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

-------