deposition,
or in an ecosystem spike, may affect the ligand
reactivity, reduction, evasion, and methylation of
Hg. However, for the full-scale application of the
amendment to the lake and its watershed, it would
be impractical to collect a sufficient quantity of
297C.L. Babiarz et al. / The Science of the Total Environment 304 (2003) 295?303
fresh precipitation to mimic a rain event. Even if
it were practical, the number of binding sites on
ligands in natural rainwater would likely be satu-
rated by the concentration of mercury (2.5
mg l ) required for delivery of the amendmenty1
in a reasonable quantity of water and time. Further,
in order to keep the spike in solution, and resistant
to photo reduction during the lengthy delivery
process, some quantity of ligand to stabilize the
spike was desirable. Logical alternatives to rain-
water included equilibrating the spike in large
volumes of lake water or an artificial matrix
containing a strong mercury-binding ligand similar
to that in rain. The specific concern to be tested
with ultrafiltration was whether the quantity and
chemical quality of the ligands dissolved in the
delivery solution would affect the initial partition-
ing of mercury in the lake.
To test the leading matrix candidates, three 5-l
Teflon bottles were filled with unfiltered surface
water from Lake 658. Each bottle was spiked to
bring the concentration of HgCl in the bottle201 2
to ;1.3 ng l (the concentration expected fromy1
a single application of the full-scale spike mixed
into a 2 m depth of the lake). One bottle was
spiked using a 130 mg l solution of HgCl iny1 201 2
a 6.5 mM solution of thiolglycolate at pH 3.5
(L658-RSH). Two bottles were spiked using a 130
mg l solution of HgCl dissolved in unfilteredy1 201 2
lake water at a pH of 3.3. One of the latter bottles
was processed immediately (L658-T0) and the
other was kept at near ambient conditions (capped,
dark, and refrigerated) for 3 days before processing
to assess equilibration effects (L658-T3). Our
hypothesis is that the kinetics of re-equilibration
of Hg with ligands in lake water depends on the
complexation of Hg in the spike (or in natural
deposition), and that re-equilibration may be slow
enough that the reduction and partitioning of Hg
originating from the spike could differ from that
originating in natural deposition.
Although the full-scale experiment will not
include MeHg as an amendment, the spiking solu-
tions described above also contained enough
MeHg to raise the Hg concentration to ;0.3199
ng l . This MeHg amendment was primarilyy1
introduced to assess reduction and evasion in
another component of the larger pilot-scale
experiment.
2.2. Does the delivery method affect colloidal
phase mobility of isotope in overland runoff?
Two application methods were used to deliver
HgCl on separate upland test patches. The201 2
upland sites, contained bedrock outcroppings, thin
soils, shrub mosses, and low-lying mixed vascular
plants. In one application, the isotope was diluted
to 4.1 ng l in 1400 l of lake water in a largey1
polyethylene tank, equilibrated, and sprayed onto
a small (120 m ) plot through a sprinkler system2
delivering water at a rate of 14 mm h , equivalenty1
to a typical rainstorm in the region. A composite
sample of runoff was collected by integrating small
subsamples of water over the full time-course of
the experiment (;30 min). The runoff was col-
lected into a 5-l bottle from the outlet of an acid-
cleaned acrylic weir. In the second application, 25
mg m of isotope was sprayed in a fine mist thaty2
evaporated shortly after contact with the ground.
Runoff from a rainstorm that occurred the follow-
ing day was diverted through a series of acid-
cleaned acrylic weirs into a 300-l polyethylene
storage tank, and the 5-l subsample used for
ultrafiltration was collected from the same tank.
2.3. Field and analytical procedures
Whole-water samples were filtered in-line using
pre-ashed quartz fiber filters rated at a 0.7 mm
pore size (Whatman QMA). Conventional filtrates
were further processed using a 10 kDa (f0.0015
mm) ultrafiltration membrane to isolate the col-
loidal and dissolved phases. All ultrafiltration sep-
arations were conducted in a clean room and fully
mass balanced as described by Babiarz et al.
(2000) and Hoffmann et al. (2000) and Babiarz
et al. (2001). Unless otherwise noted ultrafiltration
separations were performed within 2 h of sample
collection using 0.23 m spiral-wound regenerated2
cellulose membranes (Millipore model PLGC). All
sample containers, tubing and pump heads were
composed of Teflon . An exhaustive, multiple step
cleaning process was employed to prepare the
298 C.L. Babiarz et al. / The Science of the Total Environment 304 (2003) 295?303
cartridges for trace-metal applications (Babiarz et
al., 2000; Hoffmann et al., 2000).
During an ultrafiltration separation, the
membrane-passing fraction (permeate) was col-
lected at a rate of ;200 ml min . The non-y1
passing fraction (retentate) was recirculated at
;1100 ml min until the ratio of the retentatey1
volume to the feed volume (concentration factor)
was approximately 3:1. The concentration factor
was kept low to avoid changes in equilibrium
metal partitioning to organic carbon. A mass bal-
ance was conducted with each separation as part
of the QAyQC protocol. Mass balance closure was
calculated from the following fractions: (1) the
feed solution, (2) the pre-conditioning solution
(sampled after the pre-conditioning step was com-
pleted), (3) the final retentate, (4) several perme-
ate subsamples, (5) a post-separation Milli-Q
flush, and (6) a dilute sodium hydroxide rinse (0.1
N).
Mercury mass balance results were generally
excellent, averaging 97.3% (S.D.s9.4) for dis-
solved organic carbon (DOC), 96.1% (S.D.s18.4)
for Hg , and 99.8% (S.D.s9.2) for MeHg (TableT
1). These error terms are comparable to our lab-
oratory analytical protocols that require the reana-
lysis of individual samples until the relative
standard deviation is within 10% and spike recov-
eries are within 25%.
Total mercury concentrations were determined
using the bromine monochloride oxidation tech-
nique followed by stannous chloride reduction,
nitrogen purging, gold-trap pre-concentration, ther-
mal desorption and cold-vapor atomic fluorescence
spectroscopy (CVAFS) detection (Gill and Fitz-
gerald, 1987; Liang and Bloom, 1993; Babiarz et
al., 1998). MeHg was determined by distillation,
aqueous phase ethylation, nitrogen purging,
Carbotrap pre-concentration, thermal desorption,
chromatographic separation, pyrolytic conversion
to Hg8, and CVAFS detection (Bloom 1989; Liang
et al., 1994; Babiarz et al., 1998). Analytical
detection limits (three times the S.D. of the blank)
average 0.1 ng l Hg and 0.05 ng l MeHg.y1 y1T
Isotopic Mercury analysis was determined using
a Perkin-Elmer Elan 6100 that was dedicated for
mercury-only analysis and was housed in a state-
of-the-art mercury analytical facility operated by
Dr. David Krabbenhoft at the USGS in Middleton,
Wisconsin. This ICP-MS was fitted with a contin-
uous flow injector analyzer system, and an inline
gold amalgamation system that allows for rapid
sample throughput and low-level detection. The
method, modeled after the pioneering work of
Hintelmann et al. (1995) and Hintelmann and
Evans (1997), had an absolute detection limit of
approximately 1 pg Hg, or approximately 0.05
ng l . The minimum detectable amount of MeHgy1
was also 1 pg.
The ICP-MS was calibrated using native Hg
stock solutions diluted from a primary standard. A
calibration curve was determined for the following
isotopes: Hg, Hg, Hg, Hg, and Hg.198 199 200 201 202
For a given sample, native Hg concentrations were
determined by averaging the value calculated for
each isotope that was not enriched ( Hg, Hg,198 200
Hg). Isotopic (spiked) Hg concentrations were202
determined as the difference between the value
calculated for that isotope ( Hg or Hg) and199 201
the averaged native concentration, multiplied by
the natural abundance of the isotope in question.
DOC samples were collected in pre-ashed glass
bottles, and stored dark in a cold room. DOC
concentrations were determined on a Shimadzu
TOC-5000 using high temperature (680 8C) cata-
lytic oxidation. Analytical uncertainty is typically
"0.1 mg l .y1
3. Results and discussion
Results from our pilot-scale experiments using
tangential-flow ultrafiltration to probe the affects
of differing spike application methods on the
partitioning of the added mercury are shown in
Table 1. Beyond the immediate demonstration that
the techniques are viable, it is too early to draw
definitive conclusions regarding the biogeochem-
istry of newly deposited mercury. However, the
data point to several key questions that will guide
future research in the METAALICUS project.
3.1. Lake additions
The three L658 spiked waters averaged 2.4
(S.D.s0.5) ng l native Hg unfiltered, and 0.6y1 T
(S.D.s0.2) ng l isotopic Hg unfiltered. A post-y1 T
299
C
.L.B
abiarz
et
al.
/
The
Science
of
the
Total
E
nvironm
ent
304
(2003)
295?303
Table 1
Organic carbon, total mercury, and methylmercury as a percentage of the filtered value
Organic carbon Total mercury Methylmercury
Filtered UF mass Colloidal Dissolved Filtered UF mass Colloidal Dissolved Filtered UF mass Colloidal Dissolved
-0.7 mm Balance
0.7 mm?10
kDa
-10 kDa
-0.7 mm Balance
0.7 mm?10
kDa
-10 kDa
-0.7 mm Balance
0.7 mm?10
kDa
-10 kDa(mg l )y1 (%)
(mg l )y1 (%) (mg l )y1 (%)
(ng l )y1 (%)
(ng l )y1 (%) (ng l )y1 (%)
(ng l )y1 (%)
(ng l )y1 (%) (ng l )y1 (%)
Cascade
Native 8.0 102.2 3.1 38.5 5.1 63.7 7.7 73.2 2.7 35.0 2.9 38.2 0.47 91.3 0.20 43.6 0.22 47.7
Amended b b b b b b 0.8 79.2 0.5 57.3 0.2 21.9 a a a a a a
U1F
Native 7.1 80.7 2.1 30.0 3.6 50.7 21.4 110.1 12.0 56.1 11.6 54.0 0.10 121.4 0.03 32.1 0.09 89.3
Amended b b b b b b 0.11 74.2 0.0 36.1 0.0 38.1 a a a a .a a
L658 T0
Native 9.1 100.8 5.8 63.3 3.4 37.5 2.6 126.4 1.7 66.3 1.6 60.0 0.20 96.1 0.12 58.5 0.08 37.6
Amended b b b b b b 0.75 96.3 0.6 74.5 0.2 21.8 0.51 99.5 0.35 68.6 0.16 30.9
L658 T3
Native 8.7 103.3 4.7 53.8 4.3 49.4 2.0 102.0 1.0 47.6 1.1 54.3 0.12 97.6 0.08 66.9 0.04 30.8
Amended b b b b b b 0.4 88.6 0.2 62.4 0.1 26.2 0.31 93.7 0.20 64.0 0.09 29.8
L658 RSH
Native 9.0 99.5 5.4 59.7 3.6 39.8 1.8 119.2 1.2 68.3 0.9 50.9 0.10 100.4 0.07 74.6 0.03 25.8
Amended b b b b b b 0.4 91.6 2.7 71.2 2.9 20.5 0.20 98.4 0.20 71.6 0.22 26.8
Mean 8.4 97.3 4.2 49.1 4.0 48.2 3.8 96.1 2.3 57.5 2.2 38.6 0.25 99.8 0.16 60.0 0.12 39.8
S.D. 0.8 9.4 1.6 14.3 0.7 10.4 6.6 18.4 3.6 14.0 3.5 15.4 0.16 9.2 0.10 14.8 0.08 21.2
Median 8.7 100.8 4.7 53.8 3.6 49.4 1.3 94.0 1.1 59.9 1.0 38.2 0.20 98.0 0.16 65.5 0.09 30.9
Minimum 7.1 80.7 2.1 30.0 3.4 37.5 0.1 73.2 0.0 35.0 0.0 20.5 0.10 91.3 0.03 32.1 0.03 25.8
Maximum 9.1 103.3 5.8 63.3 5.1 63.7 21.4 126.4 12.0 74.5 11.6 60.0 0.51 121.4 0.35 74.6 0.22 89.3
Methylmercury isotope was not added to the solution.a
Not applicable because carbon isotope ratios were not modified.b
300 C.L. Babiarz et al. / The Science of the Total Environment 304 (2003) 295?303
Fig. 1. Results from lake water spike matrix comparison. L658-T0 used a lake water spike matrix and was processed immediately
after the spike addition. L658-T3 used a lake water spike matrix and was processed 3 days after the addition. L658-RSH used a
thioglycolate matrix and was processed 3 days after the addition.
separation 0.01 N HNO rinse of the process bottle3
recovered 13 pg of Hg, indicating a loss to the201
walls of -1% of the amendment. Of the mercury
in solution, 12% of the native Hg was in theT
particulate ()0.7 mm) phase compared to 3% of
the isotopic Hg . The particulate phase is relativelyT
less important for newly added mercury. A similar
result is true for MeHg. The three L658 spiked
waters averaged 0.16 (S.D.s0.06) ng l nativey1
MeHg unfiltered, and 0.36 (S.D.s0.15) ng ly1
isotopic MeHg unfiltered. The post-separation
nitric acid rinse recovered 2 pg of isotopic MeHg,
again indicating a loss to the walls of -1% of the
spike. Of the MeHg in solution, 10% of the native
MeHg was in the particulate ()0.7 mm) phase
compared to 5% of the isotopic MeHg. Again, the
particulate phase is relatively less important for
newly added mercury.
The filtered phase distribution between the col-
loidal and -10 kDa fractions offer several prac-
tical observations, and raise several hypotheses for
future research. First, the ultrafiltration technique
was successfully combined with stable isotopic
techniques?as validated by the good agreement
in the observed phase distribution of carbon and
the native mercury species in Fig. 1. As far as we
are aware, these are the first reported colloidal
phase data from lake waters. Second, the choice
of amendment matrix did not affect the initial
partitioning of new mercury?as evidenced by the
similar phase distributions of the isotope amend-
ments. However, within 3 days of the spike, neither
301C.L. Babiarz et al. / The Science of the Total Environment 304 (2003) 295?303
Fig. 2. Results from the upland spike application method comparison.
matrix (HgCl or Hg-thioglycolate) reached the2
same equilibrium between the dissolved and col-
loidal phases as that of native Hg in L658. These
observations suggest that (a) equilibration times
between newly delivered Hg and lake water
ligands may be days to weeks (instead minutes to
hours as assumed); (b) Hg is transported from
uplands and wetlands as non-reactive complexes;
or (c) non-reactive Hg complexes form slowly
over time in lake water.
Intriguing observations that will guide future
work include the apparent predictive capacity of
the organic carbon distribution for the native HgT
phase distribution. Presumably reflecting the long-
known association of mercury with organic carbon
(Andren and Harriss 1975; Mierle and Ingram
1991), the similar distributions may indicate the
long-term equilibrium of ?old mercury? in the
system. This observation may also reflect a subset
of Hg that has become recalcitrant and removedT
from the pool of mercury available for methylation.
This supposition is backed, in part, by the observed
phase distribution of both inorganic Hg(II) and
MeHg isotopes. Both distributions are quite similar
to each other and to the native MeHg distribution.
We hypothesize that on average, native MeHg may
be ?newer? than native Hg , or at minimum, thatT
native MeHg, isotopic Hg(II), and isotopic MeHg
may to partition to the same subfraction (ligand
quality) of DOC.
3.2. Watershed additions
Two methods were used to apply the isotope to
separate watershed test plots. The first was a
sprinkler system that delivered the isotope in 1400
l of lake water, thus simultaneously creating the
overland runoff that was sampled. The second
method was a small-volume fine mist application
accomplished using manual sprayers. The latter
method relied on a natural rain event to create the
overland runoff that was sampled.
Several practical conclusions can be drawn from
the results presented in Fig. 2, but additional
questions will also direct future research. To our
knowledge, these are the first reported colloidal
phase mercury data in overland runoff from non
contaminated sites.
302 C.L. Babiarz et al. / The Science of the Total Environment 304 (2003) 295?303
In general, very little of the added mercury was
transported in the filtered phase, regardless of the
spike application method (Sprinkler: 4.1 ng ly1
applied vs. 0.8 ng l observed. Fine Mist: 25-y1
mg m applied vs. 0.1 ng l observed). The resty2 y1
of the isotopic mercury must have either stuck to
the bedrock, evaded to the atmosphere, or traveled
with the particulate phase.
Although the source water was different, the
organic carbon phase distributions appear inde-
pendent of application method. On a concentration
basis alone, this observation may indicate that the
source of the organic matter is largely from the
upland surface itself. However, the color of the
two samples of runoff indicates a different ligand
composition within the organic carbon in each
experiment. The fine mist was applied to an upland
plot with deeper soils, and the runoff was darker
in color.
The native Hg phase distribution appears fol-T
low that of organic carbon for both experiments,
but there may be evidence of enrichment in the
colloidal phase (Fig. 2). This enrichment could be
due to several mechanisms including the stripping
of native Hg from the -10 kDa phase as theT
overland runoff drains the watershed. That is, the
depletion of more reactive species (weak complex-
es and free ions) by photoreduction and evasion,
or by adhering to the watershed surface. Alterna-
tively, direct colloidal-phase enrichment of native
Hg could result from disintegration products ifT
the mercury mainly traveled with the particulate
phase. Unfortunately, we do not have particulate
phase data from these samples to test this alternate
hypothesis.
For the fine mist application, the isotopic HgT
appears strongly depleted in the -10 kDa fraction.
This observation is likely due to the contact and
drying time between deposition and the rain event
that created the runoff, leaving the colloidal phase
to emerge as an important vector for transport
within the filtered fraction. For the sprinkler appli-
cation, the striking enrichment of native MeHg in
the -10 kDa may be a result of flushing from the
landscape.
To some extent, these differences in phase dis-
tribution for both the isotopic Hg and the nativeT
MeHg likely reflects differing ligands in the DOC
from each experiment. In addition, the phase dis-
tribution of the feed water for the experiments
should also be determined to assess questions of
source apportionment in each compartment. In
general, our results suggest that it is important to
further characterize DOC quality.
Acknowledgments
The authors express their gratitude to Karen
Scott, Ken Sandilands, Kris Rolfhus, Steve Hoff-
mann, Dave Owens, Georgia Riedel, Tyler Bell,
Mark Olson, Shane Olund, Carl Mitchell, and John
DeWilde for their help with fieldwork, analysis,
and sampling design. This work was funded, in
part, by a grant from the Electric Power Research
Institute, Rick Carlton project officer; the United
States Geological Survey; the United States Envi-
ronmental Protection Agency, STAR grant R82-
7631; and the Natural Sciences and Engineering
Council of Canada.
References
Andren A, Harriss R. Observations on the association between
mercury and organic matter dissolved in natural waters.
Geochim Cosmochim Acta 1975;39:1253 ?1257.
Babiarz CL, Hoffmann SR, Shafer MM, Hurley JP, Andren
AW, Armstrong DE. A critical evaluation of tangential-flow
ultrafiltration for trace metal studies in freshwater systems.
Part 2. Total mercury and methylmercury. Environ Sci
Technol 2000;34:3428 ?3434.
Babiarz CL, Hurley JP, Benoit JM, Shafer MM, Andren AW,
Webb DA. Seasonal influences on partitioning and transport
of total and methylmercury in rivers from contrasting water-
sheds. Biogeochemistry 1998;41:237 ?257.
Babiarz CL, Hurley JP, Hoffmann SR, Andren AW, Shafer
MM, Armstrong DE. Partitioning of total mercury and
methylmercury to the colloidal phase in freshwaters. Environ
Sci Technol 2001;35:4773 ?4782.
Bloom N. Determination of picogram levels of methylmercury
by aqueous phase ethylation, followed by cryogenic gas
chromatography with cold vapor atomic fluorescence detec-
tion. Can J Fish Aquat Sci 1989;46:1131 ?1140.
Cai Y, Jaffe R, Jones RD. Interactions between dissolved?
organic carbon and mercury species in surface waters of the
Florida Everglades. Appl Geochem 1999;14:395 ?407.
Gill GA, Fitzgerald WF. Picomolar mercury measurements in
seawater and other materials using stannous chloride reduc-
tion and two-stage gold amalgamation with gas phase
detection. Mar Chem 1987;20:227 ?243.
Harris R, Rudd JWM, Amyot M, Babiarz C, Beaty K,
Blanchfield P, Bodaly RA, Branfireun B, Gilmour CC,
303C.L. Babiarz et al. / The Science of the Total Environment 304 (2003) 295?303
Heyes A, Hintlemann H, Hurley J, Kelly C, Krabbenhoft
D, Lindberg S, Paterson M, Podemski C, Rolfhus K,
Sandilands K, Scott K, Southworth G, St. Louis V.
METAALICUS: A whole-ecosystem study to determine the
relationship between atmospheric mercury deposition and
concentration in fish. In: Yamamoto M, editor. Abstracts of
the Sixth International Conference on Mercury as a Global
Pollutant, 15 October?19 October 2001, Minamata, Japan;
2001. p. 123.
Hintelmann H, Harris R, Heyes A, Hurley JP, Kelly CA,
Krabbenhoft DP, Lindberg S, Rudd JWM, Scott KJ,
St. Louis VL. Reactivity and mobility of new and old
mercury in a boreal forest ecosystem during the first year
of the METAALICUS study. Environ Sci Technol
2002;36:5034 ?5040.
Hintelmann H, Evans RD. Application of stable isotopes in
environmental tracer studies?measurement of monomethyl-
mercury (CH Hg ) by isotope dilution ICP-MS and detec-q3
tion of species transformation. Fresenius J Anal Chem
1997;358:378 ?385.
Hintelmann H, Evans RD, Villeneuve JY. Measurement of
mercury methylation in sediments by using enriched stable
mercury isotopes combined with methylmercury determina-
tion by gas chromatography?inductively coupled plasma
mass spectrometry. J Anal At Spectrom 1995;10:619 ?624.
Hoffmann SR, Shafer MM, Babiarz CL, Armstrong DE. A
critical evaluation of tangential-flow ultrafiltration for trace
metal studies in freshwater systems. Part 1. Organic carbon.
Environ Sci Technol 2000;34:3420 ?3427.
Liang L, Bloom NS. Determination of total mercury by single-
stage gold amalgamation with cold vapor atomic spectro-
metric detection. J Anal At Spectrom 1993;8:591 ?594.
Liang L, Horvat M, Bloom NS. An improved speciation
method for mercury by GCyCVAFS after aqueous phase
ethylation and room temperature precollection. Talanta
1994;41:371 ?379.
Mierle G, Ingram R. The role of humic substances in the
mobilization of mercury from watersheds. Water Air Soil
Pollut 1991;56:349 ?357.
Renner R. Follow the mercury. Environ Sci Technol
2001;35:229A?230A.
Rouhi AM. Whereto, mercury? Unique experiment to track
fate of mercury added to an ecosystem is under way in
Canada. Chem Eng News 2001;79:35 ?38.
USEPA. Mercury Report to Congress. US Government Printing
Office, Washington DC; 1998.