United States
Environmental Protection
Agency
Office of
Research and Development
Washington, DC 20460
EPA/600/3-91/046
August 1991
xvEPA
Application of a Water
Quality Assessment
Modeling System at a
Superfund Site
-------
EPA/600/3-91/046
APPLICATION OF A WATER QUALITY ASSESSMENT
MODELING SYSTEM AT A SUPERFUND SITE
by
Kendall P. Brown1, Edward Z. Hosseinipour1,
James L. Martin1,
and Robert B. Ambrose, Jr.2
lAScI Corporation
Athens, Georgia 30613
Environmental Research Laboratory
U.S. Environmental Protection Agency
Athens, Georgia 30613
ENVIRONMENTAL RESEARCH LABORATORY
OFFICE OF RESEARCH AND DEVELOPMENT
U.S. ENVIRONMENTAL PROTECTION AGENCY
ATHENS, GEORGIA 30613
Printed on Recycled Paper
-------
DISCLAIMER
The information in this document has been funded wholly or in
part by the United States Environmental Protection Agency under
Contract Number 68-03-0355 to AScI Corporation. it has been
subject to the Agency's peer and administrative review, and it has
been approved for publication as an EPA document. Mention of trade
names or commercial products does not constitute endorsement or
recommendation for use by the U.S. Environmental Protection Agendy.
ii
-------
FOREWORD
EPA's Center for Exposure Assessment Modeling (CEAM) was
established in July 1987 to meet the scientific and technical
exposure assessment needs of EPA's Program and Regional Offices and
Superfund Technology Support Center for Exposure and Ecorisk
Assessment. The Center is also the focal point for a variety of
general Agency support activities related to the scientifically
defensible application of state-of-the-art exposure assessment
technology for environmental risk-based decisions. CEAM provides
analysts and decision-makers operating under various legislative
mandates with relevant exposure assessment technology, training and
consultation, technical assistance, and demonstration of new or
innovative applications. This research report describes one such
demonstration project - analysis of metals contamination of the
Upper Clark Fork River, Montana.
Rosemarie C. Russo, Ph.D.
Director
Environmental Research Laboratory
Athens, Georgia
111
-------
ABSTRACT
Water quality modeling and related exposure assessments at a
Superfund site, Silver Bow Creek-Clark Fork River in Montana,
demonstrate the capability to predict the fate of mining waste
pollutants in the environment. A linked assessment system —
consisting of hydrology and erosion, river hydraulics, surface
water quality, metal speciation, non-point source and groundwater
mixing and transport models—has been applied at the site to show
the applicability of such modeling schemes and the complexities
involved in the application. Some of the models had to be modified
to match the requirements of this project. Graphs of the water
quality parameters show good fit between the measured and predicted
concentrations at some stations whereas substantial deviations are
observed at other stations along the course of the stream.
IV
-------
CONTENTS
Disclaimer ii
Foreword iii
Abstract iv
Figures vi
Tables viii
Acknowledgments x
1. Introduction 1
2. Conclusions and Recommendations 2
3. Background 5
4. Objectives 11
5. Models and Methods 12
6. MINTEQA2 Results 32
7. PRZM Calibration 42
8. Mining Waste Mass Estimates 55
9. Removal of Metals in the Warm Spring Ponds in
NPSOUT 58
10. Hydrologic and Hydraulic Studies 62
11. Results 67
12 . References 79
-------
FIGURES
Number Page
1 Upper Clark Fork Basin of Silver Bow Creek, MT, a
Super fund S ite 6
2 Diagram of Modeling System Used in the Clark Fork River
Exposure Assessment 12
3 Oxidation of a Sulfide Particle 16
4 Profile of the River, Water Table, and Tailing Deposit.. 16
5 Geochemical Profile of the Unsaturated Tailings and
Alluvium 17
6 PRZM Evaluates the Magnitude of Metals loading to the
River for Each of the Three Pathways 22
7 Vadose Zone Characterization at Sites Along Silver Bow
Creek 33
8 Plot of Tailing Deposit Depth versus Eh 35
9 Storm Runoff Sampling Sites at Silver Bow Creek 47
10 pH Versus Pond Metal Removal Efficiency for Flow Through
Pond #3 and Pond #2 60
11 Flow Through Versus Metal Removal for Flow Through Pond
#3 and Pond #2 60
12 Cross-Section of a Typical Erodible Channel 65
13 Measured versus Predicted Copper Loadings 69
14 Measured versus Predicted Copper Loadings 69
15 Measured versus Predicted Copper Loadings 70
16 Measured versus Predicted Copper Loadings 70
17 Measured versus Predicted Copper Loadings 71
18 Measured versus Predicted Copper Loadings 72
19 Predicted and Measured Winter Floods, Segment 1 73
20 Predicted and Measured Winter Floods, Segment 11 73
vi
-------
21 Predicted and Measured Spring Floods, Segment 1 74
22 Predicted and Measured Spring Floods, Segment 11 74
23 Predicted and Measured Normal Winter Flow, Segment 1.... 75
24 Predicted and Measured Normal Winter Flow, Segment 11... 75
25 Predicted and Measured Normal Spring Flow, Segment 1.... 76
26 Predicted and Measured Normal Spring Flow, Segment 11... 76
27 Predicted Normal Fall/Summer Flow Segment 1 77
28 Predicted Normal Fall/Summer Flow Segment 11 78
VII
-------
TABLES
Number Page
1 Waste Site Areas 23
2 WASP4 Segments Along the Clark Fork River and Silver
Bow Creek 30
3 Eh versus pH for a Fixed Oxygen Content 36
4 Eh versus Oxygen Content for a Fixed pH 36
5 Demonstration of Soil/Tailings Strata and Geochemistry.. 37
6 Typical Components of the Tailings Unsaturated Zone
Pore Water 39
7 Species formed for pH=3.0, Eh=?600mV 40
8 Calculation of Kd from Mass Fractions 40
9 Copper Kd and Dependence on pH and Eh 41
10 Hydrologic Parameters for Water Flow and Mass Transport
for PRZM 44
11 Initial Parameter Grid for 2.44-cm Storm Precipitation.. 45
12 Sensitivity of PRZM Parameters 46
13 Simulated Rainfall 48
14 Antecedent Moisture Conditions: Fixed Initial Water
Content 49
15 Antecedent Moisture Conditions: Fixed Rainfall for
Previous Week 49
16 SCS Curve Number Versus Infiltration for 2.29-cm Storm.. 50
17 SCS Curve Number Versus Infiltration for 9.14-cm Storm.. 51
18 Total Suspended Solids for May 29, 1985, Storm 51
19 Erosion Parameter Grid for 0.97-cm Storm 52
20 Flow Composite Data for May 29, 1985, Storm 53
viii
-------
21 Peak Copper Fluxes into the Clark Fork Basin 56
22 Removal of Copper in Suspension and in Solution by
Precipitation and Flocculation in Warm Springs Pond 58
23 Superfund Subsite NPSOUT Final Parameters 67
IX
-------
ACKNOWLEDGMENTS
We would like to acknowledge the cooperation of the Superfund
managers in Helena, Montana and the State of Montana Water Quality
Bureau in providing us with the information on water quality. Gary
Ingman and Scott Brown were especially helpful in this regard.
This document was developed by the Center for Exposure
Assessment Modeling with the support of the Superfund Technology
Support Project (TSP), managed by the Office of Program Management.
The assistance of Robert Carsel and Dr. David Brown of the
Environmental Research Laboratory, Jerry Allison of the Computer
Science Corporation, and Timothy Wool of the AScI Corp. for
guidance on program operation and software support are appreciated.
The cooperation of Dr. Russell Erickson of EPA's Environmental
Research Laboratory, Duluth, MN, is appreciated.
-------
SECTION 1
INTRODUCTION
This project was initiated to analyze the hydrology and water
quality of the Silver Bow Creek and the Upper Clark Fork River,
Montana, in response to years of mining activity in the surrounding
areas. Heavy metal concentrations in the surface and subsurface
waters of the Clark Fork River basin have diminished aquatic life
in many of the region's streams. The principal sources of these
metals are the waste byproducts of copper mining in the Silver Bow
Creek watershed above and around the town of Butte, Montana. One of
the Superfund site, the Silver Bow Creek/Butte Area site, includes
the Butte mining areas, Silver Bow Creek, the Warm Springs Ponds,
and the Upper Clark Ford River down to the entrance of Milltown
Dam. The other Superfund site is the Anaconda Smelter site. The
Anaconda Smelter site includes the Anaconda Smelter, the
surrounding waste and slag dumps, the Anaconda Ponds, and the
Opportunity Ponds. A Superfund remedial investigation has been
conducted for the Silver Bow Creek/Butte Area site and a set of
volumes that constitute the Silver Bow Creek Remedial Investigation
Final Report (RIFR) have been used as our primary references
(Tuesday et al., 1987).
-------
SECTION 2
CONCLUSIONS AND RECOMMENDATIONS
The copper concentrations in discharges from the Warm Springs
Ponds and the concentrations in Silver Bow Creek (WASP4 Surface
Water Segments 1-7) have been simulated with accuracy, as shown in
the results for Segment 1. The variance between the measured total
copper concentrations in winter flood waters in the Upper Clark
Fork River and the predictions in Segment 8 to 11 obtained using
the methods and models described in this report, show that, for the
main stem of the Clark Fork River, copper in the river may have
unidentified sources currently not included within the model.
For example, the stream bed may hold a large reserve of
tailings deposited during floods that bypassed the Warm Springs
Ponds or deposited during floods and high flows before these ponds
were constructed. Resuspension of stream-bed sediments that carry
heavy metal contaminants may be the primary contaminant source if
flood flows have sufficient hydraulic power to strip away the
armored (cemented with lime and iron oxide floes) cobbled (cemented
over large rocks) bottom of the river and resuspend the sediments.
The primary metal source during large flows and floods may be
resuspended copper oxide-iron oxide precipitates, sulfide tailing
particles, and copper adsorbed to iron oxide precipitates and to
primary alumino-silicate sands.
It has been reported that flows in winter include ice flows
and phenomena characteristic of unstable stream beds such as "head
cutting" or large-scale bank erosion. The stream bed of both Silver
Bow Creek and the Clark Fork River down to the town of Deer Lodge
have been extensively reengineered and straightened during road
construction, railroad construction, and possibly during the
construction of the Warm Springs Ponds (Gary Ingman, personal
communication 1989). The stream bed may be unstable during large
flows and the stream bed may be widening with successive flooding
events. This may cause the stream bed to widen sufficiently to
reenter the older sections of the stream bed, causing further
instabilities in the stream flow path and stream banks.
During winter floods, major changes occur to the transport
patterns when the subsurface is frozen. Freezing of the subsurface
reduces infiltration, creates ice lenses, and affects the chemical
and microbiological activity in the frost zone. These phenomena are
not included in the chemical transport model used (PRZM).
Therefore, the runoff in winter should include a much larger
portion of the snowfall, rainfall, and snowmelt than is currently
predicted, and a larger runoff could result in a narrowing of the
hydrograph for the winter floods.
-------
A narrow hydrograph may lead to concentration versus time
peaks that are sharper than the l-day resolution permitted by PRZM.
Concentration may vary significantly over a single day for cases
when the hydrograph is narrow. An improvement in model resolution
to time scales of less than a day may be necessary for such cases.
We also can conclude that there may be substantial contaminant
transport events in the river that have not been looked for
previously. This conclusion is suggested by the fall and late
summer surface water model results. The model, therefore, could be
used to plan river sampling for heavy metals, sulfates, metal-
bearing colloids, sediments, and other components.
Resuspension of sediments in the Warm Springs Ponds is
accounted for with constant first-order metal removal efficiencies.
With current data it is difficult to determine causation in the
metal removal efficiencies in the ponds, although general trends
can be inferred by regression against the metal removal data. These
regressions indicate that pond efficiencies increase with pH and
decrease with exit flow from Pond 2, with the confidence levels for
the regression coefficients being highest for the correlation of
copper and zinc removal to the pH of Pond 2 effluent.
The exposure model could be extended to the other metals
present in the river. Although copper is the metal most in
exceedance of acute toxicity concentrations in stream surface
waters below the Warm Springs Ponds at the start of the Upper Clark
Fork River, zinc regularly exceeds acute toxicity standards in Pond
2 and 3. Arsenic also is present in the stream and in groundwater,
so application of the exposure model to these other heavy metals
would be useful. For extension to zinc, arsenic, and cadmium, data
such as Kd for the groundwater in the oxidation layer would be
needed. The pond removal efficiencies for zinc, arsenic, and
cadmium also would be needed. Data are not available for other
contaminants present in groundwater (silver,antimony,cobalt, lead).
Pond efficiencies in this model have been approximated by an
average value. The seasonal determinants of pond efficiencies and
the dependence of pond metal removal efficiency on flow and pH are
not understood well enough to incorporate efficiency predictions
into the metals loading model.
To predict the ratio of dissolved to suspended metal
concentrations in the stream, pH data and information about the
dependence of Kd on pH would be needed. If chemical equilibrium is
assumed, these relationships could be predicted by use of an option
for variable Kd within the surface water model and by use of a
predictive mechanism for pH that accounts for carbonate balances
and acid loadings over the stream course.
-------
One suggestion and conclusion for improving the quality and
predictive power of the modeling system is to modify the program
GCTRAN, which represents large colloid transport in groundwater. If
improvements are made in estimating the rate of groundwater
discharge to the stream, infiltration, changes in the water table,
and the effect of bank storage and evapotranspiration, then a more
accurate simulation of the streamside contaminant sources may
improve the predictive capability of this modeling system.
The goal of improving prediction of contaminant transport
raises the need for a software linkage between surface and ground
waters based on: (1) better equations of flow in groundwater and
surface water for this site; (2) improved equations that define the
rates of infiltration from the surface to the groundwater; (3)
equations that define kinetics of particle oxidation and changes in
metal form, including changes in contaminant distribution and form
over different particle sizes, changes in contaminant distribution
and form over different particle structures (porosity, morphology) ,
and changes in contaminant metal distribution and form over
different compositions of the particles to which the metal attaches
(such as sulfides, quartz, limestone, iron oxide, mixed silicon and
aluminum oxides (clays)); and (4) expanded equations that describe
the effect of groundwater flow on the movement of contaminants in
solution, in suspensions and within immobile solids. Having a
software linkage would be a great advantage on any other project
where the transport and flow in the saturated zone and in a
variably porous and conductive subsurface controls the release of
contaminants.
-------
SECTION 3
BACKGROUND
DESCRIPTION OF THE STUDY AREA
Silver Bow Creek is located in southwest Montana, where it
flows as a small stream from the Metro Storm Drain northwest of the
city of Butte to the Warm Springs Ponds 27 miles downstream (Figure
1). Its confluence with Warm Springs Creek near Warm Springs and
Anaconda just below Warm Spring Pond 2 forms the Upper Clark Fork
River. Several small streams and gulches feed the creek between
Butte and Warm Springs.
The drainage area of silver Bow Creek is on the order of 425
square miles. Both acidic mine drainage and contaminated
groundwater seepages enter Silver Bow Creek within the Butte town
limits before reaching the Colorado Tailings. Below the town
boundary, the creek continues to pass large streamside tailings
deposits and flood-deposited tailings banks.
Some small tributaries, including Missoula Gulch, Brown's
Gulch, and German Gulch, flow into Silver Bow Creek above the Warm
Springs Ponds, a series of three settling ponds. The creek flows
through Warms Springs Pond 3 (100 hectares of open water) into Pond
2 (32 hectares of open water) and into a small set of subsidiary
Wildlife Ponds. The lower ponds, including Pond 1 (8 hectares of
open water) are nearly filled with sediment and sediment islands
break the water surface of Pond 2 and the Wildlife Ponds. Below
Pond 2 the creek merges with the Mill-Willow Bypass, which drains
Mill and willow Creek and some of the waste subsites surrounding
the Anaconda Smelter and the seepage from the Warm Spring Ponds and
the Opportunity Ponds. Below this confluence, the creek combines
with the flow from Warm Springs Creek (which drains the remaining
Anaconda waste subsites) and becomes the Upper Clark Fork River.
Modesty Creek, Lost Creek, Dempsey Creek and Racetrack Creek
increase the river flow between the Warm Springs Creek and Deer
Lodge.
-------
UPPER CLARK FORK
RIVER
DEER
LODGE
RACETRACK
CREEK
MINING
SITES
BLACKTAIL
CREEK
YANKEE DOODLE
POND
N
WARM SPRINGS
PONDS
SILVER BOW
CREEK
WARM SPRINGS
CREEK
COLORADO
TAILINGS
BROWNS
GULCH
WILLOW
CREEK
ANACONDA
SMELTER
GERMAN
GULCH
Figure 1. Upper Clark Fork basin of Silver Bow Creek, MT, a Superfund site.
-------
More than 100 years of continuous mining operations and
related activities have changed the area's natural environment
greatly. Waste rocks, ore process tailings, acidic mine drainage
and smelting wastes are the primary sources of heavy metal loadings
to the Silver Bow Creek and Clark Fork River via surface runoff and
ground water flow. The Warm Springs settling ponds have a limited
volume and during high stream flows (discharges of greater than 700
ft3/sec) the Warm Springs Ponds are bypassed into the Mill-Willow
Bypass without any treatment. The Ponds also may be bypassed by
flows as low as 150 ft3/sec if the entry gate into Pond 3 is
blocked by debris. Further, geotechnical studies have revealed that
a flow of 4000 ft3/sec can result in the failure of diversion and
control structures such that a large amount of contaminated
sediments are released into the Clark Fork River. Hydrological
investigations estimate the 100-year flood to be 3600 ft3/sec
(CH2M-Hill 1988). The RIFR indicates that the seepage from under
the Warm Springs Ponds can act as a major source of ground water
pollution. Toxic elements within tailing deposits include arsenic,
cadmium, copper, lead, iron, and zinc. The severity of the
contamination problem was such that, in 1983, EPA declared the area
along the course of the Silver Bow Creek and Clark Fork River from
Butte to the Milltown Dam to be a high priority Superfund site. The
site hydrology, hydrogeology and geochemistry are very complex due
to the variety and magnitude of contaminant sources and the
multitude of pathways to the surface water and ground water
resources. Therefore, a detailed modeling scheme was employed to
delineate the pathways and the fate of pollutants.
HISTORY OF METALS IN THE RIVER
Since 1880, when large scale mining and smelting of copper
began, the valley and the stream have been used as dumping areas
for wastes. Wastes in Butte include tailings from the flotation
process that separates copper from the ore and waste rock that was
removed as backfill and overburden, or, was discarded as being too
low-grade to be put through a separator.
Pollution problems began early along the Silver Bow Creek. The
first industrial operations sluiced the wastes directly into the
stream. Later, the mine operators constructed settling ponds and
streamside tailings piles as part of an attempt to preserve water
quality. The wastes in the stream moved down the river, especially
during floods which caused erosion and transport of sediment, and
were widely distributed over the flood plain and the river bed, at
least as far as Milltown Reservoir (below Deer Lodge).
In 1911 and 1916, Warm Springs Ponds 1 and 2 were constructed
on Silver Bow Creek above the confluence of Silver Bow Creek with
its two principal tributaries, Warm Springs Creek and Mill-Willow
-------
Creek. The height of Warm Springs Pond 2 was raised during 1967-
1969. These ponds originally were designed to settle the metals
carried by Silver Bow Creek and prevent contamination further
downstream.
As the two ponds lost capacity due to sediment accumulation,
sedimentation efficiency declined and more particles remained
suspended in the effluent and were carried into the river. To
remedy the problem, Warm Springs Pond 3 was constructed above the
first two ponds between 1954 and 1959. Pond 3 was improved between
1959 and 1969 to increase the capacity for metal removal.
Beginning in 1967, lime has been added to the ponds to
precipitate and flocculate the heavy metal contaminants that enter
with the stream flow. This method of settling the metal colloids
and particles from Silver Bow Creek has been successful in reducing
the metal content of the Clark Fork River during periods of normal
flow. During high flows, however, the ponds are bypassed and Silver
Bow Creek enters the upper Clark Fork River without treatment.
Between 1933 and 1937 the stream itself was channelized to
prevent further erosion of the tailings from the banks. The first
alteration to the stream course of the Silver Bow Creek was a
channelization of the flow between smelter slag blocks placed along
the stream channel. This was done to prevent the downstream
transport of newly deposited mine and slag tailings on the old
banks during periods of bank overflow due to flooding.
BIOLOGICAL IMPAIRMENT
Adjacent to Silver Bow Creek and the Clark Fork River are
flood plains and low banks that have been covered with waste
sediments. Silver Bow Creek in its natural state was a meandering
brook through lush marshes and wetlands. The original environment
included wetland bogs, moss banks, and a rich floral habitat that
may have supported a wide variety of higher invertebrates (for
example, crayfish), fragile aquatic and wetland plants,
salamanders, turtles, and wetland-nesting birds. In many places in
the Upper Clark Fork River Basin, below the flood-deposited gravel
and mine waste is a thick layer of organic material, bearing
evidence to the original biological diversity and richness of the
basin.
In some areas, heavy metals from the sediments have either
limited plant growth to metal-tolerant species or killed entire
plant communities. These areas are called slickens, a term that
applies to all of the areas that are either dead or have visible
biological impairment. The past use of water from the Clark Fork
River for irrigation has led to the contamination of grazing lands.
The biological habitat within the river and creek has been
damaged as well. Silver Bow Creek does not support trout, whereas
8
-------
the Upper Clark Fork River supports a brown trout population that
suffers from both chronic and acute toxicity. The river always
contains significant metal toxicants, such as zinc, copper,
arsenic, lead, cadmium, and aluminum.
Mass fish kills occur during floods, and the acute toxicity of
the stream due to elevated metals concentrations is thought to be
the cause. During flood flows (which can occur during late winter
snow melts and during early and late summer heavy rain storms),
flow in Silver Bow Creek can bypass the Warm Springs Ponds and
enter the Upper Clark Fork River. Such events are known to have
caused mass fish kills in the Clark Fork River.
HAZARDOUS WASTE INVENTORIES IN THE UPPER CLARK FORK BASIN
The following waste inventory includes, from the top of the
basin to the town of Deer Lodge, the major waste subsites, the area
of coverage, the volume and mass of waste, the type of waste, and
the history of each subsite.
The conversions used were as follows. One hectare is 2.47
acres. The approximate density of the mill tailings(wet) used in
the RIFR is 2.6 g/cm3, or 2700 pounds/yd3. This density is
equivalent to 1.35 tons per yd3. One pound is 0.454 kg, and one ton
is 0.908 metric tonnes.
Although the effects of the seepage from the Berkeley Pit (in
the Butte Operations Active Mine Area) and the 500 mines and shafts
(with 3000 miles of underground workings) in Butte and the vicinity
are severe and contribute greatly to contamination, we have no way
currently of predicting the time dependence of those effects.
Consequently, we have focused on estimating the mass of the known
waste subsites above ground.
As a rough estimate, the overall average mass fraction of
remaining copper in the tailings and waste materials is assumed to
be 2080 mg/kg of bulk waste material, or 0.208% by weight.
Estimates for copper content in uncovered streamside mill
tailings impoundments can be made based on the Colorado Tailings
total mass fraction of copper(1829 mg/kg). The mass fraction of
copper in fluvially mixed streamside tailings deposits is
approximately 2350 mg/kg (RIFR Summary). The estimate of average
measured solid phase tailings concentrations of copper in tailings
in the Opportunity Ponds is 2030 mg/kg of bulk waste material, or
0.203% by weight (Tetra Tech, 1985).
The largest waste holding in the Upper Clark Fork River Basin
is the collection of waste dumps and leach pads in and around the
Butte Operations Active Mine Area in Butte. The area of coverage is
1400 acres or 567 hectares. The waste mass is approximately
-------
9.08X1011 kg, and is composed of mine waste rock, mill tailings, and
heap leach tailings.
The second largest dump site in the Upper Clark Fork Basin is
the Yankee Doodle Tailings and its tailings dam, in the Butte
Operations Active Mine Area, which has 2.27xlOn kg of waste
material and impounded tailings. This subsite covers 750 acres (304
ha) .
In Butte and the vicinity, there are 150 major unreclaimed and
reclaimed waste rock dumps, totaling 350 acres (141 ha) in area.
The volume is estimated to be 9.85xl06 yd3 , equivalent to a mass
of 1.21xl010 kg, which represents the fifth largest repository for
waste in the Upper Clark Fork River basin (Camp, Dresser, and
McKee, 1987).
Butte and vicinity also include small side-stream tailings
deposits that include the Colorado Tailings (30 acres, 2.5xl05
yd3 ) and the Clark Tailings (62 acres, l.OxlO6 yd3 ).
Along Silver Bow Creek from Warm Springs to Deer Lodge, it has
been estimated that there are 4.25xl06 yd3 of waste material
(equivalent to a mass of 5.21xl09 kg. This material includes mixed
tailings, mine waste rock, natural sediments, and precipitates.
This waste has been mapped based upon the visible portions, and the
visible area is 1100 acres (445 ha).
The RIFR describes the Warm Springs Ponds as being estimated
to contain "approximately 19 million cubic yards of mixed tailings,
mine waste rock, natural sediments, and precipitates" (based on
study by Hydrometrics, 1983). This volume is equivalent to a mass
of 2.33x10" kg of waste. This mass estimate ranks the wastes in the
Warm Springs Ponds as the fourth largest waste repository in the
Upper Clark Fork River basin. The surface area that these wastes
cover is the total area of the Warm Springs Ponds plus the dried
pond beds (Pond 1 and 2) and is a minimum of 346 acres (140 ha).
The third largest mass of waste at a subsite is at the
Anaconda Smelter site. Initial estimates of the waste mass at the
Anaconda Smelter (smelter wastes) and the Opportunity Ponds place
the figure for contaminants at 1.85xl08 yd3 (equivalent to a mass
of 2.27X1011 kg) and the Opportunity Ponds surface area coverage at
6000 acres (2428 ha) (Tetra Tech, 1985). Additional Anaconda site
wastes are at the Old Works near Warm Springs Creek and at the
Anaconda Ponds.
The above listing of known subsites may be incomplete, but
will be amended if more subsites become known, such as other
subsites in and around Anaconda. The approximate total mass of
copper mining, beneficiation, and smelter waste materials is
1.41x10" kg.
10
-------
SECTION 4
OBJECTIVES
The modeling effort at the Center for Exposure Assessment
Modeling (CEAM) focused on the prediction of the frequencies of
exposure of fish to toxic metals at different concentration levels
in the stream by using metal speciation and water quality models
combined with historical data on the site. Investigations to date
have focused on the typical flood events. The main objective is to
complete a description of metals exposures and anticipated effects
on the entire river during historical periods of flooding. The
mechanisms affecting the native fish, including the exposure
time(s) and concentrations that produce mortality, will be modeled
by EPA's Environmental Research Laboratory, Duluth, Minn. (ERL-
Duluth).
11
-------
SECTION 5
MODELS AND METHODS
The general conceptual model describes sources of various
metals (species) in waste dumps and on the river banks.
Chemical partitioning between water and soil during transport
and transformation of heavy metals into the stream were analyzed in
the following manner as indicated in Figure 2. The chemistry of the
tailings deposits was used to determine the form of the metals; the
flow behavior of rain on the banks as well as overland flow
determined the principal transport mechanisms. The rates of metal
transport depend on the rate of advection predicted by PRZM (Carsel
et al., 1984), GCTRAN, and NPSOUT (Brown, 1989) and the solubility
of the oxidized metal at sulfide particle surfaces indicated by
MINTEQA2 (Brown and Allison, 1987). Surface water transport is
simulated by WASP4. The hydraulic parameters for WASP4 (Ambrose et
al. 1987) are provided by a river hydrodynamic and sediment
transport model RIVERMOD (Hosseinipour, 1988b). The toxicity
evaluations will be performed by a Fish Acute Toxicity model
developed at the ERL-Duluth.
MINTEOA2
PRZM
NlMJon, WM4
•unroll
GCTRAN
NPSOUT
luMM ion. mlilng
RIVERMOD
WASP4/TOXI4
Fish Acul«
Mortality Mod*l
Figure 2. Diagram of modeling system used
in Clark Fork River exposure assessment.
12
-------
SITE GEOLOGY AND GEOCHEMISTRY
The Superfund site has been divided into geographic subsites
for the purpose of exposure assessment modeling, with each subsite
having its own geology, hydrology, and geochemistry. The processes
that create acidic effluent from mine solid wastes are summarized
below.
The ores are blasted from the rock face, mined, and crushed.
The crushed ore is then processed by beneficiation, which is a
floatation separation that removes copper minerals from the
remainder of the ore. The tailings are wastes from beneficiation
that are predominantly wet finely crushed rock with floatation
chemicals (predominantly sulfonated pine oils) adhering to the rock
surfaces. Mechanical reduction in size of the ore particles that is
intended to aid flotation-separation (beneficiation) of concentrate
from tailings results in an increase in surface area available for
oxidation of the remaining metal sulfide minerals in the tailings.
Waste waters from the beneficiation process include sulfate,
metals, and beneficiation chemicals in solution or emulsion. The
copper minerals are smelted in an oxidative roasting process that
produces copper metal; slag is left as process waste.
The streamside tailings deposits are sand-like fine particles
of metal sulfides embedded in feldspar and quartz. The primary
metal sulfide minerals in the waste rock, overburden, and processed
ore include iron pyrite (FeS2, iron sulfide), chalcopyrite
(FeCuS2,iron-copper sulfide), realgar (AsS, arsenic sulfide),
chalcocite (Cu2S, copper sulfide), and galena (PbS, lead sulfide).
These minerals originally were geologically embedded in monzonite
porphyry formations. The metal sulfide deposits were formed by
sequential hydrothermal precipitations; sulfides were deposited
from hot geothermal solutions. As the geothermal solutions flow
from hot rock into rock fractures at lower temperatures, the
solutions drop in temperature and become supersaturated. The
deposition of sulfide occurred on the surrounding parent rock
surfaces and in rock fissures.
The sources of the copper to the creek and river are waste
copper1 sulfide-bearing particles that are corroded by exposure to
water and oxygen. The limits on metal transport and solubility will
determine the magnitude of sources of contaminants to the river.
Under'certain circumstances the partitioning of metal between waste
materials and water can be simulated with MINTEQA2. MINTEQA2 is
applied with the assumption that oxidized heavy metal is always
present (metal availability is not rate-limited). Eh is set by the
balance between oxygen diffusion and consumption, and decreases
with soil and waste material depth. pH is set by advection and
diffusion of H* away from the particle. pH is decreased by limita-
tions on diffusive and advective transport of hydrogen ions and
13
-------
this increases the diffusive driving forces for oxidized metals.
pH increases with depth: with less oxygen, less H+ is created
because of the lower oxidation rate.
Oxygen Transport
Eh is controlled by the availability of oxygen. The metal
sulfides in tailings react with oxygen and water to form metal ions
(Fe3+,Cu2+), and sulfuric acid (H2SO4) . The process of oxidation
depends on a supply of reactants (water, sulfides, oxygen) and on
water as a transport medium for reactants and products. The process
of oxidation in nature or in deposits of waste materials is limited
by transport of product or reactants; otherwise metal sulfides
would not be frequently observed in a natural or disturbed setting.
Oxygen is transported from the ground surface of the tailings
where oxygen in the water phase is in equilibrium with the
atmosphere. Oxygen diffuses into the tailing deposit on the stream
bank and is transported by diffusion and advection into each
unsaturated soil stratum where sulfides are present. Transport of
oxygen to the metal sulfide particle core depends on the rate of
diffusion through metal oxide and metal sulfate layers on a
particle surface. The rate of metal oxidation and sulfate formation
is assumed to be equivalent to the oxygen transport rate through
the soil pores (that is, the oxygen transport rate is assumed to be
the rate-limiting step).
The rate of oxygen transport into the oxidized layer and
through the sulfide particle fissures to unreacted metal sulfides
is greater than or equal to the transport away of some oxidation
products (FeO or CuO). This was concluded from the observed buildup
of metal oxides on particles in the oxidized soil stratum of the
saturated zone.
Oxidation Product Transport
As the sulfides are oxidized and byproducts (such as the
sulfates, the oxides, sulfuric acid (H2SOJ , and H+) build up, they
must be carried away in order to prevent a halt to oxidation. As
corrosion of the sulfide proceeds, the pH of the particle surface
drops and the copper at the particle surface and within particle
fractures and crevices becomes more soluble. Oxidation products are
transported away by diffusion out of the particle fissures into the
soil pore water and subsequent advection. If this did not occur,
oxidation products (H+, S042' ) would build up and hinder oxidation
kinetics and thermodynamics. Some oxidation products (H+, S042" )
will be removed faster than others (solid CuO,FeO), and so a
buildup of some products (metal oxides, carbonates, and sulfates)
occurs at the oxidation front between solid oxidized metal and
metal sulfides. The rate of accumulation of waste products at the
oxidation front is determined by the kinetics of oxidation and by
14
-------
the reaction-limiting mechanisms for transport of products and
reactants (Figure 3).
Copper diffuses into the groundwater and enters other
transport pathways. The oxidizing particles are the major source
for copper entering subsequent transport pathways such as runoff,
erosion, and leaching. Less dilution of acid occurs during dry
periods, so that the H+ ion is trapped and increases the solubility
of oxidized metal. Consequently, during dry periods, there is an
increase in the mobility of particulate and soluble metal outside
the tailings particles. During wet periods, pH at the particle
surface is raised sufficiently to reduce transport outside the
particle and reduce the source term. This can serve as an
explanation why the magnitude of the source term varies seasonally
and why the calibrated Kd may have a lower value during seasons
that are concurrent with accumulation of oxidized metal in the pore
water and on the tailings surfaces. Capillary transport of oxidized
metal and acids to the tailings surface may also result in
seasonally variable calibrated Kd and seasonally variable transport
of metals and acids. Recent surface water analytic data that
encompasses a wider range of flow and seasonal conditions should
make possible the modeling of seasonally variable heavy metal
transport.
The ground pore water around the sulfide particles has a
lower hydrogen ion concentration than the water film around the
sulfide particles, and the higher pH reduces copper solubility
relative to the water film. As the copper diffuses from the sulfide
particle surface to the bulk solution, pH increases and the copper
can form a metal oxide or carbonate precipitate in suspension in
the unsaturated zone. The pH is between 3.0 and 5.0 in the
unsaturated zone of the oxidized soil layer and a pH as low as 2.5
in the water film on the surface of the oxidizing sulfide particle.
The higher pH and lower Eh can cause precipitates to form in the
groundwater. These precipitates may also constitute a major portion
of the copper transported by leaching and underflow, and copper may
be carried in colloidal form (as precipitates in suspension) in the
groundwater of the saturated zone.
Dependence of Geochemistry on Soil Depth
In the absence of a one-dimensional model that will define the
geochemical conditions for a vertical soil core, the first step in
the analysis of the waste site is to develop a general modeling
framework. The soils of the Upper Clark Fork Basin are a poorly
sorted alluvium that comprises sand, silt, and gravel. The soil
core can be simulated with a mixture of tailings and alluvium
measuring 250 cm deep and with a water table that can vary from a
depth of 0 cm adjacent to the stream bank to a depth of 250 cm at
the edge of the tailings furthest away from the surface water
(Figure 4).
15
-------
Metal Oxide
Precipitate
lowpH
H2O
Figure 3. Oxidation of a sulfide particle.
Saturated Zone
Figure 4. Profile of the river, water table and tailing deposit.
16
-------
Each soil stratum below the surface of the stream-side
tailings deposit is geochemically defined by the oxygen consumption
and oxygen concentration in that stratum. The model for the
tailings pile geochemistry that is illustrated here corresponds to
both measurements of depth versus geochemistry in the Remedial
Investigation Final Report (RIFR) (Tuesday et al., 1987) and other
literature on mine tailings geochemistry (Tetra Tech, 1985) . Oxygen
diffusion to lower strata is limited by consumption in upper soil
strata. Diffusion is especially limited for tailings that are
compacted, fine-grained, flocculated, or that have a small void
fraction. In such situations, the limited amount of diffusing
oxygen can be totally consumed in the upper soil strata. (Figure 5)
Units of Kd:
Variable
Water Table Depth
Moles/kilogram solid
Moles/liter water
Soil Surface
100 cm
100 cm
50 cm
150cm
200 cm
250 cm
50 cm
SO cm
Oxidizing zone pH » 3 - 5 Eh-.45 Kd*86-3.3E6
-.60V Cuprous ferrlte
Intermediate pH • 5 Eh-.45V Kd • 3.3E6
Mixing Zone Cuprous (errlte
Mixing zone pH » 7 Eh*.45V Kd'1.fl5E6
Cuprous lerrlle
Reducing zone pH = 7 Eh«-.2V Kd"4.3E15
Chalcopyrlte
Bedrock and/or Original Alluvium
Figure 5. Geochemical profile of the unsaturated
tailings and alluvium. Diagram shows the geo-
chemical properties of the mixed tailings and
alluvium when they are above the saturated water
level.
17
-------
In some cases the tailings have been mixed by alluvial or
mechanical processes with the common bedrock of the Upper Clark
Fork River basin. This native bedrock has large portions of calcium
carbonate that can neutralize acid produced by oxidation and
corrosion of the sulfide particles.
The conditions in the soil and tailings mixture are
represented by two geochemical parameters: Eh (electrochemical
potential) and pH (hydrogen ion abundance). The hypothesized
vertical sequence of unsaturated tailings geochemistry is described
as follows; the oxidized stratum is the top tailings layer, below
which are, in sequence, the intermediate mixing layer, the mixing
layer, and the reduced tailings layer (Figure 4).
The top 100 cm of the soil-tailings layer has excess oxygen
available, and the Eh consequently is at an oxidizing potential of
between 0.45 and 0.80 mV. Because of the rapid rate of hydrogen ion
production, the pH in this layer is represented as 3.0-5.0.
The very top of the soil is subject to the most regular
erosion and oxidation. It may be that the top 1 to 10 cm of
tailings are completely oxidized and are not acidic. However, this
has not been the assumption of the descriptions provided here, if
pH was high (>5.0) at the surface, then a separate soil layer could
be included in PRZM with a higher value for copper Kd. A high pH
would reduce copper solubility and copper transport due to runoff.
A leached surface would also have a lower mass fraction of copper
and less copper available for transport by erosion. Changes in the
model to represent a top tailings layer that has a high pH and a
low copper content would reduce the flood-mediated transport of
copper.
If capillary-driven upward movement of copper and acid results
in the surface deposition of acid, copper, and other metals on the
tailings surfaces then the opposite of surface leaching will occur.
Evidence that this is occurring comes from the dramatic and rapid
rise in the metal content and the acidity of the river during rain
storms, floods, and rapid snowmelts. An input file for PRZM would
represent the copper-enriched top layer as having a higher copper
content and a lower Kj. In the absence of site-specific evidence
neither the enriched surface or the leached surface have been
represented in the PRZM input files.
The mixing layers are depicted with an electrochemical
oxidation potential of approximately 0.45V. The intermediate mixing
layer is between 100 and 150 cm deep and with pH=5. The mixing
layer is between 150 and 200 cm and with pH=7. In these layers, the
hydrogen ions mix with the reducing compounds in the soil core, but
because some oxidation is still occurring, Eh remains at 0.45V.
Oxidation also can occur as a result of copper sulfate oxidizing
18
-------
iron sulfide to form copper sulfide and iron sulfate. Even in the
presence of hydrogen ions from the oxidation layer above, the pH is
higher in the mixing layer because of lower reaction rates and the
neutralizing effect of the carbonates and the unoxidized sulfides.
In this simulation of native soil and tailings, the bottom
layer is assumed to be a reduced stratum, where there is a
negligible concentration of oxygen. The bottom soil layer has the
least oxygen available and considerable reserves of reducing
sulfide, so that Eh is in a reducing range (Eh=-0.2V, pH=7).
The saturated zone has little oxygen because of oxygen
consumption by reaction in the upper layers and because diffusive
transport through water is slow. Oxidative conversion of sulfides
in the saturation zone, therefore, is assumed to be small relative
to the oxidation rate in the unsaturated zone. The saturated zone
is better mixed than the unsaturated zone because of the continuous
water phase. The Eh value is between -0.2 and 0.2 V, and pH is
between 6.5 and 7.5.
Geochemistry Controls Leachate
Depending on the depth of the water table and the width of the
tailings deposits, leachate will either pass only through the
shallow oxidized layer or will penetrate more deeply into the
mixing and reducing layers. The water table level affects the
quality of the leachate entering the water table and helps
determine the form and the concentrations of metals that pass into
the stream.
To characterize the leachate that enters the water table, it
is necessary to determine solubilities in the last unsaturated zone
soil and tailings layer and the metal concentrations in the
leachate in this layer. The geochemical conditions in the last
unsaturated zone layer will determine solubility, and the depth of
the soil and tailings layer will dictate the approximate values of
the principal parameters (Eh, pH).
If the water table is adjacent to the reduced layer of soil
and tailings, the model we have proposed will predict less soluble
copper in the leachate than if the saturated layer was adjacent to
the oxidation layer (the top 100 cm of the soil core) or the mixing
layer (the middle 100-200 cm of the soil core). We do not have data
that quantifies the amount of colloidal copper transported in the
unsaturated zone or in the saturated zone.
Hydrogen ions are assumed to be neutralized in the reduced
layer by the metal sulfides and calcium carbonate present in the
alluvium, and the pH is assumed to remain at about 7. The sulfuric
acid may either be neutralized by calcium carbonate (calcareous
bedrock and alluvium) or reduced by iron pyrite and other un-
oxidized sulfides. Metal sulfates and sulfuric acid are
19
-------
precipitated and neutralized.
The water table can intercept any of the strata of the
unsaturated zone, depending on the distance from the stream, the
water table gradient, and the slope of the stream bank. At the edge
of the stream, the water table and the soil/tailing deposit surface
are the same. Away from the stream edge, the water table has a more
shallow gradient than the deposit surface slope. The water table
gets deeper and the water table intercepts lower geochemical strata
as the distance from the stream increases. A lower slope for a
stream bank has a more shallow water table (Figure 5).
The framework of a water table that intercepts the leachate
from deeper unsaturated zone layers as the distance from the stream
increases provides a means of evaluating the leachate loading from
each of the layers. To do this, the intersecting area of each layer
that is bounded by the water table needs to be evaluated. Each
intersecting area represents an interface between the saturated and
unsaturated zone that carries a leachate across the boundary. In
order to characterize the leachate at this boundary, it is
necessary to determine solubilities in the unsaturated layer and
the metal concentrations in the leachate (Figure 5).
MODELING METAL SPECIATION/TRANSPORT
The needs of this project dictate that a metal speciation
model be employed to predict metal solubility and that surface and
subsurface flow and transport models be used to predict the rate of
contaminant transport from the tailings deposit. The metal
speciation model used is MINTEQA2 (Brown and Allison, 1987).
Metal Speciation Model—MINTEOA2
MINTEQA2 solves for the thermodynamic activities of all
possible compounds in aqueous solution by solving iteratively the
combined mass balance and mass action equations. The model uses a
thermodynamic database of formation constants and reaction
stoichiometries. The mass balance equations are established by the
initial amount of each component in the initial mixture. Mass
action equations are established by the formation constants for
each metal-ligand complex. The simultaneous equations are
determined by the complexes for which data exist in the
thermodynaraic database for a set of specified chemical components.
The combined set of equations is solved using the Newton-Raphson
iteration method.
In addition to solution chemistry, MINTEQA2 predicts the
formation of precipitates; it also predicts adsorption of metals
and the formation of metal-organic complexes when the sorption and
formation constants are included in the database.
20
-------
The Pesticide Root Zone Model—PRZM
The water balance on the surface and through the unsaturated
zone includes precipitation, infiltration, runoff, and
evapotranspiration. The flow in the unsaturated zone and on the
tailing deposit surfaces may be one or more of the following:
overland flooding, infiltration, underflow, and runoff from the
slopes above the stream bank. Rainfall is the source of all four
flows. Rainfall can either flow over the surface as runoff and
cause erosion, or infiltrate and flow in the subsurface through the
open pores and reach surface water flows as interflow before
reaching groundwater table. It can cause overbank flooding during
a flash flood or sudden snow melt, and ultimately it may percolate
through the entire unsaturated zone and reach the aquifer to become
deep groundwater storage and/or the underflow (baseflow) component
of streams.
PRZM was developed to model the transport of pesticides in and
below the root zone in agricultural fields. Flow of water across
the tailings deposit surface, and through the deposit is predicted
by PRZM. The PRZM model uses the Soil Conservation Service (SCS)
Curve Number method to partition the precipitation between runoff,
infiltration, and evaporation (Carsel et al. 1984).
The predicted contaminant transport modes in PRZM include
runoff (dissolved metal carried in overland flow), erosion (metal
in solid phase carried by suspension in overland flow), and
leaching (gravity-driven unsaturated zone transport of soluble
metal). Two modes of transport not presently modeled in PRZM are
overland flood transport of metals and capillary transport of
contaminants to the surface from the subsurface. Overland flood
transport of metals is distinct from transport by overland flow
erosion, and is similar to sediment transport in streams and rivers
(see Figure 6). Capillary transport occurs when the soluble metal
is carried to the surface of the soil by upward flow when
evapotranspiration is occurring rapidly. The soluble metal collects
at the soil crust and appears as a salt with white or green
crystals of zinc and copper sulfates and carbonates.
c PRZM includes the parameter, Kd, that describes the relative
amount of contaminant in solid phase versus aqueous phase. This
parameter is called the partition coefficient and is calibrated
using stream quality data.
THE INTERFACE PROGRAMS BETWEEN PRZM AND WASP4
Calibration of PRZM was performed using data on surface water
runoff from tailings sites along Silver Bow Creek and data on
sediment erosion in Butte at the headwaters of Silver Bow Creek
21
-------
/CAPILLARY I
[TRANSPORT
Figure 6. PRZM evaluates the magnitude of
metals loading to the river for each of
the three pathways.
(Brown, 1989). Using the contaminant release time series from PRZM
program runs, daily releases of metals and eroded sediments can be
determined for each site using NPSOUT and GCTRAN. The daily loading
of copper into a surface-water transport model (WASP4) is predicted
using the geography of the subsites. Output from the NPSOUT program
is transferred to WASP4 as a non-point source loadings file.
These daily loadings into each portion of the river take into
account measured slopes, areas, and aquifer types. For each
contaminated subsite, the surface area in hectares is needed.
Geographical data are sparse for the off-stream sites. The width of
stream-side tailings were measured for the entire river course, and
the average width of the stream-side tailings was used to calculate
the area of the streamside tailings subsites. The areas of these
subsites are shown in Table 1.
The NPSOUT Program
NPSOUT is a program written for this project to predict the
transport of contaminant between the plant root zone and the
22
-------
surface water by simulating mechanisms for contaminant transport
through the near-stream saturated zone (Brown and Hosseinipour,
1989). The NPSOUT program obtains and uses output data from both
PRZM and GCTRAN. PRZM output data are used to predict the movement
of small particles (less than 100 nro diameter). The GCTRAN program
simulates large particle generation and transport in the un-
saturated and saturated zone.
Predictions of metals transport in groundwater will depend on
whether the metals are in particulate or soluble form. If the
metals are in particulate form, then the phenomena that affect
particle mobility will become important for predicting contaminant
transport. Typical pH and Eh of the saturated zone (pH about 6.5,
Eh about 0.2V) indicate that the metal will be in a colloidal
precipitate form. The transport of this precipitate should not be
retarded by the adsorption of colloid onto the soil; the precipita-
tes should have a very low adsorption rate, because of electros-
tatic stabilization.
The program NPSOUT provides metals loadings from both the
streamside deposits and other sources; it adjusts the metals source
magnitudes using various parameters. The principal parameters that
can be modified in NPSOUT and PRZM input files to calibrate the
predicted surface water concentrations to known data are the metals
concentrations at the surface of the soil and tailings mixture, K,j,
and the recharge flow to streamside aquifer volume ratio Q/V. Rates
of transport of metal through groundwater also will depend upon the
hydraulic gradients and the hydraulic conductivity of the soil
through which the metals pass on the way to the river.
TABLE 1. Waste Site Areas
LocationArea (hectares)Typeofwastes
Butte surface1049waste rock
Butte subsurface not applicable acid mine
drainage
Anaconda Smelter/ 2429 smelter waste
Opportunity Ponds
streamside 648 mixed tailings
and alluvium
The principal modes of metals transport in the saturated and
unsaturated zone are transport as colloidal particles and as
solutes. Colloids are a product of precipitation from oversaturated
solutions. It is assumed that the smaller colloidal particles are
transported by advection. One simplifying assumption is that larger
23
-------
precipitates formed in the unsaturated zone are immobilized and
only soluble compounds and small colloids are carried from the
unsaturated zone to the saturated zone; this assumption leads to
one transport model for the larger colloidal particles and another
for the smaller colloidal particles.
GCTRAN output data are used to model the transport of larger
particles (between 100 and 10,000 nm in diameter). The particle
size division is arbitrary as there is no general theory for
predicting particle size and mobility in unsaturated media during
oxidation processes at this time.
One purpose of the NPSOUT program is to provide a mixing model
for the transport of solutes and the smaller classes of colloids
(100 nm to 1 nm) assuming a particle velocity identical to the
water flow velocity in both the saturated and unsaturated zones.
The rate of movement of water and contaminants in the solute and
small colloid forms through the porous media (saturated and
unsaturated) is assumed to be the same. The overall velocity for
contaminants in the unsaturated zone may be much lower than in
water because of the confined movement of large particles.
Small particle and solute transport in the groundwater is
described with a Continuous Stirred Tank Reactor mixing model.
Smaller particles and solutes are treated as first entering a well-
mixed partial volume, V, of the aquifer and then entering the
stream with a fixed flow rate, Q. The smaller particles have an
average residence time in the aquifer that depends on the ratio
Q/V. Calculation of Q is based on the hydraulic gradient, the
hydraulic conductivity, and the stream depth. Calculation of V is
based on the width of the tailings and the depth of the aquifer.
Porous media flow is governed by Darcy's Law.
Q/V is the parameter that represents the ratio of the
horizontal flow rate in the aquifer to the effective tailings
deposit width on one side of a perennial stream(for the case of a
streamside deposit) or an intermittent stream(for the case of off-
stream waste subsites).
Flow gradient and aquifer permeability are measured or
estimated. Average groundwater velocity can be calculated using the
above parameters. Local velocities can be obtained by the solution
of groundwater flow equations subject to the prevailing boundary
conditions. The velocity and stream depth are used to estimate the
total flow Q through the aquifer.
The approximate pulse lifetime is then chosen for the subsite.
Final values are shown in Table 23.
For contaminated aquifers that are at a distance from the
stream, a time of travel parameter for travel from the mixing
volume in the aquifer to the streambank where recharge is released
by the aquifer also can be included in NPSOUT. For all following
24
-------
trials, travel time from the pulse front to the surface water has
been set to 0 days, as the tailings in all cases are immediately
adjacent to the surface water.
The contaminant flux into the saturated zone was determined
using the leachate fluxes predicted by PRZM. The fluxes entering
the groundwater are treated as entering a partial aquifer of fixed
depth and a width equal to the average stream-side tailings deposit
width (i.e. 38m). The parameter used within the program NPSOUT to
represent the rate of decline in the contaminant concentration in
the aquifer is the ratio Q/V, where Q is recharge flow per day from
a partial aquifer volume and V is the maximum partial volume of the
aquifer that is we11-mixed.
Two trial values of Q/V are 0.4 for the streamside tailings
and 0.04 for the main tailings deposits in Anaconda and Butte
(based on a lower hydraulic conductivity for the off-stream sites) .
The difference in the parameters also could be attributed to the
larger waste deposit widths, which could be much larger for the
off-stream sites in Anaconda and Butte, or to greater stream depths
for on-stream tailings. A value of 0.04 leads to a time scale of
120 days for complete emptying of the aquifer of a single 1-day
contaminant pulse, and 0.4 to a time scale of 10 days for the same
scenario. Thus the contaminant pulse lifetime does not need to
exceed 120 days. Final calibrated values for Q/V are shown in Table
23.
Values for horizontal velocity of 900 to 3650 ft/year (0.75 to
3.0m/day) have been reported (Tetra Tech, 1985) in the Opportunity
Ponds alluvial aquifer(with a hydraulic gradient of 1%). These
flows are in a northeast direction, towards the Mill-Willow Bypass
and the Upper Clark Fork River.
The GCTRAN Model
The Groundwater Colloid Transport program GCTRAN was written
for this project for use in conjunction with PRZM to account for
the special transport characteristics of particles that range from
0.1 to lo.o urn in size (large-size colloids).
GCTRAN simulates how the mobility of larger colloids will
change as the water table fluctuates and as percolating water
during floods and heavy rainstorms removes significant portions of
the available colloids generated by oxidation processes in the
unsaturated zones.
The transport of large colloids as modeled by GCTRAN assumes
that these particles are moved with a velocity in the saturated
zone that exceeds the average pore water velocity and that these
particles have a velocity of zero in the unsaturated zone.
25
-------
The excess velocity of larger colloids in the saturated zone
is based on the assumption that excluded volume effects increase
large particle velocities above pore water flow velocities.
Excluded volume refers to the portions of the pore water where the
average velocity field is small and the diameter of the pores is
small. The larger a particle is, the less chance that a particle
will pass through the smaller pores during its flow through
connected passageways in the open spaces of the soil. The large
particle only passes through the pores that have a minimum required
diameter and that have higher pore water velocities. Thus the
average flow field that the large particle is suspended in exceeds
the average flow field velocity of all the soil pores. This
increase of the particle velocity to above the average velocity of
pore water can be mathematically predicted given a known distribu-
tion of pore sizes and assumptions about the pore networks. The
differences between average velocities of particles with different
sizes in porous matrices is the basis for two commonly used
chromatographic separation techniques—size exclusion chromatog-
raphy and gel permeation chromatography. Size exclusion chromatog-
raphy has been applied to both polymers in solution and to
particles, and gel permeation chromatography is a technique applied
to the characterization of polymer mixtures.
NPSOUT treats the larger particles as if they enter the
surface water at the same time as runoff and eroded sediment, once
they enter the aquifer.
The GCTRAN program uses a number of parameters to control the
way in which large colloids leach into the aquifer. The first
parameter is K^, which predicts a copper metal concentration that
corresponds to pH and Eh within a boundary layer of still water
surrounding each oxidizing sulfide particle. The parameter PTRANS
defines the magnitude of the transport of metal ions to the water
outside the boundary layer. In the unsaturated zone outside the
boundary layer, the metal ions are assumed to precipitate and then
accumulate, when the water table rises, colloids in the risen water
are advected by the saturated water flow to the surface water. The
parameter controlling water table fluctuation is RIVRIS, a constant
that relates cumulative daily rainfall to change in the water table
level WATTAB. Daily outflow and a decrement in the water table is
assumed to be constant, so that the stream recharge outflow results
in a decrement to the water table level RECHDEL, with values
ranging from 0.01 to 0.2 cm/day. Using the above hypotheses,
transport for large particles and changes in the water table level
are predicted for the soil and tailings core.
26
-------
The Simulation in NPSOUT of Removal of Metals in the Warm Spring
ponds
The section on Warm Springs Pond metal removal efficiency
concludes that there is only a limited dependence of metal removal
efficiency in the Warm Springs Ponds on flow rate. This apparent
limited dependence may be due to the correlation of flow rate with
other unknown or unstudied causative factors, such as seasonality
and flow history. For this and other reasons discussed later, we
used a hypothesis that residence time for the metal in these
liming/precipitation ponds does not consistently predict removal,
and instead used a constant removal efficiency.
The NPSOUT program removes copper from the stream at the end
of Segment 6 of the surface water model (at the exit to Warm
Springs Pond 3) . The average measured removal efficiency (mass
removed/initial contaminant mass) of total copper (suspended and
dissolved) between the inlet to Warms Springs Pond 3 and the outlet
of Warm Springs Pond 2 is 80.4%. The approach taken for removal of
copper metal from the stream in the Warm Springs Ponds was to use
NPSOUT to remove 80.4% of all copper loadings from stream loadings
to Surface Water Segments 1 through 6.
Sediment-Mediated Transport of Contaminants in Surface Water
Sediment non-point source mass loads are made available
through PRZM output. Sediment loadings are provided to the WASP
model as a non-point source from the erosion of the stream banks
The characteristics and the quantity of sediment strong!
influences transport, removal, storage, and release of heav_
metals. The WASP4 Manual (Ambrose et al.) suggests a range for K±
coefficient values from l.OxlO6 to 6.0xl03, with the range depending
on the concentration of suspended copper.
Metal distribution and adsorption parameters and sedi-
ment/eroded soil particle size distributions are not tabulated for
the Upper Clark Fork River, but can be estimated if some
simplifying assumptions can be made. For example, Kd varies with
season, surface flow, and pH. The ratio of dissolved metal to
precipitate and adsorbed metal depends on the extent to which the
metal and surface water chemistry has reached equilibrium and on
the rate of precipitation. The following estimates predict the
ratio of suspended copper metal to dissolved copper metal in the
surface waters examined.
(1) A first type of distribution coefficient could be predicted
based on conditions in the tailings piles with the assumption that
the copper in solution remains in solution in the surface water. If
there has been no precipitation, the metals in solution are
supersaturated, and if the surface water distribution coefficient
^ remains 234 liter/kg, then the model predicts that the sediment
27
-------
transports less than 1% (with values generally below 0.25%) of the
copper transported. There are other possible values of K,,, depend-
ing on the extent of non-equilibrium.
(2) If the distribution coefficient is dependent on the pH and Eh
of the surface water a different value must be used. The steps to
calculate Kj in surface water are as follows. We initially estimate
that copper mass concentration in the suspended stream sediments is
2080 mg/kg, the same as the mine tailings described previously. The
soluble fraction can be estimated with pH and Eh, using the program
MINTEQA2; if the surface water chemistry includes a pH of 7, Eh of
0.60V, then the soluble copper in surface water equals 3.7
ug/liter. T^ is 5.5x10*5 liter/kg (2080 mg/kg divided by 3.7
ug/liter equals 5.5x10*5 liter/kg ).
(3) An alternate number for the copper solids concentration could
come from measured flood values described in the RIFR. If a typical
suspended copper concentration during a flood is 200 ug/liter, and
typical suspended solids concentration for the same flood is 15
mg/liter, then the solids concentration is 200 ug/liter
copper/15,000 ug/liter sediment, or 1.33xl07 ug/kg.
The new surface water Kj is calculated by dividing 1.33xl07
ug/kg by 3.7 ug dissolved copper/liter (for a value of 3.6x10*
liter/kg). For such a large value for the ratio of copper in solids
to copper in solution, transport by precipitates and sediments
would be the predominant copper flux. The large value is a
consequence of an apparent enrichment of copper in some suspended
river sediments of copper (predominantly in a fine particle
fraction). Enrichment of copper in some suspended sediments is
consistent with the hypothesis that the surface water carries
copper in soluble or colloidal form.
(4) A very different result is obtained if we use the measurements
• J*is.solved copper from the RIFR data sets. Ka can be estimated
with dissolved and suspended copper concentrations. Distribution of
metal between sediment and solution would vary with chemical
conditions and with the amount of suspended sediment. The method
for determining surface water K,, using only measured data can be
illustrated by the following example.
The following data has been taken from the RIFR. If suspended
solids are 20 mg/liter, and suspended copper is 0.2 mg/liter, then
the copper solids concentration is 0.2 mg/liter divided by 20
mg/liter, or 10,000 mg/kg. With dissolved copper measured as o.l
mg/liter, then K,, is 1.0x10* mg/kg divided by 0.1 mg/liter, which
equals l.OxlO5 liter/kg.
(5) Ka also may be predicted as varying when pH varies as a
function of location or time. pH is not currently predicted within
WASP, although the transport of alkalinity, hardness, pH, and
related parameters is currently under development. The software to
28
-------
model variance of Kd with time is also under development. Some
spatial variability of Ka is already in WASP.
The adsorption characteristics of the sediment may also play
an important part in the transport of metals. Use of the distribu-
tion coefficient for precipitation, and the adsorption coefficient
for adsorption, are not distinguishable methods, although there is
a conceptual difference in the two ways of thinking about metal
fate and allocation.
THE WATER QUALITY AND TRANSPORT MODELING PACKAGE
The surface water quality modeling package consists of two
separate models, RIVERMOD and WASP4, linked by an interface
program .
The River Hydraulics and Sediment Transport Model — RIVERMOD
To simulate the hydraulics of combined Silver Bow Creek/Clark
Fork River and their tributaries, a fully implicit river
hydrodynamic and sediment transport model (RIVERMOD) was adopted.
This model provided the hydraulic parameters for the water quality
model WASP4/TOXI4. In the development of RIVERMOD, the Saint Venant
equations are used for the conservation of mass and momentum. The
sediment transport module uses a sediment yield equation as well as
a sediment continuity equation in the sand size range. The
hydraulics and sediment transport equations are solved uncoupled,
that is, first the Saint Venant equations are solved for hydraulic
parameters (Q,Y,V, etc.) and then these are used for the computa-
tion of sediment yield and sediment transport from a given segment.
Details of this model are given by Hosseinipour (1988a, 1989). In
this application, the hydraulic model was modified to include time-
variant lateral inflows to better match the flow characteristics of
the stream. The sediment transport module was not activated in this
project. Natural cross-sectional parameters are provided to the
model via a least square equation that relates the flow cross-
sectional area as wetted perimeters to water depth in the forms
A = a:+a2ya3 and P = b1+b2yb3
The variables in the above equation are
A = channel flow cross-sectional area
P = wetted perimeter
alf az, a3, bx, b2, b3 are the parameters obtained
using least square procedure for each cross-section.
These are the input to the hydraulic model. A separate code
was developed to calculate these parameters for the model.
RIVERMOD can be used to generate quantitative estimates of the
bed load (mobile sediment in the stream) and may be used in the
29
-------
future to predict the suspended sediment that enters the river due
to channel and stream bed instability. The alternative to using
KENUTIL (which reformats sediment non-point source load from the
PRZM time series) is to use the interface from the hydraulics
output of RIVERMOD to place the sediment load data in a WASP4
hydraulics reference file.
The Water Quality Model—WASP4
The model readily available for predicting surface water metal
concentrations was WASP4/TOXI4. It was developed for sediment and
toxic material transport and eutrophication processes in surface
water systems, mainly estuaries and wide river basins. The package
includes sub-models for hydraulics, eutrophication and toxic
constituents. Details of the model are given by Ambrose et al.
(1987) . In this project, the creek and river from Butte to Deer
Lodge was divided into model segments for the use of WASP4.
The primary source of surface water analytical data is the
Remedial Investigation Final Report study (RIFR) on the Silver Bow
Creek (Tuesday et .al., 1987); therefore, the main stream
segmentation was chosen to coincide with the measurement stations
of that study as outlined in Table 2. Application of the simplified
surface water transport model WASP4 results in a metal
concentration time series for each surface water segment. The
surface water modeling results have been compared to data from the
RIFR and from the Montana Water Quality Bureau.
TABLE 2. WASP4 Segments Along the Clark Fork River and Silver Bow
Creek
WASP4SegmentRIFR Station and'
segment length(km) location at End of Segment
1 5.0 SS-07 below Colorado Tailings
2 7.9 SS-10 near Silver Bow
3 8.1 SS-14 near Miles Crossing
4 8.0 SS-16 at Gregson Bridge
5 9.9 SS-17 at Stewart Street Bridge
6 5.6 Exit from Warm Springs Pond 3
7 8.1 SS-29 at Perkins Lane Bridge
8 5.0 Between measurement stations
9 8.2 SS-30 near Racetrack
10 8.4 SS-31 Below Dempsey Creek
confluence
11 10.6 SS-32 at Deer Lodge
30
-------
WASP4 was used to predict the surface water quality for the
entire stream reach, with 11 segments representing different
reaches of the river. Segments 1 to 5 represent Silver Bow Creek.
Segments 6 and 7 represent the Warm Springs Ponds or the Mill-
Willow Bypass, depending on the volume of the flow. Segments 8 to
11 represent the Upper Clark Fork River.
NPSOUT converts the PRZM unit copper and sediment loading
outputs into specific loadings for each segment. The time series
from NPSOUT are reformatted as WASP4 reference files. These time-
variable and space-variable source terms help determine the surface
water contaminant concentrations.
31
-------
SECTION 6
MINTEQA2 RESULTS
The metal speciation model MINTEQA2 (Brown and Allison, 1987)
was used to predict the copper soil/water distribution coefficient
Kd for PRZM, which partitions copper between the soil-tailings
mixture and the groundwater. This model has been applied to
conditions used to describe the tailings piles to determine copper
solubility at measured or estimated pH and Eh.
The program MINTEQA2 begins with an initial set of aqueous
components and minerals not at equilibrium and then uses
equilibrium constants to predict the equilibrium composition of a
final solution and the composition of precipitates. Complete
specification of the speciation model requires the provision of the
appropriate geochemical parameters for each metal complexation and
precipitation problem. For our work, the two principal model
parameters are Eh (oxidation potential) and pH. Examples of the
range of predicted effects of pH and Eh on copper solubility at
equilibrium are shown in Table 9.
Eh and pH have been investigated in the unsaturated
groundwater at tailings deposit sites along Silver Bow Creek
(Figure 7) . There is limited data in the RIFR Supplements on Eh and
pH in unsaturated zone pore water, in the form of values were
measured at several streamside tailings deposits along Silver Bow
Creek—Ramsay Flats, the Colorado Tailings, Silver Bow, and the
Manganese Stockpile. These sites were extensively studied for the
RIFR and have been used as a surrogate for the tailings deposits
along the Clark Fork River where no data is available. Some of this
data appears contradictory and varies over a wide range, or has
been measured by unspecified or questionable experimental methods.
The lowest measured pH value of unsaturated soil pore water
was 2.21, taken at between 4 to 6 inches at the Manganese Stockpile
simulated rainfall plot (September 23, 1986). The range of pH
values for the Manganese Stockpile and the Ramsay Flats simulated
rainfall plots between 0 and 6 inches in depth are a pH of 2.21 to
4.01 (at 1-2 inches at Ramsay Flats). The measurement technique is
not described in the RIFR Supplement.
The upper measured value of unsaturated soil pH was 6.83,
taken at a depth of 6 feet into the pore water sample well site at
the Manganese Stockpile (measurement on July 23, 1986). The range
of pH values for the pore water sample well sites at the Manganese
Stockpile, Silver Bow, and the Ramsay Flats between 4 and 6 feet in
depth are a pH of 5.5 (at 4 feet, on October 30, 1986 at the
Manganese Stockpile) to a pH of 6.83.
32
-------
OJ
UJ
MANGANESE
STOCKPILE
CLARK ! V-'
TAILINGS
\
Figure 7. Vadose zone characterization at sites along Silver Bow Creek.
-------
However, for a separate set of measurements of pore water
taken at the Manganese Stockpile on April 14, 1986, the result was
anomalous pH values of 2.87 to 4.21. These were considered in error
since they were outside the range of the previously measured
values. These contradictory values and the pH values as a whole
provide a possibility to estimate pH values, but not a uniform and
certain picture of how pH varies with depth in the unsaturated zone
pore water.
These values can be compared to pH for Colorado Tailings in a
range of 2.7 to 4.0 at depths of 0.0 to 100 cm. The average pH in
fluvially mixed streamside deposits is 4.73 (see RIFR Summary).
Eh measurements were obtained with a platinum electrode
measurement. This method of measuring oxidation potential has been
judged inaccurate by some inorganic analytic chemists. They argue
that the electrochemical couple measured by the platinum electrode
is the PtO/Pt redox couple and that the potential measured depends
solely on the pH of the solution. The use of a general Eh
measurement as a general solution property has been controversial
because of the presence of multiple electrochemical couples, each
of which may define a separate oxidation potential (see Lindbergh
and Runnells, 1984). The uncertainty of the Eh values increases the
expected error in Eh and expected error in copper solubility.
However, for the sake of demonstration both MINTEQA2 and a
possible first principles approach to predicting the contamination
flowing from mine waste dumps, the dependence of Eh and pH on depth
was estimated by using extrapolations of measurements within the
unsaturated zone in tailings deposits along Silver Bow Creek.
Figure 8 shows the influence of soil depth on the median values of
pore water Eh measurements from several dates and the unsaturated
zone sampling sites. As has been described previously, the
measurements of Eh made in pore water in the tailings unsaturated
zones were made with a platinum electrode. In order to average the
values of Eh in the soil pore water, the two median values of
measurements of Eh were determined, and then averaged and plotted
in Figure 8. The average values for the Silver Bow site are 344 mv
at 4 ft (122 cm) and 60 mV at 6 ft (183 cm) . The average values for
the Ramsay Flats site are 367 mV at 4 ft (122 cm) and 64 mV at 6 ft
(183 cm). Asymptotes for Eh of 450 mV at 0 cm depth and -200 mV at
250 cm depth have been indicated in Figure 8. The median values can
be compared to one of the demonstration Eh values, 450 mV. The
comparison indicates that the demonstration values for Eh may be at
the upper extreme of the actual range of values. If the reader does
not wish to regard these values as reliable, then the soil
geochemistry profile presented may be considered as primarily
illustrative in character.
Eh has been used as a substitute for oxygen concentration in
the unsaturated groundwater. Oxygen concentration was tested as a
possible parameter in MINTEQA2, but converging numerical solutions
did not result when oxygen concentration was fixed. This may be
34
-------
Surface
o
100
cm
200
300
Expected Assymptote - 0.45 V
/
Silver Bow
X
Ramsay Flats
I Expected Assymptote • -0.2 V
I.
I
I
I
I
I
-0.1 0 0.1 0.2 0.3 0.4 0.5
V (Redox Potential)
Figure 8. Plot of tailing deposit depth
versus oxidation reduction potential (Eh).
because the H2O/HVO2 electrochemical reduction/oxidation couple,
with parameters pH and partial pressure of O2, is used by MINTEQA2
to predict Eh, and the resulting Eh is at a sufficiently high value
to make difficult the convergence of the MINTEQA2 solution.
Consequently, Eh was used directly as an indicator of oxygen
concentration.
Table 3 presents a short list of predicted electrochemical
potential values for varying pH and the partial pressure of 02
fixed at 0.21 atmospheres. These results are based upon the
H2O/HV°2 reduction/oxidation couple and the general equation
(Krauskopf, 1967).
Eh m 1.23 + 0.0295 log [ppOa]l/a[H*]a
35
-------
Table 3. Eh versus pH for a
Fixed Oxygen Content of
0.21 Atmospheres
jpH Eh (mV)
27B1070
3.0 1040
4.0 980
5.0 930
7.0 800
7.5 770
Table 4 presents predicted electrochemical potential values
for pH = 7.0 and an oxygen partial pressure that varies from 0.21
atmospheres to 0.0001 atmospheres. These results are also based
upon the H2O/HVO2 reduction/oxidation couple.
Table 4. Eh Versus Oxygen Content
for a Fixed pH of 7.0
02 partial pressure
(atmospheres)
0.21
0.002
0.0001
Eh
(mV)
800
770
750
The apparent limited variance of Eh when the partial pressure
of O2 is reduced to below its probable range for even soils below
200 cm indicates that the H2O/HVO2 reduction/oxidation couple does
not predict the behavior of reduced geochemical environments. An
optimal expected range of Eh for oxidizing conditions would be
between 450 mV and 800 mV. 450 mV and 600 mV are representative
values for Eh in some of the predictions about the solubility and
mobility of heavy metals in oxidizing conditions made using
MINTEQA2.
The soil and tailings core has been divided as described in
Table 5 to represent the idealized soil core geochemistry and
because PRZM solves for transport through multiple soil layers.
Depths of the various layers have been based on unsaturated zone
36
-------
data from the tailing deposits along Silver Bow Creek and on
sulfide/sulfate profiles at the tailing deposits at Anaconda (Tetra
Tech). Descriptions of increased sulfide content with increasing
depth in tailings and smelter waste impoundments are available in
the Tetra Tech report on the Anaconda Smelter site (Tetra Tech,
1985).
Sulfide/sulfate ratio profiles are provided in the Anaconda
Smelter report, for tailings deposits that are above the water
table. The sulfide/sulfate profiles are provided in the Anaconda
report as a guide to the soil geochemistry profile. The ratio of
sulfur in sulfide form to sulfur in sulfate form decreases as the
depth of the tailings decreases and as the age of the deposit
increases. Each range of sulfide/sulfate ratios represents a
separate soil care layer.
Sampling well sites TS-14 and TS-16 were used as
representatives of the streamside tailings unsaturated zone
geochemistries because these are the two unsaturated soil sampling
wells closest to the Silver Bow Creek at the Ramsay Flats sampling
area.
TABLE 5. Demonstration of Soil/Tailings Strata and Geochemistry
Tailings layer
Oxidation layer
Intermediate layer
Mixing layer
Reducing layer
Depth
(cm)
0-100
100-150
150-200
>200
pH
3.0-5.0
5.0
7.0
7.5
Eh
(mV)
450-800
450
450
-200
If we use the idealization that the tailings deposits have a
uniform composition and porosity, we would expect the structure of
the deposit geochemical profile to be uniform, with similar oxygen
consumption rates and similar oxygen diffusion and advection rates.
The groundwater depth should not affect oxygen transport in the
unsaturated zone.
Advection of oxygen into the unsaturated zone when saturated
groundwater levels rise during river bank storage might be an
important exception. For such a case, oxygen would enter from
below as well as from above the unsaturated zone. For the
Anaconda site, the ground water is much further from bank storage
of the river and should have much less oxygen dissolved in the
groundwater aquifer.
Iron, copper, arsenic, lead, cadmium, and zinc have been
included in excess of actual groundwater concentrations in order to
allow MINTEQA2 to calculate saturated solutions for each of these
37
-------
heavy metal contaminants. The metal transport predictions will
emphasize copper transport. The system of models can be used on
other metals such as cadmium, zinc or arsenic.
The ligands in solution in the unsaturated zone groundwater
are included in the MINTEQA2 input file, based upon the vadose zone
groundwater composition measurements. Analytic data on unsaturated
groundwater in tailings deposits was used for Ca2+, Mg2+, Na+, Cl~,
and F". Ca2+ is included in excess to assure a saturated solution.
Fluoride and chloride (F", Cl") anions are included with the soluble
fraction the same as the input fraction because there is no
fluoride or chloride precipitation.
Sulfate (SO42~) has been included at large concentrations (0.03
m/liter) because of the presence of sulfides in the tailings and
waste rock, and because the resulting concentrations are close to
actual saturated zone concentrations of 0.025 m/liter. Carbonate
(C032~) has been set using C02 partial pressure at the atmospheric
value of 3.3x10"* atmospheres in order to replicate the minimum of
the expected large carbonate concentrations from the abundance of
calcareous bedrock and alluvia in the Butte and Upper Clark Fork
River basin. An expectation of larger carbonate concentrations
would lead to the fixing of a larger C02 partial pressure -
refinements of this demonstration may show that a larger fixed CO2
partial pressure is more appropriate.
Larger organic ligands are assumed not to be present; surface
water analytical data indicate this is true, although analytical
data on soluble organic ligands are not available for the vadose
zone.
The solid phases present may be determined in advance by
reasonable guesses as to metal and sulfur oxidation states and by
the saturation indices for minerals provided at the end of our
initial MINTEQA2 model runs. Solid species that form under the
conditions in the top 100 cm include gypsum (CaS04) , goethite
(FeOOH) , hematite (Fe2O3), cuprous ferrite (Fe2Cu2OJ , anglesite
(PbSO4) , franklinite (ZnFe204), and cadmium arsenate (CdAsOA) . The
above minerals are predicted to precipitate because they are
thermodynamically optimal: nevertheless, the probability that some
of these minerals will form is small. For example, while hematite
is the thermodynamically optimal iron oxide, the probable
precipitate from a solution at ambient temperature and pressure
will be goethite or an amorphous iron oxide. Hematite is
geologically unlikely to form, although a mineral with a different
crystalline structure but the same stoichiometry may be formed.
Cupric ferrite (CuFe2OJ is not the thermodynamically favored
copper oxide precipitate for the above conditions. MINTEQA2
predicts precipitation of cuprous ferrite as the primary copper
mineral in the oxidized zone. This result is dependent not only on
38
-------
TABLE 6. Non-precipitating Component Con-
centrations in the Unsaturated Zone Pore
Water of the Tailings Deposits (at Samp-
ling Well TS-16) and Precipitating Com-
ponents in Excess of Measured Concentra-
tions
ComponentComponent concentration
before precipitation
(m/liter)
Mg2+
Na+
S042"
F'
Cl"
Ca2+
C032~
Cu2+
Zn2+
Fe2+
H3As03
Pb2+
Cd2+
1.28 X 10"2
1.09 X 10'2
0.03
1.0 X 10'*
1.3 X 10'3
9.1 X 10'3
4.9 X 10'*
0.001
0.001
0.006
0.0001
0.0001
0.0001
the geochemical conditions but on the assumptions or estimates of
K8p for the precipitates, especially for cuprous ferrite. The value
for -log (Ksp) in the MINTEQA2 thermodynamic data base of MINTEQA2
is 8.92. This estimate is experimentally determined, and other
measurements may produce different solubilities; variance in Ksp
will affect the amount of copper predicted to be in solution. For
example, if - log (Ksp) is 6.77 then [Cu2+] will be increased,
possibly by as much as two orders of magnitude.
The results indicate that many species will be highly
transportable for moderate to very acidic pH and for oxidizing Eh.
From the most mobile, as detailed in Table 7, to the least mobile,
the hazardous heavy metals are arsenic, cadmium, zinc, copper,
lead, and iron. In Table 8, the MINTEQA2-predicted soluble copper
fractions are provided, along with the calculated partition
coefficient, based upon a typical solids composition of 2.08 g
copper/kg mixed tailings.
The method for calculation of K<, is to divide the total copper
in 1 kg of soil/tailings by the mass of total soluble copper in 1
liter of water. For example, for 2.08 g/kg Cu in the solid mass of
1 kg, there is (molecular weight of Cu, g/m) 63.5 x total Cu in
solution (in moles per liter) in 1 liter of groundwater. For total
39
-------
TABLE 7. Species Formed for pH=3.0,
Eh=600mV
Concentration
Species formed (m/liter)
Cu2*
Cu+
Zn2+
Pb2*
H3As03
H3As04
Cd2+
Ca2*
Mg2+
Na*
Pe2*
Fe3+
S042"
Cl"
F"
C032"
3
6
1
1
1
1
1
7
1
1
8
1
2
1
1
1
.8
.5
.0
.5
.6
.0
.0
.3
.3
.1
.2
.1
.8
.3
.0
.2
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
10
-4
-12
-3
-5
-12
-4
-4
-3
-2
-2
-7
-7
-2
-3
-4
-5
Cu in solution of 3.8 x 10"* m/liter, the result is Kd of 2.08 g/kg
/ 2.4 x 10'3 g/liter = 86.
Table 8 and Table 9 present values for the copper Kj predicted
by MINTEQA2 for idealized representations of geochemical conditions
in mine tailings piles and stream side deposits.
TABLE 8. Calculation of K,, from Mass Fractions, for Eh = 450 mV
Core layer Copper fraction Copper fraction K^j ~~
in solid phase in liquid phase (m/kg
(m/liter) / m/liter)
Oxidation layer
Intermediate
Mixing layer
0.033
0.033
1.04X10'6
1.08X10'8
31496
3.3X106
Mixing layer 0.033 l.77x!0'10
1.85X108
40
-------
TABLE 9. Copper K^ and Dependence on pH and Eh
Soil Core LayerpHEh(v)
4.3X10"
1.85X108
5.5X105
3,300,00
8,850
31,496
86
Reducing
Mixing
Mixing
Intermediate
Mixing
Intermediate
Mixing
Oxidation
Oxidation
7.0
7.0
7.0
5.0
5.0
3.0
3.0
-0.2
0.45
0.60
0.45
0.60
0.45
0.60
For the case of copper solubility, the approximate effect of
raising pH by 1.0 is to lower solubility by xO.l, and the
approximate effect of raising Eh by 150 mV is to increase
solubility by x400.
The results shown in Table 9 indicate that, for an Eh range
from 450 mV to 600 mV, and a pH range from 3.0 to 5.0, the
oxidation layer has a range of Kj from 86.2 to 3,300,000. All of
our calibrated Ka values fall in the lower portion of this range.
A caution should be applied to the use of uncalibrated
distribution coefficients that rest solely upon equilibrium
predictions. The model presented here assumes that precipitation
occurs to the full extent that is predicted by the equilibrium
case. The kinetic limit on the rate of precipitation of goethite
(or other ferric hydroxides, both amorphous and crystalline) am
the rate of precipitation of cuprous ferrite may be of sufficien
significance to caution against reliance on K,, predictions that are
outside the ranges of calibrated values.
41
-------
SECTION 7
PRZM CALIBRATION
PRZM predictions include contaminant transport and ground
surface water flow. The flow predictions were calibrated first: the
differences between the predicted flow and the measured flow in an
experiment represented the quantity to be minimized by calibration.
The sources of data for the model application included the
Silver Bow Creek Remedial Investigation Final Report, National
Oceanic and Atmospheric Administration (NOAA) meteorological data,
and US Geological Survey reports and data on stream flow.
Meteorological data in the form of daily maximum and minimum
temperature and daily precipitation have been assembled from tapes
sent to ERL-Athens from NOAA's Climatic Data Center. Topographic
data from maps provided by Superfund site managers were used to
establish stream and river bed perimeters, slopes and slope
lengths. The surface features of the stream banks were used to make
initial estimates of the area of leaching from the unsaturated soil
and tailings strata into the saturated groundwater.
GROUND SURFACE HYDROLOGY DATA
Parameters for modeling use may be uncertain or may only be
known to be within a range of values. In order to calibrate the
water flow model (linked here with the mass-transport model), PRZM
must be run with a range of values for important parameters. To
calibrate parameters that affect predictions, empirical data and
site data should be considered.
Application of the program PRZM requires ground surface and
subsurface hydrologic data for the Upper Clark Fork River subsites.
Ground surface hydrology describes the overland flow to Silver Bow
Creek and the Clark Fork River from surrounding hillocks, stream
banks, and plains. Slope and slope length are parameters for the
Universal Soil Loss method and are incorporated into PRZM. PRZM
also requires hydrologic parameters for the SCS Curve Number method
and for water balance calculations.
Soil and waste deposit properties (such as composition and
hydrology) vary over the entire Silver Bow Creek Superfund site. In
order to represent this heterogeneity, the Silver Bow Creek site
has been partitioned into subsites, each of which requires specific
site data.
Current data are available primarily for the Silver Bow CreeK
streambank tailings deposits. Data from Silver Bow Creek have been
used to represent the entire Superfund site up to Deer Lodge. We
42
-------
assumed that the streamside deposits of mining waste above the Warm
Springs Ponds are similar to those below the Ponds.
Tailings data on Silver Bow Creek drilling sites include depth
of water table, pH, composition (clay, sand, silt, rock) and
aqueous and solid geochemistry. Drilling site data were chosen
based upon the level of soluble contamination. The 11 sites that
had the highest soluble (oxidized) copper composition in any layer
were used for tailings data. These sites are in the process of
releasing metals. The sites resemble the original high sulfide ores
at the buried depths and will have a lower sulfide content at the
surface. This demonstrates that some aging and weathering of the
mining waste has occurred.
The sites that were not chosen as representative had extensive
fresh pyritic sulfur near the waste deposit surface. Such sites are
not yet oxidized sufficiently to be modeled using a transport and
solubility-limited model. A more appropriate model for fresh
tailings and waste would be a source term that increases over time
to approach the value of the weathered waste source term (with
kinetic limitations on metal oxidation product fluxes). The
unoxidized sites are unsteady increasing sources that have lower
magnitude metals fluxes than the sites we have included.
The options for representation of water content, wilting
point, and field capacity are to have the model calculate from
soil/tailings makeup or to use correlations. The PRZM model will
calculate bulk density, wilting point, and field capacity from
composition data. Both having the program calculate the input
variables and doing the calculations by hand were tried. The method
chosen was to let the program calculate the values. In this way,
calibration to fundamental composition variables is possible.
Table 10 indicates the parameters to be calibrated, their
ranges of values, and the site characteristics represented by those
values.
CALIBRATION METHODOLOGY
For a set of deviations of predictions from a set of
measurements, the distribution of deviation magnitude depends on
parameters chosen; calibration consists of minimizing some
statistical function of this set of deviations. The most logical
function to be minimized to a uniform value is the standard
deviation for relative variations in parameters, so that the
magnitude of result variance is uniform for "the same relative
change in all parameters.
Each parameter in the input file has a functional
contribution to the predicted result. Variables with smaller
effects can be left out of the calibration process because their
43
-------
TABLE 10. Hydrologic Parameters for Water Flow and Mass
Transport to be Included in PRZM
Parameter
Value
(units)
Description
scs curve number
SCS curve number
SCS curve number
Sand content
Clay content
Grade (slope)
Slope length
USLE LS parameter
USLE K parameter
Hydraulic
drainage rate
Hydraulic
draining Drainage Rate
loam
Field capacity
Initial water content
Wilting point
58
89
90
45.1%
11.8%
2.0%
100 feet
1.7
0.42
Meadow, Type B soil
Rangeland, Type D soil
(poor condition,
impermeable soil)
Type D soil
Slope of
Tailings
Colorado
1 day
-i
1.6 day
-i
Determined by slope
and grade
Soil erodability factor,
for fine sand and <0.5%
organic matter
Free draining field,
totally drained to
field capacity in l day
Restricted
field, for clay
0.232 cm water/
cm soil
0.232 cm water/
cm soil
0.057 cm water/
cm soil
Water content at -0.33
millibar capillary
pressure
Water content at -15.0
millibar capillary
pressure
44
-------
large relative variations result in a minimized standard deviation.
Instead, what has been done is to vary parameters that will change
the prediction changes by a relatively large percentage with a
small percentage change in the variable value. For such a
parameter, standard deviation from measured values will be large
for all ranges of the large effect parameter except for the range
about the value that minimizes the difference between prediction
and the measured value. This approach calibrates the model to the
variables to which it is most sensitive.
There is a subset of model parameters that have a significant
effect on the ratio of runoff to infiltration. Other parameters are
kept constant during the calibration and as close as possible to
the known values for typical sites along the river. Application of
PRZM (Table 11) shows the parameters to which the predictions are
most sensitive and indicates the parameters that can most
efficiently minimize the difference between predicted runoff and
measured runoff.
TABLE 11. Initial Calibration Parameters Grid for 2.44-cm Storm
Precipitation
Sand
content
20%
60%
20%
60%
20%
60%
20%
60%
20%
60%
20%
60%
20%
60%
20%
60%
Clay
content
8%
8%
18%
18%
8%
8%
18%
18%
8%
8%
18%
18%
8%
8%
18%
18%
Hydraulic
drainage rate
Free drainage
Free drainage
Free Drainage
Free drainage
1.6 day'1
1.6 day'1
1.6 day'1
1.6 day'1
Free drainage
Free drainage
Free drainage
Free drainage
1.6 day'1
1.6 day'1
1.6 day"1
1.6 day"1
SCS curve
number
58
58
58
58
58
58
58
58
86
86
86
86
86
86
86
86
Infiltration
(cm)
2.44
2.44
2.44
2.44
1.71
1.652
1.646
1.76
As can be seen from this grid of parameters and the
infiltration results (Table 12), the sensitivity in the model to
these parameters is (in order of decreasing sensitivity): the SCS
curve number, the clay and sand content, and the drainage rate .
45
-------
TABLE 12. Sensitivity of PRZM
Parameters and Results
Variable
SCS CN
Clay content
Drainage
Sand content
Sensitivity3
-0.669
0.055
0.034
-0.002
aAbsolute change in infiltra-
tion/absolute change in varia-
ble
The best estimate of the parameters to which the model results
are less sensitive were used for the final calibration. For
example, the value initially used for drainage rate was 1.6 /day,
the value for clay loam, although it was later set to the Free
Drainage option of 1.0 day"1. The values used for clay and sand
content were the median values for the tailings sites chosen from
the drilled and sampled sites recorded in the Remedial
Investigation Final Report(Appendix B, Part 3) and identified in
the report chapter on site geochemistry and metal speciation data
(45.1% sand, 11.8% clay). Other general numbers for tailings
composition have been produced. The Colorado Tailings have an
approximate composition of 90% sand. The fluvially mixed streamside
deposits between Butte and Deer Lodge have an average composition
of 55% sand and 10% clay (see RIFR Summary).
RUNOFF AND INFILTRATION CALIBRATION
To determine the water infiltration parameters, PRZM results
are compared with a results from simulation of rainfall
infiltration, and runoff at two sites on the banks of Silver Bow
Creek (Figure 9) . These simulated rainfall tests were conducted at
the Ramsay Flats and the Manganese Stockpile (see Supplemental
RIFR, Appendix C).
The simulated rainfall rates applied to the test plots were
0.2 in/hr, 0.8 in/hr, and 2.0 in/hr. The simulated rainfall was
applied for 1/2 hour. The test dates were September 23 to 26,1986.
Runoff volumes from the plots were measured and infiltration
volumes were calculated as applied simulated rainfall minus runoff
volume.
The following ratios of infiltration to total precipitation
have been calculated from measured runoff volumes on the site for
46
-------
rL*TS ^f
STWJK RUNOFF S4MPLC STTS
Figure 9. Storm runoff sampling sites at Silver Bow Creek.
-------
1/2 hour tests (Table 13) . The rates of infiltration were
extrapolated to hypothetical rain storms of 4.5 hours in duration.
These estimated (from measurement) infiltration volumes during rain
storms were compared with the predictions of total infiltration
from rain storms with the same total precipitation that have been
modeled by PRZM. The volume for such a storm is included into the
meteorological data for September 1986 used as input data to a PRZM
run.
TABLE 13. Simulated Rainfall
Application rate (extrapolated rainfall)
Site 0.2 in/hr 0.8 in/hr 2.0 in/hr
(2.29 cm) (9.14 cm) (22.9 cm)
Manganese Stockpile
Infiltration ratio 0.225 0.210 0.243
Extrapolated Total 0.514 cm 1.92 cm 5.56cm
infiltration
Ramsay Flats Site
Infiltration ratio 0.238 0.251 0.113
Extrapolated Total 0.545 cm 2.30 cm 2.59 cm
infiltration
THE EFFECT OF ANTECEDENT MOISTURE
The moisture present in the soil before precipitation, also
known as the antecedent moisture condition, can affect the
infiltration predicted by the SCS Curve Number method, it is
important to know if the same plot of land was used for sequential
experiments or trials, as this would affect the antecedent moisture
conditions, and should lead to an adjustment in the meteorological
data used for the calibration.
The conditions under which the experiment was performed are
not known. To determine if the antecedent moisture condition
affected the measured runoff and should be taken into account
during calibration, PRZM was tested for sensitivity to two types of
antecedent moisture conditions. The conditions used in the PRZM
input files reflected the field conditions. For example, the PRZM
input files used the same date as the runoff experiment and local
meteorological data. The PRZM results illustrate how total
infiltration reaches a minimum as antecedent moisture is increased.
This result is consistent with PRZM's use of the SCS Curve Number
equations to predict infiltration and runoff.
48
-------
The first type of antecedent moisture condition tested is an
increase in initial soil moisture content. With a storm event of
2.29 cm, the initial soil moisture content of 0.232 cm/cm was large
enough to reduce infiltration to a minimum value, as is shown in
Table 14.
TABLE 14. Antecedent Moisture
Conditions: Fixed Initial
Water Content
Fixed initial
water content Infiltration
(cm/cm) (cm)
0.232 1.21
0.45 1.21
For the second type of antecedent moisture condition, where
regular rainfall occurs with a fixed rate for the week before the
rainstorm, the amount of infiltration also reaches a minimum as the
antecedent rainfall rate increases.
TABLE 15. Antecedent Moisture
Conditions: Fixed Rainfall
for the Previous Week
Rainfall rate Infiltration
(cm/day) (cm)
0.1
0.3
1.0
3.0
1.47
1.21
1.21
1.21
The sensitivity results from Tables 14 and 15 indicate that
the antecedent moisture condition can be set at a reasonable value
such as field capacity (0.232 cm/cm) . If there is a dry soil
condition or minimal rainfall preceding the storm event, then the
infiltration will be underestimated by PRZM. For the date examined
(September 23-26, 1986), the previous week of rainfall did not
49
-------
exceed an average of 0.1 cm/day. For this reason, the estimate of
curve number based on the PRZM calibration runs may be sensitive to
any real cases where the initial soil moisture content is less than
0.232 cm/cm.
RUNOFF CALIBRATION RESULTS
Table 16 shows that for an initial calibration to a 2.29 cm
storm rainfall in 4.5 hours (refer to Table 11), total infiltration
should range between 0.514 and 0.545 cm. An SCS curve number of 97
predicts infiltration near the two measured values of 0.514 cm and
0.545 cm infiltration.
TABLE 16. SCS Curve Number
Versus Infiltration for 2.29
cm Storm Precipitation
SCS curve
number
90
94
97
99
Infiltration
(cm)
1.297
1.00
0.59
0.28
Table 17 shows the final calibration to a storm rainfall of
9.14 cm. For a storm rainfall of 9.14 cm, the expected infiltration
(see Table 13) should vary between 1.92 and 2.30 cm. A curve
number of 90 predicts infiltration of 1.96 cm (within the expected
range).
A curve number calibrated to a storm rainfall of 9.14 cm,
should predict values of infiltration that deviate from the
measured values for the 2.29 and 22.9 cm storm volumes. For
example, the expected infiltration for 2.29 cm of rainfall is 0.514
cm to 0.693 cm but the predicted infiltration with Curve Number
equal to 90 is 1.3 cm. For a storm rainfall of 22.9 cm, the
predicted infiltration with Curve Number equal to 90 is less than
the actual infiltration of 2.59 cm to 5.56 cm.
EROSION CALIBRATION
Following the calibration of runoff and infiltration from a
typical site on the river bank, the surface erosion of soil and the
PRZM input parameters must be calibrated using the relation of the
predicted erosion to known erosion.
50
-------
TABLE 17. SCS Curve Number
Versus Infiltration for
9.14 cm Storm Precipitation
SCS curve
number
89
90
94
97
Infiltration
(cm)
2.9
1.96
1.35
0.70
Erosion data are not available for the Clark Fork River.
Instead, reference is made to a storm that is accompanied by Silver
Bow Creek tributary flow and total suspended solids data. The storm
of May 29, 1985, had 0.97 cm of precipitation. Suspended solids
content in the runoff for several tributaries to the Metro Storm
Drain and Silver Bow Creek from specific drainage watersheds above
and around the Butte mines was measured.
Table 18 provides the May 29, 1985, suspended solids data for
each measuring station. Each measuring station was placed below a
watershed with a known surface area.
The tributary flows that correspond to these erosion data can
be predicted using PRZM and the parameters calibrated above, or can
be taken from flow composite data (unpublished RIFR data). The use
of two different flow rates for the tributaries leads to two
different calibrations, but the most reliable calibration is
believed to be based on the PRZM runoff predictions.
TABLE 18. Total Suspended Solids
for May 29, 1985, Storm
Measuring
station
PS-01
PS-02
PS-04
SS-02
Total
suspended
solids
(mg/liter)
1890
1440
2650
272
Eroded
mass
(tonnes/
hectare)
0.0266
0.0203
0.0373
0.0038
51
-------
The erosion parameters varied for the calibration are the
Universal Soil Loss Equation Length of Slope parameter (USLE LS)
and the Universal Soil Loss Equation erodability parameter (USLE
K). The SCS method management practice parameters P and C do not
apply, because the sites are uncultivated. The USLE K parameter
ranges from 0.05 (sand,<.5% organic content) to 0.38 (loam,<.5%
organic content). The USLE LS parameter ranges from 0.07 (0.5%
slope, 25 ft length) to 0.18 (1% slope, 300 ft length).
EROSION CALIBRATION USING FLOW BASED ON RUNOFF PARAMETERS
To use the measured total suspended solids we must determine
how many liters of runoff occurred for that day. Using PRZM
variables from the previous calibration of runoff, runoff per
hectare at 0.141 cm for May 29, 1985, is predicted by PRZM using
1984-1985 meteorological data. This volume is 1.41x10" cm3/hectare.
By conversion, this value is equivalent to 0.447 ft3/sec for
a watershed area of 77.8 hectares. The measured area of the
watershed for PS-04 is included in Table 19 and was measured from
a map using a planimeter. For 2650 mg/liter of total suspended
sediment, conversion to total eroded sediment for the 0.447 ft3/sec
flow from 77.8 hectares shows a predicted total eroded mass of 2.9
tonnes or 0.0373 tonnes/hectare for this one day (see Table 18).
The ratio of erosion per hectare to total suspended solids for
May 29, 1985 with fixed runoff versus infiltration parameters is
0.0373 tonnes/hectare for 2650 per mg/liter of suspended solids
(see Table 18).
TABLE 19. Erosion Parameter Calibration
Grid for 0.97-cm Storm Precipitation
Erosion
USLEK USLELS (tonnes/hectare)
0.38
0.38
0.05
0.25
0.24
0.12
0.18
0.18
0.07
0.07
0.18
0.10
0.07
0.09
0.0373
0.0353
0.0046
0.0601
0.0318
0.0111
0.0215
The measured value of eroded sediment per liter to be used is
an average of the two central values for total suspended solids
1440 and 1890 mg/liter, and eroded sediment per day of between
52
-------
0.0203 and 0.0266 tonnes/hectare. An average rate of suspended
solids eroded from the land surface for this event is 0.0235
tonnes/hectare-day.
The calibration of the erosion parameters USLEK and USLELS is
as follows. The best fit value for USLEK (0.18) represents half
loamy sand and half loamy fine sand. The value for USLELS (0.09)
represents a slope of 0.5% and slope length of 75 ft. An erosion
rate of 0.0215 tonnes/hectare results for the calibration date, May
29, 1985.
CALIBRATION OF EROSION BASED UPON FLOW COMPOSITE DATA
Runoff volumes from the eroded Butte sites can be determined
from either the PRZM model as was done above or by the Flow
Composite method followed by the authors of the Remedial
Investigation Final Report. Flow/surface runoff data for May 29,
1985 were gathered by the Flow Composite method. Flow composite
data uses total flows during the 90-minute storm hydrograph, and
then calculates the average flow during the storm.
Runoff in Table 20 was derived from conversion of acre-ft to
hectare-cm using the watershed areas calculated with a planimeter
from the map of the Silver Bow Creek Watershed that accompanied
the erosion study section of the Remedial Investigation Final
Report. The average value of the two central runoff depths
for the storm is 0.231 cm. This value can be compared with 0.141 cm
of runoff for May 29, 1985, as predicted by PRZM.
TABLE 20. Flow Composite Data for May 29, 1985,
Storm
Measuring
station
for
watershed
PS-01
PS-02
PS-04
SS-02
Surface
area
(hectares)
31.2
77.5
77.8
87.7
Flow
(acre-ft)
0.25
2.28
3.96
0.38
Runoff
(cm/hectare)
0.099
0.363
0.628
0.0535
More runoff coincides with more erosion because the basis of
our erosion estimates is the data available on suspended sediment
per liter. The average measurement of erosion per hectare (from a
53
-------
PRZM-generated runoff of 0.141 cm) at 0.0235 tonnes/hectare can be
compared with an erosion per hectare (for the Flow Composite
measure runoff data of 0.231 cm) at 0.0385 tonnes/hectare.
The Flow Composite data predict more erosive parameters than
the PRZM prediction runoff data. The Flow Composite data fit best
to erosion parameters that have the following values. The best fit
value for USLEK (0.38) represents loam. The value for USLELS (0.18)
represents a slope of 1% and slope length of 300 ft. These
erosive parameters predict more erosion (0.0373 tonnes/hectare on
the calibration date, May 29, 1985) than the values calibrated to
the PRZM runoff results. Nonetheless, the erosion parameters
calibrated to PRZM runoff have been used.
54
-------
SECTION 8
MINING WASTE MASS ESTIMATES
The variability of the off-stream sources over a year implies
that there are both seasonal sources and permanent sources. The
seasonal sources include the off-stream subsites in Butte and at
the Anaconda Smelter (see Table 1). The sub-surface drainages from
underground mine networks and mine pits in Butte constitute a
permanent set of drainage sources that are represented in NPSOUT as
having no seasonal variance.
PHYSICAL STRUCTURE OF THE BUTTE MINES DUMP SITES
The dump sites in the Butte vicinity total 2592 acres or 1049
hectares, based upon the estimates available from the soils
screening study work plan. The adjustment factor in NPSOUT used to
correct the area of this tailing deposits site is B(for Butte) . The
adjustment factor B can be considered a correction for data errors
in the area estimates. B also can be considered an adjustment for
differences in the contaminant source characteristics of the off-
river waste deposits versus the on-river streamside tailings
deposits.
The Butte subsite is primarily composed of 1750 acres of waste
rock dumps and 750 acres of the Yankee Doodle tailings. If this
Butte subsite area estimate is correct, then the adjustment factor
B should be 1.0. Future work in application of this model should
include examination of the physical structure of these dumps and
evaluation of the dump site areas using aerial photographs. The
final value of B settled on and used is 1.0 .
PHYSICAL STRUCTURE OF THE ANACONDA DUMP SUBSITES
In the Tetra Tech RIFR for Anaconda, the Anaconda Opportunity
Ponds are estimated to be 6000 acres (2429 hectares) in area and to
have an estimated waste mass of 2.5xl08 tons.
The use of this area in predicting the magnitude of the off-
river metal source loadings requires an adjustment parameter in
NPSOUT, A(for Anaconda), and a set of peak loadings from the off-
river sources.
To have the correct proportionality between the metals
entering the Clark Fork from Warm Springs Creek (considered to be
the entire loading from Anaconda) and metals from the Opportunity
Ponds seepage drains, German Gulch, Brown's Gulch, and Mill-Willow
55
-------
Creek, the metal fluxes for May 7, 1985, (a peak Cu loading date in
our data) were used (Table 21). The correction factor A is applied
to 6284 kg/day of peak loadings from the Anaconda subsite, and
corrects the subsite area to 1500 hectares when A=0.24. Using
calibration to determine the magnitude of A is an important step.
Future work will include examination of the physical structure of
these dumps and evaluation of area from aerial photographs. The
final calibrated value of A used is 0.24 .
TABLE 21. Peak Copper Fluxes into the Clark Fork Basin
May 7, 1985, RIFR Measurement
Source
Station TOXI4 Load
Dump site
Brown's Gulch
German Gulch
Opportunity Ponds
SS-12 145
SS-15 170
SS-23,24 46
Mill-Willow Creek SS-18 1401
Warm Springs Creek SS-28 6284
tailings
tailings
treatment
ponds
tailings
and seepages
Anaconda
smelter waste
MINE DRAINAGE FROM THE BUTTE MINES
One important source of Cu into Silver Bow Creek above the
RIFR Measuring Station SS-07 is an underground mine-shaft network
under the Butte area that drains at an approximately constant rate
into the Metro Storm Drain and other storm drains and empties
directly into Silver Bow Creek. This source plus variable sources
on the same site result in a 2800 kg/yr to 4100 kg/yr loading of Cu
into Silver Bow Creek. The mine drainage carries copper into
Silver Bow Creek at an estimated rate of 2373 kg/yr (6.5 kg/day) .
PHYSICAL DIMENSIONS OF THE STREAMSIDE TAILINGS DEPOSITS
The area covered by the tailings along the banks of the Clark
Fork River has been determined using a set of maps prepared from
aerial photographs that have been converted to show vegetation type
and extent of vegetation cover from the start of the river at the
confluence of Silver Bow Creek and Warm Springs Creek down to the
town of Deer Lodge. The width of the tailings deposits, as shown on
the maps (the Vegetation Maps 7,8,9, and 10 from Appendix D, RIFR) f
was measured at 6-cro map intervals. The tailings may extend into
veaetated areas; however, no vegetated areas were included in this
preliminary determination of the tailings deposits area.
For the four maps, the average measured visible width of the
tailings deposits is 38.8 meters on each side of the river, or 77.?
56
-------
square meters of deposits per meter of river course. This area will
be used for both calculations of the area of intersection of the
saturated zone and the oxidation (and unsaturated) zone, and for
the area from which erosion and runoff are predicted to occur by
PRZM.
C and D represent, respectively, the adjustments to the
measured surface area covered by the streamside tailings above the
Warm Springs Ponds (along Silver Bow Creek) and adjustments to
measured area below the Warm Springs Ponds (along the Upper Clark
Fork River). Both C and D have a value of 1.0 if the measured
surface area is used by the program NPSOUT in the contaminant
loading calculations. Current calibrations of these two factors
settle on a value of 1.0 for C and a value of 4.0 for D. These
values indicate that initial measurements of tailings surface area
may have not accounted for all the contamination along the Upper
Clark Fork River below the Warm Springs Ponds.
57
-------
SECTION 9
REMOVAL OF METALS IN THE WARM SPRING PONDS IN NPSOUT
To complete the model of metals transport from Silver Bow
Creek to the Clark Fork River, a predictive description must be
provided for the metal removal efficiency of the settling ponds
(Warm Springs Ponds 3 and 2). The model will simulate reduction in
metal fluxes at the segment juncture between the Silver Bow Creek
and the Clark Fork River.
The NPSOUT program models the removal of copper in the stream
at the end of Segment 6 of the surface water model (at the outlet
to Warm Springs Pond 3). The approach taken with NPSOUT to account
for removal of copper metal from the stream in the Warm Springs
Ponds was to remove 80.4% of all copper loadings from sources to
surface water segments 1,2,3,4,5, and 6.
The data shown in Table 22 are from the State of Montana
(Ingman, 1986). The measured removal efficiency (mass
removed/initial contaminant mass) of total copper (suspended and
dissolved) between the station just above Warms Springs Pond 3
(Montana Water Quality Bureau Station 03) and the outlet of Warm
Springs Pond 2 (Montana Water Quality Bureau Station 04) is 80.4%
on average.
TABLE 22. Removal of Copper in Suspension and in
Solution by Precipitation and Flocculation in
the Warm Springs Ponds
Date of pH Rate of Flow Fraction Cu removed
measurement (ft3/sec) (% of mass flux)
10/28/1985
12/11/1985
1/6/1986
2/4/1986
2/25/1986
3/10/1986
4/7/1986
4/21/1986
5/5/1986
5/19/1986
6/2/1986
6/16/1986
Average Total
Median Total
8.3
7.9
7.9
8.0
7.8
7.8
8.3
8.9
8.4
9.0
7.8
8.5
Cu Removal ;
Cu Removal:
43.7
30.5
33.0
41.3
200.0
69.9
78.5
73.7
93.9
93.9
153.3
64.3
»
»
87.4
81.5
74.6
79.1
94.0
64.8
83.3
88.9
94.5
79.1
54.3
83.8
80.4%
82.4%
58
-------
The mechanism proposed for metal removal has been the
combination of precipitation (with adsorption) and flocculation.
Precipitation or adsorption of Cu and Zn should definitely occur as
pH rises, as has been shown in the MINTEQA2 results, but the
flocculation mechanism is not well-defined. Arsenic removal has a
negative coefficient of correlation with pH, and this is possibly
due to inhibition by OH" of arsenate anion (As03~) adsorption or
precipitation.
We can consider two cases. In the first case, little
flocculation is occurring and this is not a dominant removal
mechanism. Then metals are removed by precipitation and the
settling of unflocculated precipitates. If settling is occurring at
a regular velocity (for a smaller particle size, Vz or settling
velocity is smaller), then more settling and a higher removal
efficiency should occur for longer residence times in the ponds.
For the second case, flocculation is occurring, and is caused
by a mechanism such as the counteracting (neutralization) of FeOOH
acidic surface charges with Ca(OH)2 (lime), resulting in
electrostatic repulsive forces being reduced and particle-particle
combination occurring after interparticle collision and as a result
of Van der Waal attractions between particles. Final removal rates
would increase with increasing settling velocities and residence
time in this case as well.
Regression analysis was performed on the above data set with
both flow and pH independent variables (Figures 10 and 11) . The
results indicate little or no dependence of removal efficiency on
flow. For instance, the coefficient of regression is positive for
copper and negative for zinc and arsenic (indicating less zinc and
arsenic removed with lower flow rates). In addition, the increase
in R2 with the inclusion of flow as a dependent variable is,
respectively, 0.004, 0.004, and 0.21 for Cu, Zn, and As (out of a
possible 0.879, 0.772, 0.992). These are small increases in
explained variability.
R2 is the ratio of ssR to ssT. R2 is also the ratio the sum of
squares (explained variability) explained by the use of the
variable to the total sum of squares (total variability).
The R2 terms for regression of efficiency versus pH for Cu,
Zn, As are respectively, 0.121, 0.228, 0.008. This leads to the
deduction that some of the variability in pond efficiency can be
explained with pH. However, the maximum R2 value is 1.0, so much of
the variability is left unexplained.
The seasonal determinants of pond efficiencies and the
dependence of pond metal removal efficiency on flow and pH are not
understood well enough to incorporate efficiency predictions into
the metals loading model. There is a correlation of pond efficiency
59
-------
M
3
C
M
C/l
3
«J
§
u
f 0
5\
U
J
-10
-20
-30
-40-
-50
-60;
-70
-80-
L
-90
j
1 00
L
A
,
A
A
A
A.
a
A
A A
n
A
LJ
X
S x
"an s
X
X
Coppsr
a
Zinc
j^
Arsenic
1
7.8 8 8.2 8.4 8.6 8.8 9 9.2
PH
Figure 10. pH versus pond metal removal efficiency
for flow through pond #3 and pond #2.
with pH (efficiency increases with pH for Cu and Zn and decreases
with pH for As) (Figure 10) . We can conclude that pond metal
removal efficiency does depend on pH, so if pH can be predicted
then values for pond metal removal efficiency may be improved.
A
U
o-
-10
-20-
-30-
-50-
-60
-70
-80-
-90-
_inn
A
A
A
A
A
a
A
A A i
^ a
A
X X
s x n g i
5 a x a
s ;
s
Copper
a
Zinc
A
Arsenic
i
L
20 40 60 80 100 120 140 160 180 200
Discliarge Flow from Pond#2 (cfs)
Figure 11. Flow through versus metal removal for
flow through pond #3 and pond #2.
60
-------
Since these data do not include a known flood peak, it may be
that there are unidentified reductions in metals removal
efficiencies at very high flow rates. Montana Water Quality Bureau
Data from late 1986 and 1987 and RIFR data may show that strong
correlations do exist and that the above results are anomalous.
Freezing-over of the ponds may lead to drastically reduced
pond metal removal efficiencies due to short-circuiting of flows,
although evidence that this occurs is as yet only anecdotal (Gary
Ingman, personal communication).
The variables not included in the analysis above are
seasonality, climatic history, and flow history. We believe that,
in combination with these variables, it is likely that there is
dependence of removal rate efficiency on flow rate and that
residence time for the metal in the liming/precipitation ponds
affects removal efficiency. The reason that there should be a
correlation of flow to metal removal efficiency is that time of
residence should be lower for a higher flow rate and flocculation
of metal should occur at a lower rate, if the pond volumes do not
vary significantly with flow rates or season.
61
-------
SECTION 10
HYDROLOGIC AND HYDRAULIC STUDIES
This section summarizes the available hydrologic and hydraulic
information on the site and describes the hydraulic modeling effort
on the Silver Bow Creek-Upper Clark Fork River.
HYDROLOGY OF THE AREA
The area covering the Silver Bow Creek watershed has a
continental climate, that is characterized by short, cool, dry
summers and long cold winters. Annual precipitation may vary from
6 to 20 inches with an average of approximately 12 inches. Climate
in the upper Clark Fork Valley is about the same and the average
precipitation is 13.2 inches. In general 50% of precipitation falls
in the late spring and early summer, which if augmented with
snowmelt can cause severe flooding. The majority of the annual
stream flow comes from melting of the snow packs. Streams can
easily become partially or fully ice covered during winter season.
Although maximum flow in Silver Bow Creek and Upper Clark Fork
River occurs in late spring and early summer, on smaller sub-
watersheds maximum flow may occur in winter by rainfall on frozen
ground. Cloud bursts and thunder storms may occur in late the
summer season.
Hydrologic investigation of the area (CH2M-Hill, 1988) has
shown that the Probable Maximum Flood (PMF) peak at the diversion
into Warm Springs Pond 3 is about 117,200 ft3/sec with a total
volume of 191,000 acre-feet. Also flow entering ponds in excess of
5600 ft3/sec would cause the overtopping of Warm Springs Pond 3,
floods in excess of 7000 ft3/sec would overtop Warm Springs Pond 2,
and flows in excess of 7500 ft3/sec would overtop all three ponds.
As a result of overtopping, the pond embankments would fail due to
erosion and subsequent gullying. In addition, sediment volumes
delivered to the ponds is estimated to be 25,000, 50,000, and
100,000 cubic yards for 10-, 25-, and 100-year floods,
respectively. Therefore, as time goes on, the capacity of the ponds
to hold water decreases and the probability of failure due to
floods increases. The above reference provides a table of historic
floods in the area as well as details of sub-watersheds hydrology
and peak flows for tributaries.
HYDRAULICS AND FLOW MODELING
The main path of Silver Bow Creek and the Upper Clark Fork
River can best be described as meandering with top width varying
from 10 to 50 ft. The cross sections are close to rectangular in
the main channel with deeper points near one of the banks. The
flood plains, which could be very wide at some locations, often are
62
-------
covered with small bushes, trees or grasses. The bottom materials
are mostly sand, cobbles, and boulders that result in high
roughness coefficients. The stream has a high bottom slope (on the
average 0.42%) and, consequently, its velocity is high even during
normal flows. Many perennial and ephemeral tributaries feed the
stream along its course from Butte to Deer Lodge making it a
gaining stream. In the top reaches of Silver Bow Creek above the
confluence with Blacktail Creek, the stream may go dry during mid
to late summer but the flow may still exist under the immediate
stream bottom layer in the form of underflow through the alluvial
aquifer that underlies the valley floor. Because of the high slope
and the consequent transport of finer particles even during the low
flow periods, the bottom materials are coarse except on the flood
plains and in the overbank regions where finer materials may be
found. Flow is seldom uniform or laminar and, for all practical
purposes, it can assumed as turbulent everywhere and at all times.
Many structures such as bridges, diversions, etc., which can alter
the flow regime are built in the stream path. The major structures
are the Warm Spring Ponds and their control facilities. These
structures are primarily built to treat the Silver Bow Creek waters
before entering the Upper Clark Fork River, although they could
serve as a flow regulator during flood periods. Flows in excess of
the ponds' control structure capacities are not treated and are
bypassed through Mill Willow Creek extensions.
As this brief introduction shows, the hydraulics of the Silver
Bow Creek-Clark Fork River combination is complex and the
hydrodynamic modeling effort requires a good feel for the area and
some innovation to adapt standard river models to the problem. This
difficulty is compounded further by the lack of adequate
information. In the following discussion the details of hydraulic
modeling are explained.
HYDRAULIC MODELING
The model applied in this project in an extended and modified
version of the unsteady river hydrodynamic model developed by Amein
and Fang (1970) and later updated and used by Fread and Smith
(1978) as well as others. In our application, the model is modified
to transport sediment in the sand size range and to accommodate
time-variant lateral inflows as well as interfacing with the WASP4
water quality modeling package.
The model uses the Saint Venant equations for fluid flow, and
sediment yield and sediment continuity equations for particulate
transport. The sediment yield equation used in the model is the
Yang's total load formula. Two forms of the sediment continuity
equations are used to account for scour/deposition and bed level
variations. The Four Point Implicit numerical scheme which is fully
stable, efficient and accurate is adopted for the time integration.
Nonlinearity is treated by the Newton-Raphson iterative technique.
63
-------
The flow equations are solved simultaneously for depth and
discharge at all cross sections. From these parameters flow
velocity, channel width and hydraulic radius are calculated to be
used for the computation of bed shear velocity and velocity for
incipient motion. These variables and the particle fall velocity
then are used in the Yang's equation to find the sediment flux at
a given cross-section. The sediment flux is used in the sediment
continuity equation in combination with lateral sediment inflow, to
calculate the sediment cross-sectional change from which the
determination of scour and deposition is made. Although the
sediment module was not activated in this application, it has been
used in other projects (Hosseinipour, 1989).
MODEL EQUATIONS
The river flow and sediment transport model described herein
uses two sets of equations. The first set is the Saint Venant
equations for flow and the second set consists of the sediment
yield and sediment transport continuity equations for the
longitudinal transport and bed level changes.
FLOW EQUATIONS
The one-dimensional equations of unsteady state f^ow used in
the model consist of the equations for the conservation of mass and
momentum. Derivation of the various forms of the equations are
given by Chow (1959) and Henderson (1966). Expansion of the
equations into the Finite Difference forms and implicit time
integration scheme is described in detail by Amein and Chu (1975)
and Hosseinipour (1989).
SEDIMENT TRANSPORT AND SEDIMENT YIELD EQUATIONS
The sediment yield equation used in the modification of the
hydrodynamic model is the sand transport equation proposed by Yang
(1973). The transport of the sediments in the stream is simulated
by incorporating the sediment continuity equation into the flow
model. The flow equations and the sediment transport equation are
then solved for the model network uncoupled.
Theoretical development of these equations are given by Garde
and Ranga Raju (1985), Yalin (1977) and others. Application of the
equations are reported by Chang (1982) and by Chang and Hill
(1976). These equations account for sediment cross sectional
variation with time and sediment yield variations with distance
along the stream. Figure 12 below depicts a typical erodible
channel cross section and the associated sediment profile.
64
-------
Figure 12. Cross-section of a typical erodible channel.
DATA REQUIREMENTS
The model described above requires two kinds of data:
hydraulics and sediments information. The hydraulic data include
channel morphometry, bed elevations, and initial and boundary
conditions. The sediment data include sediment median size,
sediment porosity, specific weight, water viscosity and channel
roughness. If cross-sectional topography data are available then a
separate sub-model can be used to generate exponential rating
functions for cross sections and wetted perimeters as a function of
depth in the form of A=c+a2 • yb2 and P=d+ai ybl for natural
streams. The sub-model uses the topographical coordinates tc
generate these relationships for every cross section. The variables
are defined as: A is the cross sectional area, y is the depth at
the given cross section, and the rest are constants calculated by
the sub-model. The model then uses these relationships to calculate
automatically the area and wetted perimeter as the water depth
changes. This feature allows the model to use natural cross
sections and therefore simulations are closer to the natural
behavior of the stream.
65
-------
SILVER BOW CREEK-CLARK FORK RIVER APPLICATION OF THE MODEL
The model was applied on the 52.6 mile stretch of the Silver
Bow Creek-Clark Fork River from Butte down to Deer Lodge. Since the
hydraulic characteristics of the Warm Spring Ponds is very
different than the stream channel, the flow modeling was done
through the by-pass around the ponds. Various segmentations and
cross sections were used in the initial trial runs to check the
adaptability of the model to the problem. The final runs were made
using the natural cross sectional topography from the maps of the
hydrologic investigations. According to these maps, which were also
verified on a site visit later, the main channels are from 10 to 50
feet wide and are roughly rectangular in shape. The model was run
with 30 segments initially and later the number of segmentations
were reduced to 11 to increase the efficiency as well as to match
the stations where water quality parameters were available. No
significant discrepancy in the hydraulic modeling results were
observed as a result of reduction in the number of segmentations.
The report on the hydrological investigations provided the
Manning's roughness coefficient in the range of 0.045 to 0.055 for
the main channel and 0.075 to 0.15 for over bank flooding areas.
These figures were used to calculate the composite roughness
coefficient for the high flow scenarios in the modeling effort. The
formula used to calculate the composite roughness was chosen from
the Chow's book which has the form
N
n = [Z Pi nVP]1/2
1
where p£ is the portion of the perimeter with roughness value of nif
P is the total perimeter and n is the composite roughness.
The USGS data files on the streams provided three useful
discharge measurements on the Silver Bow Creek, Warm Spring Creek,
and Clark Fork River at Deer Lodge that could be used in modeling
studies. Five scenarios were chosen in flow modeling which
represented the different flow situations throughout the year in
the streams. These included normal winter and late summer low
flows, late spring and early summer high flows, and flash floods
due to cloud bursts and thunderstorms or sudden snow melts. The
boundary conditions were assigned as upstream discharge at Butte,
estimated time variant lateral inflows, and downstream stage at
Deer Lodge. Since only discharge measurements were provided at Deer
Lodge, the Manning equation was used to estimate the stage
corresponding to the given discharge. Further, since great
variations in the discharges at the three gauging stations were
present, the time variant-tributary inflows were adjusted to match
the total flow at Deer Lodge. The incomplete data available on some
other stations along the stream provided an insight in the
adjustments of tributaries flows.
66
-------
SECTION 11
RESULTS
SURFACE WATER MODELING RESULTS
The in-stream copper concentration predictions for the Upper
Clark Fork River basin have been compared to measured surface water
concentrations. The comparison has been examined, as has the
modeling system and the assumptions used, for observations about
the usefulness of our approach.
Each subsite has a final calibrated value for K,, and for Q/V.
Other parameters and their values at subsites are shown in Table
23. These include values for time of travel to open water from the
sub-tailings aquifer and the lifetime of the contaminant pulse.
Q/V is the parameter that represents the ratio of the
horizontal velocity in the aquifer to the effective tailings
deposit width on one side of a perennial stream(for the case of a
streamside deposit) or an intermittent stream(for the case of off-
stream waste subsites).
TABLE 23. Superfund Subsite NPSOUT Final Parameters
Lifetime of
contaminant
Superfund pulse Q/V
subsite K
-------
RESULTS OF TRIBUTARY AND MAIN DUMP SITES
The surface water modeling results of this study are
summarized in the following manner. First, the copper
concentrations in Silver Bow Creek, the Mill-Willow Bypass, and
Warm Springs Creek are compared with the predicted metal loadings
from the off-stream subsites. The subsites include the Butte mine
drainages and all other subsites for segment 1 (including the
Yankee Doodle Tailings, the waste rock dumps, and the Butte Active
Mine Area, with active mining and milling sites) and the Anaconda
Smelter with the other subsites for segment 7 (including the slag
piles and the Oppportunity Ponds).
The contaminants at the end of segment 1 are taken as a
measure of the Butte source terms, such as the principal Butte
waste dump seepages and drainages. The principal drainages and
seepages from the Anaconda smelter area (Mill-willow Bypass
streamside tailings, Opportunity Ponds seepages, Mill-Willow Creek
basin, Warm Springs Creek basin) enter segment 7. The measurements
of loadings into segments l and 7 are used as reference data for
the calibration of the NPSOUT model for contaminant sources in
Butte and Anaconda. During high flows, these off-stream source
terms may not describe the entire metal flux observe^ in the stream
but they should represent 50% to 100% of the copper loadings during
medium and low flows.
Predictions of copper flux from the tributary sites (Mill-
Willow Bypass, Warm Springs Ponds) were performed with NPSOUT, and
the results are compared with a selection of 15 measurements that
date from 1984 through 1986 (Ingman, 1987) . This comparison is made
graphically by plotting loading estimates from model runs versus
tributary loading data (Brown, 1989).
The first two sets of comparisons (Figures 13 through 16) show
how increasing Kd reduces the predicted metal fluxes and improves
the resemblance between model predictions and the measured data for
segment 1 and segment 7.
Figures 16 and 17 demonstrate the results of a change in the
Q/V ratio parameter (variable TQDIW in the model) for the off-
stream subsite sites and an improved resemblance to the calibration
data for Segment 7. This change is equivalent to a large increase
in the rate of flow from the aquifer. Contaminant pulse lifetime
is kept at 120 days for all calibrations of the off-stream
subsites. Travel time of the contaminant pulse front to surface
water has been kept at 0 days.
68
-------
s
o
k
a
o
o a
Segment 1
Kd = 234, TQDIVV =0.04
D
Conrparison Oates^ 1904-1986
Figure 13. Measured versus predicted copper
loadings, segment 1.
400
b
i^
v_/
§.
T3
(O
O
per
rSegfrent "V
Kd = 234, TQDIVV =0.04
a
3 a D a D
^,TT???ifTT
T" 1
Comparison Dates, 1984-1985
Figure 14. Measured versus predicted copper
loadings, segment 7.
69
-------
Segment 1
Kd = 328, TQOIW = 0.04
8
v
50
ag
x X x
S B
x x x
Comparison Dates, 1984-1986
Figure 15. Measured versus predicted copper
loadings, segment 1.
Segment 7
328, TQOIW =0.04
« 10°
S1
1
S-
n
a
fi Q o a o D n
X W • •! m ^ XvTVy^K
lirtTTTlTTTTTiT
X
ItoasacKl
a
fr«iict«ii
{* e t; * *m ;* ;*!
Conpartson Dates, 1984-1986
Figure 16. Measured versus predicted copper
loadings, segment 7.
70
-------
Segment 7
Kd = 328, TQDIW = 0.12
D
Z
D
a
Y Y T Y T T f i T T 1
a
X
, D
3 aa TT n TT
r T T T r~
CompnrSson Dates, 1984-1986
Figure 17. Measured versus predicted copper
loadings, segment 7.
The type of groundwater transport and mixing model used
substantially controls the prediction of copper loadings into the
surface water. Other possible changes to the groundwater mixing
model would be to include more than one mixing volume. The
efficiency of mixing in the aquifer also might be changed in order
to improve the model. The Q/V parameter changes the rate of flow
represented but not the efficiency of mixing or the number of
mixing volumes in the aquifer.
The PRZM model results also demonstrate the predictions for
copper loadings depend on the partition coefficient IQ. A more
realistic model might vary the partition coefficient within a given
subsite, but the current approach does not provide for that
possibility. The use of a single soil core geochemistry (2-
dimensional isotropy) for each subsite is the most significant
assumption in the model. Figures 13, 15, and 18 compare
concentration predictions for several values of Kj with measured
surface water copper concentrations.
71
-------
Segment 1
Kd = 656, TQDIVV =0.04
8
Ul
D)
r>
S
ht
Q.
Q.
8
is S 8 §
B
§ 9 s S g
T i i i i i i i r
i i i i r
sBsggSSiSSiSsS
a 3 M * t a 2 1 » 5 a
sj s a ^ a a t
Comparison Dates, 1984-1336
Figure 18. Measured versus predicted copper
loadings, segment 1.
IN-STREAM PREDICTIONS
The final calibrated parameters for the model are shown in
Table 23. The in-stream modeling effort is based upon generating
stream hydraulics and loadings for five different meteorological
and flow scenarios. Two flood periods were examined - a typical
winter flood due to runoff and a typical spring flood due to
snowmelt. The seasonality of the flow periods and contaminant
transport rates has already been discussed. Figures 19 through 22
below show the model predictions for the above scenarios.
Parameters chosen for the streamside tailings on Silver Bow
Creek were as follows. Q/V (the NPSOUT variable QDIW) was 0.4
I/days and the contaminant pulse lifetime was set at 12.0 days.
For the streamside tailings along the Clark Fork River, Q/V was set
to i.o I/days, the pulse lifetime was set to 3.0 days for all
cases, and the travel time was set to 0.0 days.
In Figures 23 through 26, the comparisons between the
concentrations of copper measured for normal winter and normal
spring flow and model predictions are shown. The normal flows for
spring and winter were also examined.
72
-------
2500
| I I I I I I I I I ] Till
02/14/86
02/24/86
Event Dotes
-T-T r . i i i |
03/06/86
Figure 19. Predicted and measured winter floods, segment 1.
(•>
1200
1000
800
600
,\ 400
200
0 | i I i i i I I I r] i" i' r i i i i ii [ i i-ii r i i 'i
02/14/85 03/06/86
- 02/24/86
Event Dates
Figure 20. Predicted and measured winter floods, segment 11.
73
-------
3000
2500
2000
10
.S
d
1500
1000 —
o-,—
05/14/86
I I I I 1 I I I I 1 \
05/24/86
. 06/03/86
Event Dotes
06/13/86
Figure 21. Predicted and measured spring floods, segment 1.
1000
0
05/14/86
06/03/86
Event Dates
06/13/86
Figure 22. Predicted and measured spring floods, segment 1.
74
-------
0 ,-,
01/03/86
i i j i i i i ! i i i i
01/23/86 02/12/86
01/13/86 02/02/86
Event Dates
Figure 23. Predicted and measured normal winter flows,
segment 1.
20
CO
.c
o
^
3
0 I 11 11 i i i 111 11 i 11 11 i i i i ] i i . 11 i 11 11
01/03/86 01/23/86 02/12/86
01/13/86 02/02/86
Event Dates
Figure 24. Predicted and measured normal winter flow,
segment 11.
75
-------
3CC
250
E 200
k_
(A
.c 150
o
^ 100
s-
50
\J I I I I I I I 1 I I I I T T r 711 I F I
04/05/86 04/25/86 05/15/86
04/15/86 05/05/86
Event Dates
Figure 25. Predicted and measured normal spring flow,
segment 1.
0 iiT^^^H^TTT^^^PT^n^^^T^T^PPn
I i i i i i i i i i I
04/05/86 04/25/86 05/15/86
04/15/86 05/05/86
Event Dates
Figure 26. Predicted and measured normal spring flow,
segment 11.
76
-------
Finally, the low flow periods during late summer and the fall
were examined (see Figures 27 and 28) . Without having experimental
data for comparison, one can nonetheless observe the severity of
the predicted high copper concentration in the stream during low
flow periods.
0' i i i i i i i i i i i i i i i i i i
07/14/86 08/03/86 08/23/86
07/24/86 08/13/86
Event Dates
Figure 27. Predicted normal fall-summer flow, segment 1.
77
-------
in
(J fi i ifi iH^^^n^T T r rT^^
07/14/86 08/03/86 08/23/86
07/24/86 08/13/86
Event Dates
Figure 28. Predicted normal fall flow, segment 11.
78
-------
REFERENCES
Ambrose, R. B., T. A. Wool, J.P. Connolly, and R. W. Schanz.
1987. WASP4. A Hvdrodvnamic and Water Quality Model - Mod^i
Theory. User's Manual. and Programmer's Guide. U.S.
Environmental Protection Agency, Athens, GA. EPA/600/3-87/039!
297 pp.
Amein, M. and H.L. Chu, 1975. "Implicit Numerical Modeling of
Unsteady Flows", Journal of Hyd. Div.. ASCE, HY6, P.717
Brown, D. S. and J.D. Allison, 1987. MINTEOA1. An Equilibrium
Metal Speciation Model; Users Manual. U.S. Environmental
Protection Agency, Athens, GA. EPA/600/3-87/012.
Brown, K. P. 1989. Prediction of Metal Speciation and
Transport Using Models of Streamside Tailings Deposits In;
Proceedings of the Hazardous Waste and Hazardous Materials
Conference. Hazardous Materials Control Research Institute,
Silver Springs, MD. 726 pp.
Brown, K. P. and Z. Hosseinipour 1989 Water Quality Modeling
and Transport Analysis of Heavy Metal in the Clark Fork River
In: Symposium Proceedings on Headwaters Hydrology. American
Water Resources Association, Bethesda, MD. 708 pp.
Camp Dresser & McKee 1987. Final Field Operations Plan. Butte
Area Soils Screening Study. Camp Dresser & McGee. Denver, CO.
Carsel, R. F., C. N. Smith, L. A. Mulkey, J. D. Dean, P.
Jowise. 1984. Users Manual for the Pesticide Root Zone Model
fPRZMl U.S. Environmental Protection Agency, Athens, GA.,
EPA/600/3-84/109. 216 pp.
Chang, H.H. and J.C. Hill 1976. "Computer Modeling of Erodible
Flood Channels and Deltas," Journal Hvd. Div.. ASCE,
HY10, P-1461, 1976.
Chang, H. H. 1982. "Mathematical Model for Erodible Channels,"
Journal Hvd. Div., ASCE, HY5, P.678
Chow, V.T. 1959. Open-Channel Hydraulics. McGraw-Hill Book
Co., Inc., New York, NY.
CH2M-Hill. 1988. Silver Bow Creek Flood Modeling Study
(Drafts. CH2M-Hill, Boise, ID.
Fread, D.L., and G.F. Smith 1978. "Calibration Technique for
1-D Unsteady Flow Models", Journal Hvd. Div. .Ascy.r
104(HY7):1027-1044
79
-------
Garde, R.J. and K.G. Ranga Raju 1985. Mechanics of Sediment
Transportation and Alluvial Stream Problems. 2nd edition,.
John Wiley and Sons, Inc., New York, NY.
Henderson, P.M. 1966. Open Channel Flow. The McMillan Co., New
York, NY.
Hosseinipour, Z. 1988a. Review and Summary of Erosion and
Transport Processes in HSPF. Recommendation for Updating
Erosion and Transport Modules. U.S. EPA Chesapeake Bay Liaison
Office, Annapolis, MD.
Hosseinipour, Z. 1989 Fluvial Hydrodynamic and Sediment
Transport Model for the Chesapeake Bay Watershed. Int
Proceedings of the International Conference on Channel Flow
and Catchment Runoff for Centennial of the Manning's Formula
and Kuichling's Rational Formula. Charlottesville, VA.
Hosseinipour, Z. 1988b Development of a Fluvial River Flow
Routing and Sediment Transport Model for the Chesapeake Bay
Watershed. U.S. EPA Chesapeake Bay Liaison Office, Annapolis,
MD.
Hydrometrics. 1983a. Summit and Deer Lodge Valleys Long-Term
Environmental Rehabilitation Study. Butte-Anaconda. MT. Vol
VII; Warm Springs Ponds. For the Anaconda Minerals Company,
Butte, MT.
Krauskopf, K. B. 1967. Introduction to Geochemistry. McGraw
Hill Co., New York, NY.
Ingman, G. 1987 Water Quality Data for the Clark Fork River
1985-1986. Montana State Water Quality Bureau Helena, MT.
(unpublished report)
Lindberg, R.D., Runells, D.D., Ground Water Redox Reactions:
"An Analysis of Equilibrium State Applied to Eh Measurements
and Geochemical Modeling". Sciencef Volume 225, pp. 925-927
1984
Tetra Tech. 1985. Anaconda Smelter Remedial Investigation
Final Study . Tetra Tech, Bellevue, WA.
Tuesday, D. S., M. Grotbo, and W. M. Schafer 1987 Silver Bow
Creek Remedial Investigation Work Plan and Draft Final Report.
Multitech, Butte, MT., and Data Summary Report. Supplemental
Remedial Investigation. Ch2M Hill (Black and Veatch, ICF, PRC,
Ecology and Environment), Helena, MT.
Yalin, M.S. 1977. Mechanics of Sediment Transport Pergamon
Press, Inc. New York, NY.
80
-------
Yang, C. T., 1973 "Incipient Motion and Sediment Transport,"
Journal of Hyd. Div.. ASCE, HY10, P.1679
U.S. GOVERNMENT PRINTING OFFICE 1991/548-187/40561
81
------- |