United States
                    Environmental Protection
                    Agency
Environmental Research
Laboratory
Athens GA 30613
                    Research and Development
EPA/600/S3-86/045 May 1987
SEPA          Project  Summary
                    Case  Studies  and Model  Testing
                    of the  Metals  Exposure Analysis
                    Modeling  System (MEXAMS)
                    A. J. Medine and B. R. Bicknell
                      In .the EPA's wasteload allocation/
                    total maximum daily load (WLA/TMDL)
                    program, the agency must establish
                    more stringent effluent limitations and
                    guidelines for toxic chemicals (including
                    metals) if previous limitations are not
                    adequate to attain or maintain accept-
                    able water quality levels. The Metals
                    Exposure Analysis Modeling System
                    (MEXAMS) was recently developed to
                    assist in this effort. This model, linking
                    a complex  speciation model with an
                    aquatic transport/fate model, should
                    help discriminate between the fraction
                    of metal that is dissolved and in bio-
                    available form, and the fraction that is
                    complexed  and  rendered relatively
                    nontoxic.
                      The MEXAMS model has been tested
                    with data from three rivers to determine
                    its ability to simulate fate, transport,
                    and speciation of heavy metals in river
                    systems. The rivers selected for study
                    were the Naugatuck River in Con-
                    necticut and the Ten Mile River  in
                    Massachusetts, both of which receive
                    electroplating and  metal finishing
                    wastes; and the White River in Utah,
                    which receives  metal loadings from
                    natural sources. This report documents
                    the tests as case studies to guide future
                    users in application of the model.  In
                    addition, the program has been en-
                    hanced by addition of sediment settling/
                    resuspension to the aquatic transport
                    submodel EXAMS.
                      This Protect Summary was developed
                    by EPA's Environmental Research Lab-
                    oratory, Athens,  GA,  to announce key
                    findings of the research project that Is
                    fully documented In a separate report of
                    the same title (see Project Report order-
                    Ing Information at back).
 Introduction
  Transport  processes, environmental
 distributions, and  biological effects of
 heavy metals in our aquatic environments
 have been the focus of increasing concern.
 Direct toxicity to aquatic organisms and
 indirect toxicity  to humans and other
 higher organisms are at the center of this
 concern. Although heavy  metals are
 natural constituents of aquatic environ-
 ments, additional quantities introduced
 by man's presence may result in an al-
 tered chemical  composition of  the
 aqueous and sediment phases, often with
 detrimental impacts on the environment.
 In an effort to  reduce these impacts,
 environmental transport/fate models of
 heavy metals in aquatic ecosystems are
 being developed.
  One of these models  is the Metals
 Exposure Analysis Modeling System
 (MEXAMS). This program links MINTEQ,
 a geochemical model, with  EXAMS, an
 aquatic exposure  assessment model.
 MEXAMS was developed to allow assess-
 ment of the impacts of "priority pollutant"
 metals (As, Cd, Cu, Pb, Ni, Ag and Zn).
 This report presents a case study and
 tutorial for MEXAMS.
  The first evaluation consisted of cali-
 brating, testing, and assessing the model
 response for an actual site application on
 the Naugatuck  River, located in  the
 Housatonic  River Basin in western
 Connecticut.
  The second evaluation addressed the
 use of the MINTEQ geochemical model
for the White River (Utah), a dynamic
 lotic system. Water quality  parameter
 input, laboratory adsorption experiments,
chemical  analysis, MINTEQ  utilization,
and prediction of laboratory results are
described.

-------
  The third evaluation addressed applica-
tion of MEXAMS to the Ten Mile River, a
small  stream located  in  southeastern
Massachusetts, that receives inorganic
metal waste from numerous industries
along its banks. The evaluation involved
data  analysis,  preparation  of detailed
MEXAMS input, and description of the
model simulation for a wasteload alloca-
tion study of the river.
  A second purpose of this work was to
incorporate simple sediment settling and
resuspension into MEXAMS. Considera-
tion of metal losses from the water column
due to steady-state sediment settling will
add an important process to MEXAMS,
allowing more accurate calibration and
simulation of many systems.


Model Descriptions
  The Exposure Analysis Model System
(EXAMS)  describes the fate, transport
and impacts of organic contaminants in
aquatic systems. The program is a deter-
ministic simulation model,  based on a
core of mechanistic  process equations
derived from fundamental  theoretical
concepts.
  The model estimates exposure, fate,
and persistence of organic pollutant dis-
charges, using conservation  of mass to
balance loadings, transport,  and trans-
formation  of the  compound.  EXAMS
simulates  three  transport  processes:
advection, dispersion, and volatilization.
Advective and dispersive transport are
possible for dissolved species, sediment
sorbed material and bio-sorbed materials.
Transport  through  the  ecosystem com-
partments may be represented by whole
sediment bed loads, suspended sediment
washloads, exchanges with fixed volume
sediment  beds,  and   ground  water
infiltration.
  EXAMS is capable of simulating both
lotic and lentic  aquatic systems.  These
systems are initially compartmentalized
and classified  as littoral, epilimnion,
hypolimnion,  or benthic. Each compart-
ment is assumed to be completely mixed.
The computer then develops single dif-
ferential equations for each compartment.
  MINTEQ, a computer program for cal-
culating aqueous geochemical equilibria,
was developed for incorporation  into
MEXAMS.  MINTEQ is  used to  predict
metal speciation including sorption, pre-
cipitation and ion exchange  of "priority
pollutant"  metals  in  aquatic  environ-
ments. This chemical equilibrium problem
is described as a set of mass balance
equations, one for each component, and
a set  of mass action equations, one for
each species. The equilibrium constant
approach is utilized to solve the equilibrium
problem, solving nonlinear mass action
expressions  using linear mass balance
equations. The  equilibrium composition
of an aquatic environment is determined
by minimization of Gibbs free energy of
the system within the mass balance con-
straints. This chemical equilibrium will
determine aqueous  metal speciation in
addition to the  effects of precipitation/
dissolution and sorption, and is used in
EXAMS to determine fate and migration
of a metal.
  Solid phases are dealt with using  the
"transformation of basis" method. This
method reduces the number of indepen-
dent variables  to be determined  and
allows the solution  of a wider range of
chemical equilibrium problems.
  MINTEQ also is capable of modeling
adsorption  in  a number of ways.  Six
different  algorithms  are  accessible
through the use bf MINTEQ for describing
sorption  phenomena - "activity"  Kd,
"activity" Langmuir equation, "activity"
Freundlich equation, ion exchange, con-
stant capacitance surface complexation,
and triple layer surface complexation.
  EXAMS  and MINTEQ are the two in-
dividual models linked to form MEXAMS.
The MEXAMS program provides three
different modes  of  operation. The  first
mode is the  MINTEQ only, which allows
the operator to determine how changes
in water  chemistry will affect  metal
speciation and solid phase  interactions
without regard  to transport processes.
The second mode is  EXAMS only, which
deals with the ionization, sorption, trans-
port, and transformation of a given pol-
lutant.  Finally, the third  mode  links
EXAMS and MINTEQ, allowing the user
to determine  the effect of  transport
processes  and chemical interactions on
priority pollutant  concentration.  To use
the coupled mode,  the  first step is to
create an EXAMS input file that describes
the characteristics of the aquatic environ-
ment being assessed. The user then  en-
ters the MEXAMS Interactive Software
Package (MISP).  MISP  will call for a
MINTEQ file for each compartment that
contains different water quality data. The
user will also input run-specific informa-
tion that controls the number of times
MINTEQ updates  metal  concentrations.
MEXAMS  is now set to simulate metal
behavior, migration,  and fate.

Naugatuck River Case Study
  The Naugatuck River has a long history
of industrialization along its  length. Cur-
rently  approximately 30 to  300 cubic
meters/day  of  treated  electroplating
waste is being  discharged  into  the
Naugatuck River.  Most  industrial dis-
charges on the river currently adhere to
Best Available Technology (BAT) guide-
lines. Both long-term biological monitor-
ing and recent toxicity testing, however,
have shown toxic impacts of metals. The
data set available was used to illustrate
how to apply MEXAMS to field conditions
such as in the Naugatuck River.
  The  main  issue surrounding model
calibration and verification is obtaining
estimates for model parameters and
comparing predicted concentrations with
actual observed data. Ideally, several data
sets are available for independent calibra-
tion and verification. In the case of  the
Naugatuck River data, limited data per-
mitted only the calibration of the model.
  Subjective variation of  parameter
values and a qualitative comparison of
model  solution and observation is  the
most common approach to model calibra-
tion. Through this procedure, the calibra-
tion process attempts to account for (1)
spatial variations not represented by the
model formulation; (2) functional depen-
dencies of parameters that are either
non-quantifiable,  unknown and/or  not
included in the model algorithms; or (3)
extrapolation of laboratory measurements
of parameters to natural field conditions.
During  calibration,  the analyst may
choose to adjust some of the parameters
to improve model predictions or to alter
the structure of the relationships  (or
system physical representation) between
the variables in the model.
  The first step in the calibration was to
determine which model inputs have the
least reliability. These are the parameters
that will be  adjusted to perform a sen-
sitivity analysis. Any sensitivity analysis
is system dependent  because certain
input parameters may  be more or less
sensitive depending on the system being
modeled. The first decision to be made in
the utilization of MEXAMS was to select
a system configuration. There are many
possible configurations due to the range
of advective  and dispersive pathways.
Two initial  configurations,  containing
three water column  and three benthic
compartments,  were tested. Originally,
both of these configurations had the en-
tire groundflow routed  through  the
benthic compartments.  This caused a
considerable problem because the sedi-
ment residence time within the benthic
compartments was so low that the metal
was being flushed out of the system. This
problem was corrected by routing a por-
tion of the groundwater  into the  water

-------
column compartments, increasing the
sediment  residence  time  and thus  in-
creasing the concentration of adsorbed
pollutant within the bed sediments.
  The parameters that were manipulated
during the calibration  process included
the advective pathway for groundwater
accrual, the benthic Kd's, and the benthic
dispersion coefficient. Varying the percent
groundwater flow into the benthic com-
partments  between  0.01% and  1%  al-
lowed  adjustment of  the benthic par-
ticulate metal concentration. Next, varying
the benthic-water column dispersion co-
efficient from 5.00*E-05  to  5.00*E-08
(mVhr) gave  another relationship be-
tween dispersion coefficient and adsorbed
benthic metal concentrations. Finally, the
benthic partition coefficient was increased
from field-determined values, which were
considered inaccurate. It is very difficult
to obtain  a  benthic sediment sample
without entrainment of water from the
water  column. This  water dilutes sedi-
ment interstitial concentrations, resulting
in low Kd values.
  To obtain greater  spatial accuracy in
water column concentrations, a ten-com-
partment  configuration was  developed
(Figure 1). The  final configuration was
calibrated by adjusting the percent ground
water  flow into the benthic compart-
ments, the advective flow between com-
p-rtments, and  the  Kd  value for the
benthic compartments.
  The  calibration results obtained from
the final configuration are presented in
Table 1. As stated previously, three pa-
rameters  were  adjusted  to  achieve a
calibrated  model.  Table 2 displays the
original log Kd's and the  final  log Kd's
used to obtain  the  ten  compartment
calibration. The  advective  flow paths for
the  new  compartmentalizations  were
presented  in Figure  1, and the percent
ground water routed into benthic com-
partments  eight and ten  were changed
from 0.01% to 0.1%  and 0.13%, respec-
tively.  The final  calibration obtained  for
the ten-compartment model was reason-
able, and the concentrations predicted by
the model were well within the error bars
of the field data.
  The most important aspect of this model
calibration was the compartmentalization
of the river reach. Initially calibrated for a
six-compartment configuration, the model
did not predict adequately the total and
dissolved  concentrations  in the  down-
stream water column compartment due
to its length of 6.7 miles.  As a result, a
ten-compartment configuration was used
to obtain the final calibration. This final
configuration accurately calculated total,
/

.90
2
1.0
.10

3

.90
4
1.0
.10

5
1.50
6
1.0
.10

7

.24
8
1.0
.76

9
.38
10
                                                                           1.0
                                                                           .62
Figure  1.    Final calibration configuration, compartments 1 through 10. Values are advective
            proportions.
Table 1.    MEXAMS Model Results, Naugatuck River

                      Model       Total Copper       o/0
Location            Compartment  Field    Model    Diff.
              Dissolved Copper    %
               Field   Model   Diff.
Water Column lug/1)
Palmer Br. Rd.
Bogue Rd.
Rte. 118
Campville Rd.
Br. Abutment
1
3
5
7
9
12.
A//» <
24.
16.
14.
11.5
Segment L
24.6
16.6
12.6
-4.1
tat a
+2.5
+3.8
-10.0
9.5
Kin
17.
14.
11.
9.1
Segment D
16.8
14.
10.6
-4.7
-1.2
0.0
3.6
Benthic Region (mg/l)
Palmer Br. Rd.
Bogue Rd.
Rte. 118
Campville Rd.
Br. Abutment and
Campville Rd.
Paniculate Copper
2
4
6
8
10

140.
62.
73.
177.
154.*

138.
59.
74.
179.
150.

-1.5
-4.8
+ 1.4
+ 1.1
-2.6

140.
62.
72.
176.
754.*

138.
59.
73.
178.
150.

-1.5
-5.0
+ 1.4
+0.6
-2.6

  The average copper concentration measured at Campville Rd. and the Bridge Abutment was
  used for comparison (based on model compartmentalization).
Table 2.   Comparative Kd Data,
            Naugatuck River
Compartment #
7
2
3
4
5
6
7
8
9
10
Original
logKd
0.30
3.72
-0.25
2.95
-0.05
2.67
-O.O5
2.67
-0.05
2.67
Calibrated
logKd
0.30
5.02
-0.25
4.20
-0.05
4.42
-0.05
5.30
-0.05
5.17
dissolved, and paniculate concentrations
for all water column and benthic compart-
ments. The calibration of this model is
only  the initial  phase  of accurately
modeling this river reach, however, and
model validation  to another data set is
required.


White River Case Study
  Natural loadings of heavy metals into
the White River system associated with
tributary discharge from  periodic storm
events can result  in dissolved metal con-
centrations that exceed the toxicity thres-
holds.  Dissolved copper  concentrations
increased from 4 ng/\ to 340 HQ/\ as-
sociated with  an  increase of river flow
from 500 ftVsec to 1100 ftVsec due to a
storm  event  in March  1975. Elevated
dissolved concentrations  for Zn, Pb, and
Cr also have been observed.
  The  White  River  is not  extensively
developed, although deposits of oil shale
are presently  being mined. The  target
metals chosen to exercise the MINTED
adsorption model in this study (Zn,  Pb, Ni,
Cu) are  associated  with  potential con-
tamination  to the  lotic  system  from
retorted  waste  shale  leaching into
groundwater and surface waters, retort
processing waters, and mine dewatering
activities. If leachate through  ground-
water  or  accidental  discharge into the
White River occurred, water quality stan-
dards established by the State of Utah
could be exceeded, thus resulting in vary-
ing degrees of  metal stress to the aquatic
communities.
  The  upper basin  of the  White River
originates in the  Flat Tops  Wilderness
Area in  northwestern Colorado,  which
contributes the majority of the flow. The
lower basin drainage area extends from

-------
northwestern Colorado to the confluence
of the Greeen River in northeastern Utah.
The lower basin area is characterized by
a semiarid climate. The watershed con-
tains surficial  deposits of lacustrine
sediments containing calcite and dolomite
deposited by Lake Unita (Palocene Epoch).
The physical-chemical processes associ-
ated with  the geologic weathering and
erosional  transport  causes  drastic
changes to water quality and gives the
White River a strong carbonate system.
The area of study on the White River is
located  in  the lower drainage basin area
in eastern Utah, near federal  oil shale
lease tracts.
  Because of the land morphology, the
study area has various perennial washes
and one ephemerial stream, Evacuation
Creek. For several months, the flow of
this creek is low and contains very high
levels of  total  dissolved  solids. The
watershed contains anthropogenic metal
sources that are eroded and transported
during storm events. This overland runoff
can  cause  significant water quality
changes to the  White  River by  heavy
metal and solids loading.
  The flows of the White River are cor-
related  to three  flow regimes  — upper
basin runoff,  lower basin runoff, and
baseline. Concentrations of major anions
and  cations found  in  the  White River
occur in the order of:
   HCOj > S042 > Cl > C032 > Ca+2 >
           Na+ > Mg+2 > K+
Upper basin  runoff provides  the best
water quality conditions, with  low dis-
solved constituent concentrations but high
total  suspended solids concentrations.
Lower basin runoff shows intermediate
dissolved  concentrations,  where  the
baseline regime shows the highest con-
centration of dissolved substances and
lowest total suspended solids values.
  The chemical  matrix (major  anions/
cations) of the water system had to be
determined for  subsequent input into
MINTED as  Type  I components. This
matrix can be determined by a full water
quality  analysis  on field samples, from
historical data, or from correlation graphs
that show the relationship of the major
anions and cations to a physical param-
eter (TDS). At least two years of constant
monitoring of the system should be used
to formulate the correlations, to encom-
pass possible seasonal  fluctuations.
Redox chemistries were not considered
in the water quality matrix because of the
high dissolved oxygen concentrations, the
relative concentrations of ammonia,
nitrate and nitrite, and the low  levels of
ferrous  ion  observed. It  was assumed
that these conditions would reflect a high
pe with sulfide being rapidly converted to
sulfate. High sulfide levels were reported
in alluvial  groundwater  of  Evacuation
Creek and Asphalt Wash.
  To  best  utilize  the capabilities  of
MINTEQ and  MEXAMS  for describing
metal fate and transport, data are required
to describe the solid phase partitioning of
any  potential  metal loadings. In most
cases, limited data  are available for suit-
able descriptions of sorption/desorption
and  chemical precipitation. A  certain
degree of caution  must be  exercised in
using literature data to estimate metal
partitioning to suspended solids  and/or
bed  sediments,  because natural sedi-
ments are  a mixed population of  solid
types (oxides, clays, detritus, hydroxides,
silicates, etc.). Site-specific data may often
be  necessary for successful  model
application.
  To illustrate how basic data may be
acquired for  site-specific  application,
adsorption  experiments have been per-
formed and the details reported. These
laboratory reactor studies quantify equi-
librium concentrations of dissolved and
particulate metals under controlled condi-
tions. These equilibrium  concentrations
can be determined for variable conditions
of pH, total metal  concentrations, TSS
concentrations, ionic strengths and matrix
concentrations. By varying  parameters,
adsorption relationships can be analyzed
and formulated for  MINTEQ input.
  Two reactor techniques were used for
determining adsorption behavior  for tar-
get metals — small volume reactors and
large volume.  Both of these techniques
used ambient  solids concentrations and
ionic strengths. The parameters that were
varied were total metal and pH.
  In implementing MINTEQ, the  user
must first determine chemical activity by
developing  an input file without  consid-
ering adsorption and chemical precipita-
tion. Execution of this file creates output
characterizing, among other things, the
dissolved metal ion activity, the  charge
balance,  and the saturation indices for
solids. For the charge  balance computa-
tions, two values should be noted, un-
speciated and speciated. An unspeciated
charge imbalance  of greater than 30%
can indicate that one or more ionic con-
stituent may be missing. The more  reliable
indicator, however, is the charge balance
after speciation. If this charge balance is
greater than 20%,  then it is very likely
that an ionic  component is absent. If
environmental data are lacking,  an ion
concentration adjustment may be required
on a conservative Type I component.
  An output section giving saturation
indices (SI) for all minerals and solids can
indicate what solids will precipitate foi
the full MINTEQ model run. The SI values
for diaspore, argonite, calcite and fluorite
were all positive, indicating the solid  is
oversaturated in the system. Because ol
kinetic factors, however,  the solids may
not  actually  obtain  equilibrium in the
time frame  considered, or may nol
actually form due to the constraints placed
on the system, such as no fixed partial
pressure (open to atmosphere).  Some ol
these solids form  under  high partial
pressures of C02 (groundwater  environ-
ment) and if used in a system  with nc
fixed partial pressure, a Gibbs Phase Rule
Violation will occur. General guidelines
for selecting solid phases are given in the
MINTEQ Technical Manual.

  To determine which MINTEQ adsorptior
model can be implemented for the White
River, data obtained from the metal addi
tion reactor studies and MINTEQ were
plotted. The  first approach  was to plo
particulate metal concentration (moles/I
versus dissolved metal ion activity con
centration  (moles/1) obtained  frorr
MINTEQ for various total metal concen
trations for a specific pH. A linear regres
sion was performed and the slope of the
line (activity Kd) was determined. The loc
value of the Kd was then utilized as ar
input parameter when implementing the
MINTEQ adsorption model.
  In utilizing the activity  Kd approach tc
model metal-sediment adsorption, a new
MINTEQ input file had to be developed
The component for surface sites (ID 990
was added as a Type I component and it!
activity fixed as a Type III species. Fou
Type V species (calcite, dolomite, diaspore
tenorite) were  inserted.  These  specie;
are allowed to precipitate in the system i
their saturation indices are exceedec
during chemical speciation. The last addi
tion involved a Type II modification (inser
tion of species not in  data base) tc
incorporate the adsorption  "reaction'
between Zn and the surface sites (SON).
  These new MINTEQ files were devel
oped and executed for  all Zn  reactoi
experiments  involving  four total meta
levels (0.124 mg/l,  0.224  mg/l, 0.52^
mg/l, and 1.024 mg/l) at two  specific
pH  conditions  that conformed to the
activity Kd approach (pH 7.50 and 8.39)
In order to evaluate how well  MINTEC
predicted  the laboratory adsorption re
suits, the predicted molar values of the
sorbed mass (particulate) and  aqueous

-------
(dissolved) masses were compared with
the actual experimental results from the
pH = 8.39 data set that conformed to the
linear Kd approach. Figures 2 and 3 show
the comparison plots of the MINTED pre-
diction versus actual  laboratory results
for both dissolved and paniculate Zn.
  If the  system being modeled has  a
relatively constant  pH and the metal
variation is not too great (<  2x), the
activity Kd approach would give satisfac-
tory results. If pH, metal level, SS and
other parameters change dramatically,
however, then one of the other adsorption
routines should be selected. For example,
zinc adsorption in White River  samples
exhibited a curvilinear relationship more
characteristic of Langmuir behavior. In
most situations, however, the metal level
is low enough that  the  linear portion
could be used  to calculate a Kd.  Lead
adsorption  was linear in all pH ranges
tested (6.40, 6.80, 7.30, 8.12, and 8.90)
and nickel  adsorption also appeared to
show a linear relation between 1 /(Ni+2)
and particulate Ni. A summary of the log
Kd's as a function of pH for the White
River are shown in Figure 4.
  The Langmuir model can often be used
to predict metal sorption  in natural sys-
tems. At low (Cu+2), the equation is linear;
whereas at very high (Cu+2), the equation
represents a saturation of surface  sites
(CuS = ST). The equation could  be  used
over a wide range of (Cu+2)'s and could
account for different site availability by
variation of ST (maximum surface cover-
age). This equation would be very useful
on the White River where large fluctua-
tions  in suspended solids and  ambient
dissolved copper are observed. Metal
partitioning to solids  is more significant
in the water column, as opposed to bed
sediments  and  thus, dissolved metal
should be related to suspended sediment
(and responsive to ST in the Langmuir
equation). As metal partitioning becomes
more  dominated by bed sediment inter-
action, the usefulness of the Langmuir
decreases and the activity Kd might  be
entirely appropriate.
  An experiment using the White River
with pH = 7.5  and  SS  = 200 mg/l
illustrated the  lower section of  a Lang-
muir type fit. Attempts to increase the
range of the plot  by Cu addition  only
resulted in malachite precipitation. A plot
of the  linearized  equation showed  a
curvilinear  relation when a straight line
would be anticipated if data fit a Langmuir
model.
  At this point, it is emphasized that the
MINTED input files were only considering
inorganic speciation of the metal. No
                I   I  I I I I Ml     I   I  I I I I  I II      I   I   I  I I  111!
                              Actual Dissolved Zinc, m

Figure 2.   Actual dissolvedzinc versus MINTEQ prediction.
organics were  used  even though  the
White  River  contains  5 to  7 mg/l of
dissolved organic carbon (DOC, see Table
3). The curvilinear relation could result
from omission of organic complexation,
an effect that would be most pronounced
at low (Cu+2). Addition of EDTA simulates
the effect of natural organic complexation
on the  (CU"2) calculation by MINTEQ.
  The optimum fit (linear) was obtained
with 55 /ig/l of EDTA (10'666M). Using
this organic level for two other experi-
ments  at pH  =  6.7 and 8.3 gave linear
plots indicating that this artificial specifi-
cation  of an  organic complexing agent
(EDTA) might be a reasonable substitute
for actual complexing organics in the test
water. The curvilinear nature observed in
these experiments (without EDTA) is not
due to  lack of consideration of desorbed
ions in the equilibrium solution. Under
the conditions of the experiment (addition
of >1  mg/l of metal),  background  con-
centrations of Ca+2, Mg*2, Na+, etc. would
appear essentially unchanged  in  the
White River matrix by release due to an
ion exchange phenomenon.
  The slope of the lines for pH = 6.7, 7.5
and 8.3 and  estimates for ST from the
experiments allowed a determination of
the Langmuir Constant, KL. The results
(Table 4) indicate that the metal adsorption
phenomena in the White River can  be
adequately described using kinetics based
on the mass of solids, the dissolved (free)
metal, the available sites and pH,  and,
that equilibrium  metal experiments can
be fit to an "activity" Langmuir Isotherm.
The experiments at pH = 6.7 and 8.3 did
not have sufficient data to permit precise
estimates  of ST and KL and the values in
the table should be regarded as estimates.
It is essential that continued development
of  methods for  determining dissolved
metal species be encouraged  and that
organic speciation of metals be considered
in modeling systems highly transitory in
flow and  water quality, such  as  that
observed in the White  River.
Ten Mile River Case Study
  This case  study of MEXAMS  involves
application  to  a wasteload  allocation
program  on the  Ten  Mile River  in
Massachusetts. This river receives metal
wastes, treated municipal wastewater,
and urban runoff at a number of points

-------
      10- T-
  1
      10-'
                           A   I  1  I  I I I
                                                             I  I I I  I I
         10'*                          10~s


                             Actual Paniculate Zinc, m

Figure 3.   Actual paniculate zinc versus MINTEQ prediction.
       White River
       SS = 1560 mg/l
                             Lead
                            Nickel
                          Zinc
                7     a
                  pH
10
Figure 4.   Log Kd,s versus pH for Zn, Ni,
           andPb.
along its 22-mile length, and estimates
of metal speciation within the river would
be useful for determination of acceptable
effluent limits to the basin.
  In recent years, the question has been
raised  as to whether  current inorganic
industrial discharge limitations on Ten
Mile River are  adequate to attain  or
maintain acceptable water quality levels
with regard to  heavy  metals. Also, the
long history of  metal  waste discharges
has resulted in extremely high concen-
trations of heavy metals in the sediments
of impoundments along the river. These
sediments now  act as additional sources
of metals to the river during high flow
periods.
  The  basic data  required  to execute
MEXAMS for a  system such as Ten Mile
River are 1) detailed descriptions of the
hydraulics, including flows and channel
geometry;  2) dissolved  and adsorbed
concentrations  of all  major chemical
species; and 3) loading rates of metals.
The 1984  data that are described are
preliminary; the data had not been com-
pletely analyzed and verified at the time
this project was completed.
  River and tributary flow measurements
at eight stations were available for several
dates in the summer of 1984. Because
heavy rainfall had occurred prior to the
surveys, all flows were significantly higher
than base flows for the basin. In addition,
the flows were steadily decreasing during
the surveys; consequently, they do not
represent steady-state conditions. Chan-
nel depths and widths were measured at
the flow  stations.  Additional  data to
characterize the channel geometry also
were collected at several cross sections
in the  1984  surveys.  Impoundment
dimensions (depth and surface area) were
obtained from previous model estimates.
Chemical concentrations and water qual-
ity parameters  were measured at river
monitoring stations and impoundments
on several dates during the 1984 surveys.
Table 5 lists the concentrations at various
monitoring stations used in this applica-
tion. Metal loadings data (concentrations
and discharge  rates) from  industrial and
municipal point sources represent aver-
ages of the data collected during separate
surveys in July 1984.
  Available data for this study contained
problems  that limited the scope of the
application and the reliability of the model
results. Primary among them is the lack
of steady-state  hydraulic conditions — a
crucial requirement of MEXANS —which
existed during the data collection surveys.
Metal concentrations were imprecise due
to the  analytical methods.  The  flame
atomic adsorption method provided an
accuracy of 0.01  mg/l for most metals,
which generally resulted in  reported
concentrations  of one  significant  figure
and many dissolved concentrations of 0.
Calibration of MEXAMS in Ten Mile River
requires metal  concentration measure-
ments accurate to 0.001 mg/l. A related
problem was the lack of metal adsorption
data for Ten Mile River. MINTEQ adsorp-
tion  models require precise measure-
ments of dissolved and adsorbed con-
centrations using  native water  and
sediments, as well as detailed chemical
analyses of the samples. Major cationic
species such as Ca2+, Mg2+, Na+, and K+,
were  not determined individually; con-
sequently, ionic strength computations in
MINTEQ were  inaccurate. Conductivity
measurements, from which ionic strength
can be estimated, were available.
  Copper speciation and adsorption was
simulated in a 15-mile stretch of the river
from the Fuller Pond outlet downstream
to Ten Mile Reservation Pond. The EXAMS
input was based on flow data of July 24,
1984. Segmentation of the river generally
followed the gross channel morphology.

-------
Table 3.    Water Quality Data, White River

    Selected Parameters
White River Surface Water Quality (11/1974-9/1976)
Baseflow

Total Alkalinity (mgCaCo3/l)
Dissolved Solids (mg/l)
Total Hardness (mg/l)
pH (units)
Conductance (^.ohms/cm)
Calcium (mg/l)
Magnesium (mg/l)
Sodium (mg/l)
Potassium (mg/l)
Chloride (mg/l)
Sulfate (mg/l)
Sulfide (mg/l)
DOC (mg/l)
Orthophosphate (mgP/l)
Ammonia (mgN/l)
Nitrate (mgN/l)
Cu fag/ 1)
2N fag/I)
Cd (ng/l)
Cr fag/I)
Pb (ng/l)
Ni (ng/l)
Fe fag//)
Ba fag/I)
Alfag/t)
Mn fag/I)
Mean ±
199 ±
526 ±
289 ±
81 ±
813 ±
70 ±
27 ±
71 ±
23 ±
40 ±
176 +
0.2 ±
6.8 +
.02 +
.03 +
12 +
6.0 +
21 +
0
2
2
5
28 +
48 +
17 +
4
S.E.
1.7
5.7
3.3
0.05
12.7
1.1
0.4
19
0.09
1.3
2.2

1 18
01
004
02
308
507




54
158
22

Max
242
717
460
8.8
1650
120
39
180
6 1
120
230
2.5
140
.87
19
1.10
160
180
1
20
7
16
270
140
80
20
Min
147
449
240
70
625
61
22
34
.9
24
140
0
25
00
00
.00
0
0
0
0
0
0
0
0
0
0
Lower Basin Runoff
Mean ±
177 ±
588 +
301 ±
8.2 ±
874 +
70 ±
30 ±
82 +
23 ±
41 ±
209 ±
0.2 ±
5.8 +
.01 ±
.05 +
.14 ±
41 +
12 ±
1
2
4
2
33 +
38 +
30 +
9
S.E
11.2
8.8
4.8
0.12
137
096
06
1.9
0.28
1.2
38

1 39
002
013
.029
188
37




4.8
102
57

Max
221
676
340
94
1010
76
35
no
4.0
58
260
1 3
12
04
.29
52
340
60
2
10
37
4
80
100
90
30
Mm
144
513
220
6.8
700
56
20
65
20
25
180
0
30
.00
00
00
2
0
0
0
0
0
10
0
0
0
Upper Basin Runoff
Mean ±
134 ±
293 ±
184 ±
7.8 +
473 ±
46 +
16 ±
32 ±
1.7 ±
15.5 ±
88 ±
0 1 ±
58 ±
02 ±
02 ±
16 ±
4 ±
4 ±
0
0
1
3
38 +
27 +
30 ±
2
S.E.
39
137
69
0.11
23.3
1 6
0.9
25
0.07
1.5
69

0.53
004
004
02
1 2
1.5




5.6
145
44

Max
204
536
280
8.8
810
67
28
75
3.0
42
190
2
8.6
.06
04
.38
60
20
1
0
4
6
90
130
80
10
Min
107
212
140
6.5
320
35
11
17
13
3
69
0
44
.00
00
01
1
0
0
0
0
0
0
0
0
0
Table 4.    Summary of Adsorption Modeling In The White River-Langmuir Constants and
           Activity Kd For Copper Partitioning
                                   Activity Langmuir Model
                  Activity
SS, mg/l
220
220
220
1590
1590
1590
1590


pH
670
750
830
680
7 46
831
895
KLST
143
243
6090
186
800
19300
136000
ST.
M/L
3 3 £-05
KL, L/M Kd
7 36 E +06
829
575
17000
933 000
Table 5. Water Quality Data, Ten Mile River
River Station
TM01
TM02
TM03
TM04
TM05
TM06
TM07
TM07A
TM08
TM08A
TM09
TM10
TM11
TM12
TM13
TM14
pH
7.2
6.8
6.9
7 1
7.1
7.2
72
70
7 1
72
7.1
7 1
72
7.1
6.9
73
Alkalinity
39.
27
26.
28
23
27.
34.
28.
29.
32
30
30
30.
29
25.
33.
Hardness
100
71
55.
52
42.
44
50.
39
39
42
42.
38.
38
36
38.
46
C1_
17
21
22
26.
22
21.
34.
31
67
35.
35.
32
29.
27.
28.
30
SO/
6.0
3.9
33
3.4
24
2.2
3.0
2.2
2.2
24
2.7
24
2.7
1.8
2.2
2 1
NO3-N
22
2.6
1.3
1.8
0.0
09
4.4
2.2
2.7
27
3 1
27
22
22
2.2
7.1
NH3-N
006
0.05
007
0.29
006
007
0.12
0 16
0.11
0 14
0 14
012
0.08
005
008
0.10
PO4-P
0.03
0.03
0.24
0.61
0.34
0.34
12
0.61
0.43
0.61
0.55
0.61
049
0.37
040
0.89
SS
15
6.5
12.
50
5.5
20
45
55
90
6.0
7.5
6.0
10
70
3.5
4.0
with  impoundments  and  intervening
channel sections forming the segments.
Fifteen segments or 30 EXAMS compart-
ments were delineated and divided into
three sets of ten compartments, because
MEXAMS is limited to ten compartments
per  run. A complete  Ten Mile  River
simulation  required three successive
model runs, with simulated metal outflow
from  the downstream  end of  one run
providing the input loading to the up-
stream end of the next run.
  MINTEQ input was developed based on
the water chemical survey data of July 2,
1984, and  mean copper loadings from
three July  1984 dates. Each  EXAMS
compartment  requires  an associated
MINTEQ input set. Because many of the
chemical constituent concentrations ex-
hibited little spatial variation, all channel
water column and benthic compartments
in each ten-compartment section were
assigned the same  MINTEQ input file.
Also, several ionic species were omitted
from  the final MINTEQ input,  because
they are unlikely to form complexes with
metals or to affect the equilibrium cal-
culation significantly The species omitted
were Cl, N03, PO43,  Na*, 10, and NH^.
  The  adsorption model selected  was
activity Kd.  Extensive data requirements
of other MINTEQ  adsorption models

-------
precluded their use  in this  study. The
available data were inadequate for reliable
application of the activity Kd model. Pre-
liminary MINTED runs were performed to
determine the activity Kd coefficients for
copper under different conditions of water
chemistry and sediment  concentration.
These Kd's were  subsequently used ii.
the MEXAMS simulation.
  The activity Kd  model  is  defined  in
terms of the activity of uncomplexed metal
as opposed to the total dissolved metal
concentration. This assumes that  only
free (uncomplexed) metal adsorbs. Thus,
in order to obtain an activity Kd coefficient
applicable to a specific water sample,
one must have "measurements"  of this
activity.  Analytical techniques  measure
the dissolved metal concentration, which
includes uncomplexed metal plus aqueous
complexes. To obtain the activity of the
metal, preliminary MINTED  runs were
performed on the water analyses used in
this application. The activity of uncom-
plexed metal computed by MINTEQ  in
these runs was combined with the cor-
responding measured concentration  of
adsorbed metal to obtain the activity Kd
for each water sample. The activity of the
uncomplexed metal  is affected by the
total ionic composition of the water as
well as  competition  with complexes;
consequently, a complete water analysis
is necessary to develop the MINTEQ input.
  MEXAMS  was executed  using the
available input data. Calibration  of
MEXAMS was not performed in this study
due largely to the non-steady-state condi-
tions exhibited in  the 1984 survey data
and lack of critical adsorption and metal
concentration data. It is instructive, how-
ever, to examine the results of the simu-
lation to suggest further data collection
and modeling efforts. In Table 6, simulated
copper concentrations in all water column
compartments and impoundment benthic
compartments of the first ten-compart-
ment section of Ten Mile River are com-
pared with observed concentrations. As
was noted previously, the MINTEQ input
for channel  benthic compartments was
not differentiated from  the overlying
water column because channel sediments
were  not sampled,  and   impoundment
sediments were  judged  to be  more
significant in transport/fate of metals.
  The results show a definite over-pre-
diction of both dissolved and total copper
in the water column while the benthic
results  are slightly low. Since the river
flow rates used in this study were higher
than  average, the  likely  cause  of the
discrepancies is either non-representative
metal loading rates (i.e., higher than the
Table 6.    Comparison of Predicted and Observed Copper Concentrations

                   Observed                               Mexams
Station
TMOJ
Wetherells
Pond
TM03.TM04
Falls Pond

TM06


we
B

we
B

Dissolved
(mg/l)
0.
0.
—
0.02
0.02
—
0.02
Total
fmg/ir
0.47
0.07
1000."
0.04
0.06
850.*
0.02
Compart-
ment
1
3
4
5
7
8
9
Dissolved
(mg/l)
0.15
0.15
0.15
0.11
0.088
0.088
0.075
Total
(mg/ir
0.47
0.44
684.*
0.34
0.27
418.*
0.23
WC - Water column
B  - benthos
* Benthic total concentration units are mg/kg

historical average) or an incorrect model
configuration such as not considering all
transport processes from  the water
column to the benthic sediment.  It is
likely that this system could be calibrated
by varying one or more processes so that
more metal is present in  the sediment
(particularly the impoundments), and less
in the water column. This would be done
by further increasing benthic adsorption
relative to the water column and/or in-
creasing  the dispersive transfer rates.
The resulting calibration would be difficult
to defend, however, given  the  uncer-
tainties in hydraulics, metal loading, and
observed metal concentrations.
  The detailed MINTEQ speciation for the
metal of interest  (Cu) is shown in Table 7
for water column and  benthic environ-
ments. As expected for a natural system
of this type, sorption dominates copper
fate, and uncomplexed copper probably
represents a small fraction of the metal.
  The factors that most affect metal fate
in Ten Mile River  are adsorption and
transport  within and through  the im-
poundments,  including settling  of ad-
sorbed metal  during high flow. Insuf-
ficient knowledge of these factors as v /ell
as  uncertainty  regarding typical  flow
conditions, impoundment hydraulics, and
metal concentrations limited the scope of
this study.
  Based  on these conclusions, several
recommendations  would facilitate a
 Table 7.    Typical Speciation of Copper
           In Ten Mile River
Species
Cu2 (aq)
Cu-SOH
CuC03(aq)
CufOHMaq)
CuHCtf, (aq)
Water
Column
6
67
12
13
2
Benthic
100
quantitative application. Data collector
should be  performed during low flow
steady-steady  conditions,  and all  dati
should be collected concurrently when
ever possible.  Because of their  impor
tance,  impoundments should be  wel
characterized with regard to hydraulics
including  dimensions,  flows, and  esti
mated  sediment trapping  efficiencies
Metal concentration data should be ac
curate  to  at least 1 ng/\, and routine
chemical  analyses  should include  th(
following:  H+, Na+, K", Cat2,  Mg+2, SO42
Cr, P04*  NO3, NhU, alkalinity Eh, anc
suspended sediment.
  A thorough  equilibrium  adsorptior
analysis should be performed for each
metal,  including precise  metal deter
minations  under varying  conditions o
sediment  concentration and total meta
to  allow  application of the Langmuii
isotherm  adsorption model.  Composite
sampling of metal discharges should be
performed to limit the effect of variable
flow and concentration. MEXAMS shouk
be  applied in  a manner similar  to tha
described  here with a refined compart
mental configuration and consideratior
of sediment settling.
  Calibration of the system could be per
formed by varying settling rates,  water
sediment  dispersion, and  adsorptior
coefficients to  vary the water column
benthic interaction. Alternately, one coulc
use a dynamic  transport  model thai
includes adsorption and sediment trans-
port to characterize dissolved and par-
ticulate metal  transport.  Subsequenl
MINTEQ simulations would be performec
to  obtain  more detailed information o1
metal  speciation  in  selected  critica
reaches of Ten Mile River, such as near a
large point source or in an impoundment

Sediment Settling and
Resuspension In MEXAMS
  Sediment transport is likely to be im-
portant in the transport of highly adsorbed
                                    8

-------
and precipitated chemicals in rivers. The
significant transport  processes  include
advection and dispersion within the water
column, deposition (settling), and scour
(resuspension). In many rivers, net depo-
sition over time can lead to effective
burial of chemicals within the  benthic
sediments. EXAMS' steady-state  hydrau-
lic simulation does not explicitly consider
the processes of sediment settling and
resuspension. Currently, river sediment
washloads are represented in EXAMS by
advective and dispersive sediment trans-
port  between water column compart-
ments. Also, bedload transport is repre-
sented  by  flows  between  benthic
compartments which are  horizontally
adjacent. However, net sedimentation and
resulting burial of chemicals must be
represented  by an artificial first-order
degradation  rate of the chemical in the
benthic sediments.
  A  simple  settling/resuspension  al-
gorithm was added to MEXAMS to im-
prove the model's steady-state sediment
transport capabilities. This algorithm al-
lows  the user to  specify steady-state,
vertical settling  from the water  column
to benthic compartments and correspond-
ing  resuspension.  Net  deposition of
sediment to the benthic compartment is
considered; however, net erosion is not
permitted. This  section  in  the report
documents the incorporation of these new
capabilities in M EXAMS including com-
putations added to the code, input, output,
and limitations.
Allen J. Medina is with Ecosystem Research Institute. Logan, UT 84321; and
  Brian R.  Bicknell is  with AQUA TERRA Consultants,  Mountain View,  CA
  94043.
Lee A.  Mulkey is the EPA Project Officer (see below}.
The complete report, entitled "Case Studies  and Model Testing of the Metals
  Exposure Analysis Modeling System fMEXAMS)," (Order No. PB 87-141 081 /
  AS; Cost: $24.95, subject to change) will be available only from:
        National Technical Information Service
        5285 Port Royal Road
        Springfield, VA 22161
        Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
        Environmental Research Laboratory
        U.S. Environmental Protection Agency
        Athens, GA 30613

-------
to
               > me
                  A Q>

                  Is
                  T)
                  3
                  3
                  -*!-»


               §1*

                    o
               j   i.
               w   o
               NJ   3
               S   3
               00   (0
                    JO
                     oo

            2      c


-------
United States
Environmental Protection
Agency
 Environmental Research
 Laboratory
 Athens GA 30613
Research and Development
 EPA/600/S3-86/044 May 1987
Project Summary
Modeling  the  Benthos-Water
Column  Exchange  of
Hydrophobic Chemicals
P. M. Gschwend, S-C. Wu, 0. S. Madsen, J. L Wilkin, R. B. Ambrose, Jr.,
and S. C. McCutcheon
  An analysis and modeling framework
was developed to simulate and predict
the transfer  of hydrophobic organic
chemicals between bed sediments and
overlying waters. This approach entails
coupling a description of the micro-
scopic scale process of sorption kinetics
with models  of the exposure of bed
particles to adjacent waters of varying
composition (i.e.,  due to diffusion of
solutes in interstitial fluids or pore water
advection, due to  biological mixing of
surficial sediments, due to suspension
of bed solids for a  period into the over-
lying water column.) Numerical simula-
tion  routines are  developed both for
sorption kinetics and to  demonstrate
coupling of this particle-water exchange
to particle movements in the case of a
biologically mixed bed. These routines
were used to assess the sensitivity of
sorption kinetics and the overall trans-
port  to chemical  and sediment  pro-
perties. Similar computer programs can
be used as subroutines in global chemical
fate models. Also a formulation of bed-
load  transport and of sediment resus-
pension was developed which yields
the contact time of bed particles with
the overlying water column. This model
result is then combined with the sorption
kinetics subroutine  to estimate bed-
water exchange in instances where
these processes greatly facilitate bed
particle-water column contact.
  This Project Summary was developed
by EPA's Environmental Research
Laboratory, Athens,  GA, to announce
key findings of the research project that
Is fully documented In a separate report
of the same title  (see Project Report
ordering Information at back).
 Introduction
  Several  models have been recently
 devised to describe the fate and transport
 of pollutants in bodies of water. However,
 these models are based on incomplete
 descriptions of the processes that control
 the exchange of chemicals between the
 bed and water column. In the current
 project, the authors describe the im-
 portant processes and develop mathe-
 matical descriptions that should be useful
 in updating existing models and devising
 new models. In addition, the final project
 report will  be a useful reference in
 describing the conceptual framework and
 relationships between direct sorption or
 desorption, diffusion, advection; biotur-
 bation and sediment transport.
  Figure 1  gives  the conceptual frame-
 work for describing the benthic exchange
 processes. For the purpose of this study,
 the aquatic environment was envisioned
 as consisting of  a water  column/n an
 active, moving bed load transport layer
 and an immobile  bed where sediment is
 stored. The definition of the active bed
 layer is taken to be two grain  diameters
 in thickness for sediment transport, and
 about 5 to 20 cm thick for bioturbation;
 however these definitions are arbitrary
 because thickness is difficult to forecast.
 The depth of the immobile layer is to be
 governed  by burial, compaction, and
 erosion processes. The water column may
 be described with more than one layer if
 significant chemical gradients exist and
 are necessary  to describe benthic
 exchange.
  Figure 1 also ranks the processes in
terms  of process energy requirements
and expected contact time between bed
particles and the dissolved phase of a

-------
   Geo-
   Morphological
   State
   Hydraulic
   State
 Ci,w
         Deposition, Burial, Compaction
                                                    Scour
                                                                   Water
                                                                   Column
    \
                                                    Coastal Areas, Estuaries, Streams
                             Lakes, Coastal Areas, Reservoirs, Estuaries
  Laminar or
  Quiescent Flow
                                                  /-A
                                                   Ejection
                                                                                                                Cone.
 Ca.w

                      O     O
ojo     o
o
o
       ef
             o
                                   o    oo
                              o
p
b
oo
     o
Exchange of
Sediment and
Surrounding
Fluid
                                                                                       Immobile
                                                                                       Bed
Direct
Sorption
Exchange
(D
Pore Water
Diffusion of
Dissolved
Species
(ID
Advective-
Dispersive
Flow
(III)
Biotur-
bation
(IV)

Bed Load
Transport
(V)

Suspended
Transport
(VI)

                                              Increasing Energy, Velocity, and Sediment Movement

                                              Generally Increasing Contact Time
Figure 1.   Processes involved in bed-water column exchange
chemical  in  the  water column.  Direct
sorption is expected  to  be the  least
energetic and slowest exchange process,
whereas sediment transport is expected
to be the most energetic and, to involve
some of the largest fluxes of material.
However, the limiting process may involve
the slowest, least energetic process.
  Although this  work  significantly im-
proves our understanding  and modeling
capability  for bed-water  pollutant ex-
change, several other  important issues
remain incompletely  developed.  For
example, the inclusion of colloids  and
their impact on  transport. We  do not
understand the sources  and sinks of these
nonsettling sorbents, particularly in sedi-
ment beds, and our knowledge of their
mobility in porous  media  and ability to
bind pollutants is limited. Additionally,
the importance of bioturbation and other
sediment modifying activities of benthic
organisms to bed  load  transport  and
resuspension is uncertain. Finally, sus-
pension of sediment particles from cohe-
                       sive beds remains poorly understood, and
                       therefore modeling of bed-water column
                       exchange for pollutants where cohesive
                       sediment is involved is limited to diffusion
                       and bioturbation-controlled situations.

                       Sorption  Kinetics
                         The formulation for sorption kinetics is
                       a physically  based description  of the
                       microscale processes encompassing dif-
                       fusion of nonpolar hydrophobic chemicals
                       into the pore space of natural aggregate
                       particles coupled  with local  partition
                       equilibrium as illustrated in Figure 2. The
                       research conducted during this  study
                       indicates that many natural particles of
                       importance to the sorption  process can
                       be described  as porous spheres  having
                       an  intraparticle porosity of about 0.13.
                       Based  on  this  conceptual  model the
                       sorption kinetics can be described as
                            -=D.
                                'eft
                   32Csw(r) 2 aCsw(r)

                               3r  J
                                       dr
                                (D
                                              where csw(r) =  total concentration of
                                              sorbate (chemical) at a radial distance r
                                              from the center of a particle and
                                                                 Dmn'+1
                                                              (1-n)PsKp+n
                                                                                 (2)
          in which D,,, = molecular diffusivity that
          can be  determined by  the  method of
          Hayduk  and  Laudie,  n  = intraparticle
          porosity of about 0.13, ps = specific gravity
          of the particles, and Kp = partition coef-
          ficient that can be predicted from  the
          normalized octanol-water partition coef-
          ficient, Kow, and the fraction of organic
          carbon contained in the natural particles.
          Thus equation  2 provides a  physically
          based method for predicting sorption and
          desorption.
            The flux of material from a  layer of
          particles on the surface of the bed can be
          determined from  the description of the
          fraction that is sorbed or desorbed at the
          end of the residence time, tr. The fraction

-------
                                       Turbulent
                                       Flowing Exterior
                                                              Molecular Diffision
                                                              in Pore Fluid
               Impenetrable
               Mineral Grain
        Stagnant, Nonflowing
          Interior forewater

Figure 2.    Physical picture of processes controlling sorption kinetics.
                                                Bioturbation
                                                  In the case of bioturbation, mixing by
                                                benthic organisms is described using an
                                                eddy viscosity scheme. This results in a
                                                flux expression of the form

                                                             3C         z
                                                   Flux = - Eb	= wbC = / f(z) C dz (5)
                                                              dz         0
                                                where Eb = mixing coefficient, C = total
                                                concentration  of chemical  in the  dis-
                                                solved, colloidal-bound, and sediment-
                                                sorbed phases, wb = vertical  sediment
                                                velocity induced by biological mixing, and
                                                f(z) = feeding activity due to ingestion of
                                                particles.
                                                  For plow-like bioturbation  involving
                                                mixing at the surface. Equation 5 can be
                                                applied by noting that wb and f(z) are zero.
                                                Table 1 gives the known values of Eb and
                                                the depth of the mixed layer for several
                                                species of  benthic animals. The wide-
                                                spread application of the  method will
                                                require determination of Eb and mixing
                                                depth for all species of  interest. Alter-
                                                natively,  the rate of  benthic  mixing is
                                                related to individual reworking rates, r'
                                                depth of mixed layer, L, population density,
                                                and bulk  density of the sediments, pb via
                                                        Eb = L r'(population)/pb
                                                                             (6)
sorbed or desorbed from or to an infinite
volume of water is given by:
(3)
          1-(6/7r)  J {(1/m2)
                   m=1
         exp(-Deffm2(^)2tr/R2)}
where M, = mass sorbed to the layer of
surficial bed particles over time tr, Mre =
mass attached to surficial  bed particles
after  infinite time,  m = the  number of
particle sizes the sediment is arbitrarily
divided into, Def( = effective intraparticle
diffusivity that is essentially  molecular
diffusivity retarded  by sorption, and  R =
particle radius. The residence time, tr, for
sediment  particles  can be determined
from  descriptions  of bioturbation  and
sediment transport. Equation 3 for infinite
water bodies is expected to be accurate
for  many  streams,  lakes, and estuaries
where water volumes are large compared
to the volume of surficial sediments. In
cases  where  this  may not  be true, a
numerical solution is derived to compute
M,/M«, and this solution is incorporated
in a basic program included in the final
report.

Diffusion and Advection
  The diffusive and advective flux of dis-
solved and colloidal material is described
by

                Flux =
                                         (4)
          dz

where n2 = porosity of the bed, i = an
empirical  factor dependent upon n2 and
determined by the formation factor that
describes  the effect  of  tortuosity on
molecular diffusion, Dm = molecular dif-
fusivity,  Dc  - diffusivity of colloidal
material, Sc = concentration of colloidal
and nonsettling material, and w2 =  pore
water  velocity in the bed.  Here  it  is
assumed  that the size of the  pores  is
large compared to the colloidal material.
Table 2 gives estimates of individual re-
working  rates  and  mixing depths  for
benthic  ploughers and conveyor-type
species.  Figures 3, 4, and 5 show the
sensitivity of the computed flux to particle
diameter, particle porosity, and phase
partitioning. The values on which these
calculations are based are given in Section
3.3.2.3 of the final report.
  Conveyor-belt  bioturbation  involves
worms that ingest sediment at some depth
z into  the bed  and egest the reworked
sediment at the bed surface. The worms
ingest the sediment for the organic carbon
contained in the sediments and in the
process rework the sediment into pellets
or trails of inorganic sediment bound by
mucous.  The reworking rate is
                                                  w'b = (
                                                                    population
                                   (7)
                                                Figure 6 shows the sensitivity of the flux
                                                to pellet diameter and the partitioning
                                                coefficient. See Section 3.3.2.4 for more
                                                details.

                                                Sediment Transport
                                                  The description of sediment transport
                                                is based on a physical framework  for
                                                cohensionless particles where the resis-
                                                tance  to movement  derives  from the

-------
Table 1.      Biogenic Mixing Coefficients. (Source: review by Lee and Swartz)
Location
Species
                                                                     L(cm)
                                       Eb(cm2/sec)
8 Calculated from data.
b Calculated by Guinasso and Schink (1975).
c Calculated by Alter (1978).
  Vertical diffusion coefficient.
"Horizontaldiffusion coefficient.
                                                             Method
Deep Sea, various sites
Mid-Atlantic Ridge
Long Island Sound
Chesapeake Bay
New York Bight
Rhode Island
0-1 cm
2-1 Ocm
La Jo/la, California

Barnstable Harbor,
Long Island Sound
Long Island Sound
Laboratory
Laboratory
Laboratory
Laboratory
?
?
Yoldia, Nucula
?
?

Leptosynapta, Scoloplos

Euzonous mucronata
f=Thoracophelia)
Pectinaria gouldii
Yoldia limatula
Yoldia limatula
Yoldia limatula
Clymenella torquata
Clymenella torquata
Molpadia oolitica
10-48
8
4
10-15
?

1
8

30
6
2
3
3
11
11
7-9
3.6 x Iff" -3.16 x Iff8
6x1ff9
1. 2-3.5 x10'6
1 x 10~6
5 x 10'7

2.9 x Iff6 -1.6 x Iff5
8.3 x 1 ff7 -4.3x1 ffe

1.5 x Iff5
7.6x1ffa
3.2 x 10'7
2 x Iff6
1 x Iff5
2-3 x 10~4
4.5 x Iff5
5.7-9.4 x Iff5
Dimensional analysis
210 Pb pattern
234 Th pattern
Dimensional analysis?
234 Th pattern

Dimensional analysis8
Dimensional analysis8

Dimensional analysis1'
Dimensional analysis'
Dimensional analysis'
Dimensional analysis0
Pore water profiles
Pore water profiles'1
Pore water profiles"
Depth of oxidized layer8
 Table 2.    Individual Particle Reworking Rates, Annual Reworking Rates, and Depth of Reworking. (Source: review by Lee and SwartsJ


Species
Annelids
Abarenicola claparedi
Abarenicola pacifies
Abarenicola pacifies


Amphitrite ornata
Amphitrite ornata
Arenicola marina
Clymenella torquata
Clymenella torquata


Guild

FUN
FUN
FUN


SISDF
SISDF
FUN
CB
CB
Individual
Reworking
Rate
(mg/ind/day)

3.600
10.900
0-4.500
0-15.000
—
5.100
2,600-5.200
4.700
900
1,650
Total
Reworking
Rate
(g/m2/yrj

—
—
—
—
310 kg
—
—
—
54,000
73,000

Depth of
Reworking

—
—
—
—

-------
Table 2.    (continued)


Species
Callianassa major
6 species
Paraphoxus spinosus
Uca pugillator
Uca pugnax
Echinoderms
Caudina chilenses
Echinocadrium cordatum
Holothuria spp.
7 species
Leptosynapta tenuis
Leptosynapta tenuis
Scotoplanes sp.
Stichopus ntoebil
Stichopus variegatus
Enteropneust
Balanoglossus gigas


Guild
MISSDF-E

MISDF-V
MISDF-E
MISDF-E

CB
MSSDF-V
MESDF

FUN
FUN
MESDF
MESDF
MESDF

FUN
"o = original data C = calculated from data
Individual
Fteworking
Rate
(mg/md/dayl
3.500

8.910
96
75

160.000
3.000
Total
Reworking
Rate
(g/m2/yr)
12.6-630 kg

54-2,200 kg
230
820

—
—

Depth of
Reworking
	
to<10cm
0-1 cm
—
—

—
—


Comments
Amount deposited per entrance, just faces

Burrowing
Just feces. recalculated from data
Just feces. recalculated from data



25.000-220,000 — —

J0.400-J8.4OO
34,000
100.000
38.000
49,000


—
590-3,000 kg
—
—
—


0.5- 10 cm
1.15cm
1 iMni
—
—


Feces and below surface reworking
Feces and below surface reworking
Feces



200,000-250,000 — —
fP = calculated by Power (J977)


Source
0

C
C
C

P
P
P

0
C
0
P
P

P
dH = calculated by Hargrave (19721
NOTE: Guilds CB. FUN, SISDF, MISDF-V, MISSDF-E are primarily tube, funnel, or deep burrow forming species whereas MISSDF-V. MESDF and MIFF are primarily
      surface ploughing or mixing species.
weight of the individual particles rather
than through interparticle  bonds. Thus
this  component  of  the description  is
limited to silty sediment and coarser sizes.
Furthermore the formulation is limited to
particles of a uniform size, and following
the work of Einstein, assumes that several
discrete size classes can be separately
described. This ignores the effect of large
sizes on the critical  shear stress of the
small particles and vice versa. Finally, the
conceptualization assumes  that  the
transport  system is  instantaneously  in
equilibrium between the suspended, bed,
and  immobile-bed loads  illustrated  in
Figure 7.
  Based on this conceptual model, the
distribution  of sediment mass at equili-
brium between the suspended, bed, and
immobile compartments is given by
                   P23
m,  =
    m,  = •
                               M  (8)
                               M  (9)
       P23 + P32 + P21 P32/P12


               P32

       P23 + P32 + P21 P32/P12


           P21 P32/P12
          P23 + P32 + P21 P32/P12
where M is  the total mass in the three
compartments  and pnm  are exchange
coefficients for sediment between layers
n and m.
  The mean downstream velocity for the
sediment mass is given as

         MU = m,  U, + m, U2     (11)
where U1 is the velocity of the suspended
sediment mass and U2 is the velocity of
the sediment mass in the bed-load layer.
From the average velocity of the sediment
mass, it is possible to compute the ex-
posure time of the sediment particles to
the water column over reaches of given
length as tr = length/U.
  The time of exposure or residence time
is coupled  with  the  sorption kinetics
model given in Equation 3 to describe the
transfer of a contaminant to or from the
sediment  moving  in  the stream. The
solution of Equations 8 through 10 in the
downstream direction describes distribu-
tion of contaminated sediment. The final
report  illustrates  the  solution of these
equations in examples for a  river and
deep river or reservoir.

Summary and Recommendations
for Future Research
  To estimate  bed-water exchange  of
hydrophobic  organic pollutants,  a  two-
step modeling approach or description is
recommended. First, particle-water ex-
change on the microscopic scale must be
quantified; this can be done using the
retarded  radial diffusion model, which
treats each  case as a function of com-
pound  solution diffusivity and hydrop-
hobicity and sediment particle size and
organic content. Section 2 of the  final
report describes a  numerical simulation
routine to handle  such solid-water ex-
change of chemicals even in cases where
there is a spectrum of particle sizes in-
volved and  the solution  concentrations
vary in time. Second, this particle-water
exchange kinetics description must be
coupled with descriptions of the relative
translations of sediment particles and the
adjacent fluids (i.e., due to porewater
advection, bioturbation, bed-load  trans-
port, or particle resuspension). This pro-
duces a prediction of the overall exchange
of chemicals between the bed and the
water  column. Section  3  of the final
report demonstrates  the  coupling of
particle-water  pollutant exchange in
biologically mixed beds. Section 4 devel-
ops a quantitative description of the expo-
sure of  a  moving  bed  particle to the
overlying water column and then couples
this transport to sorption kinetics.  In any
case of interest, decisions concerning the
intensity of various processes facilitating
bed  particle-water column  contact are
necessary before good predictions  of pol-
lutant transfer can be expected.
  Several areas of  future research are
suggested to improve and extend these
analytical methods:
  (1) The sources and fates of colloidal
materials in sediments needs  to be
examined. Additionally, the sorbent pro-
perties of these macromolecules or
microparticles should be assessed. These
sorbents may be particularly important in
transporting very hydrophobic pollutants
from beds that are not biologically mixed.
  (2) The nature of bed particle and pore
water movements under the influence of
benthic infauna should be explored fur-
ther. Pore water pumping (or irrigation)
was neglected here for want of a general

-------
          I
           I
                      0.08 1
                      0.06
                      0.04
                       0.02
                      0.00
                  Diameter (cm)

                    0.005
                                                    Diameter (cm)
                                          100
                                                    200
                                        Time (days)
I
 s
 1
 s
0.06
      0.04
      0.02
      0.00
diffusion length scales and intra-aggre-
gate porosity for solids as they exist in a
bed should be researched further. Also,
to extend this  approach to other con-
taminants such as trace metals and polar
organic compounds, the mechanisms
controlling their sorption kinetics inter-
actions with sediment particles should
be examined.
  (6) Finally, efforts should be made to
test the accuracy of model predictions
against real world situations. Currently,
there is a dearth of field data for com-
parison with model  predictions. Thus,
bed-water fluxes must be measured at
times and  places where the prevailing
bed  mixing processes are known and
ancillary data are obtained to estimate
their intensity.
Figure 3.    Sensitivity of the plow-like bioturbation mediated pollutant flux to 5 different
            sediment particle sizes.  The values of other parameters are same as those in
            the example problem in Section 3.3.2.3 of final report.
quantitative description of this process as
a function of organisms involved.  Also,
approaches for estimating parameters and
better quantifying the mixing activities of
benthic infauna from field measurements
are needed.
  (3) The development of a basic under-
standing for the  factors and processes
governing cohesive sediment resuspen-
sion and transport  is  also  necessary.
These  cohesive organic-rich muds are
the predominant sites for  collection of
many  pollutants  discharged  to  natural
waters, yet our ability to quantitatively
describe the movements of particles in
these beds remains poor.
                                     (4) In the sediment transport models
                                   formulated here,  steady flow conditions
                                   were assumed. The impact of unsteady
                                   (e.g.,  tides  in  estuaries),  and  even
                                   catastrophic (e.g., storms) phenomena to
                                   the modeling of sediment transport still
                                   remains  an  important  area to  be
                                   examined.
                                     (5) Further  assessment  of the con-
                                   ceptualization of  the microscopic  scale
                                   particle-water  exchange  of  chemicals
                                   from particles in beds to the surrounding
                                   pore waters should be done. The retarded
                                   radial diffusion model has been tested
                                   primarily for aggregate particles in sus-
                                   pension. Issues such as the appropriate

-------
 I
CM
 5

 I
       0.30 n
0.20
       o.ro
       o.oo
                  o.w


                  0.08-


                  0.06


                  0.04-


                  0.02


                  0.00
                                                    Microporosity
                                               100
                                                                200
                                                    Time (days)
                                                                                 300
                                                                           103
Figure 4.    Sensitivity of the plow-like bioturbation mediated pollutant to 5 different sediment
             intraparticle porosities.  The values of other parameters are same as those in
             the example problem in Section 3.3.2.3 of final report.

-------
                          0.10
                          0.00

     0.80 n
     0.60
      0.40
     0.20
      0.00
                        /Cow
                                                700              200              300
                                                   Time (days)
                                                                         W3
                                      Time (days)
Figure 5.    Sensitivity of the plo w-like bioturbation mediated pollutant to chemical partitioning.
             The values of other parameters are same  as those  in the example  problem in
             Section 3.3.2.3 of final report.

-------
                Initial Concentration = 1 fjg/g
                Reworking Rate - 0.052 cm/day
        0.03 T   Bulk Density of Sediments = 0.5 g/cm3
                Microporosity = 0.13
        0.02
        0.0 1-
         000
                           10*
10s


Kow
10e     107
10"
10*
Figure 6.    Sensitivity of conveyor-belt type bioturbation mediated flux to Kovl and pellet size.
           The values of other parameters are same as those in the example given in Section
           3.3.2 4 of final report.
                                                                   ' U,
            Layer 1  Suspended Load
            ia^er 2.- Bed /.oarf
           	  bed-water interface
            Layer 3' Immobile Bed
      H-^
      I       T >>»
Figure 7.    Definition sketch: three-layer transport model.
  P. M.  Gschwend,  S-C.  Wu. 0. S. Madsen,  and J.  L  Wilken are with the
    Massachusetts Institute of Technology,  Cambridge, MA 02139; the EPA
    authors R. B.  Ambrose, Jr., and S. C. McCutcheon (also the  EPA Project
    Officer, see below) are with the Environmental Research Laboratory, Athens,
    GA 30613.
  The complete report, entitled "Modeling the Benthos- Water Column Exchange
    of Hydrophobic Chemicals," (Order No. PB 87-145 389/AS; Cost: $24.95,
    subject to change] will be available only from:
          National Technical Information Service
          5285 Port Royal Road
          Springfield, VA22161
          Telephone: 703-487-4650
  The EPA Project  Officer can be contacted at:
          Environmental Research Laboratory
          U.S. Environmental Protection Agency
          Athens,  GA 30613

-------

-------
 
  3
  O
  O
  D
  o
  3
  0)
  O
  X

  ^
  01
  N>
  O>
  00
TJ
O
CD c


-------