Journal of Great Lakes Research 40 (2014) 347–359

Contents lists available at ScienceDirect

Journal of Great Lakes Research

journal homepage: www.elsevier.com/locate/jglr

A mercury transport and fate model (LM2-Mercury) for mass budget

assessment of mercury cycling in Lake Michigan
Xiaomi Zhang a, Kenneth R. Rygwelski b,⁎, Russell G. Kreis Jr. b, Ronald Rossmann c
a
Trinity Engineering Associates, Inc Large Lakes Research Station, 9311 Groh Road, Grosse Ile, MI 48138, USA
b
U.S. Environmental Protection Agency, Office of Research and Development, National Health and Environmental Effects Research Laboratory, Mid-Continent Ecology Division,
Large Lakes and Rivers Forecasting Research Branch, Large Lakes Research Station, 9311 Groh Road, Grosse Ile, MI 48138, USA
c
Large Lakes Research Station, 9311 Groh Road, Grosse Ile, MI 48138, USA

a r t i c l e i n f o a b s t r a c t

Article history: LM2-Mercury, a mercury mass balance model, was developed to simulate and evaluate the transport, fate, and
Received 7 August 2013 biogeochemical transformations of mercury in Lake Michigan. The model simulates total suspended and
Accepted 10 February 2014 resuspendable solids (TSRS), dissolved organic carbon (DOC), and total, elemental, divalent, and methylmercury
Available online 26 April 2014
as state variables. Simplified processes among the mercury state variables including net methylation, net reduc-
tion of divalent mercury, and reductive demethylation are incorporated in the model. Volatilization of elemental
Communicated by Erik Christensen
mercury as a kinetic (phase transfer) process and partitioning of total, divalent, and methylmercury as a set of
Index words: instantaneous equilibrium processes were also simulated. The model was calibrated to data collected in 1994
Water quality modeling and 1995 and corroborated by comparing model output generated from a long-term model hindcast to total
Mass balance model mercury measured in high quality sediment profiles. Model hindcast predictions of total mercury in the water
Mass budget assessment column were within estimates of total mercury calculated from observed lake trout bioaccumulation factors.
Mercury Using the model, a mass budget assessment of mercury cycling in the lake was conducted. Atmospheric deposi-
Lake Michigan tion, including wet and dry (particle) deposition and absorption of gaseous divalent mercury, was the dominant
source of total mercury to the lake, followed by sediment resuspension, and then tributary loads. The major loss
mechanism of total mercury from the water was associated with the settling of solids, followed by net volatiliza-
tion. Methylmercury loading associated with wet deposition was the dominant source to the lake, followed by
tributary loadings, and in situ net methylation.
Published by Elsevier B.V. on behalf of International Association for Great Lakes Research.

Introduction
Mercury is a global, persistent, and bioaccumulative toxic pollutant.
Its organic form, CH3Hg (methylmercury or MeHg), is the most biolog-
ically active and toxic form of mercury. It can accumulate in organisms
within the food chain, such as fish, posing a risk to wildlife and
humans.
Smaller lakes have the potential to have more severe mercury
problems than large lakes, such as the Great Lakes in the United
States. With relatively large ratios of area/volume, smaller lakes are like-
ly to have relatively higher mercury concentrations than the larger lakes
in response to atmospheric deposition, which is usually the most
predominant source of mercury to both small and large lakes (Mason
and Sullivan, 1997). In addition, small lakes warm up more quickly
and attain higher summer temperatures than large lakes such as Lake
Michigan. Mercury methylation rate constants increase with increasing
⁎ Corresponding author. Tel.: +734 692 7641.
E-mail address: rygwelski.kenneth@epa.gov (K.R. Rygwelski).
water temperatures (Ulrich et al., 2001), thereby increasing the poten-
tial for the production of MeHg in a smaller lake.
A series of linked numerical models are often used to simulate com-
plex physical and chemical cycling of mercury in aquatic ecosystems
and to evaluate the effects of external mercury inputs on aquatic and
terrestrial organisms and humans that consume them. As an important
component in the linked modeling framework, water quality models
provide both short-term and long-term predictions on the mercury
environmental exposure concentrations in the water column and
sediments of a water body. The predicted environmental exposure
concentrations are used for further predicting the level of mercury
accumulated in the organisms such as fish and their predators.
Mercury transformation processes in an aquatic ecosystem can be
very complex, and some of the processes have not been well defined.
In general, most of the rate constants for these processes have not
been experimentally determined. However, over the last few decades,
significant modeling efforts have been made to describe and simulate
the fate, transport, and biogeochemical transformation of mercury in
aquatic ecosystems. Several notable mechanistic models have been

http://dx.doi.org/10.1016/j.jglr.2014.03.008
0380-1330/Published by Elsevier B.V. on behalf of International Association for Great Lakes Research.
348 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359

developed and applied to various aquatic ecosystems that utilized sim-

plifying assumptions such as those used in this paper (key processes
modeled, equilibrium speciation of mercury complexes, and first-
order biogeochemical kinetics among mercury components). These
models include the MCM (Mercury Cycling Model) by Hudson et al.
(1994), SERAFM by Knightes (2008), SERAFM and Wasp7 by Knightes
et al. (2009), MERC4 by Martin (1992), and WASP5 by Tsiros and
Ambrose (1998).
The United States Environmental Protection Agency (USEPA) select-
ed mercury as one of the target contaminants for the Lake Michigan
Mass Balance Project (LMMBP) to support the development of a Lake
Wide Management Plan (LaMP) for Lake Michigan. A mercury mass bal-
ance model, LM2-Mercury, was developed to: 1) describe and evaluate
the fate, transport, and biogeochemical transformations of mercury in
Lake Michigan; 2) forecast mercury concentrations in the lake (both
water column and sediment); and 3) provide insights for policy makers
to evaluate the long-term responses of the lake to various mercury load-
ing scenarios. The mercury transport and fate model independently
simulates total mercury (tHg, an analytical measurement that combines
all mercury components) and the three individual mercury components
(elemental mercury (Hg0), divalent mercury (Hg2+), and methylmer-
cury (MeHg)) and their interactions with total suspended and
resuspendable solids (TSRS) and dissolved organic carbon (DOC).
MINTEQA2 (Allison et al., 1991), an equilibrium metal speciation
model applied to Lake Michigan water with mean estimates of 22
major cation and anion concentrations for the year 1991, predicted
that N 99% of the dissolved Hg2+ and MeHg was Hg(OH)2 and MeHgOH,
respectively. The LM2-Mercury model was designed to have enough so-
phistication in overall model conceptualization to capture key processes
while maintaining simplicity of speciation processes where permitted.
This paper provides details on model development, conceptual frame-
work, and model input. The results of model calibration and corrobora-
tion as well as a mass budget are presented and discussed. A future
paper will provide results and a discussion on forecasts with the consid-
eration of impacts of global and local contributions to mercury concen-
trations in the lake and a post-audit corroboration. Fig. 1. LM2-Mercury water column segmentation. The numbers on the left side of the
boxes are segment numbers. The numbers on the right side of the boxes are average
Model development and construct segment thickness. This figure is from Zhang et al. (2008) and reused with permission
from Elsevier.

LM2-Mercury is based on the principle of conservation of mass. The

lake water and sediments are represented in the model as a series of dis-
crete computational elements (segments). The sediment compartments
are linked to the water compartments via particle settling, resuspen-
sion, and diffusion processes.
LM2-Mercury adapted the generic code and structures from the Lake
Michigan PCB model, LM2-PCBs (Zhang et al., 2008), including the
water transport fields and the sediment transport scheme. The spatial
segmentation for LM2-Mercury is the same as that used by LM2-PCBs.
The segmentation includes forty-one water segments (Fig. 1 shows
only the surface water segments) with five horizontal layers and 53
surficial sediment segments categorized to address distinct sediment
non-depositional, transitional, and depositional zones. Details of the
model segmentation scheme including sediment segmentation are
provided in the Electronic Supplementary Material (ESM Appendix B).
Several specific modifications were made in the adaptation of LM2-
PCBs for mercury components: 1) the temperature-dependent Henry's
Law constant for tHg and Hg0; 2) air–water exchange formulations for
tHg and Hg0; 3) an approach similar to that used by Vette, 1998 and
Vette et al., 2002) to estimate absorption of gaseous divalent mercury
(reactive gaseous mercury abbreviated as RGM) to Lake Michigan;
4) a distribution factor was added to specify the fraction of the dissolved
phase of total mercury allowed to be volatized; and 5) simplified
speciation processes (Tsiros and Ambrose, 1998) were added for trans-
formation among the mercury components for both the water column
and the sediment. A detailed model description including numerical
method, general mass balance equations for the state variables (total
mercury, mercury components, TSRS, and DOC) in water column and
sediments, and aggregation of the inputs of hydrodynamic fields
based on the outputs from the hydrodynamic model (POMGL —
Princeton Ocean Model for the Great Lakes) can be found in the ESM
Appendix B.
LM2-Mercury simulates TSRS, DOC, total mercury (tHg), and three
mercury components (Hg0, Hg2 +, MeHg) as model state variables.
Among the mercury state variables, tHg, Hg2 +, and MeHg exist in
both the water column and the sediment as dissolved and associated
with particulate phases. In this paper, whenever tHg, Hg2+, and MeHg
are used without a qualifier, they should be interpreted as containing
Hg in both dissolved and particulate phases. Hg0 exists in the system
only in the dissolved phase. Equilibrium partitioning is assumed be-
tween the dissolved and particulate-associated mercury components
in both the water and sediment. The time step used in the model simu-
lations is about 3 h which is assumed to be long enough for mercury or a
mercury component to reach equilibrium between the dissolved and
particulate phases in a large fresh water system. The transport and
fate of mercury is closely linked to the movement of water, TSRS, and
DOC, including processes such as advection, dispersion, settling,
resuspension, partitioning to TSRS and DOC, etc. In addition, the
transformation processes among the mercury components, including
net reduction of inorganic mercury, net methylation, and reductive de-
methylation, are critical to the fate of mercury in the lake. Conceptuali-
zations of the dynamics of TSRS, mercury components, and total
mercury in the lake are given in Appendix A. The conceptual framework
 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359 349

(Appendix A Figs. A1 to A5) is an integration of current scientific knowl-
edge and site-specific parameterization derived from mercury data col-
lected during the period of LMMBP. Water sample collection for LMMBP
was from 1994 to 1995 and sediment sample collection was from 1994
to 1996. The diagrams in Appendix A illustrate the key physical and bio-
geochemical processes that control the concentration of tHg and the
mercury components in both water and surficial sediments of Lake
Michigan.
TSRS and DOC dynamics
air, versus water. The air–water transport kinetics of Hg0 is dominated
by the water-side transfer coefficient. Partitioning to solids is not signif-
icant for Hg0. Pore-water diffusion of Hg0 occurs at the water–sediment
interface. Hg2 + net reduction and reductive demethylation of MeHg
generate Hg0 in the water. The dissolved Hg0 mass change rate (Saw)
in a surficial water segment due to the air–water exchange process
was calculated using the following equation:
  
C
Saw ¼ − kol C dw − a0 A ð3Þ
H

TSRS is one of the two mercury sorbents in the model. Loads of TSRS
from tributaries, shoreline/bluff erosion, primary production, sediment,
and the atmosphere are major sources of TSRS to Lake Michigan and are
included in the model. In the water column (both lake water and
tributary water), TSRS are fine grained, settleable and resuspendable
particles and include biotic carbon, abiotic carbon, clay, and silts. TSRS
settle through the water column to the surficial sediment layer and
can be re-introduced back to overlying water by resuspension. Water
column TSRS is operationally defined as the dried (105 °C) mass
retained on a 0.7 μm glass fiber filter per unit volume of water. Water
column TSRS is often called Total Suspended Solids. In the sediment,
TSRS includes the following resuspendable solids: silt, clay, and particu-
late organic carbon. Sediment TSRS was estimated using the measure-
ments of bulk density, porosity, and organic carbon contents of
sediment samples. The TSRS load derived from bluff erosion
(Montieth and Sonzongni, 1976) includes the following resuspendable
components: clay, silt, and particulate organic carbon (Schawb, 2000).
The estimation by Harrsch and Rea (1982) for the input of fine-
grained inorganic material from the air to Lake Michigan was used as
the TSRS load from the atmosphere in our model. TSRS does not include
a sand fraction due to the low sorption of mercury to this solid phase.
DOC is the second sorbent in the model. It is predominantly com-
posed of humic and fulvic organic acids with colloidal-sized particles
that pass through an ashed GF/F glass fiber filter with a pore size of
0.7 μm. The sources for DOC are mainly from tributaries, a byproduct
from primary production, and decay from particulate organic carbon.
The particulate organic components of TSRS and DOC decay in the
water column and the sediment (Fig. A1 in Appendix A). To avoid intro-
ducing too much complexity, we chose the generalized, temperature-
corrected, Michaelis–Menten equation for decay rate constants of
TSRS and DOC in the water column:
! and tributaries are significant sources of Hg2+ to Lake Michigan. The
kd C T−20
kdw ¼ θ ð1Þ
k1=2 þ C
where kdw is the decay rate constant of TSRS or DOC in water (d− 1),
where d is day; kd is the substrate saturated decay rate constant of
TSRS or DOC in water at 20 °C (d−1); k1/2 is the Michaelis–Menten
half-saturation constant for TSRS or DOC in water (mg/L); C is the
segment-specific concentration of TSRS or DOC in water (mg/L); θ =
1.05 is the Arrhenius temperature coefficient (dimensionless); and T is
where kol is the overall mass exchange rate coefficient (m/d), Cdw is the
segment-specific dissolved Hg0 concentration in a surficial water seg-
ment (ng/m3), Ca is the segment-specific Hg0 vapor concentration
(ng/m3), H′ is the temperature-dependent Henry's law constant for
Hg0 (dimensionless), and A is the surface area of a surficial water
segment (m2).
The overall mass exchange rate coefficient (kol) was calculated using
the Whitman two-film theory formulation (Whitman, 1923):
1
kol ¼   ð4Þ
ð1=kl Þ þ 1=kg H 0
where kl is the liquid film mass transfer rate coefficient (m/d) and kg is
the gas film mass transfer rate coefficient (m/d). Both kl and kg were cal-
culated using the O'Connor equations (O'Connor, 1983). The
temperature-dependent Henry's law constant (H′) for Hg0 was calculat-
ed using the equation of Sanemasa (1975):
H ¼ 10ð T
0 −1078
− logðT Þþ5:592Þ
ð5Þ
where T = interfacial temperature, K.
Divalent mercury dynamics
A conceptualized framework for Hg2 + dynamics can be found in
Appendix A Fig. A3. Hg2+ is reactive and very water soluble. The rate
of air–water exchange is air-side resistance limited. Hg2 + partitions
strongly to fine-grained particles (TSRS) and DOC relative to water
(Stein et al., 1996; Tsiros and Ambrose, 1998). Wet and dry (particle) at-
mospheric deposition and gaseous absorption (Landis and Keeler, 2002)
primary sinks for Hg2+ in the water compartment are through settling,
net reduction to Hg0, and net methylation to MeHg. In the sediment,
major losses for Hg2+ are through burial, resuspension, and net methyl-
ation. The transformations (net reduction and net methylation) of Hg2+
are modeled as either pseudo-first-order sinks in the water column or
first-order sinks in the surficial sediments, based on our best understand-
ing of the processes (Amyot et al., 2000; Tsiros and Ambrose, 1998). In
the water column, the generalized
  and temperature-dependent net re-
duction rate constant kRw−Hg2þ and net methylation rate constant
  d

the segment-specific temperature (°C). kMw−Hg2þ for dissolved Hg2+ are computed using the following
d
In the surficial sediment layer, the equation used for TSRS and DOC equations:
decay rate constants was formulated as:
0 1
kRw−Hg2þ  C d ðT−20Þ
kRw−Hg2þ ¼ −@ Aθ
T−20
kds ¼ kds θ ð2Þ d
ð6Þ
d kð1=2Þ−Hg2þ þ C d
d

where kds is the segment-specific and temperature-dependent decay

rate constant of TSRS or DOC in sediments (d−1) and kds is the decay
0 1
rate constant of TSRS or DOC in sediment at 20 °C. kMw−Hg2þ  C d ðT−20Þ
kMw−Hg2þ ¼ −@ Aθd
ð7Þ
0 d kð1=2Þ−Hg2þ þ C d
Elemental mercury (Hg ) dynamics d

A conceptualized framework for Hg0 dynamics can be found in where kRw−Hg2þ is the substrate saturated net reduction rate constant of
d
Fig. A2 in Appendix A. Hg0 has very strong equilibrium partitioning to Hg2+ in water column at 20 °C (d−1), kMw−Hg2þ is the substrate saturated
d
350 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359
net methylation rate constant of Hg2+ in water column at 20 °C (d−1),
kð1=2Þ−Hg2þ is half-saturation constant for Hg2+ in Lake Michigan water
d
column (ng/L), and Cd is segment-specific dissolved Hg2+ concentration
in water (ng/L).
In the surficial sediment layer, the equations used for calculating
temperature-dependent net methylation rate constants for both partic-
ulate and dissolved Hg2+ were formulated as:
ðT−20Þ
kMs−Hg2þ ¼ −kMs−Hg2þ θ
p p
ðT−20Þ
kMs−Hg2þ ¼ −kMs−Hg2þ θ
d d
where, kMs−Hg2þ
p
, kMs−Hg2þ are net substrate methylation rate constants
d
for particulate and dissolved Hg2 + at 20 °C (d− 1) in sediments,
respectively.
Methylmercury (MeHg) dynamics
Due to its bioaccumulative and toxicological potential, MeHg is an
important mercury component. MeHg can be generated from Hg2 +
under hypoxic or anoxic conditions and the presence of mercury meth-
ylating bacteria. A conceptualized framework for MeHg dynamics can
be found in Appendix A Fig. A4. Air–water exchange through volatiliza-
tion and absorption of MeHg is not significant. MeHg also has a relative-
ly high affinity to fine-grained solids (TSRS) and DOC (Stein et al., 1996;
Tsiros and Ambrose, 1998). Reductive demethylation of MeHg to Hg0 is
shown to occur in the water. The sources of MeHg to the water column
are from wet deposition, tributary loads, and in situ methylation of
Hg2+. It was reported by Glass and Sorensen (1999) that 1.5% of total
mercury wet deposition in the Upper Midwest was methylmercury
based on a six-year trend (1990–1996). Hall et al. (2005) reported
that the percentage of total mercury as MeHg in the majority of precip-
itation events collected at five US sites in the Great Lakes region ranged
from N1% to 6%. Hammerschmidt et al. (2007) reported that the mean
molar ratio of MeHg and Hg2+ measured in continental precipitation
(based on data from the peer-reviewed literature) was 0.025 ± 0.006
(2.5%). Based on these literature values, it was assumed that 2.5% of
the LMMBP-measured tHg in wet deposition was methylmercury.
Hg2+ wet deposition was calculated by subtracting MeHg wet deposi-
tion from tHg wet deposition. The primary losses for MeHg in the
water column are reductive demethylation of MeHg to Hg0 and settling
of the particle-associated fraction to the sediment. The source of MeHg
in surficial sediment is mainly from particulate-bound MeHg settled
from the overlaying water. The losses of MeHg in surficial sediment
are due to resuspension and burial to a deeper sediment layer.
The
 rate constant
of reductive demethylation for dissolved MeHg
kDw−MeHgd in the water column is temperature dependent and
calculated using the following equation:
!
kDw−MeHgd  C d
ðT−20Þ
kDw−MeHgd ¼ − θ
kð1=2Þ−MeHgd þ C d
where kDw−MeHgd is the substrate saturated reductive demethylation
rate constant of MeHg in water column at 20 °C (d−1), kð1=2Þ−MeHgd is
half-saturation constant for MeHg in the Lake Michigan water column
(ng/L), Cd is the segment-specific dissolved MeHg concentration in
water (ng/L).
Total mercury dynamics
Modeling total mercury is a simplification compared to modeling in-
dividual mercury components. Most of the mercury data from the
LMMBP and historical databases are reported as tHg, so modeling tHg
and comparing its output to sums of the individual Hg component con-
centrations from the Hg component model helped us gain confidence
that the mercury component model yielded reasonable results (see
Appendix A Fig. A5). The tHg mercury model has a major limitation
though in that it cannot predict the bioaccumulative MeHg component
in the water. In both the water column and the sediments, tHg exists in
dissolved and particulate (sorbed to TSRS and DOC) phases. Equilibrium
partitioning is assumed between dissolved and particulate tHg. The air–
ð8Þ
water exchange for dissolved tHg is governed by the same equations
used for Hg0 (Eqs. (4) and (5)). However, only the dissolved gaseous
mercury (DGM), a relative small portion (10–30%) of dissolved tHg
(Mason et al., 1995), is subjected to the air–water exchange process.
ð9Þ
Hg0 is often operationally defined as the dissolved gaseous mercury
(DGM) which includes very volatile Hg0 and other volatile mercury
components that are mostly undetectable (Poissant et al., 2000;
Rolfhus and Fitzgerald, 2001; Vette, 1998). Based on the literature
(Mason et al., 1995) and field measurements (Mason and Sullivan,
1997), a distribution factor was estimated and applied to only allow
17% (dissolved gaseous mercury) of dissolved tHg to engage in the
volatilization process. The settling and resuspension of particulate
tHg, diffusion at water–sediment interface for dissolved tHg and
DOC sorbed tHg, and burial process for tHg are also conceptualized
in the model.
Input data and parameterization
The inputs used for the LM2-Mercury calibration, including trans-
port fields, external loads, initial and boundary conditions, and parame-
ters for the processes, were assembled based on field data collected
during the project period (1994–1995), outputs from other models,
and from the literature. Transport fields including bi-directional hori-
zontal advective flows, net vertical advective flows, tributary flows,
and vertical dispersion at adjacent water segments are the same as
those used in LM2-PCBs (Zhang et al., 2008) and are derived from a so-
phisticated hydrodynamic model (Princeton Ocean Model for the Great
Lakes (POMGL), Schwab and Beletsky, 1998). Among the inputs, sources
of TSRS and mercury, field data, and parameterization of physical,
chemical, and biogeochemical processes are of particular importance
to model calibration and results.
TSRS and mercury loads
TSRS loads enter Lake Michigan from shoreline/bluff erosion, primary
production of organic particles, atmospheric deposition, and tributaries
(Table 1). The TSRS load (including fine-grained sediment = silt +
clay) from shoreline erosion was derived based on county by county es-
timates from a technical report to the International Joint Commission
(IJC) (Montieth and Sonzongni, 1976). The load of TSRS derived from
the phytoplankton was estimated based on the primary production
(carbon dry weight) generated from the high-resolution eutrophication
model LM3-Eutro (Pauer et al., 2006) and the conversion of model-
ð10Þ
Table 1
Total suspended and resuspendable sediment (TSRS) annual loads (1994–1995) to Lake
Michigan and data sources.
Loading component Annual load Data source
(ton/year)
Shoreline erosion 5.45 × 106 Montieth and Sonzongni (1976)
Primary production 5.77 × 106 Pauer et al. (2006), Rygwelski
et al. (2011)
Atmosphere 0.82 × 106 Harrsch and Rea (1982)
Monitored tributaries 0.53 × 106 Hall and Robertson (1998)
Unmonitored tributaries 0.18 × 106 Hall and Robertson (1998)
Total 12.75 × 106
 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359 351

Table 2 where, u10 is the wind velocity measured at 10 m above water surface
Mean annual external loads of total mercury and mercury components to Lake Michigan (cm/s), DRGM (Hg2 + molecular diffusivity in air, cm2 s− 1) = 0.09
for 1994–1995.
(Marsik et al., 2007), Dg_H2O (water vapor molecular diffusivity in gas
Loading Annual load (kg/year) Data source phase, cm2 s−1) = 0.26 (Schwarzenbach et al., 2003).
component
tHg Hg0 Hg2+ MeHg
Field data, initial conditions, and vapor concentrations
Tributary 236 8.6a 222b 5.8 USEPA (2004), Hurley et al.
(1996), Hurley et al. (1998)
Dry deposition 53 53 Landis and Keeler (2002) During the 1994–95 project, field data for DOC and TSRS in the water
Wet deposition 616 600c 15.4c Landis and Keeler (2002) column were collected at 41 water survey stations (ESM Appendix B
RGM absorption 651 651 Estimated using Equation (3.1) Fig. S3) during 8 field sampling cruises. The measured DOC in lake
Total 1556 8.6 1526 21.2
water averaged 1.6 mg/L and showed very little spatial or temporal var-
a
Assumes that 17% of dissolved tHg load from the tributaries is Hg0 and that 10–30% of iation (ESM Appendix B). The range of measured TSRS in lake water was
dissolved mercury in water was in the Hg0 form (Mason et al., 1995).
b between 0.11 and 6.84 mg/L with a mean concentration of 1.3 mg/L and
Hg2+ tributary load was calculated by subtracting Hg0 and MeHg tributary loads from
the tHg tributary load. a median concentration of 0.92 mg/L. There was no clear seasonal vari-
c
Based on literature (Hammerschmidt et al., 2007; Hall et al., 2005; Stein et al., 1996), ation of TSRS during the project period. However, a very distinct spatial
assumes that 2.5% of the LMMBP-measured total Hg wet deposition is methyl mercury, variation of TSRS concentration was observed for samples collected in
and then Hg2+ wet deposition was calculated by subtracting MeHg wet deposition from the open lake (0.89 mg/L) versus the nearshore (1.43 mg/L). The TSRS
tHg wet deposition.
field data also showed that at a given station, the concentration of
TSRS increased with water sampling depth.
TSRS concentrations for the sediment layer, where the sand compo-
generated phytoplankton carbon to total algal biomass via a method nent is subtracted due to low Hg partitioning, were estimated based on
from a technical report by Rygwelski et al. (2011). Tributary TSRS loads the measurements of bulk density, porosity, organic carbon contents
provided by the United States Geological Survey (USGS) included loads from the sediment samples (ESM Appendix B Fig. S3), and locational
from 11 monitored tributaries and 18 unmonitored tributaries (Hall information such as depositional and non-depositional zones. The
and Robertson, 1998). range of TSRS concentration in sediment varied from 52,000 to
The external load of total mercury to Lake Michigan is mainly from 250,000 mg/L or from 51 to 218 mg (dw)/g (ww), where dw = dry
the atmosphere and tributaries (Table 2). Direct inputs of mercury weight, ww = wet weight. ESM Appendix B Fig. S4 shows the distribu-
from the atmosphere include dry (particle) and wet deposition and tion of TSRS in the surficial sediments of the lake.
mercury (RGM) absorption. The load from tributaries, including 11 The field data for tHg were collected at 17 water survey stations
monitored and 18 unmonitored tributaries, was measured and estimat- from 6 sampling cruises. The measured tHg concentrations in the
ed for tHg and MeHg in both dissolved and total (dissolved + particu- water column ranged from 0.037 to 0.777 ng/L with a mean ± one stan-
late) forms (Hurley et al., 1998). Atmospheric dry (particle) and wet dard deviation concentration of 0.32 ± 0.12 ng/L. About 60% of tHg in
deposition for tHg was provided by the University of Michigan Air Qual- Lake Michigan is in the dissolved phase. Data analysis based on the
ity Laboratory based on measurements during 1994–95 (Landis and total mercury concentrations measured from the six sampling cruises
Keeler, 2002). It is believed that there is no significant Hg0 associated showed no clear spatial or temporal variation trends in Lake Michigan
with dry (particle) or wet deposition (Mason et al., 1994; Poissant water column during the two-year sampling period. It appears that
et al., 2000). Therefore, it is assumed that the LMMBP-measured atmo- the average tHg concentration (0.34 ng/L) for samples taken in greater
spheric dry (particle) and wet deposition fluxes were primarily Hg2+. than 30 m water depth is slightly higher than the average tHg concen-
The MeHg associated with wet deposition made up a small percentage tration (0.29 ng/L) for the ones taken from less than 30 meter water
of tHg, but it was the largest source of MeHg to the lake. Under the as- depth. In general, the results from our data analysis of tHg in the
sumption that the air–water transport of RGM is exclusively from the water column are consistent with reported values in the literature
atmosphere to water surface, absorption of RGM was treated as a direct (Mason and Sullivan, 1997).
atmospheric load to Lake Michigan. The amount of absorption of RGM Sediment samples were collected during the project (1994–1996)
during the project period was calculated using a simplified form of for mass of total solids, tHg, and MeHg at 117 sediment sample stations
Eq. (3), that included 55 box cores, 61 PONAR grabs, and 1 gravity core (ESM
Appendix B Fig. S3). Rossmann (2002) was used to define the distribu-
  tions of total and MeHg concentrations in the surficial sediments of the
Ca
Saw ¼ kol A ¼ kg C a A ð11Þ main lake (ESM Appendix B Fig. S5 and Fig. S6). The measured tHg
H0
concentrations in the surficial sediments of the main lake ranged from
260 to 35,000 ng/L (or from 2 to 264 ng/g (dw), dw = dry weight)
where Saw is mass flow rate (annual load) for RGM absorption (kg/year), with average tHg concentration of 18,000 ng/L (or 140 ng/g (dw)) for
Ca is the segment-specific Hg2+ vapor concentration (ng/m3), H′ is the box core samples and 4500 ng/L (or 37 ng/g (dw)) for Ponar samples.
temperature-dependent Henry's law constant for RGM (dimension- Based on the work of Rossmann et al. (2001), the range of measured
less), A is the surface area of a surficial water segment (m2), and kg is MeHg concentrations in the surficial sediment varied from 20 to 206
the gas film mass transfer rate coefficient (m/d). The overall mass ex- ng/L (or from 0.16 to 1.68 ng/g (dw)) with an average MeHg concentra-
change rate coefficient (m/d), kol, is approximated as the product of kg tion of 75 ng/L (or 0.57 ng/g (dw)) for box core samples.
and H′ (kg H′), assuming that the liquid film mass transfer rate coeffi- The segment-specific initial concentrations used in the model for
cient (m/d), kl, is much greater than kg H′ in Eq. (4). kg was computed TSRS, DOC, and tHg in the water were based on an interpolation of an-
using the Schwarzenbach formulation (Schwarzenbach et al., 2003) nually averaged cruise measurements collected at the water survey
with correction of Hg2 + (RGM) molecular diffusivity to water vapor stations. The segment-specific initial concentrations used in the model
molecular diffusivity across the air–water interface. Eq. (12) was chosen for TSRS, tHg, and MeHg in the surficial sediment were based on the in-
for its simplicity. terpolation of the measurements collected at the sediment sampling
stations. Due to the lack of measurements and information from the lit-
!0:67 erature for DOC concentration in the lake sediments, the initial concen-
DRGM
kg ¼ ð0:2u10 þ 0:3Þ ð12Þ trations of DOC for the sediment segments were assigned to be equal to
Dg H2 O
the average DOC concentrations in the overlaying water columns.
352 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359
Unlike tHg, there was no routine cruise measurement at the water
survey stations for mercury components (Hg0, Hg2+, and MeHg) and
no measurement at the sediment sampling stations for Hg0 and Hg2+
during the project period. The segment-specific initial concentrations
for Hg0 were estimated by assuming 17% of dissolved tHg in Lake
Michigan being Hg0 (or DGM = dissolved gaseous mercury). Mason
et al. (1995) indicated that 10–30% of total dissolved mercury in fresh-
water is Hg0. For modeling purposes, the lake-wide average initial con-
dition concentration of the estimated Hg0 in our study was 32 pg/L
which is consistent with the value 28 ± 17 pg/L reported in a previous
study of Hg in Lake Michigan (Mason and Sullivan, 1997). The segment-
specific initial concentrations for MeHg were estimated by assuming 4%
of tHg in Lake Michigan was MeHg. The percentage was derived based
on the data (MeHg/tHg = 0.063 pM/1.6 pM = 0.04, where M =
molar) reported from the study by Mason and Sullivan (1997). Given
the limited data available to estimate this value, mercury data from 13
northern freshwater lakes (Watras et al., 1998) were statistically ana-
lyzed using a multiple linear regression of methylmercury as a function
of DOC and pH (ESM Appendix B). This resulted in an estimate of 4.7%
(R2 = 0.71) for Lake Michigan, which corroborated reasonably well
with the 4% measured. These MeHg/tHg ratios are lower than those
occurring in the small 13 northern lakes. Finally, the segment-specific
initial concentrations for Hg2 + were derived by subtracting Hg0 and
MeHg from tHg.
Vapor concentrations serve as forcing functions in air–water ex-
change for tHg and Hg0. The segment-specific time functions of vapor
concentration for tHg were obtained from the University of Michigan
Air Quality Laboratory and were based on the interpolation of the mea-
surements collected at the air sampling stations (ESM Appendix B
Fig. S3). The time functions of vapor concentration for Hg0 were
estimated for Lake Michigan as 98% of tHg vapor concentration, which
is consistent with the percentage reported in the literature (Mason
et al., 1995; Poissant et al., 2000; Vette, 1998; Vette et al., 2002).
Parameterization
Defining values of parameters used for describing physical, chemi-
cal, and biogeochemical processes is very important to the model.
Table A1 in Appendix A lists the parameters associated with the major
processes conceptualized in Fig. A1 through Fig. A5 in Appendix A.
There were very few historical data available to determine the model
parameters for Hg that were specific to Lake Michigan or the Great
Lakes, especially the rate constants of mercury components transforma-
tions which are significantly influenced by various site-specific factors
and temperature (Hudson et al., 1994; Stein et al., 1996; Vette, 1998).
Most of the parameters in Appendix A Table A1 were initialized with
values from a very limited set of references and finalized during
model calibrations.
Model calibration, corroboration, and results
The parameters used in the model (Appendix A Table A1) were not
permitted to assume arbitrary values and were not automatically ad-
justed using statistical analysis in order to obtain the best “curve fits”
in a strictly mathematical sense. Instead, the parameters were initially
specified either from estimation based on field measurements or from
the literature and were then adjusted during model calibration. When
literature values were used, the calibrated parameters were constrained
to be within the reported range.
Samples and measurements gathered during the LMMBP over the
two-year period for water and three-year period for sediments (1994–
1996) produced the most comprehensive data sets to date for the
Lake Michigan system which locally defined the external inputs (such
as loads from tributaries, the atmosphere, and vapor concentrations)
and parameters related to the water transport fields (Schwab and
Beletsky, 1998). Concentrations measured for the water column and
the sediment box core samples were compared to model output during
the model calibrations and further model corroboration. Hundreds of
model calibration runs were conducted in order to finalize the set of
values for the majority of parameters listed in Table A1 in Appendix A.
The overall LM2-Mercury model calibration was carried out in sepa-
rate calibrations. It was first calibrated for TSRS and DOC, and then for
tHg, and finally for the mercury components.
Calibration of TSRS and DOC
Sorption of mercury to fine-grained particles in an aquatic system
increases with increasing organic carbon content of the particles
(Stein et al., 1996; Tsiros and Ambrose, 1998). Measured during the
LMMBP, TSRS in Lake Michigan water is relatively high in organic car-
bon contents (23%). Therefore, mercury tends to partition quite strongly
onto TSRS in Lake Michigan. Using a previous study (Zhang et al., 2008)
on modeling PCBs in Lake Michigan, the TSRS data sets provided helpful
insights and estimates for parameters related to TSRS solids cycling and
dynamics as well as TSRS and DOC biotransformation processes. In that
study, the daily resuspension rates were computed as a function of
measured wave heights and water depth of each segment. The number
of parameters needing adjustment during the calibration of TSRS and
DOC were narrowed down to only six. These included saturated decay
rate constants for the particulate organic components of TSRS and
DOC in the water, a decay rate constant for the particulate organic com-
ponents of TSRS in sediments, yield coefficients of DOC due to TSRS
decay in both water and sediments, and the TSRS settling velocity
(Appendix A Table A1).
The results of TSRS and DOC calibration are presented in Fig. 2. Here
the model-generated volume-weighted, lake-wide average TSRS and
DOC concentrations are plotted against the volume-weighted, lake-
wide average TSRS and DOC cruise measurements during the LMMBP
project period. Modeled outputs for individual water segments in the
epilimnion indicated that there were mild seasonal variations for both
TSRS and DOC due to strong seasonal algae blooms in late spring and
early summer. However, the lake-wide average TSRS and DOC concen-
trations in Fig. 2 show little variation for TSRS and DOC over the two-
year period. There is also evidence of resuspension events (small spikes)
in winter periods presented in the TSRS plot in Fig. 2.
The segment-specific concentrations generated from the model and
the observed data (segment mean concentrations derived from cruise
samples = the data from a cruise sampling averaged by segment) for
DOC and TSRS agreed well (Fig. 3) yielding the correlation coefficients
such as R2 (R2 = 0.6 for DOC; R2 = 0.44 for TSRS) and linear slopes
near 1.0 (1.15 for DOC; 1.08 for TSRS where the slope represents the
mean bias). Some of the disagreement between modeled and measured
3.0 3.0
DOC Cruise Mean
DOC Model Output
TSRS Cruise Mean
TSRS Model Output
TSRS (mg/L)
2.0 2.0
DOC (mg/L)
1.0 1.0
0.0 0.0
Jan-94 Jul-94 Dec-94 Jul-95 Dec-95
Fig. 2. Time series of the model-generated volume-weighted, lake-wide average Total
suspended and resuspendable sediments (TSRS) and DOC (dissolved organic carbon)
concentration versus the volume-weighted, lake-wide average TSRS and DOC cruise mea-
surements from the Lake Michigan Mass Balance Project.
 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359 353

4.00 4.00
y = 1.0801x - 0.0583

TSRS cruise field data mg/L

y = 1.1508x - 0.2918

DOC cruise field data mg/L

R² = 0.6016 R² = 0.4398
3.25 3.00

2.50 2.00

1.75 1.00

1.00 0.00
1 1.75 2.5 3.25 4 0 1 2 3 4
DOC model outputs mg/L TSRS model outputs mg/L

Fig. 3. Observed data (segment-specific cruise mean measurements) versus the model predictions of DOC (left panel) and TSRS (right panel) in respective segments.

values can be attributed to comparison of point measurements to model
output that represents an average over a large segment. In particular,
TSRS has a strong concentration gradient from nearshore to offshore
regions of the lake that complicates comparison of modeled values
from large segments to point measurements. Considering these caveats,
there is reasonable agreement between modeled and measured values.
Calibration of total mercury and mercury components
Due to the simplicity of the conceptualization for its mass balance
(Fig. A5 in Appendix A), tHg was calibrated first. Two parameters,
KTSRS,tHg and KDOC,tHg (partition coefficients for tHg on TSRS and DOC
in lake water), were adjusted during the calibration, and the final values
(Table A1) for these two parameters were within the range of 103.14 to
107.0 reported in the literature (Lyon et al., 1997; Mason and Sullivan,
1997; Miles et al., 2001; Tsiros and Ambrose, 1998).
The initial values of partition coefficients for Hg2 + and MeHg on
TSRS and DOC in lake water were assigned based on the final values of
partition coefficients from the calibration of tHg. The final partition co-
efficients for tHg, Hg2+, and MeHg (Appendix A Table A1) are very com-
parable to the 105.5 to 105.8 for tHg, and 105.7 for MeHg observed by
Mason and Sullivan (1997) and within the range of 102.8 to 106.6 for
MeHg reported in the literature (Lyon et al., 1997; Miles et al., 2001;
Tsiros and Ambrose, 1998). The model calibration for mercury compo-
nents (Hg2 +, Hg0, and MeHg) was conducted through iterations
where the parameters, including partition coefficients and rate

0.8 0.20

Cruise Mean Cruise Mean

0.6 Model Output 0.15 Model Output
∑Hg (ng/L)

Hg0 (ng/L)

0.4 0.10

0.2 0.05

0.0 0.00
Jan-94 Jul-94 Dec-94 Jul-95 Dec-95 Jan-94 Jul-94 Dec-94 Jul-95 Dec-95

0.8 0.04

Cruise Mean Cruise Mean

0.6 Model Output 0.03 Model Output
MeHg (ng/L)
Hg2+ (ng/L)

0.4 0.02

0.2 0.01

0.0 0.00
Jan-94 Jul-94 Dec-94 Jul-95 Dec-95 Jan-94 Jul-94 Dec-94 Jul-95 Dec-95

Fig. 4. Time series of lake-wide, volume-weighted averaged concentrations of ∑Hg (upper left panel), Hg2+ (lower left), Hg0 (upper right), and MeHg (lower right) generated from the
LM2-Mercury model versus the volume-weighted, lake-wide averaged concentrations of total mercury derived from the cruise measurements and Hg2+, Hg0, MeHg estimated using the
fractions of total mercury concentration. Note: ∑Hg = Hg0 + Hg2+ + MeHg; the “observed” data for Hg0, Hg2+, and MeHg were estimated as a fraction of tHg, assuming that 17% of
dissolved tHg in Lake Michigan is Hg0, 4% of tHg in Lake Michigan is MeHg, and rest of tHg is Hg2+. The error bars in ∑Hg plot represent one standard deviation from mean of the cruise
measurements.
354 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359
constants for mercury transformation, were adjusted simultaneously.
The objective in calibration was to achieve the best possible agreement
between model output and measurements as indicated by R2, the slope
of linear regression between model outputs and observed data, and
graphically visualizing the model calibration plots against field
observations.
The model-generated, volume-weighted, lake-wide average con-
centrations of Σ Hg (based on summing the independently modeled
mercury components), Hg2+, Hg0, and MeHg are plotted against the
volume-weighted, lake-wide average concentrations derived from the
LMMBP cruise results (Fig. 4). In general, there is good agreement be-
tween the model-generated and field-observed mercury concentra-
tions. It appears that there was some seasonal variation in model-
generated lake-wide concentrations for ∑ Hg, Hg2 + and Hg0 due to
strong seasonal variations observed in mercury atmospheric wet
deposition and RGM absorption which are the dominant sources to
the lake. The total mercury atmospheric wet deposition in the summer
was much higher than the winter wet deposition, which is likely due to
the higher concentrations of Hg in rain than in snow and the greater
precipitation amounts observed in the summer (Keeler and Dvonch,
2005; Vette et al., 2002). RGM absorption was highest for the summer
seasons and lowest for winter seasons caused by strong seasonal differ-
ences in the deposition velocities and ratios of RGM and TGM (total gas-
eous mercury) with the maximum in summer and the minimum in
winter (Shannon and Voldner, 1995; Vette et al., 2002). The seasonal
variation in Hg0 concentration in the water was related to variation of
the following: Hg2+ concentrations, reduction rates, and Hg0 saturation
percentage (higher in the summer and lower in the winter; Rolfhus and
Fitzgerald, 2001; Vette et al., 2002). The model-generated MeHg con-
centration shows slight variation throughout the LMMBP period due
to variation in wet deposition loads. No seasonal variation was observed
in the load from tributaries.
The segment-specific concentrations generated from the model and
observed data (segment mean concentrations derived from cruise sam-
ples = the data from a cruise sampling averaged by segment) for tHg
and ∑Hg are compared in Fig. 5 where tHg is an analytical measure-
ment of total mercury and ∑ Hg is an independent estimate of the
same collective mass of mercury components. The linear regression be-
tween the model prediction and observation gives reasonable statistical
results for tHg (R2 of 0.53 with a slope of 1.05) and ∑ Hg (R2 of 0.50
with a slope of 0.95).
Model corroboration
Parameters, especially water–sediment cycling rates (settling and
resuspension rates) and mercury partition coefficients determined
Model vs Observed Data for tHg
1.00
y = 1.05x + 0.01
R² = 0.53
Cruise field data ng/L
0.75
0.50
0.25
0.00
0.00 0.25 0.50 0.75
Model outputs ng/L
Fig. 5. Observed data (segment-specific cruise mean measurements) versus the model predictions of tHg (left panel) and ∑Hg (right panel) in respective segments (Note: ∑Hg = Hg0 +
Hg2+ + MeHg).
during the short-term model calibration, cannot be fully constrained.
To investigate long-term model performance after the short-term
model calibration and provide confidence in model predictability, a
long-term (146-year) tHg mercury hindcast was conducted for the
LM2-Mercury model.
As is the case with most contaminants, measured historic mercu-
ry loads for Lake Michigan do not exist. However, a record of historic
mercury loads can be obtained from the lake sediments (Rossmann,
2010). The reconstruction of a Lake Michigan historical tHg loading
function used for the mercury hindcasts was based on the average
annual tHg load measured in 1994–1995 and the information on
the time variation of mercury fluxes to sediment based on a high
quality core (core 41) collected from Lake Michigan during the
project period. Only core 41 was used in deriving the historical mer-
cury loads to the lake (the other cores were used in model corrobo-
ration). Among the 52 sediment cores collected and analyzed for
mercury, core 41 was selected to represent the time variation of mer-
cury loads to the lake due to a high sediment accumulation rate (one
of the highest rates) observed at the core location which leads to lit-
tle or no time delay for delivery of annual mercury loads to its loca-
tion. A detailed description of the protocol to reconstruct historical
tHg loadings to Lake Michigan can be found in Rossmann (2010).
The water transport fields (advection and dispersion) and meteoro-
logical condition such as wind speed, air and water temperature, etc.
for the two-year (1994–1995) project period were repeated
during the hindcast simulations. The meteorological conditions dur-
ing the 1994–1995 period are close to the means for these parame-
ters (Rossmann, 2006). Values of the parameters (Appendix A
Table A1) finalized during the short-term (two years — 1994 and
1995) model calibration were kept the same during the hindcast
simulation.
In order to corroborate the representativeness of the site-specific pa-
rameters used in the LM2-Mercury model for the important physical
and biogeochemical processes in Lake Michigan, the results of the mer-
cury hindcast must demonstrate that the model is able to generate the
temporal variation of total mercury concentrations that reasonably
match the observed historical data and profiles in both the water
column and sediments. The historical water and sediment data
are independent data sets; and, therefore, they help to evaluate model
corroboration in a more rigorous manner than comparing model
predictions to observed mercury concentrations in just one model
compartment.
For purposes of corroborating the model's predictions of tHg in lake
water (see Fig. 6, Chart a), the only data available that could be used to
make these estimates were lake trout data (whole fish). A mean bioac-
cumulation factor (BAF) was calculated for lake trout utilizing data from
Model vs Observed Data for ∑Hg
1.00
y = 0.95x + 0.02
R² = 0.50
Cruise field data ng/L
0.75
0.50
0.25
0.00
1.00 0.00 0.25 0.50 0.75 1.00
Model outputs ng/L
 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359 355

1.2 were collected from sites near Saugatuck, MI, Sheboygan, MI and Stur-
Volume-weighted, lake-wide mean (model)
Cruise mean(1994-1995)
geon Bay, WI during 1994–1995.
Total Mercury (ng/L)
Historical Estimates by using BAF Lake trout data (composites of 5 fish) collected prior to the 1994–
0.9
1995 LMMBP period along with the calculated BAF from 1994 to 1995
were used to estimate dissolved MeHg in the water. Once dissolved
in water

MeHg was estimated, it was used to estimate tHg in water assuming

0.6
the 1994–1995 ratio (0.32 ng tHg/L/0.00963 ng dissolved MeHg/L) was
constant (see ESM Appendix B for a discussion on this topic). Therefore,
0.3
tHg in the water column for any given year where fish data were available
was estimated by dividing the mean fish concentration by the average
1994–1995 BAF and then use the resultant dissolved MeHg concentration
Chart a estimate to estimate tHg from the 1994–1995 ratio. The fish data were
0.0
1850 1879 1908 1937 1966 1995 obtained from the US Geological Survey's National Contaminant Biomon-
Year itoring Program (USGS, 2013) and were collected during the period
1972–1987. The average age of these fish was estimated to be
6.1 years (range of 5.2–7.5 years). The fish were collected near
9A 15a 41a 13A 17A 19a 21A
22a 24A 25A 26A 27A 29A 31A
0.50 33A 34a 36A 37A 39A 46A 48A
53A 54A 55A 58a 61a 62A 63a some of the same sites as those associated with the 1994–1995 sam-
65A 66a 70a 78A 79a 80A 82A
pling (Saugatuck, MI and Sheboygan, WI). The third site was near
Total mercury (μg/g dw)

83A 85A 86a 87A 97A 99A 101A

0.40 103a 107A 108a 110a 112a 120a Model
Beaver Island in the northern part of the lake.
Measurements of mercury in sediment were more plentiful than his-
in sediments

0.30 torical mercury in water data, with fifty-two box core sediment samples
collected for the LMMBP. Among the box core sediment samples, forty-
0.20 eight sediment core profiles were analyzed for tHg.
Fig. 6 shows that the lake-wide average tHg concentrations generat-
0.10 ed from the model compare well to the observations in both the lake
water column (Chart a) and sediments (Chart b), respectively. As ex-
Chart b pected, the model-generated tHg concentration profile in sediment seg-
0.00
1850 1875 1900 1925 1950 1975 2000 ment #53 (Chart c) matches very well to the profiles from the two high
Year quality box cores (cores 41 and 15) collected within the same segment.
The high quality cores were defined as those having highest sediment
0.50
sed_core_15
accumulation rate, a thin mixed layer representing less than two years
of deposition, and little delay in the delivery of mercury from sources
in sediment segment 53
Total mercury (μg/g dw)

sed_core_41
0.40 to the two sites (Rossmann, 2010). In general, the hindcast results
Model
shown in Fig. 6 confirmed that the model is in reasonable agreement
0.30

0.20 Table 3
Mercury (tHg, Hg0, Hg2+, and MeHg) mass budget assessment for Lake Michigan. Mass
flow rates and inventories are annual averages of the 2-year LMMBP period (1994–1995).
0.10
Mass budget component tHg Hg0 Hg2+ MeHg
Chart c Mass flow rates (kg/year) in water
0.00
1850 1875 1900 1925 1950 1975 2000 column
Total external loads (see Table 2) 1556 8.6 1526 21
Year
Settling −1701 −1656 −24
Resuspension 438 433 2.8
Fig. 6. Comparison between the modeled and observed historic total mercury (tHg) Absorption from atmosphere 326 319
concentration in Lake Michigan water (Chart a), lake-wide sediment trends (Chart b), Gross volatilization from lake −722 −644
and two high quality cores from sediment segment #53 (Chart c). The locations of the Gain or loss (−) from Hg2+ net 325 −325
sediment cores collected during LMMBP can be found in Fig. 1 of Rossmann (2010). The reduction
bioaccumulation factor (BAF) is defined as the field measured MeHg in the whole lake Gain or loss (−) from Hg2+ net −5.4 5.4
trout divided by the field measure dissolved MeHg in the water and was used to estimate methylation
tHg in the water. Gain or loss (−) from MeHg 5.4 −5.4
reductive
demethylation
Diffusion 20 15 2.6
Input from Lake Huron 10 1.5 8 0.4
Export to Lake Huron −21 −1.4 −20 −0.8
Chicago diversion −2.14 −0.1 −1.0 −0.04
1994 to 1995, and this factor (assumed to be constant) was then applied Total net gain or loss (−) −96 14 −25 2.0
Mass flow rate (kg/year) in surficial
to historical lake trout data where only MeHg concentration in whole
sediments
fish were available. Although the predicted historical tHg mercury in Settling 1701 1656 24
the water column exhibits a relatively high variability, the central ten- Resuspension −438 −433 −2.8
dency of these predictions is captured by the model. Burial −587 −586 −0.6
Gain or loss (−) from Hg2+ net 19 −19
The BAF is defined as the field measured MeHg in the fish (Raymond
methylation
and Rossmann, 2009) divided by the field measured dissolved MeHg in Diffusion −20 −15 −2.6
the water. Based on the 1994–1995 MeHg measurements in whole lake Total net gain or loss (−) 656 0 641 −1
trout (wet weight) and dissolved MeHg concentrations in lake water Inventory (kg)
(0.00963 ng/L), a mean BAF of 11.8 × 106 was calculated. Lake trout Water column 1270 195 1120 53
Surficial sediment (0–4 cm) 18,210 0.05 18,098 77
composites (5–6 year old fish) were used to calculate the BAF and
356 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359
with historical data and provided further confidence and constraint in
the parameters determined during the short-term model calibration.
Mass budget assessment
With the inclusion of the total tHg external load and various
modeled processes, a mass budget assessment was conducted for total
mercury (tHg) and the individual mercury components in order to
quantify mercury cycling among the various compartments of Lake
Michigan (Table 3). Comparisons of tHg and ∑Hg can be made using
this table.
Using Tables 2 and 3, one can gain insight on what is controlling tHg
cycling. The gains of tHg to the water column, ordered from highest to
lowest, are RGM absorption, wet deposition, resuspension, Hg0 absorp-
tion, tributary load, dry (particle) deposition, diffusion from sediments,
and input from Lake Huron. The atmospheric contribution to the total
tHg load was 85%. The losses from the water column, ordered from
highest to lowest, are settling, gross volatilization, export to Lake
Huron, and loss through the Chicago diversion. In the surficial sedi-
ments, the only source is settling from the water column. Losses from
the sediment, ordered from the highest to the lowest, are burial, resus-
pension, and diffusion. The input of mercury from Lake Huron and ex-
port of mercury to Lake Huron are the results of bi-directional water
flow through the Straits of Mackinac.
Although not as comprehensive as the one conducted in this study,
mass budget assessment for tHg cycling in Lake Michigan has been
done by a few other researchers (Jeremiason et al., 2009; Mason and
Sullivan, 1997; Vette et al., 2002). Table 4 shows a comparison between
the results from this study and the other researchers. In general, our
assessment on tHg cycling is in good agreement with their estimates.
Using Tables 2 and 3, Hg0 gains to the water column, arranged from
the highest to the lowest, are Hg2+ net reduction, absorption from the
atmosphere, tributary loads, reductive demethylation, and input from
Lake Huron. The major loss of Hg0 from the water column was from
gross volatilization. Some minor amounts were also lost from export
to Lake Huron and through the Chicago diversion. In the surficial sedi-
ment, there were no mercury gains or losses of Hg0.
The gains of Hg2+ in the water column, arranged from highest to
lowest, are RGM absorption, wet deposition, resuspension, tributary
load, dry (particle) deposition, diffusion from the surficial sediment,
and input from Lake Huron (Tables 2 and 3). The largest losses of
Hg2+ in the water column, arranged from highest to lowest, are settling,
Hg2 + net reduction, export to Lake Huron, net methylation, and the
Chicago diversion. The loss from both net reduction and net methylation
accounted for 16% and 3% of the total loss for Hg2+ in the water column,
respectively. The major source of Hg2+ to the surficial sediments is from
settling. A minor gain of Hg2 + in the surficial sediments was from
negative net methylation meaning the mass of Hg2+ generated from
demethylation is greater than the mass of Hg2+ lost to methylation.
The sources of MeHg to the water column, ordered from the highest
to the lowest, are wet deposition, tributary loads, gain from in situ net
Table 4
Comparison of mass fluxes for reactive gaseous mercury (RGM) deposition, dissolved
gaseous mercury (DGM) emission, and tHg net settling and burial done by different
researchers for Lake Michigan.
RGM deposition (kg/year) DGM emission (kg/year) Reference
320–959 286–797 Vette et al. (2002)
520 Mason and Sullivan (1997)
380–850 Jeremiason et al. (2009)
651 644 This study
Net settling (kg/year) Burial (kg/year) Reference
742 632 Mason and Sullivan (1997)
1263 587 This study
methylation, resuspension, diffusion, and gain from Lake Huron. The
losses of MeHg from the water column, ordered from the highest to
the lowest, are associated with settling, loss from in situ reductive
demethylation, export to Lake Huron, and loss through the Chicago
Diversion. In the surficial sediments, the principal source of MeHg is
from settling. The major loss of MeHg from the sediment, ordered
from highest to lowest, are negative net methylation, resuspension, dif-
fusion, and burial. Ulrich et al. (2001) supports our finding that in well
oxygenated sediments such as those of Lake Michigan, negative net
methylation (demethylation predominates) can be a significant MeHg
loss mechanism. Cold temperatures in the lake sediments also facilitate
demethylation (Ulrich et al., 2001). The negative net methylation alone
accounted for a 76% loss of MeHg from the sediment.
The gain of MeHg from in situ Hg2 + net methylation contributed
17% of the MeHg sources to the water column. Mercury methylation
has been well documented to occur within hypoxic/anoxic zones
facilitated by methylating bacteria. Hypoxic zones in the Lake
Michigan system include Green Bay and nearshore areas of Lake
Michigan where the green, filamentous Cladophora alga are abun-
dant. There is also speculation that micro-hypoxic zones exist within
the “deep chlorophyll layer” — a zone just below the thermocline in
the lake where enhanced concentrations of MeHg have been found in
the Great Lakes (Ogorek et al., 2011). The rest of the lake is well
oxygenated and methylation would not be expected in those zones.
In the sediment, settling and demethylation were the most dominant
source and sink for MeHg, respectively.
Some other comments on mercury transport and fate during the pe-
riod 1994–1995 can be made when reviewing Table 3. The results show
that the Lake Michigan system had a net loss from the water column of
96 kg/year and was likely not at steady state for tHg during the project
period, which is in agreement with the downward trend in water
column predicted by the model hindcast (Fig. 6, Chart a) and the con-
tinuing downward trend observed from the recent post-audit data
(the mean concentration of 0.332 ng/L in the main lake for 2005
spring–summer by Jeremiason et al., 2009 and the mean concentration
of 0.253 ng/L for 2010 by Krabbenhoft and Dove, 2012). Most of the
mercury coming into the lake water is in the form of Hg2+ from external
sources, and most of it leaving the water column is also in the form of
Hg2+. The water column inventory of tHg and MeHg was 1270 kg and
53 kg, respectively, and the inventory of tHg and MeHg in the sediment
was 18,210 kg and 77 kg, respectively. Hg0 mass inventory was very low
in the sediment.
Conclusions
Results from the LM2-Mercury calibration and corroboration dem-
onstrate the representativeness of the model construct and parameter-
ization for describing mercury cycling in Lake Michigan. Atmospheric
sources, including loads from RGM absorption and wet and dry
(particle) deposition, were the primary external loads of mercury to
Lake Michigan. A mass budget assessment conducted based on the
model results illustrated that, in addition to the loading from atmo-
sphere, interaction between water and the surficial sediment, and air–
water exchange are the other key processes controlling tHg cycling
and concentrations in the lake. In situ production of Hg0 from Hg2 +
net reduction was the dominant internal source of Hg0 for Lake
Michigan. The largest Hg2+ mass loss was from the settling to the sed-
iment. The gain of MeHg from in situ Hg2+ net methylation contributed
approximately 17% of the MeHg to the water column. Most of the MeHg
gain in the water column is coming from MeHg loads associated with
wet deposition and the tributary inputs. Net methylation in the sedi-
ment is −76%. Settling was another key process influencing the concen-
trations of MeHg in both the water and surficial sediment.
Model results indicate that mercury in Lake Michigan water during
the period 1994–1995 was likely not at steady state (mechanisms that
tend to reduce the mercury water concentrations dominated over
 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359 357

mechanisms that tend to increase mercury concentrations), and this re- WHg2+ Absorption
sult is consistent with a continuing downward trend confirmed by more
recent measurements of mercury in Lake Michigan water.
Advection Advection
Bound Hg2+ KTSRS,Hg 2+ Out
In Dissolved kRw-Hg 2+
Acknowledgments Hg2+
TSRS DOC K DOC,Hg 2+ kMw-Hg 2+ Exchange
Water
Funding sources for this project came from the U.S. Environmental Column
Settling Resuspension Diffusion Diffusion
Protection Agency (Great Lakes National Program Office, Office of Re-
search and Development, and Contract No. EP-D-11-092). It has been Surface
subjected to review by the National Health and Environmental Effects Mixed Bound Hg 2+ KTSRS,Hg 2+ Dissolved
Research Laboratory and approved for publication. Approval does not Sediment
Hg 2+
kMs-Hg 2+
TSRS DOC KDOC,Hg2+
signify that the contents reflect the views of the Agency, nor does men- Layer
tion of trade names or commercial products constitute endorsement or
Burial
recommendation for use. Special thanks goes out to the agency person-
nel at the Large Lakes Research Station, Grosse Ile, Michigan for general
support and technical direction; anonymous technical reviewers (J. Fig. A3. Conceptual framework for Hg2+, where kRw–Hg2þ ¼ Hg2þ net reduction to Hg0 in
water; kMw–Hg2þ ¼ Hg2þ net methylation to MeHg in water; net methylation to MeHg in
Great Lakes Res.); and Dr. James Pauer and Dr. Mark Rowe for their help- sediments; K TSRS;Hg2þ = partitioning of Hg2+ between dissolved phase and particulate
ful comments. (TSRS) phase; K DOC;Hg2þ = partitioning of Hg2+ between dissolved phase and particulate
(DOC) phase; W = sum of all loads.

Appendix A. Conceptual framework for model state variables and

a table of parameters specified for processes conceptualized in
LM2-Mercury
WMeHg

Advection Advection
In Bound MeHg KTSRS,MeHg Dissolved kMw-Hg 2+ Out
MeHg kDw-MeHg
TSRS DOC KDOC,MeHg Exchange
Water
Column
Settling Resuspension Diffusion Diffusion
WTSRS WDOC
Surface
Mixed Bound MeHg KTSRS,MeHg
Dissolved
Sediment kMs-Hg 2+
Advection In Advection Out Layer TSRS DOC KDOC,MeHg MeHg
TSRS TSRS
DOC DOC
Decay Exchange
to
Water Burial
Column Settling Resuspension Diffusion
Fig. A4. Conceptual framework for MeHg, where kMw–Hg2þ ¼ Hg2þ net methylation to
Surface MeHg in water; kDw–MeHg = MeHg reductive demethylation to Hg0 in water; kMs–Hg2þ ¼
Mixed DOC Hg2þ net methylation to MeHg in sediment; KTSRS,MeHg = partitioning of MeHg between
Sediment
TSRS TSRS
DOC Decay dissolved phase and particulate (TSRS) phase; KDOC,MeHG = partitioning of MeHg between
Layer to dissolved phase and particulate (DOC) phase; W = sum of all loads.

Burial

Fig. A1. Conceptual framework for TSRS and DOC cycling, where TSRS = total suspended
and resuspendable solids; DOC = dissolved organic carbon; W = sum of all loads.

Volatilization
WtHg + Absorption

Volatilization
WHg0 + Absorption
Advection Advection
In Bound tHg KTSRS,tHg Dissolved Out

Advection Advection TSRS DOC KDOC,tHg tHg

Exchange
In 2+ Dissolved Out Water
k Rw-Hg
Hg0
k Dw-MeHg Exchange Column
Settling Resuspension Diffusion Diffusion
Water
Column Surface
Diffusion
Mixed Bound tHg KTSRS,tHg Dissolved
Sediment
Surface
Layer TSRS DOC K DOC,tHg tHg
Mixed Dissolved
Sediment
Layer
Hg0 Burial

Fig. A5. Conceptual framework for tHg, where tHg = total Hg; KTSRS,tHg = partitioning of
Fig. A2. Conceptual framework for Hg0, where kRw–Hg2þ ¼ Hg2þ net reduction to Hg0 in total Hg between dissolved phase and particulate (TSRS) phase; KDOC,tHG = partitioning
water; kDw–MeHg = MeHg reductive demethylation to Hg0 in water; W = sum of all loads. total Hg between dissolved phase and particulate (DOC) phase; W = sum of all loads.
358 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359

Table A1
Final parameters specified for processes conceptualized in LM2-Mercury.

Parameter Definition and units Value Data Source

kdðDOCÞ Saturated decay rate constant of DOC in water at a temperature of 20 °C, d−1, d = day 0.006 Zhang et al. (2008); then calibrated
k1/2(DOC) Half-saturation constant for DOC decay in water, mg/L 3.1 Zhang et al. (2008); then calibrated
kdsðDOCÞ Decay rate constant for DOC in sediments at a temperature of 20 °C, d−1 0.015 Zhang et al. (2008)
kdðTSRSÞ Saturated decay rate constant of TSRS in water at a temperature of 20 °C, d−1 0.023 This study1; then calibrated
k1/2(TSRS) Half-saturation constant for TSRS decay in water, mg/L 1.7 Estimate from this study
kdsðTSRSÞ Decay rate constant for TSRS in sediments at a temperature of 20 °C, d−1 10−6 This study2; then calibrated
Y(TSRS–DOC) Yield coefficient (%) of DOC during TSRS decay in water, dimensionless 40 Zhang et al. (2008); then calibrated
Ys(TSRS–DOC) Yield coefficient (%) of DOC during TSRS decay in sediment, dimensionless 50 Zhang et al. (2008); then calibrated
vs(TSRS) Lake wide average TSRS settling velocity, m/d 0.71 Calibration from this study
vb Lake wide average sediment accumulation rate, cm/year 0.17 Zhang et al. (2008)
Ds Diffusion coefficient for dissolved Hg and DOC diffusion at water and sediment interface, m2/d 1.73 × 10−4 Zhang et al. (2008)
KTSRS,tHg In situ partition coefficient for tHg on TSS, L/kg TSRS 105.95 This study3; then calibrated
K TSRS;Hg2þ Partition coefficient for Hg2+ on TSRS, L/kg TSRS 106.05 This study4; then calibrated
KTSRS,MeHg′ Partition coefficient for MeHg on TSRS, L/kg TSRS 105.30 This study4; then calibrated
KDOC,tHg Partition coefficient for tHg on DOC, L/kg DOC 104.95 This study5; Calibration
K DOC;Hg2þ Partition coefficient for Hg2+ on DOC, L/kg DOC 105.05 This study5; then calibrated
KDOC,MeHg Partition coefficient for MeHg on DOC, L/kg DOC 105.50 This study5; then calibrated
kRw−Hg2þ Net saturated reduction rate constant for dissolved Hg2+ in water at 20 °C, d−1 0.012 Vette et al. (2002); then calibrated
d
kð1=2Þ−Hg2þ Half-saturation constant for dissolved Hg2+ in water, ng/L 0.3 Estimate from this study
d
kMw−Hg2þ Net saturated methylation rate constant for dissolved Hg2+ in water at 20 °C, d−1 0.0002 Tsiros and Ambrose (1998); then calibrated
d
kDw−MeHgd Saturated reductive demethylation rate constant for dissolved MeHg in water at 20 °C, d−1 0.002 Tsiros and Ambrose (1998); then calibrated
kð1=2Þ−MeHgd Half-saturation constant for dissolved MeHg in water, ng/L 0.018 Estimate from this study
kMs−Hg2þ p
Rate constant of net methylation for particulate Hg2+ in sediment at 20 °C, d−1 −6.0 × 10−6 Tsiros and Ambrose (1998); then calibrated
kMs−Hg2þ Rate constant of net methylation for dissolved Hg2+ in sediment at 20 °C, d−1 −6.0 × 10−6 Tsiros and Ambrose (1998); then calibrated
d
1,2
Estimated initially based on saturated decay rate constant of particulate organic carbon (POC) (Zhang et al., 2008) and the ratio of POC and TSRS measured in the water and sediment of
the lake. 3Estimated initially based on field measurements from LMMBP. 4Assumed that partition coefficient was initially assigned as the one used for tHg. 5Assumed that partition coef-
ficient for mercury on DOC is an order of magnitude less than that on TSRS (Tsiros and Ambrose, 1998; Zhang et al., 2008).

Appendix B. Supplementary data
Supplementary data to this article can be found online at http://dx.
doi.org/10.1016/j.jglr.2014.03.008.
References
Allison, J.D., Brown, D.S., Novo-Gradac, K.J., 1991. MINTEQA2/PRODEFA2, A geochemical
assessment model for environmental systems. Version 3.0 User's Manual (EPA/600/
3-91/021).
Amyot, M., Lean, D.R.S., Poissant, L., Doyon, M.R., 2000. Distribution and transformation of
elemental mercury in the St. Lawrence River and Lake Ontario. Can. J. Fish. Aquat. Sci.
57 (Suppl. 1), 155–163.
Glass, G.E., Sorensen, J.A., 1999. Six-year rend (1990–1995) of wet mercury deposition in
the Upper Midwest, U.S.A. Environ. Sci. Technol. 33, 3303–3312.
Hall, D., Robertson, D., 1998. Estimation of contaminant loading from monitored and un-
monitored tributaries to Lake Michigan for the USEPA Lake Michigan Mass Balance
Study. Quality Systems and Implementation PlanU.S. Environmental Protection
Agency, Great Lakes National Program Office, Chicago, IL.
Hall, B.D., Manolopoulos, H., Hurley, J.P., Schauer, J.J., St. Louis, V.L., Kenski, D., Graydon, J.,
Babiarz, C.L., Cleckner, L.B., Keeler, G.J., 2005. Methyl and total mercury in precipita-
tion in the Great Lakes region. Atmos. Environ. 39, 7557–7569.
Hammerschmidt, C.R., Lamborg, C.H., Fitzgerald, W.F., 2007. Aqueous phase methylation
as a potential source of methylmercury in wet deposition. Atmos. Environ. 41,
1663–1668.
Harrsch, E.G., Rea, D.K., 1982. Composition and distribution of suspended sediments in
Lake Michigan during summer stratification. Environ. Geol. 4, 87–98.
Hudson, R.J.M., Gherini, S.A., Watras, C.J., Porcella, D.S., 1994. Modeling the biogeochemi-
cal cycle of mercury in lakes: The mercury cycling model (MCM) and its application
to the MTL study lakes. In: Watras, C.J., Huckabee, J.W. (Eds.), Mercury Pollution —
Integration and Synthesis. Lewis Publishers, Florida, pp. 473–523.
Hurley, J.P., Shafer, M.M., Cowell, S.E., Overdier, J.T., Hughes, P.E., Armstrong, D.E., 1996.
Trace metal assessment of Lake Michigan tributaries using low-level techniques. En-
viron. Sci. Technol. 30, 2093–2098.
Hurley, J.P., Cowell, S.E., Shafer, M.M., Hughes, P.E., 1998. Tributary loading of mercury to
Lake Michigan: importance of seasonal events and phase partitioning. Sci. Total Envi-
ron. 213, 129–137.
Jeremiason, J.D., Kanne, L.A., Lacoe, T.A., Hulting, M., Simcik, M.F., 2009. Comparison of
mercury cycling in Lakes Michigan and Superior. J. Great Lakes Res. 35, 329–336.
Keeler, G.J., Dvonch, T.J., 2005. Atmospheric mercury: a decade of observations in the
Great Lakes. In: Pirrone, N., Mahaffey, K.R. (Eds.), Dynamics of Mercury Pollution on
Regional and Global Scales. Springer, New York, pp. 611–636.
Knightes, C.D., 2008. Development and test application of a screening-level mercury fate
model and tool for evaluating wildlife exposure risk for surface waters with mercury-
contaminated sediments (SERAFM). Environ. Model Softw. 23, 495–510.
Knightes, C.D., Sunderland, E.M., Barber, M.C., Johnston, J.M., Ambrose Jr., R.B., 2009. Appli-
cation of ecosystem-scale fate and bioaccumulation models to predict fish mercury
response times to changes in atmospheric deposition. Environ. Toxicol. Chem. 28
(4), 881–893.
Krabbenhoft, D., Dove, A., 2012. Mercury and methylmercury in surface waters across the
Great Lakes. Presented at Mercury Science in the Great Lakes Workshop, May 31,
2012. U.S.EPA/GLNPO, Chicago, IL.
Landis, M.S., Keeler, G.J., 2002. Atmospheric mercury deposition to Lake Michigan during
the Lake Michigan mass balance study. Environ. Sci. Technol. 436, 4518–4524.
Lyon, B.F., Ambrose, R.B., Rice, G., Maxwell, C.J., 1997. Calculation of soil–water and
benthic sediment partition coefficients for mercury. Chemosphere 35, 791–808.
Marsik, F.J., Keeler, G.J., Landis, M.S., 2007. The dry-deposition of speciated mercury to the
Florida Everglades: measurements and modeling. Atmos. Environ. 41, 136–149.
Martin, J.L., 1992. MERC4: A mercury transport and kinetics model (beta 1.0). Sponsored
by American Society of Civil Engineers. U.S. Environmental Protection Agency, Center
for Exposure Assessment Modeling, Athens, GA.
Mason, R.P., Sullivan, K.A., 1997. Mercury in Lake Michigan. Environ. Sci. Technol. 31,
942–947.
Mason, R.P., Fitzgerald, W.F., Morel, F.M.M., 1994. The biogeochemical cycling of elemen-
tal mercury: anthropogenic influences. Geochim. Cosmochim. Acta 58, 3191–3198.
Mason, R.P., Morel, F.M.M., Hemond, H.G., 1995. The role of microorganisms in elemental
mercury formation in natural waters. Water Air Soil Pollut. 80, 775–787.
Miles, C.J., Moye, H.A., Phlips, E.J., Sargent, B., 2001. Partitioning of monomethylmercury
between freshwater algae and water. Environ. Sci. Technol. 35, 4277–4282.
Montieth, T.J., Sonzongni, W.C., 1976. U.S. Great Lakes shoreline erosion loadings. Interna-
tional Joint Commission, Windsor, Ontario.
O'Connor, D.J., 1983. Wind effects on gas–liquid transfer coefficients. J. Environ. Eng. 109,
731–752.
Ogorek, J., Krabbenhoft, D., De Wild, J., Sabin, T., Tate, M., Thompson, C., 2011. Mercury
and methyl mercury in the water column, sediments, and biota across the Great
Lakes. 10th International Conference on Mercury as a Global Pollutant, Halifax,
Canada.
Pauer, J.J., Taunt, K.W., Melendez, W., 2006. LM3-Eutro. Part 2. In: Rossmann, R. (Ed.), Re-
sults of the Lake Michigan Mass Balance Study: PCBs Modeling Report. U.S. Environ-
mental Protection Agency, Office of Research and Development, National Health and
Environmental Effects Research Branch, Large Lakes Research Station, Grosse Ile, MI
(EPA/600/R-04/167).
Poissant, L., Amyot, M., Pilote, M., Lean, D., 2000. Mercury water–air exchange over the
Upper St. Lawrence River and Lake Ontario. Environ. Sci. Technol. 34, 3069–3078.
Raymond, B., Rossmann, R., 2009. Total and methyl mercury accumulation in 1994–1995
Lake Michigan lake trout and forage fish. J. Great Lakes Res. 35, 438–446.
Rolfhus, K.R., Fitzgerald, W.F., 2001. The evasion and spatial/temporal distribution of mer-
cury species in Long Island Sound, CT-NT. Geochim. Cosmochim. Acta 65, 407–418.
Rossmann, R., 2002. Lake Michigan 1994–1996 surficial sediment mercury. J. Great Lakes
Res. 28, 65–76.
Rossmann, R., 2006. Introduction. Part 1. In: Rossmann, R. (Ed.), Results of the Lake Michigan
Mass Balance Study: PCBs modeling report. U.S. Environmental Protection Agency,
Office of Research and Development, National Health and Environmental Effects Re-
search Laboratory, Mid-Continent Ecology Division, Large Lakes and Rivers Forecasting
Research Branch, Large Lakes Research Station, Grosse Ile, MI, pp. 1–119 (EPA/600/R-
04/167, http://www.epa.gov/med/grosseile_site/LMMBP/pcb-report.html).
 X. Zhang et al. / Journal of Great Lakes Research 40 (2014) 347–359
Rossmann, R., 2010. Protocol to reconstruct historical contaminant loading to large lakes:
the Lake Michigan sediment record of mercury. Environ. Sci. Technol. 44, 935–940.
Rossmann, R., Rygwelski, K.R., Filkins, J.C., 2001. Methyl mercury in Lake Michigan surfi-
cial sediments. 44th Conference on Great Lakes Research. International Association
for Great Lakes Research, Green Bay, WI.
Rygwelski, K.R., Anstead, A.M., Zhang, X., 2011. Methods to convert LM3-Eutro BIC to TSS
for use in LM2-Hg model. Technical Report Number 886. U.S. Environmental Protec-
tion Agency, Office of Research and Development, National Health and Environmen-
tal Effects Research Branch, Large Lakes Research Station, Grosse Ile, MI.
Sanemasa, I., 1975. The solubility of elemental mercury vapor in water. Bull. Chem. Soc.
Jpn. 48, 1795–1798.
Schawb, D.J., 2000. Lake Michigan shoreline and tributary mass loadings. Email from
David J Schawb, National Oceanic and Atmospheric Administration, Great Lakes
Environmental Research Laboratory, Ann Arbor, MIU.S. Environmental Protection
Agency, Office of Research and Development, National Health and Environmental
Effects Research Branch, Large Lakes Research Station, Grose Ile, MI (EPA/REP-898,
5 pp.).
Schwab, D.J., Beletsky, D., 1998. Lake Michigan mass balance study: hydrodynamic model-
ing project. National Oceanic and Atmospheric Administration, Great Lakes Environ-
mental Research Laboratory, Ann Arbor, MI (ERL-GLERL-108).
Schwarzenbach, R.P., Gschwend, P.M., Imboden, D.M., 2003. Environmental Organic
Chemistry, second ed. Wiley Interscience, New York.
Shannon, J.D., Voldner, E.C., 1995. Modeling atmospheric concentrations of mercury and
deposition to the Great Lakes. Atmos. Environ. 29, 1649–1661.
Stein, E.D., Cohen, Y., Winer, A.M., 1996. Environmental distribution and transformation
of mercury compounds. Crit. Rev. Environ. Sci. Technol. 26, 1–43.
359
Tsiros, I.X., Ambrose, R.B., 1998. Environmental screening modeling of mercury in the
Upper Everglades of South Florida. J. Environ. Sci. Health A33, 497–525.
U.S. Environmental Protection Agency, 2004. Results of the Lake Michigan mass balance
study: Mercury data report. U.S. Environmental Protection Agency. Great Lakes Na-
tional Program Office, Chicago, IL (EPA/905/R-01/012).
U.S. Geological Survey, 2013. National Contaminant Biomonitoring Program. http://www.
cerc.usgs.gov/data/ncbp/fish.htm (Accessed on 3/20/2013).
Ulrich, S.M., Tanton, T.W., Abrashitova, S.A., 2001. Mercury in the aquatic environment: a
review of factors affecting methylation. Crit. Rev. Environ. Sci. Technol. 31 (3),
241–293.
Vette, A.F., 1998. Photochemical influences on the air–water exchange of mercury.
(Doctoral dissertation) The University of Michigan, Ann Arbor, MI.
Vette, A.F., Landis, M.S., Keeler, G.J., 2002. Deposition and emission of gaseous mercury to
and from Lake Michigan during the Lake Michigan Mass Balance Study (July 1994–
October 1995). Environ. Sci. Technol. 36, 4525–4532.
Watras, C.J., Back, R.C., Halvorsen, S., Hudson, R.J.M., Morrison, K.A., Wente, S.P., 1998. Bio-
accumulation of mercury in pelagic freshwater food webs. Sci. Total Environ. 219,
183–208.
Whitman, W.G., 1923. A preliminary experimental confirmation of the two-film theory of
gas absorption. Chem. Metall. Eng. 29, 146–148.
Zhang, X., Rygwelski, K.R., Rossmann, R., Pauer, J.J., Kreis Jr., R.G., 2008. Model construct
and calibration of an integrated water quality model (LM2-Toxic) for the Lake Mich-
igan mass balance project. Ecol. Model. 219, 92–106.